The conditions for multi-level
governance
Implementation, politics and cooperation
in Swedish active labor market policy
Martin Lundin
DISSERTATION SERIES 2007:1
Presented at the Department of Government, Uppsala University
The Institute for Labour Market Policy Evaluation (IFAU) is a research institute under the Swedish Ministry of Employment, situated in Uppsala. IFAU’s
objective is to promote, support and carry out scientific evaluations. The
assignment includes: the effects of labour market policies, studies of the functioning of the labour market, the labour market effects of educational policies
and the labour market effects of social insurance policies. IFAU shall also disseminate its results so that they become accessible to different interested parties
in Sweden and abroad.
IFAU also provides funding for research projects within its areas of interest.
The deadline for applications is October 1 each year. Since the researchers at
IFAU are mainly economists, researchers from other disciplines are encouraged
to apply for funding.
IFAU is run by a Director-General. The institute has a scientific council, consisting of a chairman, the Director-General and five other members. Among
other things, the scientific council proposes a decision for the allocation of
research grants. A reference group including representatives for employer
organizations and trade unions, as well as the ministries and authorities concerned is also connected to the institute.
Postal address: P O Box 513, 751 20 Uppsala
Visiting address: Kyrkogårdsgatan 6, Uppsala
Phone: +46 18 471 70 70
Fax: +46 18 471 70 71
[email protected]
www.ifau.se
This doctoral dissertation was defended for the degree of Doctor in Philosophy at the
Department of Government, Uppsala University, June 8, 2007. The third essay of this
thesis contains a revised version of research previously published by IFAU as Working
paper 2005:2.
ISSN 1651-4149
MARTIN LUNDIN
The Conditions for
Multi-Level Governance
Implementation, Politics, and Cooperation
in Swedish Active Labor Market Policy
! "
# $ %&&' (&)(* +
, + + -,,. , ,.
/ 0. %&&'. , 1 + 02/ . 3
- 1
, 4 /
0
-. 5' . .
36 7'$27(2*&!2(78728.
9 , ++ : 3 ++ , + , , +2
, , , , + + ;4/0-< . , = =
.
, . , 3 , + =) + ,, + , : 3 ,
, ++ + , . 3 +2 4/0- ,
, .
>
, ,
.
3 , . , 33 . + . , ,
, + + . ,
, + , ,
++
. + , +
. ? , , ,
+ .
: >
333 ,
,
. 3 , ,
. 3 , , ,, , ,
,, .
, , , + , , , + .
3
3
1
-
4 ! "#$%&' !
"
@ 0
/ %&&'
36 7'$27(2*&!2(78728
))))
2'7(! ;,)AA..A:B))))
2'7(!<
List of papers
This dissertation consists of the following essays:
I.
II.
III.
Lundin, Martin. 2007. “Political Partisanship and Entity Size: When
Parties Matter for Public Policy.” Manuscript, Department of
Government, Uppsala University.
Lundin, Martin. 2007. “Explaining Cooperation: How Resource
Interdependence, Goal Congruence, and Trust Affect Joint Actions
in Policy Implementation.” A slightly different version of this article
has been accepted for publication by the Journal of Public
Administration Research and Theory and will appear in a forthcoming issue.
Lundin, Martin. 2007. “When Does Cooperation Improve Public
Policy Implementation?” Manuscript, Department of Government,
Uppsala University.
Contents
Acknowledgments ......................................................................................................7
The Conditions for Multi-Level Governance..............................................................9
Political Partisanship and Entity Size:
When Parties Matter for Public Policy .....................................................................49
Explaining Cooperation: How Resource Interdependence, Goal Congruence,
and Trust Affect Joint Actions in Policy Implementation.........................................75
When Does Cooperation Improve Public Policy Implementation? ..........................105
Acknowledgments
A couple of months ago, I was about to get rid of some old binders. In one of
them there were documents from my very first semester as a political science
student. Among the documents was a short paper with the following title:
“Implementation problems and the Swedish Active Labor Market Policy”.
Evidently, I was the author. This was a surprising discovery since I did not
remember that labor market policy had occupied my attention as an undergraduate student. I started to read the paper and recognized that the
analysis—as most pieces of work by inexperienced students—was quite
ordinary. One of the paper’s conclusions was that labor market policy involves coordination of authorities, which could be difficult. At the time
when I formulated these unoriginal ideas, I could not dream that I would
complete a dissertation on basically the same theme eleven years later.
I think this book includes more interesting findings than my early
attempts. The progress is probably partly a consequence of my own hard
work. But many are those who have been important for the completion of
this dissertation. It is not possible to provide the complete list of contributors, so it has to be restricted to those who have influenced the project the
most.
First and foremost, I owe my supervisors a sincere gratitude. With their
sharp comments, PerOla Öberg and Jörgen Hermansson have helped me
immensely during my years as a doctoral candidate. I recommend future
postgraduate students to take advantage of their skills and enjoyable
company.
The following readers provided constructive criticism and useful
suggestions on the entire manuscript: Paula Blomqvist, Matz Dahlberg, Nils
Hertting, Barry Holmström, and Sven Oskarsson. At earlier stages of
research, Hans Blomkvist, Joakim Johansson, and Björn Lindberg were
appointed commentators on parts of the manuscript. Their proposals also
affected the final results.
The environment at the Department of Government has been inspiring.
My colleagues made my years as a doctoral candidate rewarding. Thanks to
all of you! Gunnar Myrberg deserves a special mention. Gunnar has been my
nearest companion at the department. We were roommates for some years
and we have been teaching together, studying statistics in Ann Arbor, and
discussing many personal and job-related things. Kalle Lindgren has provided crucial methodological advises. This means that I can forgive him for
7
always trying to free ride when it comes to clean and refill the coffee
machine. Sverker Gustavsson’s guidance during my first year at the department must be recognized as well. Julia Jennstål, Sara Monaco, Thomas
Persson, Andreaz Strömgren, Pär Zetterberg, Jörgen Ödalen and all of those
who have participated in the many enjoyable discussions in the department’s
lunchroom over the years have also contributed in a positive way.
Another environment of importance is the Institute for Labour Market
Policy Evaluation (IFAU). My colleagues at IFAU inspired me to apply for
the postgraduate program in the first place. And IFAU has always been a
great place to visit. The staff’s knowledge in labor market policy and
statistical methods has been useful, but just as important are all those weird
and enjoyable discussions that take place at IFAU. Perhaps surprisingly,
economists can be fun.
I owe thanks to the following foundations and institutions for contributing
with financial support to various parts of the research project: The Institute
for Labour Market Policy Evaluation, the Swedish Research Council, the
Siamon Foundation, the Wallenberg Foundation, and the Swedish Council
for Working Life and Social Research.
But life is not only work. My family and friends have been important
supporters, although most of you never received a good description of what
my work was all about and why it was important to me. Now when I have
finished my thesis you have a great opportunity to put all your interests and
hobbies aside and discover the thrilling mystery of public administration. To
give you a foretaste of the excitement, I provide a quotation from Essay I: “I
use three different approaches to counter autocorrelation and heteroscedasticity. Robust standard errors clustered on municipalities and standard
errors based Newey and West’s heteroscedasticity and autocorrelation consistent covariance matrices with two lags are the first two methods.” I guess
you are tempted to get started right away?
It remains for me to express my appreciation to the most important ones:
Ingrid, all your colds and coughs have constantly forced me to reschedule
my plans during the last half-year. But your smile, laughter, and warm
embrace compensate for this by a comfortable margin: You make all the
troubles associated with writing a thesis go away. Daniela, you have been
the closest person to me ever since that short paper eleven years ago. You
gave substantial comments on my work, but your sacrifices and encouragement were even more important. These last eleven years would not at all
have been so good without your love. Thanks!
Martin Lundin
Uppsala, April 2007
8
Introduction
The Conditions for Multi-Level Governance
Contemporary political systems comprise several tiers of government. The
reason for having more than one level is apparent: there are no grounds to
expect a certain geographical entity to be the appropriate one for all public
concerns. But bureaucracies at various levels of government are often involved simultaneously when political ideas are implemented at local level.
Thus, a perennial problem in the business of running a country is how to
manage the machinery of the state when policies are turned into actions:
How can the central state direct local public units to work effectively
towards public sector goals?
The three essays housed in this thesis have distinct purposes. But they
have a common ambition to disentangle the course of events when agencies
from different levels of government—more precisely, from the central and
local levels—are involved at the same time in public policy implementation.
Essay I tries to find out whether local political partisanship explains local
government actions. Essay II sets out to explain cooperation between
agencies. Essay III, finally, examines the impact of cooperation on policy
output.
The empirical analyses are focused on active labor market policy (ALMP)
in Sweden. Thus, the study centers on the fight against unemployment,
which is one of the major problems in advanced industrialized societies
today. ALMP is a central-government policy domain. By this I mean that it
is a policy area in which the central government is clearly intended to be the
main principal. But local governments also have a key role. By examining
local-government involvement in ALMPs, and their relationship to centralgovernment agencies, it is possible to uncover what is actually going on at
local level. In turn, this gives valuable insights into governance processes.
The aim of this introduction is to bring the separate studies into a
common framework and provide some additional information on matters that
are only briefly touched upon in the essays. The introduction consists of four
sections and a data appendix. The purpose of the first section is to present
the problems of governance and policy implementation when various levels
of government are involved in local actions. In the essays, the research
setting is concisely introduced. Thus, the second section of this overture is a
9
Introduction
rather detailed description of intergovernmental relations within Swedish
ALMPs. The main aim is to prove that it is important to take relationships
between levels of government into account in ALMPs. In the third section,
the essays are summarized. Lastly, I take some license to reflect on experiences from the study. A data appendix is also attached: Within this research
project, I have collected new data through a postal questionnaire. This information is utilized in essays II and III. The appendix describes the survey.
Levels of government, governance, and implementation
In this section, it is demonstrated that political steering can be difficult when
central and local public units are involved simultaneously in policy implementation. In addition, I outline three key research questions that are important to try to answer in order to find out the conditions for governance in
these contexts.
Levels of government
The basic pattern of political organization in modern democracies is
territorial. Almost every country in the world has more than one level of
government, with only very small states as exceptions to this rule. And the
complexities tend to increase. Petersson (2003, 150) notes, for example, that
not so long ago, the inhabitants of the city of Gothenburg in Sweden were
only represented at the local and the national levels. Nowadays, there are six
levels to take into account: the European Parliament, the Swedish
Parliament, the county of Västra Götaland, the region of Gothenburg, the
municipality of Gothenburg, and the district councils in Gothenburg.
A very important reason for having different tiers of government is that
the preferences of the inhabitants in different geographical entities are
different. In order to promote values such as democracy and efficiency, these
preferences ought to affect public policy (for example, Elazar 1972;
Molander 2003; Thomas 1979). Gallagher, Laver, and Mair (2001) use a
fictitious bridge-building project to illustrate that objectives may differ due
to geographical location.
The lives of people who live right beside the bridge may be ruined by its noise
and disruption. People who live in the region but not in the immediate locality of
the bridge may find the bridge saves them hours each week behind the wheel of
an automobile and reduces the risk of their being involved in a fatal accident.
Those who live elsewhere in the country but too far away ever to use the bridge
may be forced to pay for part of it with their taxes while getting, as they see it, no
benefits from it. (Gallagher, Laver, and Mair 2001, 136)
10
The conditions for multi-level governance
Different geographical groups of individuals tend to have different interests
in the policy process, which may imply tensions. For instance, assume that
two authorities are involved in constructing the bridge presented in the quote
from Gallagher and his colleagues. All else being equal, one would expect
the authorities’ relationship to be less conflictual if both of them are regional
government agencies than if one of them is a local government actor and the
other represents the regional level. In the latter case, local objectives become
an issue. Thus, coordinating the activities of various governmental bodies is
difficult. A lot of time and other resources have to be devoted to resolve the
dilemma and it may be complicated for the central government to reach
ambitions.1
Newspaper articles often describe problems of cooperation between
authorities from different levels of government. Two brief examples from
Sweden can serve as illustrations. Svenska Dagbladet reported in 2003
(November 16) on collaborative problems between medical services, administered by the County Councils (Landsting), and the municipal Social
Services in the treatment of mentally ill patients. The article claimed that
clients are often passed on between the authorities and that no authority
takes full responsibility. The director-general of the National Board of
Health and Welfare (Socialstyrelsen) said that local implementing agencies
have trouble getting along when economic resources have to be taken from
both the county and the local level.
The second example concerns the responsibility for underage refugees
coming to Sweden without parents. A critical situation at a refugee center,
Carslund, was reported in the media in 2002. The number of children at the
center, administrated by the Swedish Migration Board (Migrationsverket),
had increased substantially. Thus, the center was understaffed and the premises were inadequate. Many children were in poor mental health and
several suicide attempts were reported. It was claimed in the newspapers that
problems of collaboration between the Migration Board and municipal
authorities had contributed to the troubles. It was unclear when and how the
municipalities should get involved, and which level should take responsibility for costs at different stages of the process. As a consequence, the
refugee center became overcrowded (for example, Dagens Nyheter,
February 10 and 11, 2002; for similar discussions see Dagens Nyheter,
December 30, 2003).
Governance
The problems of coordinating tiers of government have also received quite a
lot of attention among political scientists. In contemporary research, studies
1
For some classical studies on intergovernmental relations, see Grodzins (1966), Oates
(1972), and Wright (1988).
11
Introduction
often address the “governance” concept (for example, Pierre 2000; Pierre
and Peters 2000; Rhodes 1997; Stoker 1998). Similar discussions are found
in studies that prefer to talk about “intergovernmental management” (Radin
2003),2 “network management” (Kickert, Klijn, and Koppenjan 1997), or
“multi-level governance” (Hooghe and Marks 2003; Peters and Pierre 2001;
Smith 2003). The central theme in studies that use one of these trendy concepts is how to steer society towards collective goals given a situation in
which “boundaries between and within public and private sectors have
become blurred” (Stoker 1998, 17; see also Pierre and Peters 2000). It is
recognized that governing is difficult and that a simple and conventional
view of political steering is not adequate. That is, the central government
cannot steer society “from above” independent of other actors (for example,
Kickert, Klijn, and Koppenjan 1997; Stoker 1998). Accordingly, the
traditional view of bureaucracies—that is, Max Weber’s legal-rational
model—is not in correspondence with reality. In line with this view, linear
political steering-models with the focus on formal, hierarchical relationships
should be replaced with models in which relations are more fluid and
different public and private actors are assumed to be interdependent in
governing processes.
The governance perspective is not an especially elaborated framework.
Rather, it is “a set of observations looking for a more comprehensive theory”
(Pierre and Peters 2000, 7). Moreover, it does not provide a causal theory.
Instead, it is a framework for describing how governing works in practice
(Stoker 1998). One might wonder how useful the framework really is and
what new insights it actually provides. But the perspective can at least
highlight some important research questions.3 The most fundamental idea of
governance research is of relevance in this thesis: the central government
cannot realize its ambitions without taking other actors into consideration.
For instance, local governments and their objectives and resources are going
to influence central government possibilities of reaching policy goals in both
federal and unitary political systems.4
Two features of the governance discourse are important to underscore
here. First, governance theory and empirical studies that relate to governance
are often concerned with the interaction of a large number of public and
2
Sometimes the term “intergovernmental relations” is reserved for relationships among
governments within federal states. Moreover, scholars examining the European Union talk
about intergovernmental relations when they analyze the relationships among member states.
But similar dynamics exist in all political systems. Radin (2003) claims that it is too restrictive to see intergovernmental relations as something that occur only in federal states or
between countries.
3
Many of these questions could, however, be addressed within political science research
without explicitly relating to the popular governance research discipline.
4
Note that prior research on political control, before the “new governance” approach became
fashionable, demonstrated that various levels of government influence the output of national
government policies (for example, Chubb 1985; Scholz and Wie 1986).
12
The conditions for multi-level governance
private actors—the network. Coordination problems in these networks are
the usual focus of analysis.5 This thesis is limited to the relationships of
central and local government units.6 I do not consider networks at all and the
separate articles are not framed within a governance perspective. I adopt
more specific theories. Thus, the essays could not be classified as governance research in the same way as the typical governance study of today.
Nevertheless, the arguments and the empirical findings are of relevance for
scholars interested in governance processes.
Second, a key discussion in governance research is the role of the central
government. According to some scholars, the central government is just one
of many actors in self-organizing networks that no single actor can control
(for example, Rhodes 1997). That is, hierarchies are more or less unimportant. Some scholars think that this is normatively acceptable. Hirst (2000)
claims that the “old” democratic model in which political representatives are
held accountable for their actions must be rejected since politicians do not
exercise a monopoly on politics. Representative institutions could and
should primarily be “watchdogs.” Thus, it is necessary to adopt a “democratic model, which involves devolving as many of the functions of the state
as possible to society (whilst retaining public funding) and democratizing as
many as possible of the organizations in civil society. The aim is to restore
limited government and to ‘politicize’ civil society” (Hirst 2000, 28).
In my view, this is a dubious position both normatively and empirically.
If the descriptions of Rhodes and Hirst are correct empirically, I think that
the first step is not to surrender. Rather, it is important to find ways to reinstate democratically elected politicians as primary governing actors. A
point of departure in this study is that the central government ought to have a
significant influence—the government should govern.7
Fortunately, elected representatives and hierarchies are still likely to make
a difference. The most prominent supporters of this view are Pierre and
Peters (2000). Networks have become more important, they claim, but networks are less significant than hierarchical relationships. The government is
at the center of governance and the hierarchy “is the benchmark against
which we should assess emerging forms of governance” (Pierre and Peters
2000, 17). Lundqvist (2001) examines the validity of the claims of Rhodes
and Pierre and Peters using Local Investment Programs for Suitable
Development in Sweden as a critical case. The evidence supports Pierre and
5
For an interesting theoretical discussion and four detailed case studies on collaboration in
networks, see Hertting (2003).
6
For a discussion of central–local government relations, see Rhodes (1999).
7
This does not, of course, mean that I defend a position in which the central government
controls all public concerns in detail. This is not likely to be very beneficial from a democratic
or an effectiveness perspective. Actors such as subnational governments in both federal and
unitary states should make a difference. But given that the central government is intended to
be the main principal, it should definitely be able to govern.
13
Introduction
Peters’ argument that government is at the center of structures of governance.
Implementation
To understand the conditions for governance, it is important to understand
implementation at local level. At this level, political ambitions become
actions. Implementation research is about the courses of events in which
political ideas are put into practice, that is, what goes on between policy expectations and policy results (deLeon 1999). Local practices are not always
the same as the aims stated in official documents. Moreover, performance
frequently varies from one local context to another and from time to time. In
relation to classical definitions of politics, such as “who gets what, when and
how” (Lasswell 1936), implementation studies become important.8
Certainly, if we do not comprehend how political ambitions and ideas fare in
reality, we cannot know who gets what, when, and how.
Jeffrey Pressman’s and Aaron Wildavsky’s 1973 (1984) book
Implementation is usually regarded as the starting-point of implementation
research.9 Pressman and Wildavsky conducted a detailed case study of how a
public program, with the aim of increasing minority employment in
Oakland, California, was carried out in practice. They found large
discrepancies between intentions and practice. According to the authors, an
important explanation for the unsatisfactory results was that many actors had
to coordinate their work in the implementation process. Pressman and
Wildavsky identified about thirty major decision points that had to be cleared when the program was carried out. Several participants were involved in
the process and making all the actors work together was a difficult task,
leading to management problems. Thus, besides bringing implementation
onto the political scientists’ research agenda, Pressman and Wildavsky
pinpointed the dilemmas of interorganizational policy implementation.
The research body that followed the Pressman/Wildavsky-study is rather
messy. As O’Toole (2004, 310) puts it, “theories about policy implementation have been almost embarrassingly plentiful, yet theoretical consensus is
not on the horizon ... After hundreds of empirical studies, validated findings
are relatively scarce.” Similar conclusions are drawn by scholars such as
Matland (1995), deLeon (1999), and Rothstein (1994, ch. 3). But one of the
8
Among the most influential implementation studies are Brehm and Gates (1997), Elmore
(1979), Goggin et al. (1990), Hjern and Porter (1981), Lipsky (1980), Mazmanian and
Sabatier (1989), and Pressman and Wildavsky (1984). There are also several overviews of the
literature, such as Lester et al. (1987), Matland (1995), and Winter (2003a; 2003b).
9
But Pressman and Wildavsky were not the first scholars to address questions of
implementation (Winter 2003a, 203; Hill 1996, 127–28). For instance, Herbert Kaufman
(1960) examined implementation in the book The Forest Ranger many years before Implementation was published. Nonetheless, it was the work of Pressman and Wildavsky that
placed the issue of implementation in a paramount place on the research agenda.
14
The conditions for multi-level governance
few things scholars agree on is the conclusion by Pressman and Wildavsky
that we have to take into account that many actors influence implementation
if we want to understand what is going on at local level (O’Toole 2003).
Interorganizational relationships were subject to intense scrutiny in early
studies. For example, Hanf, Hjern, and Porter (1978) studied local networks
carrying out adult labor market training in Sweden and Germany. They
indicated that the implementation of these training programs required
assistance from many public and private organizations. Another contribution
is an analysis by Hjern and Porter (1981), who argued that clusters of
organizations normally carry out public programs. A multiorganizational
unit of analysis is therefore appropriate when describing and evaluating
public policy implementation. O’Toole and Montjoy (1984) is yet another
pioneering study. They analyzed different types of ties between
organizations and discussed how these relationships are likely to affect
policy implementation. The focus on multiple actors has remained a recurrent theme in empirical research.10 O’Toole (2003, 237) concludes:
The topic of interorganizational relations will remain important for administrators
tasked with helping to make policy implementation succeed. Accordingly, it is
critical to understand how to make sense of such institutional settings for
improving prospects for implementation success.
Usually, interorganizational policy implementation involves an intergovernmental aspect. According to Kettl (1993, 414), “nearly 80 percent of all of
the entries for programs and organizations in one standard implementation
textbook are for programs and organizations that have an important
intergovernmental dimension.” Thus, the implications of these relations for
public policy implementation are worth paying attention to.
A seminal work in this context is Goggin et al. (1990). Goggin and his
colleagues argued that research had neglected the role of the state level in
research on implementation of federal mandates in the United States. The
authors suggested a model of intergovernmental policy implementation.
Interests and motives, incentives and constraints, and power and resources at
various levels were assumed to affect how political ideas were turned into
practice. Case studies of hazardous waste policy, waste-water treatment
projects, and family planning programs suggested that several levels of
government should be taken into account. In an earlier study, Thomas (1979)
discussed similar things. Thomas stressed the importance of working out
jurisdiction between levels of government. Implementation at local level
would otherwise become difficult.11
10
Studies such as Agranoff and McGuire (2003), Bardach (1998), Hudson et al. (1999),
Jennings (1994), Jennings and Ewalt (1998), and O’Toole (1983) contribute to the debate.
11
In a more recent study, Cho et al. (2005, 33) claim that the literatures on intergovernmental
relations and policy implementation, to a large extent, “run parallel rather than intersect.”
15
Introduction
Research questions
The journey thus far has showed that it can be difficult for central government to realize ambitions at local level and that subnational actors are likely
to have a stake in implementation. From a democratic perspective, elected
representatives should be able to govern. Thus, the problem is how to coordinate various levels and actors. These actors may be politicians or public
servants. Hence, three research questions are important to penetrate in order
to comprehend local actions.
First, we need to discern the role of local politics: Do local government
actions dependent on what political parties control the local legislature? Or
can the central government expect local units to act in a similar manner regardless of who governs? There is a large literature on the effect of political
parties on public policy. A number of studies are concerned with the subnational level (for example, Besley and Case 2003; Petterson-Lidbom 2006).
The insights from this body of research can be used to improve our
understanding of central government steering possibilities. Essay I builds on
and contributes to this literature.
Second, in order to coordinate activities of local units, it is important to
find out what factors explain cooperation between agencies: How can the
government make agencies work together? Many studies try to explain interorganizational cooperation in policy implementation (for example, Alter and
Hage 1993; Bardach 1998; O’Toole 2003). Essay II adds to this research and
reveals some conditions for cooperation across organizational and governmental boundaries.
Third, we cannot assume that cooperation always improves policy
implementation. Thus, the impact of interorganizational cooperation on
policy output and outcome is also a key topic: Does cooperation between
agencies affect policy implementation? Some studies have focused on this
question (for example, Jennings 1994; Jennings and Ewalt 1998). The purpose of Essay III is to take a close look at the effects of cooperation.
The role of local governments in Swedish active labor
market policy
What is an appropriate case for studying the questions of interest in this
thesis? Two features are of certain interest. First, two different public
authorities, representing different levels of government, must be involved in
the implementation process. Second, there has to be some potential conflict
of interest between levels of government. Additionally, an important and
high-prioritized policy area is, of course, preferred over an issue of less
importance. ALMPs in Sweden is one case that fulfills these requirements.
16
The conditions for multi-level governance
In this section, the relationship between central and local government
authorities in Swedish ALMPs is described. The section is divided into six
subsections. First, I provide a short account of the role of local governments
in the Swedish political system. Second, the central government role in
ALMPs is portrayed. Third, I show how local governments are involved in
ALMPs. Fourth, the objectives of central and local government actors are
discussed. Fifth, it is demonstrated that the agencies interact to a considerable extent. Lastly, a short summary is offered.
Local governments in Sweden
Sweden is a unitary state, but has a long constitutional tradition of local
government liberty. There is a county level (landsting), but local governments (kommuner) are more important. As of 2007, there are 290 local
governments (or municipalities) in Sweden with an average population of
approximately 30,000 inhabitants. The largest municipality is Stockholm and
the smallest is Bjurholm: Stockholm has almost 800,000 residents and
Bjurholm fewer than 3,000. The municipalities provide a lot of services of
the welfare state, including day care, care of the elderly, social welfare
services, and primary education. As a result, they employ around 20 percent
of the total Swedish workforce. To a large extent, local services are financed
by local income taxes: only around 20 percent of their income is from grants.
Municipalities can set the tax rate freely, there are no restrictions on
borrowing, and municipalities decide on their own organization.12 In a
comparative perspective, local governments in Sweden are considered to
have a high degree of autonomy (Lidström 1996; Norton 1994).
But local government liberty is not unconditional. The central government regulates activities through various steering mechanisms, such as laws
and decrees. A common criticism is that the central government has decentralized a lot of responsibilities without providing the necessary funding
(Johansson 1996; Montin 2002). Furthermore, some scholars argue that there
has never been a genuine desire to diffuse much real power down to the local
level (Dahlkvist and Strandberg 1999). But most scholars agree that there
has been an overall tendency towards decentralization and that
municipalities have a key role in the Swedish political system, even though
there are obvious restrictions on actions (Montin 2002; Peterson 1994).
ALMPs – A central government responsibility
Municipalities manage a considerable part of the Swedish welfare sector. At
the beginning of the twentieth century, labor market activities were also a
12
For a description of the Swedish municipalities, see Gustafsson (1999), Montin (2002), and
Petersson (1994).
17
Introduction
local government concern. But in 1940, intended to be a temporarily solution
in turbulent times of war around the world, the national government took
over responsibility from the local governments. It did not turn out to be a
provisional solution. The National Labor Market Administration (Arbetsmarknadsverket) was institutionalized in 1948 in order to give nation-wide
effect to labor market activities (Thoursie 1990).
Labor market policies are usually split into two broad categories: passive
(PLMP) and active (ALMP) measures. Unemployment benefit is the main
type of passive support. Labor market training, subsidized employment,
work practice, and diverse job brokering activities are typical examples of
active programs. The active policy is primarily guided by ideas introduced
by trade union economists Gösta Rehn and Rudolf Meidner in the 1940s and
1950s. An extensive and active policy was necessary in order to combine the
policy objectives of high employment, low inflation, wage solidarity, and
endorsement of economic growth. Mobility-enhancing activities, such as
labor market training, were to be used so that workers risking unemployment
could be moved from low-productivity sectors to high-productivity sectors.
The policy was therefore an important part of Sweden’s economic policy,
and national government control was considered important (Rothstein
1996, 56–64).
Today, Public Employment Service (PES) offices (Arbetsförmedlingar)
implement central government decisions at street-level. There are about
300–350 PES offices in Sweden. Some of them are specialized in the
rehabilitation of unemployed with work disabilities and others work within a
particular business sector. But the vast majority of agencies provide services
to all kinds of job-seekers and employers, and can use the full set of ALMPs
available. The average number of employees is approximately 22, but some
offices only have a couple of employment officers, and others a staff of 100
members or more.13
Historically, the recruitment policy at all levels within the administration
has been based on ensuring the staff’s ideological commitment to national
policy objectives, rather than on formal education and merits. Promotion
policy and the training of employment officers have also aimed at
guaranteeing a commitment to overall national goals. The reason behind this
kind of staff policy was that a flexible organization with large discretion at
local level was preferred in order to make the policy efficient. At the same
time, to ensure that the policy worked according to national government
13
These figures are based on the National Labor Market Administration’s personnel statistics
from 2003.
18
The conditions for multi-level governance
intentions, it was important that street-level bureaucrats understood and
approved of policy objectives (Rothstein 1996, 116–30).14
The overall goal of the policy is an effective national labor market with
full employment and good economic growth (Runeson 2004, 10). In order to
reach this goal, management by objectives (målstyrning) is the principal
mechanism for governmental control over the National Labor Market
Administration and for supervision within the administration (Lundin 2004;
Nyberg and Skedinger 1998). There are, of course, laws, rules, and guidelines, but the local employment officers are relatively free when deciding the
appropriate actions in a particular case (Lundin 2004). The following list,
based on Runeson’s (2004, 9–13) survey of official documents, summarizes
the administration’s central tasks when trying to reach the overall goal of an
effective labor market:15
•
•
•
•
•
•
•
•
•
•
•
Match job-seekers with vacancies
Improve the skills of unemployed individuals
Enhance flexibility in the labor market
Protect and take special measures for weak groups in the labor market
Counteract inequalities between men and women in the labor market
Treat clients uniformly all around the country
Adapt activities to conditions in different parts of the country
Adapt labor market programs to the individual and improve his or her
chances of getting a job
Provide financial security to job-seekers through the unemployment
insurance system, but monitor that job-seekers follow the official
requirements for receiving benefits
Programs should not distort competition and displacement effects
should be minimized
Activities should not distort the ecological balance
These general tasks are complemented by specific quantitative goals adapted
to each PES office.16 It is apparent that ALMPs are primarily focused on
macroeconomic performance, that is, an effective matching procedure bet14
This is primarily a historical account of how the labor market administration was organized
in the 1950s, 1960s, and 1970s. Systematic evidence on the contemporary organization is, to
my knowledge, lacking. Formally, recruitment should be based on skills and experience rather
than on ideological commitment today.
