Economic Studies 151
Selva Bahar Baziki
Firms, International Competition, and the Labor Market
Selva Bahar Baziki
Firms, International Competition,
and the Labor Market
Department of Economics, Uppsala University
Visiting address: Kyrkogårdsgatan 10, Uppsala, Sweden
Postal address: Box 513, SE-751 20 Uppsala, Sweden
Telephone:
+46 18 471 00 00
Telefax:
+46 18 471 14 78
Internet:http://www.nek.uu.se/
_______________________________________________________
ECONOMICS AT UPPSALA UNIVERSITY
The Department of Economics at Uppsala University has a long history.
The first chair in Economics in the Nordic countries was instituted at
Uppsala University in 1741.
The main focus of research at the department has varied over the years
but has typically been oriented towards policy-relevant applied economics,
including both theoretical and empirical studies. The currently most active
areas of research can be grouped into six categories:
*
Labour economics
*
Public economics
*Macroeconomics
*Microeconometrics
Environmental economics
*
*
Housing and urban economics
_______________________________________________________
Additional information about research in progress and published reports is
given in our project catalogue. The catalogue can be ordered directly from
the Department of Economics.
Dissertation presented at Uppsala University to be publicly examined in Lecture Hall 2,
Ekonomikum, Kyrkogårdsgatan 10, Uppsala, Thursday, 11 June 2015 at 14:15 for the degree
of Doctor of Philosophy. The examination will be conducted in English. Faculty examiner:
Docent Katariina Nilsson Hakkala (Aalto University and Government Institute for Economic
Research, Helsinki).
Abstract
Baziki, S. B. 2015. Firms, International Competition, and the Labor Market. Economic studies
151. 186 pp. Uppsala: Department of Economics. ISBN 978-91-85519-58-3.
Essay 1: We propose a model of cross-border acquisitions in which multinational
enterprises (MNEs) and private equity-firms (PE) compete over domestic assets. MNEs'
advantage lies in firm-specific synergies, whereas PE-firms are good at restructuring. Prevailing
interest rates do not work in favor of PE-firms, but a lower risk premium and better financial
market development do. Stronger firm-specific synergies favor MNEs. Performing a welfare
analysis, we show that a policy of restricting PE-firms can be counterproductive.
Essay 2: To understand how share of cross-border leveraged buyouts (LBOs) in all crossborder Mergers and Acquisitions (M&As) varies across countries and time, I use a model of
cross-border M&As similar to Norbäck and Persson (2009) and Norbäck et al. (2013) where
the share of LBOs are negatively related to transaction costs, international market integration,
and property rights. I find evidence for these predictions in a comprehensive dataset covering
all majority-owned cross-border M&As in the world.
Essay 3: This paper studies the effect of increased competition from low wage countries
on the earnings gap between skilled and unskilled workers using Swedish matched workerfirm micro data. Treating Chinese accession into WTO as an exogenous shock, the paper
shows that higher Chinese import penetration increases earnings for high-skilled workers, and
creates a significantly larger skill premium, contributing to the increase in wage inequality. One
percentage points increase in Chinese import penetration results in about 1 percent higher wages
for skilled workers, and the rise in Chinese imports explains about 10 percent of the overall rise
in skilled wages.
Essay 4: This paper studies the changes in labor allocations across firms and industries in
response to recent changes in information and communication technologies (ICT) adoption and
increase in Chinese import penetration using a detailed matched worker-firm micro data from
Swedish manufacturing. High ICT adoption industries show increased assortative matching
patterns: those affected by high change in import competition experience skill upgrading in high
end firms, and those that are relatively shielded from import competition see an increase in
low skilled workers in lower end firms. Low ICT adoption does not exhibit these patterns. We
provide a tentative model to explain the main changes observed in the data.
Keywords: International Restructuring, Private Equity, Leveraged Buyouts, Import
Competition, Assortative Matching, Skill Premium
Selva Bahar Baziki, Department of Economics, Box 513, Uppsala University, SE-75120
Uppsala, Sweden.
© Selva Bahar Baziki 2015
ISSN 0283-7668
ISBN 978-91-85519-58-3
urn:nbn:se:uu:diva-251652 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-251652)
To my family
Acknowledgements
It takes a village to raise a baby.
First and foremost, I thank my advisor Nils Gottfries for his unmatched
sense of duty, tireless dedication and support towards his students as well as
his sense of humor. I have been very lucky to have had your input and insights
during many hours of discussions on my papers which have been integral to
my understanding of economic research. I thank my second advisor Mikael
Carlsson for all his comments, direct and practical advice, and moral support.
I am indebted to Thomas Rønde and Fredrik Heyman for their detailed reading
of the papers in this thesis, and valuable suggestions during my Licentiate and
Final seminars. I thank Oskar Nordström Skans for his valuable input to my
research, and Mikael Lindahl for his open door policy.
I am thankful to my coauthors Pehr-Johan Norbäck, Lars Persson, and
Joacim Tåg from whom I have learned a great deal both when I was working for them, and later when we became coauthors. I am also lucky to count
Teodora Borota Milicevic and Rita Ginja as great coauthors, friends, and
source of continuous support during the job market. A big part of the job
market burden also fell on Ali S Onder who had to switch between being a
mentor and being a friend all too often. I thank Daria Finocchiaro for all the
valuable advice during this process as well.
I would like to thank my previous bosses Jane Sneddon Little from the
Federal Reserve Bank of Boston, and Magnus Henrekson from the Research
Institute of Industrial Economics who have both given me valuable guidance,
and support over the years starting well before my PhD studies.
The PhD program would not have been the same if it were not for my classmates. I am thankful for the many wonderful discussions and camaraderie I
felt with Blair, Jessica, Miri, Sara, and my Uppsala cohort that is both the
best rowing team and the worst rowing team at the same time. I owe Irina
thanks both for being the best officemate since the very first day and also for
gracefully putting up with all sorts of distractions coming from this side of
the office. I thank Rachatar for braving the job market with me, Lovisa for our
culinary discoveries, Mohammad and Oskar for all the laughs, and Fredrik and
Linna for the dance moves.
I thank Tobias for the great Humlan tradition and discussions, Alex for all
his advice and always positive attitude, Jon, Jovan, Mattias E, Martin, Gunnar, and Georg for the football nights, Johan and Glenn for the courses we
braved together, Erik for sharing his unparalleled knowledge and intuition in
statistics, Edoardo, Laurent, Niklas, Oscar, Jenny, Sebastian E, Gabriella, Sebastian A, Evelina, Linuz, Ylva, Lucas, Karolina, Johannes, Arizo, Mattias N,
Anna, Michi, Yuwei, Adrian and Haishan for discussions around the office,
Kicki, and Maria for being wonderful mentees, and fellow Board Members of
the Association of PhD students, as well as the Social Sciences PhD Council
members for the great fun and distraction they have provided over the years.
Thanks to the administrative staff, Ann-Sofie, Nina, Stina, Emma, Åke, and
especially Katarina whose exemplary sense of organization and management
has saved lives and visa applications more than once.
I want to thank my Stockholmers Elena, Ettore, Pamela, and Aron for being
amazing friends, concert companions, and sanity keepers. Thanks to Selcuk
and the media people Görkem, Daniel, Emma, and Sylvain for giving me the
comic relief I needed, for all the lyxig fika, and antics we took part in. You
have been an integral part in what made life in Uppsala and Ekonomikum fun.
There are several people who have been participating in my adventures,
supporting me in my aspirations, and rightfully making fun of my never ending
studentship for a very long time. My RC gang, The Kitty team; Ivette, Pritika,
Manuela, and the rest of the BMC crew; Selim, Simon, Stef; Nayoung and
Katie; and Antoine; thank you for being wonderful friends I can count on.
This thesis is dedicated with gratitude to my parents and my brother for the
unconditional love and never ending support they have given me as I chased
my dreams away from them for the past 14 years. My final thanks goes to
Sweden, for being my home away from home for the past six years, and to
Uppsala University and its beautiful city for providing the perfect scenic environment for PhD studies.
Uppsala, April 23, 2015
Selva Bahar Baziki
Contents
1
Introduction
................................................................................................
13
2
Cross-border Acquisitions and Restructuring: Multinational
Enterprises versus Private Equity-firms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2
The Difference Between MNEs and PE-Firms . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3
The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1
Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2
Stage 3: Product Market Competition . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.3
Stage 2: Acquisition Auction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.4
Stage 1: Financial Contracting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4
Equilibrium Ownership: MNEs versus PE-Firms . . . . . . . . . . . . . . . . . . . . . . .
2.4.1
Firm-Specific Synergies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.2
Financial Market Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.3
Financial Market Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5
Welfare Effects of Restricting PE-Firms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.6
Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.6.1
Different Risk Premia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7
Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
20
24
26
26
27
28
31
32
32
35
36
37
40
40
41
45
Cross-border Leveraged Buyouts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2
A Theory of Cross-border M&As and LBOs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1
Stage 4: Product Market Competition . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2
Stage 3: Sale of the Restructured Domestic Firm . . . . . . . . .
3.2.3
Stage 2: Restructuring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.4
Stage 1: M&A or LBO? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.5
Transaction Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.6
Market Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1
Investments by the incumbents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2
Property Rights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.3
IPOs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.4
Policy and Welfare Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4
Empirical Estimations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.1
Empirical Models and Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.2
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.3
Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
46
50
50
51
51
53
58
58
62
62
63
63
64
65
65
69
78
3
3.5
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A.1 Proofs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A.2 Cournot with linear demand and restructuring k in the
target firm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
83
83
84
88
4
Globalization, Chinese Imports, and Skill Premia in a Small Open
Economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.2
Theoretical Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.3
Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.4
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.5
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.5.1
Chinese Import Penetration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.5.2
IV Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.5.3
Robustness Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.6
Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.6.1
Within Firm Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.7
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5
Import Competition and Technological Changes: Mobility of
Workers and Firms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3
Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1
Basic Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.2
Industry classifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.1
Worker-firm fixed effects distribution and its change
over time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.2
Sources of distribution changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5
Potential theoretical explanations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.1
The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.2
Numerical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.6
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
142
142
147
149
149
152
153
154
156
159
159
164
168
172
1. Introduction
The rise in globalization in the past thirty years has raised a number of questions about its multifaceted effects. Integration of markets on one hand increases the level of competition which could hurt some sectors or firms that
are unable to keep up, but on the other hand makes it possible for firms to
access cheaper inputs that are used in final goods production. Globalization
in financial services increases access to lending sources, but with highly integrated financial markets risks can also spill over faster. Higher international
trade translates into a higher variety of goods for consumers, but may come at
the cost of a job loss if their employment was in a firm unable to keep up with
the higher level of competition.
This thesis is composed of four essays that deal with various aspects of
globalization mentioned. The first two essays look at cross-border mergers
and acquisitions and incorporate Private Equity firms into the process as an
alternative to multinational enterprises (MNEs). The presence of Private Equity firms in international acquisitions arena is a phenomenon that first rose
to prominence in the 80s due to increased financial globalization and it has
caused heated discussions since. Private Equity firms perform heavily leveraged buyouts (LBOs) and keep the acquired firm for a limited amount of time
during which they restructure it and usually exit by a resale or an IPO. While
some view the presence of Private Equity firms and their LBO practices as representing a shift in corporate governance practices (Jensen, 1989) and applaud
the role they play in the creation of efficient firms, others have pointed out
that they may owe some of these efficiency gains to reductions in employment
which may contribute to higher unemployment rates (Kaplan, 1989).
Essay 1: Cross-border Acquisitions and Restructuring:
Multinational Enterprises versus Private Equity-firms
In this paper we develop a theory of the determinants of a cross-border leveraged buyout, and their interaction with cross-border mergers and acquisitions. We analyse the welfare effects of Private Equity participation in firm
takeovers. Even though there is a large body of work focusing on cross-border
mergers, there is a limited number of papers that jointly consider different
types of buyers. Previous work by Horn and Persson (2001) and Head and
Ries(1997) examined the welfare effects of buyouts, and Head and Ries (2006)
13
as well as Bjorvatn (2004) emphasize that cross border acquisitions are not
perfect substitutes with greenfield investments.
Our contribution is to propose a model of cross-border acquisitions and buyouts where MNEs exploit firm specific assets and retained earnings whereas
Private Equity firms exploit restructuring skills. We then use this model to
examine how macro-financial indicators and the structure of the domestic financial market may create asymmetries in the MNE acquisition-Private Equity
buyout pattern and discuss the corresponding welfare effects. We show that
changes in interest rates do not favor one type of investor over the other, since
the borrowing costs for the Private Equity firms are taken equal to the opportunity cost of using retained earnings by the MNEs. Higher risk premia on the
other hand hurt Private Equity firms more, since they do not have access to
retained earnings as MNEs do. Better financial market development provides
an easier exit route, and lower costs for the PE firm when they resell the target. This makes the target country more attractive to Private Equity firms, and
should increase the share of LBOs.
Finally, we explore the policy dimension comparing non-discriminatory
policies to discriminatory policies that restrict access to Private Equity firms.
In the policy debate there has been a concern that private equity buyouts, in
particular buyouts by foreign private equity firms, are driven by short run gains
resulting in long run inefficiencies. Comparing consumer and total welfare under two types of regimes, we underscore that restricting buyouts can be counterproductive since private equity buyouts (and their subsequent IPOs, or exit
sale to an outsider) may prevent a concentration of the market. Linking this
result to our earlier finding, the probability of market concentration is higher
when the risk premium in the economy is high since their retained earnings
put multinational enterprises at a comparative advantage over financial type
buyers.
Essay 2: Cross-border Leveraged Buyouts
To investigate the role of country-specific indicators for the difference in the
share of LBOs in all M&As across time and countries, I build on a model
of cross-border M&As from Norbäck and Persson (2009) and Norbäck et al.
(2013) that incorporates LBOs. A domestic firm is targeted by potential buyers who are either of the incumbent type that conducts an M&A, or a Private
Equity type which conducts an LBO. Both types of buyers invest in restructuring at the newly acquired firm, but since the PE firm maximizes the resale
price of the target rather than the product market profits as incumbents do, they
restructure more intensely.
Several testable results are drawn from the model. First, since Private Equity firms buy firms to sell them, an LBO will incur transaction costs first at
the initial take over, and secondly when the Private Equity firm is reselling the
14
target firm after restructuring. In contrast, an incumbent firm which buys to
keep only incurs the transaction cost only once. Therefore higher transaction
costs in a country should make it relatively less attractive to private equity
firms, decreasing cross-border LBO shares.
Next, the market structure of the target country will also create asymmetries in how attractive the target will be for either type of buyer. In this regard
I investigate two parameters: the size of the product market and the number of
firms serving the market. I use a Cournot setup where the restructuring performed by either type of buyer reduces marginal costs of the firm. Taking the
number of firms in the economy as given, increasing market size increases the
value of Private Equity firm’s heavier restructuring. Thus, the model predicts
relatively higher presence of LBOs in larger markets. For a given market size
however, higher number of firms reduces the market share of each firm. This
reduces the strategic benefit of owning the target firm, which in turn reduces
the Private Equity firm’s valuation of it, and thus the share of LBOs should
decline as the number of firms increases in the market.
Finally, building on Baziki et al. (2015), I tie the relative likelihood of Private Equity buyouts to property rights. When property rights are protected,
MNEs will be able to restructure more efficiently since they can then use their
firm-specific assets in restructuring without fearing knowledge spillover to rivals and make full use of potential synergies with the target firm. This increases the value of the target firm for the MNE. Therefore, better property
rights should decrease the relative presence of LBOs in an economy.
This paper complements the existing literature on the industry investment
aspect of cross-border M&As in international oligopoly markets by explicitly
allowing cross-border leveraged buyouts to affect the international merger and
restructuring procedures. The features of the model developed should make it
useful for analyzing issues where the focus is on the interplay between crossborder M&A and buy outs, firm restructuring and different types of corporate
and industrial policies.
On the empirical side I test the predictions of the model using a comprehensive database on all global cross-border investments and find that they are
broadly consistent with the data. Previously La Porta et al. (1997) and (1998)
have looked at a sample of 49 countries and investigated the link between
polity measures, law enforcement and financing. My empirical section adds
to this work by looking at investments going into all countries in the world
with a multifaceted focus.
Globalization has made it possible for countries with different production
capacities to access international product markets. Previously, the trade literature focused on an uneven exchange of goods between developed and less developed economies where the developed (North) economies would sell higher
15
value added, often manufactured goods to the South, and buy goods with less
value added from them. Over time, this type of exchange would lead to a specialization in their comparative advantages; high value added goods for North,
and low value added in the South.
An interesting recent phenomena is the rise of intra-industry trade between
countries of different development levels, manifested by the higher level of imports from less developed low-wage countries to developed economies in not
just primary goods, but also in manufacturing. These low-wage imports increase competition in the domestic markets in developed economies, and raise
questions on its effects on the allocation of resources, and on labor markets.
The remaining two papers in this thesis deal with the ways rising imports from
low-wage countries in manufacturing industries affect local labor markets in
Sweden in terms of return to skill and allocation of workers across industries
and firms.
Essay 3: Globalization, Chinese Imports, and Skill Premia in a
Small Open Economy
This paper investigates the response of the wage gap between skilled and unskilled workers in the face of tougher competition from low-wage countries by
using a rich linked employer employee database for Swedish manufacturing
firms from 1996 to 2007. Since there are many other ongoing phenomena at
the domestic and global level during this time period, it is difficult to single out
the effects of globalization on wage dispersion and more specifically the skill
premium. Many previous papers in the field have attributed the change in return to skill to skill-biased technical change. To be able to identify the effect,
I follow the recent empirical literature, and treat China’s membership to the
World Trade Organization as an exogenous trade shock to Sweden. Since an
unobserved domestic shock could be driving changes in both wages and imports from China, I instrument Swedish imports with imports to other Northern
European countries to overcome potential issues of simultaneity.
With rising imports, the wages of skilled and unskilled workers should be
affected differently depending on the nature of the imported good. Imports
from low-wage countries should affect the workers that produce similar goods
negatively, creating higher levels of unemployment and lower wages for lowskilled workers. In this environment, firms may find it profitable to switch
their production to the higher end (Grossman and Rossi-Hansberg, 2008) and
increase their demand for, and therefore the wages of high skilled workers.
Controlling for a time trend in return to education, I find that higher imports
from China contribute to a higher skill premium in the economy, and therefore
a widening wage gap, but I do not find a negative wage effect on low skilled
workers. The magnitude of the positive effect of higher imports from China
16
on skilled workers could explain about 10 percent of the rise in real wages for
college educated workers in my data.
This result adds to the differing views already present in the empirical literature. Using local labor market level data in the U.S., Autor et al. (2013)
look at the effect of industry level Chinese imports on local labor markets
and do not find a significant effect on manufacturing firm wages. On the other
hand, Alvarez and Opazo (2011) use Chilean data, and see that higher Chinese
imports have a negative effect on low skill wages. Using matched employeremployee data from Denmark, Ashournia et al. (2014) investigate the effect
of both industry-level and firm-level Chinese import penetration on individual
wages. The firm-level analysis shows evidence of lower wages for low skill
workers. This paper complements these previous empirical studies in the detailed matched employer-employee data it uses, and the attention it gives to
differentiating the effect for workers who stay on the same job versus a total
effect including those who switch their jobs.
Essay 4: Import Competition and Technological Changes:
Mobility of Workers and Firms
Workers have been increasingly matching to firms that resemble their type in
the economy (high type workers to high type firms, and low to low). Previous literature has attributed the cause of this change to either innovations in
ICT technology (Autor and Dorn, 2013) or rising imports from China. Those
focusing on technological motivations have predicted a double ended concentration of workers where high skilled workers match with high type firms, and
vice versa. Papers in trade literature predict that higher trade with low wage
countries should reduce low skilled employment in competing sectors.
In a recent paper, Autor et al. (2014) attempt to disentangle the two forces,
ICT technology and import competition, in their effect on employment across
skills. They find that technological progress and import competition have
rather independent effects. The approach we follow resembles Autor et al.
(2014), but we add in three important dimensions. First, since we have access
to matched employer-employee data, we can track changes of firms over time
and control for the firm where the individual works. Next, beyond Autor et al.
(2014), we study the impacts of technological changes and trade not only on
employment, but on labor allocations, and finally we study the interaction between technology and import competition and look at movements in the wage
distribution.
Relying on the Abowd, Kramarz, Margolis (1999) methodology, we first
decompose individual wages into the contributions from the characteristics of
the individual and the firm they are employed in. We then construct the joint
distribution of these individual and firm effects, thus creating a map of the
manufacturing sector labor force allocation. We follow this distribution over
17
the span of two periods 1996-2000 and 2001-2007. The second period starts
in 2001 when China joined the WTO, which is treated as an exogenous trade
shock. Our goal is to provide evidence on the effects of technological change
and its interaction with import competition on the allocation of workers across
firms, i.e. the sorting patterns.
We find that there is an increase in sorting in high technology industries.
However, the change in the sorting pattern is not uniform across industries
within the high ICT intensity group, which points to the importance of the
interaction between technology and trade. High ICT sectors affected by an
increase in import competition experience skill upgrading in high end firms.
In contrast, ICT industries that are relatively shielded from import competition see an increase in the number of low skilled workers in lower end firms.
Finally, we propose a model which attempts to explain the effect of trade in
high ICT industries on the matching of workers with firms.
References
Abowd J.M., Kramarz F. and Margolis D.N., (1999). High Wage Workers and
High Wage Firms, Econometrica, Econometric Society, 67(2), 251-334.
Alvarez R. and Opazo L., (2011). Effects of Chinese Imports on Relative
Wages: Microevidence from Chile, Scandinavian Journal of Economics, 113,
342-363.
Ashournia D., Munch J. and Nguyen D., (2014). The Impact of Chinese Import Penetration on Danish Firms and Workers, IZA Discussion Paper Series,
8166.
Autor D.H., Dorn D. and Hanson G.H., (2013). The China Syndrome: Local
Labor Market Effects of Import Competition in the United States, American
Economic Review, 103(6), 2121-68.
Autor D., Dorn D. and Hanson, G., (2014). Untangling Trade and Technology:
Evidence from Local Labor Markets, Economic Journal, forthcoming.
Baziki S.B., Norbäck P.-J., Persson L., and Tåg J., (2015). Cross-border Acquisitions and Restructuring: Multinational Enterprises versus Private Equityfirms, IFN Working Paper No.1057.
Bjorvatn K., (2004). Economic Integration and the Profitability of CrossBorder Mergers and Acquisitions, European Economic Review, 48, 12111226.
Grossman G. and Rossi-Hansberg E., (2008). Trading Tasks: A Simple Theory of Offshoring, American Economic Review, 98(5), 1978-1997.
18
Head K. and Ries J., (2008). FDI as an Outcome of the Market for Corporate
Control: Theory and Evidence, Journal of International Economics, 74, 2-20.
Head K. and Ries J., (1997). International Mergers and Welfare Under Decentralized Competition Policy, Canadian Journal of Economics, 30, 1104-1123.
Horn H. and Persson L., (2001). The Equilibrium Ownership of an International Oligopoly, Journal of International Economics, 53, 307-333.
Jensen M., (1989). Eclipse of the Public Corporation, Harvard Business Review, Sept-Oct.
Kaplan S.N. (1989). The Effects of Management Buyouts on Operating Performance and Value, Journal of Financial Economics, 24(2), 217-254.
La Porta R., Lopez-De-Silanes F., Shleifer A., and Vishny R.W., (1997). Legal
Determinants of External Finance, The Journal of Finance, 52(3), 1113-1150.
La Porta R., Lopez-De-Silanes F., Shleifer A., and Vishny R.W., (1998). Law
and Finance, Journal of Political Economy, 106(6), 1113-1155.
Norbäck P.-J. and Persson L., (2009). The Organization of the Innovation
Industry: Entrepreneurs, Venture Capitalists, and Oligopolists, Journal of the
European Economic Association, 7(6), 1261-1290.
Norbäck P.-J., Persson L., and Tåg J., (2013). Buying to Sell: Private Equity
Buyouts and Industrial Restructuring, CESifo Working Papers, 4338.
19
2. Cross-border Acquisitions and
Restructuring: Multinational Enterprises
versus Private Equity-firms1
2.1 Introduction
It is well established that cross-border acquisitions by multinational firms
(MNEs) play a key role in the global industrial development and restructuring
process. It is less known, however, that more than 8 percent of all cross-border
acquisitions that took place during 1998-2010 were cross-border acquisitions
undertaken by private equity firms (PE-firms). PE-firms are financial buyers
of assets that acquire firms with the goal of restructuring and then reselling
them. Data from the Capital IQ database displayed in Figure 2.1 shows that
there is substantial variation over time, across countries, and across sectors
in the share of cross-border acquisitions that are acquisitions by PE-firms as
opposed to by MNEs.
Despite a burgeoning literature on cross-border acquisitions by MNEs, there
is, to our knowledge, no work on the determinants and effects of cross-border
acquisitions undertaken by PE-firms and their interaction with cross-border
acquisitions undertaken by MNEs. What are potential determinants behind
the variation we observe in Figure 2.1? And what are the welfare effects of
acquisitions by PE-firms as opposed to MNEs?
To answer these questions, we develop a theory of cross-border acquisitions by MNEs and PE-firms. The starting point is that MNEs and PE-firms
differ along three important dimensions. First, MNEs are "insiders" in an industry in the sense that they already have industry-specific knowledge and
firm-specific assets in the global marketplace. These assets take the form of
production facilities, intellectual property and retained earnings. Second, PEfirms are on the other hand "outsiders" that excel at reorganizing target firms
to improve their productivity and profitability. As such, they do not have to
rely on firm-specific assets to profit from acquisitions of target firms and they
1 This chapter is coauthored with Pehr-Johan Norbäck (Research Institute of Industrial
Economics), Lars Persson (Research Institute of Industrial Economics and CEPR), and Joacim
Tåg (Research Institute of Industrial Economics). Financial support from the Marianne and
Marcus Wallenberg Foundation, Wallander-Hedelius-Browaldh Foundation, and Tom Hedelius’
and Jan Wallander’s Research Foundation is gratefully acknowledged. We thank Nils Gottfries
and Thomas Rønde as well as seminar participants at various institutions for excellent comments
and suggestions.
20
12
10
8
Percent by PE-firms
6
4
1998
1999
2000
2001
2002
2003
France
Italy
Germany
Spain
Switzerland
Sweden
Netherlands
Denmark
Finland
Belgium
Austria
Czech Republic
United Kingdom
Norway
New Zealand
Singapore
Poland
Ireland
South Africa
Australia
Ukraine
India
Canada
United States
Hong Kong
Mexico
Brazil
Russia
China
2004
2005
2006
2007
2008
2009
2
6
2010
Consumer Discretionary
Consumer Staples
Industrials
Utilities
Healthcare
Financials
Telecommunication Services
Materials
Energy
Information Technology
0
2
4
6
8
10
12
Percent by PE-firms
14
16
0
4
8
10 12 14
Percent by PE-firms
Figure 2.1. Percent of cross-border acquisitions that are undertaken by PE-firms
Note: These figures display the percent of all cross-border acquisitions between 1998 and 2010
in the Capital IQ database that undertaken by PE-firms across time (top), country (bottom left),
and sector (bottom right). We selected all "Mergers/Acquisitions" that had the feature "Crossborder" and "Acquisition of Majority Stake" and with transaction status "Closed" or "Effective". We then characterized the transaction as a transaction by a PE-firm if the transaction
had the secondary feature of "LBO (Leveraged Buyout)". The figures display the percent of all
cross-border transactions across time, country and industry for countries with more than 500
transactions in total.
21
do not have to account for how their actions affect other potential asset holdings in the market. They are, however, required to rely to a greater extent on
external investors to finance the acquisitions they undertake. Third, PE-firms
are temporary owners of assets while MNEs are more permanent owners of
assets. This implies that PE-firms are in need of clear exit strategies and face
additional costs associated with reselling the target firms they acquire.
Using this distinction between MNEs and PE-firms, we propose the following model: initially there is a domestic product-market served by several
foreign MNEs. There is also a domestic firm which is put up for sale. The
domestic firm could be acquired either by one of the foreign MNEs or, alternatively, by one of several competing foreign PE-firms. Prior to bidding for the
domestic firm, MNEs and PE-firms may seek financing from outside investors
to finance the acquisition. Post-acquisition repayments to outside investors are
made and if a PE-firm acquired the target, the PE-firm resells the target firm.
Product market competition takes place between the firms with assets in the
market.
We first establish that firm-specific synergies between MNE assets and the
target firm assets are an important determinant of whether an MNA or a PEfirm ends up acquiring the target. Stronger synergies work in favor of MNEs,
that are then able to bid higher for the target firm compared to PE-firms. This
suggests that variation across industries in asset complementary or how strong
economies of scale are could be one important determinant of the differences
in the share of all cross-border transactions across industries that are acquisitions by PE-firms.
We then show that acquisitions by PE-firms are more likely when financial
market development in the target firm’s country is higher. The reason is that
this makes exits by PE-firms less costly. For example, in countries with developed financial markets there are more exit opportunities, and it is common that
PE-firms exit their investments by listing the target firm on a stock exchange.
Financial market development may thus be one important factor determining
the share of all cross-border acquisitions across countries that are made by
PE-firms.
In addition to synergies and financial market development, we also examine
how general financial market conditions affect whether an MNE or a PE-firm
acquires the target. We show that the overall interest rate in the economy does
not affect who ends up as the owner in equilibrium. At first glance, this result is
surprising since PE-firms lack retained earnings and must, to a greater extent,
rely on financing from banks and investors. However, with well functioning
capital markets, the cost of access to capital from outside investors to finance
the acquisition also determines the opportunity cost of retained earnings for
MNEs. When the interest rate goes up, making access to external financing
more costly, the opportunity cost of using retained earnings for the acquisition also increases, implying that the foreign MNE firms remain indifferent to
using retained earnings or external financing to finance the acquisition. Ac22
quisition prices, though, do respond to changes in the interest rate as higher
interest rates reduce the amounts that both MNEs and PE-firms are willing to
pay to make the acquisition.
However, in contrast to the general interest rate, the risk premium associated with lending affects the equilibrium ownership of the target. In effect,
since MNEs can finance part of an acquisition with retained earnings instead
of going to outside investors, MNEs have an advantage over PE-firms when
the risk premium is high. This financial advantage decreases when the risk
premium decreases. This suggests that we should observe fewer cross-border
acquisitions by PE-firms relative to cross-border acquisitions by MNEs when
economic conditions are worse and external financing for acquisitions is more
costly.
Our model also allows us to explore the policy dimension. In the policy
debate, there has been a concern that acquisitions by PE-firms—in particular by foreign PE-firms—are driven by short-run tax gains and asset stripping, resulting in job losses and long-run inefficiencies. For example, in 2005
Franz Muntefering of the ruling Social Democratic Party in Germany attacked
foreign PE-firms making acquisitions in Germany by stating to the newspaper Bild that “They stay anonymous, have no face, fall upon companies like
swarms of locusts, strip them bare and move on.”2
Performing a welfare analysis on the effects of restricting foreign PE-firms
from buying domestic target firms, we find that restricting PE-firms can be
counterproductive since PE-firms may prevent a concentration of the market
and may also—in instances when synergies between the MNEs’ assets and the
target’s assets are low—increase productivity of the target’s assets more than
MNEs do.
Our paper is a contribution to the theoretical literature on cross-border
mergers and acquisitions in oligopolistic markets. This literature has, in contrast to the traditional foreign direct investment literature, emphasized that
greenfield investments and cross-border acquisitions are not perfect substitutes.3 There is also a small theoretical literature addressing welfare aspects
of cross-border mergers in international oligopoly markets.4 Our contribution
to this body of literature is to build on Norbäck and Persson (2007, 2008) to
incorporate foreign financial bidders in the form of PE-firms along the lines
of Norbäck, Persson, and Tåg (2013). We are thus the first to propose an
2 See
also, for instance, “Testing the Model: Private Equity Faces a More Hostile World”
(Jul 9 2009, The Economist), “Editorial, New Rules for Private Equity” (August 30 2009, New
York Times) or “Private Equity Fights Tax Plan” (February 27 2009, Financial Times).
3 See, for instance, Blonigen (1997), Bjorvatn (2004), Bertrand and Zitouna (2006), Fumagalli and Vasconcelos (2006) Head and Ries (2008), Mattoo, Olarreaga and Saggi (2004), Raff,
Ryan and Stähler (2009) and Norbäck and Persson (2008)
4 This literature includes papers by, for example, Falvey (1998), Head and Ries (1997), Horn
and Persson (2001), Lommerud, Straume and Sorgard (2006), Neary (2007) and Norbäck and
Persson (2007).
23
oligopoly model of cross-border acquisitions where MNEs compete for domestic target firms with foreign PE-firms. We then use this model to derive
predictions on when acquisitions by PE-firms as opposed to by MNEs are
more likely and to study the welfare effects of restricting acquisitions by PEfirms.
Our analysis is related to the literature on industrial reorganization in the
finance literature which shows that M&A activity can be triggered by changes
in owner productivity and the cost of new capital, where more productive owners buy assets from less productive ones.5 We add by showing that financial
conditions may affect the types of mergers and how the efficiencies in these
acquisitions are affected. There is also an emerging finance literature on the
differences between strategic and financial buyers of assets in takeover auctions.6 This literature has not, however, focused on cross-border takeover auctions, oligopolistic product markets or welfare analyses.
The rest of the paper is organized as follows. The next section briefly discusses the differences between MNEs and PE-firms. In Section 2.3 we then set
up a formal model of bidding competition between MNEs and PE-firms over
domestic targets. In Section 2.4, we perform comparative statics to derive
our main propositions regarding the effect of synergies, financial market development, and financial market conditions on whether an MNE or a PE-firm
ends up buying the target. Section 2.5 discusses the welfare effects of banning
cross-border acquisitions by PE-firms, Section 2.6 discusses extensions to the
model, and we offer some concluding remarks in Section 2.7.
2.2 The Difference Between MNEs and PE-Firms
To better highlight the difference between MNEs and PE-firms, we start out
with a brief primer of the business model of MNEs and PE-firms. MNEs are
typically firms with firm-specific assets such as patents, know-how, and brand
image that they exploit internationally. They are often large in size, organized
as limited liability companies, and listed on a stock market. An MNE can
expand internationally, either through greenfield investments (setting up a new
plant) or by acquiring firms in host countries. The MNE model works as
follows:
1. A group of entrepreneurs or managers with a business idea sets up a
limited liability firm.
2. Due to proficient management and the creation of high-quality firmspecific assets such as patents and know-how, they grow and expand
internationally.
5 See,
for instance, Maksimovic and Phillips (2002).
for example, Gorbenko and Malenko (2014), Martos-Vila, Rhodes-Kropf, and Harford (2013) and Hege, Lovo, Slovin, and Sushka (2012).
6 See,
24
3. The firm then exploits its firm and industry-specific assets and retained
earnings internationally by exporting and/or undertaking foreign direct
investment. It can also engage in greenfield investments or cross-border
acquisitions at this stage.
4. This process then continues and shareholders benefit from increased
stock market value and dividends.
In contrast, PE-firms are partnerships set up to acquire, restructure and resell firms with the help of money from institutional investors and from banks.
This business model emerged in the 1980s in the United States but has since
spread out worldwide.7 The private equity business model works as follows:
1. A group of entrepreneurs or managers with restructuring skills and an
idea on how to improve the profitability of existing businesses set up a
PE-firm and an associated PE-fund with a predetermined life span (usually around 10 years).
2. The partners in the PE-firm raise capital from institutional investors such
as pension funds and wealthy individuals.
3. After the target amount of capital for the PE-fund has been raised, the
fund is closed and the PE-firm starts looking for firms to acquire, restructure, and then resell.
4. Once a firm has been identified, debt is raised to finance the acquisition.
PE-firms usually acquire multiple firms in each fund, and each acquisition is financed with 60%-90% debt.
5. The target firm is acquired and restructured. Cash flows from the firm
are used to pay off part of the debt.
6. After the firm has been restructured, the PE-firm resells the firm. The
most common exit routes are listing the company on a stock exchange
or selling it to another firm.
7. The returns from cash flows during the restructuring period and from
the sale of the firms in the fund are split on a 80/20 basis with 80%
going back to the investors in the PE-fund and 20% going to the PEfirm. The PE-firm also charges a management fee of 1%-2% of the
capital committed to the fund.
This business model gives PE-firms several advantages over publicly listed
MNEs in the restructuring process. First, concentrated ownership implies that
agency costs are lower than in publicly listed firms and that the high leverage
that PE-firms put on target firms puts pressure on managers to generate cash
flow and not waste money on unprofitable investments (Jensen 1986; 1989).
Second, PE-firms are temporary owners of the target firms and therefore have
stronger incentives to both restructure target firms and take on debt to give
management incentives to undertake restructuring activities (Norbäck, Persson
and Tåg, 2013). Finally, PE-backed firms are not listed on a stock exchange
and can therefore have an advantage over publicly traded firms due to less
7 See
Kaplan and Strömberg (2009) or Tåg (2012) for surveys of the literature on PE-firms.
25
stringent reporting requirements. Publicly traded firms are subject to tighter
bookkeeping, accounting and reporting standards which impose restrictions on
the time and effort management spends on productivity enhancing exercises.
2.3 The Model
Throughout the paper, we will focus on three main differences between MNEs
and PE-firms driven from the section above which described how PE-firms and
MNEs operate. First, MNEs are "insiders" in an industry in the sense that they
already have industry-specific knowledge and firm-specific assets in the global
marketplace. These assets take the form of production facilities, intellectual
property and retained earnings. PE-firms lack such retained earnings, assets
and industry-specific knowledge.
Second, PE-firms are on the other hand "outsiders" that excel at reorganizing target firms to improve their productivity and profitability.8 As such, they
do not have to rely on firm-specific assets to profit from acquisitions of target firms and they do not have to account for how their actions affect other
potential asset holdings in the market.
Third, PE-firms are temporary owners of assets while MNEs are more permanent owners of assets. This implies that PE-firms are in need of a clear
exit strategy and face additional costs associated with reselling the target firms
they acquire.
In the remainder of this section, we build a model that incorporates these
three differences between MNEs and PE-firms.
2.3.1 Setup
Consider an economy consisting of a domestic target firm, several foreign
MNEs also operating in the domestic market and several foreign PE-firms.
MNEs and PE-firms compete with each other to acquire the target firm, and
they both need to obtain financing from external investors. The timing of the
model is as follows:
1. Financial contracting. MNEs and PE-firms write contracts with external
investors to obtain promises of financing in stage 2 in the case of winning
the auction for the target firm. Repayments are made at stage 3.
2. The acquisition auction. A domestic target firm is up for sale through a
first price perfect information auction with externalities.9 Externalities
mean that the value of winning for a bidder is determined relative to what
8 See
Davis, Haltiwanger, Handley, Lerner and Jarmin, Miranda (2014) for recent evidence
on productivity improvements and Kaplan and Strömberg 2009 or Tåg(2010) for surveys of the
literature.
9 For an example of this type of auction, see Jehiel and Moldovanu (1999) or Norbäck,
Persson and Tåg (2013).
26
happens if the bidder loses the auction. The bidders in this auction are
the MNEs and the PE-firms. Depending on the valuations for winning
the auction, either an MNE or a PE-firm obtains the target.
3. Product market competition. Firms that have assets for producing compete in the product market. If the target was acquired by a PE-firm in
stage 2, it will need to resell the target firm at the beginning of this stage
(to, say, an independent outside investor).10 This gives n + 1 firms competing in the product market. Alternatively, if an MNE acquired the
target, there will only be n firms competing on the market. In this case,
after product market actions have been set and profits have been realized, repayments to outside investors are made, (if the MNE chooses to
resell the target, eventual costs of reselling the assets are incurred) and
the game ends.
We solve the game by backward induction starting with the product market
competition stage.
2.3.2 Stage 3: Product Market Competition
The set of potentially producing firms in the industry is J = {i1 , .., in , p1 , .., pm },
where j ∈ J is an element. The first n entries refer to the n number of MNEs
(i) and the final m entries to the m number of PE-firms (p). The set of (potential) owners of the target firm’s assets is L = J , where l ∈ L is an element. Let
π j (x, l) denote the product market profit of firm j. The vector of actions taken
by firms in product market interaction is x. Only firms that have assets for producing compete in the product market. Hence, if the target was acquired by a
PE-firm then there are n + 1 firms on the market, whereas if an MNE acquired
the target, there are n firms on the market.
Given l, firm j chooses an action x j to maximize its product market profits
anticipating the repayments to outside investors:
π j (x j , x− j : l) − Rl − El − Fl .
(2.1)
In this expression, x− j is the set of actions taken by j’s rivals, Rl refers to
payments made to outside investors and El are exit costs incurred by PE-firms
such that Ei = 0 and E p = E if firm j = l is the acquirer. Fixed restructuring
costs are Fl and are only relevant if the target is acquired (Fl > 0 if l = j and
Fl = 0 for l = j).
∗ (l)) exists, and
Assume that a unique Nash-Equilibrium x∗ (l) = (x∗j (l), x−
j
is defined as
∗
∗
+
π j (x∗j , x−
j : l) − Rl − El − Fl ≥ π j (x j , x− j : l) − Rl − El − Fl , ∀x j ∈ R . (2.2)
10 We
could also allow the PE firm to make some product market profits during the restructuring process without changing the results qualitatively.
27
Since product market actions do not affect repayments of what was previously borrowed and fixed restructuring costs do not affect product market
actions, we can define a reduced-form product market profit for firm j, taking
∗ (l) , l).
ownership l of the target firm’s assets as given, as π j (l) ≡ π j (x∗j (l) , x−
j
Thus, we have that πA (i) is the profit for an MNE (l = i) and πP (p) is the
profit for a PE-firm (l = p) in the case when an acquisition takes place. Nonacquiring MNEs’ profits are πNA (l), where l = {p, i} is the type of owner of
the target firm. Non-acquiring PE-firms do not have assets in the market and
thus have zero profits.
The potential owners of the target firm differ in terms of how efficient they
are at utilizing its assets; this is denoted by an ownership efficiency parameter
γ ∈ [0, γmax ] for γmax > 1. This parameter captures synergies between the assets of the MNE and the target and it captures the extent to which PE-firms are
able to restructure the target firm to improve profitability. We allow the ownership efficiency parameter to vary between MNEs and PE-firms, but normalize
it to unity for PE-firms. Hence, γ > 1 says that a MNE firm would be able to
make better use of the target’s assets than a PE-firm, whereas if γ < 1 then an
MNE is less efficient than a PE-firm at running the target. The ownership efficiency parameter affects the potential profits of the agents through the agent’s
optimal action x∗j . The first effect is a direct effect through improvements in
efficiencies, and cost reduction, while the second effect is a strategic effect
that materializes itself in the form of a relative price advantage compared to
rivals. Note that profits do not depend on γ when the target is PE owned.
The ownership efficiency parameter will affect the profits of MNEs and
rival MNEs in the following way:
Assumption 1
dπNA (i)
dπA (i)
> 0 and
< 0.
dγ
dγ
Assumption 1 is compatible with oligopoly models of competition (see for
example Farrell and Shapiro, 1990). A simple example is a Cournot model
where γ reduces the marginal costs for the acquirer.11
2.3.3 Stage 2: Acquisition Auction
In this stage, we determine the ownership and the acquisition price of the
target’s assets. The acquisition process is depicted as an auction where all
MNEs and PE-firms simultaneously post bids. Everyone announces a bid,
Let demand be linear, P = a− Qs ,
where a indicates consumer willingness to pay and s denotes market size. Direct product market
profits are Πh = (P − ch )qh , where qh is output for a firm of type h = {A, NA}. The marginal
cost of an acquirer is cA = c − γk and the non-acquirer has the marginal cost cNA = c. k captures
2
the quality of the initial assets. Reduced-form profits then take the form πh (l) = 1s q∗h , where
11 Take the Linear Quadratic Cournot Model as an example.
q∗A (i) =
28
a−c+nγk
n+1
and q∗NA (i) =
a−c−γk
n+1 .
Hence,
dπA (i)
dγ
> 0 and
dπNA (i)
dγ
< 0.
bi , which is either accepted or rejected by the target’s owner. Following the
announcement of bids, the target’s assets are sold at the highest bid price.
The acquisition is solved for Nash equilibria in undominated pure strategies.
We allow MNEs with retained earnings A to access financial markets to earn
interest rate r between stages 2 and 3, if they do not use the retained earnings
to (partly) pay for the target’s assets. In stage 1, each MNE and PE-firm is
offered a borrowing contract specifying a repayment R∗l (Il ) in period 3 when
borrowing an amount Il .
To solve the acquisition auction and determine bids, we need to determine
the valuations of the bidders for obtaining the assets. To aid in this, we introduce the net gain function Nl (S) which defines the net gain for a bidder of
type l if the acquisition price is S for which the acquirer only borrows the minimum amount needed. We denote this amount by Ilmin (S, A). As noted above,
the MNE will face a fixed restructuring cost of Fi when restructuring the target
firm.
An MNE will have two net gain functions defined as
Nil (S) = [πA (i) − S] + [Iimin (S, A) − R∗i (Iimin (S, A))] + A − Fi
Acquire
− [πNA (l) + (1 + r)A],
(2.3)
Do not acquire
for l ∈ {i, p}. The first term consists of product market profits net of the
acquisition price (πA (i) − S). The second term consists of the interest payments on borrowed funds, i.e. the amount borrowed net the amount repaid
(Iimin (S, A) − R∗i (Iimin (S, A)). The third term is the retained earnings (A) and
the fourth term is the fixed restructuring costs (−Fi ). The fifth term is the
product market profits in the case where the MNE does not acquire the target
and is forced to compete either with an MNE that acquires the assets (−πNA (i))
or a target acquired by a PE-firm (−πNA (p)). Note that an MNE’s maximum
willingness to pay for the target depends on what happens if another MNE
obtains the target (l = i) or if a PE-firm obtains the target (l = p) in a different
way since the two types differ in synergies γ. The final term, −(1 + r)A, is the
value of the retained earnings in the case where the MNE does not acquire the
target.
The maximum willingness to pay, vil , can be determined as vil = min S, s.t
Nil (S) ≥ 0. Solving for Nil (S) = 0, we get the maximum willingness to pay
for each of the two net gain functions as
vil = πA (i) − Fi − [R∗i (Iimin (vil , A)) − Iimin (vil , A)]
Cost of borrowing
− πNA (l) −
rA
(2.4)
Opportunity cost of retained earnings
29
Using the same argument, we see that the net gain for a PE-firm of acquiring
the assets equals
N p (S) = [π p (p) − E] − S − [R∗p (I pmin (S)) − I pmin (S)] − Fp −
Acquire
0
, (2.5)
Do not acquire
which gives a maximum willingness to pay equal to
v p = πP (p) − Fp − E − [R∗p (I pmin (v p )) − I pmin (v p )].
(2.6)
Cost of borrowing
Given the valuations vil and v p , defined in equations (2.4) and (2.6), we can
now solve the auction for the target’s assets and determine equilibrium ownership and the acquisition price. These valuations can be ranked in six ways and
the auction solved by considering each ranking in turn.
Lemma 1 The equilibrium ownership of the target and the acquisition price
S∗ in stage 1 are given in Table 2.1.
Table 2.1. Equilibrium Ownership
Inequality
I1
I2
I3
I4
I5
I6
Definition
vii > vip > v p
vii > v p > vip
vip > vii > v p
vip > v p > vii
v p > vii > vip
v p > vip > vii
Winning type
i
i (or p)
i
i
p
p
Acquisition price, S∗ .
vii
vii (or v p )
vii
vp
vp
vp
Proof. See the Appendix.