15
The National Labour Market Administration is primarily regulated by the following official
laws and decrees: Law (2000:625), Ordinance (2001:623), Ordinance (2000:628), and
Ordinance (2000:634).
16
For example, in 2003 and 2004 there were quantitative goals concerning long-term
unemployed clients, unemployed youth, part-time unemployed individuals, and unemployed
persons with work disabilities. Moreover, there were also quantitative goals to measure the
output of labor market training, the share of clients having an individual action plan, and
employers’ satisfaction with the local PES office (Lundin 2004, 15–17).
19
Introduction
ween available jobs and job-seekers. It is also obvious that local interests
must not be prioritized over the national goal of decreasing unemployment.
For instance, the Swedish government notes that “narrow municipal interests
must not be prioritized over the goal of reducing unemployment and bottlenecks on the labor market since these interests may have negative effects on
the functioning of the labor market” (Government Bill 1999/2000:98, 70).17
Nevertheless, the government recognizes that within overall macroeconomic
objectives, activities ought to be adapted to local conditions and to clients’
needs and skills. In the next subsection, I will demonstrate that local governments have a key role in this process.
Local government involvement in ALMPs
Although the national government took over responsibility for ALMPs in the
1940s, the municipalities have remained important actors. For example, they
organized public relief works throughout the twentieth century (Olli
Segendorf 2003). In several governing documents, the Swedish government
makes clear that the National Labor Market Administration must carry out
activities in close collaboration with the municipalities.18 Besides these
general exhortations, the Swedish government has also taken a number of
steps in order to increase municipal involvement in ALMPs. Most of these
decisions were made in the 1990s when there was an unemployment crisis in
Sweden. Some examples are presented below.19
•
17
At local level in Sweden there are joint cooperative and advisory
bodies with the purpose of adapting ALMPs to suit local conditions.
These bodies are called Employment Service Committees (Arbetsmarknadsnämnder). Some kind of local advisory bodies have existed
in Sweden since the 1970s, but in 1996 the committees were
reformed. The national government wanted to strengthen the
municipalities’ role in ALMPs. Thus, municipal representatives
should constitute the majority of the committee members. In addition,
the municipalities were given the right to nominate the committee
chairman. This construction is somewhat peculiar since ALMPs are
supposed to be a responsibility of the central government. But
committee representatives from the National Labor Market Ad-
This is my own translation of the following quote in Swedish: “De snävt kommunala
egenintressena får inte prioritieras framför bekämpningen av arbetslöshet och begynnande
flaskhalsar eftersom det kan ge oönskade effekter på arbetsmarknadens funktionssätt.”
18
For example, in Ordinance (2000:623) the PES offices are urged to work together with the
municipalities.
19
This list can be developed further. For additional examples, see Government Bill
(2002/03:44) and Forslund, Fröberg, and Lindqvist (2004).
20
The conditions for multi-level governance
ministration were supposed to stop decisions that run counter to
national objectives (Lundin 1999; Lundin and Skedinger 2006).20
•
Another policy change in the 1990s concerned programs for
unemployed youth. In 1994, the government declared that no young
person should remain unemployed for more than 100 days without
taking part in an active program. The government’s ambition was that
every individual under 25 years of age should be offered a labor
market program if he or she was unable to find a job within three
months of registration as unemployed at the PES. The government
encouraged the municipalities to take part in the actions by instigating
two labor market programs. The Municipal Youth Program
(Kommunala ungdomsprogrammet) was introduced in 1995. In 1998,
the UVG-guarantee (Ungdomsgarantin) was implemented. The
former program was directed towards youth below 20 years of age;
the latter concerned youth aged 20–25. Both programs implied that
the municipality should arrange activities for young unemployed
while getting some financial compensation from the PES office
(Carling and Larsson 2005).21
•
An additional policy change with implications for the municipalities’
engagement in ALMPs was a change in the Social Service Act in
1998. The reform made it possible for the municipalities to demand
those social assistance recipients whose major problem was
considered to be unemployment to participate in work related programs in exchange for social allowances. Municipalities were allowed
to reduce benefit if clients refused to take part in work related programs.22 This resulted in a lot of municipal labor market related
activities for unemployed social assistance recipients (Hjertner
Thorén 2005; Salonen and Ulmestig 2004).23
Municipalities are involved in many labor market programs. Lundin and
Skedinger (2006) estimate that in 1999, 40 percent of the participants in all
the National Labor Market Administration’s active programs were, in fact,
organized by the municipalities. The target groups are often youth (Carling
and Larsson 2005) and social allowance recipients (Salonen and Ulmestig
20
The Local Employment Service Committees are regulated by Ordinance (2001:623). See
also Government Bill (1994/95:218), Government Bill (1995/96:25), Government Bill
(1995/96:148), Government Bill (1995/96:222), and Government Bill (1999/2000:98).
21
See Law (2000:625) and Ordinance (2000:634). See also Government Bill (1994/95:218),
Government Bill (1997/98:1), and Government Bill (1999/2000:98).
22
See Law (2001:453) and Government Bill (1996/97:124).
23
Salonen and Ulmestig (2004, 35) estimate that around 12,000 persons were activated in
such measures at any given point in time in 2002. This corresponds to approximately
10 percent of all the individuals participating in labor market programs in Sweden.
21
Introduction
2004). Job-search assistance and work practice are some of the activities
municipalities arrange.
Many local governments have established a labor market administration
to handle labor market issues. This is probably a consequence of increased
municipal engagement in ALMPs in the 1990s. In 1995, about two in ten
municipalities had instigated such an administration. Three years later, in
1998, the share had increased to approximately seven in ten (Swedish
Association of Local Authorities 1999).24
Why are municipalities taking part in ALMPs? The central government
wants local governments to be involved since they can contribute with
diverse forms of local resources. From a municipal viewpoint, unemployment entails considerable problems. For instance, the local unemployment
rate impinges on municipal revenues; high local unemployment will result in
low tax revenues. Unemployment also increases municipal costs. For instance, unemployed persons not qualifying for unemployment benefits
financed by the central government often need social allowances paid for by
the municipalities (Wetterberg 1997). Unemployment may also make
citizens leave the municipality in order to get jobs in other parts of Sweden.
Thus, for reasons such as securing the local tax base, decreasing expenditure
on social allowances, and stopping emigration, the municipalities may try to
help the PES offices out.25
Central and local government objectives
By now we know that both central and local government authorities are
involved in the implementation of ALMPs in Sweden. But what objectives
do the authorities have? Should we expect conflicts of interest?
There is no reason to doubt that both the central and the local level share
the overall goal of decreasing unemployment. But interests may diverge
within this general goal. A simple example can be used as illustration:
emigration is a serious problem for municipalities in Sweden—the tax base
is endangered. From a national labor market perspective, however, geographical mobility is important. The PES offices’ task becomes easier if
unemployed clients move from a troublesome labor market environment to
areas where the chances of getting a job are better. Accordingly, whereas
municipalities probably are more focused on finding a local solution, PES
24
Note that municipalities that do not have a special labor market administration manage
labor market issues within another administration, for example, social services.
25
Recently, centralization tendencies have become more apparent in Swedish ALMPs. For
example, the municipal youth programs discussed above were abolished in spring 2007
(Government Bill 2006/07:100; Ministry of Employment 2007), since evaluations indicated
poor results (for example, Carling and Larsson 2005; Forslund and Nordström Skans 2006).
However, there is no systematic knowledge on centralization tendencies and it is too early to
say that subnational levels have lost influence.
22
The conditions for multi-level governance
offices are likely to be more open to activities that do not benefit local
governments.26 In a study of the Local Employment Service Committees,
Lundin (1999) found that 60 percent of the committee members were of the
opinion that there were conflicts between central and local interests in local
level activities.27 Thus, there are reasons to assume that a central–local
dimension gives rise to certain dilemmas when ALMPs are implemented.
Within the present research project, questionnaire data on the
relationships between PES offices and municipalities have been gathered.
These data are used in essays II and III. Details are reported in the appendix
attached to this introduction. The respondents were PES managers, managers
of municipal labor market administrations, and local politicians in charge of
municipal labor market activities. A set of questions concerned the agencies’
objectives. Table 1 shows—by denoting the share of respondents claiming
that a certain goal receives “very high” or “fairly high” priority—how the
authorities allocate priorities among different goals.
The findings are decisive. As expected, reducing unemployment is very
important to both organizations. Both of them also give high priority to
unemployed youth, clients with severe difficulties in the labor market, and
matching procedures. But the overall impression is that agencies prioritize
objectives in a different manner.
The PES offices put much more emphasis on formal directives and guidelines from the central government. While 98 percent of the PES managers
said that central government rules and guidelines have high priority, only
58 percent of the municipal managers and 31 percent of the politicians gave
the same answer. In addition, municipal representatives do not consider the
goals of the National Labor Market Administration to any significant extent.
Because subsidizes are for the most part paid by the central government, it is
not surprising that shifting persons from subsidized to unsubsidized jobs is
another objective primarily of concern for the PES offices. Furthermore,
monitoring is more important for PES offices, whereas municipal officials
give somewhat more priority to clients’ own needs and requests.
26
A concrete and real example demonstrates this: in the Local Employment Service
Committee of Arjeplog, the municipality acted in order to shift financial resources from a
central government mobility grant to local labor market programs (Protocols of Local
Employment Service Committee February 21, 1997 and October 29, 1997, municipality of
Arjeplog). These activities were halted, but it indicates that local representatives may act in
order to benefit their own commune in the implementation of ALMPs.
27
A couple of open comments may illustrate this further. A handful of representatives from
the National Labor Market Administration said that “the municipality wants to increase local
population, whereas the PES ought to increase geographical flexibility on the labor market”,
and “municipalities like to improve their budget through national labor market resources”. On
the other hand, some municipal representatives concluded “local initiatives are blocked by
national rules on how the labor market policy should work” (Lundin 1999, 45–6).
23
Introduction
Table 1. Objectives in ALMPs: Percentage Claiming that a Certain Objective is
Given “Very High” or “Fairly High” Priority
PES
Objective
Municipality
Managers
Managers
Politicians
Difference
Attaining the quantitative goals of
the National Labor Market
Administration
97
37
35
+ 61
Following central government
rules and guidelines
98
58
31
+ 53
Monitoring clients
91
53
58
+ 36
Shifting people from subsidized to
unsubsidized jobs
72
52
62
+ 15
Improving matching between
available jobs and unemployed
persons
97
70
82
+ 11
Ensuring that there are labor
market programs for groups of
unemployed with severe problems
in the labor market
94
80
86
+ 11
Ensuring that there are labor
market programs for young people
under 25
98
92
95
+4
Reducing unemployment
91
89
95
-1
Taking clients’ own requests and
needs into account
62
82
69
- 13
Improving municipal services for
the local population
12
49
52
- 38
Activating unemployed persons
living on social assistance in labor
market programs
29
79
89
- 55
Increasing or maintaining the local
population
16
68
92
- 61
6
83
92
- 81
Reducing expenditure on social
assistance
Notes: Data come from questionnaires distributed to managers of PES offices, managers of municipal
labor market administrations, and local politicians having responsibility for labor market questions.
Formulation of question: “How are the following objectives prioritized at the PES (in the municipality’s
labor market activities)?” The question directed to the politicians was slightly different: “How are the
following objectives prioritized by the political majority in the municipality?” A scale of five categories
was used: “Very high,” “Fairly high,” “Neither high nor low,” “Fairly low,” and “Very low or not at all.”
The number of respondents for each item was between 258 and 264 among the PES offices, between 238
and 242 among municipal managers, and between 231 and 242 among the politicians. The difference
column presents the percentage point difference between answers from the PES offices and the average of
managers’ and politicians’ answers in the municipalities.
24
The conditions for multi-level governance
Municipalities think that objectives of a local character are important. Two
examples underscore this. First, among the PES managers, 16 percent
regarded population goals as important. About 68 percent of the municipal
managers and 92 percent of the local politicians noted that maintaining or
increasing the local population is an important objective in local ALMPs.
Second, the PES offices do not pay for social assistance. This is probably
why decreasing these expenditures and activating unemployed persons living
on social assistance in programs are objectives to which they do not give
priority. More or less all municipal actors claimed that these matters are important.
A couple of objectives are prioritized somewhat differently among
municipal managers and politicians. The elected representatives give less
attention to central government rules and guidelines. Furthermore, they think
that population goals are even more important. By and large, however, the
two groups of municipal representatives provide a rather similar picture.
In another survey question, I asked whether there had been any conflicts
between the PES and the municipality in 2003. Around 28 percent of the
PES managers, 50 percent of the municipal managers, and 32 percent of the
local politicians reported minor or major conflicts. Most of these respondents
marked the response alternative “yes, minor conflicts.” Thus, it is probably
not correct to describe the relationship as extremely problematic. Nevertheless, I believe that it is safe to conclude that there are conflicts of interest
that emerge as a consequence of the fact that the authorities represent
different tiers of government.
Do PES offices and municipalities interact?
A final question to consider before this description of the research setting is
complete is to what extent the authorities interact in local activities. In the
questionnaire, the respondents were asked to report the level of contact between the authorities. They also reported how much they communicated with
other groups of actors involved in ALMPs. The answers to these questions
are presented in Figures 1–3.
25
Introduction
Other PES
County Labour Board
National Labour Market Board
Municipality
Social Insurance office
County Administrative Board
County Council
Other public authorities
Unions
Employers associations
Other stakeholder organizations
Private firms
0%
20%
40%
Daily contacts
60%
80%
100%
Weekly contacts
Figure 1. Frequency of Communication between PES Offices and Diverse
Categories of Organizations (Answers by PES Office Managers, n = 263)
Figure 1 indicates the level of communication according to PES managers.
More than nine in ten reported that the PES has at least weekly contacts with
the municipality. About 50 percent noted daily contacts. Only private firms
and other PES offices are contacted to a greater extent. Most potential work
opportunities are located within the private sector and actions probably need
to be coordinated among neighboring PES offices. It is therefore not
surprising that these groups of actors are contacted more often. Note that the
PES offices seem to interact more with municipalities than with key actors
such as unions, employers’ associations, and even the County Labor
Boards.28
The answers from municipal managers are reported in Figure 2. The
picture from Figure 1 is underscored. Communication within the municipal
organization is, of course, rather intense. But the municipalities’ interaction
with the PES offices is frequent as well: around nine in ten report that the
authorities contact each other at least on a weekly basis.
28
The County Labor Board (Länsarbetsnämnden) is the body within the National Labor
Market Administration located at the county level.
26
The conditions for multi-level governance
Other administrations within the municipality
Other municipalities
Local PES
County Labour Board
National Labour Market Board
Association of Local Authorities and Regions
Social Insurance office
County Administrative Board
County Council
Other public authorities
Unions
Employers associations
Other stakeholder organizations
Private firms
0%
20%
40%
Daily contacts
60%
80%
100%
Weekly contacts
Figure 2. Frequency of Communication between Municipal Labor Market
Administrations and Diverse Categories of Organizations (Answers of Municipal
Labor Market Managers, n = 248)
Figure 3 presents the answers from municipal politicians. At first glance,
Figure 3 seems to show a somewhat different picture. But unlike the
managers, the politicians were supposed to provide an answer that reflected
how often they personally communicate with other actors in labor market
related issues. This means that their answers are not an account of the
municipal labor market administrations’ communication with various groups
of actors. Very few politicians speak to civil servants at the PES office on a
daily basis. Around 20 percent indicate at least weekly communication. On
the other hand, in comparison to other groups of labor market actors, the
PES offices are important.
All things considered, there seems to be a lot of contact between PES
offices and municipalities. In general, they do not operate in isolation from
each other.
27
Introduction
Other municipalities
Association of Local Authorities and Regions
Local PES
County Labour Board
National Labour Market Board
Ministry of Ind., Employment and Com.
Social Insurance office
County Administrative Board
County Council
Other public authorities
Unions
Employers associations
Other stakeholder organizations
Private firms
Elected representatives at central level
0%
20%
40%
Daily contacts
60%
80%
100%
Weekly contacts
Figure 3. Frequency of Communication between Local Politicians in Charge of
Labor Market Activities and Diverse Categories of Actors (Answers of Municipal
Politicians, n = 237)
Governance and Swedish ALMPs
It is obvious that it is important to take relationships between levels of
government into account in Swedish ALMPs. The central government has
ambitious goals within labor market policy. A large and strong national
public authority is intended to take care of labor market operations. But at
local level, local governments are also involved when decisions are turned
into actions. Central and local government agencies interact to a great extent
in policy implementation. And even though reducing unemployment is a top
priority for both actors, objectives diverge to a considerable extent. This
means that governance may be complicated.
In the essays to follow, the conditions for governance within ALMPs in
Sweden are analyzed. It is examined whether political partisanship at local
level influences municipal policy within ALMPs: Does it matter whether the
left wing or the right wing governs at local level? Furthermore, explanations
to cooperation between PES offices and municipal administrations are
studied: What factors boost cooperation? Lastly, I take a look at the impact
28
The conditions for multi-level governance
of cooperation on policy output: When does cooperation improve public
policy implementation? In the next section, the essays are summarized.
Lessons from the essays
The three essays of this thesis pose general research questions, with distinct
purposes beyond ALMPs in Sweden. This means that the study provides
insights both to the particular case and to common problems in most policy
areas.
Lesson one: Political partisanship explains local government policies,
but only in large local entities
Do political parties make a difference for public policy? This question
constitutes a large political science literature. Essay I adds to this body of
research. The focus is local governments and it is hypothesized that political
partisanship has a greater impact on public policy in large local entities than
in small ones. The hypothesis rests on the idea that politics is more pragmatic, and party antagonisms less pronounced, in small subnational entities.
Empirically, local government involvement in ALMPs in Sweden is
analyzed. There are strong reasons to suspect leftist local governments to be
more involved in ALMPs than right-wing ones since these policies primarily
benefit the core voters of the left. But if politics are more pragmatic in small
entities, and if party politicization becomes more important as entity size increases, the effect should be larger in more sizeable entities.
The analysis is based on quantitative panel data from 1998 to 2005. The
amount of resources used for ALMPs per local inhabitant is employed as the
dependent variable. These data allow for hard empirical tests. In accordance
with the hypothesis, the effect of political partisanship turns out to be contingent on local population size. In the smallest entities, it is not important
who governs—ALMP expenditures are not affected by political partisanship.
As population size increases, partisanship begins to play a much more
essential role. Left-wing governments spend, on average, more money on
ALMPs than right-wing governments. Thus, forthcoming studies on the
importance of local political partisanship for public policy should consider a
possible interaction term between partisanship and size.
The study also reveals that local needs explain local government involvement in ALMPs to a large extent. A factor such as unemployment is, not
surprisingly, an important explanation.
The findings can primarily be generalized to countries in which local
governments have a lot of autonomy and to policy areas in which there is an
apparent left–right political dimension. But there is no strong reason to
29
Introduction
assume that the idea does not apply in other settings as well since the
argument is general. More research could reveal whether this is correct.
Lesson two: Trust and similar goals increase cooperation only when
they exist simultaneously
A perennial problem of public administration is how to make agencies work
together. In Essay II, the task is to explain levels of cooperation between
agencies. Previous research shows that resource interdependence, goal congruence, and mutual trust boost interorganizational cooperation. These three
factors are also examined in Essay II. The empirical findings show that all
three are important. But it is argued that interaction effects must be considered in order to fully understand the relationships. The argument is not
restricted to the study of ALMPs in Sweden, or to intergovernmental
relations. It is a general claim concerning how cooperation between organizations and between individuals ought to be understood.
The study is based on 203 dyads of Swedish PES offices and municipal
labor market administrations in 2003. I find that mutual trust is necessary if
goal congruence is to increase cooperation between agencies. It is argued
that this is because objectives cannot be communicated in a credible way
when one party does not trust one another. Furthermore, mutual trust has a
positive effect only if organizations have similar objectives. The idea is that
trust can make cooperation easier, but if actors do not want the same thing it
does not matter how reliable they find one another; trust is not sufficient by
itself. Thus, the common claims that trust and goal congruence affect
cooperation should be modified. Cooperation will increase when trust and
goal congruence exist simultaneously. Thus, if a management strategy aimed
at increasing cooperation only focuses on organizational objectives or the
level of trust, it is likely to fail.
But the study indicates that the effects of trust and resource interdependence are not contingent on each other. My argument that resource
interdependence and trust should be interacted is therefore not supported by
evidence from ALMPs in Sweden. This may be a result of methodological
problems, but it is also possible that resource interdependence is enough to
guarantee that actors are honest and stick to an agreement. Research projects
with the aim of solving this puzzle would be valuable.
The findings of the study are robust, but it necessary to be cautious when
it comes to causal statements since the analysis is conducted within the
limits of a cross-sectional design. One fundamental question is the causal
order between trust and cooperation. Accordingly, additional studies seeking
to discern the importance of interaction terms between trust on the one hand,
and resource interdependence and goal congruence on the other hand, would
be valuable.
30
The conditions for multi-level governance
Lesson three: Cooperation improves implementation of complex tasks
Essay III is concerned with the fundamental question of policy implementation: What factors explain policy output? Interorganizational cooperation is
often considered valuable in the public sector. If cooperation increases, so
the argument goes, political objectives will be reached to a greater extent.
But in Essay III it is proposed that the impact of cooperation is contingent on
the type of policy being carried out.
It is suggested that the effect of cooperation on implementation
performance increases with task complexity. It is likely that the benefits
from cooperation are rather low when the task does not demand a wide range
of intense actions. In these situations, the capacity to carry out decisions is
not enhanced that much through cooperation—the most important resources
are already available within a certain agency. Working across organizational
boundaries implies costs. For instance, time and other resources are
necessary to establish and maintain a productive relationship. If cooperation
does not imply that many benefits, there is a significant chance that costs
will be high in relation to benefits. But as complexity increases, so do the
benefits. Cooperation becomes a more value adding activity. Accordingly, I
expect no or very little impact from cooperation when the task is not complex and a stronger positive effect in the case of complex tasks. To my
knowledge these arguments have not been discussed and studied in the
context of local policy implementation before.
Two policies within Swedish ALMPs are examined: programs for
unemployed youth and activities for clients with an especially demanding
situation on the labor market. The fundamental difference is that the youth
policy corresponds to standard duties, while the activities for clients in a
particularly difficult situation are much more complex. The prediction is that
cooperation is a more fruitful strategy in the latter case.
In agreement with the hypothesis, the empirical test suggests that policy
matters. A positive effect of cooperation is indicated only when the policy is
complex. I argue that Swedish ALMPs constitute a critical case in an important sense. There are strong reasons to expect that cooperation generally
is a good strategy. First, information requirements are high. Second, both
authorities share the overall goal of reducing unemployment. Third, the
client groups of the two policies are highly prioritized by both agencies (see
Table 1 in this Introduction). Fourth, both the PES offices and the
municipalities have resources that can be valuable in the implementation of
measures for both client groups. Given the empirical findings, there is no
reason to expect a positive effect of cooperation in situations in which the
conditions are less favorable if the task is not complex.
The main lesson of Essay III is that we should not take for granted that
cooperation makes implementation better. Thus, scholars should develop and
test theories on when interorganizational cooperation is a good strategy for
31
Introduction
making political ambitions come true. And practitioners should perhaps not
see cooperation as a panacea for improving public sector performance.
Routes to better governance
I started this overture with a question: How can the central state direct public
bodies to work effectively towards public sector goals? Obviously, this
thesis cannot decisively answer this grand question. The essays deal with
questions that ultimately are about causality—and causality is always intricate in the social sciences. All ideas introduced in the thesis must be
analyzed in additional contexts before distinct, generalized knowledge is
possible.
The thesis indicates a number of ways that research can be improved. For
instance, more quantitative studies on policy implementation are needed,
especially from European countries. Pooled cross-sectional time-series data
sets would be particularly welcome. Closer looks at the causal mechanisms
that are suggested in the three essays are also valuable. Moreover, additional
indicators of key variables—for example, implementation output and resource interdependence—would also be helpful. A last important lesson for
future research is the significance of interaction effects: All three essays
examine whether the effect of a certain variable differs depending on other
factors. In all three cases, interactions were important. For example, political
partisanship is more important for public policy in large than in small local
entities. Thinking in terms of conditional impacts can be helpful for improving both theory and empirical analysis. General law-like theories such as
“partisanship matters for public policy” are, of course, more appealing. But
integrative theories can often make our claims more precise. The fortunes
and problems of interaction effects have been discussed in many prior
studies (for example, Lindgren 2006; Oskarsson 2003). But in the literatures
of central interest for this thesis, studies that focus on interaction effects are
rather scarce. In my view, a promising path for future research would be to
consider possible interactions more seriously.
Although more research is needed, this thesis has provided important insights to governance processes. The main implications can be summarized as
follows:
•
32
Local agencies from different levels of government interlock in policy
implementation at local level. Although they share overall objectives,
they will have different priorities as a consequence of the fact that
they represent different levels of government. The objectives affect
cooperation between units of government. Goal congruence boosts
cooperation. But goal congruence is not a sufficient condition for increasing joint actions. If there is no trust, cooperation will be com-
The conditions for multi-level governance
paratively low, even though agencies want the same thing. Nevertheless, trust does not make a difference if the actors have different
objectives. Thus, it is important to consider trust and objectives at the
same time.
•
Local government actions are motivated by local needs to a large
extent. In addition, when agencies need one another, cooperation
between them increases. This means that steering mechanisms should
probably look different, depending on the local situation.
•
Political ideology does not play an important role in local government
actions in small entities. But in large entities partisanship has a profound impact. Thus, coordination procedures should most likely take
different forms in large and in small entities.
•
A high level of cooperation between units of government can improve
public policy implementation. But whether this actually is the case
will depend on how complex the task is. Policies that are not
particularly complex are not carried out better as a consequence of
collaborative endeavors. As complexity increases, it becomes more
important that agencies actually work together if the central government wants to achieve its policy ambitions.
The governance literature claims that multi-level and multi-jurisdictional
governance is the norm rather than the exception. Whether this is true or not,
and whether this is something really new, is debatable. But in any case, the
scope and focus of governing will often involve an intergovernmental
dimension. I have provided information on what is actually going on at local
level. How these facts should be valued from a public management perspective is up to others. However, the findings can hopefully help us find the
routes to better governance we all seeking so desperately.
33
Introduction
References
Literature
Agranoff, Robert, and Michael McGuire. 2003. Collaborative public
management. New strategies for local governments. Washington, DC:
Georgetown Univ. Press.
Alter, Catherine, and Jerald Hage. 1993. Organizations working together.
Newbury Park, CA: Sage.
Bardach, Eugene. 1998. Getting agencies to work together. The practice and
theory of managerial craftsmanship. Washington, DC: Brooking Institution
Press.
Besley, Timothy, and Anne Case. 2003. Political institutions and policy
choices: Evidence from the United States. Journal of Economic Literature
41(1):7–73.
Brehm, John, and Scott Gates. 1997. Working, shirking, and sabotage.
Bureaucratic response to a democratic public. Ann Arbor, MI: Univ.
Michigan Press.
Carling, Kenneth, and Laura Larsson. 2005. Does early intervention help the
unemployed youth? Labour Economics 12:301–19.
Cho, Chung-Lae, Christine A. Kelleher, Deil S. Wright, and Susan Webb
Yackee. 2005. Translating national policy objectives into local achievement
across planes of governance and among multiple actors: Second-order
devolution and welfare reform implementation. Journal of Public
Administration Research and Theory 15(1):31–54.
Chubb, John E. 1985. The political economy of federalism. American
Political Science Review 79(4):994–1015.
Dahlkvist, Mats, and Urban Strandberg. 1999. Kommunal sjävlstyrelse som
maktspridningspojekt? Den svenska statstraditionen och den lokala
politiska styrelsen. In Maktdelning, SOU 1999:76, ed. E. Amnå, 257–318.
deLeon, Peter. 1999. The missing link revisited: Contemporary implementation research. Policy Studies Review 16(3/4):311–38.
Elazar, Daniel J. 1972. American federalism: A view from the states. New
York: Thomas Crowell.
Elmore, Richard F. 1979. Backward mapping: Implementation research and
policy decisions. Political Science Quarterly 94(4):601–16.
34
The conditions for multi-level governance
Forslund, Anders, Daniela Fröberg, and Linus Lindqvist. 2004. The Swedish
activity guarantee. OECD social, employment and migration working paper
no. 16. Paris: OECD.
Forslund, Anders, and Oskar Nordström Skans. 2006. Swedish youth labour
market policies revisited. Working paper 6, Institute for Labour Market
Policy Evaluation, Uppsala, Sweden.
Gallagher, Michael, Michael Laver, and Peter Mair. 2001. Representative
government in modern Europe. 3rd edition. New York: McGraw-Hill.
Goggin, Malcolm L., Ann O’M. Bowman, James P. Lester, and Laurence J.
O’Toole Jr. 1990. Implementation theory and practice. Toward a third
generation. New York: HarperCollins.
Grodzins, Morton. 1966. The American system: A new view of government in
the United States, ed. D. J. Elazar. Chicago, IL: Rand McNally.
Gustafsson, Agne. 1999. Kommunal självstyrelse. Stockholm, Sweden: SNS.
Hanf, Kenneth, Benny Hjern, and David O. Porter. 1978. Local networks of
manpower training in the Federal republic of Germany and Sweden. In
Interorganizational policy making. Limits to coordination and central
control, ed. K. Hanf and F. W. Scharpf, 303–41. London: Sage.
Hertting, Nils. 2003. Samverkan på spel. Rationalitet och frustration i nätverksstyrning och svensk stadsdelsförnyelse. Stockholm, Sweden: Égalité.
Hill, Michael. 1997. The policy process in the modern state. 3rd ed. London:
Prentice Hall.
Hirst, Paul. 2000. Democracy and governance. In Debating governance, ed.
J. Pierre, 13–35. Oxford, UK: Oxford Univ. Press.
Hjern, Benny, and David O. Porter. 1981. Implementation structures: A new
unit of administrative analysis. Organization Studies 2(3):211–27.
Hjertner Thorén, Katarina. 2005. Municipal activation policy: A case study
of the practical work with unemployed social assistance recipients. Institute
for Labour Market Policy Evaluation, Working Paper 20, Uppsala, Sweden.
Hooghe, Liesbet, and Gary Marks. 2003. Unraveling the central state, but
how? Types of multi-level governance. American Political Science Review
97(2):233–43.
Hudson, Bob, Brian Hardy, Melanie Henwood, and Gerald Wistow. 1999. In
pursuit of inter-agency collaboration in the public sector. Public Management 1(2):235–60.
35
Introduction
Jennings, Edward T., Jr. 1994. Building bridges in the intergovernmental
arena: Coordinating employment and training programs in the American
states. Public Administration Review 54(1):52–60.
Jennings, Edward T., Jr., and Jo Ann G. Ewalt. 1998. Interorganizational
coordination, administrative consolidation, and policy performance. Public
Administration Review 58(5):417–28.
Johansson, Anders. 1996. Staten i den kommunala verksamheten. In SOU
1996:169.
Kaufman, Herbert. 1960. The forest ranger. Baltimore, MD: Johns Hopkins
Univ. Press.
Kettl, Donald F. 1993. Public administration: The state of the field. In
Political science: State of the discipline II, ed. A. Finifter, 407–27.
Washington, DC: Brookings Institution Press.
Kickert, Walter J. M., Eeik-Hans Klijn, and Joop Koppenjan. 1997. Managing complex networks: Strategies for the public sector. London: Sage.
Lasswell, Harold D. 1936. Politics: Who gets what, when, and how. New
York: McGraw-Hill.
Lester, James P., Ann O’M Bowman, Malcolm L. Goggin, and Laurence J.
O’Toole, Jr. 1987. Public policy implementation: Evolution of the field and
agenda for future research. Policy Studies Review 7:200–16.
Lidström, Anders. 1996. Kommunsystem i Europa. Stockholm, Sweden:
Publica.
Lindgren, Karl-Oskar. 2006. Roads from unemployment. Institutional
complemenatities in product and labor markets. Department of
government, Uppsala University, Sweden.
Lipsky, Michael. 1980. Street-level bureaucracy: Dilemmas of the individual
in public services. New York: Russel Sage.
Lundin, Daniela. 2004. Vad styr arbetsförmedlarna? Rapport 16, Institute for
Labour Market Policy Evaluation, Uppsala, Sweden.
Lundin, Martin. 1999. Arbetsförmedlingsnämnderna – en implementeringsstudie. Stencil 6, Institute for Labour Market Policy Evaluation, Uppsala,
Sweden.
Lundin, Martin, and Per Skedinger. 2006. Decentralisation of active labour
market policy: The case of Swedish local employment service committees.
Journal of Public Economics 90(4-5):775–98.
Lundqvist, Lennart J. 2001. Implementation from above: The ecology of
power in Sweden’s environmental governance. Governance 14(3):319–37.
36
The conditions for multi-level governance
Matland, Richard E. 1995. Synthesizing the implementation literature. The
ambiguity-conflict model of policy implementation. Journal of Public
Administration Research and Theory 5(2):145–74.
Mazmanian, Daniel A., and Sabatier, Paul. 1989. Implementation and public
policy. Lanham, MD: Univ. Press of America.
Molander, Per. 2003. Staten och kommunerna. Stockholm, Sweden: SNS
Förlag.
Montin, Stig. 2002. Moderna kommuner. Malmö, Sweden: Liber.
Nyberg, Sten, and Per Skedinger. 1998. Arbetsförmedlingarna – mål och
drivkrafter. Ds 1998:16. Stockholm, Sweden: Ministry of Finance.
Norton, Alan. 1994. International handbook of local and regional government. A comparative analysis of advanced democracies. Aldershot: Edward
Elgar.
Oates, Wallace E. 1972. Fiscal federalism. New York: Harcourt Brace
Jovanovich.
Olli Segendorf, Åsa. 2003. Arbetsmarknadspolitiskt kalendarium II.
Rapport 9, Institute for Labour Market Policy Evaluation, Uppsala,
Sweden.
Oskarsson, Sven. 2003. The fate of organized labor. Explaining
unionization, wage inequalities, and strikes across time and space. Department of government, Uppsala University, Sweden.
O’Toole, Laurence J. Jr. 1983. Interorganizational co-operation and the
implementation of labour market training policies: Sweden and the Federal
Republic of Germany. Organization Studies 4(2):129–50.
———. 2003. Interorganizational relations in implementation. In Handbook
of public administration, ed. J. Pierre and B. G. Peters, 234–44. London:
Sage.
———. 2004. The theory-practice issue in policy implementation research.
Public Administration 82(2):309–29.
O’Toole, Laurence J., Jr., and Robert S. Montjoy. 1984. Interorganizational
policy implementation: A theoretical perspective. Public Administration
Review 44(6):491-503.
Petersson, Olof. 1994. Kommunalpolitik. Stockholm, Sweden: Publica.
Petersson, Olof. 2003. Folkstyrelse och självstyrelse. In Staten och
kommunerna, ed. P. Molander, 150–58. Stockholm, Sweden: SNS Förlag.
37
Introduction
Pettersson-Lidbom, Per. 2006. Do parties matter for economic outcomes? A
regression-discontinuity approach. Unpublished manuscript. Stockholm,
Sweden: Department of Economics, Stockholm Univ
Peters, B. Guy, and Jon Pierre. 2001. Developments in intergovernmental
relations: toward multi-level governance. Policy & Politics 29(2):131–5.
Pierre, Jon (ed.). 2000. Debating governance. Oxford, UK: Oxford Univ.
Press.
Pierre, Jon, and B. Guy Peters. 2000. Governance, politics and the state.
Basingstoke, UK: MacMillan Press Ltd.
Pressman, Jeffery L., and Aaron Wildavsky. 1984. Implementation. How
great expectations in Washington are dashed in Oakland; or, why it's
amazing that federal programs work at all, this being a saga of the
economic development administration as told by two sympathetic observers
who seek to build morals on a foundation of ruined hopes. Los Angeles,
CA: Univ. California Press.
Radin, Beryl A. 2003. The instruments of intergovernmental management. In
Handbook of public administration, ed. J. Pierre and B. G. Peters, 607–18.
London: Sage.
Rhodes, R.A.W. 1997. Understanding governance. Policy networks, governance, reflexivity and accountability. Buckingham, PA: Open Univ. Press.
Rhodes, R.A.W. 1999. Control and power in central-local government
relations. Aldershot, UK: Ashgate.
Rothstein, Bo. 1994. Vad bör staten göra? Om välfärdsstaten moraliska och
politiska logik. Stockholm, Sweden: SNS Förlag.
———. 1996. The social democratic state. The Swedish model and the
bureaucratic problems of social reforms. Pittsburgh, PA: Univ. Pittsburgh
Press.
Runeson, Caroline. 2004. Arbetsmarknadspolitisk översikt 2003. Rapport
12, Institute for Labour Market Policy Evaluation, Uppsala, Sweden.
Salonen, Tapio, and Rickard Ulmestig. 2004. Nedersta trappsteget. En
studie om kommunal aktivering. Växjö, Sweden: Växjö Univ.
Scholz, John T., and Feng Heng Wei. 1986. Regulatory enforcement in a
federalist system. American Political Science Review 80(4):1249–70.
Smith, Andy. 2003. Multi-level governance: What it is and how it can be
studied. In Handbook of public administration, ed. J. Pierre and B. G.
Peters, 619–28. London: Sage.
38
The conditions for multi-level governance
Stoker, Gerry. 1998. Governance as theory: five propositions. International
Social Science Journal 50(155):17–28.
Swedish Association of Local Authorities. 1999. Kommunerna i arbetsmarknadspolitiken: en enkätundersökning. Stockholm, Sweden: Swedish
Association of Local Authorities.
Thomas, Robert D. 1979. Implementing federal programmes at the local
level. Political Science Quarterly 94(3):419–35.
Thoursie, Ragnar. 1990. Den offentlliga arbetsförmedlingens utveckling
under 1900-talet. In Perspektiv på arbetsförmedlingen, SOU 1990:31, 107–
237. Stockholm, Sweden: Allmänna Förlaget.
Wetterberg, Gunnar. 1997. Kommunerna. Stockholm, Sweden: SNS Förlag.
Winter, Søren C. 2003a. Implementation: Introduction. In Handbook of
public administration, ed. J. Pierre and B. G. Peters, 205–11. London:
Sage.
———. 2003b. Implementation perspectives: Status and reconsideration. In
Handbook of public administration, ed. J. Pierre and B. G. Peters, 212–22.
London: Sage.
Wright, Deil S. 1988. Understanding intergovernmental relations. Pacific
Grove, CA: Brooks/Cole.
Newspapers
Dagens Nyheter. 2002, 2003
Svenska Dagbladet. 2003
Governing documents and other official documents
Government Bills (propositioner)
1994/95:218
1995/96:25
1995/96:148
1995/96:222
1996/97:124
1997/98:1
1999/2000:98
39
Introduction
2002/03:43
2006/07:100
Laws (lagar)
2000:625 (Lag om arbetsmarknadspolitiska program)
2001:453 (Socialtjänstlagen)
Municipality of Arjeplog, Protocol of Local Employment Service Committee
February 21, 1997
October 29, 1997
Ordinances (förordningar)
2001:623 (Förordning med instruktion för Arbetsmarknadsverket)
2000:628 (Förordning om den arbetsmarknadspolitiska verksamheten)
2000:634 (Förordning om arbetsmarknadspolitiska program)
Other documents
Ministry of Employment. En jobbgaranti för ungdomar. Memorandum
April 11, 2007.
40
The conditions for multi-level governance
Appendix: A description of the questionnaire data
The essays housed in this thesis are based on large-N data. Essays II and III
rely heavily on new questionnaire data collected within the research project.
The survey is only sketchily described in the essays. Thus, a somewhat more
detailed description is offered in this appendix.
Design
In February 2004, postal questionnaires were distributed to all PES offices
and municipalities in Sweden.29 Three groups were approached: (i) PES
managers, (ii) managers in charge of municipal labor market activities, and
(iii) municipal politicians with formal responsibility for labor market issues.
The preparation time was about six months and several persons commented
on the questions and the design.30 The questions focused especially on the
relationship between the PES offices and the municipalities, but various
issues of local labor market activities were considered.
The addresses of the PES offices were taken from the webpage
(http://www.ams.se) of the National Labor Market Administration: 366
workplaces were identified. The 290 municipalities were contacted in
advance by telephone in order to locate the responsible politicians and
managers. In the municipality of Stockholm, labor market operations are
handled in 18 different offices organized geographically (kommundelar).
This was recognized after data collection had started. Thus, I administered
the survey by e-mail to the city districts of Stockholm. Three different forms
were designed, one for each group of respondents. Three postal reminders
and one reminder by phone were conducted. Data collection ended in May
2004.
Response rates
The response rates were satisfactory (see Table A1). I removed eight PES
offices from the original population. Some of these offices had been closed
down or merged with other PES offices, while others only had support
functions without responsibility for clients. About 75 percent of the 358
offices remaining in the population participated in the survey. In Stockholm,
56 percent of the municipal managers answered the questionnaire, whereas
85 percent of the managers in other parts of Sweden participated. Finally,
84 percent of the municipal politicians took part in the study.
29
The Institute for Labor Market Policy Evaluation (IFAU) financed the survey. ARS
Research AB collected and registered the data.
30
Many thanks to Hanna Bäck, Hans Ekholm, Jörgen Hermansson, Vivi Libiets, Karl-Oskar
Lindgren, Linus Lindquist, Daniela Lundin, Gunnar Myrberg, PerOla Öberg, Thomas
Persson, and Michael Söderström for valuable comments.
41
Introduction
Table A1. Response Rates
PES
Offices
Managers
Population 1
Municipalities
Managers in
Stockholm
Managers in
other parts
of Sweden
289
Politicians
366
18
8
0
0
0
Population 2
358
18
289
290
Responses
268
10
246
245
Non-responses
90
8
44
45
Response rates (%)
75
56
85
84
Removed from population
290
Are non-responses worrisome?
Although the response rates are good, it is worthwhile taking a closer look at
respondents and non-respondents in order to find out whether it is reasonable
to assume that the missing values have only a negligible impact on the
findings. Below, I examine the four groups presented in Table A1.
A first step was to compare those who actually responded with the
population in terms of background characteristics. If the respondents are
similar to the populations, data are likely to be trustworthy. Table A2 presents information on PES offices, while Table A3 deals with the
municipalities.
Table A2. Background Characteristics: PES Offices (%)
Population of
PES offices
83.9
Responding
PES offices
87.2
23.8
19.2
PES located in rural areas
8.0
9.0
Municipal unemployment rate
(including participants in ALMPs)
5.4
5.5
PES organized by area
PES located in large city (Stockholm,
Gothenburg, or Malmoe)
PES offices may be organized by area (geographically) or by function.
Table A2 reveals that the responding group and the population are organized
in a similar manner, although offices organized by function participate somewhat less. Location is another factor to consider. The sample consists of
fewer offices located in Stockholm, Gothenburg, and Malmoe. There is also
a slight overrepresentation of offices in rural areas. These differences are,
42
The conditions for multi-level governance
however, not that large. Lastly, the average unemployment rate is very much
alike in the respondent group and in the population.
Table A3. Background Characteristics: Municipalities
Population
Managers
Politicians
Large city areas
13.5
14.9
12.8
Rural areas
20.4
20.7
19.3
Municipal unemployment rate
(including participants in ALMPs)
5.3
5.3
5.3
Socialist chairman of the municipal
executive board
59.3
58.4
59.0
Population over 65 years of age
19.0
18.9
19.0
1.9
1.8
1.9
30,830
28,367
32,499
Population with foreign citizenship
Number of inhabitants
Note: The figures are reported in percentages except in the case of “Number of inhabitants.”
Table A3 provides variables that make it possible to compare the
municipalities participating in the survey with the whole population of
municipalities. The responding groups are very similar to the population on
all characteristics.
The telephone reminder made it possible to learn more about the nonresponses (see Table A4). The interviewers’ first task was to remind the
managers and politicians of the questionnaire. Officials that declared that
they were not willing to participate in the survey were given the opportunity
to answer a couple of questions by telephone instead. Around 40 percent of
the PES managers, 43 percent of the municipal managers, and 31 percent of
the politicians agreed to answer telephone questions (these answers are
discussed later on; see Tables A5–A7). Thus, only 54 PES managers, 25
municipal managers, and 31 municipal politicians did not participate at all.
The non-participants are divided into four subgroups. The interviewers
were unable to reach some individuals (no contact), even though they tried
repeatedly. The elected representatives were hardest to get in touch with, but
note that 56 percent only implies 25 persons. Some officials were reminded
about the study and said that they were going to participate, but then never
did (reminded but no questionnaire sent in). Another group consists of those
who did not want to participate at all (not willing to participate). A very frequent reason for not taking part was lack of time. There is no indication that
a high level of disputes between the authorities is an important reason for not
participating, which is good. The last group of non-participants is tiny: vacant position, newly employed, or illness.
43
Introduction
Overall, a heavy workload is probably a common reason for not
participating. Admittedly, this could be a problem for data quality. It is
difficult to have a specific opinion on whether, and if so in what direction,
this implies biases. But recall that the non-participants are few. Accordingly,
it is highly unlikely that the results are distorted to a significant extent.
Table A4. Information on Non-responses Collected through Telephone Calls (%)
PES offices
Managers
Municipalities
Managers
Politicians
Non-participants
60
57
69
No contact
(17)
(0)
(56)
Reminded, but no questionnaire sent in
(20)
(21)
(2)
Not willing to participate
(20)
(36)
(11)
(3)
(0)
(0)
40
43
31
100
n = 90
100
n = 44
100
n = 45
Vacant position, newly employed, or
illness
Responding to phone questions
In Tables A5–A7, the results from the telephone questions are reported. The
purpose was to be able to compare the answers from the telephone interviews with the answers to the same questions in the postal questionnaire: if
the responses differed systematically, the reasons for being skeptical towards
the data would increase. It is reasonable to assume that the answers are relatively comparable, although telephone interviews and postal surveys are
not exactly the same thing.
Table A5 shows the results in the group of PES managers (36 managers
answered the telephone questions). The first question concerned the number
of employees. The offices are about the same size. The respondents were
also asked if a certain caseworker was assigned the task of handling youth
clients and the Activity Guarantee.31 The telephone calls and the postal
questionnaire showed similar results. Respondents reported frequency of
communication and the answers, once again, did not diverge to any noticeable extent. Moreover, the share of offices that had signed a collaborative
contract with the municipality concerning youth and the Activity Guarantee
was also more or less identical. Lastly, the respondents in both groups
reported the same levels of conflict.
31
The Activity Guarantee is a labor market program for clients with severe difficulties on the
labor market; see Essay III for details on youth programs and the Activity Guarantee.
44
The conditions for multi-level governance
Table A5. PES Managers’ Responses to Some Questions on Local Labor Market
Policy (%)
Postal
questionnaire
22
82
Telephone
interviews
25
81
Caseworker with responsibilities for the
Activity Guarantee
92
86
Daily contacts with the municipality
51
61
Cooperative contract with the
municipality: youth clients
77
78
Cooperative contract with the
municipality: Activity Guarantee
78
72
No conflicts between the PES and the
municipality in 2003
72
78
Average number of employees
Caseworker with responsibilities for
unemployed youth
Table A6 reports the answers from municipal managers (19 managers
participated). Three questions were asked.32 The findings show that the share
of municipalities with a special labor market administration is lower,
communication between the municipality and the PES is less frequent, and
conflicts are fewer according to the participants in the telephone interviews.
One logical interpretation is that municipalities that do not engage that much
in labor market operations decided not to participate in the study. This is not
a cause for concern since the non-responses in this group are so few.
Table A6. Municipal Managers’ Responses to Some Questions on Local Labor
Market Policy (%)
Postal
questionnaire
75
Telephone
interviews
47
Daily contacts with the PES
56
21
No conflicts between the PES and the
municipality in 2003
48
74
Existence of a labor market
administration
32
Actually, a fourth question concerned the number of municipal employees working on labor
market issues. It is obvious that the respondents interpreted this question very differently in
the postal questionnaire. Thus, I have decided not to use this information in the essays within
this thesis. But it can be mentioned that the average number of employees noted in the postal
questionnaire is 12, while in the telephone interviews it is 9.
45
Introduction
Table A7. Municipal Politicians’ Responses to Some Questions on Local Labor
Market Policy (%)
Postal
questionnaire
56
Social democrat
Telephone
interviews
79
Communicates with the PES at least every month
84
79
Average number of hours devoted to labor market issues
every month
Political parties at local level agree on labor market issues
to a “fairly high” or “very high” extent
17
14
96
100
No conflicts between the PES and the municipality in
2003
67
79
Note: The figures are reported in percentages except in the case of “average number of hours devoted to
labor market issues every month”.
Table A7 shows the answers from municipal politicians (14 persons
answered the telephone questions). The telephone respondents reported less
communication with the PES, fewer hours per month devoted to labor
market issues, and a lower level of conflict between the PES and the
municipality. Thus, politicians who did not participate in the postal
questionnaire were probably less involved in ALMPs. The differences are
not as obvious as in Table A6 and there is no strong reason to assume that
the non-responses imply a considerable setback.
Table A8. Background Characteristics of Districts within Stockholm (All 18
Districts and the 10 Responding Districts)
Population
Respondents
Average age (years)
38.7
38.2
Unemployment (%)
3.1
3.3
Percentage of inhabitants receiving welfare
benefits
6.7
8.2
11.1
13.0
42,119
38,358
Percentage of inhabitants with foreign citizenship
Number of inhabitants
Note: The calculations are based on figures taken from City of Stockholm (2004), Statistisk årsbok för
Stockholm 2004, available at www.stockholm.se. Participants in ALMPs are not included in the unemployment rate.
Because the data collection in Stockholm was conducted separately,
telephone reminders were not possible. Nevertheless, it is possible to compare some important characteristics of the population of city districts and the
districts participating in the survey. Table A8 indicates that the respondents
are quite similar to the population as a whole. To some extent, districts with
46
The conditions for multi-level governance
many welfare benefit recipients and inhabitants with foreign citizenship
participated more.
To sum up, response rates were good. A detailed analysis shows that nonresponses probably do not constitute a major problem. In the group in which
non-responses are most frequent—among the PES offices—all analyses
indicate that missing values are not a problem. Among the municipalities, it
seems that those who did not answer the questionnaire are a little bit less
involved in ALMPs. But since response rates are very good, this is not
particularly worrisome.
47
Essay I
Political Partisanship and Entity Size: When
Parties Matter for Public Policy*
Do political parties make a difference for public policy? Does it matter
whether the left wing or the right wing governs? This puzzle has occupied
the attention of numerous scholars in political science and related research
disciplines for a long time. Cross-national, national, and subnational data
from various policy areas have been employed to evaluate the importance of
political parties. Some studies indicate an effect (for example, Besley and
Case 2003; Korpi and Palme 2003; Petterson-Lidbom 2006), while others
reject the idea that partisanship matters (for example, Ashworth and
Heyndels 2002; Faust and Irons 1999; Garand 1988). It is sometimes
suggested that the effect of partisanship is dependent on other factors
(Brown 1995; Imbeau, Pétry, and Lamari 2001; Schmidt 1996). That is,
parties may in some settings be of minor importance for output, but make a
significant difference given other circumstances. This article contributes to
the literature by studying one structural factor that might influence the effect
of political partisanship on local government actions—the size of the local
entity.
Local political entities within a political system have different sizes. For
instance, in Sweden the lowest governmental tier (the municipal level) comprises between approximately 2,500 and 775,000 inhabitants. It is reasonable
to assume that politics looks rather different in small and large localities. The
basic idea in this article is that politics, for several reasons, is usually more
pragmatic, and party antagonisms are less pronounced, in small subnational
entities. Thus, it is hypothesized that the effect of partisanship is larger in
sizeable localities.
The empirical analysis concerns local government involvement in
Swedish active labor market policy (ALMP). As in most other Western
democracies, ALMPs are primarily a central government affair in Sweden. A
large central government administration (the National Labor Market
Administration) with local branches (Public Employment Service offices,
*
Helpful comments were given by Paula Blomqvist, Matz Dahlberg, Jörgen Hermansson,
Nils Hertting, Karl-Oskar Lindgren, PerOla Öberg, and Sven Oskarsson.
49
Essay I
PES) implements political decisions on the ground. But local governments
are also involved in local actions to a considerable extent (Carling and
Larsson 2005; Lundin and Skedinger 2006; Lundin 2007; Salonen and
Ulmestig 2004). ALMPs mainly benefit voters of left-wing parties. Thus, it
is reasonable to expect local governments in which socialist parties are
strong to be more engaged in the efforts to decrease unemployment. But if
politics are more pragmatic in small entities, and if party politicization becomes more important as entity size increases, a larger partisanship effect is
probable in areas in which many people live.
The analysis is based on quantitative panel data, 1998–2005. This data set
makes it possible to carry out rigorous empirical tests. The analysis indicates
that the effect of partisanship on local government involvement in ALMPs
really is contingent on entity size. Political partisanship makes no difference
in very small localities. As population size increases, partisanship begins to
play a more significant role. In the group of local governments in which the
population is really large, it matters a great deal who governs.
This research contributes to at least two important discussions in the
literature. First, it suggests and tests one condition that can help us improve
theory on when political partisanship makes a difference for public policy. It
seems as if entity size—at least at local level—should be interacted with
political partisanship. Ignoring interactions between partisanship and size
could be a reason why some scholars have failed to identify a partisan effect.
Second, within the “partisanship matters”-discourse some studies have
analyzed the effects of partisanship on ALMPs using cross-national data
(Hicks and Kenworthy 1998; Rueda 2005; Swank and Martin 2001; Vernby
2006). The findings are mixed. This might be a consequence of methodological problems. In the present study, the number of observations is increased and I am able to hold country-specific attributes constant. Thus, the
empirical test is probably more reliable. The findings suggest that scholars
such as Rueda (2005) are wrong to claim that political partisanship does not
affect the use of ALMPs.
The study is organized as follows: First, the theory is outlined. Second,
the involvement of local governments in ALMPs is introduced. Third, I
present the statistical model and measures. Fourth, findings are reported. I
conclude by summing up the results and discussing possible implications.
Political partisanship and the size of local entities
Political partisanship has been at the core of political science research for a
long time. Studies such as Alt (1985), Downs (1957), Hibbs (1977), Korpi
(1974), and Strom (1990) discuss why and to what extent we should expect
parties to make a difference for public policy. It is generally agreed that the
left–right political dimension is the most important aspect of political life.
50
Political partisanship and entity size
Basically, the left–right dimension boils down to views on the role of the
government versus the market (Blais, Blake, and Dion 1993). Downscale
social groups are the core constituency of the left wing, while the core
constituency of the right wing is upscale groups. Parties are assumed to have
goals related to those of their main voters.1 Accordingly, leftist governments
will endorse the interests of labor, whereas right-wing ones will promote the
interests of the self-employed and employed managers.
Left-wing parties and voters are normally assumed to be more positive
towards government intervention (cf. Hibbs 1977). But why does the left
have a more positive view of government intervention? Welfare policy can
serve as an example. Korpi (2006, 173) holds that “while some life-course
risks are generated by universal processes, most risks, such as those
associated with aging, illness, work accidents, unemployment, poverty, and
rearing of children, are instead unequally distributed among individuals
occupying different positions in socioeconomic class structures.” The core
voters of socialist parties are more exposed to most of these risks. In
addition, they also tend to have fewer resources to cope with them.
Accordingly, reducing the importance of market capacities for handling lifecourse risks is in the interest of the core constituency of the left. Government
intervention, in the form of ambitious welfare programs, is therefore likely to
be put into practice when the left wing holds office. On the other hand,
upscale groups control major economic resources. They are expected to
prefer market solutions to handle distributive processes since they have the
economic capacity to participate in such markets. Consequently, low taxes
and a relatively small welfare state are objectives of right-wing parties
(Korpi 1974; 2006; see also Allan and Scruggs 2004 for empirical
evidence).2
According to official documents, such as party manifestos, parties
officially pursue different political objectives (Budge 2001). But empirical
research seeking to estimate the impact of partisanship has produced mixed
results. This is especially true if we consider studies focusing on national
politics.3 On the other hand, recent research conducted on subnational data
seems to corroborate the “partisan matters”-argument to a greater extent. For
1
Notice that the median voter theorem proposed by Downs (1957) suggests that votemaximizing parties converge to the center in a two-party system (but not necessarily in multiparty systems). Thus, policies will be characterized only by the preferences of the median
voter and the partisanship impact on public policy is likely to be tiny. But others have
suggested that parties are not only vote seekers, but also policy seekers (for example, Strom
1990). Accordingly, in the words of Blais, Blake, and Dion (1993), it is “probably fair to
predict that in most cases parties will propose moderately different policy packages” (41), and
we should expect that “parties matter, but only moderately” (42).
2
Kato and Rothstein (2006), Pierson (1996), Swenson (1991), and Ross (2000) provide
alternative views.
3
For an overview see Blais, Blake, and Dion (1993). For a meta-analysis, see Imbeau, Pétry,
and Lamari (2001).
51
Essay I
example, analyzing Swedish municipalities, Petterson-Lidbom (2006) convincingly shows that left-wing governments spend and tax more than rightwing governments. Blom-Hansen, Monkerud, and Sorensen (2006) examine
revenue policies in Danish and Norwegian municipalities and find that the
left–right dimension explains local revenue policies to some extent, especially in the Danish case. Allers, De Haan, and Sterks (2001) also find an effect
of political partisanship on local taxes based on data from Dutch local
governments. Besley and Case (2003) go through studies on US state data.
They suggest that whereas older studies usually reject the partisanship
hypothesis, recent studies support the argument to a greater extent.
An interesting contribution is Brown’s (1995) study of welfare efforts in
US states. Brown argues that the parties’ voter basis varies among the states;
partisan cleavages look different in different environments. He predicts and
finds a large impact of partisanship on welfare efforts only in the group of
states where the Democratic Party are supported mainly by Catholic, lowincome, union, and female groups. What is interesting about the study for
our present research is that Brown suggests that political realities are different in different subnational entities. In turn, this should lead us to assume
that the influence of partisanship varies with other structural factors.4
In this article, in which the focus is the local political level, our main
interest is a possible interaction term of partisanship and local entity size. By
size I mean the number of local inhabitants. My claim is that politics is more
pragmatic in small localities; parties agree to a greater extent and we should
not assume that political partisanship makes a major difference. The impact
of partisanship is expected to increase with population size; pragmatism decreases, opinions diverge, and, as a consequence, it matters more how power
is distributed along the left–right political dimension.
Why should we assume that pragmatism is more frequent in small local
entities? A number of arguments are relevant. First, in large localities the
representatives decide on bigger issues that concern more people and involve
more money. Party organizations at local level—and at higher levels of
government as well—are likely to be more interested in these distributive
processes than when decisions are less important. Greater activity within
political parties makes it more difficult to agree across party boundaries.