For I1 − I3, a preemptive MNE acquisition takes place. MNEs bid against
each other in the fear of being forced to compete with an MNE that acquired
the target and obtained efficiency gains from synergies, γ. The acquisition
price is bid all the way up to S∗ = vii . For I2, there can potentially be two
equilibria. In the first, MNEs do not expect other MNEs to outbid the PE-firms
so that a PE-firm wins the auction. In the second, MNEs expect other MNEs to
bid for the target and that a preemptive acquisition takes place. For simplicity,
we will focus on the second equilibrium (a preemptive acquisition occurs). For
I4, a concentrating MNE acquisition takes place at price S∗ = v p . One MNE
finds it profitable to bid slightly above the PE-firms in order to reduce the
number of firms on the market from n + 1 to n. For I5 − I6, a PE-firm acquires
the assets since a PE-firm’s valuation exceeds an MNEs’ valuations. Finally,
note that the equilibrium borrowing Il∗ (S∗ , A) and repayment R∗l (Il∗ (S∗ , A)) are
determined by S∗ .
30
2.3.4 Stage 1: Financial Contracting
In this stage, each MNE and PE-firm is offered a borrowing contract specifying a repayment Rl (Il ) in stage 3 when borrowing an amount Il in stage 2 if
(Ilmax ). Note that
acquiring the target firm, up to the maximum amount Rmax
l
funding is conditional on making an acquisition, and we assume perfect information and bidding competitions among symmetric investors, implying that
all investors break even. Consequently, each MNE and PE-firm is promised a
borrowing contract in equilibrium.
Since acquiring firms are typically associated with an idiosyncratic risk, the
investor will in reality require a risk premium to lend money to the acquirer.
This risk is typically not present in the deposit (lending) rate r. To capture differences in borrowing costs between agents, we introduce an additional borrowing cost labeled ρ. For simplicity, we call this fixed extra borrowing cost a
"risk premium" even though our model does not have any uncertainty in it.12
Given perfect financial markets, it follows that investors will require an
interest rate of r +ρ to lend in equilibrium. At a lower interest rate, the investor
will prefer putting its money in the risk free investment to get r. And perfect
financial markets mean that there are always investors willing to lend to the
MNEs and the PE-firms at r + ρ. This implies that the equilibrium repayment
will be R∗l = (1 + r + ρ)Il .
We can now state the following lemma.
Lemma 2 In the borrowing equilibrium, we have that:
(i) A PE-firm can borrow of I p∗ max = (1 + r + ρ)−1 (πP (p) − Fp − E) with an
associated repayment of R∗pmax = πP (p) − Fp − E.
(ii) An MNE can borrow Ii∗ max = (1 + r + ρ)−1 (πA (i) − Fi ) with an associated
repayment of R∗pmax = πA (i) − Fi .
This lemma pins down the maximum amount that MNEs and PE-firms can
borrow to finance the acquisition in stage 2. The proof goes a follows. For
PE-firms, we know that the maximum payoff in stage three is πP (p) − Fp − E.
Thus, R∗pmax = πP (p)−Fp −E and I p∗ max = (1+r +ρ)−1 (πP (p)−Fp −E). For
an MNE, we know that its maximum repayment in stage three is πA (i) − Fi .
Thus, R∗i max = πA (i) − Fi and Ii∗ max = (1 + r + ρ)−1 (πA (i) − Fi ).
12 The risk premium will account for a comprehensive credit risk that includes but is not
limited to the risk of loan restructuring, moratorium, and other changes in the payment plan
which are more common among PE-firms than the MNE. Additionally, ρ could reflect additional
expectations of returns investors in PE-funds would require to compensate for liquidity risks.
Thus, it would make sense to introduce differences in the credit risk of both types of borrowers.
However, to simplify this, we here assume it to be the same for MNEs and PE-firms. We relax
this assumption in Section 2.6.
31
2.4 Equilibrium Ownership: MNEs versus PE-Firms
Having set up a model of bidding competitions between MNE firms and PEfirms, let us now turn to comparative statics on the determinants of when an
MNE will acquire the target and when the target will be bought by a PE-firm.
We focus on three aspects: firm-specific synergies (γ), financial market conditions (interest rates, r and ρ) and financial market development (exit costs,
E).
2.4.1 Firm-Specific Synergies
We start with how equilibrium ownership depends on firm-specific synergies,
γ. To this end, we need to derive the maximum willingness to pay, vil and
v p , given optimal behavior in the financial market. We can use that R∗i =
(1 + r + ρ)Ii and I min = S − A in equation (2.3) to obtain
Nil (S) = [πA (i) − Fi − S] + [(S − A) − (1 + r + ρ)(S − A)] + A
Acquire
− [πNA (l) + (1 + r)A] = 0,
(2.7)
Do not acquire
for l ∈ {i, p}. Solving for Nil (S) = 0, we get the maximum willingness to pay
for each of the two net gain functions:
vil = [1 + r + ρ]−1 [πA (i) − Fi − πNA (l) + ρA].
(2.8)
Then, we use that R∗l = (1 + r + ρ)Il and I min = S in equation (2.5) to obtain
N p (S) = [π p (p) − E] − S − [(1 + r + ρ)S − S] − Fp −
Acquire
0
.
(2.9)
Do not acquire
Solving for N p (S) = 0, we get:
v p = [1 + r + ρ]−1 [π p (p) − Fp − E].
(2.10)
Having obtained the valuations vii , vip and v p , we can now turn to examining
how synergies γ affect who will acquire the target firm.
The valuations of MNEs, vip and vii , increase monotonically in γ, whereas
the PE-firm valuation v p is independent of γ. Thus, we can state the following
Lemma.
Lemma 3 There exists a unique γPE defined from vip (γPE , ·) = v p and a unique
γPRE defined from vii (γPRE , ·) = v p .
To explain and illustrate our results, we will make use of the following
assumption which, for instance, holds in the LQC model.
32
Assumption 2 γPRE > γPE > 0.
The intuition behind this assumption is the market power effect. The assumption holds if vip > vii or if πNA (p) < πNA (i). This latter inequality holds
at low synergies as profits for a non-acquiring incumbent will be greater when
another incumbent buys the target and prevents an increase in the number of
firms on the market.
This assumption allows us to derive a simple graphical solution where all
types of relevant equilibria are present. In Figures 2.2(i) and 2.2(ii), we derive
the equilibrium ownership for which the size of synergies γ varies. We can
state the following proposition.
Proposition 2.1 Equilibrium ownership depends on firm-specific synergies, γ,
as follows:
(i) A PE-firm will acquire the target at price S∗ = v p if γ ∈ (0, γPE ).
(ii) An MNE concentrating acquisition will take place at price S∗ = v p if γ ∈
[γPE , γPRE ).
(iii) An MNE preemptive acquisition will take place at price S∗ = vii if γ ≥
γPRE .
To see the intuition behind this proposition, start with considering Figure
2.2. The top part of the figure illustrates the net value for an MNE of deterring
a PE-firm from acquiring the target (vip − v p ) and the net value for an MNE of
deterring another MNE from acquiring the target (vii − v p ). These functions
are given by
vip − v p = [1 + r + ρ]−1 [πA (i) − Fi − πNA (p) + ρA − π p (p) + Fp + E], and
(2.11)
vii − v p = [1 + r + ρ]−1 [πA (i) − Fi − πNA (i) + ρA − π p (p) + Fp + E]. (2.12)
When the synergies are low γ ∈ (0, γPE ), an MNE’s valuation is lower than a
PE-firm’s. This is illustrated in Figure 2.2(i) where both the vip − v p and the
vii − v p curve is below zero. Thus, without sufficient firm-specific synergies,
an MNE acquisition is not profitable and instead a PE-firm will acquire the
target firm (Figure 2.2(ii))
From Assumption 1, the valuation vip increases in synergies, γ. Indeed, the
profit as an acquirer, πA (i), increases in γ:
d(vip − v p )
dπA (i)
= (1 + r + ρ)−1
> 0.
dγ
dγ
(2.13)
A further increase in synergies γ will thus make an MNE concentrating acquisition strictly profitable as vip − v p > 0. The equilibrium sales price is then
S∗ = v p . This is illustrated in Figure 2.2(ii), where MNE concentrating acquisitions occur in the region γ ∈ [γPE , γPRE ).
33
(i) The acquisition game
Net value
v ip
I6 & I5
I4
0
γP E 1
(ii) Equilibrium ownership
PE-firm
MNE
(concentrating)
∗
∗
S = vp
−
vp
vp
−
v ii
I1 & I2 & I3
γ P RE
Synergies
MNE
(preemptive)
S ∗ = vii
S = vp
γP E 1
γ max
γ P RE
γ max
Synergies
(iii) Equilibrium ownership
Exit cost
MNE
MNE
(concentrating)
∗
S = vp
(preemptive)
PRE
S ∗ = vii
PE
PE-firm
S ∗ = vp
0
E P RE (γ)
E P E (γ)
γP E 1
γ P RE
γ max
Synergies
Figure 2.2. Illustrating equilibrium ownership as a function of synergies and exit costs
34
Finally, we turn to the case of high levels of firm-specific synergies γ ∈
(γPRE , γmax ). Using Assumption 1, we can note that the preemptive valuation of MNEs vii will increase more than the concentrating valuation vip since
increasing synergies do not only increase the product market profit as an acquirer but also decrease the product market profit as a non-acquirer. Thus,
the preemptive valuation vii is not only driven by the benefits of obtaining a
strong position in the product market as an acquirer, but also by the preemptive
motive for avoiding a weak position as a non-acquirer:
d(vii − v p )
dπNA (i)
−1 dπA (i)
> 0.
(2.14)
= (1 + r + ρ)
−
dγ
dγ
dγ
It then follows that a further increase in firm-specific synergies into the region
γ ∈ (γPRE , γmax ) will make a preemptive acquisition strictly profitable as vii −
v p > 0 (Figure 2.2(i)). A fierce bidding competition among MNEs then drives
the equilibrium sales price to S∗ = vii (Figure 2.2(ii)).
2.4.2 Financial Market Development
Financial market development (as captured by E) is likely to favor PE-firms
over MNEs in the competition for the target firm. PE-firms which own the target and its assets temporarily incur exit costs E when reselling the target firm.
Financial market development in a country is important when reselling assets
as PE-firms often list the target on a stock exchange or resell it to other incumbent firms. These resales are easier to undertake if financial markets are better
developed. In contrast, this cost is not present for MNEs which are permanent
owners of the target and its assets. We have the following proposition.
Proposition 2.2 More developed financial markets (lower E) favor PE-firms
over MNEs.
To see how exit costs E affect equilibrium ownership, consider Figure 2.2(iii).
This figure shows how equilibrium ownership is jointly determined by the
synergies γ and the exit cost, E, faced by PE-firms. Let E PE (γ) be the PEcondition defined from vip = v p , and let E PRE (γ) be the preemption condition
defined from vii = v p . Solving for E in each equation we have:
E PE (γ) = −πA (i) + πNA (p) + πP (p) − ρA − Fp + Fi , and
(2.15)
(γ) = −πA (i) + πNA (i) + πP (p) − ρA − Fp + Fi .
(2.16)
E
PRE
The loci associated with both the PE-condition E PE (γ) and the preemption
condition E PRE (γ) are downward-sloping in the γ−E space. Since the profit of
the acquirer increases in γ, the profits of a non-acquirer πNA (l) are decreasing
in γ, and a lower exit cost for the PE-firm is needed to balance the incumbent’s
higher value of obtaining the target.
35
Preemptive acquisitions occur to the right of the preemption locus E PRE (γ).
Here, the exit costs and synergies for the MNEs are too high for PE-firms
to compete. High levels of synergies make the acquisition very attractive to
MNEs, which is why they compete with each other for the ownership of the
target firm. Below this point for lower values of both E and γ, PE-firms can
compete with MNEs for ownership of the target firm. Until we hit the E PE (γ)
border, MNEs are winning the auction and performing entry deterring acquisitions between the PE-locus E PE (γ) and the preemption locus E PRE (γ). This
is since for any given E, the synergies γ are still high enough (and likewise,
for any given γ, E are still too high). Below the line E PE (γ), the combinations
of the costs and synergies make it more profitable for the PE-firms to own the
target firm, so we have PE-firm acquisitions to the left of the locus E PE (γ).
Thus, it follows that for higher exit costs E, MNEs are more likely to end up
acquiring the target firm.
2.4.3 Financial Market Conditions
Financial market conditions (r and ρ) vary over the business cycle and refer
to firms’ cost of acquiring outside funding for their projects and acquisitions.
PE-firms are dependent on the availability of cash from outside investors to be
able to make acquisitions. MNEs on the other hand, have retained earnings,
A, which they can use to finance acquisitions. Thereby, financial market conditions will affect the two types of buyers differently. Our model provides the
following predictions.
Proposition 2.3 Financial market conditions affect equilibrium ownership as
follows:
(i) Changes in the interest rate r do not affect who ends up buying the target,
but it does affect the equilibrium acquisition price S∗ .
(ii) A lower risk premium, ρ, makes an acquisition by a PE-firm more likely.
To see how r affects equilibrium ownership, note that both E PE (γ) and
are independent of the interest rate r so r does not affect equilibrium ownership of the target firm. The acquisition price, however, is affected
by r. To see this, note that
E PRE (γ)
dS∗ (i) dvii
=
= −(1 + r + ρ)−2 [πA (i) − Fi − πNA (i)] < 0, and
dr
dr
dS∗ (p) dv p
=
= −(1 + r + ρ)−2 [πP (p) − Fp − E] < 0.
dr
dr
(2.17)
(2.18)
The higher r is, the lower are the equilibrium acquisition prices. Intuitively,
while MNEs have retained earnings A to use as payment in acquisitions, these
retained earnings have an opportunity cost dictated by the same interest rate r
that determines the cost of borrowing. Hence, having retained earnings does
36
not lower the cost for an MNE to make the acquisition because the MNE accounts for the opportunity cost of using these funds. Perfect financial markets
make an MNE indifferent between using retained earnings and financing the
acquisition using outside financing. The equilibrium acquisition prices, however, do respond to financial market conditions since higher costs associated
with obtaining financing for an acquisition will lower the profitability of making the acquisition and, hence, lower the valuations that MNEs and PE-firms
put on the target.
Let us now turn to how the risk premium ρ affects equilibrium ownership.
Inspecting E PE (γ) and E PRE (γ), it follows that both E PE (γ) and E PRE (γ) shift
downwards as the relative value of the MNE’s retained earnings used as payment in the acquisition increases (when ρ increases). With a higher risk premium ρ, MNEs gain an advantage over PE-firms that do not have access to
retained earnings to be used to partly finance the acquisition. In effect, an
MNE "saves" ρA by being able to use retained earnings instead of going to
outside investors. This implies that in the event of high risk premia, MNEs
are able to pay more for the target firm in a given acquisition making it more
likely that an acquisition by an MNE takes place.
2.5 Welfare Effects of Restricting PE-Firms
To assess the welfare effects of restricting PE-firms for the domestic country,
we will compare welfare for domestic consumers and the target firm in different market structures. These are (i) a non-discriminatory (ND) policy where
both PE-firms and MNEs are allowed to compete for the ownership of the target firm and (ii) a discriminatory (D) policy which prohibits PE-firms from
buying domestic assets.
To this end, we let PS(l) and CS(l) denote the producer and consumer surpluses when the owner of the target firm is either the MNE (l = i) or the PEfirm (l = p). Welfare under ownership l will then be the sum of the producer
(target) and consumer surpluses, that is
W (l) = PS(l) +CS(l) = S∗ (l) +CS(l).
(2.19)
In each of the six potential equilibrium ownership outcomes in Table (2.1), we
can investigate the difference between ND and D policies (denoted by subscripts).
I1(vii > vip > v p ): MNEs compete against each other for ownership of the target. Restricting PE-firms from participating in the auction for the target
neither affects equilibrium ownership nor the acquisition price. We have
WND = WD = vii +CS(i).
I2(vii > v p > vip ): There are two equilibria under the ND policy. There is
one where MNEs compete with each other, and here PE-firms do not
affect the outcome. There is also another equilibria where no bidding
37
competition between MNEs occur since all MNEs coordinate on bidding
below v p . For this equilibrium a PE-firm may end up owning the target.
Then, under the D policy, PE-firms would be restricted from taking over
the target. As before, we focus on the equilibrium with an MNE as the
acquirer. We have WND = WD = vii +CS(i).
I3(vip > vii > v p ): Here v p is too low for the MNEs to respond by bidding up
the price to vip . Competition is again between MNEs. The absence of
PE-firms in the auction under a D policy has no effect on the outcome.
We have WND = WD = vii +CS(i).
I4(vip > v p > vii ): The PE-firm acts as an active threat in the auction as its
valuation has an edge over vii . This is since the MNE’s valuation of the
firm when the alternative buyer is a PE-firm, vip , is the highest valuation
and the sales price S∗ under MNE ownership goes up from vii to a strictly
higher v p under the ND policy. Under the D policy, the threat of a PEfirm taking over the target is missing, and so the sales price remains at
vii . We have WND = v p +CS(i) and WD = vii +CS(i).
I5(v p > vii > vip ): Under the ND policy, the target is acquired by a PE-firm
at price S∗ = v p as opposed to by an MNE at price S∗ = vii under the D
policy. We have WND = v p +CS(p) and WD = vii +CS(i).
I6(v p > vip > vii ): Once more, the ND policy allows for a PE-firm to acquire
the target instead of an MNE at a price of v p rather than vii . We have
WND = v p +CS(p) and WD = vii +CS(i).
Thus, depending on the rankings of the valuations, the D policy will have
different effects. The domestic target firm gets the same sales price S∗ = vii
for Inequalities I1 − I3 under both policies, but under I4 − I6 the target firm
gets a lower sales price of vii instead of v p . Consumer surplus depends on
two components: the number of firms that service the market (n or n + 1)
and the synergies between the owners and the target firm (γ). Both of these
will affect product market prices on the domestic market. Under I1-I4 with
an MNE acquisition, there are fewer firms serving the market since the target
is acquired by the MNE (we go from n + 1 to n firms in the market). This
reduces consumer surplus. However, the owner will operate the target with
synergies equal to γ which, for high values, can overturn the negative effect
market concentration may have on consumer surplus. Note, however, that the
D policy does not affect the outcome, so in the comparison between the ND
and the D policies, consumer surplus remains unchanged for I1 − I4. In sum,
we have:
⎧
⎪
⎨CS(p) −CS(i) + v p − vii for I5 − I6.
(2.20)
WND −WD = v p − vii
for I4.
⎪
⎩
0
for I1 − I3.
Figure 2.3 illustrates this outcome graphically. Note first that Figure 2.3(i)
shows that concentrating MNE acquisitions can occur for synergies below
38
(i) The acquisition game
Net value
v ip
I6 & I5
0
S ∗ = vp
vp
v ii
I4
γP E 1
(ii) Equilibrium ownership
PE-firm
−
MNE
(concentrating)
S ∗ = vp
γP E 1
−
vp
I1 & I2 & I3
γ P RE
γ max
Synergies
MNE
(preemptive)
S ∗ = vii
γ P RE
(iii) Consumer surplus
CS
γ max
Synergies
( i)
CSN D
CS(p)
CS(p) − CS(i)
C SD
0
(iv) Total surplus
plus
γP E 1
γ P RE
γP E 1
γ P RE
Synergies
W
( i)
WN D
W (p)
γ max
CS(p) − CS(i)
vp − vii
WD
0
γ max
Synergies
Figure 2.3. Illustrating changes in consumer and total surplus
39
unity, i.e. where γ is above the threshold γPE but below one. Welfare inefficient MNE acquisitions, compared to acquisitions by PE-firms, can therefore
lead to decreased consumer surpluses for two reasons. First, productivity in
the MNE acquired firm may be lower than the productivity after an acquisition
by a PE-firm which will lead to an increase in equilibrium consumer prices.
Second, an MNE acquisition leads to a market concentration compared to an
acquisition by a PE-firm (n firms instead of n + 1 firms). This induces an additional decline in the consumer surplus. The aggregate effect is illustrated in
Figure 2.3(iii) as a discrete change downwards by CS(p) − CS(i) in the consumer surplus at γPE . Figure 2.3(iv) illustrates the effect on total surplus by
accounting for the change in producer surplus as well. For regions I4 − I6,
producer surplus is reduced by v p − vii . This leads to an additional discrete
drop in total surplus equal to v p − vii at γPE . Note that for I1 − I3 the WD and
WND curves run in parallel. We thus have the following proposition:
Proposition 2.4 Restricting PE-firms from bidding for domestic targets can
decrease consumer and total surplus when synergies between the MNEs and
the target’s assets (γ) are sufficiently low.
Thus, acquisitions by PE-firms can be welfare improving both because PEfirms could be more efficient than foreign MNEs at running the target firm, and
because an acquisition by a PE-firm prevents a concentration of the market.
Hence, a policy restricting foreign PE-firms from acquiring domestic targets
could be counterproductive.
It is worthwhile noting, however, that it is possible that there exists parameter values such that acquisitions by PE-firms also can reduce consumer
surplus. Suppose that the fixed costs of private equity firms Fp are low, and
that synergies γ are high. Then a PE-firm could use their fixed cost advantage
to increase the threshold associated with an acquisition by an MNE. Since consumer surplus is increasing in synergies arising in MNE acquisitions it follows
that consumers may be worse off if a PE-firm acquires the target. The effect
on total surplus however remains ambiguous as the total effect on producer
surplus is ambiguous.
2.6 Extensions
2.6.1 Different Risk Premia
As stated above, we use the additional borrowing cost risk premium to account
for credit risks that could be different for the two types of borrowers (MNEs
and PE-firms). In this section we allow for banks to charge the same underlying interest rate, but attach a different risk premium to the MNEs (denoted
as ρMNE ) and PE-firms (denoted as ρPE ) to formulate the difference in total
40
borrowing costs between PEs and MNEs. We then get
Nil (S) = [πA (i) − Fi − S] + [(S − A) − (1 + r + ρMNE )(S − A)] + A
Acquire
− [πNA (l) + (1 + r)A] = 0
(2.21)
Do not acquire
for l ∈ {i, p}. To see how equations E PE (γ) and E PRE (γ) would then be
updated to reflect the differences we first calculate the different valuations in
this setting to evaluate E PE (γ):
vip = [1 + r + ρMNE ]−1 [πA (i) − Fi − πNA (p) + ρMNE A] .
(2.22)
The valuation for the PE-firm then takes the value:
v p = [1 + r + ρPE ]−1 [π p (p) − Fp − E].
(2.23)
The expression for E PE (γ) would then become:
[πA (i) − Fi − πNA (p) + ρMNE A][1 + r + ρPE ]
− [π p (p) − Fp ].
[1 + r + ρMNE ]
(2.24)
Let us consider the situation where we move from a setting where the premia
are equal to the case of ρPE > ρMNE . We have
E PE (γ) = −E =
∂E PE (γ) [πA (i) − Fi − πNA (p) + ρMNE A]
=
> 0.
∂ρPE
[1 + r + ρMNE ]
(2.25)
Thus, if we allow for different risk premia for different types of buyers in a
setting where ρPE > ρMNE , we will have the E PE (γ) curve shift down, thus,
decreasing the combinations of γ and E that correspond to an acquisition made
by a PE-firm. This implies that MNEs gain an advantage over PE-firms compared to in the setting where risk premia are equal.
2.7 Concluding Remarks
The globalization process implies that new business models spread wider and
faster over the world than ever before. More than 8 percent of all cross-border
acquisitions that took place during 1998-2010 were cross-border acquisitions
undertaken by PE-firms. And there was substantial variation over time, across
countries and across sectors in the share of cross-border acquisitions that were
acquisitions by PE-firms as opposed to those by MNEs.
In this paper we have developed a formal model of competition for domestic
assets between MNEs and PE-firms to better understand possible causes for
41
this variation. In our model, MNE advantage lies in firm-specific synergies
and retained earnings, whereas PE-firms are good at reorganizing target firms.
We showed that lower interest rates do not work in favor of PE-firms, but that
lower risk premia and better financial market developments do. Stronger firmspecific synergies, however, favor MNEs. Performing a welfare analysis, we
show that a policy of restricting PE-firms from buying domestic assets can be
counterproductive.
While it is well established that cross-border mergers and acquisitions play
a key role in the global industrial development and restructuring process, our
model is, to the best of our knowledge, the first to incorporate PE-firms that
compete with MNEs for domestic firms that are up for sale. As such, we
identify an important role of PE-firms as challengers of existing international
oligopolies. These results suggest that policies improving the international
market for corporate control would be preferred to policies restricting crossborder acquisitions by PE-firms.
In the analysis, we have assumed that the PE-firm exits its investment by
means of listing the firm on a stock exchange or through a sale to an outside
investor. However, a substantial share of all exits by PE-firms are exits through
trade sales to incumbents. In that setting, while the main results of the paper
will still be valid, an acquisition by a PE-firm, from a welfare point of view,
will not restore the intensity of competition in the product market in the long
run. However, as shown by Norbäck, Persson and Tåg (2013), exits by trade
sales to incumbents lead to incentives to invest more in restructuring to enhance bidding competitions for target firms in the exit stages. This tends to
benefit consumers but can hurt producers.
42
References
Bertrand O. and Zitouna H., (2006). Trade Liberalization and Industrial Restructuring: The Role of Cross-Border Mergers and Acquisitions, Journal of
Economics and Management Strategy, 15(2), 479-515.
Bjorvatn K., (2004). Economic Integration and the Profitability of CrossBorder Mergers and Acquisitions, European Economic Review, 48(6), 12111226.
Blonigen B.A., (1997). Firm-Specific Assets and the Link between Exchange
Rates and Foreign Direct Investment, American Economic Review, 87(3), 447465.
Davis S.J., Haltiwanger J., Handley K., Jarmin R., Lerner J., and Miranda J.,
(2014). Private Equity, Jobs and Productivity, American Economic Review,
104(12), 3956-3990.
Falvey R., (1998). Mergers in Open Economies, The World Economy, 21(8),
1061-1076.
Farrell J. and Shapiro C., (1990). Asset Ownership and Market Structure in
Oligopoly, RAND Journal of Economics, 21(2), 275-292.
Fumagalli E. and Vasconcelos H., (2006). Sequential Cross-Border Mergers,
International Journal of Industrial Organization, 27(2), 175-187.
Gorbenko A.S. and Malenko A., (2014). Strategic and Financial Bidders in
Takeover Auctions,The Journal of Finance, 69(6), 2513-2555.
Head K. and Ries J., (2008). FDI as an Outcome of the Market for Corporate
Control: Theory and Evidence, Journal of International Economics, 74(1),
2-20.
Head K. and Ries J., (1997). International Mergers and Welfare Under Decentralized Competition Policy, Canadian Journal of Economics, 30(4b), 11041123.
Hege U., Lovo S., Slovin M.B., and Sushka M.E., (2012). Asset Sales and the
Role of Buyers: Strategic Buyers versus Private Equity, MPRA Paper 39549,
University Library of Munich, Germany (2012).
Horn H. and Persson L., (2001). The Equilibrium Ownership of an International Oligopoly, Journal of International Economics, 53(2), 307-333.
Jehiel P. and Moldovanu B., (1999). Resale Markets and the Assignment of
Property Rights, Review of Economic Studies, 66(4), 971-991.
43
Jensen M.C., (1989). Eclipse of Public Corporation, Harvard Business Review, 67, 61-74.
Jensen M.C., (1986). Agency Costs of Free Cash Flow, Corporate Finance,
and Takeovers, American Economic Review, 76(2), 323-329.
Kaplan S. and Strömberg P., (2009). Leveraged Buyouts and Private Equity,
Journal of Economic Perspectives, 23(1), 121-146.
Lommerud K.E., Straume O.R., and Sorgard L., (2006). National versus International Mergers in Unionized Oligopoly, RAND Journal of Economics, 37(1),
212-233
Maksimovic V. and Philips G., (2002). Do Conglomerate Firms Allocate Resources Inefficiently Across Industries? Theory and Evidence, Journal of Finance, 57(2), 721-767.
Martos-Vila M., Rhodes-Kropf M., and Harford J., (2013). Financial vs.
Strategic Buyers, NBER Working Papers 19378, National Bureau of Economic Research.
Mattoo A., Olarreaga M. and Saggi K., (2004). Mode of Foreign entry, Technology Transfers, and FDI Policy, Journal of Development Economics, 75(1),
95-111.
Neary J.P., (2007). Cross-Border Mergers as Instruments of Comparative Advantage, The Review of Economic Studies, 74(4), 1229-1257.
Norbäck P.-J. and Persson L., (2007). Investment Liberalization - Why a Restrictive Cross-Border Merger Policy can be Counterproductive, Journal of
International Economics, 72(2), 366-380.
Norbäck P.-J. and Persson L., (2008). Globalization and Profitability of CrossBorder Mergers and Acquisitions, Economic Theory, 35(2), 241-266.
Norbäck P.-J., Persson L., and Tåg J., (2013). Buying to Sell: Private Equity
Buyouts and Industrial Restructuring, CESifo Working Paper Series No. 4338
Raff H., Ryan M., and Stähler F., (2009). The Choice of Market Entry Mode:
Greenfield Investment: M&A and Joint Venture, International Review of Economics and Finance, 18(1), 3-10.
Tåg J., (2012). The Real Effects of Private Equity Buyouts. In The Oxford
Handbook of Private Equity (2012), edited by Douglas Cumming. Oxford:
Oxford University Press.
44
Appendix
Proof of Lemma 1
First, bi ≥ max{v p , vii , vip } is a weakly dominated strategy. No owner wants
to post a bid above its valuation of obtaining the assets, and the assets will
always be sold.
Inequality I1 (vii > vip > v p ): Since vip > v p , an MNE will always have an
incentive to outbid PE-firms. The MNEs will then bid up the price to vii to
prevent a rival from obtaining the assets. An MNE will obtain the assets.
Inequality I2 (vii > v p > vip ): Since v p > vip , the outcome depends on what
an MNE believes will happen if it does not win. If it believes that another
MNE will win, MNEs will then bid up the price to vii and an MNE will obtain
the assets. If it believes that a PE-firm will win, then since v p > vip , the PEfirms will bid up the price to v p and an MNE will obtain the assets.
Inequality I3 (vip > vii > v p ): Since vip > v p , an MNE will always have
an incentive to outbid PE-firms. The MNEs will then bid up the price to vii
to prevent a rival from obtaining the assets. An MNE will obtain the assets.
Since MNEs realize that a PE-firm will never obtain the assets (vip > v p ), the
price will not be bid up to vip .
Inequality I4 (vip > v p > vii ): Since vip > v p , an MNE will always have an
incentive to outbid PE-firms and bid up the price to slightly above v p . However, only one MNE has this incentive, since no other MNE wants to outbid
him or her (v p > vii ). An MNE will then obtain the assets at price v p .
Inequality I5 (v p > vii > vip ): Since v p > vip , no MNEs will want to outbid
the PE-firms. The PE-firms will then bid up the price to v p and a PE-firm will
obtain the assets.
Inequality I6 (v p > vip > vii ): Since v p > vip , no MNEs will want to outbid
the PE-firms. The PE-firms will then bid up the price to v p and a PE-firm will
obtain the assets.
45
3. Cross-border Leveraged Buyouts1
3.1 Introduction
Cross-border leveraged buyouts (LBOs) add up to a substantial share (8.3%)
of all cross-border M&As that took place globally during 1998-2008. These
LBOs are orchestrated by private equity (PE) firms which compete with established incumbents to buy, restructure, and sell firms generating heated discussions regarding their role in the economy, as well as potential benefits for
the countries receiving LBO investments. Most of the concerns regarding the
presence of LBOs in a country are centered on immediate job losses and the
short length of their investment horizon.2 At the same time, there are empirical studies that emphasize the role of LBOs in the creation of efficient firms.
Baker and Wruck (1989), and Luehrman (2007) present case studies on the
various ways LBOs can improve firm performance. More recently Davis et al.
(2013) present results from a more comprehensive data analysis and confirm
productivity gains at LBO targets as a result of restructuring within the firm.
The share of LBOs in all the global M&As varies substantially over time
and across countries. Figure 3.1 shows that for the time period 1998-2008 the
share varies between 3.8% and 16.3%. Figure 3.2 shows that the share also
varies between countries, from as low as 1% up to 20% in Luxembourg. The
purpose of this paper is to improve our understanding of the relative importance of LBOs in different countries. I use a theory of cross-border M&As
that incorporates LBOs in order to derive testable hypotheses which are then
investigated using a comprehensive dataset covering all globally performed
majority-owned cross-border M&As.
1 Financial
support from the Wallander-Hedelius-Browaldh Foundation is gratefully
acknowledged. I thank Nils Gottfries for valuable discussions and supervision, in addition
to Pehr-Johan Norbäck, Lars Persson, Thomas Rønde, Joacim Tåg, and the participants of
Sudswec 2012 for their comments.
2 See Kaplan (1989) and Lichtenberg and Siegel (1990) for case studies on the negative
effects with regards to employment. Davis et al. (2013) find that while LBOs are associated
with job separations, they also create higher job turnover. Unlike previous studies that have
found lower averages, Strömberg (2007) uses data with better coverage to show that the median
length of LBO investment is around 9 years.
46
Figure 3.1.
The number of Cross-border M&As and LBOs over time.
Figure 3.2. The mean share (1998-2008) of all cross-border M&As that are LBOs per country, for a
select number of countries.
The model setup follows closely Norbäck and Persson (2009), Norbäck et al.
(2013), and Baziki et al. (2015). A domestic firm is up for sale in a country
where several foreign Multinational Enterprises (MNEs) compete in the product market. These incumbent MNEs and foreign PE firms enter an auction for
the ownership of the domestic firm. After the acquisition, the owner has the
opportunity to restructure the domestic firm to increase its competitiveness.
This restructuring affects owner’s profits positively and competitors’ profits
negatively. Next, if the target was initially bought by a private equity firm, it
47
is put on the market again and sold to an incumbent.3 All firms servicing the
industry then compete in the product market.
As in the models of Norbäck and Persson (2009) and Norbäck et al. (2013),
the target firm acquired by a PE firm undergoes a strictly higher level of restructuring compared to a firm acquired by an MNE. This could contribute
to a better understanding of the role of LBOs in international restructuring,
and suggest a mechanism behind the existing empirical findings that suggest
higher productivity gains for target firms that undergo LBOs. Private equity
firms buy the domestic firm with the intention to sell, which implies that they
would like to maximize the sales price of the domestic firm. The sale price
increases with higher level of restructuring due to two reasons. First because
it increases the value of owning the restructured target firm, and secondly because the incumbent would like to prevent becoming rivals with the restructured target firm in the product market competition. As an outcome, private
equity firms restructure the target firm more intensely compared to foreign
incumbents, which further reduces expected profits for competing incumbent
firms once the domestic firm is sold by the private equity firm back into the
market. This expectation of intense restructuring by private equity firms motivates incumbents to bid more aggressively in the initial acquisition auction, so
as to prevent the private equity firm from acquiring the firm in the first place.
I use this model to derive testable implications. First, I show that high transaction costs in a country should make it relatively less attractive for private equity firms to operate there, thereby decreasing the share of cross-border LBOs.
PE firms buy firms with the intention to sell them, so they will incur transaction costs first at the initial take over, and a second time when they resell the
target firm after restructuring. In contrast, a firm buying to keep the target only
incurs the transaction cost once. Thus, rising transaction costs should affect
the different types of buyers asymmetrically, and therefore reduce the share of
LBOs.
Next, I find that cross-border LBO share is also dependent on the market
structure of the domestic economy. Using a Cournot setup, I show that for a
given number of firms in the market, the value of the PE firm’s restructuring
advantage goes up with increasing sales revenue per firm, which makes LBO
outcomes more prevalent for high market sizes. For given market sizes, the
larger the number of active firms in the market, the smaller the market share
of each firm, including the target, will be. This makes the negative externality
of not owning the target smaller, which removes the PE firm’s advantage. As
a result a higher number of firms results in a smaller share of LBO outcomes.
Since the number of firms in the product market increase with higher international market integration, I conclude that a higher degree of international
openness should decrease LBO shares in the target economy.
3 See
48
Section 3.3 for a discussion on an alternative exit strategy .
Finally, making use of a similar setup in Baziki et al. (2014), I link a
stronger level of property rights protection to a lower likelihood of private
equity buyouts relative to incumbent firm buyouts. When property rights are
well protected, MNEs will be able to restructure more efficiently, since they
can then use firm-specific assets both when restructuring and during the product market competition without the fear of potential knowledge spillovers to
rivals. Therefore, better property rights will make the target relatively more
attractive to MNEs, and decrease the share of takeovers that end up resulting
in LBOs.
Using data on majority-owned and completed cross-border M&As and crossborder LBOs from the Capital IQ database for the years 1996-2008, I test
these predictions for the share of LBO outcomes. The explanatory variables
are country level polity measures. As a proxy for transaction costs, I use an
indicator from the World Bank Development Indicators that captures the red
tape and corruption in each of the countries in the form of business registration costs. For international market integration, I use an openness index from
the United Nations that measures imports plus exports as a share of GDP. For
property rights, I use a measure constructed by the Heritage Foundation that
measures how well private property is protected under the law and how wellenforced these laws are. I find that in line with the theory, higher transaction
costs, international openness, and property rights all decrease the cross-border
LBO shares of target countries.
Next, panel level logit regressions are performed to predict whether a given
cross-border transaction is a cross-border LBO using the three main variables
and country level controls. The finding here is that higher transaction costs decrease the probability of an LBO incident, while higher levels of international
openness increase LBO incidents.
Navaretti and Venables (2004) summarize the previous literature on the activities of MNEs. This paper adds to the previous literature on M&As by introducing an international cross-border investment setting, where the predictions
are empirically tested on a globally comprehensive database.
La Porta et al. (1998) consider whether differences in investor protection
across countries are due to different commercial law traditions and if variations in the application of law explain the diversity in financing or corporate
structures in 49 different countries. In contrast, this paper looks at the effect
of time-varying polity indicators on all cross-border investments in the world.
La Porta et al. (1998) do not find cross-country differences in the presence of
different types of investors, whereas this paper shows that there are asymmetries in how attractive a country may be to one type of investor over another.
This could explain differences in the share of PE type investors in different
countries.
La Porta et al. (1997) examine the impact of legal systems on external
financing in 49 different countries and find that better legal protection leads to
more external financing. To relate this result to the paper at hand, there are two
49
issues to consider. The CapitalIQ data does not provide details on financing
structures for the M&As performed, so it is not possible to identify takeovers
performed by MNEs that are relatively more dependent on debt financing.
Next, since I focus on global cross-border investments, the investors are not
likely to be tied down by the particulars of the financial conditions of the
target country, but instead may get their financing from outside. Abstracting
from these differences, if better property rights could be thought of as a proxy
for legal protection, then this paper differs from La Porta et al. (1997) by
showing that better property rights reduce the presence of LBOs performed by
PE firms, which are more dependent on external financing than incumbents.
Section 3.2 details the setup of the auction and presents the details of the
theory. Section 3.3 discusses variations on the assumptions of the setup. Section 3.4 presents both the empirical models and their results, and, finally, Section 5.6 concludes this paper.
3.2 A Theory of Cross-border M&As and LBOs
The model follows closely Norbäck and Persson (2009) and Norbäck et al.
(2013) and is used here to derive testable implications for variations in LBO
presence across countries. There is a domestic firm up for sale in an oligopoly
market where there are I = 1, 2, .., i, ..., N foreign firms taken to be ex ante
symmetric competing in the product market. In Stage 1, the domestic firm is
either sold to one of the incumbents, or to a private equity firm, from the set
P = 1, 2, .., p, ..., P.
The acquiring firm invests k in the restructuring of the target firm in Stage
2. Restructuring increases the profits of the owner, but decreases the profits
of the competitors. Both types of owners face a cost for restructuring, but PE
firms enjoy a cost advantage.
In Stage 3, after restructuring, the private equity firm p exits its investment
by selling the domestic firm back to one of the foreign firms.
Finally, in Stage 4, the firms on the market compete. Since all the firms servicing the market are MNEs, they all have to incur costs to access the domestic
market.
3.2.1 Stage 4: Product Market Competition
In the product market competition, incumbent firms, denoted by i, will choose
an optimal action xi to maximize their product market profit, Πi (xi , x−i , k). The
profit of an incumbent firm is a function of three factors; the firm’s action xi ,
its N − 1 competitors’ optimal strategies x−i , and the restructuring performed
on the target firm k.
50
The following first order condition gives profit maximizing equilibrium actions for all MNE firms participating in the product market competition given
the level or restructuring:
∂Πi ∗ ∗
(x , x ; k) = 0.
(3.1)
∂xi i −i
Denote the firm acquiring the target firm as A and all the non-acquiring firms
as NA, since the incumbent firms are taken to be ex-ante symmetric. This
also implies that the optimal actions are symmetric. Define the reduced-form
profits which depend on the amount of restructuring as follows:
RA (k) ≡ ΠA (xA∗ (k) , x∗NA (k) , k), RNA (k) ≡ ΠNA (x∗NA (k) , xA∗ (k, ) , k). (3.2)
Assume that the amount of restructuring performed on the target increases the
acquirer’s equilibrium profits in Stage 4, and decreases competitors’ profits4
3.2.2 Stage 3: Sale of the Restructured Domestic Firm
If the domestic firm was acquired by a private equity firm in Stage 1, then it
will be sold to one of the N incumbent firms in this stage where the firms will
declare their offers for the target simultaneously. The offers, and therefore
the sale price for the target S∗ (k) will be equivalent to their valuation of the
target5 , which will be symmetric across incumbents:
S∗ (k) = RA (k) − RNA (k) − T.
(3.3)
This valuation depends on the difference between acquirer and non-acquirer
profits, as well as T , a transaction cost associated with an acquisition. This
cost includes the cost of identifying and evaluating the domestic firm, legal
and administrative costs incurred during the acquisition process, as well as the
costs of potential appropriation problems.
3.2.3 Stage 2: Restructuring
The restructuring intensity will vary depending on who acquires the domestic
firm in Stage 1. Therefore, restructuring by incumbents will be treated separately from restructuring by private equity firms. If a private equity firm p
obtains the domestic firm in Stage 1, the private equity firm invests kP in the
restructuring phase. If, on the other hand, a foreign firm i obtains the domestic
4I
present a Linear-Quadratic Cournot model below where this assumption holds.
a proof of the result, see Appendix 3.5.
5 For
51
firm in Stage 1, the acquiring firm invests kA in the restructuring stage.6 Assume that the cost function C(k) for the restructuring k performed is a strictly
convex function (i.e. C (k) > 0 and C (k) > 0).
Incumbent Acquisition in Stage 1 (M&A)
An incumbent firm would like to maximize their product market profits given
its type-specific restructuring cost CA (k):
Max : RA (k) −CA (k).
{k}
(3.4)
Taking RA (k) −CA (k) as strictly concave in the amount of restructuring k, the
optimal restructuring by an incumbent is then determined by:
dRA
= CA (kA∗ ).
dk
(3.5)
Private Equity Acquisition in Stage 1 (LBO)
Assume that the private equity firm faces a similar type-specific strictly convex
cost function CP (k), but has a smaller restructuring cost in this stage, since
they are restructuring experts (i.e. CP (k) < CA (k)). In this model the private
equity firm does not participate in the product market competition, but instead
seeks to restructure the target firm and exit with a sale in Stage 3. Therefore
the private equity firm maximizes the sales price from Stage 3, S∗ (k) instead
of the reduced form profits of an acquirer, R(k). Using the result on this sales
price from Section 3.2.2, the optimization facing the private equity firm is as
follows:
Max : S∗ (k)−CP (k)
s.t : S∗ (k) = RA (k) − RNA (k) − T.
(3.6)
And the optimal level of restructuring for the private equity is given by:
dS∗ (k) dRA dRNA
=
−
= CP (kP∗ ).
dk
dk
dk
(3.7)
Again, the difference S∗ (k) − CP (k) is taken to be strictly concave in restructuring.
Recall that the amount of restructuring performed benefits the acquiring
firm, whereas it hurts all the non-acquirers. This gives the result that the level
of restructuring performed under private equity ownership should be strictly
higher compared to the level of restructuring performed by an incumbent, kP∗ >
kA∗ . The incumbents in Stage 3 do not only consider owner’s profits in their
valuation of the restructured target firm, but also the profits from being a rival
6 The
optimal levels of restructuring is the same for all i at kA and for all p at kP due to
symmetry.
52
to the restructured firm, which are decreasing in k. This creates a strategic
motivation for the private equity firm to increase the amount of restructuring
performed at the target firm to drive the Stage 3 sales price up. Figure 3.3(i)
demonstrates how the optimal restructuring performed by an incumbent and
private equity firm differ in this setting.
Marginal
Profit and
Marginal
Cost
࡯Ʋ࡭ ሺ࢑ሻ
࡯Ʋࡼ ሺ࢑ሻ
ࡼ
(i) Restructuring
investment in
Stage 2
ࢊࡿ‫כ‬
ൌ ࡯Ʋ࡭ ሺ࢑ሻ
ࢊ࢑
Profits
ࡼ
ࡾ࡭ ࢑ െ ࡯ࡼ ሺ࢑ሻ
(ii) Net
Valuations in
Stage 1
ࡾ࡭ ࢑ െ ࡯࡭ ሺ࢑ሻ
ࢂࡼ
ࢂࡵࡼ െ ࢂࡵࡵ ൐ ૙
݇௉‫כ‬
Figure 3.3. Optimal restructuring and net valuations for the target firm.
3.2.4 Stage 1: M&A or LBO?
The initial acquisition in Stage 1 determines the equilibrium ownership structure. To be able to determine whether an M&A or an LBO takes place and
to evaluate the initial acquisition price paid for the domestic firm under either case, Stage 1 valuations of the MNE and private equity firms need to be
derived and ranked.
At this stage, assume that a private equity firm has transaction costs TP ,
where TP < T . Private equity firms can be expected to face lower transaction
costs for several reasons. First, they are likely to possess unique assets and
insights in terms of identifying underperforming firms in need of restructuring.
53
Second, private equity firms are experts when it comes to buying and selling
companies and are likely to have developed efficient routines and connections
for lowering transaction costs in comparison to incumbents that are likely to
be less accustomed to buying and selling firms.
Denote the value of the target firm to an incumbent firm if the alternative
buyer would be another incumbent as vII . This value will have to consider the
difference in the profits under ownership of and rivalry with the target firm in
addition to accounting for the potential costs of acquisition and restructuring:7 :
vII = RA (kA∗ ) −CA (kA∗ ) − RNA (kA∗ ) − T.
(3.8)
The private equity firm will base its valuation vP on the Stage 3 sales price
of the target:
vP = S∗ (kP∗ ) −CP (kP∗ ) − TP
= RA (kP∗ ) −CP (kP∗ ) − RNA (kP∗ ) − T − TP .
(3.9)
The final valuation, vIP , is the value of the target firm for an incumbent if
alternatively the firm would have been acquired by a private equity firm:
vIP = RA (kA∗ ) −CA (kA∗ ) − RNA (kP∗ ) − T.
(3.10)
Recall that restructuring decreases rivals’ profits and also that the private equity firm will restructure the target more heavily compared to an incumbent.
This means that the profits of the incumbent as a rival to the target owner will
be lower if the target was initially restructured by a private equity firm in Stage
2. This implies that vIP > vII , as depicted in Figure 3.3(ii), and gives three possible cases depending on whether vP is below vIP and vII , between those values
or above them. The resulting equilibrium ownership, l ∗ , and acquisition price
A∗ are described in Table 3.1.
Table 3.1. The equilibrium ownership structure and the acquisition price A.
Definition:
Ownership l ∗ : Acquisition price, A∗ :
Case 1: vIP > vII > vP
I
vII
Case 2: vIP > vP > vII
I
vP
Case 3: vP > vIP > vII
P
vP
In Case 1, vP is the lowest valuation, and therefore the private equity ownership
of the target firm is not an active threat. The incumbents compete against each
other to acquire the target firm, and the winning incumbent pays vII . In Case
7 The first letter in the subscript refers to the type of the acquiring firm (here I for incumbent),
and the second letter refers to the firm who would have obtained the target instead (here I for
another incumbent).
54
2, since vP > vII , private equity firms pose an active threat in the auction, but
the incumbent will be willing to pay at least vP in the auction in order to outbid
the private equity firms. In Case 3, the private equity firms value the target the
most, and will therefore compete against each other, with the winner paying
vP to take over the firm.
Participation Constraints
The presence of both the incumbents and the private equity firms will be determined by the level of transaction costs present in the domestic market. Private Equity firms will participate in the auction as long as the following PE
participation-constraint is satisfied:
VP = RA (kP∗ ) −CP (kP∗ ) − RNA (kP∗ ) > T + TP .
(3.11)
Similarly, the Incumbent participation-constraint can be defined as follows:
VII = RA (kA∗ ) −CA (kA∗ ) − RNA (kA∗ ) > T, or,
VIP = RA (kA∗ ) −CA (kA∗ ) − RNA (kP∗ ) > T
(3.12)
depending on whether the PE takeover is an active threat or not.
PE Threat and Takeover
Going from Case 1 to Case 2 presents a change in the status of PE firms, as
they switch from not actively participating in the auction for the target firm
to becoming an active threat for the incumbents in the auction. Then, the PE
threat-condition defines the border between these two cases where vP = vII
holds (i.e. VP − VII = TP ). The lower the transaction cost TP , the higher the
threat of PE in the auction for the target firm. In this setting, the incumbents
have incentives to prevent an LBO from taking place by outbidding private
equity firms, as it prevents aggressive restructuring by PE (kA∗ < kP∗ ) .
Next, moving from Case 2 to Case 3, is the reversal in the order of valuations vIP and vP where the PE firm starts to value the target firm more than
any other incumbent. For the border between these two cases, define the PE
takeover-condition where vP = vIP (i.e. VP −VIP = TP ).
Given the participation and PE threat- and takeover-conditions, the equilibrium ownership can be depicted in Figure 3.4. Private Equity firms can only
exist in the auction if they are facing combinations of transaction costs such
that T at most can be equal to VP when TP = 0, or when TP = VP when T = 0.
Connecting the two lines gives us the PE participation constraint; a combination of (TP , T ) that defines the base of an isosceles triangle on the lower corner.
For PE to be active participants in the auction, combinations of (TP , T ) have to
fall on, or below this constraint.
For these low enough combinations of transaction costs where PE firms are
actively participating in the auction, the incumbent participation constraint is
55
PE takeover
condition
PE threat
condition
ܶ
ࢂࡼ
Region 4:
No Sale
ࢂࡵࡼ
Incumbent
participation
constraint
ࢂࡵࡵ
Region 1:
PE performs
LBO
Region 3:
Incumbent
acquisition
ࢂࡼ -ࢂࡵࡼ
ࢂࡼ -ࢂࡵࡵ
Region 2:
Incumbent
Outbids PE
ࢂࡼ
Private Equity
participation
constraint
ܶ௉
Figure 3.4. The Equilibrium Ownership Structure with asymmetries in transaction
costs.
56
VIP > T . For combinations of costs where PE firms are no longer a threat, the
incumbent participation constraint will be VII > T . This produces a stepwise
incumbent participation constraint function in the (TP , T ) plane, where the incumbents participate only if the transaction cost T is below this constraint.
Along the TP -axis, rising values of TP decrease vP but leaves vII and vIP
unchanged. As we go along the axis and increase TP , we first hit the PE
takeover-condition at TP = VP − VIP and then the PE threat-condition, which
sits at a higher value TP = VP − VII (since vIP > vII ). For values of TP beyond this point, there will not be any PE takeovers. The resulting graph gives
four regions with distinct outcomes for sales, ownership, price and the optimal
amount of restructuring performed on the target firm.
Region 1 offers low enough combinations of transaction costs such that
the PE valuation of the target is the highest of all three possible valuations
until we hit the PE participation-constraint above and PE takeover-condition
vP = vIP to the right. More specifically, in this region, PE transaction costs TP
are lower than the difference VP −VIP . This region then corresponds to Case 3
in Table 3.1, where the PE firm’s valuation of the target will exceed that of the
relevant valuation by incumbent firms (i.e vP > vIP ). The PE performs a LBO
and obtains the target firm at sales price vP and performs its optimal level of
restructuring at kP∗ .
Beyond the PE takeover-condition in Region 2 where TP > VP − VIP , PE
firms do not have sufficient transaction cost advantages to take over the firm.
However, they are still an active threat in the auction for the target firm until
TP gets too high for the PE to participate and the PE threat condition is hit
(i.e. TP = VP − VII ). In the region where transaction costs T are also low
enough (i.e. T ≤ VP ) incumbents that would like to prevent the high level of
restructuring value the firm higher at vIP > vP and outbid the PE firms from the
auction. This gives us a region bounded by PE threat and takeover conditions,
as well as the Incumbent participation constraint, which corresponds to Case 2
from Table 3.1. The target firm is obtained by the incumbent that matches the
valuation of PE firms at sales price vP , and is restructured at the lower level
kA∗ .
Beyond the PE threat condition, PE firms are no longer active participants
in the auction, and do not pose a threat to the incumbent. The incumbents
compete among themselves for the ownership of the target firm, for values of
transaction costs that are low enough (i.e. T < VII ), corresponding to Case 1
from Table 3.1. The incumbent purchases the target at sales price vII and again
performs the lower level of restructuring at kA∗ .
In Region 4 in Figure 3.4, no sale and therefore no restructuring of the target
occurs. This point is well above what the incumbent finds profitable to pay for
the target firm, given the high level of T above the incumbent participation
constraint. It is also beyond the levels of TP that would make it profitable for
the PE firm to justify its aggressive restructuring of the firm, given the high
level of its relevant transaction costs TP .
57
To sum up: to be active in equilibrium, private equity firms must face sufficiently low transaction costs compared to incumbent firms. Since PE firms
perform strictly higher levels of restructuring, the setting with sufficiently low
transaction costs for private equity firms leads to the highest level of restructuring.
3.2.5 Transaction Costs
By using Table 3.1 and Figure 3.4, testable implications can be derived from
the model. Both types of firms will have to incur an initial transaction cost if
and when they buy the target firm. In addition, if the target firm is purchased by
the PE firm in Stage 1, then the incumbent purchasing the PE-restructured firm
in Stage 3 will have to incur this initial transaction cost in that stage affecting
the PE firm’s valuation of the firm. As a result, the PE firm’s valuation will be
affected twice by transaction costs, as opposed to MNE firms that only have to
incur this cost once. Thus, PE firms will be relatively more sensitive to higher
costs - high transaction costs lower the private equity firm’s valuation of the
target more than it decreases either of the incumbent valuations. Therefore,
the possibility that a private equity owner could undertake a buyout decreases.
Consequently, the following proposition holds:
Proposition 3.1 Higher transaction costs decrease the likelihood of private
equity buyouts relative to buyouts by incumbent firms.
3.2.6 Market Structure
This section studies how the equilibrium pattern of cross-border private equity
buyouts depends on the market structure present in the domestic economy. To
this end, I utilise a Linear-Quadratic Cournot model (LQC) in order to derive
explicit solutions for optimal restructuring levels and ownership decisions by
agents in all stages of the game as a function of the market size and number of
firms present in the product market competition.8
For the LQC model, assume that firms face a linear demand and are in
Cournot competition with each other over homogeneous goods. The demand
function is of the form P = a − Xs , where a indicates consumer willingness to
pay, X is the sum of all quantities xi produced by firms that access the market,
and s denotes market size. We think of the number of firms in the market as
being inversely related to the degree of openness of the domestic economy to
the global economy. If there are barriers to trade, then fewer foreign firms can
access the domestic industry.
8 This type of framework, typically modelling an investment game followed by a stage with
oligopoly interaction, has been applied in, for instance, d’Aspremont and Jacquemin (1988),
Leahy and Neary (1997) and with regards to competition policy in Neary (2002). The paper at
hand adds to this body of work by incorporating an acquisition game into the framework.
58
There are two types of incumbent firms that need to be identified: the
acquirer of the target firm, which is marked by A, and all the other nonacquirer firms, which are denoted by NA. The firms face the profit function
Πi = (P − ci )xi , where xi is the output for a firm of type i = {A, NA}. The
acquiring firm A also faces a cost function C(kA ) = μA kA2 for the restructuring
kA they will perform, where μA is a constant multiplier. I assume the same
cost structure for the private equity firm if the target was private equity owned
in Stage 1, C(kP ) = μP kP2 , where μP < μA . Investments in development k are
assumed to reduce the acquirer’s marginal production cost (i.e. cA = c − kA ).
The non-acquirer is therefore assumed to have the marginal cost cNA = c.
The model is solved backwards for the optimal level of restructuring performed by an incumbent in Stage 2 defined in Equation 3.5, and also by a
private equity firm defined in Equation 3.7. The details of the solution are
presented in the Appendix. The optimal levels of restructuring for both the
incumbent A and the PE firm P are then given as:
kP∗ =
s(n + 1)Λ
μP (n + 1)
2
− s(n2 − 1)
, kA∗ =
snΛ
μA (n + 1)2 − sn2
,
(3.13)
where Λ = a−c. Figure 3.5 shows the difference between the optimal levels of
restructuring kP − kA and the resulting Equilibrium Ownership Structure in the
domestic economy as a function of the number of firms serving the product
market n and the market size s.9 Subfigure 3.5(i) shows the difference in
optimal levels of restructuring. First note that the difference is always positive
in the entire range: highest for the combination of a large market and small
number of firms; and lowest for a small market and high number of firms.
Subfigures 3.5(ii) and (iii) show in more detail that for any given level of n,
the difference is increasing in market size s, with the difference being largest
for small values of n. This is due to the fact that for a given n, a larger s would
translate into larger sales represented by the target firm. And, for a given s,
smaller n means a larger market share. Knowing that both these factors will
make the target firm very attractive to MNE firms in Stage 3, the PE firm
would like to increase its restructuring intensity even more with higher s and
fewer n. Consequently we have the largest difference in the optimal level of
restructuring on the low n-high s corner, and the smallest difference on the
high n-low s corner.
These differences in the restructuring intensity translate to the Equilibrium
Ownership Structure as demonstrated in Subfigure 3.5(iv)10 . An increase in
the number of firms in the product market always translates into the target
parameters in the equations are set to the following levels: nmax = 100, s ∈ {0.01, 1},
Λ = 1.5, μA = 1.1, and μP = 1.07.
10 To be able to illustrate EOS correctly, I assumed that the transaction costs were an average
of the 25th percentile of the MNE valuations VII and VIP , whereas the PE firm’s transaction
costs TP are set at half of the 25th percentile of VP .
9 The
59
݇௉‫ כ‬െ ݇஺‫כ‬
(i) Difference in
optimal
restructuring
݊
‫ݏ‬
݇௉‫ כ‬െ ݇஺‫כ‬
(ii) Difference in
optimal
restructuring,
base ࢔
‫ݏ‬
݊
݇௉‫ כ‬െ ݇஺‫כ‬
(iii) Difference in
optimal
restructuring,
base ࢙
݊
‫ݏ‬
‫ݏ‬
(iv) Equilibrium
Ownership
Structure
Region 2:
Incumbent
outbids PE
Region 1:
PE performs
LBO
Region 3:
Incumbent
acquisition
Region 4:
No Sale
݊
Figure 3.5. Optimal level of restructuring performed by the PE and MNE firms in Stage 2, and the
corresponding Equilibrium Ownership Structure illustrated as a function of the number of firms n and
market size s with the following parameters: nmax = 100, s ∈ {0.01, 1}, Λ = 1.5, μA = 1.1, and μP = 1.07.
To be able to get the Equilibrium Ownership Structure in subfigure (iv), transaction cost T was assumed to
be an average of the values VII and VIP at their 25th percentiles, and TP was set equal to half of the 25th
percentile of valuation VP .
60
firm getting a smaller share of the product market. For small market sizes, this
small share also means a small sales revenue for the firm. This lowers both
the value of owning the firm, as well as the negative externality of not owning
it. Thus, for a low range of market size, none of the MNE firms would find it
profitable to obtain the target, be it in Stage 1 or Stage 3. As a consequence,
the PE firm has no incentive to enter the market if its chances of being able to
sell back in Stage 3 are as slim. This gives a region increasing in n, marked as
Region 4, for low market shares where there is no sale of the target firm, and
where no restructuring takes place in the target.
Above this region, the target is either MNE or PE owned in Stage 1. For
a given n, as market size s increases, the value of the PE firm’s advantage
in restructuring goes up. Its higher level of restructuring decreases marginal
cost further compared to an MNE restructuring. This becomes increasingly
valuable as the market size increases, since the sales volume of the firm is
increasing in s. In this case, the PE firm will value the target firm higher, since
with enough restructuring performed at a lower cost, it is possible to make
the firm an attractive option for the MNEs to purchase in Stage 3. Thus, for
high enough values of s PE firms will dominate in the equilibrium ownership
outcome. This is marked as Region 1 in Figure 3.5 and depicted by Case 3 in
Table 3.1. The sales price will be vP and the optimal level of restructuring is
kP∗ .
In a product market served by a very small number of firms, the market
share of each firm is larger. In this case, there are very large negative externalities for the MNEs that end up being rivals with the heavily restructured firm
in the product market, as this region corresponds to the highest difference in
optimal levels of restructuring between MNEs and PE firms. As a result, the
incumbents’ valuation of the target if the alternative buyer is a PE firm gets
larger, and they outbid the PE firm as they are motivated to prevent becoming rivals with a heavily restructured firm in the product market, as shown in
Region 2. This corresponds to Case 2 in Table 3.1: the sales price will be
vP under the ownership of an incumbent that will perform its optimal level of
restructuring kA∗ on the target firm.
For a given s, an increase in the number of firms in the market will eat
away at the negative externality of now owning the target firm, since the market share of the target firm goes down with higher n. This makes the auction unattractive for the PE to enter, since they will no longer be able to use
this strategic point to their advantage while re-selling the target in Stage 3.
Therefore we will see an increasingly large number of MNE takeovers as n
increases, shown as Region 3 in Figure 3.5(iv). In this region, the sales price
will equal the valuation vii , and the optimal level of restructuring performed
on the target will be kA .
The interaction betwen the two forces results in a large share of LBOs for a
lower number of firms in the market, and higher values of market share. In this
region, the PE firm uses its restructuring advantage given the market size, in
61
addition to the potential negative externality given the small number of firms
already servicing the market. As the number of firms in the market increases
to the right, sales per firm are lower. This will lower the Stage 3 sales value of
the firm, and as a result the PE will not find it as profitable to over-restructure
for some combinations towards the south of this region. This will result in a
lower share of LBOs as the number of firms servicing the market increases.
So, to sum up, the following result on the equilibrium ownership in the
LQM model holds:
Proposition 3.2 In the Linear-Quadratic Cournot Model, the share of private
equity LBOs is increasing in proportion to market size, and decreasing in proportion to the number of firms.
3.3 Discussion
The sections below explore the effects of allowing for investment by nonacquiring incumbents, a different exit strategy by the Private Equity firm in
Stage 3 after the restructuring is completed, and, finally, a welfare analysis of
restrictive policies against PE participation.
3.3.1 Investments by the incumbents
The setup of the model assumed that only the acquiring incumbent or PE
firm gets the opportunity to restructure before the product market competition takes place. This section discusses the effect of possible investments by
non-acquiring rivals. At any given point, all firms should be operating at their
own profit maximizing levels. Any incentives to restructure by other (nonacquiring rivals) parties should come from a change in the market structure; in
this setting, when a domestic firm is acquired through an acquisition. Hence,
the acquiring firm is endowed with a first-mover advantage in this setting.
Since the acquiring firm will no longer have a monopoly on restructuring
rights, it will now have to account for a reaction from the non-acquirers in
their valuations of the target firm. For an incumbent, this means that its profits
of acquiring the target, as well as its negative externalities from not owning
the target will be lower. This will decrease the incumbent’s valuation of the
target firm. As a result, the Stage 3 sales price will also be lower, which will
dampen the private equity firm’s valuation. The sales price should therefore be
lower in all three cases of equilibrium ownership structure depicted in Table
3.1, without affecting the subsequent results.
62
3.3.2 Property Rights
Baziki et al. (2014) analyze a similar setup with a target firm up for sale, assuming that PE firms are restructuring experts, while MNEs make use of synergies between themselves and the target firm. Their model differs from the
present one in that if the domestic firm is acquired by a PE firm, it will operate
in the product market as an independent firm. If the target and the MNE have
a high level of synergies, then the target will be valued higher and acquired by
an MNE. In this case, the MNE will transfer firm-specific information, which
could take the form of patents, production methods, technologies or innovations in input utilization that they utilize when expanding internationally to
increase their competitive advantages.
These transfers will increase the efficiency with which the target firm is
operated, increase profits, and decrease the rivals’ product market revenues.
But the MNE can only transfer profit-enhancing knowledge over to the target
firm to the extent that they feel this transfer will be secured by property rights
protection in the target country. In other words, the MNE will not be able
to reach the full extent of its potential synergies with the target if property
rights are not well protected. The MNE would then fear that trade secrets
would be picked up by rivals. 11 However, since PE firms do not have such
industry-specific trade secrets, they would not be equally affected by the level
of property rights protection in the target country.
MNEs should then prefer to invest in countries with higher levels of property rights protection, which allow them to fully reap the benefits they could
get from their synergies with the targets. In a country with low levels of property rights protection, even if the level of synergies between the MNE and the
target are high, the MNE will realize that they will not be able to fully exploit these synergies. This would lower their valuation of the target firm, and
should increase the probability of an LBO in Stage 1. In the reverse scenario,
it then follows that an increase in property rights will decrease the likelihood
of cross-border LBOs.
Proposition 3.3 Stronger property rights decrease the likelihood of private
equity buyouts.
3.3.3 IPOs
Instead of exiting by a sale to incumbents, the PE firm could also exit by performing an initial public offering.12 Using backwards induction, the private
equity firm will evaluate its exit options prior to deciding the level of restructuring, and consider whether it is more profitable to exit by IPO or by selling to
11 See
Navaretti and Venables (2004) for a detailed discussion on this.
to data presented in Kaplan and Strömberg (2009), in 2009, 38% of PE firms
left their investments by a sale.
12 According
63
an incumbent. Gans et al. (2002) and Gans and Stern (2003) show that a firm
will chose to service the product market rather than sell the firm if their intellectual property rights are not well protected, if initial costs to investment
are low and intermediaries aiding in trade are not available. In a cross-border
setting other particulars of the industry such as the competitive environment in
the market could also affect the decision on the type of exit. An investigation
of these additional factors is left to future research.
3.3.4 Policy and Welfare Effects
This section considers the domestic welfare effects of cross-border buyouts.
Governments in many countries are skeptical when it comes to leveraged buyouts and PE firms. One of the major concerns among policy makers is that
private equity firms acquire domestic firms mainly for short run gains and invest less in the future of the company than other types of owners.
Following the conventional approach in an international oligopoly setting,
domestic welfare is equated to the sum of consumer surplus and the target
sale price.13 In this setting, a Non-discriminatory (ND) policy where both
cross-border private equity buyouts and incumbent cross-border acquisitions
are allowed to compete for the ownership of the target firm is compared to
a Discriminatory (D) policy that does not allow cross-border private equity
buyouts to take place. Let W (i) = S∗ (i) +CS(i) denote welfare under incumbent ownership, where S∗ (i) and CS(i) denote the equilibrium sales price and
consumer surplus. Define welfare under PE ownership in a similar manner,
W (p) = S∗ (p) +CS(p). The difference in welfare between these two policies
would then be evaluated as W ND−D = W (p) −W (i), which is equivalent to:
W ND−D = [S∗ (p) − S∗ (i)] + [CS(p) −CS(i)] ,
{z
}
|
(3.14)
Sales premium
The first term in (3.14) accounts for the difference in sales prices under different regimes, and the second term evaluates the potential difference in consumer surplus. In the cases where PE is actively present in the auction for
the target firm (Cases 2 and 3 depicted in Table 3.1), the sales price is driven
up by the valuation of the PE firm. Thus, the first term should be positive.
Next, PE’s higher restructuring intensity should benefit the consumers, which
should lead to a positive second term. Therefore welfare will be higher under
the non-discriminatory policy than under the discriminatory one.
13 In line with the theory, this section abstracts from the effects of tougher competition on the
domestic economy in evaluating domestic welfare.
64
3.4 Empirical Estimations
Proposition 3.1 suggests that private equity buyouts should be more likely
when transaction costs in the target firm’s country are low. Moreover, by
Proposition 3.2 I expect that private equity buyouts should be less likely in
economies that are very open to international trade. To test these predictions,
an empirical analysis is performed to assess how the pattern of cross-border
private equity buyouts and cross-border acquisitions depend on transaction
costs, property rights protection and the target country’s openness. The prediction emerging from Proposition 3.3 suggests that private equity buyouts
should be more likely when property rights in the target country are not wellprotected.
3.4.1 Empirical Models and Data
The predictions of the model are tested using the global mergers and acquisitions database CapitalIQ. Much of the existing empirical literature uses survey
data with attrition or coverage issues, or data that is not clearly identified to
either of the particular type of the acquisition investment (for example management buyouts versus leveraged buyouts), or types of investors (for instance
PE firms versus angel investors). In contrast, Capital IQ database has global
coverage on all cross-border takeovers and makes it possible to clearly identify
when a takeover is performed as a leveraged buyout. In the empirical analysis, this paper only considers majority investments where the acquiring firm
invests in more than 50% of the target firm, as those represent a large invested
stake in the target country. I further restrict the sample to include only completed or currently active acquisitions, excluding those that have merely been
announced or are in the setup phase. This leaves more than 72,000 observations covering the years 1998 to 2008 14 .
The data is then reorganized into a panel indexed by the recipient country
(i.e. the country of the target firm).15 The assignment of the recipient country
in the CapitalIQ database often follows that of the headquarters of the target
firm, and not necessarily the actual country where the target firm is located.16
14 The CapitalIQ database has substantially higher coverage rates in data from 1996 onwards.
The current choice of years maximizes the country coverage for this study, while there are still
some missing years for some developing countries.
15 One way of looking at FDI flows between countries would be to calculate the net amount
(UNCTAD, 2009). While this is a useful method for evaluating cross-country performance,
I am able to identify the real recipient country of the investment which makes the net flow
information redundant. Furthermore, this paper is interested in the motivating factors for each
and every cross-borders transaction, and therefore would see the alternative as an unfit method
for these purposes.
16 For instance, a firm in country X could be tied to the regional office in country Y and
corporate headquarters could be located in country Z. The transaction may look like from buyer
country to country Y or Z, but the actual investment is in X.
65
Since this paper looks at the impact of domestic polity factors on the presence of one type of investor over another, it is essential to correctly identify
the recipient country. To this end, I employed an address search system that
assigned the country to be the actual target country in such cases.
The polity measures are described below. The measures for international
openness, and property rights are accessed directly from The Quality of Government Dataset from the University of Gothenburg (Teorell et al., 2009). This
dataset is then supplemented with data from the World Bank (2010) where
needed.
Cross-border Share Regressions
Transaction Costs
To examine Proposition 3.1, I will estimate a reduced-form model of how the
share of cross-border PE buyouts is affected by transaction costs in a country.
• To proxy for the red tape, corruption, and other transaction costs that
businesses operating in the target country face, I use a World Development Indicator measure from the World Bank (2010) of the cost of
starting a business. The measure accounts for the number of visits, paperwork and other interactions required to obtain permits/licenses, and
meet verification requirements. To harmonize the costs across countries,
the total cost is then normalized by Gross National Product per capita in
each country for each given year. This rescaling allows for a comparison of the importance of the business start-up costs given the average
incomes in the country.
For country i, in time t, the following equation is estimated:
Share_LBOi,t = α0 + α1 BusinessCosti,t + Xi,t β + γi + γt + εi,t
(−)
(3.15)
where Share_LBOi,t is the share of cross-border private equity buyouts over
cross-border acquisitions, Xi,t is a vector of controls, γi is a country-specific
effect, γt a time-specific effect and εi,t is a stochastic shock. From Proposition
3.1, since PE incur costs twice, the share of cross-border private equity buyouts
should be negatively correlated to higher values of business start up costs with
α1 < 0.
Market Structure
To examine Proposition 3.2, I estimate a reduced-form model of how the economic integration of a country into the global economy affects the share of
cross-border private equity buyout investments. The proposition states that
66
the share of LBOs should decline with a rising number of firms in the market. The number of firms servicing the market should increase as the market
becomes more integrated with international markets. To this end, rising international openness should decrease the proportion of takeovers that end up as
LBOs in Stage 1.
• As a measure of openness of the domestic economy to world markets,
I use the United Nations openness to trade in constant (1990) price, defined as the ratio of total trade (exports plus imports) to GDP in constant
prices (1998-2008).
• In addition to the regular country-specific controls, I also use the number
of companies that are listed annually on the domestic stock exchange
for each of the countries from the World Bank as a proxy for existing
domestic competition level in the target country. Since the number of
firms n is a measure of both openness to the rest of the world, as well as
internal competition, including this variable will help us single out the
effect of external competition better.
For country i, in time t:
Share_LBOi,t = α0 + α1 TradeOpennessi,t + Xi,t β + γi + γt + εi,t
(−)
(3.16)
where Share_LBOi,t is the share of cross-border private equity buyouts over
cross-border acquisitions, Xi,t is a vector of controls, γi is a country-specific
effect, γt a time-specific effect and εi,t is a stochastic shock. According to
Proposition 3.2, since higher international openness make the PE valuation
lower, I expect a negative relationship.
Property Rights
To examine Proposition 3.3, I will estimate a reduced-form model of how the
property rights protection index value of a country affects the share of crossborder private equity buyouts over total M&As.
• To measure rule of law and property rights protection in a given country, I make use of the Heritage Foundation measure on property rights.
This is a composite measure that rates the coverage, as well as the enforcement, of the law protecting property. Judicial independence and
corruption are also included in the measure to the extent they also affect
the enforcement of property protection. It would have been ideal to have
a separate measure for Intellectual Property Rights protection, but there
is no time series IPR data comprehensive enough to complement the
data at hand.17 For countries where IPR data is available, the Heritage
17 The
benchmark study of IPR in developing nations is Park (2001), but the scope of that
data is rather outdated for this study. Several current studies looking at foreign investment and
IPR only focus on developed/industrializing countries.
67
Foundation property rights measure is highly correlated. Still, I refrain
from drawing conclusions about IPR in particular, but focus on how the
presence or lack of property rights could influence investment decisions
on the extensive margin.
For country i, in time t, the relationship is identified as:
Share_LBOi,t = α0 + α1 PropertyRightsi,t + Xi,t β + γi + γt + εi,t
(−)
(3.17)
where Share_LBOi,t is the share of cross-border private equity buyouts over
cross-border acquisitions, Xi,t is a vector of controls, γi is a country-specific
effect, γt a time-specific effect and εi,t is a stochastic shock. From Proposition 3.3, since better property rights are more preferred by MNEs that perform
M&As, the share of cross-border private equity buyouts in all global transactions should be negatively correlated to higher values of property rights. Thus,
the expected sign on the coefficient on the Property Rights variable is α1 < 0.
Panel Logit Regressions
Next, this section will make use of the same three variables to predict the
likelihood of a cross-border LBO in the data given the polity measures of
interest. An LBO dummy as the dependent variable will be defined as:
LBOdummyi,t =
1 if
0 if
Number of LBOs > 0
Number of LBOs ≤ 0
(3.18)
for each year t, and country i. Since it is not possible to observe negative
number of LBOs, the dummy is effectively equal to zero when there were no
cross-border LBOs in country i year t. If the probability of getting 1 is p, then
the following three equations will be estimated:
log(pi,t /1 − pi,t ) = α0 + α1 PropertyRightsi,t + Xi,t β + γi + γt + εi,t
(3.19)
log(pi,t /1 − pi,t ) = α0 + α1 BusinessCosti,t + Xi,t β + γi + γt + εi,t
(3.20)
(−)
(−)
log(pi,t /1 − pi,t ) = α0 + α1 TradeOpennessi,t + Xi,t β + γi + γt + εi,t (3.21)
(−)
as well as versions of these equations including a lagged dependent variable.
Although the theory section makes predictions regarding the relative likelihood of cross-border LBOs over all cross-border investments, the expected
signs on the variables of interest remain the same as in the share regressions.
68
3.4.2 Results
Cross-border Share Regressions
To test each of the three hypotheses, in addition to running panel regressions
as specified above, I report results on a variant model with an additional one
period lagged independent variable as a regressor to allow for dynamic adjustment. The base regressions are reported with clustered robust errors in
Column 1 and and an additional lagged dependent variable in Column 4.
Fixed effect regressions are reported in Columns 2 and 5, the latter including a
lagged dependent variable. Fixed effect models with year effects are reported
in Columns 3 and 6, again the latter with a lagged dependent variable.
Table 3.2 presents the results for transaction costs. The model suggests that
the share of LBOs should be negatively affected by higher costs, since transaction costs are incurred by PE firms twice in comparison to the incumbents that
only incur them once. The results show that one standard deviation increase in
these costs lead to a decline of 0.0286 and 0.0367 standard deviations in LBO
shares respectively.
Table 3.3 Panel A shows the results of running the last equation on trade
share of GDP in constant prices, calculated by the United Nations. In all
specifications, with or without lags, the openness measure is of the expected
sign when significant. One standard deviation increase in openness results
in a decline of 0.252 standard deviations in the share of LBOs in the lagged
model. Results from the set of regressions without controlling for domestic
registered firms are reported in Table 3.3 Panel B, where a decline of one
standard deviation in the openness measure is equivalent to 0.325 standard
deviations of decline in LBO shares in the lagged model, and 0.299 standard
deviations of decline in the base model.
Table 3.2. Panel Regression Results of Share of LBO on Business Startup Costs
1
Business Cost
-2.50e-05***
(7.82e-06)
Log Population
Base Model
2
-5.69e-06**
(2.72e-06)
-0.337
(0.355)
3
4
-2.44e-05***
(5.57e-06)
-2.51e-05
(1.73e-05)
Share LBOt−1
Observations
R-squared
N of Countries
Country Control
Year Control
0.147**
(0.0590)
918
166
622
0.005
160
918
0.021
166
Yes
Yes
Yes
820
152
Model with Lag
5
6
-7.36e-06***
(2.03e-06)
-0.0883
(0.280)
-0.137**
(0.0606)
-3.13e-05***
(6.45e-06)
538
0.028
139
820
0.025
152
Yes
Yes
Yes
-0.0678
(0.0519)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Columns 1-3 present Base model results, and columns 4-6 present results from regressions including
the lagged dependent variable. Within R-squared values reported. Robust standard errors are reported in
parentheses and all standard errors are clustered by country.
69
Table 3.3. Panel Regression Results of Share of LBO on International Openness
1
Base Model
2
3
4
-0.000408
(0.000287)
-3.25e-05*
(1.76e-05)
8.85e-07
(2.00e-05)
1.38e-06
(2.46e-06)
Model with Lag
5
6
Panel A: with Listed Companies
Openness
Listed Comp
-0.000147
(9.96e-05)
-9.26e-06
(8.43e-06)
Log Population
-0.00154**
(0.000616)
-5.69e-05**
(2.32e-05)
-0.826***
(0.230)
Share LBOt−1
Observations
R-squared
Number of ctry
0.312***
(0.0536)
997
112
Country Control
Year Control
997
0.083
112
997
0.336
112
Yes
Yes
Yes
-0.00182***
(0.000504)
-0.700***
(0.198)
-0.000590*
(0.000312)
839
107
-0.00109***
(0.000340)
-1.07e-05
(1.58e-05)
-0.212
(0.172)
0.154***
(0.0562)
-0.000497**
(0.000223)
-5.26e-06
(1.86e-05)
839
0.083
107
839
0.310
107
Yes
Yes
Yes
-0.00121***
(0.000357)
-0.287
(0.194)
0.152***
(0.0553)
-0.000642**
(0.000267)
990
0.088
149
990
0.272
149
Yes
Yes
Yes
-0.0184
(0.0489)
Panel B
Openness
7.03e-06
(4.29e-05)
Log Population
Share LBOt−1
Observations
R-squared
Number of ctry
Country Control
Year Control
4.11e-06
(1.90e-05)
0.303***
(0.0495)
1,281
177
1,281
0.081
177
1,281
0.315
177
Yes
Yes
Yes
990
149
0.00238
(0.0508)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Columns 1-3 present Base model results, and columns 4-6 present results from
regressions including the lagged dependent variable. Within R-squared values reported. Robust
standard errors are reported in parentheses and all standard errors are clustered by country.
70
Table 3.4. Panel Regression Results of Share of LBO on Property Rights
1
Property Rights
0.00159***
(0.000333)
Log Population
Base Model
2
-0.00172
(0.00155)
-1.201***
(0.291)
3
4
-0.00270**
(0.00118)
0.000979***
(0.000221)
Share LBOt−1
Observations
R-squared
N of Countries
Country Control
Year Control
1,054
150
1,045
0.042
149
1,054
0.335
150
Yes
Yes
Yes
0.288***
(0.0512)
817
132
Model with Lag
5
6
-0.000932
(0.00116)
-0.600**
(0.253)
0.175***
(0.0636)
809
0.061
131
-0.00195*
(0.00103)
Yes
Yes
Yes
-0.0209
(0.0596)
817
0.300
132
*** p<0.01, ** p<0.05, * p<0.10.
Note: Columns 1-3 present Base model results, and columns 4-6 present results from
regressions including the lagged dependent variable. Within R-squared values reported. Robust
standard errors are reported in parentheses and all standard errors are clustered by country.
Results for property rights are presented in Table 3.4. The main specifications are in Columns 3 and 6. Both specifications show that a higher level of
property rights protection leads to a lower LBO share. In the main model with
year- and country-specific controls and robust standard errors that are clustered by country, a point increase in the property rights index leads to a 0.27%
decline in the share of LBOs (Column 3). When the model is run with lags,
the effect declines to 0.195% (Column 6). This is equivalent to one standard
deviation change in property rights leading to 0.311 standard deviations of decline in the cross-border LBO share. For the model with a lag, this effect goes
down to -0.225 standard deviations in the outcome variable.
I would like to further investigate intellectual property rights (IPR) protection, which is only partially included in the composite measures available.
For companies that have firm-specific knowledge, patents, or technology that
they would like to protect, this would be an important factor to consider when
investing abroad. In line with the model and Proposition (3.3), intellectual
property rights should be of relatively higher concern to incumbent Mergers
and Acquisitions compared to private equity LBOs. However, a reliable index that covers most of the developing countries in the current dataset is not
available. For the countries that have IPR measures, the correlation of PR and
IPR is at 0.78, which is in line with the argument that one of the reasons why
M&As prefer higher property rights is IPR protection alone.
Tables 3.5 and 3.6 show results of regressions with all independent variables
of interest included. Columns 3 and 6 show that all the variables of interest
carry the expected sign, but only the openness variable has a significant effect
on the share of LBOs. In the base model on standard deviation, an increase in
openness translates to a decline of 0.0681 standard deviations in the share of
LBOs in the country. And, in the specification with lags, the negative effect
71
is reduced to a 0.0681 to 0.0582 standard deviations decline in the outcome
variable. Unfortunately, due to a lack of country and year coverage overlap
between the variables, this regression has the least number of observations,
and therefore its results are not the main focus of discussion.
72
73
361
0.037
99
-8.43e-06**
(3.77e-06)
-1.68e-05
(1.95e-05)
0.000882***
(0.000225)
2.00e-06
(4.70e-06)
*** p<0.01, ** p<0.05, * p<0.10.
97
334
0.096
97
-5.59e-07
(1.54e-06)
-0.000575***
(0.000188)
-0.00294
(0.00211)
-3.20e-05
(2.73e-05)
-0.687
(0.520)
-0.206***
(0.0679)
Model with Lag
5
334
0.096
97
-0.195***
(0.0698)
-1.96e-06
(2.74e-06)
-0.000552**
(0.000233)
-0.00246
(0.00196)
-3.67e-05
(2.44e-05)
6
Note: Columns 1-3 present Base model results, and columns 4-6 present results from regressions including the lagged dependent variable. Within R-squared
values reported. Robust standard errors are reported in parentheses and all standard errors are clustered by country. Columns 2 and 5 have Country, Columns 3
and 6 have Country and Year controls.
99
334
361
0.039
99
-3.45e-08
(2.57e-06)
-0.000646**
(0.000272)
-0.00227
(0.00175)
-5.63e-05
(4.17e-05)
4
Observations
R-squared
Number of ctry
1.34e-06
(2.01e-06)
-0.000600**
(0.000242)
-0.00320
(0.00206)
-5.28e-05
(4.75e-05)
-0.957
(0.746)
3
0.109
(0.0864)
361
-6.37e-06**
(2.73e-06)
-5.17e-05*
(3.01e-05)
0.000849***
(0.000253)
-7.82e-07
(5.39e-06)
Base Model
2
Share LBOt−1
Log Population
Listed Comp
Property Rights
Openness
Business Cost
1
Table 3.5. Panel Regression Results of Share of LBO on All Variables - with Listed Companies
74
441
0.030
133
-1.01e-05
(7.12e-06)
-7.05e-06
(2.27e-05)
0.00111***
(0.000208)
*** p<0.01, ** p<0.05, * p<0.10.
120
384
0.086
120
-5.11e-07
(1.52e-06)
-0.000447*
(0.000265)
-0.00275
(0.00178)
-0.671
(0.427)
-0.205***
(0.0659)
Model with Lag
5
384
0.081
120
-0.194***
(0.0678)
-1.66e-06
(2.53e-06)
-0.000447
(0.000302)
-0.00209
(0.00161)
6
Note: Columns 1-3 present Base model results, and columns 4-6 present results from regressions including the lagged dependent variable. Within R-squared
values reported. Robust standard errors are reported in parentheses and all standard errors are clustered by country. Columns 2 and 5 have Country, Columns 3
and 6 have Country and Year controls.
133
384
441
0.030
133
4.94e-07
(2.18e-06)
-0.000539*
(0.000319)
-0.00201
(0.00141)
4
Observations
R-squared
Number of ctry
1.31e-06
(1.53e-06)
-0.000518*
(0.000271)
-0.00301*
(0.00172)
-0.888
(0.564)
3
0.105
(0.0771)
441
-8.66e-06
(6.19e-06)
-4.38e-05
(2.97e-05)
0.00111***
(0.000228)
Base Model
2
Share LBOt−1
Log Population
Property Rights
Openness
Business Cost
1
Table 3.6. Panel Regression Results of Share of LBO on All Variables
Panel Logit Regressions
The panel logit regressions test the effect of the independent variables on the
likelihood of observing at least one cross-border LBO in the target country.
The baseline case performs a pooled panel logit, supplemented by fixed effects
logits with a country fixed effect and an additional year effect. As before, the
specifications also present the results of the same regressions with a lagged
dependent variable. All regression columns are accompanied with their corresponding odds ratio effects immediately to the right. The discussion below
will focus on the logit specifications with country fixed effects.
As shown in Table 3.7, an increase in business costs has a negative effect on
the likelihood of LBO outcomes. The pooled model specifies this effect to be
about a 1.5% decline in the likelihood of an LBO in response to a unit increase
in business costs. Unfortunately, even though the effect is of the expected sign,
the variable of interest does not lead to significant results in the fixed effects
models.
Table 3.7. Panel Logit Results of LBO instance on Business Costs
Base Models
Business Costs
(1)
(Base)
(2)
(OR)
(3)
(FE)
(4)
(OR)
(5)
(FE i.y)
(6)
(OR)
-0.0153***
(0.00335)
0.9848
-0.000537
(0.00141)
21.70***
(5.848)
319
65
Yes
0.9995
-0.00107
(0.00217)
0.9989
Log Population
Observations
N of Countries
Country Control
Year Control
1,174
179
2.66e+09
Yes
533
80
Yes
Yes
Yes
Yes
Models with lagged dependent variable
Business Costs
-0.0128***
(0.00264)
0.98724
Log Population
LBOt−1
Observations
N of Countries
Country Control
Year Control
1.541***
(0.310)
1,173
178
4.6687
-0.000556
(0.00147)
22.95***
(6.017)
-0.342
(0.279)
0.9995
-0.00111
(0.00222)
0.998891
-0.463*
(0.244)
.629251
9.25e+09
0.7106316
319
65
Yes
533
80
Yes
Yes
Yes
Yes
Yes
*** p<0.01, ** p<0.05, * p<0.10.
Note: Top panel presents Base model results, and the bottom panel presents results from regressions
including the lagged dependent variable. RE stands for Random Effects, FE for (conditional) Fixed Effects
estimations, and OR stands for Odds Ratio. Standard errors are reported in parentheses.
75
76
2,469
190
0.00798***
(0.00184)
1,335
115
1.0080
1.0014
1,325
102
0.0127***
(0.00322)
10.22***
(1.507)
Yes
1,028
85
0.0170***
(0.00447)
0.000675*
(0.000375)
11.15***
(1.883)
27349.4
1.0128
Yes
69419.7
1.0007
1.0171
Base Models
(FE)
(OR)
0.9972
Yes
Yes
1.0003
0.9981
(OR)
(RE)
2,280
190
1.421***
(0.201)
0.00391**
(0.00158)
1,245
115
1.521***
(0.218)
0.00263**
(0.00116)
0.00129***
(0.000294)
*** p<0.01, ** p<0.05, * p<0.10.
1,325
102
-0.00279
(0.00278)
Yes
Yes
1,028
85
-0.00187
(0.00366)
0.000346
(0.000354)
(FE Y)
4.1421
1.0039
4.5760
1.0013
1.0026
1,188
99
0.00883***
(0.00310)
9.273***
(1.620)
0.387**
(0.179)
Yes
924
82
0.0121***
(0.00422)
0.000605
(0.000455)
10.56***
(2.054)
0.341*
(0.195)
1.4730
10649.9
1.0089
Yes
1.4059
38720.6
1.0006
1.0122
1,188
99
-0.0858
(0.201)
-0.00349
(0.00296)
Yes
Yes
924
82
-0.167
(0.221)
-0.00220
(0.00371)
0.000375
(0.000425)
Models with lagged dependent variable
(OR)
(FE)
(OR)
(FE Y)
0.9178
0.9965
Yes
Yes
0.8463
1.0004
0.9978
(OR)
Note: Columns 1-6 present Base model results, and columns 7-12 present results from regressions including the lagged dependent variable. RE stands for Random Effects, FE for
(conditional) Fixed Effects estimations, and OR stands for Odds Ratio. Standard errors are reported in parentheses.
Observations
N of Countries
LBOt−1
Log Population
Openness
Panel B
CountryControl
Year Control
Observations
N of Countries
LBOt−1
Log Population
Listed Comp
1.0060
Openness
0.00594***
(0.00164)
0.00137***
(0.000275)
(OR)
(RE)
Panel A: with Listed Companies
Table 3.8. Panel Logit Results of LBO instance on International Openness
Table 3.9. Panel Logit Results of LBO instance on Property Rights
Base Models
Property Rights
(1)
(Base)
(2)
(OR)
(3)
(FE)
(4)
(OR)
(5)
(FE i.y)
(6)
(OR)
0.0131
(0.00885)
1.0132
-0.0317**
(0.0125)
6.380***
(1.857)
999
85
Yes
0.9688
-0.0110
(0.0136)
0.9890
Log Population
Observations
N of Countries
Country Control
Year Control
1,837
162
590.207
Yes
1,011
86
Yes
Yes
Yes
Yes
Models with lagged dependent variable
Property Rights
0.0332***
(0.00621)
1.0338
Log Population
LBO
Observations
N of Countries
Country Control
Year Control
1.531***
(0.233)
1,697
162
4.6236
-0.0378***
(0.0137)
4.747**
(2.038)
0.151
(0.205)
0.9629
0.98318
-0.231
(0.222)
0.79376
115.2492
1.1626
888
82
Yes
-0.0170
(0.0143)
899
83
Yes
Yes
Yes
Yes
Yes
*** p<0.01, ** p<0.05, * p<0.10.
Note: Top panel presents Base model results, and the bottom panel presents results from regressions
including the lagged dependent variable. RE stands for Random Effects, FE for (conditional) Fixed Effects
estimations, and OR stands for Odds Ratio. Standard errors are reported in parentheses.
Table 3.8 presents the results for the probability of an LBO occurrence
through an increase in the openness indicator. In the models with country
fixed effects, we see that one unit increase in international openness makes it
about 1% more likely to observe any LBOs in the country in all specifications.
In Table 3.9, the fixed effects model shows that an increase in property
rights by one point decreases the probability of observing a LBO by about
3% in the regular model, and by 3.7% in the model with a lagged dependent
variable. This is in line with Proposition (3.3), which proposes a decline in
LBO likelihood with increases in property rights protection.
Overall, the approach in this section is suggestive, but not necessarily a
very good fit for the purposes of this paper. The logistic model treats the
outcome variable as binary, which does not directly follow the predictions in
the paper on the relative presence of LBOs. A lot of valuable information
on the frequency of these cross-border LBOs coming into a country is lost
in these regressions of binary outcomes. Furthermore, since the fixed effects
logit models drop observations that do not display a change in the outcome
77
variable over time, a large share (for example 2/3 of the data for the business
costs) of the data is out of the sample when using this method.
3.4.3 Robustness
In this paper, the motivation for selecting the particular variables used have
been, in order of application, the fit of the variable definition, coverage of the
number of countries, and, finally, the coverage over the number of years. There
are several other sources of polity measures for property rights, transaction
costs and trade openness measures most of which are well summarized in the
University of Gothenburg database.
Some of these measures are composites, which makes it difficult to specifically identify the relationships this paper seeks to identify. Freedom House
has two rule of law variables that have incorporated individual freedoms into
the composite effect, which make it not as suitable as a proxy for property
rights. World Bank supplies Worldwide Governance Indicators updated by
Kaufmann et al. (2009) which publishes a rule of law indicator that includes
contract enforceability, but also includes criminal and judicial aspects.
Brady et al. (2014) have put together a Comprehensive Welfare States
Dataset that has three suitable capital and current account openness variables,
which could stand in for both transaction costs and international openness.
However, it only covers 20 countries for a limited number of years. Almost all
other available international openness variables are constructed as total foreign
trade weighted by domestic output, similar to the variable from the United Nations used in this study. Easterly (2001) provides an interesting exception with
a variable that calculates the size of taxes on international trade and transactions. This variable is a direct measure of trade costs present in the domestic
economy. Unfortunately, this variable covers only 46 countries, and therefore
proves to be a less than good choice for this study.
Some countries are known to be tax havens of runaway capital where companies may set up a branch, or relocate their headquarters in order to obtain
certain tax benefits.18 Since this study looks at how differences in countryspecific indicators across target countries derive asymmetric results in the type
of investor, the tax havens present a challenge when correctly assigning the target country to the cross-border investment. I address this concern by using the
following two steps. First, since I reassign the recipient country to be the location of the actual target through an address search, the number of instances
where one of these tax havens is actually the target country in my final sample
is quite small as shown in Table 3.11. Next, I repeat the empirical investiga18 The list of tax haven countries in the CapitalIQ database are: British Virgin Islands, The
Cayman Islands, Antigua and Barbuda, Barbados, Channel Islands, Falkland Islands, French
Polynesia, Netherlands Antilles, Seychelles, Saint Kitts & Nevis, Saint Lucia, and Turks and
Caicos Islands.
78
tions outlined in Section 3.4.1 on the sample of countries excluding the tax
havens, and find very similar results, presented in Tables 3.12 to 3.14 in the
Appendix.
3.5 Conclusion
The number of leveraged buyouts performed by private equity companies exhibit big differences both across time and countries. To understand these differences, this paper uses a model, similar to Norbäck and Persson (2009) and
Norbäck et al. (2013), of cross-border M&As, which incorporates LBOs in
a setting where a domestic target firm that is up for sale has potential buyers
that are either of the incumbent type, which conducts an M&A, or a PE type,
which conducts an LBO. The acquiring firm, be it an incumbent or a PE firm,
gets the chance to restructure the newly acquired firm. If acquired by a PE
firm, the target firm is put up for sale again after the restructuring phase, and is
bought by an incumbent. At this point, all incumbents compete in the product
market.