Second, Dahl and Tufte (1973) argue that an increase in size implies an
increase in attitudinal diversity. Accordingly, the population is expected to
be more homogenous in small entities. This is likely to imply less antagonism among political parties. Most certainly, feelings of togetherness
produce pragmatism. Third, specialization and internal diffusion of responsibilities are greater in large localities (see for example, Petersson 1994,
4
In addition, Schmidt (1996) and Imbreau, Pétry, and Lamari (2001) propose that interactions
between partisanship and other structural factors should have a more prominent place in
future research on the effects of partisanship.
52
Political partisanship and entity size
49), and it is therefore reasonable to assume that elected representatives are
less specialized in small entities, which probably increases the influence of
the bureaucracy (and perhaps also organized interests). This may lead to
party ideology having less impact on decisions. Fourth, in small entities the
politicians are closer to the level at which things actually happen. Pragmatism may evolve as a consequence of greater contact with inhabitants
affected by decisions. In areas in which many people live, there is a greater
distance between elected representatives and street-level actions. Fifth, decision making normally involves fewer individuals in small entities. This
could generate political processes that resemble rational deliberation rather
than negotiations. If decisions are made in a deliberative manner, it is likely
that the influence of party ideology will be somewhat less.
There is empirical evidence indicating that party-politicization is greater
in large local entities. For instance, in Sweden the number of local governments has decreased significantly, from about 2,500 in 1952 to 290 in 2007.
An extensive research program has evaluated the major reform between
1962 and 1974: one of the most important conclusions was that partypoliticization had increased, for example, in the form of more conflicts of
interest between political parties (Strömberg and Westerståhl 1983).
Moreover, a recent study on size and democracy on Swedish municipal data
shows that clashes of opinion are more frequent among political parties in
large than in small localities (Johansson and Karlsson, forthcoming). Given
the theoretical arguments and the empirical evidence, a greater impact of
political partisanship should be expected in sizeable entities.
Local governments and active labor market policy in
Sweden
Swedish local governments (or municipalities) have the constitutional right
of local self-government and an essential political and economic role. In
comparative perspective, local governments in Sweden are often considered
important (Lidström 1996). The municipalities are relatively large, with an
average population of approximately 30,000 citizens, and provide a lot of
services of the welfare state, including day care, care of the elderly, social
welfare services, and primary education. Their revenues come mostly from a
proportional income tax, which they can set freely. In addition,
municipalities decide on their own organization and have no restrictions on
borrowing. Municipalities also employ a large share of the Swedish
workforce, approximately 20 percent.5
5
For descriptions of Swedish local politics, see Hanna Bäck (2003), Henry Bäck (2003), and
Montin (2002).
53
Essay I
When it comes to local party politics in Sweden, Henry Bäck (2003)
argues that parties have an important role in local governance and that the
left–right dimension dominates political life. In addition, Hanna Bäck (2003,
63–65) shows convincingly that a left–right policy dimension describes local
politics well, although there is also a green dimension. Nevertheless, some
case studies indicate that ideological overtones sometimes are not that
pronounced; consensus among politicians across party boundaries is not rare
(for example, Henry Bäck 2003).
Municipalities implement many policies within the welfare sector. But
ALMPs are mainly a responsibility of the PES offices, located within the
National Labor Market Administration. ALMP refers to activities, such as
labor market programs and job search assistance, targeted at the unemployed
with the aim of increasing employment. A range of different programs are
available, including labor market training, work practice, and subsidized
employment (Calmfors, Forslund, and Hemström 2002). In comparative
perspective, Sweden spends a lot on ALMPs (Martin and Grubb 2001).
Local resources, such as information and expertise, are important when
ALMPs are carried out in practice (cf. O’Toole 1983). For instance, in order
to be successful the PES offices need to know local development plans and
what kind of training to provide. The PES offices also need places of work
where unemployed clients can apply for jobs or participate in on-the-job
training. In addition, labor market activities have to be coordinated with
other policy areas, such as social policy, policies for immigrants, local
development, and education—policies that are all municipal responsibilities.
In short, PES offices are dependent on the local environment in carrying out
their tasks. In these local environments, municipalities are central actors.
The municipalities are engaged in ALMPs to a considerable extent, and
most of them have instigated labor market administrations (Swedish
Association of Local Authorities 1999). Municipalities take part in various
collaborative endeavors together with the PES offices (Lundin 2007) and, in
fact, they arrange the activities for approximately 40 percent of the participants in the National Labor Market Administration’s active measures (Lundin
and Skedinger 2006). Diverse youth programs (Carling and Larsson 2005)
and programs for social allowance recipients (Salonen and Ulmestig 2004)
are examples of activities in which the municipalities are heavily involved.
But why are municipalities engaged in ALMPs? Wetterberg (1997, 91)
provides three key reasons. First, unemployment has a significant impact on
the municipalities’ finances since the local tax base decreases as unemployment increases. This can be especially critical if unemployment also leads to
out-migration. Second, the municipalities have to bear costs when the unemployment rate increases, for example, in the form of increased expenditures for social allowances. Third, unemployment means a worsening
of national public finances. Wetterberg claims that the central government
tries to pass on costs to the municipalities when public finances are bad. This
54
Political partisanship and entity size
0
10
Percent
20
30
gives the municipalities incentives to fight unemployment. And it would be
easy to come up with additional problems for municipalities brought about
by unemployment.6
0
500
1000
1500
Expenditures (crowns per inhabitant)
2000
Figure 1. The Variation in Municipal Expenditures on ALMPs in 2005 (Source:
Statistics Sweden).
The extent to which municipalities are involved in ALMPs is not the same
everywhere in Sweden. For example, Lundin (2007) shows that cooperation
between PES offices and municipalities varies. Moreover, local government
expenditures on ALMPs fluctuate. Figure 1 shows the variation in local
government spending on labor market programs in 2005. The bulk of
municipalities use between 500 and 750 crowns per capita on ALMPs, but
there is a large variation, ranging from approximately 20 crowns to 2,220
crowns per citizen.
It is reasonable to expect that political partisanship is one factor explaining the pattern in Figure 1. ALMPs have always been associated with social
democracy in Sweden. The National Labor Market Administration was built
up by the social democratic government around 1950, and the recruitment
and training policy secured a staff that was committed to the overall
objectives stipulated by the Social Democrats (Rothstein 1996). Leftist
6
For instance, evidence from Sweden indicates that unemployment leads to an increase in
crime (Nilsson and Agell 2003).
55
Essay I
governments have emphasized large-scale programs and a lot of financial
resources have been allocated to the National Labor Market Administration.
With assistance from the Left Party, the Social Democrats have been in
power at central level for long periods in Sweden: between 1956 and 2006
the Prime Minister was a social democrat in 41 out of 50 years. Consequently, they have had a great deal of influence over labor market policy.
On the other hand, right-wing parties, mostly in opposition, have argued
for cutting down on ALMPs, privatizing some parts of the policy, increasing
incentives to take a job by decreasing unemployment benefits and taxes on
work, lowering payroll taxes, and making laws on job security more flexible.
This can be illustrated by the actions of the right-wing coalition that came to
power after the election success in autumn 2006. Quickly after gaining
office, the coalition of the Moderate Party, the Centrist Party, the Liberal
Party, and the Christian Democratic Party decreased unemployment
insurance benefits (Government Bill 2006/07:15) and the budget for ALMPs
(Government Bill 2006/07:1).
Table 1. Opinions on ALMP Expenditures in Sweden 2002 (%)
Party sympathy
The amount of tax money used
for ALMPs should be...
Increased
Unchanged
Decreased
Total
Left-wing
46
46
8
100
(370)
Right-wing
25
40
35
100
(274)
Note: The table is based on the data set Opinions on the public sector and the taxes 2002 (Åsikter om den
offentliga sektorn och skatterna 2002) collected by Stefan Svallfors and Jonas Edlund, Department of
Sociology, Umeå University. Data have been made available through the Swedish Social Science Data
Service (SSD). The original researchers and SSD are not responsible for the analysis and the
interpretation of the results. The survey was conducted among 2,000 individuals and the response rate
was about 57 percent. Left is defined as persons who think that the Social Democratic Party or the Left
Party is the best party. Right is defined as persons who think that the Moderate Party, the Centrist Party,
the Liberal Party, or the Christian Democratic Party is the best party. Individuals who do not have
sympathy for a certain party and individuals who adhere to a party other than those mentioned in the table
are excluded from the analysis. The chi-square statistics indicate that a significant relationship exists (at
least at the .01 level) between party sympathy and opinions about ALMP spending. Another question in
the survey concerned ALMPs: “Would you be ready to pay more tax if you were sure that the money
would be devoted to creating new jobs?” The findings are very similar if the analysis is based on this
question instead.
Another indication that ALMP is a policy area involving a left–right dimension is provided in Table 1. The table displays opinions on ALMP spending
among a representative sample of Swedish inhabitants in 2002. It is obvious
that persons who sympathize with the left are more positive towards
ALMPs. More than four times as many of the persons that adhere to the right
wing think that expenditures on ALMP should be reduced. In the next
56
Political partisanship and entity size
section, I examine whether partisanship influences policy in reality and if the
effect is contingent on population size.
Method and measures
The empirical tests are based on a pooled cross-sectional panel design of 276
Swedish municipalities from 1998 to 2005.7 During the time period at issue,
Sweden had a social democratic government. It was a minority government
supported on most questions by the Left Party and the Green Party. Elections
were held at the central and local levels in autumn 1998 and 2002. The total
number of observations is 2,208 and I used Ordinary Least Squares (OLS)
regression to estimate the model. In what follows, I present the variables and
the statistical modeling strategy. Descriptive statistics are reported in
Table 2.8
To assess local government involvement in ALMPs, the amount of
financial resources the municipalities devote to labor market activities is
utilized. The variable is measured in Swedish crowns per inhabitant.9 It is
reasonable to assume that the more money the municipalities expend on
ALMPs per capita, the more involved they are in the policy area. The
measure includes all municipal expenditures on ALMPs. The municipalities
can arrange diverse forms of programs. It is not possible to illustrate the
typical municipal program since the duration of programs and their content
vary to a considerable extent. But job search assistance and work practice
programs are common examples. Unemployed youth and social allowance
recipients are often the target groups.
Figure 1 shows that the expenditures on ALMPs are positively skewed.
The observations are not evenly distributed around the mean value—some
municipalities spend much more than others. The pattern is not altered if
other years within the time frame are examined. This means that OLS
assumptions may be violated. In the analyses to follow, I use the natural
logarithm of the expenditures as the dependent variable. This variable, which
I call Log(ALMP spending), is more normally distributed. Nevertheless, the
analyses have been conducted on the unmodified variable as well—the conclusions remain the same, although the findings are somewhat less robust.
7
There are 290 municipalities in Sweden: 14 municipalities are not included in the analysis
due to missing values on some key variables in some years.
8
The local unemployment rate is based on data from the National Labor Market Administration. All other variables are taken from Statistics Sweden.
9
Unfortunately, I do not have data on municipal expenditures on ALMPs in 1998. But
information from 1997 and 1999 is available. By intrapolation—that is, taking the mean value
of 1998 and 1999—reasonable estimations of expenditures in 1998 are obtained. In addition, I
have conducted a number of analyses using data from 1999–2005, and also analyses using
data from 1997 and 1999–2005. The conclusions are the same, although some findings are
less robust.
57
Essay I
The main independent variable is partisanship. If we assume that parties
can give fuller expression to their ideology as their position becomes
stronger, the proportion of seats held in the local council by the left seems
like a reasonable operationalization of the main independent variable (cf.
Blom-Hansen, Monkerud, and Sorensen 2006). But there are other ways of
measuring the strength of the left and the right. One possibility is to focus on
whether the left wing or the right wing has the majority of seats in the local
council. Another option is to concentrate on what party the chairman of the
local executive board represents. In this article, the percentage of seats in the
local council is utilized. The other two possible measurements are employed
as robustness checks.
But what parties should be considered leftist parties? The standard
approach when studying Swedish municipalities is to treat the Social
Democratic Party and the Left Party as socialist parties (cf. PetterssonLidbom 2006). The right-wing bloc consists of the Conservative Party, the
Centrist Party, the Liberal Party, and the Christian Democratic Party. I rely
on this classification. Accordingly, partisanship is defined as the percentage
of seats in the local council held by the Social Democratic Party and the Left
Party.10 The more seats these parties control, the stronger the left wing is
assumed to be. Another thing must be noted: elections were held in 1998 and
2002. Thus, there are only two temporal shifts in the variable. This means
some restrictions on the analyses to follow, which I will return to later.
Population size is an important variable in this study. Thus, the natural
logarithm of the number of citizens (in thousands of inhabitants) is included
in the empirical model. It is logical to use the logarithm since a one-unit
change in the number of inhabitants is probably more important when the
municipality is small than when it is large. I call the variable Log(size). In
order to find out whether the effect of partisanship is dependent on entity
size, an interaction term of the variables partisanship and log(size) is also
computed.
Several control variables are added. Local needs may explain municipal
engagement in ALMPs. The most obvious control is therefore unemploymentt-1. This is the percentage of municipal citizens, aged 16–64, who are
unemployed or taking part in labor market programs. It is an annual average
based on monthly data. In order to lessen the problems of causal order, the
10
The Green Party is also represented at the local level. Although the Green Party has
collaborated with the Social Democratic Party at the central level to a large extent, it is not a
socialist party in the same sense as the Social Democratic Party and the Left Party. But I have
conducted a number of analyses in which the Green Party is treated as a socialist party. Other
analyses include a dummy variable when the Green Party holds the balance of power in the
local council. Results are robust. In some local councils, local parties control some of the
seats. These parties are usually small and often one-issue parties. Sometimes they hold the
balance of power. I am not able to classify the local parties along the left–right dimension, but
including a dummy variable measuring whether there is a local party that holds the balance of
power does not change the results substantially.
58
Political partisanship and entity size
unemployment rate is lagged one year.11 The unemployment rate is normally
included in studies on expenditures on ALMPs based on cross-national data
(for example, Rueda 2005; Swank and Martin 2001; Vernby 2006).
Table 2. Descriptive Statistics
Variables
ALMP Spending
Mean /
Proportion
555.52
(Log)ALMP Spending
Partisanship
SD
Min
Max
327.23
20.00
3036.00
6.13
0.65
3.00
8.02
47.25
11.44
11.11
82.86
(Log)Size
2.92
0.89
0.94
6.67
Unemployment
7.18
3.24
1.30
23.81
Low education
23.30
6.01
4.85
41.75
Elderly inhabitants
19.03
3.68
7.74
29.99
Female
50.00
0.83
45.99
52.96
Foreign
3.97
2.77
0.70
29.26
Metropolitan
0.14
–
–
–
Sparsely populated
0.11
–
–
–
Population change
99.81
0.97
94.01
104.48
Social allowances
7.68
4.13
1.01
34.78
108.21
19.03
71.45
256.75
6.61
4.50
-15.52
23.94
Tax income
State subsidy
Note: Proportions are given in italics.
A number of measures assess the structure of the local population. The
demand for services may differ depending on what inhabitants live in the
municipality. The various political parties also have different levels of
support in different population groups. I incorporate low education, which is
the percentage of citizens aged 25–64 with an educational level lower than
high school. The percentage of the population 65 years of age or older
(elderly inhabitants), the percentage of females, and the percentage of
inhabitants with foreign citizenship are additional controls.
In the districts of the three largest cities in Sweden—Stockholm,
Gothenburg, and Malmoe—the labor market looks rather different than in
11
Note that the population, rather than the work force, is the denominator when the National
Labor Market Administration calculates unemployment figures based on their administrative
system. The official unemployment figures, provided by Statistics Sweden, are based on the
work force. But these figures are not available at the municipal level. The population 16–64
years of age is, of course, larger than the work force. Thus, the unemployment rate (including
program participants) is smaller than the figures usually discussed. This does not affect the
analysis in this study.
59
Essay I
the rest of the country. A considerable part of the Swedish population lives
in these metropolitan areas and there are probably many other local actors
besides the local government with the ability and interest to engage in
ALMPs. This might imply that local governments are less involved in
ALMPs here than in the rest of the country, all else being equal. A dummy
variable metropolitan is therefore entered into the model. Metropolitan areas
include Stockholm, Gothenburg, and Malmoe, and suburbs in which at least
50 percent of the population commutes to one of these cities. On similar
grounds, I control for sparsely populated areas. In these areas there are
probably very few alternative actors to the local government. Municipalities
with less than 20,000 inhabitants, with less than five inhabitants per square
kilometer, are considered sparsely populated.
Population change is another control variable. It is possible that
municipalities facing a decline in population engage in ALMPs in order to
stop out-migration (for a discussion, see Lundin and Skedinger 2006).
Population change is defined as the population in a given year divided by the
population the preceding year, multiplied by 100. Thus, a value of 100
implies no change in population, whereas values above 100 mean an increasing, and values below 100 a decreasing population.
I also incorporate expenditures for social allowancest-1 in hundred crowns
per citizen as a control. ALMPs may be used to reduce the number of persons living on social allowances and, as a consequence, decrease such costs
for the municipalities (Lundin and Skedinger 2006). Accordingly, large
social allowance expenditures one year may trigger municipal involvement
in labor market activities the next year.
The control variables presented above can together be classified as
diverse forms of local needs. Besides needs, municipal capacity to engage in
ALMPs may be important. More resources imply a greater ability to get
involved. The variables tax incomet-1 (the tax income per inhabitant in 1,000
Swedish crowns) and state subsidyt-1 (general subsidies from the central
government in 1,000 Swedish crowns per inhabitant) are therefore entered
into the model specification. Both variables are lagged one year to reduce
problems of causal order.
The basic statistical model also includes year dummies to deal with time
effects. The year dummies take care of inflation, macroeconomic shocks,
and other problems of contemporaneous correlation.
The pooled cross-sectional design implies several advantages. For example, more observations can be utilized and rigorous model specifications
are possible. Another benefit is that it is possible to control for unit specific
effects by including a full set of municipal dummy variables (that is, the
fixed effects approach). Fixed effects take all hard-to-measure variables that
60
Political partisanship and entity size
do not change over time into account.12 One problem is that fixed effects
consume a lot of degrees of freedom, which is especially cumbersome if key
variables do not vary that much over time (see, for example, Gujarati 2003;
King 2001). This is precisely the case here. It is therefore possible that there
is not enough variation over time in the political partisanship to detect a
statistically significant effect. In this article, I present models both with and
without fixed effects.
Two common problems with pooled cross-sectional data are
autocorrelation and panel heteroscedasticity. This means that significance
testing becomes unreliable. I use three different approaches to counter autocorrelation and heteroscedasticity. Robust standard errors clustered on
municipalities (Froot 1989; Kezdi 2004) and standard errors based Newey
and West’s heteroscedasticity and autocorrelation consistent covariance
matrices (Newey and West 1987) with two lags13 are the first two methods.
The third method is to control for the amount of resources devoted to
ALMPs in the preceding year, that is, the lagged dependent variable. This
eliminates problems of autocorrelation. In the specifications that include the
lagged dependent variable, robust standard errors14 are used to reduce the
problems of heteroscedasticity.
The model specifications that include the lagged dependent variable
imply very hard statistical tests, especially since the temporal variation
within municipalities in the main independent variable is very small. This
means that we are running a large risk of rejecting a hypothesis that we
should not reject, only because there is too little variation in partisanship.
Furthermore, adding a lagged dependent variable means that the analysis
focuses on short time effects, despite the fact that we are more interested in a
long-term equilibrium.15 It is likely that partisanship does not account for
small annual changes, but makes a difference in the long run. Including the
lagged dependent variable is therefore probably too demanding a test. But by
12
An alternative to fixed effects is the so-called random effects approach. This approach
demands additional assumptions that are rarely tenable. A Hausman test indicated that random
effects is not an appropriate method in the present research setting.
13
A rule of thumb suggested by Greene (2003, 267) is used to decide the number of lags.
Findings are robust if the number of lags is set to three instead of two.
14
Instead of using robust standard errors, panel-corrected standard errors may be employed
(Beck and Katz 1995). But simulation studies reveal that the panel-corrected standard errors
perform poorly when there are few time periods and many cross-sections—robust standard
errors are preferred in these situations (Moene and Wallerstein 2003). The present study is
based on a relatively short time series. Accordingly, robust standard errors are more reliable.
Note that findings are not changed if panel-corrected standard errors are used instead.
15
It is, however, possible to calculate the long-run effect of a permanent change in leftist
mandates. This effect is given by βx /(1-βy-1), where βx is the regression coefficient of
partisanship and βy-1 is the regression coefficient of the lagged dependent variable. For
instance, given the results in Model 3a in Table 3, the long-run effect of partisanship is 0.9
percent.
61
Essay I
showing the results from various specifications we can get a sense of the
robustness of the findings.16
Findings
The empirical results are reported in Tables 3 and 4. In Table 3, it is assumed
that the effect of partisanship is not affected by the size of the local entity.
Table 4 examines whether the impact of political partisanship differs among
small and large municipalities by including an interaction term between
political partisanship and entity size. The regression coefficients times 100
denote the average change in percent in ALMP spending, given a one-unit
increase in the independent variable, when all other variables in the model
are held constant. Note that the specifications that include the lagged dependent variable illustrate short-run effects.
In Models 1a–3a, there is a positive and statistically significant (at least at
the .05 level) impact of partisanship. This means that the effect is statistically significant even when I control for the expenditures of the preceding
year. Models 1a and 2a show that an increase of socialist seats in the local
council by one percentage point increases average expenditures by about
0.8 percent, all else being equal. It is always difficult to evaluate the
substantial importance of regression coefficients. If we focus on the mean
expenditures on ALMPs—555 crowns per inhabitant—and assume a quite
large increase of leftist seats by 5 percentage points, the predicted amount of
money used on ALMPs increases to 577 crowns per inhabitant. I think that
this is an important difference, although it probably not should be considered
a very large impact. In Models 4a and 5a, unit specific effects are
incorporated. The effect of political partisanship is close to zero—in fact, the
coefficient is negative, although not statistically significant. As noted in the
previous section, the fixed effects approach is very demanding. Tests
indicated problems of multicollinearity in Models 4a and 5a. This is a
probable reason why many coefficients are statistically insignificant. In sum,
the overall impression from Models 1a–5a is that the impact of political
partisanship is not completely robust.
I should not dwell on control variables, but some brief notes are
appropriate. Not surprisingly, local needs are important. This is clearly
illustrated by the fact that high unemployment boosts municipal involvement
in ALMPs. An increase of unemployment by one percentage point increases
the predicted expenditures by as much as 4–10 percent, depending on model
16
Note that including a lagged dependent variable in a model with a full set of unit specific
dummies provides biased estimates (Nickell 1981). There are various procedures that can be
used to reduce these problems, such as GMM estimators. But since the specifications that
include the lagged dependent variable will be troublesome anyway, I do not make use of
analyses that incorporates both the previous year’s expenditures and municipal fixed effects.
62
Political partisanship and entity size
specification, holding all other variables constant. The effect is statistically
significant at least at the .05 level in all specifications. Surprisingly,
municipalities with more money use fewer resources per inhabitant on
ALMPs, all else being equal. The effect of tax incomes is especially robust
and is statistically significant at the .01 level in all model specifications. One
possible explanation is that large tax incomes imply that the individuals
living in the area are prosperous and that the demand for ALMPs is not that
high.
I have conducted a large number of robustness tests. Some of these
checks have been noted in the course of the article, while some others are
briefly brought up here. First, dropping control variables does not make a
substantial difference. Second, diagnostic plots showed some influential
outliers. I have no obvious reason to exclude the outliers. Nevertheless,
regressions where these observations are removed have been carried out and
the results stay about the same. Third, various operationalizations of partisanship have been examined. If a dummy variable for a socialist majority is
used as the indicator of political partisanship, instead of the percentage of
seats, the effect becomes larger and even more robust. On the other hand,
focusing on the chairman of the local executive board renders small and
usually statistically insignificant regression coefficients.
In Table 3, I have not taken into account that the effect of partisanship
may be contingent on population size. This may be an important reason why
the findings are not robust. The purpose of the analyses presented in Table 4
is to find out whether this is the case.
The effects of the control variables are approximately the same in
Tables 3 and 4. In addition, robustness checks indicate the same strengths
and weaknesses as prior robustness tests. Thus, it is possible to go directly to
the heart of Table 4, that is, the interaction term of partisanship and size. If
the coefficient is positive and statistically significant there are good reasons
to assume that politics matter more in large than in small municipalities. This
is exactly the case. In all six models, the interaction term is positive and
significant at least at the .05 level. This means that even the hard tests of
Models 3b–5b indicate that the impact of partisanship varies with the size of
the local entity.
But interaction effects, and statistical tests with interactive hypotheses are
always conditional (Allison 1977; Friedrich 1982). In order to find out the
details, it is therefore necessary to take the analysis a step further and calculate conditional effects and standard errors. That is, the impact of partisanship should be determined at various municipality sizes.
63
Foreign
Yes
0.42
Year dummies
Adjusted-R²
* p < .10; ** p < .05; *** p < .01.
No
Clustered
Municipal dummies
Standard errors
0.42
Yes
No
Newey-West
0.83
Yes
No
Robust
1.658 (1.387)
7.297 (2.756)***
–0.011 (0.004)**
0.855 (0.029)***
7.297 (3.413)**
–0.043 (0.012)***
–0.004 (0.001)***
–0.006 (0.012)
0.003 (0.003)*
0.067 (0.032)**
–0.067 (0.028)**
0.003 (0.002)
–0.003 (0.010)
–0.005 (0.004)
–0.002 (0.002)
0.016 (0.004)***
–0.011 (0.012)
3a
0.001 (0.001)**
Constant
–0.043 (0.017)**
State subsidyt–1
–0.017 (0.004)***
–0.005 (0.021)
0.017 (0.005)***
0.365 (0.092)***
–0.274 (0.069)***
0.021 (0.009)**
0.005 (0.026)
–0.019 (0.009)**
–0.011 (0.005)**
0.094 (0.009)***
–0.054 (0.031)*
2a
0.008 (0.002)***
Log(ALMP spending t–1)
–0.005 (0.025)
–0.017 (0.005)***
Tax incomet–1
Social allowancest–1
Population change
0.365 (0.139)***
0.017 (0.008)**
Sparsely populated
–0.274 (0.101)***
0.005 (0.035)
0.021 (0.014)
Female
Metropolitan
–0.011 (0.008)
–0.019 (0.013)
Elderly population
0.094 (0.013)***
Unemploymentt–1
Low education
–0.054 (0.046)
1a
0.008 (0.002)***
Log (Size)
Partisanship
0.81
Yes
Yes
Clustered
11.856 (3.547)***
–0.011 (0.017)
–0.020 (0.007)***
–0.019 (0.016)
–0.004 (0.009)
0.289 (1.961)
0.056 (0.791)
–0.017 (0.032)
–0.006 (0.031)
–0.032 (0.030)
–0.055 (0.023)**
0.043 (0.016)**
–0.098 (0.710)
4a
–0.002 (0.004)
0.81
Yes
Yes
Newey-West
11.856 (2.939)***
–0.011 (0.013)
–0.020 (0.005)***
–0.019 (0.014)
–0.004 (0.006)
0.289 (1.476)
0.056 (0.609)
–0.017 (0.023)
–0.006 (0.026)
–0.032 (0.022)
–0.055 (0.017)***
0.043 (0.012)***
–0.098 (0.548)
5a
–0.002 (0.003)
Table 3. The Effects of Partisanship on Municipal Involvement in ALMPs in Sweden, 1998–2005 (n = 2,208, Standard Errors in Parentheses): Additive
Regression Models with Log(ALMP Spending) as the Dependent Variable
–0.263 (0.070)***
0.007 (0.034)
0.022 (0.014)
–0.263 (0.103)***
Female
Foreign
0.44
Year dummies
Adjusted-R²
* p < .10; ** p < .05; *** p < .01.
No
Yes
Municipal dummies
Clustered
7.650 (2.721)***
0.44
Yes
No
Newey-West
0.82
Yes
No
Robust
1.750 (1.384)
Standard errors
Constant
–0.011 (0.005)**
0.851 (0.029)***
7.650 (3.350)**
–0.040 (0.012)***
–0.040 (0.017)**
–0.004 (0.001)***
–0.006 (0.012)
0.010 (0.005)**
0.077 (0.033)**
–0.067 (0.028)**
0.004 (0.002)
–0.003 (0.010)
–0.005 (0.004)
–0.001 (0.002)
0.017 (0.004)***
0.002 (0.001)**
–0.098 (0.038)**
3b
–0.003 (0.002)*
Log(ALMP spending t–1)
State subsidyt–1
–0.013 (0.004)***
–0.013 (0.005)**
Tax incomet–1
0.014 (0.005)***
–0.001 (0.021)
0.014 (0.007)**
–0.001 (0.024)
Population change
0.410 (0.090)***
0.007 (0.025)
–0.015 (0.009)*
Social allowancest–1
0.410 (0.135)***
Sparsely populated
Metropolitan
0.022 (0.009)**
–0.015 (0.013)
Elderly population
0.095 (0.009)***
–0.010 (0.005)*
0.095 (0.012)***
–0.010 (0.008)
Unemploymentt–1
0.010 (0.002)***
–0.536 (0.098)***
2b
–0.019 (0.006)***
Low education
0.010 (0.003)***
–0.536 (0.141)***
Log (Size)
Partisanship × Log (Size)
1b
–0.019 (0.008)**
Partisanship
0.81
Yes
Yes
Clustered
12.314 (3.514)***
–0.011 (0.017)
–0.020 (0.007)***
–0.019 (0.016)
–0.003 (0.009)
0.784 (2.367)
0.029 (0.783)
–0.019 (0.032)
–0.007 (0.031)
–0.033 (0.030)
–0.053 (0.023)**
0.043 (0.016)***
0.005 (0.002)**
–0.267 (0.689)
4b
–0.015 (0.008)*
0.81
Yes
Yes
Newey-West
12.314 (2.924)***
–0.011 (0.013)
–0.020 (0.005)***
–0.019 (0.014)
–0.003 (0.006)
0.784 (1.802)
0.029 (0.604)
–0.019 (0.023)
–0.007 (0.026)
–0.033 (0.022)
–0.053 (0.017)***
0.043 (0.012)***
0.005 (0.002)**
–0.267 (0.548)
5b
–0.015 (0.007)**
Table 4. The Effects of Partisanship on Municipal Involvement in ALMPs in Sweden, 1998–2005 (n = 2,208, Standard Errors in Parentheses): Regression
Models with Log(ALMP Spending) as the Dependent Variable and Including an Interaction Term between Partisanship and Log(Size)
Essay I
Table 5. Conditional Effects of Political Partisanship on ALMP Spending
Model
Municipality size
Small: 5,000 citizens
Log(size) = 1.609
Medium: 30,000 citizens
Log(size) = 3.401
Large: 100,000 citizens
Log(size) = 4.605
Very large: 450,000 citizens
Log(size) = 6.109
1b and 2b
3b
4b and 5b
–0.003
(0.004)
[0.003]
–0.001
{0.001}
–0.007
(0.005)
[0.004] *
0.015
(0.003) ***
[0.002] ***
0.003
{0.001} ***
0.002
(0.004)
[0.002]
0.027
(0.006) ***
[0.004] ***
0.005
{0.002} ***
0.008
(0.005)
[0.005]
0.042
(0.009) ***
[0.006] ***
0.008
{0.003} ***
0.015
(0.009) *
[0.008] *
Note: Clustered robust standard errors are provided in parentheses, standard errors based on the NeweyWest procedure are given in square brackets, and robust standard errors are presented within curly
brackets. The conditional effects are given by βpartisanship + βpartisanship x Log(size) × Log(size). See Allison (1977)
and Friedrich (1982) for more thorough discussions on conditional effects and how to compute the
standard errors of these coefficients.