In this model setup, the incumbents maximize their product market profits,
whereas the PE firm maximizes the resale price of the target firm after restructuring. As a result, PE firms restructure the target more intensely, which
creates a negative externality to incumbents that end up as rivals of the restructured target in the product market. This negative externality is on the one
hand a driving force for PE to enter the auction for a lucrative resale price in
Stage 3, but it also creates a motivation for the incumbents to try to outbid the
PE firms from the auction in Stage 1 in order to avoid becoming rivals of the
restructured target later on.
The paper makes several empirical predictions. Since PE firms incur costs
associated with ownership change twice, higher transaction costs should lower
LBO presence in the economy. Given the size of the domestic market, a higher
number of firms serving the market reduces the negative externality of not
owning the target firm, and therefore reduces the share of takeovers that are
LBOs. Finally, making use of a similar model setup from Baziki et al. (2015),
I suggest that unlike PE firms that do not possess industry-specific know-how,
MNEs may want to utilize firm-specific knowledge in the target firm, which
is possible if property rights are well-protected. Therefore, a higher degree
of property rights protection in the country should make it more attractive for
MNEs, and decrease the share of M&As that end up as LBOs.
These predictions are put to test with a comprehensive cross-border investments data base that covers all the majority-owned M&As in the world that
are either completed or ongoing. The empirical outcomes are in the expected
direction postulated by the theory, and show that higher levels of transaction
costs, trade openness, and property rights protection lead to lower cross-border
LBO shares. The paper differs from the current empirical literature by look79
ing at the effect of time varying polity indicators on all cross-border takeovers
in the world, and by showing that there are asymmetries in how attractive a
country may be to one type of investor over the other based on the strength of
these variables.
80
References
d’Aspremont C. and Jacquemin A., (1988). Cooperative and Noncooperative
R&D in Duopoly with Spillovers, American Economic Review, 78(5), 11331137.
Baker G.P. and Wruck K., (1989). Organizational Changes and Value Creation
in Leveraged Buyouts: The Case of the O.M. Scott & Sons Company, Journal
of Financial Economics, 25(2), 163-190.
Baziki S.B., Norbäck P.-J., Persson L., and Tåg J., (2015). Cross-border Acquisitions and Restructuring: Multinational Enterprises versus Private Equityfirms, IFN Working Paper No.1057.
Brady D., Huber E., and Stephens J.D., (2014). Comparative Welfare States
Data Set, University of North Carolina and WZB Berlin Social Science Center.
Davis S.J., Haltiwanger J., Handley K., Jarmin R., Lerner J., and Miranda
J., (2013). Private Equity, Jobs, and Productivity, NBER Working Papers
N.19458.
Easterly W., (2001). Global Development Network Growth Database. Washington D.C.: The World Bank.
Gans J., Hsu D.H., and Stern S., (2002). When Does Start-Up Innovation
Spur the Gale of Creative Destruction?, RAND Journal of Economics, 33(4),
571-586.
Gans J. and Stern S., (2003). When Does Funding Research by Smaller Firms
Bear Fruit?: Evidence from the SBIR Program. Economics of Innovation and
New Technology, 12(4), 361-384.
Kaplan S.N., (1989). The Effects of Management Buyouts on Operating Performance and Value, Journal of Financial Economics, 24(2), 217-254.
Kaplan S.N. and Strömberg P., (2009). Leveraged Buyouts and Private Equity,
Journal of Economic Perspectives, 23(1), 121-146.
Kaufmann D., Kraay A., Mastruzzi M., (2009). Governance Matters VIII: Aggregate and Individual Governance Indicators 1996-2008, World Bank Policy
Research Working Paper, WPS4978.
Leahy D. and Neary J.P., (1997). Public Policy towards R&D in Oligopolistic
Industries, American Economic Review, 87(4), 642-662.
81
Lichtenberg F.R. and Siegel D., (1990). The Effects of Leveraged Buyouts
on Productivity and Related Aspects of Firm Behavior, Journal of Financial
Economics, 27(1), 165-194.
Luehrman T., (2007). The Hertz Corporation (A), Harvard Business School
Case no. 208-030.
Navaretti G.B. and Venables A.J., (2004). Multinational Firms in the World
Economy. Princeton University Press.
Neary J.P., (2002). The Road Less Travelled - Oligopoly and Competition Policy in General Equilibrium. Working Papers, School of Economics, University
College Dublin 200222.
Norbäck P.-J. and Persson L., (2009). The Organization of the Innovation
Industry: Entrepreneurs, Venture Capitalists, and Oligopolists, Journal of the
European Economic Association, 7(6), 1261-1290.
Norbäck P.-J., Persson L., and Tåg J., (2013). Buying to Sell: Private Equity
Buyouts and Industrial Restructuring, CESifo Working Papers, 4338.
Park W.G., (2001). Intellectual Property and Patent Regimes. Economic Freedom of the World: 2001 Annual Report 1, 101-108.
La Porta R., Lopez-De-Silanes F., Shleifer A., and Vishny R.W., (1997). Legal
Determinants of External Finance, The Journal of Finance, 52(3), 1113-1150.
La Porta R., Lopez-De-Silanes F., Shleifer A., and Vishny R.W., (1998). Law
and Finance, Journal of Political Economy, 106(6), 1113-1155.
Strömberg P., (2007). The New Demography of Private Equity, Swedish Institute for Financial Research Working Paper.
Teorell J., Charron N., Samanni M., Holmberg S., and Rothstein B., (2009).
The Quality of Government Dataset, version Jun09. University of Gothenburg:
The Quality of Government Institute, http://www.qog.pol.gu.se.
The World Bank, (2011). World Development Indicators.
82
Appendix
A.1 Proofs
Sales Price from Section 3.2.2
Take two firms i and h, both elements of the set of foreign firms I . Suppose
that incumbent i acquires the target firm at Si∗ which is equivalent to its valuation of the firm. If the firm i where to make an alternate offer where Si < Si∗ ,
firm h benefits from posting a bid at Sh = Si + ε < Si∗ , for a small ε > 0 since
it then obtains the target. Realizing this, firm i will increase its offer, and firm
h will respond with an even higher offer until they both reach their valuation
of acquiring the firm Si∗ = Sh∗ . In this case, the firms get the target firm with
probability 1/2. Finally, note that Si > Si∗ is not an optimal strategy, since
then the incumbent will be paying too much for the target firm. Thus, offering
to pay Si∗ is a Nash equilibrium and the only Nash equilibrium where firm i
obtains the target firm.
Table 3.1 from Section 3.2.4
In Case 3, consider an equilibrium where private equity firm j obtains the target firm with sales price S∗j . Note that S∗j < vP is a dominated strategy because
then for a small ε > 0 another firm could make an offer S∗ = S∗j + ε < vP and
obtain the target. Firm j would not make an offer such that S∗j > vP since
paying above the target’s maximum worth to them would be suboptimal. By
the same argument, since vP > vIP , none of the MNE firms have an incentive
to pay more than vIP to compete with the PE firm. This leaves S∗j = vP as the
only undominated strategy for the PE firm.
In Case 2, consider an equilibrium where incumbent firm i obtains the target
firm at a sales price Si∗ and where private equity j has the second highest offer,
denoted S∗j . For a small value of ε > 0, Si∗ = vP − ε is not an equilibrium, since
firm j would then benefit from deviating to S j = vP − ε/2 to eliminate the
incumbent firm from the auction. However, in that setting firm j would then
have to face other PE firms whose valuations would equal vP . As in Case 3,
the PE firm would offer S∗j = vP as its best offer. Then, since vIP > vP , an
incumbent firm would benefit from raising its offer to vP + ε for the smallest
possible unit of ε to eliminate all PE firms from the auction and claim the
target. Since ε is taken as a negligibly small amount, in the calculations I treat
the equilibrium offer coming from incumbent i as Si∗ = vP .
83
In Case 1, suppose an incumbent firm i makes the offer Si∗ = vII − ε < vP
for a small ε > 0. Then the private equity firm j would make an offer at
S∗j = vP and eliminate the incumbent firm i from the auction, acquiring the
firm. However, since vIP > vP , MNE firms will bid at least vP to outbid the PE
firms from the auction. Next, if firm i makes any offer Si∗ < vII , then another
incumbent h will pay Sh∗ = Si∗ +ε to obtain the target. This will prompt firm i to
raise its offer until it reaches vII . Any offer above this valuation at Si∗ = vII + ε
would be overpaying for the target firm. This leaves Si∗ = vII as the incumbent
firm’s optimal strategy.
A.2 Cournot with linear demand and restructuring k in the target
firm
There is a target firm that is up for sale in the market, and either one of the
n MNE firms, or m PE firms can obtain the target. If the target is obtained
by an MNE, the MNE invests kA in its restructuring, and then it competes
in the product market with other MNEs. If the PE firm obtains the target,
it restructures it by kP and sells it back to one of the MNE firms who then
compete with each other on the product market. Since revenues are increasing
in the amount of restructuring, there is an added negative externality of not
owning the target firm when the restructuring level is high. The problem facing
all actors in this setup are solved backwards, starting with Stage 4.
The profit of firm i in a setting with n MNE firms is:
πi = [P − ci ]xi − μA kA2
(3.22)
where the second term stands for the cost of restructuring the target firm, only
incurred by the MNE firm A that obtains the firm. The amount of restructuring
firm A has performed on the firm, kA is subject to a type specific cost multiplier
μA . Firms face the following inverse demand function:
P = a−
X
s
(3.23)
n
where X = ∑ xi and s is a measure of market size.
i=1
Product market competiton, Stage 4
The first-order condition taking quantities of other firms as given:
(P − ci ) +
From Equation (3.23) we know that
to
84
dP ∗
x =0
dxi i
dP
dxi
(3.24)
= − 1s . This updates Equation (3.24)
xi∗
=0
s
Updating Equation (3.25) with Equation (3.23):
(P − ci ) −
X∗
x∗
− ci − i = 0
s
s
a−
(3.25)
(3.26)
Define Λi = a − ci :
X ∗ xi∗
− =0
s
s
∗
Sum over all n firms and solve for X
Λi −
∑ Λi − n
i
This gives:
∗
X =
X∗ X∗
−
=0
s
s
s ∑ Λi
i
n+1
=
sΛ̄
n+1
(3.27)
(3.28)
(3.29)
Where Λ̄ = ∑ Λi . From (3.27), we now have:
i
xi∗ = sΛi − X ∗
(3.30)
And finally using (3.22) and (3.25) the profit function can be updated to:
π∗i
[xi∗ ]2
=
− μA kA2
s
(3.31)
Restructuring in Stage 2
At this stage, the target firm was either purchased by one of m Private Equity
firms denoted by (P), or by one of n incumbent MNEs denoted by (A). The
owner of the target can invest in restructuring at the target firm (kl , l ∈ {A, P})
which will reduce its unit costs. All the other (n − 1) MNE firms who have
not acquired the target firm (NA) will not be able to make such cost reducing
investments.
Incumbent Acquisition in Stage 1
A direct MNE acquisition in Stage 1 results in one restructured firm owned by
the acquiring MNE A. All the other (n − 1) MNE firms who are denoted NA.
Their cost structures are as follows:
cA = c − kA
cNA = c
85
Total quantity produced in this setting is:
X∗ = s
(n − 1)(a − c) + a − c + kA
nΛ + kA
Λ̄
=s
=s
n+1
n+1
n+1
where Λ = a − c as above. The production of the acquiring (A) and nonacquiring (NA) MNE firms are:
nΛ + kA
Λ + nkA
=s
n+1
n+1
nΛ
+
k
Λ
−
k
A
A
∗
= sΛ − s
xNA
=s
n+1
n+1
xA∗ = s(Λ + kA ) − s
(3.32)
The corresponding profits are:
[xA∗ (i)]2
Λ + nkA 2
2
=
− μA kA2
− μA kA = s
s
n+1
∗
2
xNA (i)
Λ − kA 2
∗
πNA =
.
=s
s
n+1
π∗A
(3.33)
where values under PE restructuring are defined analogously. To find the optimal investment level kA∗ , we take the first derivative of the optimal profit
function π∗i from Equation (3.33):
2s
Λ + nkA∗ n
= 2μA kA∗
n+1 n+1
snΛ + sn2 kA∗ = μA (n + 1)2 kA∗
kA∗ =
snΛ
μA (n + 1)2 − sn2
.
(3.34)
(3.35)
(3.36)
For this expression to be defined, the following inequality should hold:
μA (n + 1)2 − sn2 > 0
(3.37)
Which suggests:
μA >
sn2
(n + 1)2
.
(3.38)
Private Equity Acquisition in Stage 1
We know that the optimal level of restructuring performed by PE firms kP is
determined by maximizing the sales price of the target firm in Stage 3, S∗ (k)
subject to own restructuring costs as demonstrated in Equation (12) in the main
text. This valuation depends on the profits of the MNE firms (both acquiring
86
and non-acquiring) in the new setting where the level of restructuring is higher,
at the PE optimal level kP .
The corresponding profits will be:
[xP∗ (p)]2
Λ + nkP 2
=s
s
n+1
∗
2
xNA (p)
Λ − kP 2
∗
πNA (p) =
=s
s
n+1
π∗A (p) =
(3.39)
The Private Equity firm will then maximize the following:
Λ + nkP
Max : s
n+1
{k}
2
Λ − kP
−s
n+1
2
− T − μP kP2
(3.40)
The F.O.C. with respect to kP is then:
2s Λ − kP∗
2sn Λ + nkP∗
+
= 2μP kP∗
n+1 n+1
n+1 n+1
(3.41)
sn (Λ + nkP∗ ) + s (Λ − kP∗ ) = μP (n + 1)2 kP∗
(3.42)
sΛ(n + 1) + s(n2 − 1)kP∗ = μP (n + 1)2 kP∗
(3.43)
Which gives:
kP∗ =
s(n + 1)Λ
μP (n + 1)2 − s(n2 − 1)
(3.44)
For this to be defined, the following has to hold:
μP (n + 1)2 − s(n2 − 1) > 0
(3.45)
which suggests:
μP >
s(n2 − 1)
(n + 1)2
.
(3.46)
87
88
538
142
-0.0283***
(0.00617)
-0.00163
(0.00159)
0.0635***
(0.0123)
394
101
1.0656
0.9984
0.9721
1.0023
1.0496
0.9993
190
48
-0.0333*
(0.0188)
-0.00534
(0.00675)
-0.0913
(0.0687)
-5.112
(10.09)
174
44
Yes
-0.0342
(0.0211)
-0.00957
(0.00902)
-0.0701
(0.0671)
-0.00863
(0.00584)
1.046
(10.93)
0.0060
0.9127
0.9947
0.9672
Yes
2.8467
0.9914
0.9323
0.9905
0.9663
Base Models
(FE)
(OR)
0.9508
0.9853
0.9982
Yes
Yes
0.9865
0.9634
0.9763
0.9959
(OR)
(RE)
1.992***
(0.258)
538
142
-0.0163***
(0.00369)
-0.00108
(0.000829)
0.0294***
(0.00673)
1.499***
(0.278)
394
101
-0.0135***
(0.00447)
-0.000690
(0.000844)
0.0280***
(0.00736)
0.00162***
(0.000557)
*** p<0.01, ** p<0.05, * p<0.10.
190
48
-0.00180
(0.0171)
-0.0148
(0.0106)
-0.0505
(0.0658)
174
44
Yes
Yes
-0.00411
(0.0186)
-0.0240
(0.0165)
-0.0372
(0.0646)
-0.0136*
(0.00756)
(FE Y)
7.3278
1.0298
0.9989
0.9838
4.4774
1.0016
1.0284
0.9993
0.9866
-0.0372*
(0.0197)
-0.00456
(0.00662)
-0.0813
(0.0655)
-4.457
(10.26)
-0.712**
(0.347)
190
48
-0.0381*
(0.0224)
-0.00940
(0.00894)
-0.0581
(0.0650)
-0.0105
(0.00638)
3.207
(11.23)
-0.797**
(0.379)
174
44
Yes
0.4909
0.0116
0.9219
0.9954
0.9635
Yes
0.4505
24.7022
0.9896
0.9435
0.9906
0.9626
-1.036***
(0.382)
190
48
-0.000575
(0.0172)
-0.0177
(0.0116)
-0.0411
(0.0677)
-1.258***
(0.433)
174
44
Yes
Yes
-0.00244
(0.0191)
-0.0315*
(0.0170)
-0.0185
(0.0665)
-0.0186**
(0.00835)
Models with lagged dependent variable
(OR)
(FE)
(OR)
(FE Y)
0.3550
0.9598
0.9824
0.9994
Yes
Yes
0.2844
0.9816
0.9816
0.9690
0.9976
(OR)
Note: Columns 1-6 present Base model results, and columns 7-12 present results from regressions including the lagged dependent variable. RE stands for Random Effects, FE for
(conditional) Fixed Effects estimations, and OR stands for Odds Ratio. Standard errors are reported in parentheses.
Observations
N of Countries
LBOt−1
Log Population
Property Rights
Openness
Business Cost
Panel B
Observations
N of Countries
CountryControl
Year Control
LBOt−1
Log Population
Listed Comp
Property Rights
Openness
0.9793
Business Cost
-0.0210***
(0.00648)
-0.000725
(0.00134)
0.0484***
(0.0110)
0.00231***
(0.000783)
(OR)
(RE)
Panel A: with Listed Companies
Table 3.10. Panel Logit Results of LBO instance on All Variables of Interest
A.3 Results
Table 3.11. The Number of Tax Haven Outcomes
Country
Antigua and Barbuda
Barbados
British Virgin Islands
Cayman Islands
Channel Islands
Falkland Islands
French Polynesia
Netherlands Antilles
Seychelles
Saint Kitts and Nevis
Saint Lucia
Turks and Caicos Islands
LBOs
Other M&As
Share of LBOs
0
2
10
8
20
0
0
2
0
1
0
0
22
25
233
73
81
1
5
29
8
1
4
7
0
0.07
0.04
0.10
0.20
0
0
0.06
0
0.50
0
0
Note: LBOs are the number of LBOs in the given country for the period, other M%As are what the paper
considers to be incumbent takeovers. Share of LBOs is the outcome variable in the empirical analysis.
89
90
Yes
*** p<0.01, ** p<0.05, * p<0.10.
Yes
Yes
Yes
534
0.028
137
-7.36e-06***
(2.03e-06)
-0.0883
(0.280)
-0.137**
(0.0606)
Model with Lag
5
Yes
Yes
812
0.025
149
-0.0678
(0.0520)
-3.14e-05***
(6.47e-06)
6
Note: Columns 1-3 present Base model results, and columns 4-6 present results from regressions including the lagged dependent variable for countries who are not tax havens.
Within R-squared values reported. Robust standard errors are reported in parentheses and all standard errors are clustered by country.
Country Control
Year Control
162
N of Countries
149
812
906
0.021
162
-2.57e-05
(1.73e-05)
906
615
0.005
156
-2.45e-05***
(5.59e-06)
4
Observations
-5.69e-06**
(2.72e-06)
-0.337
(0.355)
3
0.140**
(0.0591)
-2.53e-05***
(7.97e-06)
Base Model
2
Share LBO
Log Population
Business Cost
1
Table 3.12. Panel Regression Results of Share of LBO on Business Startup Costs without Tax Havens
91
987
0.331
110
8.16e-07
(2.01e-05)
1.23e-06
(2.45e-06)
1,26
0.308
172
6.90e-06
(1.96e-05)
106
172
Yes
Yes
*** p<0.01, ** p<0.05, * p<0.10.
Yes
146
Yes
978
0.088
146
-0.00122***
(0.000365)
-0.286
(0.194)
0.152***
(0.0554)
Yes
832
0.083
106
-0.00108***
(0.000338)
-1.07e-05
(1.57e-05)
-0.212
(0.172)
0.155***
(0.0563)
Model with Lag
5
Yes
Yes
978
0.272
146
0.00272
(0.0509)
-0.000646**
(0.000273)
Yes
Yes
832
0.311
106
-0.0180
(0.0491)
-0.000490**
(0.000223)
-5.25e-06
(1.86e-05)
6
Note: Columns 1-3 present Base model results, and columns 4-6 present results from regressions including the lagged dependent variable for countries who are not tax havens.
Within R-squared values reported. Robust standard errors are reported in parentheses and all standard errors are clustered by country.
Country Control
Year Control
978
1,26
0.083
172
-0.000679**
(0.000315)
Yes
Yes
Observations
R-squared
Number of ctry
-0.00189***
(0.000518)
-0.656***
(0.196)
Yes
0.301***
(0.0497)
1,26
1.97e-05
(3.77e-05)
110
Share LBOt−1
Log Population
Openness
Panel B
Country Control
Year Control
832
987
0.084
110
-0.000453
(0.000291)
-3.28e-05*
(1.74e-05)
4
Observations
R-squared
Number of ctry
-0.00157**
(0.000621)
-5.71e-05**
(2.33e-05)
-0.807***
(0.229)
3
0.312***
(0.0537)
987
-0.000158
(0.000106)
-1.02e-05
(8.86e-06)
Base Model
2
Share LBOt−1
Log Population
Listed Comp
Openness
Panel A: with Listed Companies
1
Table 3.13. Panel Regression Results of Share of LBO on International Openness without Tax Havens
Table 3.14. Panel Regression Results of Share of LBO on Property Rights without Tax
Havens
1
Property Rights
0.00167***
(0.000329)
Log Population
Base Model
2
-0.000834
(0.00130)
-1.124***
(0.284)
4
-0.00223**
(0.00112)
0.00101***
(0.000223)
Share LBOt−1
Observations
N of Countries
Country Control
Year Control
1,046
149
Model with Lag
5
3
1,037
0.040
148
1,046
0.328
149
Yes
Yes
Yes
*** p<0.01, ** p<0.05, * p<0.10.
0.287***
(0.0513)
811
131
6
-0.000934
(0.00116)
-0.600**
(0.253)
0.176***
(0.0637)
803
0.062
130
-0.00195*
(0.00103)
Yes
Yes
Yes
-0.0206
(0.0597)
811
0.301
131
Note: Columns 1-3 present Base model results, and columns 4-6 present results from regressions including
the lagged dependent variable for countries who are not tax havens. Within R-squared values reported.
Robust standard errors are reported in parentheses and all standard errors are clustered by country.
92
4. Globalization, Chinese Imports, and Skill
Premia in a Small Open Economy1
4.1 Introduction
The decrease in trade costs due to innovations in communication technologies
and lower transportation costs has allowed goods from developing countries
to increase their presence in the markets in developed countries, especially
in the past 20 years. Higher import levels from low-wage countries increase
competition in domestic markets, which may put downward pressure on wages
of workers performing tasks that are in direct competition with production
in low-wage countries, as well as increase the relative demand and wages of
workers who perform work that is complementary to such imports. This may
have contributed to widening wage inequality in developed economies, which
has been widely observed and studied during the same period.2 Figure 4.1
shows that while real wages of both college and non-college educated workers
in Sweden have been on the rise, wages for the college educated have risen
faster in the period 1996-2007.
However, it is difficult to single out the effect of globalization on wage dispersion, as there are other simultaneous forces at play (Freeman, 2009 and
Feenstra and Hanson, 1999). One such potential force is skill-biased technological change, which increases the relative productivity of skilled workers
and thus contributes to rising wage inequality between workers of different
skill types.3 In this regard, the rapidly rising import levels from China present
an interesting opportunity to isolate the effects of increased competition from
low-wage countries, as the Chinese accession to the World Trade Organization in 2001 may be taken as an exogenous trade shock. China has increased
its exports to the rest of the world seven fold in the nine years following its
membership compared to doubling it in the nine years prior. Thus, looking at
Chinese import penetration into an economy will allow for the estimation of
the response of skill premia of college educated workers to a large exogenous
import shock.
1 I thank Nils Gottfries for useful discussions; as well as Teodora Borota Milicevic, Mikael
Carlsson, Rita Ginja, Georg Graetz, Fredrik Heyman, Francis Kramarz, Omar Licandro,
Mikael Lindahl, Oskar Nordström Skans, Alex Solis, and Edoardo di Porto for comments
and suggestions. The author gratefully acknowledges financial support from the WallanderHedelius-Browaldh Foundation.
2 See Acemoglu and Autor (2011) for a review of the literature.
3 See Griliches (1969) for the theoretical contribution on skill-biased technological change,
and Autor et al. (2003) for an application using computerization.
93
.2
12.4
.3
Premium
Log Real Wage
12.6
12.8
.4
13
Choosing China as a representative developing country also makes it possible to use Chinese imports in similar economies as a measure of the exogenous
shock to domestic product and labor markets. This approach helps address a
potential simultaneity problem in the estimation, where an unobserved domestic shock could be affecting wages and the level of Chinese imports at
the same time, or a reverse causality problem where an initial rise in manufacturing wages could be causing the increase in manufacturing imports from
China.
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Non-College
Premium - Right Axis
College
Figure 4.1. Log real wages by education groups, and the college premium over the years. For the
period 1996-2007 wages for both college and non-college educated workers have risen, but the difference (Premium) shows a positive trend, implying a rising college premium for the period. The sample
includes workers bown between the years 1920 and 1991 earning a real (in 2010 values) income of at least
SEK120,000 annually, employed at firms who employ at least 5 workers each year they are active.
I use a rich linked employer-employee database covering all privately owned
Swedish manufacturing firms from 1996 to 2007 to study the effects of import
penetration on the wage gap between college educated (high-skilled) and noncollege educated (low-skilled) workers in Sweden.4 The paper contributes to
the existing literature using matched employer-employee data in estimating
both the wage effect of an import shock on the total population of workers, including job switchers, and also on workers who remain employed at the same
firm. Also, I allow for separate exogenous trends for skilled and unskilled
workers in the estimations in order to isolate the wage effects of Chinese imports from changes in the return to skill caused by technology and other unobserved factors.
As cheaper Chinese goods become available in the domestic market, workers who produce goods that are in direct competition with Chinese imports are
4 Nordström
94
Skans et al. (2009) provide a good summary of the Swedish labor market.
the ones whose jobs and wages are threatened, while workers who are complementary to these imported goods stand to benefit from rising imports. For
the first group, as the demand for these low-skilled workers goes down, wages
offered to them will decline as well.5 This effect could manifest itself in lowskilled wages in two ways. One part of the adjustment mechanism could be
that the declining demand for low-skilled workers would lower their wages
on the same job. A specification with match fixed effects is used to measure
changes in return to skill as a response to changes in the competitive environment during a specific job spell. However, another part of the adjustment could
take place via the displacement of low-skilled workers. Similarly, an increase
in the wages of workers who are complementary to imports may occur either
on the same job or it can be related to job switches. Thus, an estimation with
only person fixed effects is introduced in order to capture both of these mechanisms by looking at the response of wages in the sector to changes in import
competition levels controlling for the workers’ observable and unobservable
characteristics.
The analysis is repeated using an instrumental variables approach to allow
a causal interpretation and to measure the magnitude of the impact of an exogenous change in imports on wages of skilled and low-skilled workers.
I find that higher Chinese imports have contributed to the rise in the wage
gap between skill types in the labor force. The effect comes from significantly
higher wages for highly skilled workers in the overall economy. This result
remains robust even after controlling for a separate time trend in the return to
higher education. One percentage point increase in Chinese import penetration results in about 1 percent higher wages for the college educated. Since the
cross industry average change for the period is about 2.9 percent, this translates into an increase in the average college premium of about 2.8 percent in
manufacturing jobs. Given that real wages for college educated workers in the
sample went up by about 27 percent in this period, the rise in Chinese import
penetration explains roughly 10 percent of the rise in wages of skilled workers
in Sweden.
When it comes to the effects of Chinese imports on the wages of low-skilled
workers, I find less robust effects. Without education-specific trends, I find
negative effects on the wages of low-skilled workers in response to an increase
in Chinese imports. However, when I include education-specific trends, rising
Chinese imports do not have a statistically significant effect on low-skilled
wages. A potential explanation for the lack of an effect on low-skilled wages
could be the institutional setting in the Swedish labor market with union contracts imposing a lower limit on wages.
5 See
Katz and Autor (1999) for details on the supply-demand framework and its effects on
changes in the structure of wages. In their recent work, Goldin and Katz (2008) confirm that the
change in wage differences between workers with different education levels in the U.S. since
the 1980s are explained by a combination of higher demand for skilled labor and changes in the
supply of different skill types to the labor market.
95
Recent empirical work looking at the effect of rising Chinese import penetration has not reached a consensus regarding the effect of increased Chinese
imports on wages. Focusing on local labor market outcomes for the U.S., Autor et al. (2013) examine the effect of industry-level Chinese imports on local
labor markets. They find that increased import exposure reduces wages in
non-manufacturing sectors for both college and non-college workers, but they
do not find a significant effect on wages in manufacturing. On the other hand,
Alvarez and Opazo (2011) use Chilean data on firm-level wages, and find that
higher Chinese imports have a negative effect on low-skill wages in manufacturing industries. This paper differs from these studies by using a detailed
matched employer-employee database and by finding a robust positive wage
effect of higher imports on skilled wages in manufacturing industries. With
worker-level wages that cover the entire population of manufacturing workers
in a small open economy, the Swedish data allows for a better identification of
the effects of an exogenous trade shock on individual wages.
The paper that is closest to this study is Ashournia et al. (2014), which
uses similarly linked employer-employee Danish data to examine the effects of
changes in both industry-level and also firm-level measures of Chinese import
penetration on individual wages. Employing the within job spell specification,
they find a negative effect of Chinese import penetration for low-skilled workers using the firm-level measure, but their estimates do not show any negative
effect when using the industry-level measure. I use an industry wide measure of Chinese import penetration, and my results are similar to Ashournia et
al. (2014) when considering wage changes of low-skilled workers within job
spells. When job switchers are included, I find negative effects on the wages
of low-skilled workers when I exclude education-specific trends. However,
this effect disappears when education-specific trends are included. Thus, in
line with the results of Ashournia et al. (2014), I conclude that there are no
robust negative effects on low-skilled wages in response to rising imports from
China.
The rest of the paper proceeds as follows. Section 4.3 presents the empirical
strategy, and Section 4.4 describes the data. Section 4.5 presents the results
and performs robustness checks. Section 4.6 presents extensions, and Section
4.7 concludes the paper.
4.2 Theoretical Motivation
A rise in imports from a low-wage country will affect a domestic manufacturing firm differently depending on whether the firm is in direct competition with
the imports or if the firm is using imports as intermediary inputs for its production of a final good. In the first scenario, domestic firms respond to tougher
competition in the product market by either shutting down, or cutting costs in
order to be able to stay in the market. As a result, the demand for low-skilled
96
workers who produce these goods should go down, and wages offered to them
should follow suit. Perfect competition in the labor market would imply that
wages for low-skilled workers everywhere in the sector should move in the
same direction. This would suggest not only a reduction in the wages offered
to job seekers, but also a reduction in the wages of low-skilled workers who
remain employed at the same firm. However, in the presence of labor market
frictions such as compositional differences (Bustos, 2011), downward wage
rigidities (Akerlof et al.,1996 and Dickens et al., 2007), search and matching
frictions (Helpman et al., 2010), or efficiency wages (Shapiro and Stiglitz, 198
and Amiti and Davis, 2011), it is possible to observe wage differences between
workers with the same observable skills who are employed at different firms.
Workers may seek and find jobs through a job search process. One of the
significant contributors of the wage for the worker could be the specific match
that she has with her current firm. If a worker were to lose her job, she would
have to seek employment elsewhere and could end up with a worse match
where the contribution to her wage will be lower reflecting the lower value of
the match. Meanwhile, workers of the same skill type and observable characteristics who keep their employment (i.e. particular match) will have a higher
wage profile, creating a wedge between the wages of movers and stayers.
Another mechanism that could result in this wedge is the presence of firmspecific human capital, which increases the value of a worker in the current
firm. In order to keep the worker and minimize costs associated with turnover,
the firm would then be willing to pay a premium to the worker above their
outside options in competing firms. In the event the firm closes down, the
worker loses the firm-specific capital and the wage premium that came with
it6 .
Finally, there could be contractual/institutional frictions in the market, as
well as motivational reasons (Campbell and Kamlani, 1997), which make
wages for workers already employed downward sticky. Any one of these factors, or a combination of them, would manifest itself in the data as a difference
in the wage developments between workers within the same skill group who
change jobs and those who keep their jobs.
In the second scenario, firms that produce final goods could substitute either their own production of intermediate inputs with imports from China
(which would reduce their demand for low-skilled workers), or switch their
current purchases of intermediate goods to cheaper inputs from China. Either approach would reduce production costs, increase the marginal product
of high-skilled workers who are complementary to these inputs (whereas the
low-skilled workers are substitutes), and result in higher wages for them.7
6 Jacobson
et al. (1993) find evidence of this in displaced long-tenured workers using administrative data from Pennsylvania, U.S.
7 This is also in line with the trade-induced technological change literature, where higher
competition from low-wage countries can lead to productivity increases in firms in developed
countries. The forthcoming work Bloom et al. (2015) finds that higher import competition
97
Firms may even be prompted to specialize and expand their production on the
higher end as a response to the supply of cheaper inputs, adding to the demand
for, and therefore increase in, the wages of high-skilled workers. One such avenue that firms could follow would be to offshore their low-skill production
overseas to a low-wage country, instead of producing their own intermediate
inputs domestically, as discussed in Grossman and Rossi-Hansberg (2008).8
Firms could then refocus their domestic activities on high-skill production,
which should increase their demand for high-skilled workers.
High-skilled workers will face high demand both in their current place of
work, and also in their outside option. This may create an ambiguous effect on
the relative wages of high-skilled workers who move compared to those who
stay on at the same firm. On the one hand, high-skilled workers can collect
return to tenure if they stay in the firm where they are already employed. In
the presence of labor market frictions and costs associated with posting, advertising and filling vacancies, the firm may be willing to pay higher wages to
high-skilled employees in order to keep them employed, which could result in
an additional premium for those who stay employed in the same firm. On the
other hand, workers who choose to switch firms when they are facing high demand both in and outside their place of work could be the ones who are offered
a better salary, beyond the tenure contribution, in their new firms. Whether the
wages of those who stay on the same job increase more or less compared to
the population that includes switchers in response to an exogenous trade shock
will depend on how these forces will compare with each other.
4.3 Empirical Strategy
To investigate the effects of higher imports from low-wage countries on the
wage dynamics of low- and high-skilled laborers, I take Chinese imports to
Sweden to represent low-wage developing countries that have increased their
imports to the domestic economy. The Chinese share of manufacturing imports to Sweden has grown about fourfold in the period analyzed. Currently,
China is Sweden’s largest trade partner among developing economies, and was
the 9th largest import source overall in 2012, accounting for about 4% of total
imports. The Chinese membership in the WTO in 2001 presents one with a
clear cutoff point, which can be taken as the source of an exogenous shock
to local markets in Sweden, thereby creating an excellent opportunity to study
from China leads to a higher technological change and a lower share of unskilled workers in
firms. Since higher technological change is complementary to high-skilled workers, it therefore
decreases the demand for, and the share of, low-skilled workers in these firms.
8 In their paper, Grossman and Rossi-Hansberg (2008) present a model of offshoring in
which a firm outsources low productivity jobs and can therefore focus on high-end specialization, which yields a higher level of productivity at the firm.
98
the effects of imports from low-wage countries on the wages of the low-skilled
and on the skill premia of college educated workers.
However, wages could be affected by shocks to the domestic economy that
simultaneously affect imports from China, thereby creating a spurious correlation. Further, reverse causality may arise if the true reason for an increase in
imports from China was an initial increase in domestic wages for some group
of workers. To control for this, instead of using Chinese import penetration
figures for Sweden, I use Chinese import penetration into a group of northern
European countries as a measure of the exogenous trade shock to the Swedish
economy.
These countries - Denmark, Finland and Germany - have been exposed to
trade with China in a similar way in manufacturing industries as shown in
Table 4.1. Figure 4.2 shows that the share of Chinese imports as a share of
apparent consumption in both Sweden and the group of countries follow a
similar pattern over time.
16-Tobacco
17-Textiles
18-WearingApparel
19-Leather
20-Wood
21-Pulp&Paper
22-Publishing&Media
23-Coke&Petrol
24-Chemicals
25-Rubber&Plastic
26-NonMetalMineral
27-BasicMetal
28-Metal
29-Machinery&Equipment
30-OfficeMachinery&Computer
31-ElectricMachinery
32-RadioTvCommunication
33-Medical&Optic
34-MotorVehicle
35-OtherTransport
36-Furniture
0 .2 .4 .6
0 .2 .4 .6
0 .2 .4 .6
0 .2 .4 .6
15-Food
2000
2004
2007 1996
2000
2004
2007 1996
2000
2004
2007
0 .2 .4 .6
1996
1996
2000
2004
2007 1996
2000
2004
2007
year
Composite
Sweden
Graphs by Industry
Figure 4.2. Swedish imports from China as a share of respective apparent consumptions compared to the composite measure of Denmark, Finland, and Germany, by
industry, 1996-2007
99
100
0.0001
0.0000
0.0081
0.0736
0.0035
0.0009
0.0001
0.0002
0.0018
0.0011
0.0004
0.0021
0.0005
0.0025
0.0003
0.0013
0.0035
0.0029
0.0008
0.0002
0.0001
0.0038
Share in
1996
0.0005
0.0000
0.0714
0.2495
0.0922
0.0053
0.0031
0.0018
0.0003
0.0049
0.0096
0.0222
0.0048
0.0210
0.0070
0.0337
0.0542
0.0674
0.0056
0.0028
0.0107
0.0684
0.04
0.00
6.33
17.59
8.86
0.44
0.30
0.16
-0.15
0.37
0.93
2.01
0.43
1.84
0.66
3.24
5.07
6.45
0.48
0.27
1.05
6.46
Sweden
Share in
Change
2007
% points
359
.
779
239
2524
519
3117
716
-84
333
2620
976
889
728
1938
2456
1440
2247
610
1641
7354
1692
Growth
%
0.0004
0.0008
0.0129
0.1224
0.0041
0.0021
0.0010
0.0007
0.0008
0.0051
0.0006
0.0034
0.0011
0.0059
0.0012
0.0184
0.0132
0.0234
0.0089
0.0006
0.0002
0.0035
Share in
1996
0.0006
0.0030
0.0670
0.5344
0.0346
0.0121
0.0055
0.0072
0.0011
0.0101
0.0069
0.0202
0.0147
0.0241
0.0127
0.3110
0.0771
0.1261
0.0160
0.0096
0.0253
0.0556
0.01
0.22
5.42
41.20
3.05
1.00
0.45
0.64
0.03
0.49
0.62
1.67
1.36
1.82
1.15
29.26
6.39
10.26
0.71
0.90
2.52
5.22
Composite Measure
Share in
Change
2007
% points
27
279
420
337
753
489
448
908
31
96
981
485
1263
307
942
1590
485
438
80
1509
14327
1508
Growth
%
Note: 1996 and 2007 figures are calculated using Equation 4.1. Composite figure is an average of Share of Chinese Imports in each of the manufacturing industries in Denmark,
Finland, and Germany. Growth figures are calculated as the percentage growth over the 1996 values, Change figures take the difference of the shares in these two years, and reports
it in percentage points.
15-Food
16-Tobacco
17-Textiles
18-WearingApparel
19-Leather
20-Wood
21-Pulp&Paper
22-Publishing&Media
23-Coke&Petrol
24-Chemicals
25-Rubber&Plastic
26-Non-metallicMineral
27-BasicMetals
28-Metal
29-Machinery&Equipment
30-OfficeMachinery&Comp
31-ElectricMachinery
32-RadioTvCommunic
33-Medical&Optic
34-MotorVehicle
35-OtherTransport
36-Furniture
Industries
Table 4.1. Chinese Import Share as a Share of Apparent Consumption in Sweden, and Composite Countries by Industries
In assessing the shocks to Chinese imports, my specific import penetration
measure uses the average of the share of Chinese imports in apparent consumption in Denmark, Finland and Germany in the style of Bernard et al.
(2006). The advantage of this measure is that it takes the size of the domestic
industry relative to imports into account:
CMPkt =
China
Mkt
.
MktTotal + Qkt − Xkt
(4.1)
Qkt and Xkt represent domestic total output and exports in industry k and year
t, respectively. As a robustness check, the analysis is replicated by using alternative measures of Chinese import penetration, such as (i) the share of Chinese
imports in total imports in Denmark, Finland and Germany, and (ii) Chinese
exports to the rest of the world (excluding Sweden). The results are reported
in Section 4.5.3.9
The effect of higher import competition on wages and the relative skill premium of college educated workers is estimated in two ways, building on the
Mincer wage equation.
Total Effect
The first approach aims to measure the overall effect of Chinese import penetration on manufacturing wages. Controlling for person fixed effects and
observables, this reduced form approach allows for identification of the effect
of an exogenous import shock on the return to college education. The equation
of interest for person i, firm j, industry k, and year t is:
Composite
Composite
logwi jt = θ1CMPkt
+ θ2CMPkt
+ αi + xit βt + τt + εi jt ,
XCollegeit
(4.2)
where wi jt is the log real earnings for individual i who works in firm j at
time t, τt captures the common time component, αi captures the return to the
worker’s unobserved characteristics regardless of place of employment, xit is
a vector of time-varying individual observables, such as age and education.
The import penetration measure is entered as is, and is also interacted with an
indicator for some college education or more in order to single out the effect
on the skill premium. θ1 will then capture the effect of a change in Chinese
import penetration on the wage of non-college educated workers, and θ2 will
show the effect on the return to some college education or more (i.e. the effect
on the skill premium).
Within Job Spell Variation
The alternative approach introduces a fixed component for each worker and
firm match instead of the person fixed effect in the previous specification. This
9 Table
4.25 shows that these measures are highly correlated with each other.
101
makes it possible to measure how much of the wage adjustment takes place
during the same job spell in response to rising imports from China. Match
specific effects take account of the fact that wages could be affected not only
by person and firm characteristics, but also by complementarities that could be
present in the particular match of a worker to a firm that are unobservable in
the data. For example, there could be co-worker complementarities due to the
present composition of workers in a firm, which could have an effect on wages
and returns to skill at the firm. The coefficient of interest will then capture the
effect of Chinese import penetration on the wage within the job spell, that is,
when the worker is employed at the same firm:
Composite
Composite
logwi jt = θ1CMPkt
+ θ2CMPkt
+ xit βt + z jt γ + ΨJ(i,t) + τt + εi jt ,
XCollegeit
(4.3)
Composite
is the Chinese
for person i, firm j, industry k, year t, where CMPkt
import penetration in industry k and year t. The term ΨJ(i,t) is the firm-person
effect for firm j that worker i is employed at in year t and captures the average
wage during a job spell of the worker in a particular firm. Person observables
are denoted by xit and firm observables are denoted as zjt .
In both of these approaches, the sector assignment k refers to the sector
where the worker is employed at time t. To eliminate potential endogeneity
problems due to endogenous movements between sectors, an additional set
of estimates are performed, where sector k is fixed as the industry where the
worker first worked in when she appeared in the data.
Instrumental Variables Strategy
To allow for an interpretation of the magnitude of the effects of Chinese import
penetration on wages I use an instrumental variables strategy. This method
uses a Swedish measure of Chinese import penetration, CMPktSwe as a right
hand side variable. This Swedish measure is calculated as the share of Chinese
imports over Swedish apparent consumption in the style of Equation 4.1.
Part of the increase in Chinese imports to Sweden could have been caused
by rising manufacturing wages in Sweden, or some other unobserved shock
that affects both Chinese imports and wages in Sweden. To overcome these
potential simultaneity and reverse causality problems, I instrument CMPktSwe
with the composite measure of foreign Chinese import penetration. Since imports to all the countries of interest were affected by Chinese accession to the
WTO, I expect the instrument to be correlated with the Swedish import penetration measure. For the instrument to be valid, I need to assume that the
Composite
is uncorrelated with unobforeign import penetration measure CMPkt
served shocks to wages in Sweden. In this estimation, first the following two
102
first stage equations are evaluated:
Composite
Composite
+ θ2CMPkt
CMPktSwe = θ1CMPkt
+ Ω + xit βt + τt + εi jt ,
XCollegeit
(4.4)
where the composite Chinese import measure is an instrument for the Swedish
import measure, and,
Composite
Composite
+ θ2CMPkt
CMPktSwe XCollegeit = θ1CMPkt
+ Ω + xit βt + τt + εi jt ,
XCollegeit
(4.5)
where the composite interaction term is an instrumental variable for the Swedish
interaction term.10 The term Ω stands for the covariates that are unique to each
of the two specifications: for the reduced form it stands for the person fixed
effect αi , and for the person-firm match approach, it represents zjt γ + ΨJ(i,t) .
Next, predicted values from Equations 4.4 and 4.5 are used in place of the
import penetration and the interaction terms in the second stage regressions of
log real wage, in the style of Equations 4.2 and 4.3.
4.4 Data
The linked employer-employee data is accessed through a private, but nonexclusive, link from Statistics Sweden (SCB), which provides a confidential
database where both individuals and firms are reassigned anonymous identifiers. The components of this database, as well as supplementary data from
other sources, are detailed below:
Firm data
Firm-level data on wage sums, sales, profit, capital, number of employees,
firm age, and industry classification come from Statistics Sweden’s Business
Statistics (FEK), supplemented by legal and controlling ownership of the firm
from the Business Register Database (Företagsregistret). Although FEK data
is available from 1980 onwards, I focus on the period after 1996 in order to
have the full sample of non-imputed firms. Since the Swedish industry codes
(SNI) were reported in three different systems (1992, 2002 and 2007) in the
period focused on, I converted all the codes to the SNI2002 system in order to
obtain one continuous industry index at the four digit level. The conversions
primarily utilized correspondence keys obtained from SCB, but in the few
instances where the key was not successful in matching splitting industries,11
10 Ashournia
et al. (2014) use a similar approach, but their instrument is Chinese exports
to the World, excluding Denmark. I discuss and present results using alternative measures in
Section 4.5.3.
11 These are a total of 3 cases at the two digit level in the 1992 to 2002 conversion, and 2
cases at the two digit level in the 2007 to 2002 conversion
103
I assigned a match based on the best fit of the description between sectors.
Specifics of the match are detailed in the Appendix.
Worker data
The worker side of the database, containing information on annual taxed wage
income, age, gender, highest level and field of education, is originally collected by the Swedish Tax Authority (Skatteverket) and is linked to our database
under the name Register Based Labor Statistics (RAMS) maintained by Statistics Sweden.12 Each individual is linked to a firm/plant in accordance with
the International Labour Organization’s definition of being an employee in
the third week of November. The field of education classification follows the
UN’s International Standard Classification of Education (ISCED).
Industry Trade, Output, and Exchange Rate data
Data on international trade between China and Denmark, Finland, Germany,
Sweden and the rest of the world comes from UN COMTRADE database
(http://comtrade.un.org/), which classifies trade based on product level codes
using the Standard International Trade Classification, 3rd revision. The manufactured goods that are of interest for our analysis are mainly indexed by
material. Some subheadings under the classification "Manufactured goods"
in SITC correspond one-to-one with the Swedish industry codes (SNI), but
for the majority of instances the overlap is not exact. In those cases, I have
matched these product-level trade figures to the Swedish SNI codes used in
the firm-level data based on category descriptions (see Table 5.3 in the Appendix for details). As the UN data is based on materials, it is not possible to
identify international trade in Recycling. Therefore, I have excluded Swedish
recycling firms from the analysis (52 firms in 2007).
Danish, Finnish, German, and Swedish gross output and value added data
come from the OECD STAN database13 . Finnish and German values are reported in Euros, and converted to U.S. Dollars in order to match the COMTRADE data, using Eurostat’s annual exchange rates14 . The period prior to
1999 uses ECU to U.S. Dollars instead of Euro to U.S. Dollars exchange from
the same source. World GDP and manufacturing (value added) in current U.S.
Dollars are both from the World Bank’s World Development Indicators15 . Ex12 While
the education levels in RAMS are detailed into 5 groups from pre-high school to
graduate work, I grouped individuals into the following three educational categories: less than
high school diploma, high school diploma holders, and at least some college based on the more
detailed classification.
13 The database is accessible through the link
http://www.oecd.org/industry/ind/stanstructuralanalysisdatabase.htm
14 For more on the methodology and access to data see http://ec.europa.eu/eurostat/web/
exchange-rates
15 Historical database accessible at
http://data.worldbank.org/data-catalog/world-development-indicators
104
change rates from Swedish Kronor to U.S. Dollars are obtained as annual averages from the Swedish Central Bank, the Riksbank16 . Danish Krones are
converted to U.S. Dollars using the Danish Central Bank Nationalbanken’s
annual exchange rates17 . Current dollar values are put in 2010 dollars, using
CPI from the Bureau of Labor Statistics (http://www.bls.gov/cpi/). To convert all other monetary Swedish Kronor values to base year 2010, I use SCB’s
publicly available CPI series18 .
Sample Selection
Manufacturing firms that are either quasi or fully public sector owned/controlled
may not be as free to respond in their wage or hiring/firing decisions, and they
may also not be profit maximizing firms. Since this paper concerns itself with
the reactions of firms to changes in the trading environment, in the analysis I
only consider firms that are identified as either Limited Liability Partnerships
or Limited Liability Companies in the data.19
I focus the analysis on the population of limited liability manufacturing
firms with more than 5 employees in each year that they were active between
the years 1996 and 2007, where the years of focus are dictated by data availability.20 The matched employer-employee database covers the full population
of workers and firms starting in 1996, and data on industry-level gross output
from OECD is only available in a compatible industry classification system
until 2007.
Since the database does not provide information on whether the workers is
employed full- or part-time, I restrict the baseline sample to those who earn at
least SEK 120,000 a year (SEK 10,000 ≈ USD 1,570 a month). I also drop
individuals whose education level is unknown, those who are born before 1920
or after 1991, as well as those who are only present in the data for a single
year.21
16 The
data is accessible at http://www.riksbank.se/en/Interest-and-exchange-rates/
more on the methodology and series, please see http://www.nationalbanken.dk/en/
statistics/exchange_rates/Pages/Default.aspx
18 Indices and related statistics are available at
http://www.scb.se/en_/Finding-statistics/Statistics-by-subject-area/Prices-and-Consumption/
Consumer-Price-Index/Consumer-Price-Index-CPI/
19 It would have been ideal to further control for foreign/domestic ownership of the firm, as
the firm could be paying an additional foreign ownership premium to skill. Unfortunately, in this
dataset, firm ownership data is only available from 2002 onwards. Heyman et al. (2007), using
Swedish matched employer-employee data, show that at the individual level, foreign ownership
premium in Sweden is close to zero.
20 This restriction drops about 83% of all manufacturing firms, but about 72% of those firms
have an average employee of less than 1 during their presence in the data.
21 These restrictions on the individual drop about 400,000 people from the sample, about a
quarter of whom earned at most SEK 10,000 (≈ USD 1,570) annually. The restriction of being
present in the data for at least two years drops an additional 175,892 people.
17 For
105
The final sample is composed of about 10,500 manufacturing firms and
about 850,000 workers. Further details on the sample and data can be found
in Tables 4.9 and 4.10, and in the Appendix.
4.5 Results
4.5.1 Chinese Import Penetration
The regressions are run for the period 1996-2007. The period is determined by
the availability of industry-level gross output data to construct the measure of
Chinese import penetration in Equation 4.1. Since I employ estimations with
fixed effects for workers on matches, the exercise is performed only for workers who are active for at least two years in the data in each specification. In the
following tables the results are presented in two separate panels depending on
the specification used. Both panels show results from estimates with or without firm observables, and with or without year trends in return to investment
in college education.
Total Effect
The baseline results are presented in Table 4.2 Panel A. For the overall population, Chinese import penetration has no significant effect on wages (Column
1). But when the import penetration and college education interaction term
is added in Column 2, we see that manufacturing workers with no college
education are negatively affected by rising Chinese imports, whereas wages
of college educated workers rise with an increase in Chinese imports into the
sector. This is in line with the expectation that rising import penetration from
a low-wage country should hurt the workers whose jobs are in direct competition with those in the developing country. On the flip side, the demand for,
and therefore the relative wages of, high-skilled workers go up with higher
Chinese import penetration, as can be seen from the marginal effect on the
college educated population at the bottom of Panel A. Imports appear to benefit workers whose work is complementary to Chinese imports. The effects on
the non-college and college educated population remain robust to the inclusion
of firm observables in Column 4.
As an additional check, I introduce a year-education dummy into the equation in order to capture any time trend in the return to education that may
be present in the economy, motivated by factors that go beyond trade effects.
Columns 5 to 8 in Panel A show that such an adjustment takes away the significant negative overall wage effect on non-college workers, which suggests
that increased competition with China does not have any additional effects on
the wages of non-college workers beyond the time trend. Higher import levels
from low-wage countries contribute positively to having college education in
the overall economy, as the coefficient on the skill premium is still positive.
106
While most of the debate on the wage effects of higher import penetration
from low-wage countries have focused on the negative wage outcomes of lowskilled workers, the data shows that the significant effect is actually of a higher
skill premium for skilled workers.
Within Job Spell Variation
This section examines the within job spell response of high- and low-skilled
worker wages to changes in competition from China. Table 4.2 Panel B shows
results where the person fixed effects from the total effects specification have
been replaced by a person-firm match fixed effect, where we have a slightly
different story. The positive effect on the skill premium is still present and
the marginal effect is larger for these college educated workers. The negative
effect for the low-skilled workers is halved in magnitude, and is no longer significant. This suggests that for the low-skilled worker, significant downward
adjustment in wages as a response to higher Chinese competition in the sector takes place primarily through job switches, and not necessarily on the job.
Perfect competition in the labor market would imply similar wage changes for
each skill type across the sector; however, comparing the results in Columns 2
and 4 in Panels A and B show that the wages of the employees who stay at the
same firm do not fall nearly as much. This is in line with previous literature
which finds downward wage rigidity on the job, and as detailed in Section 4.2
could be due to contractual rigidity, presence of search frictions, or efficiency
wages in the market. Comparing Panels A and B, we can conclude that in the
face of higher Chinese import penetration into the sector, low-skilled workers
who have to switch to another firm get a negative wage return from the move,
whereas those who stay on in the same match are less affected.
The regressions in Columns 3 and 4 control for firm and individual observables that could affect the wage outcome for the worker, such as age, firm’s
employment size, capital intensity, or the market presence of the firm. Being
employed at larger firms, compared to firms that employ between 5 to 10 employees, contribute positively to wages. Firms with higher sales also affect
wages positively, but, surprisingly, capital intensity does not play a significant
role. The main coefficients of interest are again immune to the inclusion of
these variables, just as in Panel A.
Columns 5 to 8 present results with year-education dummies. Similar to
the total effect estimations, when year-education controls are introduced, both
the return to education and the marginal effect are still positive for college
educated workers, but the results are no longer statistically significant. Worker
wages of either skill type are not affected by rising Chinese competition in the
sector on the job.
107
Table 4.2. Log Real Earnings and the Return to Skill - Yearly Industry for the Worker
(1)
——————–Baseline——————(2)
(3)
(4)
Panel A: Total Effect (Equation 4.2)
Chinese Import P.
0.220
(0.183)
Chinese Import P.*College
-0.194*
(0.103)
1.095***
(0.265)
Firm Size : 10-49
250-499
500+
log Capital Intensity
log Sales
0.829
0.829
0.900***
(0.249)
Panel B: Within Job Spell (Equation 4.3)
Chinese Import P.
0.289
-0.0883
(0.242)
(0.110)
Chinese Import P.*College
1.002***
(0.315)
Firm Size : 10-49
50-249
250-499
500+
log Capital Intensity
log Sales
R-squared
MarginalEff
StdErr
0.858
0.858
0.913**
(0.338)
—————Education Trend————–
(6)
(7)
(8)
-0.194*
(0.107)
1.087***
(0.264)
0.0286***
(0.00290)
0.0486***
(0.00683)
0.0559***
(0.00947)
0.0566***
(0.0125)
0.00110
(0.00109)
0.00465
(0.00290)
0.116
(0.122)
0.829
0.829
0.893***
(0.251)
0.831
0.831
0.336
(0.215)
0.831
0.831
0.328
(0.217)
0.284
(0.232)
-0.0838
(0.108)
0.983***
(0.307)
0.0238***
(0.00343)
0.0365***
(0.00666)
0.0444***
(0.00997)
0.0518***
(0.0121)
-0.00144
(0.00226)
0.0218***
(0.00648)
0.196
(0.172)
0.0810
(0.0814)
0.325
(0.241)
0.196
(0.168)
0.0870
(0.0821)
0.310
(0.235)
0.0246***
(0.00345)
0.0379***
(0.00685)
0.0457***
(0.0101)
0.0525***
(0.0117)
-0.000790
(0.00207)
0.0214***
(0.00649)
0.859
0.900**
(0.331)
0.859
0.0281***
(0.00288)
0.0477***
(0.00682)
0.0545***
(0.00940)
0.0541***
(0.0124)
0.000494
(0.00120)
0.00524*
(0.00284)
50-249
R-squared
MarginalEff
StdErr
0.213
(0.181)
(5)
0.0235***
(0.00344)
0.0362***
(0.00673)
0.0440***
(0.0101)
0.0511***
(0.0123)
-0.00211
(0.00281)
0.0223***
(0.00655)
0.858
-0.00665
(0.0629)
0.343*
(0.180)
0.113
(0.124)
0.0290***
(0.00287)
0.0488***
(0.00682)
0.0558***
(0.00971)
0.0557***
(0.0122)
0.00137
(0.00113)
0.00513*
(0.00276)
0.0246***
(0.00345)
0.0379***
(0.00687)
0.0456***
(0.0101)
0.0524***
(0.0118)
-0.000967
(0.00224)
0.0215***
(0.00651)
0.859
0.406
(0.305)
0.860
-0.00377
(0.0677)
0.331*
(0.178)
0.0291***
(0.00287)
0.0490***
(0.00681)
0.0562***
(0.00972)
0.0564***
(0.0123)
0.00153
(0.00112)
0.00496*
(0.00278)
0.860
0.397
(0.300)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany,
detailed in Equation 4.1. All columns have year and person fixed effects for 850,048 people (5,866,337 observations) in Panel A and year and
person-firm match fixed effects for a total of 1,016,790 matches (5,687,870 observations) in Panel B, and both panels include Age, Age2 and
Age3 as covariates. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Sample includes
only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least
two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories follow
Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+)
with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size equal or larger
than 5, but less than 10 employees. Columns (1) and (2) have the total population, (3) and (4) have workers who are employed at the same firm
during their entire tenure in the database.
108
4.5.2 IV Strategy
The first stage results of Equations 4.4 and 4.5 can be seen in Tables 4.11 to
4.14. I construct and employ a composite measure of Chinese import penetration that takes the average of Chinese imports over apparent consumption
in Denmark, Finland and Germany. These northern European counties are all
developed economies that face similar costs to trade with China, and the composite measure is highly correlated with the equivalent Swedish measure, as
can be seen in Table 4.25. The first stage regressions marked (a) correspond to
the regression without the interaction term. Next, columns marked (b) show
how good an instrument the composite Chinese import penetration is for the
Swedish measure, and columns marked (c) show how well the composite measure interacted with college education instruments the Swedish measure interacted with college education. From the estimates, we see that the instruments
satisfactorily identify the effects.
The second stage equations are run analogous to Equations 4.2-4.3 with log
person wages as the dependent variable, and with the predicted values from
Equations 4.4 and 4.5 replacing the Chinese import penetration measure and
interaction terms respectively. These results are discussed below.
Total Effect
Table 4.3 Panel A baseline estimates in Columns 1 to 4 show that in the total
population, an increase in Chinese imports translates into a negative wage outcome for low-skilled workers, and an increase in the wage premium for skilled
workers that is about seven times larger than the decline in low-skilled wages.
Manufacturing industries in Sweden as a whole saw an average increase of
2.9 percentage points in the Chinese penetration measure between 1996 and
2007. For the total population of workers, this translates into a decline of 1
percent in the wages of low-skilled workers, and a rise of 8 percent in the skill
premia of high-skilled workers for a total marginal effect of about 6.8 percent
in wages of the college educated.
Panel A columns 5 to 8 show the results of the estimates with controls for
trends in return to education. Here, the significance of the decline in lowskilled wages disappears, but the college educated workers still enjoy a significantly higher wage in response to higher imports from China. For an average
increase of 2.9 percentage points in Chinese imports, skilled workers enjoy
about 2.8 percent higher wages in the overall economy. Given that real skilled
wages for the sample in this period have gone up by about 27 percent, the
contribution of rising competition from low-wage countries amounts to about
10 percent of the rise in skilled wages.
Within Job Spell Variation
The results presented in Table 4.3 Panel B show a similar pattern as in the OLS
regressions. Both with and without controlling for firm characteristics, lowskilled workers do not see a significant decline in their wages within the job
109
Table 4.3. Log Real Earnings and the Return to Skill using the 2SLS Strategy - Yearly
Industry for the Worker
———————-Baseline———————
(1)
(2)
(3)
(4)
Panel A: Total Effect (Equation 4.2)
CMPSwe
0.591
-0.412*
(0.391)
(0.229)
CMPSwe *College
2.792***
(0.310)
Firm Size : 10-49
50-249
250-499
500+
log Capital Intensity
log Sales
R-squared
MarginalEff
StdErr
0.047
0.050
2.380***
(0.320)
Panel B: Within Job Spell (Equation 4.3)
CMPSwe
0.825
-0.149
(0.511)
(0.233)
CMPSwe *College
2.691***
(0.276)
Firm Size : 10-49
50-249
250-499
500+
log Capital Intensity
log Sales
R-squared
MarginalEff
StdErr
0.040
0.042
2.542***
(0.373)
0.578
(0.393)
——————-Education Trend——————
(5)
(6)
(7)
(8)
-0.412*
(0.240)
2.772***
(0.319)
0.0290***
(0.00277)
0.0493***
(0.00666)
0.0569***
(0.00883)
0.0573***
(0.0114)
0.00104
(0.000756)
0.00451*
(0.00245)
0.311
(0.275)
0.048
0.052
2.361***
(0.331)
0.058
0.058
0.966**
(0.424)
0.060
0.060
0.943**
(0.439)
0.816
(0.497)
-0.139
(0.234)
2.644***
(0.290)
0.0240***
(0.00284)
0.0370***
(0.00547)
0.0452***
(0.00824)
0.0517***
(0.00978)
-0.00141
(0.00183)
0.0213***
(0.00516)
0.559
(0.358)
0.232
(0.179)
0.999**
(0.419)
0.564
(0.355)
0.250
(0.180)
0.960**
(0.428)
0.0245***
(0.00288)
0.0379***
(0.00565)
0.0459***
(0.00833)
0.0520***
(0.00949)
-0.000560
(0.00160)
0.0215***
(0.00519)
0.044
2.506***
(0.384)
0.048
0.0283***
(0.00264)
0.0486***
(0.00670)
0.0561***
(0.00892)
0.0557***
(0.0114)
0.000803
(0.000827)
0.00492**
(0.00242)
0.0232***
(0.00287)
0.0360***
(0.00551)
0.0439***
(0.00839)
0.0500***
(0.00998)
-0.00155
(0.00206)
0.0228***
(0.00537)
0.042
-0.00851
(0.158)
0.975***
(0.310)
0.307
(0.282)
0.0291***
(0.00269)
0.0493***
(0.00655)
0.0567***
(0.00897)
0.0566***
(0.0110)
0.00154*
(0.000846)
0.00496**
(0.00231)
0.0244***
(0.00290)
0.0377***
(0.00568)
0.0456***
(0.00838)
0.0516***
(0.00949)
-0.000585
(0.00167)
0.0219***
(0.00527)
0.048
1.231**
(0.558)
0.050
-1.99e-05
(0.168)
0.943***
(0.321)
0.0292***
(0.00272)
0.0494***
(0.00654)
0.0569***
(0.00894)
0.0569***
(0.0110)
0.00159*
(0.000830)
0.00482**
(0.00233)
0.050
1.209**
(0.567)
*** p<0.01, ** p<0.05, * p<0.10.
Note: The table reports the results of the second stage regressions. All columns have year and person fixed effects for 850,048 people (5,866,337
observations) in Panel A and year and person-firm match fixed effects for a total of 1,016,790 matches (5,687,870 observations) in Panel B, and
include Age, Age2 and Age3 as covariates. Standard errors are reported in parenthesis, and are adjusted for clustering at the industry level for
the first industry of the worker. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values)
income of SEK120,000 working for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and
capital intensity. Firm Size categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium
size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms
by size are microfirms of size equal or larger than 5, but less than 10 employees. Columns (1) and (2) have the total population, (3) and (4)
have workers who are employed at the same firm during their entire tenure in the database.
110
spell contrary to the result in Panel A, but high-skilled workers have a positive
return associated with higher Chinese imports. Both the skill premium and the
marginal effect on college educated workers are between 7 to 8 percent of a
wage increase in response to an average of 2.9 percentage points increase in
Chinese imports. Panel B Columns 5 to 8 show that the estimations with yearcollege controls are very comparable to the total population effect in Panel A,
with upto 3.5 percent higher wages in response to an increase of about 2.9
percentage points in imports from China.
To conclude, the IV strategy shows that the increase in the college premium
resulting from the rise in Chinese import penetration is between four to seven
times the negative effect on the low-skilled wages, depending on the specification used. The marginal effect of college attendance is significantly positive,
around 6.7 to 7.2 percent in response to an increase of 2.9 percentage points
in Chinese imports in the baseline model, and between 2.7 to 3.5 percent in
the models with year-college controls. The effects are consistently larger for
the within job spell estimates compared to the total effect specification, suggesting downward wage stickiness on the job in manufacturing, since the total
effect specification is performed on the total population capturing both the on
the job and the between jobs effects.
4.5.3 Robustness Checks
First Industry for the Worker
The workers may respond to the increasingly competitive environment created
by higher Chinese imports by switching their industries, which would be an
endogenous change. To control for this, the estimations so far are repeated
where the industry of the worker k is fixed to the first industry they appear
with in the data.
Table 4.4 performs an analogous study on the panel to Table 4.2. The results
are very similar to the original study, where the baseline estimates show a negative wage outcome for the non-skilled workers, and a positive skill premium
for the college educated workers in Panel A, and only a positive skill premium
effect on college educated workers in Panel B. The inclusion of a year trend
in return to investment in education removes the statistical significance of the
effects, while the magnitudes remain comparable to the original study.
Table4.5 performs the instrumental variables approach, and shows even
higher magnitudes on the coefficients of interest compared to Table 4.3, which
uses the annual industry assignment. In the baseline estimates on the total effect, higher Chinese penetration by 2.9 percentage points translate into lower
wages by 1.2 percent for the low-skilled, and 8.7 percent higher premium for
the college educated. The within job spell effect again only shows a positive
return to college education in response to a trade shock of the same magnitude
(about 7.6 percent). Including time trends, the resulting effect in both estima111
Table 4.4. Log Real Earnings and the Return to Skill - First Industry for the Worker
———————-Baseline———————
(1)
(2)
(3)
(4)
Panel A: Total Effect (Equation 4.2)
Chinese Import P.
0.209
(0.152)
Chinese Import P.*College
-0.200*
(0.110)
1.134***
(0.332)
Firm Size : 10-49
250-499
500+
log Capital Intensity
log Sales
0.829
0.829
0.934***
(0.278)
Panel B: Within Job Spell (Equation 4.3)
Chinese Import P.
0.276
-0.0990
(0.230)
(0.104)
Chinese Import P.*College
1.004***
(0.339)
Firm Size : 10-49
50-249
250-499
500+
log Capital Intensity
log Sales
R-squared
MarginalEff
StdErr
-0.204*
(0.112)
1.135***
(0.334)
0.0282***
(0.00311)
0.0473***
(0.00740)
0.0537***
(0.00969)
0.0535***
(0.0123)
0.000442
(0.00109)
0.00544*
(0.00267)
0.115
(0.0913)
0.829
0.829
0.931***
(0.283)
0.831
0.831
0.331
(0.209)
0.831
0.831
0.329
(0.218)
0.285
(0.222)
-0.0820
(0.102)
0.990***
(0.333)
0.0237***
(0.00340)
0.0361***
(0.00652)
0.0434***
(0.00963)
0.0505***
(0.0114)
-0.00153
(0.00223)
0.0224***
(0.00612)
0.185
(0.160)
0.0730
(0.0724)
0.319
(0.249)
0.200
(0.158)
0.0908
(0.0722)
0.312
(0.245)
0.0246***
(0.00343)
0.0376***
(0.00665)
0.0451***
(0.00965)
0.0517***
(0.0110)
-0.000826
(0.00200)
0.0218***
(0.00606)
0.859
0.908**
(0.344)
0.859
0.0281***
(0.00311)
0.0475***
(0.00744)
0.0540***
(0.00972)
0.0533***
(0.0124)
-1.36e-05
(0.00129)
0.00552*
(0.00267)
50-249
R-squared
MarginalEff
StdErr
0.204
(0.157)
0.858
0.858
0.905**
(0.347)
—————–Education Trend—————(5)
(6)
(7)
(8)
0.0234***
(0.00341)
0.0358***
(0.00656)
0.0434***
(0.00970)
0.0501***
(0.0115)
-0.00215
(0.00267)
0.0227***
(0.00618)
0.858
0.00330
(0.0559)
0.327
(0.215)
0.113
(0.0968)
0.0290***
(0.00310)
0.0487***
(0.00721)
0.0556***
(0.00976)
0.0553***
(0.0119)
0.00112
(0.00114)
0.00527**
(0.00252)
0.0245***
(0.00343)
0.0376***
(0.00666)
0.0452***
(0.00966)
0.0517***
(0.0110)
-0.000989
(0.00215)
0.0219***
(0.00607)
0.859
0.392
(0.299)
0.860
0.00172
(0.0560)
0.328
(0.218)
0.0290***
(0.00310)
0.0486***
(0.00722)
0.0554***
(0.00975)
0.0552***
(0.0119)
0.00121
(0.00111)
0.00527**
(0.00253)
0.860
0.403
(0.297)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany,
detailed in Equation 4.1. All columns have year and person fixed effects for 850,158 people (5,868,061 observations) in Panel A and year
and person-firm match fixed effects for a total of 1,017,056 matches (5,689,217 observations) in Panel B, and include Age, Age2 and Age3
as covariates.Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of
SEK120,000 working for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital
intensity. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Firm Size categories follow
Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+)
with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size equal or larger
than 5, but less than 10 employees.
112
Table 4.5. Log Real Earnings and the Return to Skill, IV Results using First Industry
———————-Baseline———————
(1)
(2)
(3)
(4)
Panel A: Total Effect (Equation 4.2)
CMPSwe
0.565
-0.398*
(0.372)
(0.215)
CMPSwe *College
3.043***
(0.398)
Firm Size : 10-49
50-249
250-499
500+
log Capital Intensity
log Sales
R-squared
MarginalEff
StdErr
0.047
0.050
2.645***
(0.316)
Panel B: Within Job Spell (Equation 4.3)
CMPSwe
0.777
-0.165
(0.521)
(0.226)
2.770***
CMPSwe *College
(0.320)
Firm Size : 10-49
50-249
250-499
500+
log Capital Intensity
log Sales
R-squared
MarginalEff
StdErr
0.040
0.042
2.605***
(0.336)
0.553
(0.381)
—————–Education Trend—————(5)
(6)
(7)
(8)
-0.407*
(0.219)
3.044***
(0.394)
0.0284***
(0.00281)
0.0476***
(0.00667)
0.0543***
(0.00874)
0.0540***
(0.0111)
0.000524
(0.000975)
0.00536**
(0.00240)
0.309
(0.215)
0.049
0.052
2.637***
(0.305)
0.058
0.058
1.029***
(0.376)
0.060
0.060
1.025***
(0.394)
0.807
(0.501)
-0.120
(0.222)
2.741***
(0.326)
0.0238***
(0.00288)
0.0365***
(0.00550)
0.0442***
(0.00823)
0.0507***
(0.00977)
-0.00135
(0.00190)
0.0221***
(0.00521)
0.521
(0.354)
0.203
(0.169)
1.033**
(0.404)
0.565
(0.345)
0.251
(0.166)
1.022***
(0.396)
0.0245***
(0.00291)
0.0376***
(0.00565)
0.0454***
(0.00830)
0.0516***
(0.00950)
-0.000658
(0.00165)
0.0218***
(0.00520)
0.044
2.621***
(0.334)
0.048
0.0282***
(0.00278)
0.0477***
(0.00666)
0.0544***
(0.00868)
0.0537***
(0.0111)
5.26e-05
(0.00113)
0.00540**
(0.00238)
0.0232***
(0.00290)
0.0357***
(0.00553)
0.0434***
(0.00834)
0.0500***
(0.00996)
-0.00191
(0.00218)
0.0230***
(0.00531)
0.042
0.0195
(0.127)
1.009***
(0.349)
0.305
(0.227)
0.0290***
(0.00278)
0.0488***
(0.00648)
0.0558***
(0.00877)
0.0555***
(0.0107)
0.00116
(0.00101)
0.00521**
(0.00227)
0.0244***
(0.00292)
0.0375***
(0.00567)
0.0452***
(0.00833)
0.0516***
(0.00951)
-0.000817
(0.00176)
0.0220***
(0.00523)
0.048
1.236**
(0.534)
0.050
0.0160
(0.128)
1.009***
(0.358)
0.0290***
(0.00279)
0.0487***
(0.00650)
0.0556***
(0.00879)
0.0554***
(0.0107)
0.00125
(0.000978)
0.00522**
(0.00229)
0.050
1.273**
(0.524)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany,
detailed in Equation 4.1. All columns have year and person fixed effects for 850,158 people (5,868,061 observations) in Panel A and year
and person-firm match fixed effects for a total of 1,017,056 matches (5,689,217 observations) in Panel B, and include Age, Age2 and Age3
as covariates.Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of
SEK120,000 working for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital
intensity. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Firm Size categories follow
Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+)
with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size equal or larger
than 5, but less than 10 employees.
113
tion methods ranges from a 2.9 to 3.6 percent increase in wages for the skilled
workers in response to the average increase in Chinese imports.
Dividing the Sample
A potential cause of concern over the estimates could be the presence of an
unobserved and simultaneous shock that could be the real motivating reason
behind the observed results that have been attributed to changes in imports
from China. The IV strategy aims to resolve the issue of simultaneity at the
domestic level, like as a shock that could influence domestic wages and Chinese imports at the same time. This section aims to target the presence of a
simultaneous unobserved shock to the world economy that could be a driving
force behind the results. Given the years of focus in the study at hand, the particular change in the progress of informatics and communication technologies
would be a likely candidate that could affect both Chinese and Swedish firms
at the same time, as well as the world economy in general. The share of Chinese imports in the industries producing, or heavily using these technologies
(Office Machineries; Electrical Machineries; and Radio, TV, and Communications Equipment Industries) in Swedish apparent consumption grew between
15 to almost 25 times between 1996 and 2007 (Table 4.1). To address this
issue, the regressions are performed on a sample excluding these industries.
Results are reported in Table 4.6 and show that the positive premium of the
college educated workers is also present in the industries isolated from a direct
influence from informatics related technological shocks. Regressions including year-education controls, presented in columns 5 to 8, show that the positive effect is no longer statistically significant. In comparison to the results
from the full sample, this suggests that college educated workers gain more
in the informatics and communication technologies sectors than in other sectors. Relatively speaking, the college educated population is better rewarded
in industries with a high degree of informatics and technology. The exercise
is repeated using the first industry assignment of the worker, and, as shown in
Table 4.7, exhibit very similar patterns.
Different Chinese Import Penetration Measures
I consider two alternative measures that have been used in the literature. The
measure in Equation 4.6 looks at the pure share of Chinese imports over total
imports in each industry-year, which will gauge the change in the importance
of China as a trade partner. Due to the simultaneity issue discussed above, this
measure is also computed for as an average of the following formulation for
Denmark, Finland, and Germany to proxy the Swedish experience with high
imports from developing countries.
CMPkt =
114
China
Mkt
,
M Total
jt
(4.6)
Table 4.6. Log Real Earnings in the Non-ICT Industries - Yearly Industry for the
Worker
———————Baseline——————–
(1)
(2)
(3)
(4)
Panel A: Total Effect (Equation 4.2)
Chinese Import P.
-0.0932
(0.140)
Chinese Import P.*College
-0.330
(0.231)
1.619*
(0.912)
Firm Size : 10-49
250-499
500+
log Capital Intensity
log Sales
0.825
0.825
1.288*
(0.714)
Panel B: Within Job Spell (Equation 4.3)
Chinese Import P.
-0.0359
-0.241
(0.145)
(0.216)
Chinese Import P.*College
1.461
(0.887)
Firm Size : 10-49
50-249
250-499
500+
log Capital Intensity
log Sales
R-squared
MarginalEff
StdErr
-0.306
(0.229)
1.630*
(0.920)
0.0285***
(0.00331)
0.0474***
(0.00798)
0.0543***
(0.0111)
0.0543***
(0.0141)
0.00150
(0.00134)
0.00509
(0.00347)
-0.0423
(0.0956)
0.825
0.825
1.324*
(0.723)
0.826
0.826
0.00408
(0.164)
0.827
0.827
0.0345
(0.167)
0.00577
(0.128)
-0.196
(0.201)
1.437
(0.877)
0.0238***
(0.00373)
0.0369***
(0.00761)
0.0447***
(0.0114)
0.0500***
(0.0138)
0.00106
(0.00161)
0.0230***
(0.00793)
0.0193
(0.106)
0.0375
(0.0938)
-0.125
(0.123)
0.0612
(0.0943)
0.0830
(0.0796)
-0.148
(0.128)
0.0245***
(0.00384)
0.0382***
(0.00792)
0.0458***
(0.0116)
0.0511***
(0.0136)
0.00147
(0.00165)
0.0229**
(0.00804)
0.853
1.241
(0.724)
0.853
0.0284***
(0.00328)
0.0472***
(0.00795)
0.0541***
(0.0110)
0.0540***
(0.0141)
0.00154
(0.00138)
0.00506
(0.00346)
50-249
R-squared
MarginalEff
StdErr
-0.0670
(0.138)
0.852
0.852
1.220
(0.725)
—————–Education Trend—————–
(5)
(6)
(7)
(8)
0.0237***
(0.00372)
0.0367***
(0.00758)
0.0446***
(0.0114)
0.0498***
(0.0139)
0.00113
(0.00160)
0.0232***
(0.00795)
0.853
-0.0506
(0.0922)
0.0547
(0.130)
-0.0147
(0.0941)
0.0288***
(0.00333)
0.0476***
(0.00806)
0.0541***
(0.0115)
0.0543***
(0.0142)
0.00182
(0.00131)
0.00543
(0.00339)
0.0245***
(0.00384)
0.0382***
(0.00792)
0.0458***
(0.0116)
0.0511***
(0.0136)
0.00146
(0.00165)
0.0229**
(0.00804)
0.853
-0.0872
(0.197)
0.854
-0.0235
(0.0902)
0.0581
(0.135)
0.0288***
(0.00334)
0.0476***
(0.00805)
0.0541***
(0.0115)
0.0543***
(0.0142)
0.00181
(0.00131)
0.00543
(0.00339)
0.854
-0.0653
(0.194)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany,
detailed in Equation 4.1. All columns have year and person fixed effects for 779,067 people (5,306,175 observations) in Panel A and year and
person-firm match fixed effects for a total of 904,405 matches (5,153,077 observations) in Panel B, and both panels include Age, Age2 and
Age3 as covariates. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Sample includes
only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least
two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories in this
sample follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large
firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size
equal or larger than 5, but less than 10 employees.
115
Table 4.7. Log Real Earnings in the Non-ICT Industries - First Industry for the Worker
———————Baseline——————–
(1)
(2)
(3)
(4)
Panel A: Total Effect (Equation 4.2)
Chinese Import P.
-0.0346
(0.0888)
Chinese Import P.*College
-0.230
(0.184)
1.838
(1.190)
Firm Size : 10-49
0.0279***
(0.00338)
0.0465***
(0.00840)
0.0531***
(0.0112)
0.0526***
(0.0139)
0.000930
(0.00117)
0.00558
(0.00324)
50-249
250-499
500+
log Capital Intensity
log Sales
R-squared
MarginalEff
StdErr
0.823
Panel B: Within Job Spell (Equation 4.3)
Chinese Import P.
-0.0278
(0.102)
Chinese Import P.*College
0.0309
(0.0633)
0.0255
(0.0525)
0.0481
(0.142)
0.0180
(0.0644)
0.0287***
(0.00340)
0.0477***
(0.00820)
0.0544***
(0.0112)
0.0547***
(0.0135)
0.00152
(0.00122)
0.00544*
(0.00308)
0.0135
(0.0554)
0.0400
(0.142)
0.0287***
(0.00340)
0.0477***
(0.00820)
0.0544***
(0.0112)
0.0547***
(0.0135)
0.00151
(0.00122)
0.00544*
(0.00308)
0.824
0.824
1.593
(1.020)
0.825
0.825
0.0737
(0.179)
0.825
0.825
0.0535
(0.175)
-0.204
(0.181)
1.461
(0.985)
0.00610
(0.0999)
-0.168
(0.177)
1.442
(0.976)
0.0239***
(0.00355)
0.0370***
(0.00716)
0.0450***
(0.0107)
0.0502***
(0.0128)
0.000815
(0.00156)
0.0228***
(0.00723)
0.0222
(0.0786)
0.0350
(0.0678)
-0.101
(0.133)
0.0559
(0.0788)
0.0704
(0.0684)
-0.115
(0.136)
0.0247***
(0.00366)
0.0383***
(0.00740)
0.0463***
(0.0108)
0.0517***
(0.0125)
0.00125
(0.00160)
0.0225***
(0.00724)
0.853
1.275
(0.835)
0.853
0.0238***
(0.00354)
0.0368***
(0.00712)
0.0448***
(0.0107)
0.0500***
(0.0129)
0.000884
(0.00155)
0.0230***
(0.00725)
50-249
250-499
500+
log Capital Intensity
log Sales
0.852
-0.243
(0.191)
1.836
(1.186)
0.0279***
(0.00336)
0.0465***
(0.00836)
0.0530***
(0.0111)
0.0527***
(0.0138)
0.000877
(0.00118)
0.00558
(0.00322)
0.823
1.608
(1.031)
Firm Size : 10-49
R-squared
MarginalEff
StdErr
-0.0483
(0.0924)
—————–Education Trend—————(5)
(6)
(7)
(8)
0.852
1.258
(0.841)
0.853
0.0247***
(0.00366)
0.0383***
(0.00740)
0.0463***
(0.0108)
0.0517***
(0.0125)
0.00124
(0.00160)
0.0225***
(0.00724)
0.853
-0.0665
(0.181)
0.854
0.854
-0.0446
(0.182)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany,
detailed in Equation 4.1. All columns have year and person fixed effects for 758,851 people (5,307,055 observations) in Panel A and year and
person-firm match fixed effects for a total of 903,759 matches (5,157,008 observations) in Panel B, and both panels include Age, Age2 and
Age3 as covariates. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Sample includes
only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least
two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories follow
Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+)
with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size equal or larger
than 5, but less than 10 employees.
116
The final measure is Chinese exports (CX) to the rest of the world (RoW),
excluding Sweden, weighted by World GDP. This measure will see how important Chinese goods have become relative to World production over the
years.
CXktWorld −CXktSwe
CMPktRoW = log(
),
(4.7)
GDPWorld − GDPSwe
The results presented in Tables 4.15 to 4.22 confirm the presence of a positive return to skill for workers with college education in response to higher
imports from China. However, neither one of these measures take the size
of the (proxy countries’) domestic production into account in their calculation, and therefore they are not considered suitable measures of Chinese import penetration for this study. Theoretically, this paper considered the change
in Chinese imports affecting wages of different types of workers through the
mechanism of changing demand for these workers, and it would be simplistic
to assume that the demand for workers in an industry are satisfactorily measured by changes in trade shares. For given levels of total imports and domestic exports, if the size of the domestic production is small, a change in Chinese
imports will have a larger effect on the change in demand for workers in that
industry compared to an industry with a large domestic output level, whereas
specifications 4.6 and 4.7 would treat these two industries the same way. A
correct definition of the import penetration measure is crucial for capturing
the effect of interest. For this reason, the main specification uses a measure of
Chinese imports relative to domestic apparent consumption.
4.6 Extensions
4.6.1 Within Firm Variation
The previous sections established that rising Chinese imports translated into
higher wages for high-skilled workers in both the overall and within job spell
specifications, and also showed some support for downward wage rigidity in
the within job spell estimates. This within job spell estimate takes the specific match between the worker and the firm into account and computes its
contribution to the individual’s wage. Another method building on Abowd et
al.(1999) introduces unique time-invariant effects for the person and the firm.
This allows for the separation of the person component, and the firm component to see whether this would produce a different relationship in the return to
college education in response to a trade shock from a low-wage country.
In this method, the person component can be interpreted as a skill factor
that indicates the worker’s outside option or return to skill in the labor market.
Relative to the within job spell estimate, the firm component is a weaker form
of workplace complementarity factor: it is a component of your wage that is
rewarded equally across all the workers in that given firm. In other words, this
117
effect captures the firm premium that the individual gets from working in that
particular firm. By taking into account the overall return to the person’s timeinvariant qualities from potentially a longer time period, this method could
improve the precision on the point estimates. Compared to the within job
spell specification, one sure gain of using this method is in the number of
observations. A person needs to be employed for two years in the same firm
for the match effect to be identified under the within job spell specification.
In the within firm specification, this restriction is lifted; instead, the person
needs to be in the sample for two years, but not necessarily employed in the
same firm. This allows for an expansion of the number of observations in this
section.
The methodology relies on workers who switch between firms in order to
identify the firm fixed effect in the estimation equation (4.8). Switchers generate a large network of firms that are connected to each other through at least
one other firm in the group through at least one worker who moves between
them. In the analysis, the largest such network, called the Mobility Group, is
determined by maximizing the number of firms that are connected. The analysis that follows is strictly restricted to this group of interconnected firms (and
therefore their employees). This group includes 99.9% of all the workers and
98.9% of all firms in the manufacturing industries.22
Composite
Composite
logwi jt = θ1CMPkt
+ θ2CMPkt
XCollegeit
+ αi + xit βt + θJ(i;t) + z jt γ + τt + εi jt ,
(4.8)
where θJ(i;t) is a firm fixed effect that captures the premium that all the workers in that firm receive regardless of personal characteristics. Here, the coefficient θ2 will capture how the within firm return to skill changes in response
to increased Chinese imports into the sector. The year-college interaction is
included among the time varying controls to take any time trend in the return
to college education into consideration in all the specifications in this section.
OLS Results
Table 4.8 Columns 1 and 2 show the return to skill in a setting where both
person and firm fixed effects are controlled for. When firm fixed effects are
introduced to capture any firm wage premia that may be present in wages, the
remaining response in the coefficients of interest will be within firm changes
in the wages for skilled and unskilled workers. The table presents the point
estimates and standard errors of a bootstrap sampling method of 100 replications, where firms could be drawn with replacement from the population of
all firms. This section also controls for other observables on the firm level.
22 Total effect and within job spell variation estimates were also performed on the individuals
composing the mobility group to allow for comparison across estimates. Details on the Mobility
Group are presented in Table 4.10.
118
The results on the Chinese import variables are rather similar to those using
the within job specification using Equation 4.3 reported in Table 4.2, Columns
7 and 8. A higher level of competition from low-wage countries translates
into a widening of the wage distribution within the firm. Low-skilled workers are negatively affected. This could be because low-skilled workers arrive
into lower paying jobs, or, in a less likely scenario given the previous section,
on the job downward wage adjustment. Skilled employees enjoy a positive
premium in response to the higher competition from China, which suggests a
change in the firm’s production activities that shift worker demand away from
low-skilled to high-skilled employees.
Table 4.8. Log Real Earnings and the Return to Skill Within the Firm: Baseline, and
2SLS Results for Equation 4.8 with Year-College control, Yearly industry
VARIABLES
Chinese Import P.
(1)
Base
(2)
Base
(5)
2SLS
(6)
2SLS
0.1827
(0.1188)
0.0638
(0.0578)
0.3276
(0.1667)
0.0229
(0.0024)
0.0345
(0.0037)
0.0411
(0.0053)
0.0472
(0.0070)
-0.0007
(0.0015)
0.0190
(0.0044)
0.4941
(0.3270)
0.1977
(0.1807)
0.9434
(0.5211)
0.0229
(0.0023)
0.0346
(0.0035)
0.0413
(0.0054)
0.0468
(0.0065)
-0.0006
(0.0015)
0.0189
(0.0040)
5,866,337
0.8336
10,349
850,048
0.3915
(0.2127)
5,866,337
0.8335
10,349
850,048
Chinese Import P.*College
Firm Size : 10-49
50-249
250-499
500+
log Capital Intensity
log Sales
Observations
R-squared
NoFirms
Individuals
MarginalEff
StdErr
0.0229
(0.0024)
0.0344
(0.0037)
0.0410
(0.0053)
0.0470
(0.0070)
-0.0009
(0.0016)
0.0191
(0.0044)
5,866,337
0.8335
10,349
850,048
0.0227
(0.0024)
0.0343
(0.0037)
0.0410
(0.0053)
0.0464
(0.0071)
-0.0006
(0.0015)
0.0193
(0.0044)
5,866,337
0.8336
10,349
850,048
1.1410
(0.6826)
Note: Chinese Import Share defined as Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed
in Equation 4.1. All columns have year, person, and firm fixed effects and control for year-college interactions as well as Age, Age2 and
Age3 . Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000
working for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Firm
Size categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and
large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of
size equal or larger than 5, but less than 10 employees. Columns 1 and 2 have standard errors obtained from standard deviations of the relevant
coefficients after 100 repetitions of the regression following a bootstrap sampling, and are reported in parenthesis. Columns 3 and 4 report
bootstrap standard errors obtained from 100 replications with replacement based on 10349 possible firm clusters in parenthesis.
IV results
This approach will allow for the quantification of the within-firm wage adjustment of high-skilled employees. Results presented in Table 4.8 Columns 3
119
and 4 resembles that of the within job spell IV estimates reported in Panel B
of Table 4.3 Columns 7 and 8: in response to an average 2.9 percentage points
increase in Chinese imports, low-skilled workers receive 0.6 percent lower
wages, and skilled workers receive a premium of 2.7 percent higher wages as
a positive return to their educational investment.
4.7 Conclusion
This paper has analyzed the effect of increased competition from low-wage
countries on manufacturing wages in a developed economy using a detailed
matched employer-employee database from Sweden, which covers the total
population of manufacturing workers in privately owned (LLP and LLC) firms
between 1996 and 2007. The increase in Chinese imports was taken as an
example of higher competition from all low-wage countries in this period,
and the Chinese membership in the WTO in 2001 is seen as the source of an
exogenous shock. Thus, the paper focused on the changes in returns for lowskilled and high-skilled workers in response to higher imports from China.
Estimations show that the skill wage gap has increased due to a significant increase in the premium for college educated workers. This effect remains even after taking a time trend in the return to college education into
account. The effect on high-skilled workers in response to an average of 2.9
percentage point increase in Chinese import penetration is 2.7 percent higher
wages. Since real skilled wages have risen about 27 percent in this period,
this suggests that rising competition from low-wage countries has contributed
to around 10 percent of the increase in skilled wages. Skilled workers appear
to be complementary to the goods imported from low-wage countries, and
therefore command a higher wage with rising imports from China. Previous
literature suggests that firms may upgrade their production technologies in response to rising competition from China, which should increase the demand
for and wages of these workers. Wages for non-college workers are not significantly affected by changes in Chinese imports. One factor contributing to
the lack of a significant negative adjustment in wages could be the institutional
setting in Sweden, where contracts negotiated with labor unions impose lower
limits on wages.
The results of the paper differ from previous work using U.S. and Chilean
data by utilizing a detailed matched employer-employee database and by finding a robust and significant positive effect on the return to college education
even after controlling for a time trend in return to education.
Ashournia et al. (2014) use a similar database from Denmark to find lower
wages for low-skilled workers, using a firm-level Chinese import penetration
measure, and a positive effect on the return to college education using an
industry-level Chinese penetration measure, both using the within the job-spell
specification. Their specification does not utilize year-education interactions
120
so as to take other factors that may create a trend in the return to education
into account. I add to this study by considering not only a within job-spell,
but also a total effect. I find a significant positive return to college education
across specifications. In some estimations including job switchers, I find a
negative effect of higher Chinese imports on the wages of low-skilled workers. However, this result is not robust to the inclusion of a year trend in return
to college education. In an extension, Ashournia et al. (2014) look at the
long-term effects of change in Chinese imports to cumulative earnings for the
sub-sample of workers who work for all years in the data, and do not find a
significant wage premium. In comparison, the total effect specification in this
paper identifies that the wage effects of the increase in Chinese import penetration constitutes about 10 percent of the increase in skilled wages observed
in the total period, without restricting the data to a sub-sample.
Local labor markets could be an additional factor that could influence hiring/firing and wage decisions at the firm through effects on labor market tightness, especially with respect to the skill type of the worker. As it stands now,
this paper abstracts from this issue, since the municipality information in the
matched data is on the headquarters of the firm, which does not necessarily
reflect where the actual manufacturing unit is located geographically.
The measure of Chinese import penetration for this paper was detailed at
the industry level. Ashournia et al. (2014) use firm-level import penetration data on products, which may capture the competitive environment of the
firms better. With product-level imports information, it would be possible to
identify whether the firm is in direct competition with the Chinese good (if
imported goods are final goods), or is benefiting from complementarities that
exist between Chinese imports and production in the Swedish firms. Likewise,
additional data on the occupation or task assignment of the worker would help
identify which type of jobs in particular are hurt due to higher import competition, or benefit from complementarities from rising imports at the individual
level. Finally, in future studies it could be worthwhile to allow the response
of the college premium to a change in Chinese imports to differ depending on
the particular sets of skills by interacting with education fields for the college
educated manufacturing workers.
121
References
Abowd J.M., Kramarz F., and Margolis D.N., (1999). High Wage Workers and
High Wage Firms, Econometrica, Econometric Society, 67(2), 251-334.
Acemoglu D. and Autor D., (2011). Skills, Tasks, and Technologies: Implications for Employment and Earnings, in Handbook of Labor Economics ed. O.
Ashenfelter and D. Card, IV(B), 1043-1171.
Akerlof G.A., Dickens W.T., and Perry G.L., (1996). The Macroeconomics
of Low Inflation, Brookings Papers on Economic Activity, Microeconomics,
1996(1), 1-52.
Alvarez R. and Opazo L., (2011). Effects of Chinese Imports on Relative
Wages: Microevidence from Chile, Scandinavian Journal of Economics, 113(2),
342-363.
Amiti M. and Davis D.R., (2012). Trade, Firms, and Wages: Theory and
Evidence, Review of Economics Studies, 79(1), 1-36.
Ashournia D., Munch J., and Nguyen D., (2014). The Impact of Chinese Import Penetration on Danish Firms and Workers, IZA Discussion Paper Series,
8166.
Autor D.H., Levy F., and Murnane R.J., (2003). The Skill Content of Recent
Technological Change: An Empirical Exploration, The Quarterly Journal of
Economics, 118(4), 1279-1333.
Autor D.H., Dorn D., and Hanson G.H., (2013). The China Syndrome: Local
Labor Market Effects of Import Competition in the United States, American
Economic Review, 103(6), 2121-68.
Bernard A.B., Jensen J.B., Schott P.K., (2006). Survival of the Best Fit: Exposure to Low-Wage Countries and the (Uneven) Growth of U.S. Manufacturing
Plants, Journal of International Economics, 68(1), 219-237.
Bloom N., Draca M., and Van Reenen J., (2015). Trade Induced Technical
Change? The Impact of Chinese Imports on Innovation, IT and Productivity,
Review of Economic Studies, forthcoming.
Bustos P., (2011). Trade Liberalization, Exports, and Technology Upgrading: Evidence on the Impact of MERCOSUR on Argentinian Firms, American
Economic Review, 101(1), 304-40.