* p < .10; ** p < .05; *** p < .01.
Table 5 reports the importance of partisanship as municipality size varies. I
present the effects at four arbitrarily selected unit sizes between 5,000 and
450,000 citizens. The table makes clear that the effect of partisanship
increases significantly with size of municipality. In small municipalities the
average impact is more or less zero. In middle sized municipalities, the
effect is quite small and in some specifications statistically insignificant at
conventional levels. In municipalities with a population size of 100,000
inhabitants, the average impact is more substantial although not statistically
significant in Models 4a and 5a. Lastly, in very large municipalities political
partisanship plays a considerable role. An increase in the number of socialist
seats in the local council of one percentage point increases the average
expenditures on ALMPs by around 1.5–4.2 percent, holding everything else
constant. This implies that if the leftist parties strengthened their position in
the local council by five percentage points—a quite large but not uncommon
change between two elections—the predicted amount of resources devoted
to ALMPs would increase by 7.5–21.0 percent, all else being equal.
To conclude, the overall impression is that the hypothesis that political
partisanship matters more in large local entities receives quite empirical
strong support. Nevertheless, it must be acknowledged that the fixed effects
approach suggests a smaller effect than the other model specifications, and
demands a greater population size before the effect is statistically sig66
Political partisanship and entity size
nificant.17 But there are problems of multicollinearity in Models 4b and 5b.
This is a common setback in specifications including interaction terms. The
large number of dummy variables amplifies the problem. This is probably
the reason why the effect is statistically significant only given a very large
population.
Conclusion
It is difficult to imagine a modern democratic state without political parties.
Parties structure the political world, recruit political elites, offer a linkage
between ruled and rulers, and aggregate citizens’ interests. In this article, I
have examined the hypothesis that political partisanship affects public policy
to a greater extent in large local entities than in small ones. Empirically, the
study has considered municipal involvement in labor market activities at
local level in Sweden. I show that partisanship makes no important
difference in small entities, but as population size increases partisanship
becomes a relevant factor. A strong left wing boosts expenditures on labor
market programs, especially when the number of local inhabitants is very
large. Thus, the study offers conditional support for the “partisan matters”argument.
The pooled cross-sectional design provides great opportunities to control
for alternative explanations and the results appear to be quite robust (although there are some caveats). In future research, scholars interested in the
effects of political partisanship on public policy should recognize the
possibility of an interaction effect between partisanship and population size.
Previous failures to identify an impact may have been a consequence of a
misspecified model. And scholars who have indicated an effect could perhaps have refined their claim by examining interactions.
More research in various settings is needed, of course, before we can be
certain that political partisanship has a larger effect on public policy when
the population is sizeable. But the theoretical statements are general, so there
is no strong reason to assume that the idea does not apply in other settings.
However, the empirical findings in this study can be generalized primarily to
countries in which local governments are relatively autonomous—for
example, other Nordic countries—and to policy areas in which there is a
clear left–right political dimension.
Nevertheless, it would be interesting to see if the findings hold in
countries in which the local level has less autonomy and in policy areas in
17
Model 1b suggests that there is a positive and statistically significant effect at the .10 level
when the population consists of approximately 11,000 citizens or more. The corresponding
number of citizens if the estimates from the other models are used is 10,000 (Model 2b),
13,000 (Model 3b), 232,000 (Model 4b), and 116,000 (Model 5b).
Essay I
which the left–right dimension is not as evident. Another thing worth paying
attention to is the causal mechanism. I have outlined a couple of arguments
for why partisanship should make a bigger difference in large local entities.
But I have not examined these mechanisms empirically. A relevant task for
future research is therefore to find out what mechanisms are at work.
Moreover, size is not the only variable that might interact with partisanship.
Further development of theoretical and empirical research would involve
searching for additional factors that impinge on the impact of partisanship.
ALMPs make up a large part of the welfare state in many advanced
industrialized countries today (Martin and Grubb 2001). Thus, some studies
have tried to spot the factors that affect ALMP expenditures using crossnational data. Political partisanship is a key variable. The findings are not
decisive. Hicks and Kenworthy (1998) suggest a positive impact. But the
conclusion is based on a small sample, and the effect is only statistically
significant given a one-tailed significance test at the .10 level. Swank and
Martin (2001), Rueda (2005), and Vernby (2006) provide more reliable
evidence. Only the study by Swank and Martin indicates a positive and
significant effect. Thus, it seems that we do not really know if partisanship
influence ALMPs. This is troublesome. But the findings in the present study
are helpful. If we assume that policy decisions in large municipalities can be
compared to national level proceedings, there are reasons to assume that
partisanship impinges on ALMP expenditures, even at the national level.
To my knowledge, the effect of political partisanship on ALMP spending
has not been evaluated using local government data. I realize that local
governments are not directly comparable with central governments, but this
approach has a number of advantages. The Swedish case is probably
especially useable since the municipalities are rather large and have a lot of
autonomy (cf. Hanna Bäck 2003; Petterson-Lidbom 2006). Local level data
imply that I can hold all country specific effects, such as institutional and
cultural characteristics, constant. The number of observations becomes much
greater too. This improves the possibilities of hypothesis testing and examining rigorous model specifications. Furthermore, measurement errors are
probably fewer if we focus on one country instead of many.
Lastly, it is worth briefly discussing the findings from a democratic perspective. Local self-government is often considered valuable in promoting
democracy. Local citizens should be allowed to affect activities in their own
districts. If it does not matter who governs at local level, as in the case of
small entities, this might be a problem. On the other hand, it is possible to
contend that it is good that local partisanship does not matter in this
particular case. ALMP is a central government policy domain which is
supposed to be controlled by the central government. How the policy should
look should be decided in national elections. And policies ought to be
implemented in a consistent manner in order to guarantee that the ideas of
the central government are realized and to promote uniform treatment of
68
Political partisanship and entity size
clients around the country. According to this perspective, local variations
should be dependent on local needs rather than on local politics. Thus, it
could be argued that the problem is the partisanship effect in large entities
rather than lack of partisanship effect in small entities. I conclude with these
normative implications. The study contributes to the discussion by providing
the necessary facts. How the facts should be evaluated is up to others.
Essay I
References
Allan, James P., and Lyle Scruggs. 2004. Political partisanship and welfare
state reform in advanced industrial societies. American Journal of Political
Science 48(3):496–512.
Allers, Maarten, Jacob De Haan, and Cees Sterks. 2001. Partisan influence
on the local tax burden in the Netherlands. Public Choice 106:351–63.
Allison, Paul D. 1977. Testing for interaction in multiple regression.
American Journal of Sociology 83(1):144–53.
Alt, James E. 1985. Political parties, world demand, and unemployment:
Domestic and international sources of economic activity. American
Political Science Review 79(4):1016–40.
Ashworth, John, and Bruno Heyndels. 2002. Tax structure turbulence in
OECD countries. Public Choice 111:347–76.
Beck, Nathaniel, and Jonathan N. Katz. 1995. What to do (and not to do)
with time-series cross-section data. American Political Science Review
89(3):634–47.
Besley, Timothy, and Anne Case. 2003. Political institutions and policy
choices: Evidence from the United States. Journal of Economic Literature
41(1):7–73.
Blais, Andre, Donald Blake, and Stephane Dion. 1993. Do parties make a
diffenrence? Parties and the size of government in liberal democracies.
American Journal of Political Science 37(1):40–62.
Blom-Hansen, Jens, Lars Christian Monkerud, and Rune Sorensen. 2006. Do
parties matter for local revenue policies? A comparison of Denmark and
Norway. European Journal of Political Research 45:445–65.
Brown, Robert D. 1995. Party cleavages and welfare effort in the American
states. American Political Science Review 89(1):23–33.
Budge, Ian. 2001. Validating party policy placements. British Journal of
Political Science 31(1):210–24.
Bäck, Hanna. 2003. Explaining coalitions: Evidence and lessons from
studying coalitions in Swedish local government. Uppsala, Sweden: Acta
Universitatis Upsaliensis.
Bäck, Henry. 2003. Party politics and the common good in Swedish local
government. Scandinavian Political Studies 26(2):93–123.
Calmfors, Lars, Anders Forslund, and Maria Hemström. 2002. Does active
labour market policy work? Lessons from the Swedish experience.
70
Political partisanship and entity size
Working Paper 4, Institute for Labour Market Policy Evaluation, Uppsala,
Sweden.
Carling, Kenneth, and Laura Larsson. 2005. Does early intervention help the
unemployed youth? Labour Economics 12:301–19.
Dahl, Robert A., and Edward R. Tufte. 1973. Size and democracy. Stanford,
CA: Stanford Univ. Press.
Downs, Anthony. 1957. An economic theory of democracy. New York:
Harper and Row.
Faust, Jon, and John S. Irons. 1999. Money, politics and the post-war
business cycle. Journal of Monetary Economics 43:61–89.
Friedrich, Robert J. 1982. In defense of multiplicative terms in multiple
regression equations. American Journal of Political Science 26(4):797–
833.
Froot, Kenneth A. 1989. Consistent covariance matrix estimation with crosssectional dependence and heteroskedasticity in financial data. Journal of
Financial and Quantitative Analysis 24(3):333–55.
Garand, James C. 1988. Explaining government growth in the US states.
American Political Science Review 82:837–53.
Government Bill 2006/07:1. Förslag till statsbudget 2007. Utgiftsområde 13
(Proposition 2006/07:1, utgiftsområde 13).
Government Bill 2006/07:15. En arbetslöshetsförsäkring för arbete (Proposition 2006/07:15).
Gujarati, Damodar N. 2003. Basic Econometrics. Boston, MA: McGrawHill.
Greene, William H. 2003. Econometric analysis. Upper Saddle River, NJ:
Prentice Hall.
Hibbs, Jr. Douglas A. 1977. Political parties and macroeconimc policy.
American Political Science Review 71(4):1467–87.
Hicks, Alexander, and Lane Kenworthy. 1998. Cooperation and political
economic performance in affluent democratic capitalism. American Journal
of Sociology 103(6):1631–72.
Imbeau, Louis M., Francois Pétry, and Moktar Lamari. 2001. Left–right
party ideology and government policies: A meta-analysis. European
Journal of Political Research 40:1–29.
Johansson, Folke, and David Karlsson. Forthcoming. Kommunstorlek och
deomkrati. Center for Public Sector Research (CEFOS), Göteborg
University, Sweden.
Essay I
Kato, Junko, and Bo Rothstein. 2006. Government partisanship and
managing the economy: Japan and Sweden in comparative perspective.
Governance 19(1):75–97.
Kezdi, Gabor. 2004. Robust standard error estimation in fixed-effects panel
models. Hungarian Statistical Review 9:95–116.
King, Gary. 2001. Proper nouns and methodological propriety: Pooling
dyads in international relations data. International Organization 55:497–
507.
Korpi, Walter. 1974. Conflict, power and relative deprivation. American
Political Science Review 68(4):1569–78.
———. 2006. Power resources and employer-centered approaches in
explanations of welfare states and varieties of capitalism. Protagonists,
consenters, and antagonists. World Politics 58:167–206.
Korpi, Walter, and Joakim Palme. 2003. New politics and class politics in
the context of austerity and globalization: Welfare state regress in 18
countries, 1975–95. American Political Science Review 97(3):425–46.
Lidström, Anders. 1996. Kommunsystem i Europa. Stockholm, Sweden:
Publica.
Lundin, Martin. 2007. Explaining cooperation: How resource interdependence, goal congruence, and trust affect joint actions in policy
implementation. Journal of Public Administration Research and Theory;
doi: 10.1093/jopart/mul025.
Lundin, Martin, and Per Skedinger. 2006. Decentralisation of active labour
market policy: The case of Swedish local employment service committees.
Journal of Public Economics 90(4-5):775–98.
Martin, John P., and David Grubb. 2001. What works and for whom: A
review of OECD countries’ experiences with active labour market policies.
Swedish Economic Policy Review 8(2):9–56.
Moene, Karl Ove, and Michael Wallerstein. 2003. Earnings inequality and
welfare spending. A disaggreted analysis. World Politics 55:485–516.
Montin, Stig. 2002. Moderna kommuner. Stockholm, Sweden: Liber.
Newey, Whitney K., and Kenneth D. West. 1987. A simple, positive, semidefinite, heteroskedasticity and autocorrelation consistent covariance
matrix. Econometrica 55:703–788.
Nickell, Stephen. 1981. Biases in dynamic models with fixed effects.
Econometrica 49:1417–26.
72
Political partisanship and entity size
Nilsson, Anna, and Jonas Agell. 2003. Crime, unemployment and labor
market programs in turbulent times. Working Paper 14, Institute for Labour
Market Policy Evaluation, Uppsala, Sweden.
O’Toole, Laurence J. Jr. 1983. Interorganizational co-operation and the
implementation of labour market training policies: Sweden and the Federal
Republic of Germany. Organization Studies 4(2):129–50.
Petersson, Olof. 1994. Kommunalpolitik. Stockholm, Sweden: Fritzes.
Pettersson-Lidbom, Per. 2006. Do parties matter for economic outcomes? A
regression-discontinuity approach. Unpublished manuscript. Stockholm,
Sweden: Department of Economics, Stockholm Univ.
Pierson, Paul. 1996. The new politics of the welfare state. World Politics
48(2):143–79.
Ross, Fiona. 2000. “Beyond left and right”: The new partisan politics of
welfare. Governance 13(2):155–83.
Rothstein, Bo. 1996. The social democratic state. The Swedish model and
the bureaucratic problems of social reforms. Pittsburgh, PA: Univ.
Pittsburgh Press.
Rueda, David. 2005. Insider–outsider politics in industrialized democracies:
The challenge to social democratic parties. American Political Science
Review 99(1):61–74.
Salonen, Tapio, and Rickard Ulmestig. 2004. Nedersta trappsteget. En
studie om kommunal aktivering. Växjö, Sweden: Växjö Univ.
Schmidt, Manfred G. 1996. When parties matter: A review of the
possibilities and limits of partisan influence on public policy. European
Journal of Political Research 36:87–118.
Strömberg, Lars, and Jörgen Westerståhl. 1983. De nya kommunerna: En
sammanfattning av kommunaldemokratiska forskningsgruppens undersökningar. Stockholm, Sweden: Liber.
Strom, Kaare. 1990. A behavioral theory of competitive political parties.
American Journal of Political Science 34(2):565–98.
Swank, Duane, and Cathie Jo Martin. 2001. Employers and the welfare state.
The political economic organization of firms and social policy in
contemporary capitalist democracies. Comparative Political Studies
34(8):889–923.
Swedish Association of Local Authorities. 1999. Kommunerna i arbetsmarknadspolitiken – en enkätundersökning. Stockholm, Sweden: Swedish
Association of Local Authorities.
Essay I
Swenson, Peter. 1991. Bringing capital back in, or social democracy
reconsidered: Employer power, cross-class alliances, and centralization of
industrial relations in Denmark and Sweden. World Politics 43(4):513–44.
Vernby, Kåre. 2006. Strikes and Electoral Systems in 17 OECD Countries
1960–1998. In Essays in Political Economy. Uppsala, Sweden: Uppsala
Univ.
Wetterberg, Gunnar. 1997. Kommunerna. Stockholm, Sweden: SNS Förlag.
74
Essay II
Explaining Cooperation: How Resource
Interdependence, Goal Congruence, and Trust
Affect Joint Actions in Policy Implementation*
To understand what is going on when public policies are carried out at local
level, interorganizational relationships have to been taken into account (see
for example, Bardach 1998; Hjern and Porter 1981; O’Toole and Montjoy
1984). Many challenges facing modern societies—such as fighting poverty
or reducing unemployment—are difficult to manage within a single public
authority. The virtues of interorganizational cooperation have therefore been
emphasized in implementation research ever since Pressman and
Wildavsky’s (1984) pioneering study of a public labor market program in
Oakland in the 1960s. Laurence O’Toole (2003, 237) concludes that “the
topic of interorganizational relations will remain important for administrators tasked with helping to make policy implementation succeed.
Accordingly, it is critical to understand how to make sense of such institutional settings for improving prospects for implementation success.”
This article contributes to the discourse by presenting evidence on how
resource interdependence, goal congruence, and trust affect cooperation
between agencies in local policy implementation.
According to research on interorganizational relationships, mutual resource dependence and congruent goals are among the most important
antecedents to cooperative behavior. In a situation of mutual dependence,
organizations will cooperate in order to exchange resources that make it
*
A slightly different version of this article has been accepted to be published by the Journal
of Public Administration Research and Theory (Oxford University Press) and will appear in a
forthcoming issue. I would like to thank Paula Blomqvist, Matz Dahlberg, Jörgen
Hermansson, Björn Lindberg, Nils Hertting, Andreas Lindemann, Karl-Oskar Lindgren,
Daniela Lundin, PerOla Öberg, Sven Oskarsson, and Soren Winter for comments on earlier
drafts of this article. Several participants at the seminars held at the XIV Nordic Political
Science Association (NOPSA) Conference, Reykjavik 11–13 August 2005, and at the
Department of Government, Uppsala University, 26 September 2005 contributed to this
research as well. I would also like to thank Carolyn Heinrich, the editor for Journal of Public
Administration research and Theory, and the two anonymous referees for very useful
suggestions.
75
Essay II
possible to achieve organizational goals, whereas shared interests and a
similar commitment to a policy make it easier to get along and can generate
joint struggles (O’Toole 2003, 239–42). Another important finding is that
mutual trust increases cooperation (Smith, Carroll, and Ashford 1995, 10–1)
since it “facilitates interpersonal acceptance and openness of expression”
(Zand 1972, 229).
In previous research, the direct additive effects of mutual resource
dependence, congruent goals, and trust have been examined and corroborated. However, it has not been studied whether the effects of resource
interdependence and goal congruence are different in situations characterized by high levels of trust compared to situations when agencies do not trust
each other. In this article, it is examined whether a credible commitment,
that is, mutual trust, is a necessary condition for goal congruence and mutual
resource dependence to affect cooperation. Moreover, the effect of trust is
studied when the levels of resource interdependence and goal congruence
vary. It is hard to see why trust should increase cooperation if actors are not
interdependent or have similar objectives. In sum, interaction effects might
explain levels of cooperation better than the additive effects suggested in
previous research.
The study focuses on the relationships between agencies that operate at a
local level but representing different tiers of government. Policy implementation often involves an intergovernmental dimension (Jennings and
Ewalt 1998; Radin 2003; Thomas 1979) and “so long as there have been
systems of dividing decisions and functions of governments there have been
attempts to organize and manage them” (Agranoff 2004, 443). Consequently, studies such as this one are of great interest.
I make use of new large-N data on the Swedish active labor market policy
and examine dyads of Public Employment Service (PES) offices and
municipal labor market agencies. The Swedish PES offices are central
government agencies carrying out labor market policies at a local level.
However, in recent years, a parallel local government system has evolved for
labor market activities since municipalities now take an active part as well
(Hjertner Thorén 2005; Lundin 2005; Lundin and Skedinger 2006; Salonen
and Ulmestig 2004). This article takes a closer look at the interaction of the
two tiers of government.
The empirical results should be interpreted with some caution, given a
couple of methodological pitfalls. But the empirical results indicate that the
effect of goal congruence is dependent on mutual trust. If organizations do
not trust each other, similar priorities do not matter. In addition, if the
authorities’ objectives diverge to a large extent, trust does not increase
cooperation. However, the results do not support the idea that an interaction
term between resource interdependence and trust should be included in
models trying to explain collaborative behavior. In order to discern whether
these effects really are causal, I encourage more empirical research.
76
Explaining cooperation
The remainder of this article is organized as follows: In the next section,
theory and hypotheses are presented. The Swedish labor market case is then
briefly discussed. In the two following sections, data and measures are
described. The results are then reported and, lastly, in a concluding section,
the findings are summarized and their implications are discussed.
Explaining cooperation: Interdependence, goal
congruence, and trust
Even in the most well-designed political system, authorities’ responsibilities
will overlap. Accordingly, development and implementation of public policy
frequently “demands multilateral cooperation, blurs or eliminates traditional
boundaries and jurisdictions, and needs the deployment of many actors”
(Alexander 1995, xvi). Diverging organizational goals and operational
routines thus make policy implementation difficult (see for example,
O’Toole 2003). A primary task of public sector management is to get the
various agencies to cooperate when policies are carried out.1 Thus, detecting
antecedents to cooperation is of imperative interest. In line with previous
interorganizational research, cooperation in the public sector is defined as
the interactions among actors aiming at solving public problems by working
together rather than by working separately (cf. Smith, Carroll, and Ashford
1995, 10). Collaboration is used synonymously to cooperation throughout
the article.
Resource interdependence
There is an extensive and heterogeneous literature on collaboration between
organizations (Oliver and Ebers 1998). However, exchange theory provides
a foundation for explaining cooperation (Blau 1964; Levine and White 1961;
Scharpf 1978). In this framework, cooperation is a consequence of resource
interdependence. Organizations are assumed to be rational actors making
conscious and intentional decisions in order to achieve their objectives. The
motivation for working together is the need to overcome a lack of resources.
An organization will avoid interactions with others if the benefits of cooperation do not exceed the costs since cooperation is complicated, is costly,
and involves a loss of autonomy. But if organization A needs resources from
organization B and organization B needs resources from organization A,
1
Note that cooperation involves costs and that implementation performance is not
automatically improved by interorganizational cooperation. It is an empirical question
whether the positive aspects outweigh the negative. Results in Jennings and Ewalt (1998)
demonstrate that cooperation improves some aspects of policy performance, whereas there is
no effect on other aspects. The effects of cooperation are not examined in this article.
77
Essay II
there is a good chance that cooperation will take place. Financial resources
are, of course, of significant interest, but staff, premises, information,
legitimacy, and legal authority are examples of other significant resources
that can be obtained from other organizations.
Empirical evidence support the idea that mutual resource dependence
increases cooperation. Levine and White (1961) indicated that health-related
organizations highly dependent on resources from the local health system
interacted more and had fewer disagreements with local agencies than
organizations less resource dependent. In a quantitative longitudinal study of
relationships among child care and health organizations in Texas, Van de
Ven and Walker (1984) found that the need for resources was the most important variable stimulating coordination of activities. More recent evidence
exists as well. As an example, Gulati and Gargiulo (1999) examine alliance
formation using longitudinal data from a sample of American, European and
Japanese organizations; the effect of interdependence on alliance formation
is positive and statistically significant.
Goal congruence
Besides mutual dependence, goal congruence may boost collaboration.
Given the assumption that organizations try to reach certain goals, we should
not only focus on the need for external resources but also investigate the
extent to which the organizations’ goals are similar. A shared interest can be
a powerful facilitator of cooperation, whereas diverging objectives may decrease cooperation (O’Toole 2003, 239).
Empirical results confirm that objectives are important. For example,
Levine and White (1961) indicated that agreement on goals is an important
aspect of interorganizational relationships. Compatibility of goals is also
found to be positively associated with cooperation in Schmidt and Kochan’s
(1977) study of community organizations and local offices of the US
Training and Employment Service. O’Toole (1983) indicated that perceived
common interest increases cooperation among local actors in a study of
interorganizational implementation of labor market policies in Sweden and
the Federal Republic of Germany. Moreover, Luo (2001) finds positive
effects of goal congruity on personal attachment in international cooperative
ventures. Personal attachment is defined as “the degree to which boundary
spanners (resident board members and senior venture managers) from each
party are socially bound through having developed personal relationships
and interpersonal learning” (Luo 2001, 178). Luo’s operationalization
include the extent to which information and skills have been transferred
between the partners. Personal attachment can be regarded as a form of
cooperation or, at least, as a concept closely related to cooperation.
78
Explaining cooperation
Trust
Trust is a key word in the social sciences. Scholars like Elster (1989),
Ostrom (1998), Putnam (2000), and Rothstein (2000) accentuate trust as a
mechanism for overcoming social dilemmas. In research on policy implementation, Bardach (1998) stresses the importance of mutual trust in order to
make agencies work together. There are various definitions of and
perspectives on trust. Fritz Scharpf (1997, 137) uses the concept of weak
trust.2 By weak trust, Scharpf means the expectation that another actor’s
communicated preferences are honest rather than misleading and the expectation that the actor will stick to a commitment as long as the conditions
under which it was entered are not altered dramatically (see also, Svensson
and Öberg 2005). In this article, I use trust in the same sense as Scharpf.
There is an agreement among organization theory scholars that trust is an
important antecedent to cooperation. In an experimental analysis using data
on business executives, Zand (1972) shows that trust enhances cooperation.
Muthusamy and White (2005) focuses on alliances between business firms
in the United States from 1994 to 1998 and find that trust has a positive
effect on transfer of knowledge between partners. The analysis of Zaheer
and Venkatraman’s (1995) of insurance agencies also indicate a positive
relationship between mutual trust and cooperation. In an overview article,
Smith, Carroll, and Ashford (1995, 10–1) note that “although research has
identified many determinants of cooperation, virtually all scholars have
agreed that one especially immediate antecedent is trust.”
Hypotheses
I suggest that we need to be more specific about under what conditions
resource interdependence, goal congruence, and trust affect interorganizational cooperation. The idea is that interaction effects should be
considered in order to fully understand the relationships. Previous empirical
research seems to have overlooked the possibility of interaction terms.
The literature implicitly assumes that the effects of resource interdependence and goal congruence are the same when trust is low as when trust is
high. How reasonable is this assumption? Why should an organization
2
Trust, goal congruence, and resource interdependence are treated as separate variables in the
present study. It is likely that goal congruence increases trust, but mutual trust may exist even
when objectives diverge. In addition, congruent goals will not automatically imply high trust
since such a definition of trust is reliant upon an unrealistic assumption of perfect information
about preferences. Moreover, in a situation without mutual resource dependence actor A can
still think that actor B (and vice versa) has honest intents and will stick to an agreement. Some
scholars, for example, Hardin (2002), seems to collapse resource interdependence, congruent
goals, and trust: actor A trusts actor B when it is in B’s interest to fulfil A’s expectation. This
position is, in my view, not as beneficial as treating resource interdependence, goal congruence, and trust as separate variables (for a good discussion see Rothstein 2000, 484–8).
79
Essay II
choose a collaborative strategy in a situation of resource interdependence if
it cannot rely on the other organization’s commitment to the relationship?
Blau (1964, 98) describes trust as an important condition for the exchange of
resources to occur: “since social exchange requires trusting others to
reciprocate, the initial problem is to prove oneself trustworthy”. Thus, resource interdependence is only excepted to be a relevant factor when there is
mutual trust. Below, a first hypothesis is outlined.
H1
The effect of resource interdependence on interorganizational
cooperation is dependent on the level of trust; a positive effect is
expected only when organizations trust each other.
Furthermore, objectives cannot be communicated in a credible way when
one does not trust one another. When communication is not regarded as
reliable, how can organizations be sure that they have congruent goals?
Hence, I argue that trust must exist in order to make objectives a relevant
factor.
H2
The effect of congruent (or diverging) objectives on interorganizational cooperation is dependent on the level of trust. How
similar or different their priorities are is only expected to affect
cooperation when organizations trust each other.
How about trust then? Could we expect trust to affect cooperation regardless
of the level of resource interdependence and goal congruence? I argue that it
is unlikely for organizations to work together only because they trust one
another. Mutual trust is not a sufficient reason. Trust can make cooperation
easier, but it is not something that boost cooperation if there are no other
reasons. The effect of trust ought to be dependent on the similarities between
organizations’ goals and the level of resource interdependence. This suggests
two additional hypotheses.3
H3
The effect of trust on interorganizational cooperation is dependent
on the level of resource interdependence; the effect increases as the
level of interdependence increases.
H4
The effect of trust on interorganizational cooperation is dependent
on the level of goal congruence; the effect increases as objectives
become more similar.
3
Note that H3 and H4, in fact, correspond directly to H1 and H2, respectively. That is, it is not
possible to empirically separate H3 from H1, and H4 from H2. For example, the analysis cannot
support the first proposition and then reject the third proposition. Nevertheless, I am interested
in the effects of all three variables, and to make the presentation of the findings convenient,
the theoretical assertions are stated as four separate hypotheses.
80
Explaining cooperation
Intergovernmental cooperation in local policy
implementation: The Swedish labor market case
The hypotheses outlined above are examined using data on dyads of central
and local government agencies carrying out the active labor market policy in
Sweden. Unemployment is a major problem in most countries, and governments use active programs to reduce the number of persons without a job.