Campbell C.M. and Kamlani K.S., (1997). The Reasons for Wage Rigidity:
Evidence from a Survey of Firms, Quarterly Journal of Economics, 112(3),
759-789.
122
Dickens W.T., Goette L., Groshen E.L., Holden S., Messina J., Schweitzer
M.E., Turunen J., and Ward M.E., (2007). How Wages Change: Micro Evidence from the International Wage Flexibility Project, Journal of Economic
Perspectives, 21(2), 195-214.
Feenstra H. and Hanson G., (1999). The Impact of Outsourcing and HighTechnology Capital on Wages: Estimates for the United States, 1979-1990,
Quarterly Journal of Economics, 114(3), 907-940.
Freeman R.B., (2009). Globalization and Inequality in The Oxford Handbook
of Economic Inequality, ed. W. Salverda, B. Nolan, and T.M. Smeeding, Oxford University Press, 575-598.
Goldin C. and Katz L.F., (2008). The Race Between Education and Technology. Cambridge: Harvard University Press.
Griliches Z., (1969). Capital-Skill Complementarity, Review of Economics
and Statistics, 51(4), 465-468.
Grossman G. and Rossi-Hansberg E., (2008). Trading Tasks: A Simple Theory of Offshoring, American Economic Review, 98(5), 1978-1997.
Helpman E., Itskhoki O., and Redding S., (2010). Inequality and Unemployment In a Global Economy Econometrica, 78(4), 1239-1283.
Heyman F., Sjöholm, F., and Gustavsson Tingvall, P., (2007). Is There Really
a Foreign Ownership Wage Premium? Evidence from Matched EmployerEmployee Data, Journal of International Economics, 73(2), 355-376.
Jacobson L.S., LaLonde R.J., and Sullivan D.G., (1993). Earning Losses of
Displaced Workers, American Economic Review, 83(4), 685-709.
Katz L.F. and Autor D., (1999). Changes in the Wage Structure and Earnings
Inequality, in Handbook of Labor Economics ed. O. Ashenfelter and D. Card,
III(A),1463-1555.
Nordström Skans O., Edin P-A. and Holmlund B., (2009). Wage Dispersion
Between and Within Plants: Sweden 1985-2000, NBER Chapters, in: The
Structure of Wages: An International Comparison, 217-260.
Shapiro C. and Stiglitz J.E., (1984). Equilibrium Unemployment as a Worker
Discipline Device American Economic Review, 74(3),433-444.
123
Appendix
Table 4.9. Trading Firm Characteristics
Exporter
Non-Exporter
Importer
Non-Importer
2000
Share of Females
Share of College
Average Age
Average Size
Capital Intensity
0.23
0.15
41
111
312,968
0.23
0.13
41
29
336,700
0.24
0.16
41
120
320,740
0.17
0.1
41
22
286,851
2004
Share of Females
Share of College
Average Age
Average Size
Capital Intensity
0.22
0.17
43
107
346,180
0.22
0.15
42
29
302,486
0.23
0.18
43
117
352,923
0.17
0.11
43
22
292,623
2007
Share of Females
Share of College
Average Age
Average Size
Capital Intensity
0.22
0.19
43
111
353,236
0.24
0.16
42
26
294,717
0.24
0.20
43
121
357,226
0.18
0.12
44
24
305,619
Note: Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in
2010 values) income of SEK120,000 working for at least two years in manufacturing firms with more than
5 employees and values above zero for sales and capital intensity. Foreign Trade information is available
from 2000 onwards at the firm level.
124
Table 4.10. Firm Characteristics, 1996-2007
Data
Mobility Group
Number of Firms
10,463
10,349
Number of People
851,264
850,048
12.495
(0.173)
12.495
(0.173)
Average Firm Real Wage
Average Share of Females
0.22
0.22
Average Share of College
0.14
0.14
Note: Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in
2010 values) income of SEK120,000 working for at least two years in manufacturing firms with more than
5 employees and values above zero for sales and capital intensity. Mobility Group restricts the sample to
workers who work in firms interconnected by a network of movers. Wage standard deviations are reported
in parenthesis.
125
126
5,866,337
0.514
850,048
0.373***
(0.0587)
5,866,337
0.516
850,048
0.390***
(0.0677)
-0.0461
(0.0508)
5,866,337
0.640
850,048
-0.0120***
(0.00340)
0.385***
(0.0519)
—————-Base—————1(a)
1(b)
1(c)
5,866,337
0.517
850,048
-0.000245
(0.000251)
-0.00155**
(0.000667)
-0.00281**
(0.00113)
-0.00271*
(0.00141)
-0.000535*
(0.000259)
0.000550
(0.000356)
0.368***
(0.0594)
5,866,337
0.519
850,048
0.386***
(0.0682)
-0.0483
(0.0509)
-0.000267
(0.000253)
-0.00159**
(0.000664)
-0.00288**
(0.00112)
-0.00282*
(0.00142)
-0.000562**
(0.000259)
0.000577
(0.000357)
5,866,337
0.640
850,048
-0.0125***
(0.00376)
0.385***
(0.0521)
-0.000153
(0.000102)
-0.000485*
(0.000235)
-0.000798**
(0.000374)
-0.000652*
(0.000350)
-6.02e-05
(8.87e-05)
0.000135
(9.66e-05)
———–Base with controls———–
2(a)
2(b)
2(c)
5,866,337
0.514
850,048
0.374***
(0.0593)
5,866,337
0.516
850,048
0.390***
(0.0670)
-0.0463
(0.0491)
5,866,337
0.663
850,048
-0.00342***
(0.000770)
0.352***
(0.0546)
———-Education Trend———3(a)
3(b)
3(c)
5,866,337
0.518
850,048
-0.000249
(0.000252)
-0.00155**
(0.000666)
-0.00282**
(0.00113)
-0.00272*
(0.00141)
-0.000540**
(0.000259)
0.000552
(0.000355)
0.369***
(0.0600)
5,866,337
0.520
850,048
0.386***
(0.0675)
-0.0487
(0.0491)
-0.000265
(0.000254)
-0.00158**
(0.000662)
-0.00287**
(0.00112)
-0.00282*
(0.00143)
-0.000564**
(0.000259)
0.000577
(0.000356)
5,866,337
0.664
850,048
-0.00399***
(0.00117)
0.351***
(0.0548)
-0.000132
(9.49e-05)
-0.000456*
(0.000222)
-0.000778**
(0.000362)
-0.000616*
(0.000339)
-6.69e-05
(8.70e-05)
0.000143
(9.49e-05)
——Education Trend with controls——
4(a)
4(b)
4(c)
*** p<0.01, ** p<0.05, * p<0.10.
Note: The table reports the results of the second stage regressions. All columns have year and person fixed effects and include Age, Age2 and Age3 as covariates. Standard errors are reported in parenthesis, and are adjusted for clustering at
the industry level for the first industry of the worker. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years in manufacturing
firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249),
and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size equal or larger than 5, but less than 10 employees. Columns (1) and (2) have the total
population, (3) and (4) have workers who are employed at the same firm during their entire tenure in the database.
Observations
R-squared
Individuals
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
CMPSwe *College
Total Effect (Equation 4.2)
CMPSwe
VARIABLES
Table 4.11. Log Real Wages and the Return to Skill - First Stage and Reduced Form Regressions for Total Effect Estimates
127
5,687,870
0.458
1,016,790
5,687,870
0.460
1,016,790
5,687,870
0.592
1,016,790
-0.0127***
(0.00340)
0.370***
(0.0745)
5,687,870
0.461
1,016,790
0.000386
(0.000264)
0.000273
(0.000558)
8.78e-05
(0.000724)
0.00140
(0.000895)
-0.000676*
(0.000387)
-0.000628
(0.000467)
0.348***
(0.0773)
5,687,870
0.463
1,016,790
0.362***
(0.0856)
-0.0374
(0.0622)
0.000373
(0.000257)
0.000260
(0.000547)
7.19e-05
(0.000711)
0.00137
(0.000913)
-0.000702*
(0.000385)
-0.000609
(0.000457)
5,687,870
0.592
1,016,790
-0.0127***
(0.00352)
0.370***
(0.0746)
-6.22e-05
(8.22e-05)
-0.000180
(0.000166)
-0.000282
(0.000224)
0.000112
(0.000145)
-5.00e-05
(0.000111)
0.000143
(0.000139)
———–Base with controls———–
2(a)
2(b)
2(c)
5,687,870
0.459
1,016,790
0.351***
(0.0777)
5,687,870
0.460
1,016,790
0.364***
(0.0846)
-0.0387
(0.0608)
5,687,870
0.619
1,016,790
-0.00341***
(0.000734)
0.334***
(0.0774)
———-Education Trend———3(a)
3(b)
3(c)
5,687,870
0.462
1,016,790
0.000382
(0.000257)
0.000268
(0.000547)
8.55e-05
(0.000714)
0.00139
(0.000901)
-0.000678*
(0.000388)
-0.000622
(0.000462)
0.348***
(0.0780)
5,687,870
0.463
1,016,790
0.362***
(0.0848)
-0.0398
(0.0597)
0.000376
(0.000255)
0.000265
(0.000541)
7.84e-05
(0.000707)
0.00137
(0.000920)
-0.000701*
(0.000386)
-0.000609
(0.000456)
5,687,870
0.619
1,016,790
-0.00351***
(0.000920)
0.334***
(0.0775)
-2.52e-05
(7.03e-05)
-0.000111
(0.000144)
-0.000224
(0.000216)
0.000182
(0.000169)
-5.74e-05
(0.000115)
0.000126
(0.000140)
——Education Trend with controls——
4(a)
4(b)
4(c)
*** p<0.01, ** p<0.05, * p<0.10.
Note: The table reports the results of the second stage regressions. All columns have year and firm-person match fixed effects and include Age, Age2 and Age3 as covariates. Standard errors are reported in parenthesis, and are adjusted for
clustering at the industry level for the first industry of the worker. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years in
manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50
to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size equal or larger than 5, but less than 10 employees. Columns (1) and (2) have the
total population, (3) and (4) have workers who are employed at the same firm during their entire tenure in the database.
Observations
R-squared
Matches
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
0.364***
(0.0855)
-0.0363
(0.0633)
—————-Base—————1(a)
1(b)
1(c)
Within Job Spell (Equation 4.3)
CMPSwe
0.350***
(0.0770)
CMPSwe *College
VARIABLES
Table 4.12. Log Real Wages and the Return to Skill - First Stage and Reduced Form Regressions for Within Job Spell Estimates
128
5,868,061
0.547
850,158
0.370***
(0.0955)
5,868,061
0.554
850,158
0.397***
(0.0895)
-0.0747
(0.0580)
5,868,061
0.605
850,158
-0.0137**
(0.00603)
0.363***
(0.104)
—————-Base—————1(a)
1(b)
1(c)
5,868,061
0.548
850,158
-2.42e-05
(0.000219)
-0.000363
(0.000470)
-0.000624
(0.000706)
-0.000689
(0.000755)
-0.000120
(0.000156)
0.000214
(0.000144)
0.369***
(0.0958)
5,868,061
0.554
850,158
0.396***
(0.0897)
-0.0750
(0.0580)
-2.75e-05
(0.000218)
-0.000353
(0.000467)
-0.000605
(0.000697)
-0.000702
(0.000768)
-0.000150
(0.000159)
0.000219
(0.000148)
5,868,061
0.605
850,158
-0.0139**
(0.00615)
0.363***
(0.104)
-7.20e-05
(5.79e-05)
-0.000153
(0.000121)
-0.000283
(0.000220)
-0.000260
(0.000220)
-4.70e-05
(9.17e-05)
5.80e-05
(5.07e-05)
———–Base with controls———–
2(a)
2(b)
2(c)
5,868,061
0.548
850,158
0.371***
(0.0958)
5,868,061
0.554
850,158
0.398***
(0.0881)
-0.0813
(0.0572)
5,868,061
0.632
850,158
-0.00442**
(0.00178)
0.326***
(0.110)
———-Education Trend———3(a)
3(b)
3(c)
5,868,061
0.548
850,158
-2.93e-05
(0.000219)
-0.000369
(0.000476)
-0.000631
(0.000712)
-0.000705
(0.000765)
-0.000128
(0.000155)
0.000216
(0.000147)
0.370***
(0.0962)
5,868,061
0.555
850,158
0.398***
(0.0884)
-0.0815
(0.0571)
-2.14e-05
(0.000215)
-0.000341
(0.000464)
-0.000590
(0.000692)
-0.000679
(0.000759)
-0.000151
(0.000157)
0.000217
(0.000148)
5,868,061
0.632
850,158
-0.00459**
(0.00192)
0.326***
(0.110)
-3.63e-05
(4.59e-05)
-8.90e-05
(9.94e-05)
-0.000203
(0.000186)
-0.000137
(0.000175)
-3.96e-05
(8.45e-05)
4.48e-05
(4.63e-05)
——Education Trend with controls——
4(a)
4(b)
4(c)
*** p<0.01, ** p<0.05, * p<0.10.
Note: The table reports the results of the second stage regressions. All columns have year and person fixed effects and include Age, Age2 and Age3 as covariates. Standard errors are reported in parenthesis, and are adjusted for clustering at
the industry level for the first industry of the worker. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years in manufacturing
firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249),
and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size equal or larger than 5, but less than 10 employees. Columns (1) and (2) have the total
population, (3) and (4) have workers who are employed at the same firm during their entire tenure in the database.
Observations
R-squared
Individuals
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
Chinese Import P.*College
Total Effect (Equation 4.2)
Chinese Import P.
VARIABLES
Table 4.13. Log Real Wages and the Return to Skill Using - First Stage Regressions, First Industry, Total Effect Estimates
129
5,689,217
0.491
1,017,056
5,689,217
0.495
1,017,056
5,689,217
0.579
1,017,056
-0.0133**
(0.00508)
0.359***
(0.105)
5,689,217
0.492
1,017,056
0.000306
(0.000226)
0.000173
(0.000451)
-8.42e-05
(0.000627)
0.000126
(0.000684)
-0.000299
(0.000311)
-0.000294
(0.000394)
0.354***
(0.103)
5,689,217
0.496
1,017,056
0.376***
(0.103)
-0.0597
(0.0588)
0.000294
(0.000221)
0.000159
(0.000444)
-8.88e-05
(0.000617)
0.000101
(0.000665)
-0.000336
(0.000313)
-0.000274
(0.000382)
5,689,217
0.579
1,017,056
-0.0134**
(0.00529)
0.359***
(0.105)
-5.46e-05
(7.48e-05)
-0.000141
(0.000145)
-0.000289
(0.000234)
-7.39e-05
(0.000174)
-7.89e-05
(0.000152)
0.000102
(0.000115)
———–Base with controls———–
2(a)
2(b)
2(c)
5,689,217
0.492
1,017,056
0.356***
(0.103)
5,689,217
0.496
1,017,056
0.379***
(0.102)
-0.0651
(0.0558)
5,689,217
0.607
1,017,056
-0.00379***
(0.00126)
0.322***
(0.109)
———-Education Trend———3(a)
3(b)
3(c)
5,689,217
0.493
1,017,056
0.000302
(0.000218)
0.000167
(0.000442)
-8.74e-05
(0.000619)
0.000122
(0.000663)
-0.000304
(0.000311)
-0.000289
(0.000386)
0.354***
(0.103)
5,689,217
0.497
1,017,056
0.377***
(0.102)
-0.0658
(0.0555)
0.000301
(0.000217)
0.000172
(0.000436)
-7.42e-05
(0.000609)
0.000118
(0.000654)
-0.000338
(0.000313)
-0.000279
(0.000379)
5,689,217
0.607
1,017,056
-0.00400**
(0.00152)
0.322***
(0.110)
-1.21e-05
(6.04e-05)
-6.47e-05
(0.000118)
-0.000210
(0.000202)
1.93e-05
(0.000153)
-8.12e-05
(0.000149)
7.25e-05
(0.000103)
——Education Trend with controls——
4(a)
4(b)
4(c)
*** p<0.01, ** p<0.05, * p<0.10.
Note: The table reports the results of the second stage regressions. All columns have year and firm-person match fixed effects and include Age, Age2 and Age3 as covariates. Standard errors are reported in parenthesis, and are adjusted for
clustering at the industry level for the first industry of the worker. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years in
manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50
to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by size are microfirms of size equal or larger than 5, but less than 10 employees. Columns (1) and (2) have the
total population, (3) and (4) have workers who are employed at the same firm during their entire tenure in the database.
Observations
R-squared
Matches
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
0.377***
(0.103)
-0.0591
(0.0591)
—————-Base—————1(a)
1(b)
1(c)
Within Job Spell (Equation 4.3)
Chinese Import P.
0.355***
(0.102)
Chinese Import P.*College
VARIABLES
Table 4.14. Log Real Wages and the Return to Skill Using - First Stage Regressions, First Industry, Within Job Spell Estimates
130
0.829
0.829
0.00692***
(0.000928)
0.829
0.0281***
(0.00294)
0.0477***
(0.00702)
0.0549***
(0.00987)
0.0543***
(0.0130)
0.000547
(0.000887)
0.00533*
(0.00290)
0.00142
(0.00164)
0.830
0.00695***
(0.000943)
-0.00197*
(0.00112)
0.00893***
(0.00105)
0.0288***
(0.00290)
0.0486***
(0.00693)
0.0558***
(0.00972)
0.0567***
(0.0127)
0.00107
(0.000817)
0.00462
(0.00285)
0.831
0.831
0.00261**
(0.00123)
-0.000285
(0.000840)
0.00289***
(0.000592)
0.831
0.0290***
(0.00291)
0.0489***
(0.00700)
0.0562***
(0.0101)
0.0560***
(0.0127)
0.00149
(0.000895)
0.00513*
(0.00281)
0.000879
(0.00112)
0.831
0.00262**
(0.00124)
-9.59e-05
(0.000850)
0.00272***
(0.000582)
0.0291***
(0.00290)
0.0491***
(0.00698)
0.0563***
(0.0101)
0.0565***
(0.0127)
0.00161*
(0.000875)
0.00495*
(0.00281)
—————–Education Trend—————(6)
(7)
(8)
0.000760
(0.00114)
(5)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed in Equation 4.1. All columns have year and person fixed effects for 850,048 people
(5,866,337 observations) and include Age, Age2 and Age3 as covariates. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Sample includes only manufacturing workers born between
years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories follow
Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by
size are microfirms of size equal or larger than 5, but less than 10 employees.
R-squared
MarginalEff
StdErr
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
-0.00212*
(0.00112)
0.00904***
(0.00107)
———————-Baseline———————
(2)
(3)
(4)
Total Effect (Equation 4.2)
Import Share
0.00135
(0.00169)
Import Share*College
(1)
Table 4.15. Log Real Earnings and the Return to Skill using Import Share of China - Yearly Industry for the Worker, Total Effect
131
0.858
0.859
0.00702***
(0.00100)
0.858
0.0231***
(0.00332)
0.0352***
(0.00641)
0.0430***
(0.00979)
0.0494***
(0.0118)
-0.00181
(0.00275)
0.0227***
(0.00644)
0.00234
(0.00179)
0.859
0.00695***
(0.00104)
-0.000669
(0.00129)
0.00762***
(0.000943)
0.0239***
(0.00329)
0.0367***
(0.00635)
0.0446***
(0.00965)
0.0517***
(0.0115)
-0.00150
(0.00263)
0.0218***
(0.00635)
0.859
0.00171
(0.00135)
0.859
0.00305*
(0.00168)
0.000938
(0.00101)
0.00211*
(0.00102)
0.860
0.0243***
(0.00331)
0.0371***
(0.00654)
0.0450***
(0.00981)
0.0512***
(0.0113)
-0.000652
(0.00214)
0.0218***
(0.00640)
0.00178
(0.00130)
0.860
0.00302*
(0.00167)
0.00106
(0.000976)
0.00196*
(0.00102)
0.0244***
(0.00331)
0.0374***
(0.00652)
0.0453***
(0.00980)
0.0516***
(0.0113)
-0.000600
(0.00210)
0.0217***
(0.00639)
—————–Education Trend—————(5)
(6)
(7)
(8)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed in Equation 4.1. All columns have year and person-firm match fixed effects for a total
of 1,016,790 matches (5,687,870 observations), and include Age, Age2 and Age3 as covariates. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Sample includes only manufacturing
workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity.
Firm Size categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In
this table, reference firms by size are microfirms of size equal or larger than 5, but less than 10 employees.
R-squared
MarginalEff
StdErr
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
-0.000760
(0.00131)
0.00778***
(0.000942)
———————-Baseline———————
(2)
(3)
(4)
Within Job Spell (Equation 4.3)
Import Share
0.00232
(0.00190)
Import Share*College
(1)
Table 4.16. Log Real Earnings and the Return to Skill using Import Share of China - Yearly Industry for the Worker, Within Job Spell Effect
132
0.829
0.829
0.00739***
(0.00131)
0.829
0.0280***
(0.00308)
0.0471***
(0.00737)
0.0536***
(0.00966)
0.0527***
(0.0123)
-7.79e-06
(0.00123)
0.00552*
(0.00267)
0.00165
(0.00145)
0.830
0.00732***
(0.00125)
-0.00211
(0.00127)
0.00944***
(0.00155)
0.0284***
(0.00311)
0.0476***
(0.00742)
0.0540***
(0.00975)
0.0539***
(0.0123)
0.000414
(0.00114)
0.00543*
(0.00268)
0.831
0.831
0.00258**
(0.00123)
0.000238
(0.000875)
0.00234**
(0.000959)
0.831
0.0289***
(0.00307)
0.0485***
(0.00719)
0.0554***
(0.00975)
0.0551***
(0.0118)
0.00116
(0.00108)
0.00525*
(0.00253)
0.00104
(0.000993)
0.831
0.00253*
(0.00129)
0.000149
(0.000869)
0.00238**
(0.000962)
0.0290***
(0.00308)
0.0485***
(0.00722)
0.0554***
(0.00979)
0.0551***
(0.0119)
0.00121
(0.00106)
0.00526*
(0.00255)
—————–Education Trend—————(6)
(7)
(8)
0.00111
(0.000950)
(5)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed in Equation 4.1. All columns have year and person fixed effects for 850,158 people
2
3
(5,868,061 observations) and include Age, Age and Age as covariates. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years
in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Firm Size categories follow
Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by
size are microfirms of size equal or larger than 5, but less than 10 employees.
R-squared
MarginalEff
StdErr
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
-0.00201
(0.00122)
0.00939***
(0.00154)
———————-Baseline———————
(2)
(3)
(4)
Total Effect (Equation 4.2)
Import Share
0.00175
(0.00139)
Import Share*College
(1)
Table 4.17. Log Real Earnings and the Return to Skill using Chinese Import Share - First Industry for the Worker Total Effect
133
0.858
0.859
0.00711***
(0.00107)
0.0231***
(0.00325)
0.0350***
(0.00620)
0.0427***
(0.00936)
0.0491***
(0.0112)
-0.00195
(0.00264)
0.0229***
(0.00599)
0.858
0.00251
(0.00172)
-0.000736
(0.00143)
0.00787***
(0.00117)
0.0238***
(0.00324)
0.0363***
(0.00622)
0.0438***
(0.00927)
0.0507***
(0.0110)
-0.00155
(0.00258)
0.0223***
(0.00599)
0.859
0.00713***
(0.00113)
0.859
0.00181
(0.00127)
0.859
0.00298*
(0.00166)
0.00108
(0.00101)
0.00191
(0.00113)
0.0243***
(0.00326)
0.0370***
(0.00631)
0.0447***
(0.00933)
0.0510***
(0.0106)
-0.000752
(0.00205)
0.0220***
(0.00592)
0.860
0.00193
(0.00128)
0.00123
(0.00106)
0.00183
(0.00113)
0.0244***
(0.00325)
0.0372***
(0.00631)
0.0448***
(0.00932)
0.0512***
(0.0107)
-0.000690
(0.00203)
0.0219***
(0.00592)
0.860
0.00305*
(0.00165)
—————–Education Trend—————(5)
(6)
(7)
(8)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed in Equation 4.1. All columns have year and person-firm match fixed effects for a total of
1,017,056 matches (5,689,217 observations), and include Age, Age2 and Age3 as covariates. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working
for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Firm Size
categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table,
reference firms by size are microfirms of size equal or larger than 5, but less than 10 employees.
R-squared
MarginalEff
StdErr
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
-0.000871
(0.00141)
0.00798***
(0.00116)
———————-Baseline———————
(2)
(3)
(4)
Within Job Spell (Equation 4.3)
Import Share
0.00244
(0.00177)
Import Share*College
(1)
Table 4.18. Log Real Earnings and the Return to Skill using Chinese Import Share - First Industry for the Worker Within Job Spell Effect
134
0.829
0.829
0.00853
(0.00592)
-0.00410
(0.00525)
0.0126***
(0.00257)
0.829
0.0279***
(0.00294)
0.0467***
(0.00681)
0.0529***
(0.00933)
0.0519***
(0.0122)
-0.000556
(0.00189)
0.00583**
(0.00278)
-0.000907
(0.00521)
0.829
0.00829
(0.00555)
-0.00428
(0.00509)
0.0126***
(0.00256)
0.0281***
(0.00293)
0.0472***
(0.00680)
0.0538***
(0.00936)
0.0533***
(0.0121)
-0.000370
(0.00175)
0.00556*
(0.00277)
0.831
0.831
-0.00655
(0.00460)
-0.00119
(0.00413)
-0.00536***
(0.00114)
0.831
0.0288***
(0.00288)
0.0481***
(0.00675)
0.0548***
(0.00958)
0.0544***
(0.0119)
0.000664
(0.00148)
0.00549*
(0.00267)
-0.00251
(0.00410)
0.831
-0.00663
(0.00441)
-0.00116
(0.00402)
-0.00547***
(0.00116)
0.0288***
(0.00288)
0.0481***
(0.00677)
0.0546***
(0.00962)
0.0542***
(0.0120)
0.000624
(0.00148)
0.00556*
(0.00268)
—————–Education Trend—————(6)
(7)
(8)
-0.00252
(0.00422)
(5)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed in Equation 4.1. All columns have year and person fixed effects for 850,048 people
(5,866,337 observations) and include Age, Age2 and Age3 as covariates. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Sample includes only manufacturing workers born between
years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Firm Size categories follow
Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by
size are microfirms of size equal or larger than 5, but less than 10 employees.
R-squared
MarginalEff
StdErr
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
Total Effect (Equation 4.2)
Chinese Ex.
-0.000665
(0.00552)
Chinese Ex.*College
———————-Baseline———————
(1)
(2)
(3)
(4)
Table 4.19. Log Real Earnings and the Return to Skill using Chinese Exports to RoW - Yearly Industry for the Worker Total Effect
135
0.858
0.858
0.00689
(0.0103)
-0.00709
(0.00927)
0.0140***
(0.00314)
0.858
0.0232***
(0.00343)
0.0351***
(0.00669)
0.0423***
(0.00995)
0.0484***
(0.0118)
-0.00363
(0.00428)
0.0232***
(0.00638)
-0.00725
(0.00936)
0.859
0.00313
(0.00969)
-0.0106
(0.00879)
0.0138***
(0.00308)
0.0233***
(0.00342)
0.0353***
(0.00664)
0.0426***
(0.00986)
0.0488***
(0.0117)
-0.00334
(0.00396)
0.0229***
(0.00633)
0.859
-0.00489
(0.00762)
0.859
-0.00666
(0.00801)
-0.00435
(0.00755)
-0.00232*
(0.00125)
0.860
0.0242***
(0.00339)
0.0369***
(0.00671)
0.0441***
(0.00986)
0.0501***
(0.0113)
-0.00213
(0.00320)
0.0225***
(0.00630)
-0.00830
(0.00725)
0.860
-0.0102
(0.00757)
-0.00773
(0.00719)
-0.00246*
(0.00123)
0.0243***
(0.00339)
0.0369***
(0.00671)
0.0441***
(0.00987)
0.0501***
(0.0113)
-0.00215
(0.00321)
0.0225***
(0.00632)
—————–Education Trend—————(5)
(6)
(7)
(8)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed in Equation 4.1. All columns have year and person-firm match fixed effects for a total
2
3
of 1,016,790 matches (5,687,870 observations) and include Age, Age and Age as covariates. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Sample includes only manufacturing
workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity.
Firm Size categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In
this table, reference firms by size are microfirms of size equal or larger than 5, but less than 10 employees.
R-squared
MarginalEff
StdErr
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
Within Job Spell (Equation 4.3)
Chinese Ex.
-0.00356
(0.00995)
Chinese Ex.*College
———————-Baseline———————
(1)
(2)
(3)
(4)
Table 4.20. Log Real Earnings and the Return to Skill using Chinese Exports to RoW - Yearly Industry for the Worker Within Job Spell Effect
136
0.829
0.829
0.00402
(0.0115)
-0.00876
(0.0111)
0.0128***
(0.00290)
0.829
0.0279***
(0.00311)
0.0466***
(0.00734)
0.0528***
(0.00959)
0.0519***
(0.0123)
-0.000576
(0.00170)
0.00585**
(0.00265)
-0.00566
(0.0114)
0.829
0.00402
(0.0114)
-0.00880
(0.0109)
0.0128***
(0.00288)
0.0279***
(0.00311)
0.0468***
(0.00733)
0.0531***
(0.00962)
0.0524***
(0.0122)
-0.000440
(0.00159)
0.00579**
(0.00265)
0.831
0.831
-0.0117
(0.00837)
-0.00581
(0.00800)
-0.00593***
(0.00103)
0.831
0.0288***
(0.00307)
0.0481***
(0.00710)
0.0547***
(0.00961)
0.0545***
(0.0118)
0.000739
(0.00137)
0.00548**
(0.00250)
-0.00703
(0.00811)
0.831
-0.0116
(0.00839)
-0.00570
(0.00799)
-0.00587***
(0.00103)
0.0289***
(0.00307)
0.0482***
(0.00709)
0.0548***
(0.00959)
0.0546***
(0.0118)
0.000728
(0.00138)
0.00545**
(0.00249)
—————–Education Trend—————(6)
(7)
(8)
-0.00715
(0.00813)
(5)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed in Equation 4.1. All columns have year and person fixed effects for 850,158 people
2
3
(5,868,061 observations) and include Age, Age and Age as covariates. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working for at least two years
in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Firm Size categories follow
Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table, reference firms by
size are microfirms of size equal or larger than 5, but less than 10 employees.
R-squared
MarginalEff
StdErr
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
Total Effect (Equation 4.2)
Chinese Ex.
-0.00562
(0.0116)
Chinese Ex.*College
———————-Baseline———————
(1)
(2)
(3)
(4)
Table 4.21. Log Real Earnings and the Return to Skill using Chinese Exports to the RoW - First Industry for the Worker Total Effect
137
0.858
0.858
0.00461
(0.0135)
-0.00943
(0.0126)
0.0140***
(0.00318)
0.858
0.0232***
(0.00333)
0.0351***
(0.00635)
0.0422***
(0.00942)
0.0484***
(0.0112)
-0.00367
(0.00405)
0.0231***
(0.00591)
-0.00960
(0.0130)
0.859
0.00120
(0.0132)
-0.0127
(0.0124)
0.0139***
(0.00314)
0.0233***
(0.00333)
0.0352***
(0.00634)
0.0424***
(0.00936)
0.0486***
(0.0110)
-0.00336
(0.00375)
0.0230***
(0.00587)
0.859
-0.00723
(0.0101)
0.859
-0.00910
(0.0104)
-0.00673
(0.00998)
-0.00237*
(0.00120)
0.860
0.0243***
(0.00331)
0.0369***
(0.00637)
0.0441***
(0.00930)
0.0502***
(0.0106)
-0.00215
(0.00305)
0.0224***
(0.00580)
-0.0103
(0.0102)
0.860
-0.0122
(0.0105)
-0.00979
(0.0101)
-0.00238*
(0.00118)
0.0243***
(0.00331)
0.0370***
(0.00637)
0.0442***
(0.00930)
0.0502***
(0.0106)
-0.00218
(0.00306)
0.0224***
(0.00580)
—————–Education Trend—————(5)
(6)
(7)
(8)
*** p<0.01, ** p<0.05, * p<0.10.
Note: Chinese Import Share defined as the average of Chinese imports as a share of apparent consumption in Denmark, Finland and Germany, detailed in Equation 4.1. All columns have year and person-firm match fixed effects for a total of
2
3
1,017,056 matches (5,689,217 observations) and include Age, Age and Age as covariates. Sample includes only manufacturing workers born between years 1920 and 1991 earning a real (in 2010 values) income of SEK120,000 working
for at least two years in manufacturing firms with more than 5 employees and values above zero for sales and capital intensity. Robust standard errors are reported in parenthesis, and are adjusted for clustering at the industry level. Firm Size
categories follow Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49), medium size firms (50 to 249), and large firms (250+) with the addition of very large firms of larger than 500 employees. In this table,
reference firms by size are microfirms of size equal or larger than 5, but less than 10 employees.
R-squared
MarginalEff
StdErr
log Sales
log Capital Intensity
500+
250-499
50-249
Firm Size : 10-49
Within Job Spell (Equation 4.3)
Chinese Ex.
-0.00623
(0.0132)
Chinese Ex.*College
———————-Baseline———————
(1)
(2)
(3)
(4)
Table 4.22. Log Real Earnings and the Return to Skill using Chinese Exports to the RoW - First Industry for the Worker Within Job Spell
Estimates
15-Food
.0006
16-Tobacco
17-Textiles
.003
.002
.001
0
.0003
0
1996
2000
2004 2007
.08
0
1996
20-Wood
.015
.01
.005
0
2000
2004 2007
.006
.004
.002
0
1996
2000
2004 2007
1996
25-Rubber&Plastic
.01
2000
0
2000
2004 2007
1996
2000
0
1996
2000
2004 2007
1996
35-OtherTransport
.04
2000
2004 2007
2004 2007
1996
23-Coke&Petrol
.01
.01
.005
0
1996
2000
2004 2007
27-BasicMetal
2000
2004 2007
28-Metal
.03
.02
.01
0
1996
2000
2004 2007
2000
2004 2007
1996
2000
2004 2007
29-Machinery&Equipment
.015
.01
.005
0
1996
2000
2004 2007
1996
33-Medical&Optic
2000
2004 2007
34-MotorVehicle
.01
.015
.01
.005
0
1996
2004 2007
0
1996
32-RadioTvCommunication
2000
24-Chemicals
.005
.15
.1
.05
0
.04
2000
0
2004 2007
31-ElectricMachinery
0
1996
.004
.015
.01
.005
0
.08
.3
.2
.1
0
2004 2007
19-Leather
.1
.05
.008
26-NonMetalMineral
30-OfficeMachinery&Computer
2000
22-Publishing&Media
2004 2007
.02
.01
0
.005
1996
1996
21-Pulp&Paper
18-WearingApparel
.6
.4
.2
0
.04
.005
0
1996
2000
2004 2007
1996
2000
2004 2007
36-Furniture
.06
.04
.02
0
.02
0
1996
2000
2004 2007
1996
2000
2004 2007
year
Composite
Sweden
Graphs by Industry
Figure 4.3. Swedish and Finish Imports from China as a share of respective apparent
consumptions, by industry, individual scales, 1996-2007
138
Table 4.23. Data Description
Firm Data
Total Wages
Sum of personnel costs for the year (Summa personalkostnader)
Total Sales
Sum of revenues for the year (Nettomsättning)
Capital (K)
Sum of the following reported tangible assets for the year:
Land and Buildings
Machinery and Equipment
Ongoing Construction and Advance payments for tangible fixed
assets
Total Employees (N)
Total employees (Antal Anställda)
Capital Intensity
Calculated as K/N
Industry Classification
Industry Codes are reported in four different systems (SNI1969, 1992,
2002, 2007) which all have been converted to SNI2002 at the 5-digit
and 2-digit level
Business Register
Legal Form
Classification by type of legal entity
Controlling Ownership
Standard Classification by ownership control
Employee Data
Annual Wage
Taxed wage income (Kontant Bruttolön)
Age
As reported
Gender
As reported
Level of Highest Education
Under the old SUN code, the following categories:
Pre High School
Some High School without a diploma
High School diploma
Less than 2 years of University
More than 2 years of University, includes those with diploma
Postgraduate Studies
Targeted Field of Education
Targeted diploma subject
Notes: Firm Data source is Account Statistics (FEK). Data for 1980-1996 are for a sample of companies.
Data comes with a 2 year lag. Only non-imputed companies included. Business Register data is sourced
from the Business Register Database (Företagsregistret). Data available from 1980 onwards. Employee
Data source is Register Based Labor Statistics (RAMS). Data available from 1985 onwards. Each individual
is linked to a firm, and a plant where applicable.
139
Table 4.24. Sample Selection and Data Handling
Firm Data:
Total Employees
Sample only includes firms who have never had fewer than 5 employees
in all the years they were active in the data. For some of the regressions,
the firms are categorized into size groups following Eurostat classification as microfirms (less than 10 employees), small firms (10 to 49),
medium size firms (50 to 249), and large firms (250+). I augment this
scale with the addition of a fifth group of very large firms of larger than
500 employees.
Business Register:
Legal Form
I pool together firms that fall under the following legal forms:
(i) Limited Partnerships (Handelsbolag),
(ii) Limited Liability Companies other than banking and insurance companies. (Aktiebolag)
Employee Data:
Annual Wage
I only look at individuals who have earned at least SEK 120,000 (≈
USD 16,250) annually to exclude extreme part-time workers.
Age
Sample is restricted to workers born between years 1920 and 1992.
Education
I exclude workers whose education levels are unknown. While in the
data is available in 5 level detail for educational attainment, the individuals are grouped into the following three educational groups: less than
high school diploma, high school diploma holders, and at least some
college based on the more detailed classification.
Table 4.25. Chinese Import Measure Correlations, 1996-2007
Swedish CMP
Composite CMP
Composite Import Share
Chinese Exports
(1)
Swe CMP
(2)
Comp. CMP
(3)
Comp. M Share
(4)
Chinese Exports
1.0000
0.8717
0.7891
0.4797
1.0000
0.8522
0.5667
1.0000
0.6418
1.0000
Note: Swedish CMP and Composite CMP (rows and columns 1 and 2) are constructed using Equation (4.1).
Composite import share uses Equation (4.6). Chinese Exports looks at the share of Chinese exports to the
rest of the world over World GDP, as in Equation (4.7). All measures are at the industry level.
140
Table 4.26. Conversion of SITC product classifications under UN Comtrade to
Swedish industry codes SNI for manufacturing industries.
SITC
SITC Name
SNI
SNI Name
1
4
6
7
9
11
12
65
15
Manufacture of food products and beverages
16
17
Manufacture of tobacco products
Manufacture of textiles
84
85
61
Meat and meat preparations
Cereals and cereal preparations
Sugars, Sugar preparations and honey
Coffee, tea, cocoa, spices, and manufactures thereof
Miscellaneous edible products and preparations
Beverages
Tobacco and tobacco manufactures
Textile yarn, fabrics, made-up articles, n.e.s., and related
products
Articles of apparel and clothing accessories
Footwear
Leather, leather manufactures, n.e.s., and dressed furskins
18
Manufacture of wearing apparel; dressing and dyeing of fur
19
63
Cork and wood manufactures (excluding furniture)
20
64
Paper, paperboard and articles of paper pulp, of paper or of
paperboard
Printed matter
Musical instruments and parts and accessories thereof;
records, tapes and other sound or similar recordings
Coke and semi-coke (including char) of coal, of lignite or
of peat, whether or not agglomerated; retort carbon
Petroleum, petroleum products and related materials
21
Tanning and dressing of leather; manufacture of luggage,
handbags, saddlery, harness and footwear
Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting
materials
Manufacture of pulp, paper and paper products
22
Publishing, printing and reproduction of recorded media
23
Manufacture of coke, refined petroleum products and nuclear fuel
Chemicals and related products, n.e.s.
24
Manufacture of chemicals and chemical products
Rubber manufactures, n.e.s.
Plastics in primary forms
Plastics in non-primary forms
Non-metallic mineral manufactures, n.e.s.
Iron and steel
Non-ferrous metals
Manufactures of metals, n.e.s.
25
Manufacture of rubber and plastic products
26
27
Manufacture of other non-metallic mineral products
Manufacture of basic metals
28
General industrial machinery and equipment, n.e.s., and
machine parts, n.e.s.
Office machines and automatic data-processing machines
Electrical machinery, apparatus and appliances, n.e.s., and
electrical parts thereof
Telecommunications and sound-recording and reproducing
apparatus and equipment
Photographic apparatus, equipment and supplies and optical
goods, n.e.s.; watches and clocks
Instruments and appliances, n.e.s., for medical, surgical,
dental or veterinary purposes
Road vehicles (including air-cushion vehicles)
Other transport equipment
Furniture, and parts thereof; bedding, mattresses, mattress
supports, cushions and similar stuffed furnishings
—
29
Manufacture of fabricated metal products, except machinery and equipment
Manufacture of machinery and equipment n.e.c.
30
31
Manufacture of office machinery and computers
Manufacture of electrical machinery and apparatus n.e.c.
32
Manufacture of radio, television and communication equipment and apparatus
Manufacture of medical, precision and optical instruments,
watches and clocks
892
898
325
33
5 excl
57&58
62
57
58
66
67
68
69
74
75
77
76
88
872
78
79
82
—
33
34
35
36
Manufacture of motor vehicles, trailers and semi-trailers
Manufacture of other transport equipment
Manufacture of furniture; manufacturing n.e.c.
37
Recycling
Notes: The Product level trade information in UN Comtrade does not allow for the identification of the
Recycling industry to match the Swedish industry category.
141
5. Import Competition and Technological
Changes: Mobility of Workers and Firms1
5.1 Introduction
Previous literature identifies substantial differences in wages across firms, within
narrowly defined industries and for equally skilled workers. This points to
sorting of workers across firms as an important factor in shaping the labor
market outcomes. In a recent study, Card et al. (2013) analyze wage distribution trends in Germany in the period 1985-2009 and show that around one
third of the overall increase in wage dispersion in Germany during the analyzed period can be attributed to increased assortativeness in the matching of
workers with firms. Therefore it is important to study how workers are allocated across firms and to analyze the main factors that contribute to a stronger
link between the characteristics of firms and individuals.
In this paper, we study the sorting of workers across firms in the Swedish
manufacturing sector. We apply the methodology developed by Abowd, Kramarz and Margolis (1999) (hereafter AKM) on detailed matched worker-firm
micro data of the Swedish manufacturing industries for the period of 19962007. Using workers’ and firms’ fixed effects estimated in the AKM model,
we construct their joint distribution and analyze how this distribution changes
over time.
We consider two potential causes of these changes: technology and trade.
The rise in the adoption of information and communication technologies (ICT)
has been intense in Sweden since the late 1990’s, and industries in the manufacturing sector are heterogeneous in their adoption of ICT. To account for
this, we rank the manufacturing industries by their ICT intensity and explore
the differences between them. Parallel to these changes, the Swedish economy has experienced a rapid increase in international trade, measured by both
export and import values and driven by international integration policies. Of
particular interest is the brisk change in the share of trade with less developed
1 This chapter is coauthored with Teodora Borota Milicevic (Uppsala University) and Rita
Ginja (Uppsala University). We thank Nils Gottfries for invaluable discussions. We also
thank Aron Berg, Mikael Carlsson, Fredrik Heyman, Francis Kramarz, Renata Narita, Oskar
Nordström Skans for comments, as well as the participants at the 15th NOITS, the 2014 Nordic
Workshop on Matched Data, and the Riksbank GSMG seminars for insightful suggestions.
The authors gratefully acknowledge financial support from the Wallander-Hedelius-Browaldh
Foundation and the Ragnar Söderbergs Stiftelse.
142
and labor-intensive countries.2 Specifically, we observe a significant increase
in trade with China, which accelerated after China joined the WTO in 20013 .
Manufacturing industries exhibit different degrees of exposure to increasing
Chinese import penetration, and this allows us to study the effects of import
penetration on the sorting of workers, as well as to point to possible interactions between ICT technology and trade with developing countries.
In order to examine the role of technology and trade, in addition to the interactions between these factors, we divide the data into two periods: 1996-2000
(Period 1) and 2001-2007 (Period 2). We choose this division because China
entered the WTO in 2001, which we take to be an exogenous trade shock.
Furthermore, we classify manufacturing industries according to their ICT intensity (low/high) and the change in Chinese imports penetration (low/high)
between the two periods. Examining the changes in sorting patterns within
each group of industries, we find two main results.
First, increased sorting is a phenomenon that appears primarily in the ICTintensive industries. We do not find significant changes of the sorting pattern
in the less ICT intensive industries between Period 1 and Period 2. This result
holds also if we differentiate industries within the low ICT intensive group into
those with and those without a strong increase in the Chinese import penetration. However, when analyzing industries with high ICT intensity, we observe
an increase in the share of low fixed effect workers in the low fixed effect
firms from Period 1 to Period 2, and a reduction in the share of low fixed effect workers in the high fixed effects firms. At the same time, the share of
high fixed effect workers in high fixed effects firms increases. This allocation
pattern may have been caused by the nature of the ICT technologies and their
non-uniform adoption across firms.
Second, we find that the change in the sorting pattern is not uniform within
the high ICT intensity group. In high ICT industries with a high increase in
Chinese import penetration, we observe a strong increase in the share of high
fixed effect workers in high fixed effect firms, and a reduction in the share of
low fixed effect workers in the high fixed effect firms. Thus, we see stronger
sorting on the high end of the firm distribution in these industries, while there
are no significant changes on the low end. To the extent that worker and firm
effects represent their skills and quality, we observe strong skill upgrading in
the high quality firms within these industries. In high ICT industries with a low
Chinese import penetration, we also observe increased sorting, but primarily
at the other end of the distribution. The most important change here is that
2 Given
the different specialization patterns of developing labor-intensive countries relative
to developed economies, their stronger economic (trade) integration may be viewed as a form
of technological change in developed economies (e.g. through trade in intermediates or task
offshoring).
3 Many developed countries’ imports from China compose the bulk of the growth of imports
from developing countries (for more extensive reasons for selecting China, see section 5.3).
143
there is an increase in the share of low fixed effects workers in the low fixed
effects firms.
In Section 5.5, we present a tentative theoretical model that could potentially explain the effects of trade that we see within the ICT-intensive industries. We consider an economy where there are two types of workers, low-skill
and high-skill workers. Firms differ in their productivity and they can post one
of two types of jobs: an unqualified job, which can be performed by any type
of laborer, or a qualified job, which can only be performed by high-skill workers. The latter jobs are more productive, but the fixed costs are also higher. In
equilibrium, there will be a partitioning of firms: highly productive firms post
qualified jobs, less productive firms post unqualified jobs and the lowest productivity firms exit. We use the model to analyze the effect of an increase in
trade with low-skill countries (decrease in the productivity of domestic unqualified jobs) within the ICT intensive industries (where the relative productivity
of qualified job is high).
In order to allow for a differential change in trade, we assume two heterogeneous industries but a common labor market. The two industries, T and
N, are ex-ante symmetrical but we expose industry T to an increase in import
competition. We assume that this reduces the productivity of unqualified jobs
in industry T. As a result of the shock, the least productive firms in industry T will exit, while some firms with higher productivity will upgrade their
posts from unqualified to qualified jobs. Consequently, low-skilled workers
are pushed out of the T sector, low-skill unemployment increases and wages
of low-skill workers decrease. Most low-skill workers who leave the T industry are hired in the N industry where the number of unqualified jobs increases.
In this sense, the model explains increased sorting at the high end (upgrading)
in the sector affected by increased import competition and a concentration of
low-skill workers in unqualified jobs in the sector that is not affected by the
trade shock.
On the other hand, in low ICT intensive industries where the relative productivity of the qualified jobs is lower, the described responses to a trade shock
are significantly weaker. These industries have a wider range of operating
firms (lower exit threshold) and a high share of unqualified jobs, as more firms
find it non-profitable to post high-cost vacancies. A trade shock thus makes
the job type trade-off and the cross-industry reallocations less pronounced in
low ICT industries.
The effects of technological changes and trade on industry and labor market
dynamics have been analyzed extensively in previous literature, both theoretically and empirically. With respect to technology, a branch of literature places
skill-biased technological change at the center of the theoretical approach and
models a sorting mechanism where firms that use different types of technolo-
144
gies employ labor input of different skill levels.4 Autor and Dorn (2013) analyze changes in employment across skill groups and they find evidence of an
increase in the employment share of high- and low-skilled workers relative to
the mid-skilled group, which they argue may be linked to the advances in and
adoption of ICT related technology. They do not analyze the changes in allocation patterns across firms, but to the extent that these employment changes
are linked to particular types of firms, they may have an impact on the distribution of workers across firms (i.e. sorting patterns).
Import competition from low-wage countries, on the other hand, may cause
stronger competitive pressures in the less productive end of the firm distribution if the production technologies and the type of goods produced are more
similar to the low-wage country’s technology and exports. Moreover, heterogeneous firm trade models would predict that import competition may cause
pressures on the low-skilled labor as firms upgrade their skill composition in
response to this pressure.5 Several recent empirical studies have focused on
the effects of increased Chinese import penetration on labor market outcomes,
such as employment, wages, and welfare payments.6 These papers, however,
do not study the effects on the mobility of workers across firms and industries,
nor the sorting of workers.
Several studies on heterogeneous firms and trade imply a link between import competition (both in general and from developing countries in particular)
and technological and labor input choices of heterogeneous firms, which has
not been studied extensively in the empirical literature.7 In their theoretical
work, Davidson et al. (2008) and Davidson and Matusz (2012) analyze the
effect of exporting on technology choice and the resulting labor market outcomes (high end sorting in exporting industries), but they also address the
import-competing industries and both studies arrive at similar results8 . In the
model by Davidson et al. (2008), import competition reduces the gap in revenues of different types of workers, and thus may result in increased nega4 See Acemoglu (1999) and Caselli(1999), among the first. Albrecht and Vroman (2002)
arrive at a similar prediction in the model with skill-job type complementarities and unemployment.
5 For a review of the literature, see e.g. Ashournia et al. (2012)
6 Autor et al. (2013) analyze the effect of industry-level Chinese imports on U.S. local labor
markets and find a negative effect on wages and employment in import-competing markets.
Ashournia et al. (2012) explore both industry-level and firm-level effects of Chinese import
penetration on Danish firms and find that it causes low-skill wage declines at the firm-level.
Alvarez and Opazo (2011) study the effect on Chilean average firm-level wages, and find similar
negative wage effects in a developing economy as well.
7 See e.g. Kugler and Verhoogen (2011), Bas and Berthou (2013). Among the literature that
is interested in globalization, technological choice and sorting, but does not necessarily focus
on import competition, we note Grossman et al. (2014), Grossman and Maggi (2000), Costinot
(2009), Costinot and Vogel (2010) and Yeaple (2005), among others.
8 Relative to Davidson et al. (2008), the latter paper introduces firm heterogeneity as in
Melitz (2003) and monopolistic competition, but preserves the main framework.
145
tive matching (i.e. high-skill workers accepting jobs in low-performing firms
within the import-competing industries). We focus on the similar trade channel (import competition), but originating from developing, low-skill countries,
which may be expected to affect domestic low-skill labor more than the highskilled. Furthermore, we explore the trade and technology channels as exogenous to each other, as well as their interactions, in order to identify the exact
sorting patterns that these two sources may produce.
In a recent paper, Autor et al. (2014) attempt to disentangle the effects
of two forces - the ICT technology and import competition - on employment
across skills. They find that technological progress and import competition
have rather independent effects, as opposed to some previous hypotheses of
the two being just two faces of the same phenomenon. We follow a similar
approach, but we add in three important dimensions: (1) since we have access
to firm data, we can track changes of firms over time and control for the firm
where the individual works, (2) beyond Autor et al. (2014), we study the
impacts of technological changes and trade not only on employment, but also
on labor allocations across firms, and (3) we document the sorting effects of
the interactions between technology and import competition.9
Within the empirical literature that focuses on the sorting phenomena in particular, Davidson et al.(2014) explore the matching patterns between workers
and firms in Swedish manufacturing industries and find evidence for the effect
of trade liberalization on this aspect of the labor market. Greater openness in
comparative-advantaged industries increases the degree of positive assortative
matching, measured by the correlation between the individual and firm components of the wage. This effect is not present in the comparative-disadvantaged
industries (import-competing industries). Their results are robust to the inclusion of the controls for the technical change across industries, which may have
also contributed to the assortativeness of worker-firm matching. We follow a
similar approach in the empirical part, but attempt to document the sorting
phenomenon in greater detail, isolating and interacting the effects of trade and
technology. Håkanson et al. (2013) analyze the Swedish data as well and also
find a significant increase in assortative matching. They contrast two potential explanations - off-shoring and skill-biased technical change - and find that
the latter seems to have been more important. However, none of the existing
literature explore the interactions between different forces shaping the labor
distribution across firms, nor does it characterize the sorting patterns in detail
(e.g. which parts of the distribution are affected).
The choice of Sweden as the country of study is interesting for three main
reasons. First, the availability of longitudinal data on characteristics of firms
and workers (information about workplace and periods out of the labor force)
9 Previous
work on the industry effects of globalization and technology have been placing
the two side by side, attempting to determine the relative importance of the two, given that their
effects could be disentangled. There are few attempts to define the allocation outcomes of their
interactions.
146
allows us to study in detail the transitions of workers across firms and in-andout of the labor market. Second, most of the studies on the effects of exposure
to trade and technological changes on the wages and employment status of
individuals are conducted using U.S. data. The United States is a large open
economy capable of influencing world prices of goods and it has an independent trade policy. On the contrary, Sweden is part of the EU and has limited
power in international trade agreements. Therefore sharp changes in international trade flows, such as Chinese exports to the world, are mostly exogenous
shocks to Swedish firms. Finally, we focus our study on manufacturing firms,
which represent about 1/3 of the total GDP and occupy just over 1/3 of the
total of workers in the country, similar to other developed EU countries. For
this reason, the Swedish manufacturing sector may be used as representative
of the manufacturing sector of the EU, and the conclusions drawn from this
paper are probably relevant to other EU countries.
We describe our data in Section 5.2. Section 5.3 describes our empirical
strategy. Section 5.4 presents the results. In Section 5.5 we present a tentative
model, which suggests potential mechanisms behind our findings. In Section
5.6 we summarize our results and conclude.
5.2 Data
We use firm- and worker-level data from databases either collected or maintained by Statistics Sweden (SCB). The data is confidential as original worker
and firm identifiers are stripped and reassigned by SCB, but access to the
database is not exclusive. We convert all monetary values to 2010 Kronor
using the publicly available Consumer Price Index information from SCB.
Information about Chinese trade figures comes from UN Comtrade (see http:
//comtrade.un.org/). ICT classifications are based on those set by Van Ark et
al. (2003).
Firm data
Firm-level balance sheet data is available in Statistics Sweden’s Account Statistics (FEK - Företagsekonomisk Statistik). While most of the variables are
available from 1980 onwards, it only covers a selective sample of large companies until 1996, which motivates our focus on the period from 1996 onward.
This database includes information on total wage spending, sales, profit, capital, number of employees, and industry classification at the firm level. The data
is released with a two year lag, and it only includes non-imputed companies
and data. Industry classification code systems were updated three times during
the entire time period of the series, changing the industry code index system
of a firm after each update (the index systems used are SNI1969, SNI1992,
SNI2002 and SNI2007). In an effort to have a continuous industry classification under one index, we first use the conversion keys supplied to us by
147
Statistics Sweden where available. If the conversion key was not successful
in producing a match between two indices for a particular industry, we then
make use of overlapping years in different code systems to generate our own
conversion key. In instances where an industry has been split up into parts,
we assign the firms to the new industry whose description best matches the
old industry description.10 We do this exercise at the four digit detail level
industry code, whereas we only use the two digit codes in the regressions.
We supplement this database with the Business Register Database (Företagsregistret), which includes information on the legal form and controlling ownership of the firm and municipal location from 1980 onwards.
Firm-level trade statistics of exporter/importer status, and the associated trade
value and destination are available from the Foreign Trade Database (Utrikeshandel) from 2000 onwards. For intra-European Union trade, the database
has a minimum requirement of SEK 4.5 million (≈ USD 610,000) in value
to be registered as an importer or exporter. As a result, we do not observe any
within-EU trade less than this cutoff in the data.
Worker data
Matched employer-employee data comes from the Swedish Tax Authority
(Skatteverket) and is available in Register Based Labor Statistics (RAMSRegisterbaserad Arbetsmarknadsstatistik) maintained by Statistics Sweden.
The data is available from 1985 onwards, and each individual is linked to a
firm (and a plant where applicable). In this database, an individual is tied to
a place of work if he/she was employed there in the third week of November,
in line with International Labour Organization’s definition. For each worker
registered, we have information on the annual labor income, main place of
employment according to the definition stated above (firm and plant where
applicable), age, gender, highest level of education and field of education. We
group individuals into educational groups: less than high school diploma, high
school diploma holders, and at least some college based on detailed classification about the education level of individuals.
Trade and ICT data
We use UN Comtrade data for international trade between Sweden and its
partners. Comtrade data classifies trade based on product (not industry) level
codes, and manufactured goods are indexed by material. To be able to match
these product codes to Swedish industry codes from the firm-level data, we
performed a match between the two indices based on index descriptions as
shown in the Appendix. For all the other industries, we have taken the share
10 The
number of industries subject to this assignment are: 3 industries from SNI1992 to
SNI2002 matching, and 2 industries from the SNI2007 to SNI2002 matching.
148
of Chinese imports over all imports into Sweden and computed the change in
this share between 1996 and 2001 as can be seen in Table (5.6) in Appendix.11
The ICT classifications are based on Van Ark et al. (2003) and included ICT
producing, ICT intensive and less intensive ICT categories. Between ICT producing and intensive categories, two industries are split into two at the three
digit level of detail. Since we are using industry classifications at the two
digit detail level, we merge the ICT producing and intensive categories into
the same group in our classification of high ICT industries, while keeping low
ICT industries exactly the same as Van Ark et al. (2003). Details of the classification can be seen in Table (5.2) in the Appendix.
Sample Selection
Our data is composed of limited liability partnerships and companies that are
active from 1996 to 2007, since the firm-level data is based on a sample of
companies before 1996. We keep firms with at least 5 employees per year
during their entire presence in this range. While we mostly focus on manufacturing firms, we also consider all the other sectors in the descriptive analysis.12
The data does not contain information regarding full-time or part-time employment status of individuals. Therefore, we restrict the baseline sample to
those with labor earnings of at least SEK 120,000 a year (SEK 10,000 ≈ USD
1,570 a month). Next we drop individuals whose education level is unknown
and those who are born before 1920 or after 1991.13
More information about the databases and individual series can be found in
the Appendix.
5.3 Empirical Strategy
5.3.1 Basic Setup
We now present our basic econometric framework for disentangling the components of wage variation attributable to worker-specific and employer-specific
heterogeneity. We follow Abowd et al. (1999) and Card et al. (2013) in our
empirical exercise and assume that log real annual earnings yit of individual
i in year t can be modelled as an additively separable model of the worker
11 Since
Recycling is not an industry that product-level trade information allows us to identify, we have left out this industry (67 firms in Period 2) from our analysis.
12 There are a total of 60,907 firms in the database identified as manufacturing firms in this
period. Our restriction of minimum 5 employees drops about 51,000 firms, 72% of which
reported an average employment count below one employee.
13 The income restriction drops 401,074 employees, 51,343 of whom do not have an educational level assigned to them. Of the workers whose income is below the cutoff, about 26% of
them earned at most a total of SEK 10,000 (≈ USD 1,570) in a year, and about 67% of them
earned at most SEK 50,000 (≈ USD 7,850) annually. Following the income restriction, the age
restriction drops an additional 63 people from the final sample.
149
time-invariant characteristics αi , a component specific to the firm j where the
individual works in year t (denoted θJ(i;t) ), a set of time-varying observable
characteristics of the individual xit β, and an error component εit . Then, we
estimate the following model:
yit = αi + θJ(i;t) + xit β + εit .
(5.1)
In equation (5.1), αi subsumes a combination of skills and other time invariant factors specific to the worker i that are rewarded equally regardless of the
employer. xit β includes lifecycle components and aggregate shocks that affect
a worker’s wage in all jobs. In particular, xit includes year fixed effects and
cubic polynomial on age fully interacted with maximum lifetime educational
attainment. We consider two indicators of completed education of an individual: an indicator for high school degree and an indicator for some college
education or more, thereby making high school dropouts the excluded category. The firm effect θJ(i;t) is a proportional wage premium paid by firm j to
all employees (for example, rent-sharing).
We use this simple specification to obtain some descriptive features of the
wage dynamics between 1996 and 2007 in Sweden. In particular, we start by
presenting some descriptive statistics for three estimates from model 5.1: αi ,
θ
J(i;t) and εit . The residual of equation (5.1) is of particular interest to motivate
an additively separable model. We follow Low, Meghir and Pistaferri (2010)
and write εit as
εit = ψiJ(i,t) + φit + uit
(5.2)
where the match effect ψiJ(i,t) represents an idiosyncratic wage premium (or
discount) earned by individual i at firm j, relative to the baseline level αi + θ j .
We assume that ψiJ(i,t) has mean zero for all i and for all j in the sample interval. The match specific wage component arises in models in which there
is an idiosyncratic productivity component associated with each potential job
match where workers receive some share of the rents from a successful match
(e.g. Mortensen and Pissarides, 1994). As is typical in the earning dynamics
literature (see Meghir and Pistaferri, 2004), we assume that φit has mean zero
for each person in the sample interval, but contains a unit root. φit is a unit
root component that captures a drift in individuals’ earnings power. Innovations to this component could reflect employer learning, unobserved human
capital accumulation, promotions/demotions, health shocks, or the arrival of
outside offers. Finally, the transitory component uit represents any left-out
mean reverting factors. We assume that uit has mean zero for each person in
the sample interval.
Estimation and assumptions about εit
We estimate equation (5.1) by OLS. The firm fixed effects in equation (5.1)
are identified by individuals who move between firms and generate a large
150
firm network in which each firm is tied to at least one another firm in the
group through at least one worker who moves between them.14 We construct
the largest of such networks, which is called the Mobility Group in each period
and restrict our analysis to this group of interconnected firms. While we base
the choice of the largest mobility group to maximize the number of interconnected firms, this also gives us the largest network of interconnected workers.
Table 5.6 shows that this method includes between 82.7-94.9% of the firms,
and 97.5-99.5% of all the workers. For most of our analysis, we focus on firm
fixed effects, but we also consider plant-level fixed effects. When we study
plant level fixed effects, we drop single person plants from our largest Mobility Group. This means that the mobility group configuration disregards 594
and 525 workers in periods 1 and 2. In a few rare cases, the plant assignment
for the employee is unknown and these individuals are dropped from our last
block (7 and 2 employees respectively for periods 1 and 2). The worker effects
are estimated from repeated observations per worker.
The person and firm fixed effects in equation (5.1) are identified by OLS if
the three components in εit are (1) orthogonal to the individual and firm fixed
effects and (2) if they are orthogonal to the year fixed effects and to the cubic
polynomial on age interacted with maximum educational attainment. The assumption (2) is standard, whereas assumption (1) holds since the hypotheses
for ψiJ(i,t) , φit and uit stated above ensure that εit is orthogonal to the individual
fixed effects αi .
By conditioning on individual fixed effects αi and on θJ(i;t) , we allow for
systematic mobility of workers across firms to be correlated with individual
time invariant characteristics and firm specific wage-premia. However, εit may
not be orthogonal to the firm fixed effects, since there are forms of endogenous
mobility that could bias the estimate of firm fixed effects. Here we briefly
discuss here three forms of endogenous mobility that may arise because of
each component in εit (also see Card et al., 2013).
First, individuals may sort into firms based on an individual worker-firm
match component ψiJ(i,t) . To address this concern, we estimate a fully saturated model, which includes an indicator variable for each individual-job combination. The bottom rows of Table (5.7) compares the variance explained by
the two different models, and shows that the improvement in the fit with the
individual-job match model is relatively small compared to our baseline specification.
14 Abowd et al. (2004) point out that the estimated fixed effects may be subject to a downward
bias if the number of workers who switch between firms in the sample are too few; a problem
that they call "limited mobility bias." To address this issue and potentially improve the precision
of the estimated person and firm fixed effects, the analysis is repeated on two separate samples
of firms where the minimum number of movers are restricted either to at least 5 and at least 10.
This practice creates samples where the firms are interlinked with a higher degree of mobility,
but produces minimal changes in the joint distribution of the person and firm fixed effects, and
as a result does not alter any of the following conclusions.
151
Second, φit will be correlated with the firm fixed effects if wage growth predicts transitions across jobs. In other words, if permanent shocks to the wage
growth are correlated to job-to-job transitions. To address this concern, we
perform a basic event-study as suggested in Card et al. (2013). In particular,
we study the change in the mean earnings of workers who change jobs within
each interval and who were employed in their old and new firms for two years
in a row before and after the switch. We then classify the firms into highand low-paying firms based on the mean earnings of co-workers. In general,
workers leaving low-pay firms were already in an increasing earnings trend in
Period 1 before switching jobs. However, in Period 2 we do not detect any
pre-switch trend in the earnings of workers who leave either high- or low-pay
firms, regardless of the type of firm where they end up.15
Finally, if uit is correlated with job-to-job transitions, firm fixed effects will
be biased. In particular, there will be attenuation bias if individuals facing
positive (negative) transitory income shocks are more likely to move to high
(low) wage firms. By using the same event-study described above, we can
address this concern. Since, for Period 2, we are unable to detect a pre-switch
trend in the earnings of workers who leave either high- or low-pay firms independently of the type of firm in which they end up, it is likely that transitory
shocks are not correlated with job-to-job transitions. However, the seemingly
pre-existing trend in Period 1 among workers leaving low-pay firms specifically for high-pay firms requires a more thorough investigation of the validity
of this assumption.16
5.3.2 Industry classifications
For our analysis, we divide the data into two periods. Period 1 is defined as the
years before the Chinese membership in the WTO (1996-2000) and Period 2
as the post-Chinese membership years (2001-2007). We perform our analysis
on each period separately, as well as on the whole period.
In order to analyze the patterns of workers sorting by type into different
types of firms, we construct the joint distribution of the person and firm effects obtained from the baseline regressions for each of the two periods. We
classify industries according to their ICT intensity and their exposure to Chinese import competition, and track the changes in the joint firm-worker effects
distribution between Period 1 and Period 2.
15 These
results are available upon request.
recent papers criticize the methodology of Abowd et al. (1999) on the grounds
that the economic interpretation of the estimated worker and firm fixed effects is unclear; see
Hagedorn, Law and Manovskii (2012), Eeckhout and Kircher (2011) and Lise, Meghir and
Robin (2013). In light of this, we see the AKM decomposition into worker and firm fixed
effects primarily as a description of the covariance structure of the wages/earnings. We do not
take a stand on the underlying economic factors (complementarities, matching, individual and
collective bargaining, etc.) that generate these correlations.
16 Some
152
We rank the industries based on their ICT adoption into high and low groups.
The aggregate data shows a substantial increase in the share of ICT capital in
total capital of the manufacturing sector in Sweden within the analyzed period.
Since the data on firm-level ICT adoption is not available before 200417 , we
assume that most of the observed aggregate change in the ICT adoption has
been a result of its adoption within the industries that are the intensive users
of these technologies. In that sense, our distinction between the low and high
ICT intensive industries may be regarded as a distinction between the industries with relatively lower and higher change in the ICT adoption, respectively.
In order to define exposure to Chinese trade competition, we do the following. For each industry we obtain the share of Chinese imports to Sweden in
1996 and in 2001. We then rank industries according to the percentage change
in the share of imports between 1996 and 2001.18 Overall, we observe data
for 21 industries (see table 5.4). We then define High China Industries (or Tindustries) as the 10 industries with the largest change in the share of Chinese
imports and we define Low China Industries (or N-industries) as the 11 sectors
with the smallest change in the share of Chinese imports. The choice of China
as a focus of analysis presents two advantages. First, Chinese imports into
many developed countries compose the bulk of the growth of imports from
developing countries and that makes it a good proxy of cheap imported goods
from the rest of the world. And secondly, we follow the literature that has focused on empirical analysis of trade competitions (e.g., Autor and Dorn, 2014
and Balsvik et al., forthcoming).
5.4 Results
We now present the results from our main specification (equation 5.1). The
model of wage determination that includes the worker and firm fixed effects
is capable of explaining 88.20% of the variation in the data. The largest connected set includes 99.77% of the sample workforce and provides similar results: 49.5% the worker effect variation contribution and 50.5% the firm effect
variation contribution.19
17 EU-KLEMS
database provides continuous measures of consumption and gross fixed capital formation in ICT assets for the period at hand. However, they use a higher level of aggregation for industries, where it is not possible to translate their 13 industry classification to our
industries.
18 We classify the industries as high or low Chinese exposure depending on their change from
the first year of Period 1 to the first year of Period 2 to abstract from the effect of any ongoing simultaneous forces within the periods. An alternative approach that looks at the change
in period averages provides a different classification for three high-China and four low-China
industries. For the high ICT industries, this reclassification only concerns about 16 percent of
firms, and 12 percent of workers.
19 These results are available upon request.
153
Variance decomposition20 of log real wages reveals the changes in the components of the total variation, as well as their relative contributions to the total
wage dispersion in each period (Table 5.7 in the Appendix). The dispersion of
person effects increases across Periods 1 and 2, while the firm effect dispersion
is constant. At the same time, between Period 1 and 2, the covariance of the
two components is increasing. Since the covariance term captures the sorting
of workers across firms by their person and firm fixed components, we view
this evidence as supportive of our main objective of focusing on the sorting
patterns in the manufacturing sector, within and across its different industries.
5.4.1 Worker-firm fixed effects distribution and its change over
time
To illustrate workers sorting by type into different types of firms, we start by
mapping the joint distribution of the person and firm effects obtained from the
baseline regressions for each period as well as for the total period. We first
rank the firm and person effects, and then group them into deciles.21 Next,
for each firm and person effect decile bin intersection, we calculate the share
of worker-year matches to firms that fall into that particular bin, as a share of
total possible firm-worker-year outcomes in the period. This is represented
by a bar in the graph. Within each period, the sum of the shares adds up to
100. The ranking method allows us to focus on the relative positioning of the
firm and person effects compared to the pool of other workers and firms rather
than the absolute value of these effects. We are in other words focusing on the
shape of the joint firm-work effects distribution.
The first two panels of Figure 5.1 present the joint distribution of the workerfirm effects over the two periods (left:1996-2000, right:2001-2007) and the
difference (bottom panel) in the share of workers in each worker-firm bin between the periods. We document a significant change in the distribution of
worker-firm effects, with noticeable changes in the share of workers allocated
to particular bins. The difference graph reveals that the very lowest and the
very highest paying deciles of firms do not exhibit any change in the share of
workers. However, in the remaining ranges of firm characteristics, we observe
both positive sorting (larger masses of workers in the bins with high worker
and firm effects correlation, and particularly in the low worker-low firm effect
20 We
decompose the variance of observed wages for workers in a given sample interval as:
Var(yit ) = Var(αi ) +Var(θJ(i;t) ) +Var(xit β) + 2Cov(αi , θJ(i;t) )
+ 2Cov(xit β, θJ(i;t) ) + 2Cov(αi , xit β) +Var(εit )
21 Each
(5.3)
firm bin therefore contains 10% of all the firms in that period, and the person bins
are constructed analogously.
154
(a) Period 1: 1996-2000
(b) Period 2: 2001-2007
(c) P1 to P2 difference
Figure 5.1. Worker-firm fixed effects distributions.
bins), as well as the overall losses and gains in the employment shares of the
middle deciles of the firm effect. When interpreting the results, one should
note that these comparisons are not in absolute terms as the support sets of
the person and firm fixed effects may be different in range and dispersion.
These are merely an exploration of the changes in the shape of the distribution
relative to their respective supports.
When analyzing the between-periods transition patterns of workers and
firms across the fixed effects distribution, we do not observe significant transitions of workers across different person effect deciles, nor of firms across
different firm effect deciles, which points to the stability of the ranking of the
relative returns to individual skills and firm characteristics. 22
22 However, we observe a larger movement of firms across the firm effect deciles. This points
to a higher volatility in the estimated firm performance component in the workers wage (to the
extent that there has been no significant change in the wage setting regime). Nevertheless, when
we restrict the sample to exclude firms that have less than 5 (or 10) movers, the firm effects become more stable across periods and the joint distributions look qualitatively similar. Therefore,
we use these findings as a justification that both worker and firm effects are a reasonably stable
representation of their earning and paying potentials (i.e. their skills and productivity). These
results are available upon request. Furthermore, in order to understand to which extent the firm
fixed effects correlate with observable characteristics, we regress the estimated fixed effects on
a set of firm characteristics. In particular, we take one observation per firm and we correlate the
firm estimated fixed with the average firm’s capital intensity (log capital per worker), exporter
intensity, log profits per worker, share of high school graduates and the share of college graduates in the labor force of the firm. We find that when we control for industry indicators, all of
these variables correlate positively with the firm fixed effects, except export intensity. Since a
firm’s information on export it is only available after 2000, we performed this inspection only
for the second period in our sample (results available upon request).
155
In Figure (5.7) in the Appendix, we document the distribution of residuals
over the worker-firm effect deciles constructed, using the same methodology
as the joint distributions for the two periods, as well as the difference between
the first two periods. The residual component contains a part representing the
idiosyncratic match quality in the worker-firm pairs. The exploration of the
significance in this component and how it changes over time reveals little in
terms of evidence on the importance of the idiosyncratic match component.
There is a certain rise in the match effect for the low-paying firms with certain
worker effect groups, and for the high-paying firms that seem to exhibit a
decrease in the match effect when coupled with employees with higher worker
effects. However, there are no significant patterns of the residual behavior
in the remaining areas of the distribution lending support to the specification
additive of firm and person fixed effects.
In the Appendix (see Figure 5.8), we also provide the dissection of the distribution in Period 2 by the two education groups in the workforce (by highest
education level obtained during the working life). The results show that high
school graduates are distributed over the whole support of the worker-firm
effects and their distribution closely resembles the distribution of the overall
workforce, with some degree of positive assortative matching on both ends.
College graduates, on the other hand, concentrate in the highest paying firms
and a large share of these workers also have high person effects. This may add
additional strength to the high end positive matching explanation.
5.4.2 Sources of distribution changes
We now focus on the empirical and economic explanations of the effects that
we observe. We attempt to provide some evidence on the potential sources of
the changes in firm-worker effect distributions.
We first focus on the role of the adoption of new technologies. When analyzing changes for the group (and also within the group) of industries with
low ICT intensity from Period 1 to Period 2 (Figure 5.2), we do not observe
any significant changes in the fixed effects joint distribution. There is a slight
increase in the shares of high fixed effect workers in the high fixed effect firms,
and a reduction in the shares of low fixed effect workers in the same type of
firms.
However, when we turn to analyzing changes for the group of industries
with high ICT intensity from Period 1 to Period 2 (Figure 5.3), the outcomes
become more pronounced. We observe a large increase in the share of low
fixed effect workers in the low fixed effect firms, and a reduction in their shares
in the high fixed effect firms. At the same time, the shares of high fixed effect
workers in the high fixed effect firms increases significantly. This particular
allocation pattern (sorting) could be caused by the nature of the ICT technologies and their non-uniform adoption across firms. Although this finding may
156
(a) Period 1: 1996-2000
(b) Period 2: 2001-2007
(c) P1 to P2 difference
Figure 5.2. Low ICT Industries worker-firm fixed effects distributions.
be in line with the theoretical predictions of the skilled-biased technological
change literature, we are concerned as to whether this phenomenon occurs
uniformly across all industries with high adoption of the ICT. In particular,
our main question is whether there are other factors - i.e. the increase in Chinese import penetration (intermediate or final goods) - that contribute to these
mobility patterns when interacted with the change in technology. Although
we do not investigate these interactions further, but focus on their identification only, the background mechanisms of the particular specialization patterns,
compatible with the new technologies, may indeed only be possible under the
increased import of goods from developing countries, regarded as competitively disadvantaged in the new domestic environment.
Technology and import competition interactions
To analyze this phenomenon in more detail, we proceed with the analyses
within the group of the high ICT intensive industries. We distinguish between
two subgroups of industries, depending on their exposure to changes in competition from China.
The results reveal that the observed aggregate pattern for the high ICT intensive industries is not uniform across industries within this group, which
suggests some interaction between technology and trade. In the first group high ICT intensive industries with a high change in the Chinese import penetration - we observe a strong increase in the share of high fixed effect workers
in the high fixed effect firms, and a reduction in the shares of low fixed effect
workers in the high fixed effect firms (Figure 5.4). These industries experi157
6
6
5
5
4
4
3
3
2
2
9
1
5
0
1
2
4
5
6
Firm Bin
7
8
1
Person Bin 1
9
7
5
0
3
3
9
1
7
2
3
3
10
(a) Period 1: 1996-2000
4
5
6
Firm Bin
7
8
Person Bin 1
9
10
(b) Period 2: 2001-2007
(c) P1 to P2 difference
Figure 5.3. High ICT Industries worker-firm fixed effects distributions.
ence a stronger than average23 sorting on the high end of the firm distribution,
while there are no significant changes on the low end. The interaction of
import competition and technological change is not merely producing intensification or dampening of either one of the factors’ effects, but a qualitatively
different pattern. We view this result as an indication of a joint contribution of
the two forces in skill upgrading of the high quality firms, while leaving the
employment shares at the low end of the distribution unchanged.
In the second group of industries - high ICT intensive industries with a low
change in the Chinese import penetration - we observe an increase in the share
of low fixed effect workers in the low fixed effect firms (Figure 5.5). We also
observe smaller changes in the share of high fixed effect workers in the high
fixed effect firms, and the shares of low fixed effect workers in the high fixed
effect firms. This effect resembles the average for the ICT industries, but with
a much smaller change at the high end, and a much larger change at the low end
of the firm distribution, compared to the average. The increase in the low-skill
employment shares at the low end of the firm distribution indicates that these
types of firms, within industries with less exposure to import competition, may
have served as the shelter firms for the low-skill labor of their own as well as
of other more exposed industries. We wish to stress that the presence of the
high ICT intensity still remains important, as we do not observe the similar
"shelter" effects in non-exposed low ICT intensive industries.
In order to strengthen our results with an alternative investigation, we divide
the plane of worker and firm effects into low (bins 1 through 5) and high
(bins 6 through 10) areas, giving us 4 quadrants: Low Firm-Low Person, Low
23 We
take the magnitude of the effects that we observe for the aggregate of all high ICT
intensity industries as the average.
158
8
8
7
7
6
5
4
3
2
1
0
6
5
4
3
2
1
0
10
7
4
1
2
3
4
5 6
7
Firm Bin
8
9
7
5
1
2
Person bin 1
9
3
3
10
(a) Period 1: 1996-2000
4
5
6
Firm Bin
7
8
Person Bin 1
9
10
(b) Period 2: 2001-2007
(c) P1 to P2 difference
Figure 5.4. High ICT High China Industries worker-firm fixed effects distributions.
Firm-High Person, High Firm-Low Person, and High Firm-High Person. In
Appendix (Tables 5.11 and 5.12) we present logit regression results where we
control for a set of firm and worker observables, as well as interactions of
Chinese Import Penetration level with ICT level.
Table 5.11 shows that, compared to the "High China-High ICT" scenario,
all the other combinations of China and ICT levels become more likely to
have a Low Firm-Low Person outcome in Period 2, relative to Period 1, where
the differences were smaller. Low China-High ICT industries are most likely
to produce an Low Firm-Low Person outcome in Period 2, which fits well
with our expectations on the positive sorting on the low end for this group of
industries.
On the other hand, all industries are less likely to produce a High FirmHigh Person outcome compared to High China-High ICT industries, and again
these differences become more pronounced in Period 2 relative to Period 1.
Relatively, the least likely contribution comes from the Low China-High ICT
industries, which supports the findings presented in the distribution graphs.
5.5 Potential theoretical explanations
5.5.1 The model
In this section, we present a theoretical model, which we use to provide a
possible explanation of the observed industries dynamics and labor market
outcomes. We present a simple labor market matching model with both firm
and worker heterogeneity, which relies on Albrecht and Vroman (2002), but
159
5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
9
7
5
1
2
3
4
5
6
Firm Bin
7
8
3
Person Bin 1
9
9
7
5
1
2
3
10
4
5
6
7
8
Firm Bin
(a) Period 1: 1996-2000
3
Person Bin 1
9
10
(b) Period 2: 2001-2007
(c) P1 to P2 difference
Figure 5.5. High ICT Low China Industries worker-firm fixed effects distributions.
which introduces productivity differences across firms within heterogeneous
industries.
We assume that there are two types of workers that differ in their skill levels.
Both live forever and are risk neutral. We normalize the population measure
to 1 and assume that a fraction p of the population has low skill of level s1 ,
while a fraction (1 − p) has a high skill level s2 .
To study the potentially heterogeneous effect of exogenous factors across
industries, we construct a two-industry model. This allows us to analyze the
changes in the affected industry, as well as the implications for the other industry and potential cross-industry reallocations.
We consider two industries k, with k = T, N, which are ex-ante identical.
There is a measure zmax of firms in each industry, where firms differ in their
productivity level, each taking up a productivity value z (which we use to
index the firms) from a uniform distribution in the range [0, zmax ]. Each firm is
represented by one job position and filled jobs break up at an exogenous rate δ.
There are two types of jobs that each firm may choose among: an unqualified
job with low skill requirements given by y1k and a qualified job that requires
a higher skill level given by y2k . When a job in sector k is filled, the resulting
output f (s, yk , zk ) is a function of the job skill requirement yk , worker’s skill s
and firm productivity zk , and is given by
y zk
f (s, yk , zk ) = k
0
if s ≥ yk
if s < yk .
(5.4)
Skill requirement thus presents the skill input or productivity of the worker
hired for the job and cannot be higher than the worker’s own skill level. If
160
producing, firms pay their worker a wage w(s, yk , zk ) and also incur a fixed
cost c(yk ). We assume that the same fixed cost is also incurred when the job
is vacant and that it is higher for the qualified jobs (i.e. c(y1k ) < c(y2k )). Firms
choose the job skill requirements to maximize the value of the vacancy, and
will thus require y1k = s1 and y2k = s2 for the two types of jobs, respectively.
The labor market is not segmented and open jobs and unemployed workers
meet randomly. The number of meetings is determined by a matching function
m(u, v), where u represents the unemployment rate and v stands for vacancies.
Matching function exhibits constant returns to scale and can be expressed as
m(θ)u where θ = uv stands for the labor market tightness.24 Given that, lowskill workers meet vacancies at the effective rate φm(θ) where φ stands for the
share of vacancies that accept the low-skill worker, while high-skill workers
meet the available vacancies at the rate m(θ). We define φk to be the share of ksector’s unqualified vacancies in the total number of vacancies in the economy.
Likewise, unqualified vacancies meet unemployed workers at the rate m(θ)
θ ,
m(θ)
while this rate is (1 − γ) θ for qualified vacancies, with (1 − γ) representing
the share of high-skill workers in the pool of unemployment.
Matches between vacancies and unemployed workers are formed whenever
the total surplus created by the match is non-negative. Denoting the value of
unemployment for a worker of type s by U(s), the value of k-sector’s vacancy
of type yk for the firm with productivity zk by V (yk , zk ), the value of employment for a worker of type s at job yk in firm zk by N(s, yk , zk ) and the value of
filled job yk with worker s for a firm zk by J(s, yk , zk ), we can write the above
condition as
N(s, yk , zk ) + J(s, yk , zk ) ≥ U(s) +V (yk , zk ).
(5.5)
We define the expressions for the value functions below. r represents the interest rate (common for workers and firms), δ the exogenous match dissolution
rate and b the unemployment benefit.
24 We
assume m (θ) > 0 and limθ→0 m(θ) = 0, as well as limθ→0
m(θ)
θ
= ∞.
161
rN(s, yk , zk ) = w(s, yk , zk ) + δ[U(s) − N(s, yk , zk )]
(5.6)
rJ(s, yk , zk ) = f (s, yk , zk ) − w(s, yk , zk ) − c + δV (yk , zk )
(5.7)
1
1 1
1
rU(s ) = b + φN m(θ)[N̄(s , yN , zN ) −U(s )]
+ φT m(θ)[N̄(s1 , y1T , zT ) −U(s1 )]
(5.8)
rU(s2 ) = b + m(θ)[φN max{N̄(s2 , y1N , zN ) −U(s2 }, 0)
vN
+ ( − φN )(N̄(s2 , y2N , zN ) −U(s2 ))]
v
+ m(θ)[φT max{N̄(s2 , y1T , zT ) −U(s2 }, 0)
vT
(5.9)
+ ( − φT )(N̄(s2 , y2T , zT ) −U(s2 ))]
v
m(θ)
rV (y1k , zk ) = −c1 +
[γ(J(s1 , y1k , zk ) −V (y1k , zk ))
θ
(5.10)
+ (1 − γ)max{J(s2 , y1k , zk ) −V (y1k , zk ), 0}]
m(θ)
rV (y2k , zk ) = −c2 +
(1 − γ)[J(s2 , y2 , z) −V (y2 , z)] (5.11)
θ
The max operator in the value of unemployment for a high-skill worker and the
value of unqualified vacancy stands for the choice of this worker and firm with
this type of job of forming the match depending on their respective surpluses.
N̄(s, yk , zk ) stands for the expected value of employment for the worker of a
certain skill value and is a function of the expected (average) productivity of
the firm that the worker may be matched to. We will focus on the equilibria in
which the parameters of the model are such that these matches are profitable
and the high-skill workers accept the unqualified jobs.
Substituting the value functions into (5.5), the match is formed if and only
if
(5.12)
f (s, yk , zk ) − ck ≥ r(U(s) +V (yk , zk )).
The wages for each sector, job type, firm and worker type are determined by
Nash bargaining condition
N(s, yk , zk ) −U(s) = β[N(s, yk , zk ) + J(s, yk , zk ) −U(s) −V (yk , zk )], (5.13)
with β as the worker’s share of surplus, which yields the wage expression as
w(s1,2 , yk , zk ) = β( f (s1,2 , yk , zk ) − ck − rV (yk , zk )) + (1 − β)rU(s1,2 ).
In the steady-state equilibrium, the flows into and out of unemployment are
equal for each type of worker, respectively,
δ(p − γu) = φm(θ)γu
δ((1 − p) − (1 − γ)u) = m(θ)(1 − γ)u,
162
(5.14)
and the flows into and out of vacancy pools for each type of jobs are equal. Furthermore, the vacancy dynamics condition holds in each industry (i.e. flows
into and out of vacancy pools for each type of jobs within an industry are
equal).
To define the flow conditions for vacancy pools, we need to define the remaining two steady state conditions in each industry for the cutoff productivity levels that determine the k-industry’s selection of firms into exit, firms that
open only unqualified jobs, and, finally, the firms that open the qualified jobs
in equilibrium. Provided that the value of unqualified vacancy is larger than
the value of qualified vacancy for lower z firms, the marginal exiting firm z1k
in sector k is defined as the one for which the value of opening the unqualified
vacancy equals zero,
(5.15)
V (y1k , z1k ) = 0.
For higher values of productivity, there exists a firm z2k for which the value
of opening an unqualified vacancy is equal to the value of opening a qualified
vacancy (i.e. it is indifferent between the two types of vacancies),
V (y2k , z2k ) = V (y1k , z2k ).
(5.16)
This condition then defines the second productivity cutoff, which together with
the exit cutoff productivity determines the firms partitioning in each sector.
We solve for the nine equilibrium variables: unemployment rate u, labor
market tightness θ, industry share of unqualified vacancies φk , share of lowskill workers in unemployment pool γ, industry exit cutoff productivity z1k and
the industry job-type cutoff productivity z2k .
The effect of an increase in Chinese import penetration
Following our empirical strategy, we analyze the effect of the change in Chinese import penetration within the group of ICT intensive industries, which
we characterize by skill being important (i.e. ss21 is high). Employing our twoindustry framework, we focus on the two ex-ante identical industries where
we expose one industry to an increase in the import competition (non-exposed
industry denoted by N and trade exposed industry denoted by T ). We assume that a decrease in the productivity of the unqualified jobs in industry T ,
(y1T ), may be used to represent the change brought about by a stronger Chinese
presence in the industry market, which substitutes the local unqualified jobs25 ,
while it leaves the productivity of the qualified jobs unchanged. The results
of the numerical exercise are presented in the next section and we discuss the
relation of the model’s prediction to the empirical findings on the changes in
the labor distribution within the high ICT intensity industries group.
25 In
other words, lowers their productivity, rendering these types of jobs less valuable
163
5.5.2 Numerical analysis
Model parameters
The purpose of the numerical exercise is not to present an attempt to match the
data moments, but is rather used to provide the theoretical framework underlying the effects that we observe in the data. In that sense, we do not employ a
formal full model calibration, but set most of the model parameters based on
their empirical counterparts and calibrate the remaining ones to match a few
aggregate data moments.
Based on the empirical facts, the interest rate (r) is set to 0.05 and the share
of workers with low skills in the total population (p) is set to 0.6. Following
Albrecht and Vroman (2002), we set the following parameters: β = 0.5 (workers bargaining power), δ = 0.2 (job separation rate) and b = 0.1 (unemployment benefits). We assume a matching function of the form m(θ) = 2 ∗ θ0.9 .
The highest value of firm productivity in both sectors is set to zmax = 2.5.
2
2
We calibrate the relative skill ss1 and the relative vacancy cost cc1 in order to
match the labor market tightness of 0.6-0.9 (Hagedorn and Manovskii (2008)
use the average market tightness of 0.634) and the unemployment rate of 8%.26
2
2
The calibration yields ss1 = 2 and cc1 = 1.25. To represent high ICT intensive
2
industries with higher than average relative skill ss1 , we then set s2 = 2.0 and
s1 = 0.86. We compare the results of that specification to the results for low
2
ICT intensive industries with a lower than average ss1 (we will use s2 = 1.6 and
s1 = 0.86). Finally, c2 is set to 0.5 (Hagedorn and Manovskii (2008) calculate
the average vacancy cost of 0.584) and c1 = 0.4.
Numerical results
As noted above, to represent the effect of an increase in imports sourced from
low-skill intensive countries, we employ the two-industry framework (two
high ICT intensive industries, N and T ) in which only one industry is exposed to an increase in the Chinese import penetration. This is represented by
a decrease in the productivity of unqualified jobs in industry T below the productivity of the low-skill labor (i.e. y1T falls below s1 ). Figure (5.6) illustrates
the effects on each industry’s equilibrium variables and on the wages different
worker types earn at both types of jobs.
26 With the applied parameters, we obtain a higher unemployment rate of around 17% depending on the specification. We regard this as acceptable given that the model does not allow
for the reallocation of workers outside of the manufacturing sector that is documented in the
data.
164
0.855
Labor market tightness
0.21
Unemployment rate
Low−skill share in unemployment
0.805
gamma
0.208
u
theta
0.85
0.845
0.84
0.206
0.8
0.795
0.835
0.204
2.4
z low
z high
2.05
1.95
2.1
0.85 0.84 0.83 0.82
y1,T
0.6
0.2
0.85 0.84 0.83 0.82
y1,T
0.6
0.4
0.2
0
0.85 0.84 0.83 0.82
y1,T
0.85 0.84 0.83 0.82
y1,T
Av.low−skill wage, N and T (−−) Av.high−skill wage at qual.−jobs, N and T (−−)
2.3
0.59
0.6
W s1
qualified jobs
Share of qual.−jobs in N and T (−−)
0.8
0
2.3
2.2
0.4
0.85 0.84 0.83 0.82
y1,T
Qualified job treshold, N and T (−−) Sh of unq.−jobs with low−skill in N and T(−−)
2.5
0.8
Exit treshold, N and T (−−)
2
0.79
0.85 0.84 0.83 0.82
y1,T
unqualified jobs
2.1
0.85 0.84 0.83 0.82
y1,T
2.25
W s2 at y2
0.83
0.58
0.57
2.2
2.15
0.56
2.1
0.55
2.05
0.85 0.84 0.83 0.82
y1,T
0.85 0.84 0.83 0.82
y1,T
Figure 5.6. Effect of a decrease in the productivity of unqualified jobs in sector T, y1T
A decline in the productivity of unqualified jobs in industry T results in an
increase in the exit cutoff productivity (z1T ) as only more productive firms now
find it optimal to open vacancies. The wage of both types of workers in the
unqualified jobs falls and unemployment goes up. An initially lower tightness
makes the qualified job vacancies more valuable, which pushes the job-type
cutoff productivity z2T down. The wages of high-skill workers in these jobs
rise. In industry N, the results show opposite movements. A higher share of
low-skill workers in the unemployed pool and a reduction in tightness makes
unqualified jobs more profitable, and thus z1N falls while z2N increases. The
movements in the cutoff productivities in turn affect the employment shares
across skills and job types. Wages for unqualified jobs experience a stronger
decline compared to industry T , but as opposed to industry T , the wages for
qualified jobs decline as well. The main results are summarized below:
Result 1. A decrease in productivity of the unqualified jobs in industry T
produces the following effects:
i. Productivity thresholds in two sectors are moving in the opposite directions - the range of firms opening unqualified jobs in industry T is being
reduced from both ends (z1T increases and z2T falls), while the range of
firms opening qualified jobs in industry N expands on both ends (z1N falls
while z2N increases).
165
ii. In industry N, the share of low-skill employment in low productivity
firms (unqualified jobs) in total industry employment rises, while it falls
in industry T .
iii. In industry T , the share of high skill-qualified job type employment
in total industry employment increases. The share of this employment
category decreases in industry N.
iv. In industry T , the total employment drops, while it increases in industry
N.
Table (5.1) and (5.13, in the Appendix) summarize the quantitative effect
of a 3% decrease in y1T on employment shares and wages across skills in the
two industries, and also report their empirical counterparts.
Table 5.1. Employment effect of a 3% increase in Chinese import penetration, exposed
(T) and non-exposed (N) high ICT intensive industries
y1T
(1)
(2)
(3)
(4)
es1
N
eN
es1
T
eT
es2
y2,N
eN
es2
y2,T
eT
Panel A: Model
0.550 0.306
0.454 0.204
0.343 0.115
0.860
0.847
0.834
0.550
0.633
0.705
change (% point)
15.5
-20.7
-19.1
26.4
Period 1
Period 2
0.275
0.315
Panel B: Data
0.237 0.049
0.238 0.055
0.088
0.135
change (% point)
14.0
0.1
0.6
Note: The figures from the data are constructed as follows.
es2
y2,k
ek ,
0.306
0.428
0.570
4.7
es1
k
ek ,
k = N, T , is the share
k = N, T , is the share of high type
of low type workers employed in sector k.
workers on high type jobs in sector k.
In the data, we define low (high) type workers are those workers that in each period
have an estimated individual effect on the bottom (top) 25 percent of the distribution
of effects. We define low (high) type jobs are those jobs/firms that in each period have
an estimated effect on the bottom (top) 25 percent of the distribution of effects.
Relating the model to the data, the person fixed effect from the empirics is
given by the worker skill level in the model (only two levels in the model, s1
and s2 ). In the model, the firm component in the wage is a function of the
productivity z and the choice of the job type, y1 or y2 . We focus on the type of
job (which is a function of the productivity z in equilibrium) to represent the
firm effect, so there are two values, y1 and y2 , and we focus on the employment
166
and skill shares on these two types of jobs. Alternatively, one could divide the
productivity measure into low and high values of z and take that measure as
the underlying firm effect. However, given that the model does not account
for multiple jobs per firm, we decide to focus on the job types to represent
the firm effects, as this measure captures both the intensive (unqualified vs.
qualified jobs) and extensive margins (cutoff productivity z1,2 movements) of
labor allocations.