Active labor market policy refers to “measures to raise employment that are
directly targeted at the unemployed” (Calmfors, Forslund, and Hemström
2002, 5). This includes job-brokering activities and labor market programs,
such as labor market training and subsidized employment.
In Sweden, the National Labor Market Administration (Arbetsmarknadsverket), a national government authority, carries out political decisions at
local level through local PES offices (arbetsförmedlingar). There is a PES in
nearly all municipalities, and in larger cities, there is often more than one
office. Municipalities are involved in implementation of labor market
activities as well, although overall responsibility is retained at the national
level and the PES remain the most important local actor (Hjertner Thorén
2005; Lundin 2005; Lundin and Skedinger 2006; Salonen and Ulmestig
2004). It is estimated that in 1999 around 80,000 persons were activated in
labor market programs arranged by the municipalities. This accounts for
about 40 percent of the participants in all the National Labor Market
Administration’s active measures (Lundin and Skedinger 2006). In addition,
municipalities organize labor market programs (“active social policy”) for
persons who are not “job ready” and live on social welfare benefits.
Approximately 12,000 individuals were activated in such programs at any
given point in time in 2002 (Salonen and Ulmestig 2004). A majority of
municipalities have special labor market administrations, while others incorporate labor market activities in their social services administration
(Lundin 2005). The empirical analysis focuses on the relationship between
the PES offices and the municipalities’ labor market administrations.
Besides unemployment being one of the largest problems in modern
welfare states, the current research setting is suitable for at least five reasons.
First, the high information requirements imply that cooperating across
organizational boundaries becomes a viable strategy in labor market
activities (O’Toole 1983). Second, similar dyads of central-local government
agencies are analyzed, and the same political decisions are implemented
around Sweden. This means that it is possible to hold a lot of characteristics
constant when testing the hypotheses of interest. Third, cooperation varies
around Sweden, which results in variation in the dependent variable that can
be explored. Fourth, new and suitable data is available and reliable. Fifth,
quantitative studies on implementation and cooperation between agencies
have mainly been carried out on data from the United States. Evidence from
a unitary European context is therefore of interest.
81
Essay II
In the late 1970s and the beginning of the 1980s, research was conducted
on diverse aspects of interorganizational implementation in the realm of
active labor market training in Sweden and West Germany. For instance,
Hanf, Hjern, and Porter (1978) showed that implementation of programmers
was dependent on interorganizational relationships to a considerable extent,
and O’Toole (1983) indicated that common interest in the locale was an
important factor stimulating interorganizational cooperation. PES offices
specializing in labor market training were key actors in these studies.
Municipalities were involved as well, although labor market training usually
does not engage municipalities to any greater extent. These days, local
governments have a much more profound role in the active labor market
policy in Sweden than in the 1970s and the 1980s. This makes it interesting
to study the relationship between PES offices and municipalities making use
of contemporary data.
Data
Cross-sectional data on dyads of PES offices and municipalities in 2003
have been collected. A questionnaire is the main data source, but
administrative data from the National Labor Market Administration and
official municipal statistics from the database KFAKTA03 supplement the
survey.
Postal questionnaires were distributed to the managers of all Swedish PES
offices in February 2004. The response rate was good: 268 managers replied,
which implies a response rate of 75 percent. Dyadic data require information
from both sides of the dyad. Thus, questionnaires were also distributed to
managers of municipal labor market administrations and municipal
politicians with responsibility for labor market activities. The response rates
were even better in these cases: 85 percent of the managers (246 persons)
and 84 percent of the politicians (245 persons) replied. In Stockholm,
Sweden’s largest municipality, labor market operations are handled in 18
different offices organized geographically (kommundelar). The managers of
these offices answered the Stockholm questionnaires. The response rate was
somewhat lower here: 56 percent (10 persons). The non-responses are probably not that problematic since the respondents resemble the population on
characteristics such as location, size, and unemployment rate in all groups.
The PES managers’ answers form one part of the dyad. However, either
the managers’ or the politicians’ answers could make out the other side. It is
reasonable to assume managers to have more accurate information about
operations. Thus, the first choice was to make use of information from this
82
Explaining cooperation
group. The local politicians’ responses were employed when the managers’
answers were missing to maximize the number of dyads.4
Even though the response rates are very high, the number of observations
that can be used decreases in dyadic data. If one party in an agency pair is
missing, it is not possible to use that information. Internal missing data on
some variables also reduces the number of useable observations. In the end,
the analysis is conducted on 203 cases.
Like most other researchers investigating interorganizational ties, I have
to rely on reports from key informants. Managers and politicians were
instructed to give generalized information about their organization. The
assumption that the responses from one person represent the whole
organization can be questioned. However, it is reasonable to assume that the
respondents in this case are well informed and that data are trustworthy.
Measures
In this section operationalizations of cooperation, interdependence, goal congruence, trust, and control variables are described. Table 1 lists descriptive
statistics for each of the measures.
Cooperation
The dependent variable cooperation is measured using five 0/1-indicators
obtained from the PES questionnaire. The managers were asked if the PES
office and the municipality have established cooperative groups where (a)
caseworkers from the two organizations collaborate and (b) managers from
the two organizations collaborate. Moreover, the managers indicated
whether (c) caseworkers contact each other on a daily basis or more seldom.
They also provided information about whether the PES office and the
municipality have formal collaborative contracts concerning two policies:
(d) actions to prevent long-term unemployment among young people and (e)
a program called the “Activity Guarantee” where the target group is
individuals who have been unemployed for a considerable time period. In
both policies, the Swedish government encourages collaboration between
PES offices and municipalities, but the authorities are not compelled to sign
a collaborative contract.5
4
As a test, all analyses in the article have been performed including a dummy variable
measuring whether the municipal response comes from a manager or politician. Furthermore,
the analyses have been carried out using a sample solely consisting of responses from the
managers. These additional tests did not alter the conclusions.
5
For more information about the Activity Guarantee, see Forslund, Fröberg, and Lindqvist
(2004). The youth policy is described and evaluated in Carling and Larsson (2005).
83
Essay II
Table 1. Descriptive Statistics
Variables
Cooperation
Mean /
Proportion
3.60
SD
Min
Max
1.27
0.00
5.00
Non-Nordic clients
7.28
6.73
1.17
61.96
Long-term clients
11.65
4.95
0.11
30.04
Goal congruence
7.43
0.97
4.38
9.85
Trust
0.45
–
–
–
Big cities
0.07
–
–
–
Suburbs
0.10
–
–
–
Larger cities
0.12
–
–
–
Middle-sized cities
0.14
–
–
–
Industry
0.19
–
–
–
Rural
0.09
–
–
–
Sparsely populated
0.09
–
–
–
Other larger
0.11
–
–
–
Other smaller
0.09
–
–
–
Unemployment
3.31
1.27
1.10
8.20
Non-Swedish citizens
2.21
2.78
0.00
36.25
99.84
2.33
93.59
105.99
0.01
–
–
–
PES size
17.84
16.51
2.00
105.00
PES cooperative orientation
12.36
5.96
0.50
18.00
Municipality type
Population change
Work rehabilitation PES
PES finances
0.66
0.34
0.00
3.01
Mun. LMP spending
2.99
1.97
0.04
13.19
47.65
10.86
11.11
82.86
4.77
4.54
0.00
18.00
Left parties
Mun. cooperative orientation
Mun. tax base
124.00
18.46
99.29
250.58
State subsidy
6.77
5.32
–15.70
22.14
Equal terms
0.51
–
–
–
Note: Mun., municipality; LMP, labor market program. Proportions in italics.
84
Explaining cooperation
Table 2. Principal Component Factor Analysis: Cooperation
Item
Factor Loadings
(a) Joint collaborative groups at the caseworker level
0.74
(b) Joint collaborative groups at the manager level
0.70
(c) Daily contacts between caseworkers
0.43
(d) Youth policy collaborative contract
0.56
(e) Activity Guarantee collaborative contract
0.53
Note: Entries are factor loadings in a principal component analysis. The retention of factors is based on
the Kaiser criterion (that is, eigenvalues greater than 1.00). The eigenvalue for the first dimension is 1.82.
This is the only dimension where the eigenvalue exceeds 1.00. The factor explains 36.3 percent of the
variance in the variables. Bartlett’s test of sphericity indicates that we can reject the hypothesis that the
correlation matrix is an identity matrix at the .05 level. Kaiser-Meyer-Olkin Measure of Sampling
Adequacy is .65 (.50 or lower is usually considered unacceptable).
In Table 2, a factor analysis is presented. All items load positively on the
first dimension. Since the items form a single dimension, a collaboration
index is constructed ranging from zero to five. A score of zero implies that
none of the activities (a)–(e) are taking place, whereas five means that the
organizations have collaborative groups at the caseworker and manager
levels, they contact each other on a daily basis, and have signed cooperation
contracts concerning the youth policy and the Activity Guarantee. I use a
simple additive index in order to make interpretation easy. All analyses in
the article have been performed using the factor scores obtained from the
factor analysis as well, and the conclusions remain the same.
Cooperation has been measured in various ways in prior research.
Sometimes a single item is employed; other times indices are used. Measures
vary depending on the context. The frequency of communication, how often
clients are transferred between organizations, how much support a focal
organization receives from another organization, the use of various forms of
coordination techniques such as interagency committees and task forces are
examples of ways to assess to what extent organizations work together (see,
for example, Alter and Hage 1993; Jennings and Ewalt 1998; Schmidt and
Kochan 1977; Van de Ven and Walker 1984; Wholey and Huonker 1993).
The items in Table 2 are similar to the measures used in other settings,
although I am not aware of any study using exactly the same approach.
Interdependence
There are at least two approaches available to assess interdependence, none
of them perfect. A common approach is to ask respondents directly how dependent they are on each other (see, for example, Provan and Skinner 1989).
A potential drawback is the risk that the responses are a consequence of the
85
Essay II
level of cooperation rather than a cause. Another problem is that respondents
may have trouble interpreting survey questions and response alternatives.
This leads to problems with both validity and reliability.
Instead of assessing resource interdependence directly, it is possible to
look for situations in which it is realistic to assume that organizations are
mutually resource dependent. This approach is employed by Gulati and
Gargiulo (1999) studying alliances in three industries in a sample of American, European, and Japanese organizations. Gulati and Gargiulo assume that
firms benefit from cooperation between niches since they have complementary resources. Moreover, firms that are located in different regions are
understood to be interdependent; a firm needs a partner in another market.
The latter method depends on the critical assumption that it is possible to
identify circumstances when organizations are mutually dependent. I argue
that such variables are available in the present research setting. My claim is
that a main determinant of resource interdependence is the composition of
the unemployed in the locale. At the heart of the argument is that there is a
net mutual benefit from collaboration when the clientele consists of many
individuals in need of assistance both from the PES and from the
municipality.
If a large part of the unemployed face some additional hindrance besides
being unemployed, the authorities are likely to need one another. Difficulties
in speaking and understanding the Swedish language, various work disabilities, and drug addiction problems are examples of problems that
intertwine with difficulties in finding a job. Municipalities implement and
fund social assistance, policies aiming at giving support to those with a
handicap, and the introduction of immigrants into the Swedish society.
Municipal resources, such as expertise and information, are important for the
PES to obtain when many of the unemployed have problems lying within the
jurisdiction of the local governments. At the same time, the PES offices’
knowledge in labor market issues and, above all, their financial resources are
of municipal interest. Thus, mutual dependence is likely to be high when a
large part of the unemployed have multiple problems.
I cannot observe when several clients are in need of both authorities’
actions, but there are rather good indicators available. Two variables based
on register data from the National Labor Market Administration are
employed. Long-term clients, measured as the percentage of clients
registered with the PES as openly unemployed for more than six consecutive
months in 2003, is one of the indicators (this is the official definition of
long-term unemployment in Sweden). The clients should not have
participated in a labor market program during this time period. Among these
long-term unemployed, I expect to find a relatively large share of persons
with multiple problems. The second variable, non-Nordic clients, is the
percentage of the unemployed clients without a Nordic citizenship that are
86
Explaining cooperation
registered with the PES in 2003.6 It is reasonable to expect that a relatively
large proportion of these clients are in need of support from both agencies.
For example, according to the National Board of Health and Welfare (2004),
11 percent of those born in other countries than Sweden, 18 years or older,
received municipal welfare benefits 2003. This can be compared with the
2 percent among citizens born in Sweden. Moreover, municipalities are
responsible for introduction of immigrants into the Swedish society,
including teaching the Swedish language.
I believe that long-term clients and non-Nordic clients are reasonable
indicators of resource interdependence, but at least two caveats should be
mentioned. First, the variables provide no direct evidence of dependence on
each other’s resources. Thus, I cannot be completely sure that the indicators
measure something else other than interdependence. Second, even if we
accept the measures as valid indicators, one problem remains: the agencies
may be mutually dependent for other reasons. This implies that the empirical
study is limited to a certain kind of resource interdependence and that
conclusions should be made with some reservations.
Goal congruence
A direct question is often used to assess goal congruence: “to what extent are
the goals of the other organization compatible with the goals of your
organization?” (see, for example, Schmidt and Kochan 1977). However, to
reassure that answers are exogenous, I believe it is more appropriate to let
the respondents estimate the importance of certain goals and compare the
answers of the actors within each dyad.
The questionnaire respondents were asked to grade, on a scale from one
to five, the importance of 13 different objectives. The ratings were transferred into a ranking of objectives for each agency. This approached is
preferred since a ranking is more comparable between respondents. If two
objectives were given the same rating by a respondent, they got the same
ranking. The difference between the PES office’s and the municipality’s
ranking in each dyad was then calculated.7 I inverted the measure in order to
make the interpretation of the findings easier. That is, higher scores imply
more congruence. In Table 3, objectives and averages are presented. The
6
A large share of the immigrants are coming to Sweden from the neighbor countries. Nordic
immigrants are likely to have fewer boundary-spanning problems than other immigrants since
language and cultural barriers are smaller, for example.
7
The difference between the agencies’ ranking of a certain objective is given by
( PES rank − MUN rank ) 2 .
i
i
PESrank is the PES office’s ranking of the objective from 1-13 in each dyad (1 is the most
important objective, 13 is the least important objective), whereas MUNrank is the
corresponding municipality’s ranking of the same objective. This measure was inverted: 0
becomes 12, 1 becomes 11, and so on. Thus, a high value means similar objectives.
87
Essay II
minimum conceivable value is zero, which implies very different priorities.
The maximum is 12, that is, both agencies in all dyads rank the objective in
the same way.
Table 3. How Similar Various Objectives Are Ranked by the PES offices and the
Municipalities (0 = Minimum Congruence; 12 = Maximum Congruence)
Objective
Ensuring that there are labor market programs for young people under 25
Congruence
10.56
Reducing unemployment
9.85
Ensuring that there are labor market programs for groups of unemployed
with severe problems in the labor market
8.90
Improving matching between available jobs and unemployed persons
8.70
Taking clients’ own requests and needs into account
7.92
Improving municipal services for the local population
7.70
Monitoring clients
7.45
Shifting people from subsidized to unsubsidized jobs
7.30
Following central government rules and guidelines
6.56
Activating unemployed persons living on social assistance in labor market
programs
6.36
Increasing or maintaining the local population
5.75
Attaining the quantitative goals of the National Labor Market
Administration
5.27
Reducing expenditure for social assistance
3.94
The agencies agree most on the importance of arranging labor market
programs for unemployed youth and the overall goal of reducing unemployment. Disagreement is highest concerning the importance of decreasing
social assistance expenditures and the quantitative goals of the National
Labor Market Administration. The overall goal congruence, which is the
measure to be used in the analysis, is the average score of the objectives in
Table 3. Descriptive statistics are reported in Table 1.
The selected goals are chosen because they are highly relevant in the
active labor market policy in general, especially for the relationship between
PES offices and municipalities. Objectives of importance in some relationships are probably missing, but I find it more reliable and more interesting to
focus on certain goals rather than asking a general question on goal
similarity.
88
Explaining cooperation
Trust
Two survey questions were asked in order to cover trust. The purpose of the
first question was to measure perceptions of the other actor’s intents: “To
what extent do you assume that the PES office (municipality) reveals true
motives to you?” The second question is aimed at capturing the agencies’
assumptions about whether the other agency will actually keep promises:
“To what extent do you assume that the PES office (municipality) performs
in line with its promises?” There were four possible response alternatives:
“completely”, “mostly”, “partly” and “not at all”. Trust is coded as a
dichotomous variable. A relationship is characterized by mutual trust when
both actors respond that they believe the other actor have honest intents and
will keep their promises all or most of the time (that is, the alternatives
completely and mostly), otherwise the level of trust is low. Sometimes researchers instead employ a battery of items (for example, using a Likert-type
scale) to measure trust. Some researchers also distinguish between diverse
forms of mutual trust. It is not possible for me to use such measures since I
have no access to that kind of data.
Trust is a somewhat problematic variable to deal with. It could be argued
that cooperation affects trust rather than the other way around. Thus,
conclusions should be made with some reservations. The problem of causal
order between trust and other variables is a recurrent problem in the social
sciences. Studies trying to establish time order between trust and cooperation
using, for example, panel data would be important contributions to the
research field. However, in the literature on interorganizational relationships
one usually treats trust as an independent variable and cooperation as the
dependent variable.
Control variables
The research setting makes it possible to hold characteristics of the policy
and the authorities that do not vary around Sweden constant. But control
variables should be added to account for variations among dyads that could
affect both the level of cooperation and some of the main explanatory
variables.
The local context is important to hold constant for at least two reasons.
First, contextual pressure is stronger in some places than in others. Evidence
reveals that when the times are turbulent and there is external pressure,
organizations tend to join together (Schermerhorn 1975). Contextual pressure can also affect the main explanatory variables. Second, alternative
collaborative partners and/or external resources are more readily available in
some areas than in others. This can affect the relationship between the PES
and the municipality.
89
Essay II
I have added four control variables to account for important variation
among local contexts. All four of them are based on data from official
statistical sources. A set of dummy variables are included to determine the
type of municipality. The context is likely to differ among big cities, small
cities, rural areas and so on but be quite alike within these groups. For
instance, the number of potential partners is expected to be larger in metropolitan areas than in sparsely populated municipalities. Municipality type is
based on a classification made by the Swedish Association of Local
Authorities. The scheme consists of nine categories covering, for example,
big cities, industry municipalities, and rural areas.8
The unemployment rate in percent in 2003 is another control variable. I
also incorporate the population change between 2000 and 2003, measured as
an index where the local population in 2000 is the base. High unemployment
and/or decreasing population may imply external pressure. Moreover,
additional resources from, for example, the Swedish Agency for Economic
and Regional Growth or from the European Union often become available
when the population decreases or when unemployment is high.
The last context variable is non-Swedish citizens, which is the percentage
of inhabitants having foreign citizenship (in 2001, since I do not have data
from 2003). Non-Swedish citizens is added mainly because I want to make
sure that one of the key explanatory variables, non-Nordic clients, only
account for the share of unemployed clients without Nordic citizenship. I expect a relatively large share with multiple boundary-spanning problems in
this group.
There a several features of the authorities that are important to take into
account. As previously noted, labor market policy is a national government
responsibility in Sweden. Municipalities can involve themselves when they
find it appropriate. Some municipalities are therefore more active in the
labor market policy than others. The level of cooperation and the main
explanatory variables could be affected by municipal engagement. The
agencies may, for example, trust each other as a consequence of municipal
activity in the field. The municipalities’ spending on labor market policies,
excluding financial contributions from the National Labor Market
8
Big cities (reference category in the regression analyses): Municipalities with more than
200,000 inhabitants. Suburbs: More than 50 percent of the employed in the municipality
travel to a big city to get to their job. Larger cities: More than 50,000 inhabitants in the
municipality and less than 40 percent employed in the industry sector. Middle-sized cities:
20,000–50,000 inhabitants, with less than 40 percent employed in the industry sector, and
more than 70 percent living in densely populated areas. Industry: Municipalities that are not
sparsely populated, with more than 40 percent employed in industry. Rural: Municipalities
that are not sparsely populated, where more than 6.4 percent are employed in agriculture and
forestry, and where more than 70 percent are living in densely populated areas. Sparsely
populated: Municipalities with less than 5 inhabitants per square kilometer and with less than
20,000 inhabitants. Other larger: Other municipalities with 15,000–50,000 inhabitants. Other
smaller: Other municipalities with less than 15,000 inhabitants. This classification is also used
in, for example, Dahlberg and Johansson (2002). Data source: KFAKTA03.
90
Explaining cooperation
Administration, in 100 crowns per inhabitant in 2003 is employed as an
indicator of municipal engagement (mun. LMP spending). What is more, a
control variable for the percentage of seats in the local council held by
socialist parties is incorporated (left parties).9 The active labor market policy
is to a large extent a socialist project, and it is reasonable to assume that
socialist municipalities are more involved in labor market programs. Both
variables are based on official statistics.
Some PES offices and/or municipalities may, for some reason, be more
open to cooperation with others (cf. Alexander 1995, 16). This general
orientation may affect the relationship between the PES and the
municipality. How can I make sure that it is not a general openness towards
collaboration that drives the results? The coping strategy is to include
variables measuring the level of communication between the PES and the
Regional Social Insurance offices (PES cooperative orientation), and between the municipalities and the Regional Social Insurance offices (mun.
cooperative orientation). The Social Insurance offices (Försäkringskassan)
are common partners to the PES offices (Lundin 2005, 15) and to the
municipalities (Swedish Association of Local Authorities 1999, 23). The
idea is that by including these two controls, factors that impinge on the
authorities’ relationship to both the Social Insurance offices and to each
other can be held constant. The measures are based on questionnaire data.
The respondents were asked how often they communicate with the Social
Insurance office. The ordinal response alternatives were recoded into a continuous variable. Various ways of doing this was tested, without any substantial change in the results. The following operationalization is used:
“daily contacts” is set to 18 times per month, “at least once a week” is set to
eight times per month, “at least once a month” is set to two times per month,
“at least once a year” is set to every second month, and “more seldom or
never” equals zero times per month.
The agencies’ own resources need to be taken into account. An agency
having a lot of financial resources is probably not that dependent on external
resources, all else equal. Two variables based on official figures for 2003 are
included to control for the municipalities’ financial strength: tax incomes
(mun. tax base) in 1,000 crowns per inhabitant and state subsidies (state
subsidy) in 1,000 crowns per inhabitant. Data to calculate the PES offices’
financial resources (PES finances) comes from the questionnaire. The
available benefits to clients participating in active programs in 1,000 crowns
per week, divided by the number of clients per week in 2003, is employed as
an indicator of the PES office’s financial strength.
9
As a robustness check, I have used an alternative operationalization of left parties.
According to the alternative assessment, a municipality is considered socialist if the chairman
of the executive board represents the Social Democrats. Results are not substantially altered.
91
Essay II
The size of the PES (PES size), operationalized as the number of
caseworkers at the office, is an additional control variable. The reason for
including this variable is primarily that the dependent variable is not relative
to office size. One of the items in the cooperation index is about frequency
of communication: Do the authorities have daily contacts? A PES with
around 100 employees is more likely to have daily contacts than an office
with only a couple of caseworkers. Thus, cooperation is somewhat biased
towards large agencies. By including size, this deficiency is taken into
account. I do not have reliable data on the size of the municipal labor market
administration. This does not, however, impose a big problem. Several of the
already mentioned control variables are likely to account for this possible
effect. Large administrations are, for example, expected to be found in large
cities, in municipalities governed by left parties, and in municipalities
spending a lot on labor market programs. As an additional test, I have
included the number of inhabitants in the municipality as a proxy of the size
of the labor market administration; the results are not substantially changed.
Some PES offices are specialized on work disabilities, and they have
rather different objectives than the ordinary offices. The results may, in some
way or another, be affected by this. Thus, I include a dummy variable called
work rehabilitation PES that takes on a value of one if the PES is specialized
on work disabilities, otherwise its value is zero.
Lastly, it is possible that the level of cooperation increases if the actors
perceive the relationship to be on equal terms. If one of them think that the
other is trying to dominate them, the incentives to collaborate may decrease.
A dummy variable (equal terms) is constructed from the questionnaire
responses as a measure. If both parties declare the relationship to be on equal
terms the variable gets a value of one, otherwise zero.10
Findings
Ordinary Least Squares (OLS) regression analysis is used to evaluate the
hypotheses. Table 4 reports the results from six different model specifications, with more controls added successively. In Models 1–3, interaction
terms are left out; the purpose is to demonstrate how the effects would look
like if the interactions are not taken into account. Models 4–6 include
interaction terms. In Models 1 and 4, environmental characteristics are used
as control variables. Agency features, except finances and whether the
relationship is set on equal terms, are added in Models 2 and 5. Models 3 and
10
The following question was asked: “How are the terms of your organization’s interaction
with the PES office (municipality) set?” The response alternatives were as follows:
“completely on the municipality’s terms”, “mostly on the municipality’s terms”, “equal
terms”, “mostly on the PES office’s terms” and “completely on the PES office’s terms”.
92
Explaining cooperation
6 include all control variables. Internal missing values, especially concerning
PES finances, imply a decrease of useable observations in the specifications
including all controls. Diagnostic plots and the Breusch-Pagan test indicated
problems of heteroscedasticity. For this reason, robust standard errors are
used in all specifications.
The coefficients in Models 1–3 are rather similar, and interpretation can
therefore be focused solely on the coefficients presented in Model 3. Longterm clients, non-Nordic clients, goal congruence, and trust affect cooperation along the lines of previous research; the signs of the b-coefficients
are expected, and the effects are statistically significant at conventional
levels. In a dyad where the authorities trust each other, compared to an
identical dyad without mutual trust, the predicted number of collaborative
activities will be about 0.45 higher, all else equal. If goal congruence
improves by one unit, the average level of cooperation changes by roughly
0.23 units, holding everything else constant. An increase of one percentage
point of clients being long-term unemployed or being non-Nordic citizens
increases cooperation by approximately 0.05 units on average, all else equal.
To get a better indication of the joint impact of the main independent
variables, it is possible to compare two fictitious dyads. In the first dyad, the
agencies put mutual trust in each other, goals are rather similar (goal
congruence held at the third quartile), and interdependence is relatively high
(long-term clients and non-Nordic clients held at the third quartile). In the
second dyad, the agencies do not trust each other, goals diverge (goal
congruence held at the first quartile), and mutual dependence is low (longterm clients and non-Nordic clients held at the first quartile). In all other
respects, the dyads are identical. The predicted level of cooperation will, on
average, be around 1.36 units higher in the first dyad. This seems like a
relatively large difference, given the scale of the dependent variable (0-5).
Note that it is always difficult to know if an effect is large or small. In
addition, the estimated coefficients are average effects. Comparing predicted
values is therefore a bit hazardous. Thus, caution is recommended. The
impacts of the variables non-Nordic citizens and long-term clients are the
most disputable ones. But we must recall that these variables only account
for a small part of resource interdependence. This means that it is not reasonable to expect them to make an enormous difference.
93
* p < .10; ** p < .05; *** p < .01.
Non-Nordic clients
Long-term clients
Goal congruence
Trust
Trust × Non-Nordic clients
Trust × Long-term clients
Trust × Goal congruence
Unemployment
Non-Swedish citizens
Population change
Work rehabilitation PES
PES size
PES cooperative orientation
PES finances
Mun. LMP spending
Left parties
Mun. cooperative orientation
Mun. tax base
State subsidy
Equal terms
Constant
Municipality type, dummies
Standard error of regression
Number of observations
–10.48 (6.55)
Yes
1.19
20
3
0.02 (0.09)
0.04 (0.02)*
0.04 (0.07)
–1.68 (0.54)***
0.02 (0.01)**
0.06 (0.02)***
0.12 (0.09)
0.03 (0.02)*
0.10 (0.06)
–6.90 (7.06)
Yes
1.08
202
–0.03 (0.04)
0.02 (0.01)*
0.03 (0.02)
2
0.05 (0.02)***
0.04 (0.02)*
0.29 (0.09)***
0.37 (0.15)**
1
0.04 (0.02)**
0.04 (0.02)*
0.25 (0.09)***
0.40 (0.17)**
-0.01 (0.10)
0.03 (0.02)*
0.05 (0.08)
–1.22 (0.36)***
0.02 (0.01)***
0.05 (0.02)***
–0.46 (0.39)
–0.01 (0.04)
0.01 (0.01)
0.01 (0.02)
–0.01 (0.01)
–0.05 (0.02)***
–0.26 (0.18)
–3.93 (8.53)
Yes
1.03
187
3
0.05 (0.02)***
0.05 (0.02)**
0.23 (0.09)***
0.45 (0.16)***
–8.96 (6.75)
Yes
1.19
203
4
0.03 (0.02)*
0.04 (0.03)
0.12 (0.13)
–1.82 (1.43)
0.00 (0.02)
–0.00 (0.04)
0.30 (0.18)*
0.12 (0.09)
0.03 (0.02)
0.09 (0.06)
–5.08 (7.42)
Yes
1.08
202
–0.03 (0.04)
0.02 (0.01)*
0.03 (0.02)*
5
0.05 (0.02)***
0.05 (0.03)*
0.12 (0.12)
–1.95 (1.30)
–0.00 (0.02)
–0.02 (0.03)
0.35 (0.15)**
0.02 (0.09)
0.04 (0.02)**
0.04 (0.07)
–1.87 (0.64)***
0.02 (0.01)***
0.05 (0.02)***
Table 4. OLS Regression Analysis with Cooperation as Dependent Variable (Robust Standard Errors in Parentheses)
6
0.04 (0.02)**
0.06 (0.02)**
0.09 (0.12)
–1.65 (1.27)
0.00 (0.02)
–0.01 (0.03)
0.30 (0.15)**
–0.01 (0.10)
0.03 (0.02)
0.04 (0.08)
–1.27 (0.39)***
0.02 (0.01)***
0.04 (0.02)***
–0.46 (0.37)
–0.00 (0.04)
0.01 (0.01)
0.01 (0.02)
–0.01 (0.01)
–0.04 (0.02)***
–0.27 (0.18)
–3.05 (8.85)
Yes
1.02
187
Explaining cooperation
Models 4–6 contain the relevant information to evaluate the four hypotheses.