When we are relating the moments of the model to the moments of the data,
we look at the top 25 and bottom 25 percentile of firm fixed effects in the data,
as it is not clear where the border between high and low fixed effects should
be. It is important to note that our numerical exercise is not attempting to
match the moments from the data, but to relate to the qualitative changes (i.e.
the relative movements within and across the two industries).
In both the model (Result 1) and the data, we observe an absolute and a relative27 increase in the share of high-skill employment in the high quality firms
in the industry exposed to import competition (T ). On the other hand, in the
non-exposed industry (N), we observe an absolute and a relative increase in
the share of low-skill employment in the low quality firms28 . Comparing the
exposed and non-exposed industries, the model results confirm the observed
right tail and left tail sorting in the exposed and non-exposed industries, respectively, both being in the group of high ICT intensive industries.
In the final exercise, we test whether similar effects are produced in the low
ICT intensive industries upon an increase in the import penetration. We use
the same two-industry framework (N and T ), but assume a lower ex-ante dif2
ference between the productivity of the two skills (i.e. a lower ss1 ). While the
effects are of the same nature and sign as in the high ICT intensive industries
exercise, the magnitude of the changes is substantially lower. These results
point to the interactions of the ICT technology and Chinese import penetration, as defined in our theoretical exercise.
To conclude, as a result of the trade shock within the high ICT intensive industries, the lowest productive firms in industry T will exit, while the unqualified job firms with higher productivity will upgrade their posts to qualified
jobs. Consequently, low-skill workers are pushed out of the T sector, low-skill
unemployment increases and wages of low-skill workers decrease. Most lowskill workers who leave the T industry are hired in the N industry, where they
work in unqualified jobs. In this sense, the model explains increased sorting at
the high end (upgrading) in the sector affected by increased import competition and a concentration of low-skill workers in unqualified jobs in the sector
that is not affected by the trade shock. On the other hand, in low ICT inten27 Relative to the other employment category (i.e. low-skill employment in the low quality
firms).
28 Again, relative to the other employment category (i.e. high-skill employment in the high
quality firms).
167
sive industries where the relative productivity of the qualified jobs is lower,
the described responses to a trade shock are significantly weaker.
5.6 Conclusion
In this paper, we presented the analysis of the industry labor dynamics in response to recent changes in technology and import competition, using detailed
matched worker-firm micro data on manufacturing industries in Sweden. We
focused on the worker-to-firm sorting phenomena, which we capture in the
data and which may have contributed to the rise in the wage inequality in the
1996-2007 period. We concentrated on the effects of the increase in Chinese
import penetration and the ICT adoption as potential culprits for the sorting
phenomena, and, in particular, we investigated the outcomes of the interactions of these two forces. In the group of the high ICT intensity industries, we
observe an increase in the share of low fixed effect workers in the low fixed
effect firms, and a reduction in their shares in the high fixed effects firms. At
the same time, the shares of high effect workers in high effect firms increases.
This particular allocation pattern, not observed in the low ICT intensive industries, corresponds to the theoretical predictions of the skilled-biased technological change literature, caused by the nature of the ICT technologies and
their non-uniform adoption across firms.
However, the documented pattern is not uniform across industries within
the high ICT intensity group, which points to the interaction between technology and trade. In the group of high ICT industries with a high change in the
Chinese import penetration, we observe a strong increase in the share of high
fixed effect workers in the high fixed effect firms, and a reduction in the shares
of low effect workers in the high effect firms, while there are no significant
changes on the low end of the firm distribution. To the extent that worker
and firm effects represent their skills (i.e. quality) we observe a strong skill
upgrade in the high quality firms within this industry type and no change on
the low quality end. However, in the second group (high ICT industries with
a low change in the Chinese import penetration), we observe an increase in
the share of low fixed effects workers in the low fixed effects firms (i.e. "sheltering" effects of the low-skill workers). This last finding again points to the
importance of the interactions between the two factors when explaining the
aggregate outcomes. These issues certainly deserve further attention in both
the theoretical and empirical literature.
168
References
Abowd J.M., Kramarz F., and Margolis D.N., (1999). High Wage Workers and
High Wage Firms, Econometrica, Econometric Society, 67(2), 251-334.
Abowd J.M., Kramarz F., Lengermann P., and Perez-Duarte S., (2004). Are
Good Workers Employed by Good Firms?: A Test of a Simple Assortative
Matching Model for France and the United States, CREST,mimeo.
Acemoglu D., (1999). Changes in Unemployment and Wage Inequality: An
Alternative Theory and Some Evidence, American Economic Review, 89(5),
1259-1278.
Albrecht J. and Vroman S., (2002). A Matching Model with Endogenous Skill
Requirements, International Economic Review, 43(1), 283-305.
Alvarez R. and Opazo L., (2011). Effects of Chinese Imports on Relative
Wages: Microevidence from Chile, Scandinavian Journal of Economics, Wiley Blackwell, 113(2), 342-363.
van Ark B., Inklaar R., and McGuckin R. H. (2003), ‘Changing Gear’ Productivity, ICT and Service Industries: Europe and the United States, in: Christensen, Jens F. and Maskell, Peter (eds.), The Industrial Dynamics of the New
Digital Economy, Edward Elgar, Cheltenham, 56-99.
Ashournia D., Munch J., and Nguyen D., (2012). The Impact of Chinese
Import Penetration on Danish Firms and Workers, University of Copenhagen,
mimeo.
Autor D. and Dorn D. (2013) The Growth of Low Skill Service Jobs and the
Polarization of the U.S. Labor Market, American Economic Review, 103(5),
1553-1597.
Autor D., Dorn D., and Hanson G., (2013). The China Syndrome: Local
Labor Market Effects of Import Competition in the United States, American
Economic Review, 103(6), 2121-2168.
Autor D., Dorn D., and Hanson, G., (2014). Untangling Trade and Technology: Evidence from Local Labor Markets, Economic Journal, forthcoming.
Balsvik R., Jensen S., and Salvanes K.G. (2014). Made in China, Sold in
Norway: Local Labor MArket Effects of an Import Shock, Journal of Public
Economics, forthcoming.
Bas M. and Berthou A., (2013). Does Input-Trade Liberalization Affects
Firms’ Foreign Technology Choice?, CEPII Working Paper, No 2013-11.
169
Becker G.S., (1973). A Theory of Marriage: Part I, The Journal of Political
Economy, 81(4), 813-846.
Bustos P., (2011). Trade Liberalization, Exports, and Technology Upgrading: Evidence on the Impact of MERCOSUR on Argentinian Firms, American
Economic Review, 101(1), 304-40.
Card D., Heining J., and Kline P., (2013). Workplace Heterogeneity and the
Rise of West German Wage Inequality, The Quarterly Journal of Economics,
Oxford University Press, 128(3), 967-1015.
Caselli F., (1999). Technological Revolutions, American Economic Review,
89(1), 78-102.
Costinot A., (2009). An Elementary Theory of Comparative Advantage, Econometrica, 77(4), 1165-1192.
Costinot A. and Vogel J., (2010). Matching and Inequality in the World Economy, Journal of Political Economy, 118(4), 747-786.
Davidson C., Matusz, S.J., and Shevchenko, A., (2008). Globalization and
Firm Level Adjustment with Imperfect Labor Markets, Journal of International Economics, 75(2), 295-309.
Davidson C. and Matusz S.J., (2012). A Model of Globalization and FirmWorker Matching: How Good is Good Enough?, International Review of Economics & Finance, 23, 5-15.
Davidson C., Heyman F., Matusz S., Sjöholm F., and Chun Zhu S., (2014).
Globalization and Imperfect Labor Market Sorting, Journal of International
Economics, 94(2), 177-194.
Eeckhout J. and Kircher P., (2011). Identifying Sorting–In Theory, Review of
Economic Studies, Oxford University Press, 78(3), 872-906.
Grossman G. and Maggi G., (2000). Diversity and Trade, American Economic
Review, 90, 1255-1275.
Grossman G., Helpman, E., and Kircher, P., (2014). Matching, Sorting, and
the Distributional Effects of International Trade, mimeo.
Hagedorn M., Law T.H., and Manovskii I., (2012). Identifying Equilibrium
Models of Labor Market Sorting, NBER Working Papers 18661, National Bureau of Economic Research, Inc.
170
Hagedorn M. and Manovskii I., (2008). The Cyclical Behavior of Equilibrium
Unemployment and Vacancies Revisited, American Economic Review, 98(4),
1692-1706.
Håkanson C., Lindqvist E., and Vlachos J., (2013). Sorting by Skill over Time:
Evidence from Sweden 1986-2008, IIES, mimeo
Kugler M. and Verhoogen, E., (2011). Prices, Plant Size, and Product Quality,
Review of Economic Studies, 79(1), 307-339.
Lise J., Meghir C., and Robin J.-M., (2013). Mismatch, Sorting and Wage
Dynamics, NBER Working Paper No. 18719.
Low H., Meghir C., and Pistaferri L., (2010). Wage Risk and Employment
Risk over the Life Cycle, American Economic Review, American Economic
Association, 100(4), 1432-67.
Meghir C. and Pistaferri L., (2004). Income Variance Dynamics and Heterogeneity, Econometrica, Econometric Society, vol. 72(1), pages 1-32, 01.
Mortensen D. and Pissarides C.A., (1994). Job Creation and Job Destruction
in the Theory of Unemployment, The Review of Economic Studies, 61(3), 397415.
Yeaple S.R., (2005). A Simple Model of Firm Heterogeneity, International
Trade, and Wages, Journal of International Economics, 65(1), 1-20.
171
Appendix
Table 5.2. Information and Communication Technology Classifications
Van Ark et al.(2003) Classifications
Our own ICT Classifications
ICT Producing Industries
30-OfficeMachinery&Comp
313-Insulated Wire
32-RadioTvCommunication
331-3-Scientific Instruments
High ICT Industries
18-Wearing Apparel
22-Publishing&Media
29-Machinery&Equipment
30-OfficeMachinery&Comp
31-Electric Machinery
32-RadioTvCommunication
33-Scientific Instruments
35-Other Transport
36-Furniture
Intensive ICT-using Industries
18-Wearing Apparel
22-Publishing and Media
29-Machinery and Equipment
31(ex313)-Electric Machinery
334-5-Other Instruments
35-Other Transport
36-Furniture
37-Recycling
Less Intensive ICT-using Industries
15-Food
16-Tobacco
17-Textiles
19-Leather
20-Wood
21-Pulp and Paper
23-Coke&Petrol
24-Chemicals
25-Rubber&Plastic
26-Non-metallicMineral
27-BasicMetals
28-Metal
34-MotorVehicle
Low ICT Industries
15-Food
16-Tobacco
17-Textiles
19-Leather
20-Wood
21-Pulp and Paper
23-Coke&Petrol
24-Chemicals
25-Rubber&Plastic
26-Non-metallicMineral
27-BasicMetals
28-Metal
34-MotorVehicle
Notes: For our own classification, we keep the Less intensive ICT industries from van Ark et
al.(2003) as Low ICT Industries, and group the rest together into High ICT Industries. Since
Recycling is not an industry we can identify with Chinese imports at the product level from
Comtrade, we drop it from our ICT grouping.
172
Industry classifications
Table 5.3. Matching UN Comtrade SITC Codes to Swedish Industries
SITC
SITC Name
SNI
SNI Name
1
4
6
7
9
11
12
65
15
Manufacture of food products and beverages
16
17
Manufacture of tobacco products
Manufacture of textiles
84
85
61
Meat and meat preparations
Cereals and cereal preparations
Sugars, Sugar preparations and honey
Coffee, tea, cocoa, spices, and manufactures thereof
Miscellaneous edible products and preparations
Beverages
Tobacco and tobacco manufactures
Textile yarn, fabrics, made-up articles, n.e.s., and related
products
Articles of apparel and clothing accessories
Footwear
Leather, leather manufactures, n.e.s., and dressed furskins
18
Manufacture of wearing apparel; dressing and dyeing of fur
19
63
Cork and wood manufactures (excluding furniture)
20
64
Paper, paperboard and articles of paper pulp, of paper or of
paperboard
Printed matter
Musical instruments and parts and accessories thereof;
records, tapes and other sound or similar recordings
Coke and semi-coke (including char) of coal, of lignite or
of peat, whether or not agglomerated; retort carbon
Petroleum, petroleum products and related materials
21
Tanning and dressing of leather; manufacture of luggage,
handbags, saddlery, harness and footwear
Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting
materials
Manufacture of pulp, paper and paper products
22
Publishing, printing and reproduction of recorded media
23
Manufacture of coke, refined petroleum products and nuclear fuel
Chemicals and related products, n.e.s.
24
Manufacture of chemicals and chemical products
Rubber manufactures, n.e.s.
Plastics in primary forms
Plastics in non-primary forms
Non-metallic mineral manufactures, n.e.s.
Iron and steel
Non-ferrous metals
Manufactures of metals, n.e.s.
25
Manufacture of rubber and plastic products
26
27
Manufacture of other non-metallic mineral products
Manufacture of basic metals
28
General industrial machinery and equipment, n.e.s., and
machine parts, n.e.s.
Office machines and automatic data-processing machines
Electrical machinery, apparatus and appliances, n.e.s., and
electrical parts thereof
Telecommunications and sound-recording and reproducing
apparatus and equipment
Photographic apparatus, equipment and supplies and optical
goods, n.e.s.; watches and clocks
Instruments and appliances, n.e.s., for medical, surgical,
dental or veterinary purposes
Road vehicles (including air-cushion vehicles)
Other transport equipment
Furniture, and parts thereof; bedding, mattresses, mattress
supports, cushions and similar stuffed furnishings
—
29
Manufacture of fabricated metal products, except machinery and equipment
Manufacture of machinery and equipment n.e.c.
30
31
Manufacture of office machinery and computers
Manufacture of electrical machinery and apparatus n.e.c.
32
Manufacture of radio, television and communication equipment and apparatus
Manufacture of medical, precision and optical instruments,
watches and clocks
892
898
325
33
5 excl
57&58
62
57
58
66
67
68
69
74
75
77
76
88
872
78
79
82
—
33
34
35
36
Manufacture of motor vehicles, trailers and semi-trailers
Manufacture of other transport equipment
Manufacture of furniture; manufacturing n.e.c.
37
Recycling
Notes: Recycling is not an industry that product level trade information from UN Comtrade allows us to
identify.
173
174
199.9
221.9
259.0
267.3
273.4
351.8
404.1
492.5
496.6
901.7
0.04
6.99
0.27
0.24
0.25
0.09
1.16
0.05
1.77
0.35
1.60
Share
(P1,%)
0.04
8.12
0.46
0.43
0.53
0.19
2.59
0.12
4.84
0.99
4.57
Share
(P2,%)
-3.6
16.2
67.9
80.2
108.8
110.6
123.9
161.5
173.6
182.9
185.1
Change
%
Note: The imports data is from UN Comtrade at the product level. The Authors have performed a match between imported products and Swedish industry codes.
3.17
4.17
0.26
0.25
2.94
0.68
0.57
1.88
0.71
6.53
35-OtherTransport
18-WearingApparel
23-Coke&Petrol
22-Publishing&Media
24-Chemicals
15-Food
17-Textiles
34-MotorVehicle
36-Furniture
33-Medical&Optic
20-Wood
1.06
1.29
0.07
0.07
0.79
0.15
0.11
0.32
0.12
0.65
19-Leather
28-Metal
25-Rubber&Plastic
21-Pulp&Paper
26-Non-metallicMineral
29-Machinery&Equipment
27-BasicMetals
31-ElectricMachinery
30-OfficeMachinery&Comp
32-RadioTvCommunic
Change
%
Low China Industries - N sector
Share
(P2,%)
High China Industries - T sector
Share
(P1,%)
Table 5.4. Share of Chinese Imports, and Changes over the Periods, by Industries
Table 5.5. Data Description
Firm Data
Total Wages
Sum of personnel costs for the year (Summa personalkostnader)
Total Sales
Sum of revenues for the year (Nettomsättning)
Profit
Reported profit for the year (Redovisat Resultat)
Firm age
Calculated from years active in the dataset
Capital (K)
Sum of the following reported tangible assets for the year:
Land and Buildings
Machinery and Equipment
Ongoing Construction and Advance payments for tangible fixed
assets
Total Employees (N)
Total employees (Antal Anställda)
Capital Intensity
Calculated as K/N
Industry Classification
Industry Codes are reported in four different systems (SNI1969, 1992,
2002, 2007) which all have been converted to SNI2002 at the 5-digit
and 2-digit level
Business Register
Legal Form
Classification by type of legal entity
Controlling Ownership
Standard Classification by ownership control
Municipality
Municipality where the firm (headquarters) is registered. Municipality
of the main plant is only available from 2000 onwards.
Employee Data
Annual Wage
Taxed wage income (Kontant Bruttolön)
Age
As reported
Gender
As reported
Level of Highest Education
Under the old SUN code, the following categories:
Pre High School
Some High School without a diploma
High School diploma
Less than 2 years of University
More than 2 years of University, includes those with diploma
Postgraduate Studies
Targeted Field of Education
Targeted diploma subject
Trade Data
Imports and Exports
Indicator available for each year when there is foreign trade
Value
Reported value of foreign trade
Country
Code for the destination or source country
Notes: Firm Data source is Account Statistics (FEK). Data for 1980-1996 are for a sample of
companies. Data comes with a 2 year lag. Only non-imputed companies included. Business
Register data is sourced from the Business Register Database (Företagsregistret). Data available from 1980 onwards. Employee Data source is Register Based Labor Statistics (RAMS).
Data available from 1985 onwards. Each individual is linked to a firm, and a plant where applicable. Trade Data source is the Foreign Trade Database (Utrikeshandel). Data available from
2000 onwards. Partial data available between 1997-2000. For intra-EU trade, minimum of SEK
4.5 million (≈ USD 610,000) required to register as importing or exporting.
175
Table 5.6. Summary Statistics of The Manufacturing Industries, and Largest Mobility
Groups
No of
Firms
Period 1: 1996-2000
9838
Total Population
No of
Log Real
People
Wage
769461
12.407
(0.358)
Percent of Total (%)
Period 2: 2001-2007
Percent of Total (%)
9782
831740
12.511
(0.340)
Largest Mobility Group - Firms
No of
No of
Log Real
Firms
People
Wage
8300
752618
84.37
97.81
8769
820603
89.64
98.66
12.410
(0.359)
12.513
(0.340)
Note: Wage standard deviation in parentheses. The Population is restricted to agents born after
1920 and before 1992 holding a full time job (defined as earning at least SEK 120,000 in 2010
values annually) in manufacturing firms. Percent of Total indicates what percentage of the total
population of firms, or people the mobility group captures.
176
Variance decomposition
Table 5.7. Log Real Wage Variance Decomposition.
Variance of Log Annual Real Wages
Period 1
Period 2
1996-2000
Variance
Share
2001-2007
Variance
Share
0.120
0.138
Breaking down the variance:
Variance of Person Effect
0.083
69.26
0.091
66.46
Variance of Firm Effect
0.006
4.87
.006
4.17
Variance of Covariates
.015
12.40
0.012
8.86
Variance of the Residual
.014
11.48
0.018
12.63
2xCovariance of Person and Firm Effects
-0.002
-1.46
0.001
0.62
2xCovariance of (Person+Firm,Covariates)
0.004
3.45
0.012
8.27
Comparing Models:
Double Fixed Effect Model R2:
Double Fixed Effect Model AdjR2:
Std Dev of Person FE:
Std Dev of Firm FE:
0.8852
0.8439
0.28865
0.0765384
0.8737
0.8365
0.3018947
0.0762228
Match Fixed Effect Model R2:
Match Fixed Effect Model AdjR2:
Std Dev of Match FE:
0.9001
0.8541
2.553796
0.8881
0.8462
0.4695153
Note: The data is restricted to the total period mobility group employees with single jobs born
between 1920 and 1991, earning a work salary of at least SEK 120,000 a year.
177
Fixed effects joint distributions
0,006
0,005
0,004
0,003
0,002
0,001
0
Ͳ0,001
Ͳ0,002
Ͳ0,003
0,006
1
2
3
4
5
6
7
0,005
0,004
0,003
0,002
0,001
0
Ͳ0,001
Ͳ0,002
Ͳ0,003
9
7
5
3
PersonBin1
8
9
10
1
2
3
4
5
6
7
9
7
5
3
PersonBin1
8
9
10
Firm Bin
FirmBin
(a) Period 1: 1996-2000
0,006
0,005
0,004
0,003
0,002
0,001
0
Ͳ0,001
Ͳ0,002
Ͳ0,003
1
(b) Period 2: 2001-2007
2
3
4
5
6
7
9
7
5
3
PersonBin1
8
9
10
Firm Bin
(c) P1 to P2 difference
Figure 5.7. Match-effect distribution.
(a) High School graduates
(b) College graduates
Figure 5.8. Worker-firm effects distribution by education, 2001-2007.
178
179
831740
12.511
(0.340)
8300
89.64
92.69
13203
88.39
12195
98.66
820603
97.81
752618
12.513
(0.340)
12.410
(0.359)
89.14
8720
84.13
8277
85.46
12174
80.29
11078
98.28
817411
97.26
748385
12.513
(0.340)
12.410
(0.359)
Largest Mobility Group - Plants
No of
No of
No of
Log Real
Firms
Plants
People
Wage
Note: Wage standard deviation in parentheses. The Population is restricted to agents born after 1920 and before 1992 holding a full time job (defined as earning
at least SEK 120,000 in 2010 values annually) in manufacturing firms. We have dropped single person plants from our Largest Mobility Group by Plants,
amounting to 594, 525, and 387 plants for periods 1-3 and 830 plants in the total period. We also have very rare cases where the plant assignment for the
employee is unknown. We have also dropped these people from our last block, amounting to 7, 2, 7, and 16 employees respectively for periods 1-3 and the total
period.
Percent of Total(%)
14245
12.407
(0.358)
8769
9782
769461
Largest Mobility Group - Firms
No of
No of
No of
Log Real
Firms
Plants
People
Wage
Period 2: 2001-2007
13797
Log Real
Wage
84.37
9838
Total Population
No of
No of
Plants
People
Percent of Total (%)
Period 1: 1996-2000
No of
Firms
Plant level analysis
Table 5.8. Summary Statistics of The Manufacturing Industries, and Largest Mobility Groups
Table 5.9. Distribution of Plants, by Firms
Period 1 : 1996-2000
No of
Percent
Firms
(%)
Number of Plants
in the Same Firm
1
2
3
4
5
6
7
8
9
10
11+
8160
685
413
175
96
75
54
28
26
26
100
82.94
6.96
4.20
1.78
0.98
0.76
0.55
0.28
0.26
0.26
1.02
Period 2: 2001-2007
No of
Percent
Firms
(%)
8044
707
410
189
110
71
57
34
29
22
109
82.23
7.23
4.19
1.93
1.12
0.73
0.58
0.35
0.30
0.22
1.11
Note: Wage standard deviation in parentheses. The Population is restricted to agents born after 1920 and
before 1992 holding a full time job (defined as earning at least SEK 120,000 in 2010 values annually) in
manufacturing firms. We have dropped single person plants from our Largest Mobility Group by Plants,
amounting to 594, 525, and 387 plants for periods 1-3 and 830 plants in the total period. We also have very
rare cases where the plant assignment for the employee is unknown. We have also dropped these people
from our last block, amounting to 7, 2, 7, and 16 employees respectively for periods 1-3 and the total period.
180
3
3
2,5
2,5
2
2
1,5
1,5
1
1
9
0,5
7
0
1
2
3
4
5
6
FirmBin
7
8
0
3
PersonBin1
9
9
7
5
3
PersonBin1
0,5
5
1
2
3
10
4
5
6
FirmBin
(a) Period 1: 1996-2000
7
8
9
10
(b) Period 2: 2001-2007
1
0,5
9
7
5
3
PersonBin1
0
1
Ͳ0,5
2
3
4
5
6
7
8
9
10
Ͳ1
FirmBin
(c) P1 to P2 difference
Figure 5.9. Worker-firm effects distribution, plant level analysis.
181
182
2808
1766
1740
8300
Low ICT High China
High ICT Low China
High ICT High China
Total
749446
183527
123710
201431
240778
749446
181824
125361
198689
243572
8731
1802
1752
3084
2093
805761
203342
133963
204172
264284
805761
196169
143325
202801
263466
Period 2 : 2001-2007
No of People No of People
No of Firms
First Job
Last Job
Note: The data is restricted to the total period mobility group employees born between 1920 and 1991, earning a work salary of at least SEK 120,000 a year and firms of at least 5
employees each year they are active in the database. Low and High Chinese penetration industries are constructed using our Chinese penetration measure between Periods 1 and
2. Since we identify high and low Chinese Import Penetration as the change in Chinese imports over periods, we group Period 1 industries into the same low and high Chinese
Import Penetration groups as those in Period 2 for comparison. ICT classifications are fixed across all time periods. As workers may switch jobs across sectors within the period,
we present the population breakdown across industry types of their first job and also their last job within the period in the given ICT and China interaction sector.
1686
Low ICT Low China
Period 1: 1996-2000
No of People
No of People
No of Firms
First Job
Last Job
Across-industry distribution differences
Table 5.10. Employment by Industry Groups, All Periods
Distributions in groups of industries, Period 1, 1996-2000
4
4
3
3
2
2
10
1
10
1
7
7
0
0
4
1
2
3
4
5
6
7
8
FirmBin
1
PersonBin1
9
4
2
3
4
10
5
6
7
PersonBin1
8
9
FirmBin
(a) Low ICT Low China
(b) Low ICT High China
4
4
3
3
2
10
2
10
1
7
0
4
1
2
3
4
5
6
Firm Bin
7
8
PersonBin1
9
10
(c) High ICT Low China
10
1
7
0
4
1
2
3
4
5 6
7
FirmBin
8
Personbin1
9
10
(d) High ICT High China
Figure 5.10. Industries with low (top) / high (bottom) ICT intensity and low (left) /
high (right) change in Chinese import penetration, Period 1
183
Table 5.11. Logit on the outcome of being in the Low Firm-Low Person quadrant
VARIABLES
Period 1
Age
Firm Tenure
(1)
LF-LP
(2)
LF-LP
(3)
LF-LP
(4)
LF-LP
0.0298***
(0.000377)
-0.0609***
(0.00339)
0.0296***
(0.000376)
-0.0637***
(0.00339)
0.0298***
(0.000377)
-0.0621***
(0.00339)
0.0297***
(0.000377)
-0.0617***
(0.00339)
0.0249**
(0.0126)
0.236***
(0.0128)
-0.0529***
(0.0125)
0.149***
(0.00866)
-0.127***
(0.00951)
3.678***
(0.0843)
3.580***
(0.0850)
LowChina*LowICT
LowChina*HighICT
HighChina*LowICT
LowChina in Period 1
0.142***
(0.00865)
LowICT in Period 1
Constant
3.748***
(0.0841)
-0.116***
(0.00949)
3.725***
(0.0843)
Observations
474,878
474,878
474,878
474,878
0.00537***
(0.000327)
-0.102***
(0.00205)
0.00466***
(0.000327)
-0.104***
(0.00205)
0.00492***
(0.000327)
-0.104***
(0.00205)
0.00421***
(0.000329)
-0.106***
(0.00206)
0.0554***
(0.0117)
0.598***
(0.0120)
0.140***
(0.0115)
2.558***
(0.0517)
-0.195***
(0.00832)
2.637***
(0.0513)
0.194***
(0.00771)
-0.201***
(0.00833)
2.391***
(0.0523)
1.968***
(0.0532)
714,857
714,857
714,857
714,857
Period 2
Age
Firm Tenure
LowChina*LowICT
LowChina*HighICT
HighChina*LowICT
LowChina in Period 1
0.187***
(0.00770)
LowICT in Period 1
Constant
Observations
*** p<0.01, ** p<0.05, * p<0.10.
Note: Regressions include year dummies and control for the individual’s gender and education level as well as capital per worker, profit per
worker, share of high school and college graduates on the firm side. The sample is restricted to manufacturing workers born after 1920 and
before 1992. Period 2 workers are restricted to those who were present in Period 1. Reference interaction group is High China High ICT.
In Period 2, compared to Period 1 all the other combinations of China and ICT levels
become more likely to have a LFLP outcome relative to High China-High ICT industries. Low China-High ICT industries are most likely to produce an LFLP outcome in
Period 2, fitting with our expectations on the positive sorting on the low end for this
group of industries.
184
Table 5.12. Logit on the outcome of being in the High Firm-High Person quadrant
VARIABLES
Period 1
Age
Firm Tenure
(1)
HF-HP
(2)
HF-HP
(3)
HF-HP
(4)
HF-HP
-0.0414***
(0.000321)
0.0493***
(0.00254)
-0.0411***
(0.000321)
0.0508***
(0.00254)
-0.0414***
(0.000321)
0.0497***
(0.00255)
-0.0413***
(0.000322)
0.0496***
(0.00255)
-0.112***
(0.00942)
-0.181***
(0.0105)
0.0226**
(0.00974)
-2.289***
(0.0478)
0.0210***
(0.00754)
-2.241***
(0.0486)
-0.153***
(0.00658)
0.0435***
(0.00760)
-2.236***
(0.0487)
-2.208***
(0.0494)
620,465
620,465
620,465
620,465
-0.00257***
(0.000260)
0.105***
(0.00155)
-0.00212***
(0.000260)
0.104***
(0.00155)
-0.00248***
(0.000260)
0.105***
(0.00155)
-0.00192***
(0.000261)
0.106***
(0.00156)
-0.234***
(0.00842)
-0.544***
(0.00924)
-0.199***
(0.00887)
-4.600***
(0.0430)
0.00613
(0.00678)
-4.646***
(0.0432)
-0.242***
(0.00586)
0.0408***
(0.00685)
-4.558***
(0.0435)
-4.336***
(0.0439)
714,857
714,857
714,857
714,857
LowChina*LowICT
LowChina*HighICT
HighChina*HighICT
LowChina in Period 1
-0.148***
(0.00652)
LowICT in Period 1
Constant
Observations
Period 2
Age
Firm Tenure
LowChina*LowICT
LowChina*HighICT
HighChina*HighICT
LowChina in Period 1
-0.237***
(0.00581)
LowICT in Period 1
Constant
Observations
*** p<0.01, ** p<0.05, * p<0.10.
Note: Regressions include year dummies and control for the individual’s gender and education level as well as capital per worker, profit per
worker, share of high school and college graduates on the firm side. The sample is restricted to manufacturing workers born after 1920 and
before 1992. Period 2 workers are restricted to those who were present in Period 1. Reference interaction group is High China High ICT.
In Period 2, all industries are less likely to produce a HFHP outcome compared to
High China-High ICT industries. Relatively, the least likely contribution comes from
the Low China-High ICT industries, supporting the predictions of theory.
185
Numerical analysis, wage effect of import competition
Table 5.13. Wage effect of a 3% increase in Chinese import penetration, exposed (T)
and non-exposed (N) high ICT intensive industries
186
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
y1T
ws2
y1,N
ws1
N
ws2
y1,T
ws1
T
ws2
y1,N
ws2
y2,N
ws2
y1,T
ws2
y2,T
ws2
N
ws1
N
ws2
T
ws1
T
ws2
ws1
ws2
T
ws2
N
0.860
0.847
0.834
2.985
3.011
3.028
2.985
2.980
2.967
0.802
0.805
0.808
3.354
3.373
3.384
3.354
3.374
3.386
0.500
0.507
0.515
change (%)
1.4
-0.6
0.7
0.9
0.9
3
Period 1
Period 2
1.616
1.783
1.600
1.645
0.619
0.698
2.141
2.351
2.070
2.221
1.045
1.076
change (%)
10.3
2.8
12.8
9.8
7.3
3.0
Panel A: Model
0.802 3.354
0.791 3.376
0.781 3.388
-2.6
1.0
Panel B: Data
0.640 2.119
0.632 2.281
-1.2
7.6
Economic Studies
____________________________________________________________________
1987:1 1989:1 1989:2 1989:3 1989:4 1990:1 1990:2 1991:1 1992:1 1992:2 1992:3 12 13 14 15 16 17 Haraldson, Marty: To Care and To Cure. A linear programming approach to national health planning in developing countries. 98 pp. Chryssanthou, Nikos: The Portfolio Demand for the ECU: A Transaction Cost Approach. 42 pp. Hansson, Bengt: Construction of Swedish Capital Stocks, 1963-­‐‑87: An Application of the Hulten-­‐‑Wykoff Studies. 37 pp. Choe, Byung-­‐‑Tae: Some Notes on Utility Functions Demand and Aggregation. 39 pp. Skedinger, Per: Studies of Wage and Employment Determination in the Swedish Wood Industry. 89 pp. Gustafson, Claes-­‐‑Håkan: Inventory Investment in Manufacturing Firms. Theory and Evidence. 98 pp. Bantekas, Apostolos: The Demand for Male and Female Workers in Swedish Manufacturing. 56 pp. Lundholm, Michael: Compulsory Social Insurance. A Critical Review. 109 pp. Sundberg, Gun: The Demand for Health and Medical Care in Sweden. 58 pp. Gustavsson, Thomas: No Arbitrage Pricing and the Term Structure of Interest Rates. 47 pp. Elvander, Nils: Labour Market Relations in Sweden and Great Britain. A Com-­‐‑
parative Study of Local Wage Formation in the Private Sector during the 1980s. 43 pp. Dillén, Mats: Studies in Optimal Taxation, Stabilization, and Imperfect Competition. 1993. 143 pp. Banks, Ferdinand E.: A Modern Introduction to International Money, Banking and Finance. 1993. 303 pp. Mellander, Erik: Measuring Productivity and Inefficiency Without Quantitative Output Data. 1993. 140 pp. Ackum Agell: Susanne: Essays on Work and Pay. 1993. 116 pp. Eriksson, Claes: Essays on Growth and Distribution. 1994. 129 pp. Banks, Ferdinand E.: A Modern Introduction to International Money, Banking and Finance. 2nd version, 1994. 313 pp. 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 Apel, Mikael: Essays on Taxation and Economic Behavior. 1994. 144 pp. Dillén, Hans: Asset Prices in Open Monetary Economies. A Contingent Claims Approach. 1994. 100 pp. Jansson, Per: Essays on Empirical Macroeconomics. 1994. 146 pp. Banks, Ferdinand E.: A Modern Introduction to International Money, Banking, and Finance. 3rd version, 1995. 313 pp. Dufwenberg, Martin: On Rationality and Belief Formation in Games. 1995. 93 pp. Lindén, Johan: Job Search and Wage Bargaining. 1995. 127 pp. Shahnazarian, Hovick: Three Essays on Corporate Taxation. 1996. 112 pp. Svensson, Roger: Foreign Activities of Swedish Multinational Corporations. 1996. 166 pp. Sundberg, Gun: Essays on Health Economics. 1996. 174 pp. Sacklén, Hans: Essays on Empirical Models of Labor Supply. 1996. 168 pp. Fredriksson, Peter: Education, Migration and Active Labor Market Policy. 1997. 106 pp. Ekman, Erik: Household and Corporate Behaviour under Uncertainty. 1997. 160 pp. Stoltz, Bo: Essays on Portfolio Behavior and Asset Pricing. 1997. 122 pp. Dahlberg, Matz: Essays on Estimation Methods and Local Public Economics. 1997. 179 pp. Kolm, Ann-­‐‑Sofie: Taxation, Wage Formation, Unemployment and Welfare. 1997. 162 pp. Boije, Robert: Capitalisation, Efficiency and the Demand for Local Public Services. 1997. 148 pp. Hort, Katinka: On Price Formation and Quantity Adjustment in Swedish Housing Markets. 1997. 185 pp. Lindström, Thomas: Studies in Empirical Macroeconomics. 1998. 113 pp. Hemström, Maria: Salary Determination in Professional Labour Markets. 1998. 127 pp. Forsling, Gunnar: Utilization of Tax Allowances and Corporate Borrowing. 1998. 96 pp. Nydahl, Stefan: Essays on Stock Prices and Exchange Rates. 1998. 133 pp. Bergström, Pål: Essays on Labour Economics and Econometrics. 1998. 163 pp. 40 41 42 43 44
45
46
47
48
49
50
51
52
53
54
55 56 57 58 59 60 61 Heiborn, Marie: Essays on Demographic Factors and Housing Markets. 1998. 138 pp. Åsberg, Per: Four Essays in Housing Economics. 1998. 166 pp. Hokkanen, Jyry: Interpreting Budget Deficits and Productivity Fluctuations. 1998. 146 pp. Lunander, Anders: Bids and Values. 1999. 127 pp. Eklöf, Matias: Studies in Empirical Microeconomics. 1999. 213 pp. Johansson, Eva: Essays on Local Public Finance and Intergovernmental Grants. 1999. 156 pp. Lundin, Douglas: Studies in Empirical Public Economics. 1999. 97 pp. Hansen, Sten: Essays on Finance, Taxation and Corporate Investment. 1999. 140 pp. Widmalm, Frida: Studies in Growth and Household Allocation. 2000. 100 pp. Arslanogullari, Sebastian: Household Adjustment to Unemployment. 2000. 153 pp. Lindberg, Sara: Studies in Credit Constraints and Economic Behavior. 2000. 135 pp. Nordblom, Katarina: Essays on Fiscal Policy, Growth, and the Importance of Family Altruism. 2000. 105 pp. Andersson, Björn: Growth, Saving, and Demography. 2000. 99 pp. Åslund, Olof: Health, Immigration, and Settlement Policies. 2000. 224 pp. Bali Swain, Ranjula: Demand, Segmentation and Rationing in the Rural Credit Markets of Puri. 2001. 160 pp. Löfqvist, Richard: Tax Avoidance, Dividend Signaling and Shareholder Taxation in an Open Economy. 2001. 145 pp. Vejsiu, Altin: Essays on Labor Market Dynamics. 2001. 209 pp. Zetterström, Erik: Residential Mobility and Tenure Choice in the Swedish Housing Market. 2001. 125 pp. Grahn, Sofia: Topics in Cooperative Game Theory. 2001. 106 pp. Laséen, Stefan: Macroeconomic Fluctuations and Microeconomic Adjustments: Wages, Capital, and Labor Market Policy. 2001. 142 pp. Arnek, Magnus: Empirical Essays on Procurement and Regulation. 2002. 155 pp. Jordahl, Henrik: Essays on Voting Behavior, Labor Market Policy, and Taxation. 2002. 172 pp. 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 Lindhe, Tobias: Corporate Tax Integration and the Cost of Capital. 2002. 102 pp. Hallberg, Daniel: Essays on Household Behavior and Time-­‐‑Use. 2002. 170 pp. Larsson, Laura: Evaluating Social Programs: Active Labor Market Policies and Social Insurance. 2002. 126 pp. Bergvall, Anders: Essays on Exchange Rates and Macroeconomic Stability. 2002. 122 pp. Nordström Skans, Oskar: Labour Market Effects of Working Time Reductions and Demographic Changes. 2002. 118 pp. Jansson, Joakim: Empirical Studies in Corporate Finance, Taxation and Investment. 2002. 132 pp. Carlsson, Mikael: Macroeconomic Fluctuations and Firm Dynamics: Technology, Production and Capital Formation. 2002. 149 pp. Eriksson, Stefan: The Persistence of Unemployment: Does Competition between Employed and Unemployed Job Applicants Matter? 2002. 154 pp. Huitfeldt, Henrik: Labour Market Behaviour in a Transition Economy: The Czech Experience. 2003. 110 pp. Johnsson, Richard: Transport Tax Policy Simulations and Satellite Accounting within a CGE Framework. 2003. 84 pp. Öberg, Ann: Essays on Capital Income Taxation in the Corporate and Housing Sectors. 2003. 183 pp. Andersson, Fredrik: Causes and Labor Market Consequences of Producer Heterogeneity. 2003. 197 pp. Engström, Per: Optimal Taxation in Search Equilibrium. 2003. 127 pp. Lundin, Magnus: The Dynamic Behavior of Prices and Investment: Financial Constraints and Customer Markets. 2003. 125 pp. Ekström, Erika: Essays on Inequality and Education. 2003. 166 pp. Barot, Bharat: Empirical Studies in Consumption, House Prices and the Accuracy of European Growth and Inflation Forecasts. 2003. 137 pp. Österholm, Pär: Time Series and Macroeconomics: Studies in Demography and Monetary Policy. 2004. 116 pp. Bruér, Mattias: Empirical Studies in Demography and Macroeconomics. 2004. 113 pp. Gustavsson, Magnus: Empirical Essays on Earnings Inequality. 2004. 154 pp. 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 Toll, Stefan: Studies in Mortgage Pricing and Finance Theory. 2004. 100 pp. Hesselius, Patrik: Sickness Absence and Labour Market Outcomes. 2004. 109 pp. Häkkinen, Iida: Essays on School Resources, Academic Achievement and Student Employment. 2004. 123 pp. Armelius, Hanna: Distributional Side Effects of Tax Policies: An Analysis of Tax Avoidance and Congestion Tolls. 2004. 96 pp. Ahlin, Åsa: Compulsory Schooling in a Decentralized Setting: Studies of the Swedish Case. 2004. 148 pp. Heldt, Tobias: Sustainable Nature Tourism and the Nature of Tourists'ʹ Cooperative Behavior: Recreation Conflicts, Conditional Cooperation and the Public Good Problem. 2005. 148 pp. Holmberg, Pär: Modelling Bidding Behaviour in Electricity Auctions: Supply Function Equilibria with Uncertain Demand and Capacity Constraints. 2005. 43 pp. Welz, Peter: Quantitative new Keynesian macroeconomics and monetary policy 2005. 128 pp. Ågren, Hanna: Essays on Political Representation, Electoral Accountability and Strategic Interactions. 2005. 147 pp. Budh, Erika: Essays on environmental economics. 2005. 115 pp. Chen, Jie: Empirical Essays on Housing Allowances, Housing Wealth and Aggregate Consumption. 2005. 192 pp. Angelov, Nikolay: Essays on Unit-­‐‑Root Testing and on Discrete-­‐‑Response Modelling of Firm Mergers. 2006. 127 pp. Savvidou, Eleni: Technology, Human Capital and Labor Demand. 2006. 151 pp. Lindvall, Lars: Public Expenditures and Youth Crime. 2006. 112 pp. Söderström, Martin: Evaluating Institutional Changes in Education and Wage Policy. 2006. 131 pp. Lagerström, Jonas: Discrimination, Sickness Absence, and Labor Market Policy. 2006. 105 pp. Johansson, Kerstin: Empirical essays on labor-­‐‑force participation, matching, and trade. 2006. 168 pp. Ågren, Martin: Essays on Prospect Theory and the Statistical Modeling of Financial Returns. 2006. 105 pp. 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 Nahum, Ruth-­‐‑Aïda: Studies on the Determinants and Effects of Health, Inequality and Labour Supply: Micro and Macro Evidence. 2006. 153 pp. Žamac, Jovan: Education, Pensions, and Demography. 2007. 105 pp. Post, Erik: Macroeconomic Uncertainty and Exchange Rate Policy. 2007. 129 pp. Nordberg, Mikael: Allies Yet Rivals: Input Joint Ventures and Their Competitive Effects. 2007. 122 pp. Johansson, Fredrik: Essays on Measurement Error and Nonresponse. 2007. 130 pp. Haraldsson, Mattias: Essays on Transport Economics. 2007. 104 pp. Edmark, Karin: Strategic Interactions among Swedish Local Governments. 2007. 141 pp. Oreland, Carl: Family Control in Swedish Public Companies. Implications for Firm Performance, Dividends and CEO Cash Compensation. 2007. 121 pp. Andersson, Christian: Teachers and Student Outcomes: Evidence using Swedish Data. 2007. 154 pp. Kjellberg, David: Expectations, Uncertainty, and Monetary Policy. 2007. 132 pp. Nykvist, Jenny: Self-­‐‑employment Entry and Survival -­‐‑ Evidence from Sweden. 2008. 94 pp. Selin, Håkan: Four Empirical Essays on Responses to Income Taxation. 2008. 133 pp. Lindahl, Erica: Empirical studies of public policies within the primary school and the sickness insurance. 2008. 143 pp. Liang, Che-­‐‑Yuan: Essays in Political Economics and Public Finance. 2008. 125 pp. Elinder, Mikael: Essays on Economic Voting, Cognitive Dissonance, and Trust. 2008. 120 pp. Grönqvist, Hans: Essays in Labor and Demographic Economics. 2009. 120 pp. Bengtsson, Niklas: Essays in Development and Labor Economics. 2009. 93 pp. Vikström, Johan: Incentives and Norms in Social Insurance: Applications, Identification and Inference. 2009. 205 pp. Liu, Qian: Essays on Labor Economics: Education, Employment, and Gender. 2009. 133 pp. 118 Glans, Erik: Pension reforms and retirement behaviour. 2009. 126 pp. 119 Douhan, Robin: Development, Education and Entrepreneurship. 2009. 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 Nilsson, Peter: Essays on Social Interactions and the Long-­‐‑term Effects of Early-­‐‑life Conditions. 2009. 180 pp. Johansson, Elly-­‐‑Ann: Essays on schooling, gender, and parental leave. 2010. 131 pp. Hall, Caroline: Empirical Essays on Education and Social Insurance Policies. 2010. 147 pp. Enström-­‐‑Öst, Cecilia: Housing policy and family formation. 2010. 98 pp. Winstrand, Jakob: Essays on Valuation of Environmental Attributes. 2010. 96 pp. Söderberg, Johan: Price Setting, Inflation Dynamics, and Monetary Policy. 2010. 102 pp. Rickne, Johanna: Essays in Development, Institutions and Gender. 2011. 138 pp. Hensvik, Lena: The effects of markets, managers and peers on worker outcomes. 2011. 179 pp. Lundqvist, Heléne: Empirical Essays in Political and Public. 2011. 157 pp. Bastani, Spencer: Essays on the Economics of Income Taxation. 2012. 257 pp. Corbo, Vesna: Monetary Policy, Trade Dynamics, and Labor Markets in Open Economies. 2012. 262 pp. Nordin, Mattias: Information, Voting Behavior and Electoral Accountability. 2012. 187 pp. Vikman, Ulrika: Benefits or Work? Social Programs and Labor Supply. 2013. 161 pp. Ek, Susanne: Essays on unemployment insurance design. 2013. 136 pp. Österholm, Göran: Essays on Managerial Compensation. 2013. 143 pp. Adermon, Adrian: Essays on the transmission of human capital and the impact of technological change. 2013. 138 pp. Kolsrud, Jonas: Insuring Against Unemployment 2013. 140 pp. Hanspers, Kajsa: Essays on Welfare Dependency and the Privatization of Welfare Services. 2013. 208 pp. Persson, Anna: Activation Programs, Benefit Take-­‐‑Up, and Labor Market Attachment. 2013. 164 pp. Engdahl, Mattias: International Mobility and the Labor Market. 2013. 216 pp. Krzysztof Karbownik. Essays in education and family economics. 2013. 182 pp. 141 142 143 144 145 146 147 148 Oscar Erixson. Economic Decisions and Social Norms in Life and Death Situations. 2013. 183 pp. Pia Fromlet. Essays on Inflation Targeting and Export Price Dynamics. 2013. 145 pp. Daniel Avdic. Microeconometric Analyses of Individual Behavior in Public Welfare Systems. Applications in Health and Education Economics. 2014. 176 pp. Arizo Karimi. Impacts of Policies, Peers and Parenthood on Labor Market Outcomes. 2014. 221 pp. Karolina Stadin. Employment Dynamics. 2014. 134 pp. Haishan Yu. Essays on Environmental and Energy Economics. 132 pp. Martin Nilsson. Essays on Health Shocks and Social Insurance. 139 pp. Tove Eliasson. Empirical Essays on Wage Setting and Immigrant Labor Market Opportunities. 2014. 144 pp. 149 Erik Spector. Financial Frictions and Firm Dynamics. 2014. 129 pp. 150 151 Michihito Ando. Essays on the Evaluation of Public Policies. 2015. 193 pp. Selva Bahar Baziki. Firms, International Competition, and the Labor Market. 2015. 183 pp.