However, it is not possible to assess the details directly from Table 4 since
interaction effects are estimates conditional on the values of the other
interaction variables. The results in Table 4 can only provide a rough indication. The b-coefficients of long-term clients and non-Nordic clients in
Models 4–6 are more or less unchanged in comparison with Models 1–3, and
the interaction terms that include these two variables are small and
statistically insignificant. This indicates that there is no interaction between
long-term clients and non-Nordic clients, on the one hand, and trust, on the
other hand. Thus, there is no support for hypotheses 1 and 3. However, the
effects of trust and goal congruence changes dramatically. Moreover, the
interaction term of trust and goal congruence is positive, large, and statistically significant at conventional levels. This suggests that there are interaction effects in accordance with hypotheses 2 and 4.
Table 5. Conditional Effects: The Effects of Non-Nordic Clients, Long-term
Clients, and Goal Congruence as the level of Trust Changes (Robust Standard Errors
in Parentheses)
Trust
Low
High
Non-Nordic clients
0.04 (0.02)**
0.05 (0.02)**
Long-term clients
0.06 (0.03)**
0.04 (0.03)
Goal congruence
0.09 (0.12)
0.39 (0.10)***
Note: The conditional effects are based on the estimates from Model 6 in Table 4.
* p < .10; ** p < .05; *** p < .01.
As noted above, interaction effects are conditional effects, and the analysis
should not stop at this point. Allison (1977), Braumoeller (2004), and
Friedrich (1982) thoroughly discuss the commonly made mistakes when
interaction variables are included in regression analysis. The main thrust of
their argument is that the b-coefficient and the standard error of a variable
included in an interaction term depend on the value of the other interaction
variable. It is therefore necessary to calculate the impacts of long-term
clients, non-Nordic clients, and goal congruence when trust is low and high,
and to compute the effects of trust at various levels of long-term clients, nonNordic clients, and goal congruence to find out the details. Table 5 and 6
report conditional effects based on the coefficients in Model 6 in Table 4.
The results stay about the same if the coefficients from Model 4 or 5 are
used instead. 11
11
See Allison (1977), Braumoeller (2004), and Friedrich (1982) for details on how to
calculate conditional effects and significance in interaction models.
95
Essay II
In Table 5, we can see that the effect of non-Nordic clients is almost the
same when trust is high and low: a one percentage point increase of
unemployed clients from outside the Nordic countries yields, on average, an
increase of cooperation by roughly 0.05 units in both cases (significant at the
.05 level). The effect is marginally larger when trust is high, but the
difference is negligible and not robust to model specification. A similar story
applies to long-term clients, but in this case, the b-coefficient is a bit larger
when trust is low and insignificant when trust is high. This is opposite to
what hypothesis 1 suggests. But the results are not robust to model
specification. In some specifications, the variable long-term clients has a
significant and positive effect of about the same magnitude when trust is
high as when it is low. In sum, the effect of resource interdependence does
not rely on mutual trust and hypothesis 1 is therefore rejected.
However, Table 5 shows that hypothesis 2 is supported. The impact of
goal congruence is statistically insignificant at any conventional level in
cases of low trust. But when trust is high, the effect is significant at least at
the .01 level and the size of the b-coefficient is more than three times as
large as when trust is low. Thus, the effect of goal congruence of 0.23 in
Model 3 hides that the impact is much stronger when organizations trust
each other and is insignificant when they do not. Given mutual trust,
cooperation increases by 0.39 units, on average, when goal congruence
increases by one unit, all else equal.
To see how trust affects cooperation it is necessary to turn to Table 6. The
Table consists of two panels. In panel A, goal congruence is set to 8.15. This
is the third quartile of the variable’s distribution, and it is (arbitrarily) chosen
to reflect a situation where the agencies’ objectives are rather similar. In
panel B, goal congruence is set to 6.77, which is the first quartile.
Consequently, panel B shows the effects of trust, given diverging objectives.
In both panels, long-term clients and non-Nordic clients vary. In these cases,
the third quartile represents high resource interdependence and the first
quartile represents low resource interdependence.
In order to evaluate hypothesis 3—that is, that the effect of trust increases
when resource interdependence increases—coefficients within each panel
should be compared. There is no indication that the impact of trust becomes
larger when non-Nordic clients and long-term clients take on high values.
The size of the coefficients and the significance levels are about the same
within each panel. If anything, the evidence is opposite to hypothesis 3.
Hypothesis 3 is therefore rejected.
96
Explaining cooperation
Table 6. Conditional Effects: The Effects of Trust as Non-Nordic Clients, LongTerm Clients, and Goal Congruence Changes (Robust Standard Errors in Parentheses)
Long-term clients
Few (8.14)
Many (14.70)
Panel A. Effects when goals are similar
(goal congruence = 8.15)
Non-Nordic clients
Few (3.28)
0.70 (0.23) ***
0.61 (0.18) ***
Many (8.61)
0.71 (0.20) ***
0.61 (0.18) ***
Few (3.28)
0.29 (0.27)
0.19 (0.23)
Many (8.61)
0.29 (0.25)
0.19 (0.24)
Panel B. Effects when goals diverge (goal
congruence = 6.77)
Non-Nordic clients
Note: The third quartile (8.15) of goal congruence represents a situation in which goals are similar,
whereas the first quartile (6.77) implies a situation when the agencies’ objectives diverge to a large
extent. The first quartile of non-Nordic clients (3.28) and long-term clients (8.14) represents low resource
interdependence, and the third quartile of non-Nordic clients (8.61) and long-term clients (14.70) means
high resource interdependence. The conditional effects are based on the estimates from Model 6 in
Table 4. The effects and the standard errors have been calculated in the same way as in Table 5; see
Allison (1977), Braumoeller (2004), and Friedrich (1982).
* p < .10; ** p < .05; *** p < .01.
If we instead compare figures between panel A and panel B, hypothesis 4
can be evaluated. That is, the impact of trust is examined at different levels
of goal congruence. The findings are conclusive: panel A shows statistically
significant and large positive effects of trust when objectives are similar,
whereas panel B indicates statistically insignificant and small effects of trust
given diverging objectives. This is precisely what hypothesis 4 suggests.
Thus, the effects of trust presented in Models 1–3 hide the fact that the
impact is much larger when goals are similar and that we cannot expect a
positive impact if organizations have very different agendas.
To examine the robustness of the results, several diagnostic tests have
been performed.12 The most important tests can be mentioned. First,
numerous model specifications including different controls have been
examined—results are robust. Second, I have examined many diagnostic
plots looking for nonlinear relationships and outliers. Some outliers are detected, but I find no substantial reason to exclude them. Nonetheless, the
models have been reestimated without them. The results remain more or less
unchanged. Third, one concern when analyzing dyadic data is possible
12
See Fox (1991) for an overview of most of the diagnostic tests presented here.
97
Essay II
interdependence across observations: when one actor is present in several
dyads there is a risk of autocorrelation (see, for example, Mizruchi 1989).
Some municipalities are included in more than one dyad. Since they are few,
interdependence across observations should not be a big problem. As a
precaution, regression analysis excluding dyads consisting of cases that
appear more than once has been performed. In some specifications, the
variable long-term clients turns out to be insignificant (but not in Model 6),
which implies some uncertainty regarding the effect of this variable. Fourth,
I have also tested excluding dyads in which the response from the municipal
side of the dyad is provided by a politician. Again, the variable long-term
clients comes out insignificant in a couple of specifications. Fifth, tolerance
tests for multicollinearity have been carried out. In Models 1–3, it is
definitely not a problem. However, the diagnostic statistics indicate that
multicollinerarity is present in Models 4–6. This is a usual setback in
interaction models. But I do not worry that much in this case. The problem
appears to apply mainly to the variables trust and the interaction of trust and
goal congruence. The interaction term is statistically significant anyway,
which indicates that the effect of trust is lower when objectives diverge and
that an interaction ought to be included. It must be noted, however, that the
multicollinearity problem related to the variable trust could imply that trust
has a significant effect at somewhat lower levels of goal congruence than
what is indicated by Table 6.
To sum up, the most important problems indicated by the robustness
checks is that the impact of long-term clients is a bit hazardous and perhaps
some problems of multicollinearity. Otherwise, the findings show that
mutual resource dependence enhances collaboration regardless of whether
the authorities trust each other or not. The findings also imply that the logic
of interorganizational cooperation suggested in additive models should be
developed: similar goals and mutual trust do promote cooperation but
probably only when they occur simultaneously.
Conclusion
Implementation of political decisions will continue to rely on joint efforts of
various authorities. Governance in these contexts is challenging and knowledge on how cooperation can be explained is of significant interest. In this
article, new data on the relationships between Swedish PES offices and
municipalities has been utilized in order to evaluate potential explanations of
interorganizational cooperation.
Prior research suggests that authorities cooperate because they are resource interdependent, have congruent objectives, and trust each other. This
explanation holds in the present case as well. Cooperation between PES
offices and municipalities is at least partly a consequence of a situation
98
Explaining cooperation
where the authorities need one another to solve problems, give priority to
similar objectives, and trust one another.
However, the argument in the article is that theory should be adjusted to
incorporate interaction effects. The interpretation of the empirical findings
must be made with some reservations since it is difficult to be certain that
causal effects have been identified. But the results suggest that congruent
goals do not promote cooperation if authorities do not trust each other.
Furthermore, it seems like mutual trust does not enhance cooperation when
authorities have very different priorities. That is, mutual trust and goal congruence must exist simultaneously. The empirical evidence indicates that
including interaction terms probably is going to improve our understanding
of interorganizational cooperation. On the other hand, the impact of resource
interdependence is not dependent on a high level of mutual trust. Moreover,
the effect of trust is not reliant on resource interdependence. To understand
these results better, more research is required.
What are the management implications of the findings? What should
public sector managers that would like to increase collaboration do? This is
hard to tell and it is probably wise not to draw far-reaching conclusions.
However, some speculations would do no harm. One management strategy
would be to highlight what agencies have in common and introduce
transparent reporting systems between the organizations. By doing this
congruent objectives can perhaps be communicated in a trustworthy way. It
is not a feasible strategy to build up trust if the authorities’ goals diverge to a
large extent or to focus solely on objectives without establishing trust. That
is, the strategy must include both objectives and trust. On the other hand,
situations of resource interdependence seem, more or less, to enhance
cooperation by itself.
The hypotheses examined in the article are general, although the
empirical tests are focused on labor market policies in the unitary Swedish
system. There is no strong reason to assume that the research setting is
atypical to other situations of local policy implementation. We can expect
the results to hold for other similar situations as well. Nevertheless, some
cautionary points should be mentioned. First, the empirical setting implies a
rather truncated range on goal congruence by comparison with other types of
collaborations. Disagreements might be more severe in other settings, for
example, between public agencies and private enterprises. A good way to
test the robustness of the results is thus to examine how they would fare
under other conditions, in other countries, and in other policy areas. Second,
the measures of resource interdependence are far from being perfect since
interdependence is not observed directly. The fact that the results do not
indicate an interaction effect between trust and mutual resource dependence
might be a consequence of the somewhat problematic measures. Consequently, in future research, it would be beneficial if other indicators of resource interdependence are examined. Third, I have taken some license to
99
Essay II
speculate on causal relationships while operating within the confines of
cross-sectional data. Obviously, a longitudinal study would be suitable to
test the propositions put forward in this article. In-depth case studies, or
quantitative research, focusing on potential causal mechanisms would also
be valuable. Even though the mechanisms are logical, they are assumed
rather than tested in this particular case. In sum, and as always, work
remains to be done. However, this study has moved the research agenda
forward and provided insights about interorganizational cooperation when
public policies are carried out at local level.
100
Explaining cooperation
References
Agranoff, Robert. 2004. JPART Symposium introduction: Researching
intergovernmental relations. Journal of Public Administration Research
and Theory 14(4):443–6.
Alexander, Ernest R. 1995. How organizations act together.
Interorganizational coordination in theory and practice. Luxembourg:
Gordon and Breach Publishers.
Allison, Paul D. 1977. Testing for interaction in multiple regression.
American Journal of Sociology 83(1):144–53.
Alter, Catherine, and Jerald Hage. 1993. Organizations working together.
Newbury Park, CA: Sage.
Bardach, Eugene. 1998. Getting agencies to work together. The practice and
theory of managerial craftsmanship. Washington D.C.: Brookings
Institution Press.
Blau, Peter M. 1964. Exchange and power in social life. New York: John
Wiley.
Braumoeller, Bear F. 2004. Hypothesis testing and multiplicative interaction
terms. International organization 58:807–20.
Calmfors, Lars, Anders Forslund, and Maria Hemström. 2002. Does active
labour market policy work? Lessons from the Swedish experience.
Working Paper 4, Institute for Labour Market Policy Evaluation, Uppsala,
Sweden.
Carling, Kenneth, and Laura Larsson. 2005. Does early intervention help the
unemployed youth? Labour Economics 12:301–19.
Dahlberg, Matz, and Eva Johansson. 2002. On the vote-purchasing behavior
of incumbent governments. American Political Science Review 96(1):27–
40.
Elster, Jon. 1989. The cement of society. A study of social order. Cambridge,
UK: Cambridge Univ. Press.
Forslund, Anders, Daniela Fröberg, and Linus Lindqvist. 2004. The Swedish
activity guarantee. OECD Social, Employment and Migration Working
Paper 16, OECD, Paris, France.
Fox, John. 1991. Regression diagnostics. Newbury Park, CA: Sage.
Friedrich, Robert J. 1982. In defense of multiplicative terms in multiple
regression equations. American Journal of Political Science 26(4):797–
833.
101
Essay II
Gulati, Ranjay, and Martin Gargiulo. 1999. Where do interorganizational
networks come from? American Journal of Sociology 104(5):1439–93.
Hanf, Kenneth, Benny Hjern and David O. Porter. 1978. Local networks of
manpower training in the Federal republic of Germany and Sweden. In
Interorganizational policy making. Limits to coordination and central
control, ed. K. Hanf and F. W. Scharpf, 303–41. London: Sage.
Hardin, Russell. 2002. Trust and trustworthiness. New York: Russell Sage
Foundation.
Hjern, Benny and David O. Porter. 1981. Implementation structures: A new
unit of administrative analysis. Organization Studies 2(3):211–27.
Hjertner Thorén, Katarina. 2005. Municipal activation policy: A case study
of the practical work with unemployed social assistance recipients.
Working Paper 20, Institute for Labour Market Policy Evaluation, Uppsala,
Sweden.
Jennings, Edward T., Jr., and Jo Ann G. Ewalt. 1998. Interorganizational
coordination, administrative consolidation, and policy performance. Public
Administration Review 58(5):417–28.
Levine, Sol, and Paul E. White. 1961. Exchange as a conceptual framework
for the study of interorganizational relationships. Administrative Science
Quarterly 5(4):583–601.
Lundin, Martin. 2005. Does cooperation improve implementation? Centrallocal government relations in active labour market policy in Sweden.
Working Paper 2, Institute for Labour Market Policy Evaluation, Uppsala,
Sweden.
Lundin, Martin, and Per Skedinger. 2006. Decentralisation of active labour
market policy: The case of Swedish local employment service committees.
Journal of Public Economics 90(4-5):775–98.
Luo, Yadong. 2001. Antecedents and consequences of personal attachment
in cross-cultural cooperative venture. Administrative Science Quarterly
46(2):177–201.
Mizruchi, Mark S. 1989. Similarity of political behavior among large
American corporations. American Journal of Sociology 95(2):401–24.
Muthusamy, Senthil K., and Margaret A. White. 2005. Learning and
knowledge transfer in strategic alliances: A social exchange view.
Organization Studies 26(3):415–41.
National Board of Health and Welfare. 2004. Ekonomiskt bistånd,
årsstatistik 2003. Utbetalda belopp samt antal biståndsmottagare och antal
102
Explaining cooperation
biståndshushåll. Socialtjänst 2004:8. Stockholm, Sweden: National Board
of Health and Welfare (Socialstyrelsen).
Oliver, Amalya L., and Mark Ebers. 1998. Networking network studies: An
analysis of conceptual configurations in the study of inter-organizational
relationships. Organization studies 19(4):549–583.
Ostrom, Elinor. 1998. A behavioural approach to the rational choice theory
of collective action: Presidential address, American political science
association, 1997. American Political Science Review 92(1):1–22.
O’Toole, Laurence J. Jr. 1983. Interorganizational co-operation and the
implementation of labour market training policies: Sweden and the Federal
Republic of Germany. Organization Studies 4(2):129–50.
———. 2003. Interorganizational relations in implementation. In Handbook
of public administration, ed. J. Pierre and B. G. Peters, 234–44. London:
Sage.
O’Toole, Laurence J., Jr., and Robert S. Montjoy. 1984. Interorganizational
policy implementation: A theoretical perspective. Public Administration
Review 44(6):491–503.
Pressman, Jeffery L. and Aaron Wildavsky. 1984. Implementation. How
great expectations in Washington are dashed in Oakland; or, why it's
amazing that federal programs work at all, this being a saga of the
economic development administration as told by two sympathetic observers
who seek to build morals on a foundation of ruined hopes. 2nd ed. Los
Angeles: Univ. California Press.
Provan, Keith G., and Steven J. Skinner. 1989. Interorganizational
dependence and control as predictors of opportunism in dealer-supplier
relations. Academy of Management Journal 32(1):202–12.
Putnam, Robert D. 2000. Bowling alone. The collapse and revival of
American community. New York: Simon & Schuster.
Radin, Beryl A. 2003. The instruments of intergovernmental management. In
Handbook of public administration, ed. J. Pierre and B. G. Peters, 607–18.
London: Sage.
Rothstein, Bo. 2000. Trust, social dilemmas and collective memories.
Journal of Theoretical Politics 12(4):477–501.
Salonen, Tapio, and Rickard Ulmestig. 2004. Nedersta trappsteget. En
studie om kommunal aktivering. Växjö, Sweden: Växjö Univ.
Scharpf, Fritz W. 1978. Interorganizational policy studies: Issues, concepts
and perspectives. In Interorganizational policy making. Limits to
103
Essay II
coordination and central control, ed. K. Hanf and F. W. Scharpf 345–70.
London: Sage.
———. 1997. Games real actors play. Actor-centred institutionalism in
policy research. Oxford, UK: Westview.
Schermerhorn, John R. Jr. 1975. Determinants of interorganizational
cooperation. Academy of Management Journal 18(4):846–56.
Schmidt, Stuart M., and Thomas A. Kochan. 1977. Interorganizational
relationships: Patterns and Motivations. Administrative Science Quarterly
22(2):220–34.
Smith, Ken G., Stephen J. Carroll, and Susan J. Ashford. 1995. Intra- and
interorganizational cooperation: Toward a research agenda. Academy of
Management Journal 38(1):7–23.
Svensson, Torsten, and PerOla Öberg. 2005. How are co-ordinated market
economies co-ordinated? Evidence from Sweden. West European Politics
28(5): 1075–100.
Swedish Association of Local Authorities. 1999. Kommunerna i arbetsmarknadspolitiken: en enkätundersökning. Stockholm, Sweden: Swedish
Association of Local Authorities (Kommunförbundet).
Thomas, Robert D. 1979. Implementing federal programs at the local level.
Political Science Quarterly 94(3):419–35.
Van de Ven, Andrew H. and Gordon Walker. 1984. The dynamics of
interorganizational coordination. Administrative Science Quarterly
29(4):598–621.
Wholey, Douglas R. and John W. Huonker. 1993. Effects of generalism and
niche overlap on network linkages among youth service agencies. Academy
of Management Journal 36(2):349–71.
Zaheer, Akbar and N. Venkatraman. 1995. Relational governance as an
interorganizational strategy: An empirical test of the role of trust in
economic exchange. Strategic Management Journal 16(5):373–92.
Zand, Dale E. 1972. Trust and managerial problem solving. Administrative
Science Quarterly 17(2):229–39.
104
Essay III
When Does Cooperation Improve Public
Policy Implementation?*
Relationships between authorities impinge on policy implementation
(O’Toole 2003) and it is often emphasized that public policy can be carried
out better if cooperation increases among agencies. Some scholars claim that
interorganizational cooperation is inherently good (for example, Hudson et
al. 1999; Jones, Thomas, and Rudd 2004), although empirical evidence
suggests that cooperation only sometimes enhances performance (Jennings
and Ewalt 1998). The purpose of this article is to enrich our understanding of
policy implementation by examining when a cooperative strategy actually
makes implementation output better. More precisely, it is argued that the
effects of cooperation vary with the complexity of the policy carried out.
An agency that cooperates with others can make use of additional
resources, such as expertise and information. Activities can hopefully also be
better coordinated. This suggests that interorganizational cooperation improves an agency’s ability to put policy into practice. On the other hand, it is
difficult to work across organizational boundaries. For example, the collaborating authorities have to devote a lot of time and other resources to
establishing and maintaining a productive relationship. Thus, we cannot be
sure that cooperation improves implementation in every situation. In this
article, it is suggested that task complexity is a key factor that explains why
the effect of cooperation on implementation output will be greater in certain
situations. Basically, the benefits of cooperation are likely to more than
offset the costs if the task is complex. In contrast, when the policy is not
complex cooperation adds very little value, but still involves some costs. To
my knowledge, this idea has previously not been thoroughly discussed and
empirically evaluated in implementation research.
*
I gratefully acknowledge comments from Hans Blomkvist, Paula Blomqvist, Matz Dahlberg,
Jörgen Hermansson, Nils Hertting, Anders Lidström, Daniela Lundin, Eva Mörk, PerOla
Öberg, Sven Oskarsson, and seminar participants at the Department of Government at
Uppsala University, the Institute for Labour Market Policy Evaluation (IFAU), and the
Annual Meeting of the Swedish Political Science Association 2004 for comments on early
drafts of the article. I also thank Linus Lindqvist for some very valuable data management and
IFAU for funding.
105
Essay III
The empirical focus is relations between different units of government.
More precisely, the analysis concerns Public Employment Service (PES)
offices and their relationship to local governments (municipalities) in the
Swedish active labor market policy (ALMP). The PES is the main local
labor market actor. But in recent years, municipalities have become an increasingly important factor (Lundin and Skedinger 2006; Salonen and
Ulmestig 2004). The Swedish government encourages cooperation between
these actors. But will implementation of labor market activities actually be
improved if the PES cooperates a lot with the municipality? Is it possible to
anticipate a positive effect on a broad range of policies, or is it only certain
labor market activities that are affected positively?
Two policies are examined. One of the policies—activities for unemployed youth—is not that complex. The other—activities for individuals with
especially long spells of unemployment—is more intricate. Thus, I expect
cooperation to be a more efficient strategy in the latter case. The quantitative
analysis is based on recent data consisting of information from several
sources. I make use of a questionnaire addressed to chief managers of the
PES offices, and administrative data recorded by the Swedish National
Labor Market Administration and Statistics Sweden.
The findings indicate that cooperation improves the implementation of
activities for individuals unemployed for an especially long time. On the
other hand, cooperation does not enhance the implementation of the youth
policy. This indicates that complex tasks can be carried out better if a cooperative approach is employed, but that it is not reasonable to assume that
interorganizational cooperation will always have a positive impact on how
policies are implemented.
The rest of the article is structured as follows. First, the theoretical
discourse is outlined. An introduction of the research setting comes next,
followed by a section that discusses methodological issues and measures.
Empirical findings are reported. I conclude by summing up the results and
discussing their possible implications.
Implementation and interorganizational cooperation
Local practices are not always the same as the intentions stated in official
documents endorsed by politicians. In addition, performance frequently
varies from one local context to another. As a result, it is wise not to assume
that the study of statutes, government bills, and regulations will be enough to
understand what political decisions imply “in the real world.”1
1
For overviews of implementation research, see, for example, Winter (2003b) and deLeon
(1999).
106
When does cooperation improve public policy implementation?
An implementation problem occurs when a political decision is not
carried out in accordance with what the decision-maker wants. We assume
that local practice should be in line with the elected officials’ intentions, that
is, agents should follow the principal’s instructions. This perspective is easy
to endorse, based on normative democratic theory. In a modern democracy,
citizens freely elect representatives who can be held to account at the ballot
box. Politicians cannot implement a policy all by themselves; they have to
rely on a civil service to do this. But the citizens cannot replace the bureaucracy by casting their vote. Consequently, a prerequisite for satisfactory
democracy is that politicians control and govern a civil service that respects
their decisions (Sannerstedt 2001).
There are arguments in favor of another point of view, which argues that
the local civil service is more receptive to local desires and needs. Discrepancies between a decision and its implementation might therefore in
practice mean a “better” policy and a greater responsiveness to citizens’
wishes. Depending on which normative starting-point is assumed, variation
in—or insufficient—implementation performance is thus not always necessarily a disadvantage (deLeon 1999). Nonetheless, most scholars would
concur that it is troublesome if there is a gap between the law and practice
(Keiser and Soss 1998). In this article, I focus on the top-down aspect of
implementation, that is, implementation as compliance. It is, however,
necessary to acknowledge that other angles of approach are just as relevant
in implementation research.
Certain political decisions are quite easy to carry out and can be managed
almost exclusively by a unitary public administration. A change in a tax rate
or the level of a general welfare benefit is virtually self-implemented. But
implementation is normally more complicated and involves several participants, who may be public or private. They may also be at different levels
of government or operate in different policy areas. Hence, an important
component in almost every contemporary framework explaining implementation success and failure has to do with how interorganizational relationships are managed (for example, Bardach 1998; Goggin et al. 1990; Hjern
and Porter 1981; O’Toole 2003; O’Toole and Montjoy 1984; Pressman and
Wildavsky 1984; Winter 2003a). Furthermore, scholars have paid a lot of
attention to the problem of multi-level governance in recent years. That is,
how actors at various levels of government can be managed to work
effectively towards public sector goals (Hooghe and Marks 2003; Peters and
Pierre 2001). A particularly important subject, given a focus on multi-level
governance, is how relationships between levels of government impinge on
public policy.
An agency assigned the task of carrying out a political decision may use
different techniques to implement a policy. For instance, the agency may try
to cooperate as much as possible with other organizations. Cooperation
(collaboration)—that is, all the interactions among organizations aiming at
107
Essay III
solving public problems by working together (cf. Smith, Carroll, and
Ashford 1995)—is one of the golden words in public sector management.
Through partnerships and other collaborative endeavors public sector
performance is often considered to be improved. All else being equal, one
would expect a public agency with access to significant resources to carry
out a policy or program better than an agency which lacks resources (for
example, Keiser and Soss 1998; Meier and McFarlane 1996). Organizations
possess resources. An agency carrying out a political decision may thus
enhance its own capacity by collaborating with other organizations (for
example, Jennings 1994). The surrounding organizations may contribute
with information, and they could have staff, knowledge, money, and
premises, making the business of putting ideas into practice easier. By coordinating activities across organizational boundaries, the implementing
agency can also avoid resource-consuming and conflicting activities that
may result in socially perverse outcomes (Hooghe and Marks 2003). Thus, it
is easy to see why interorganizational cooperation is often assumed to improve implementation.
But there are a number of prerequisites. First, the potential partner must
have additional resources that can be of use to the focal agency. Second, the
partner must be willing to share resources. Third, there are always costs
associated with collaboration (for example, Schermerhorn 1975; Van de Ven
1976; Weiss 1987). Although a potential partner is willing to share valuable
resources with the focal agency, cooperation may be a complicated process.
To at least some extent, all organizations have different agendas and various
routines to handle things. As a consequence, interorganizational cooperation
may be a complicated process involving bargaining, and therefore call for
considerable time and other resources on the part of the agents concerned.
Another cost to be considered is that the principal’s objectives—for
example, central government intentions—may be put on one side when the
local agents concerned are trying to reach their best joint solution. In other
words, instead of pursuing central government directives, local actors may
work for other goals they can agree on. Since this study adopts a top-down
perspective on implementation, such processes are not desired. In sum, the
costs of cooperating might well outweigh the benefits and we should not
assume that more cooperation implies that political intentions are always
realized to a greater extent.2
Cooperation can, of course, have important implications for other aspects
than implementation output. For instance, a public program might become
2
In an overview article on cooperation (not specifically about implementation issues) Smith,
Carroll, and Ashford (1995, 17) note that research usually has a positive attitude towards
cooperative activities, although cooperation in some cases can have harmful consequences.
The authors underscore that “additional research is needed on the potential drawbacks of
cooperation and the conditions under which a very high degree of cooperation is not
desirable.”
108
When does cooperation improve public policy implementation?
more effective if authorities cooperate, even if implementation is not
enhanced.3 Cooperation might also imply that a decision becomes more
legitimate in the eyes of the target group for a certain policy. But this study
is limited to the question of how to make a political decision come true in
accordance with officials’ intentions.
Jennings and Ewalt (1998) note that there is only anecdotal evidence that
coordination—which in practice is measured as the level of cooperation—
actually improves public services. Jennings and Ewalt examine the
accomplishment of policy goals in employment and training services in the
United States. In a prior article, Jennings (1994) indicated positive effects of
coordination on the administrators’ subjective perceptions of performance.
But in the study from 1998 objective outcome measures were employed as
dependent variables. The analysis shows that coordination has a limited
positive effect; most of the performance indicators were unaffected by the
level of coordination. The findings suggest that although interorganizational
cooperation may sometimes be a good strategy, we cannot expect it always
to improve performance. There is some indication that there is a positive
effect in the long run, but the evidence is not decisive.
On the other hand, Hudson et al. (1999, 238) note that “while recognizing
that there are other positions, this article takes the normative position that
collaboration is generally a ‘good thing’ – a stance which is consistent with
the rather long history of collaboration in organization theory and public administration.” In many ways this is certainly true, but if we want to
understand public administration and public policy implementation it must
be better to improve theory on what the balance of benefits and costs would
look like in various situations. This can tell us when cooperation actually is a
good strategy.
There seems to be a lack of research on these matters. Pressman and
Wildavsky (1984) suggest that the number of actors involved in an interorganizational setting determines whether implementation will be successful;
more actors mean greater probability of failure. O’Toole and Montjoy (1984)
refine this argument and hold that a large number of actors makes implementation worse in cases of reciprocal interdependence (when actors
poses contingencies for each other) and in sequential interdependence (when
the output of one actor is the input of another). Contrarily, in cases of pooled
interdependence (when actors contribute to a task without dealing with each
other) and in sequential competitive interdependence (when there are actors
competing to do the same thing in the implementation process) a larger number of actors improves policy output. These scholars focus on the effects of
3
Even if a decision is carried out perfectly in line with the politicians’ intentions, that does
not entitle one to say that the policy is effective. The decision itself could be based on a causal
theory that is not accurate, that is, it is not certain that desired outcomes are achieved by the
formula the statute or regulation suggests. This article focuses exclusively on policy outputs,
not outcomes.
109
Essay III
the organizational setting per se, not the extent to which actors work
together, which is the focus in this article. The question here is rather
whether a lot of cooperation generates a better pay-off, given certain tasks.
More specifically, is a high degree of cooperation especially rewarding for
certain policy types?
The discussion is restricted to one policy dimension that deserves
particular attention: task complexity. Task complexity is defined as the product of the scope of actions and the intensity required to carry out a policy.
By scope I mean the extent to which many areas of expertise are required to
fulfil the goals or, put another way, “the degree to which tasks are variable
and require a multidisciplinary or multidimensional approach” (Alter and
Hage 1993, 117). Intensity is a matter of how much work has to be done in
order to realize ambitions; a task that requires many man-hours on the part of
an agency is considered intense. Thus, a policy that has large scope and
involves intense actions is complex. In research on interfirm partnerships, it
is quite common to talk about joint task complexity as a function of the
scope and depth (intensity) of interactions between firms (White 2005). This
definition parallels the one used in the present study. Most organization
theory scholars define complexity in a related vein, although there are
alternatives on the same theme (see, for example, Alter and Hage 1993, 116–
127).
Task complexity is an incentive for cooperation (cf. Alter and Hage
1993). A need for external resources boosts interorganizational cooperation
(for example, Van de Ven and Walker 1984; Weiss 1987) and complex tasks
require resources that cannot be found within a single agency. Thus, working
across organizational boundaries becomes important when complex tasks are
being carried out. In this perspective, actors cooperate when cooperation is
needed, otherwise not.
But the extent of cooperation between two agencies, A and B, is certainly
not only a direct function of the task complexity of policy X that A implements. Cooperation is a consequence of the agencies’ perceived total
value of working together, which may be something rather different from the
actual value that cooperation adds to the task of realizing policy X in
accordance with official intentions. First, the agencies do not have perfect
information and cannot take in all benefits and costs of cooperation. A does
not know exactly what it can accomplish without a lot of cooperation and the
greater value cooperation actually implies. Second, A cannot set cooperation
precisely at the optimal level on its own. If B does not want cooperation to
increase, A cannot do much about this since cooperation is a relational
attribute. Third, the incentives and disincentives for A to cooperate with B
are numerous.4 External pressure for cooperation is one example. Political
representatives or other bureaucratic levels may exhort A to collaborate with
4
For an overview, see Alexander (1995, ch. 1).
110
When does cooperation improve public policy implementation?
B. In some cases, a legal mandate may require A to cooperate with B. Norms
and values in both organizations are other things that determine what level of
cooperation will be established. Moreover, working together with B can
make it easier for A to accomplish other goals than those related to policy X.
Accordingly, although the net benefit of cooperation for A can be positive,
this does not mean that policy X automatically becomes better implemented.
What this discussion has shown is that an agency carrying out a policy
may use interorganizational cooperation to a lesser extent than another
agency implementing the same policy, although both agencies would have
made the same use of cooperation for realizing the specific policy in
accordance with political intentions. This means that it is too restrictive to
assume that task complexity impinges only on cooperation levels. It can also
tell us something about the effect on policy output. Whereas the other factors
mentioned above affect cooperation, it is only the cost-benefit analysis related to the policy being implemented that determines the effect on implementation in the specific case.
So the question is now why we should assume that the effect of cooperation would be greater as policy complexity increases. The impact of
cooperation is contingent on the benefits and the costs. In a less complex
policy, it is likely that the benefits of cooperation would be quite low. In
these situations, the agency’s capacity is not going to increase that much by
using other organizations’ resources since the most important resources are
already available within the agency. Organizations are formed to perform
certain tasks and in a less complex policy the most important resources are
probably already available. This means that increasing levels of cooperation
do not add that much value. As complexity rises, cooperation provides more
benefits. The agency needs more expertise, information, money, premises,
and so on. Through cooperation these resources can be attained.
The costs must also be considered before we can conclude that cooperation should be more rewarding in the case of a complex issue. As
complexity increases, the costs of cooperation are also likely to increase. For
instance, the risk of disputes and impediments becomes greater. But while
the benefits of cooperation are very low given little complexity, there will
always be costs. In fact, many costs are independent of the level of
complexity. The costs of arranging and participating in a meeting are, for
example, not much larger in the case of a complex policy. Cooperation is a
dynamic process and decisions and activities within a policy are not independent of each other. Given that actors agree on some basic things, the
additional problems that arise with increasing complexity will not be that
high.
In another context—interfirm alliances—White (2005) considers the costs
associated with cooperation. Even though he claims that costs are higher
given a complex task, he repeatedly notes that the benefits are often likely to
be even greater. He claims that “additional costs of a more complex interface
111
Essay III
may be more than offset by the benefits possible from more extensive
interaction” (1388) and if the partners have relatively similar objectives and
values, “the additional marginal cooperation costs resulting from greater
scope or depth-related coordination costs should be more than offset by
increased benefits” (1395).
In sum, it seems likely that in a situation characterized by low task
complexity there will be very little (or no) benefit from cooperation, but
some costs. These costs increase somewhat as complexity increases, but the
increase will be lower (and probably also diminishing) than the increase in
benefits. If complexity is high, the large benefits will prevail over the costs.
Cooperation is therefore a more value adding activity given complex tasks.
Accordingly, we should expect the effect of cooperation on policy output to
be greater given complex tasks. These arguments are perhaps not brand new,
but to my knowledge they have not been discussed and evaluated empirically
in the context of local policy implementation.
Intergovernmental cooperation in Swedish active labor
market policy
To discern whether policy complexity is an important variable, we need to
know the effects of cooperation on the implementation of at least two
policies: one complex task and another less complex task. Swedish ALMP
provides rather good opportunities to examine this.
Unemployment is a huge problem in most OECD countries and the
unemployment rate has, on average, doubled since 1973 (Martin and Grubb
2001). A government can use ALMPs to reduce unemployment. Job
brokering activities and labor market programs are examples of such active
measures. In Sweden, the National Labor Market Administration (Arbetsmarknadsverket) implements the policy. Local PES offices (Arbetsförmedlingar) carry out most of the programs in practice. There is a PES in
nearly all cities, and in larger cities there is usually more than one office.
Although the National Labor Market Administration is clearly the most
important authority managing ALMPs in Sweden, other actors are involved.
One feature in recent years has been more involvement on the part of local
governments in the implementation of ALMPs (for example, Lundin and
Skedinger 2006; Salonen and Ulmestig 2004). The municipalities take an
active part by, for example, organizing labor market programs. About 40
percent of the clients involved in programs administered by the National
Labor Market Administration are participating in activities in which the
municipalities are involved (Lundin and Skedinger 2006). This is a large
proportion. A lot of the municipal activities are targeted at social welfare
benefit recipients (Salonen and Ulmestig 2004) and at unemployed youth
112
When does cooperation improve public policy implementation?
(Carling and Larsson 2005). The content of the activities varies, but jobsearch assistance and work practice are quite common. A majority of
municipalities have recently established special labor market administrations. There was an increase in the number of municipal labor market administrations in the 1990s, indicating an increased municipal
involvement in Swedish ALMPs. Seven in ten municipalities had a special
department for labor market issues in 1998, compared to two in ten in 1995
(Swedish Association of Local Authorities 1999).
The municipalities have a premier position within the Swedish political
system. For example, they have the constitutional right of self-government;
they are comparably large; they can decide on their own organization; and
their incomes come mostly from a proportional income tax they can set
freely.5 In comparative perspective, local governments in Sweden are often
considered to have a powerful position (Norton 1994).
The municipalities provide most of the services of the Swedish welfare
state. For instance, they supply day care, care of the elderly, social welfare
services, introduction of immigrants into Swedish society, and primary
education. Local development is another issue that concerns local governments. Several of these policy areas—the most obvious example being social
welfare services—clearly intertwine with ALMPs. This suggests that
collaboration with the municipality could be beneficial for the PES. In
addition, municipalities are also very large employers and employ 20 percent
of the total Swedish workforce. Cooperating with large employers is, of
course, also potentially very beneficial for the PES.
In the empirical analysis, effects of cooperation between the PES and the
municipality on implementation of ALMPs are examined. The study concerns two policies. One of the policies is definitely more complex. The
positive effect of cooperation is anticipated to be greater in this policy.
In August 2000, the Swedish government launched a new labor market
program called the Activity Guarantee (Aktivitetsgarantin). The target group
was individuals who had been unemployed for a considerable time period.
By means of intense job-search assistance and close monitoring, persons
who were, or who risked being, long-term unemployed were to be given a
place in the program. All traditional ALMPs could be used within the
Activity Guarantee. Accordingly, the content of the activities varies to a
large extent. The political ambition was that jobseekers should be enrolled in
the program after 27 months of unbroken PES registration, at the latest.
What distinguishes the Activity Guarantee from other labor market programs
is its intensity. All activities are assumed to be full-time activities and the
participants should meet their personal PES supervisor on a regular basis,
and more frequently than before they entered the program. Since the persons
5
For some descriptions of the Swedish municipalities, see Bäck (2003) and Gustafsson
(1999).
113
Essay III
participating in the guarantee are “hard cases” and often have multiple
problems, the government calls for intense treatment and encourages PES
offices to collaborate with other local actors. The local governments are, in
practice, the PES offices’ most common cooperation partners (Forslund,
Fröberg, and Lindqvist 2004).
In 1994, the Swedish government declared that no young person should
remain unemployed for more than 100 days. The government regarded longterm unemployment as devastating for future labor market prospects. Young
people are defined as long-term unemployed if they are unemployed for
more than 100 days. Thus, the ambition was that every individual of below
25 years of age should be offered a labor market program if they were
unable to find a job within three months of registration at the local PES. The
central government encouraged the municipalities to take part in actions to
reduce long-term unemployment among young people in different ways. For
instance, in 1995 the government introduced the Municipal Youth Program
(Kommunala ungdomsprogrammet) and three years later the UVG-guarantee
(Ungdomsgarantin). The PES offices are expected to collaborate with local
governments in both programs (Carling and Larsson 2005).
Recall that task complexity is defined in terms of the scope and intensity
of a policy. The Activity Guarantee is more intense and has greater scope.
Thus, it is a good representative of a complex policy. First, consider the
client groups. In the Activity Guarantee, the clients are comparatively old
and they are obviously not particularly attractive on the labor market since
they have been looking for a job for such a long time. It is also reasonable to
assume that a relatively large part of the clients have multiple problems since
the program is directed at those with the weakest position on the labor
market. In comparison, youth clients have not been unemployed for such a
long time. Consequently, they are probably more motivated, more homogenous, and easier to handle. Accordingly, the treatment of clients in the
Activity Guarantee needs more areas of expertise and more intense activities.
Second, although various youth activities are probably not so simple to
implement, the Activity Guarantee is more complex. The activities are
supposed to be rigorous and carried out in small groups. Contacts between
clients and personal supervisor are also expected to be more frequent than is
generally the case. In short, these activities are more intense. Moreover, the
government’s intention was that the program should be flexible and contain
various activities suited for the individual client. This means that the
Activity Guarantee is intended to have greater scope than other ALMPs. On
balance, it is safe to conclude that the Activity Guarantee is much more complex that the youth policy, even though some youth activities can also be
said to contain complexities. Thus, I predict a greater effect of cooperation
on implementation output in the Activity Guarantee.
The selected research setting is suitable for several reasons. First,
unemployment is a large societal problem and a lot of resources are allocated
114
When does cooperation improve public policy implementation?
to diminishing the unemployment rate all around world. Studies that can
bring some clarity to what is going on when ALMPs are put into practice are
therefore important. Second, this study is focused on the relationships
between the same types of actors who implement the same policies around
Sweden. This implies that features of the agencies and the policies that do
not vary can be held constant. From a methodological point of view this is
very beneficial. Third, Swedish ALMPs can be seen as a “critical case”,
meaning that here, if anywhere, we ought to expect that cooperation will
improve implementation on a broad range of activities. O’Toole (1983)
claims that there are very high information and expertise requirements in
ALMPs. Interorganizational cooperation is therefore likely to be a good
strategy for gathering the necessary resources. Furthermore, there are
obvious connections between the responsibilities of the PES offices and the
municipalities. This means that we have good reasons for believing that
cooperation is a good strategy when it comes to both policies. If cooperation
does not enhance the implementation of the youth policy and not even the
Activity Guarantee we should probably not anticipate general positive
effects of interorganizational cooperation either.
Method and measures
Whereas quantitative approaches have become increasingly common and
more sophisticated in the implementation literature from the United States,
studies based on large-N data in the European context are rare. The lack of
quantitative studies implies at least two problems. First, it is difficult to
generalize the findings when studying only one or a couple of cases. Second,
it is extremely difficult to control for a large number of potential
explanations in case studies; almost inevitably we end up in a situation in
which several variables account for the variation in the dependent variable
equally well. Goggin (1986) pointed this out several years ago and Winter
(2003b) accentuated the importance of statistical approaches to implementation problems in an overview article in Handbook of Public Administration.
Qualitative studies are still important, but right now quantitative research
based on European data is probably more needed.6
To test whether cooperation is related to implementation performance,
cross-section data covering 2003 were collected. The PES offices are the
unit of analysis. Questionnaires were distributed to the chief managers of all
PES offices in February 2004: 268 managers answered the questionnaire,
which implies a response rate of 75 percent. An analysis of the nonresponses showed no noticeable difference between respondents and non6
Daley and Layton (2004), Brehm and Gates (1997), Keiser and Soss (1998), and Meier and
McFarlane (1996) are examples of studies using large-N data from the United States.
115
Essay III
respondents on background characteristics. Register data from the National
Labor Market Administration supplement the survey. Lastly, official statistics, in the form of municipal characteristics, are taken from the KFAKTA03
database. Variables are discussed below (see also Appendix).
Cooperation
In the questionnaire, the PES managers were asked whether the PES and the
municipality had set up regular cooperative groups in which (a) caseworkers
from the two authorities work together and (b) the managers collaborate. The
managers also indicated whether (c) caseworkers contact each other on a
daily basis or more seldom. In addition, they provided information about
whether the agencies had formal collaborative contracts concerning (d) the
youth policy and (e) the Activity Guarantee.
Table 1. Cooperation between PES Offices and Municipalities, Various Indicators
Mean/
proportion
Standard
deviation
Minimum
Maximum
Cooperation, five items
3.50
1.36
0.00
5.00
Cooperation, three items
1.96
0.96
0.00
3.00
Youth contract
0.77
–
–
–
AG contract
0.80
–
–
–
A factor analysis showed that all items load high on a single dimension (see
Lundin 2007 or Essay II of this thesis). Thus, an index was constructed. It is
a simple additive index: a value of zero implies that none of the above
mentioned ways of cooperating was used, whereas agencies that use all five
activities got a score of five. To make interpretation easier, the factor scores
obtained from the analysis are not utilized in the article. But the analyses
have been carried out based on the factor scores as well, and the conclusions
are unchanged.
One problem with the index is that two of the items are directly associated
with the youth policy and the Activity Guarantee, respectively. The other
three items are more general. I have therefore decided to test the robustness
of the results. I use the complete index with all five indicators, but also an
alternative index ranging from zero to three in which the collaborative
contracts in (d) and (e) are excluded. In addition, the contracts are employed
as indicators of cooperation. Summary statistics are reported in Table 1.
116
When does cooperation improve public policy implementation?
Implementation performance
Three measures of implementation performance are employed in the article.
I consider enrollment in programs among youth (Y-Enrollment) and
enrollment in the Activity Guarantee (AG-Enrollment). I also study whether
the PES arranges full-time activities within the Activity Guarantee (AG
Activity). Obviously, there are many other aspects of the performance that
ideally should be taken into account. The quality of the programs is not
directly assessed in this way and it is possible that activities to some extent
are meaningless. But the performance measures have certain appealing
features. The variables concern vital parts of the policies, the elected
representatives’ directives are also relatively clear, and quantitative data are
available and reliable. These three factors make them suitable dependent
variables and they are definitely important things to consider if we want to
know how government ambitions have been realized.
Two variables address the implementation of the Activity Guarantee.
First, unemployed persons are supposed to be enrolled in the Activity
Guarantee after 27 months of unbroken PES registration at the latest. I
calculate how large a proportion of the individuals with 835 days (27.5
months) or more of PES registration was activated in the Activity Guarantee
on four occasions (15 February, May, August, and November). The average
value indicates how well the agency manages to enroll clients. The reason
for not using precisely 27 months is to avoid problems that a delay in registration might entail. This variable is named AG-enrollment and a higher
score implies better implementation. Data come from the National Labor
Market Administration.
Second, the activities should be full-time for all participants in the
Activity Guarantee. In the questionnaire, the managers were asked whether
the PES arranges full-time activities for all (or almost all) of the participants
in the Activity Guarantee. The information from the questionnaire is used as
a second measurement of the implementation of the program. This variable
can take on two values (0 = no, 1 = yes) and a score of one thus implies
better implementation. The variable is referred to as AG Activity.
The central government ambition for youth unemployment is clear-cut: if
individuals under the age of 25 have not found a job after 100 days of unemployment the PES should activate them by applying labor market
programs. Using data from the National Labor Market Administration administrative system, I investigated the stock of persons of 20–25 years of age
on four occasions (15 February, May, August, and November) in 2003 for
each PES office. I counted the number of individuals unemployed for 110
days or more in sequence, without participating in a program, directly before
the point of measurement. By setting the point of calculation to 110 days
instead of 100, I made sure that the results were not affected by a possible
delay in registration. To standardize the measure, I divided the values by the
117
Essay III
number of youth clients registered at the PES for 110 days on the same
occasions. In order to make interpretation easier, the computed values were
subtracted from one. The average of the four points of measurement was
calculated. This gives the variable Y-enrollment. A higher value means better
output.
Table 2. Implementation of the Youth Policy and the Activity Guarantee
Mean/
proportion
Standard
deviation
Minimum
Maximum
Y-Enrollment
0.66
0.14
0.10
0.95
AG-Enrollment
0.49
0.16
0.00
0.97
AG Activity
0.71
–
–
–
Table 2 describes how the offices have succeeded in implementing the youth
policy and the Activity Guarantee. There are youth who do not participate in
programs but who should, according to the government’s intentions. The
Activity Guarantee has not been fully implemented either since around half
of the target group is not engaged in activities and only seven in ten
managers say that they arrange full-time activities. Note that there is a lot of
variation in implementation performance.
Why should we expect that cooperating with the municipality makes it
easier for the PES to improve the implementation aspects mentioned above?
There are many possible reasons and I shall briefly mention two: If the
municipality can contribute with money, staff, and premises it becomes
easier to enroll clients and make sure that the activities are conducted on a
full-time basis. If the municipality helps the PES to find work places where
participants can participate in work practice, the PES can devote more time
to making sure that the target group is reached and that programs are conducted full-time.
Control variables
Unfortunately, the implementation research discourse has not been able to
develop a generally accepted theory that pinpoints the precise variables to include when explaining implementation. As O’Toole (2004, 310) puts it,
“theories about policy implementation have been almost embarrassingly
plentiful, yet theoretical consensus is not on the horizon … After hundreds
of empirical studies, validated findings are relatively scarce.” But there are
some broad categories of factors that merit attention. External and internal
characteristics of an agency carrying out public policies may influence implementation (Gill and Meier 2001; Keiser and Soss 1998; Winter 2003a).
118
When does cooperation improve public policy implementation?
External characteristics that should be taken into account are the client
group treated and the implementation environment (Winter 2003a). A demanding clientele and a difficult labor market situation might affect implementation positively or negatively. On the one hand, the task becomes
more challenging, which may make implementation more difficult. On the
other hand, the incentives to arrange appropriate programs are probably
stronger since it is less likely that clients can manage on their own. There
should also be greater demand for services when the labor market situation is
problematical.
Three variables that describe clients’ characteristics are used as control
variables. Some offices are responsible for vocational rehabilitation of unemployed individuals. These agencies’ clientele is quite different from the
client group of the standard PES, and therefore a dummy variable for rehabilitation PES is incorporated in the analyses. The other two controls are
the proportion of long-term unemployed clients and the proportion of clients
who are not Nordic citizens (Non-Nordic clients). By including these
variables, I am able to hold important clientele characteristics constant.
The characteristics of the PES offices’ local context are measured by the
following variables. The municipal unemployment rate, including participants in active labor market measures, provides an assessment of the local
labor market situation. The size of the local population is added as a control
in order to address other socio-economic factors; the local labor market is
very different in a large, urban area in comparison with a thinly populated
municipality with a small population. A dummy variable (socialist
government) indicating the presence of a Social Democratic government or a
non-socialist government captures the local political context. This political
variable’s importance for implementation has been stressed in earlier studies
(Keiser and Soss 1998). Leftist governments are usually more positive towards governmental interventions such as labor market programs (Korpi
2006). Operating in a context in which socialist values are strong can make
the task of implementing policy ambitions easier.
Performance could also vary due to factors internal to an agency. Keeping
external circumstances constant, agencies which have employees who are
more willing to carry out a policy and who have a high capacity to
implement decisions will, on average, perform better (Sannerstedt 2001).
There is solid evidence that the incentives, preferences, and attitudes of
bureaucrats often affect implementation (Brehm and Gates 1997). Implementation is likely to be improved if an agency gives a policy high
priority. The questionnaire supplies data on the PES managers’ attitudes towards different labor market goals. Among thirteen labor market objectives,
the managers rated the importance of arranging programs for unemployed
young people and for unemployed individuals with severe problems on the
labor market. From this rating, I have constructed a ranking of the objectives
of each PES office. The rating is composed so that the goal that receives the
119
Essay III
highest priority gets a score of thirteen; the other goals are arranged in descending order so that the least important objective gets a score of one. Thus,
the variables priority of youth clients and priority of difficult clients assess
how important the two relevant client groups are to the PES offices.
It is difficult to obtain good indicators of an agency’s capacity. But staff
and financial resources are the main factors influencing capacity. If the
agencies do not have these resources they will certainly run into difficulties
when policies are implemented (Gill and Meier 2001; Sannerstedt 2001).
The National Labor Market Board allocates resources so that each agency
receives an amount reflecting the local labor market situation and the client
group (Nyberg and Skedinger 1998). Thus, if resources are allocated
adequately, capacity should not vary due to staff and financial resources.
Nevertheless, in the event that resources are distributed poorly, the number
of clients per staff member and the amount of financial resources per client
reserved for benefits for participants in programs are used as control
variables.
Several organizational factors could shape the agency’s capacity (Winter
2003a). Identifying the important organizational characteristics is difficult,
and the capacity may impinge on various variables we cannot observe. In
this article, organizational size is employed as a control variable. This is
primarily motivated by the fact that large agencies have greater opportunities
to cooperate with other organizations. For instance, a large staff increases the
chance that an agency will communicate with the municipality on a daily basis. Another reason for including size is that this variable could be correlated
with many organizational aspects that may affect implementation. The
logarithm of the number of employees at the PES is utilized as a measure of
organizational size. I use the logarithm since it is reasonable to assume that a
one-unit change in the number of employees is more important when staff
size is small than when it is large.
Findings
In the following section, I examine whether the degree of cooperation can
account for the differences in implementation performance depicted in
Table 2. The empirical results are reported in two sub-sections.
Effects of cooperation: The Activity Guarantee
Estimates from ordinary least squares (OLS) regression models are presented
in Table 3. The dependent variable is AG-Enrollment. The number of observations is between 201 and 212, although the number of questionnaire
respondents was 268. Some agencies do not handle the Activity Guarantee
120
When does cooperation improve public policy implementation?
and there are also internal missing values for some variables. Thus, the number of valid cases is reduced.
Table 3. OLS Regression Analysis with AG-Enrollment as Dependent Variable
(Standard Errors in Parentheses)
1
Cooperation, five items
2
0.030 ***
(0.010)
Cooperation, three items
0.027 **
(0.012)
AG contract
Rehabilitation
Long-term unemployed clients
Non-Nordic clients
Unemployment
Local population
Socialist government
Priority of difficult clients
Clients per staff member
Financial resources
Organizational size
Constant
Adjusted R²
Standard error of regression
Number of observations
3
–0.091
(0.083)
0.481 *
(0.256)
–0.432 **
(0.183)
–0.021 ***
(0.005)
0.000
(0.000)
–0.007
(0.023)
0.004
(0.004)
–0.012 *
(0.007)
0.046
(0.035)
–0.049 **
(0.017)
0.593 ***
(0.083)
0.180
0.139
201
–0.092
(0.074)
0.482 *
(0.247)
–0.348 *
(0.179)
–0.020 ***
(0.005)
0.000
(0.000)
–0.003
(0.022)
0.003
(0.004)
–0.013 *
(0.007)
0.040
(0.034)
–0.043 **
(0.017)
0.635 ***
(0.081)
0.152
0.139
212
0.090 ***
(0.026)
–0.081
(0.073)
0.541 **
(0.245)
–0.375 **
(0.176)
–0.021 ***
(0.005)
0.000
(0.000)
–0.008
(0.022)
0.004
(0.004)
–0.010
(0.008)
0.054
(0.034)
–0.032 **
(0.016)
0.553 ***
(0.083)
0.187
0.137
208
* p < .10; ** p < .05; *** p < .01.
In Model 1, the index that includes all five indicators of collaboration is used
as the main independent variable. The results indicate a positive effect of
cooperation on implementation. When controlling for internal and external
characteristics, the share of clients engaged in the Activity Guarantee increases by roughly 3 percentage points, on average, if cooperation increases
by one step. The effect is significant at the .01 level. The impact is substantial but not huge. An increase in cooperation by one step is quite a large
121
Essay III
increase (almost a standard deviation) and this yields an increase of clients
enrolled in the program by a moderate 3 percentage points. But it is not
reasonable to expect that a lot of cooperation between the PES and the
municipality should imply a dramatic difference. And cooperation seems not
unimportant: going from the sample minimum to the maximum increases
enrollment by 15 percentage points, on average.
If the cooperative contracts are excluded from the collaboration index (see
Model 2) the effect becomes a little bit lower. But it is still of importance
(0.027) and it is significant at the .05 level. Model 3 indicates that agencies
having a formal cooperative contract, on average, manage to engage about 9
percentage points more of the target group, all else being equal. The effect is
statistically significant at the .01 level. The contract seems to account for a
noticeable part of the overall effect, which is not surprising since the contract
is linked directly to the Activity Guarantee. To sum up, cooperation has a
positive effect on implementation no matter how the variable is operationalized.
Control variables are only commented upon briefly. Client group
characteristics and the labor market situation affect implementation. Moreover, large organizations perform worse, whereas agencies having few
clients per staff member seem to achieve objectives to a greater extent.
To examine the robustness of the results, several diagnostics tests have
been carried out. I find no problem of heteroscedasticity or multicollinearity.
There are some influential outliers but the results are not substantially altered
if outliers are excluded from the regression model. Diagnostic plots did not
indicate any non-linear relationships. There might be some problem of endogeneity regarding two control variables: the share of long-term unemployed
clients and the unemployment rate. I have estimated models excluding these
variables. Conclusions remain the same. In addition, numerous model
specifications including various control variables have been tested. The
results are quite robust, although the two cooperation indices turn out to be
statistically insignificant in a small number of specifications. The effect of
the contract is very robust.
The level of activation within the Activity Guarantee is a dichotomous
variable. The estimation method when using AG Activity as the dependent
variable is therefore binary logit. Results are reported in Table 4. The logit
model yields information about whether the relationship is positive or
negative and if it is statistically significant but the coefficients are not readily
interpretable. Thus, predicted probabilities, based on logit coefficients, are
reported in Figure 1.
122
When does cooperation improve public policy implementation?
Table 4. Binary Logit Regression Analysis with AG Activity as Dependent Variable
(Standard Errors in Parentheses)
1
Cooperation, five items
2
0.345 **
(0.175)
Cooperation, three items
0.241
(0.223)
AG contract
Rehabilitation
Long-term unemployed
clients
Non-Nordic clients
Unemployment
Local population
Socialist government
Priority of difficult clients
Clients per staff member
Financial resources
Organizational size
Constant
Pseudo R² (McFaddens R²)
3
–0.205
(1.536)
–3.468
(4.591)
3.170
(4.250)
0.450 ***
(0.137)
–0.003 *
(0.002)
–0.276
(0.413)
0.146 **
(0.073)
–0.097
(0.122)
0.081
(0.612)
–0.664 **
(0.330)
–1.514
(1.433)
0.185
Log likelihood
–98.037
Number of observations
197
–0.998
(1.355)
–2.068
(4.422)
3.610
(4.176)
0.447 ***
(0.134)
–0.004 **
(0.002
–0.225
(0.403)
0.151 **
(0.070)
–0.094
(0.119)
–0.021
(0.583)
–0.556 *
(0.326)
–1.152
(1.389)
0.187
–102.276
208
1.539 ***
(0.470)
–1.289
(1.388)
–2.210
(4.594)
3.701
(4.405)
0.464 ***
(0.139)
–0.004 **
(0.002)
–0.229
(0.418)
0.191 **
(0.075)
–0.085
(0.129)
0.167
(0.620)
–0.472
(0.313)
–2.821 *
(1.540)
0.217
–96.456
203
* p < .10; ** p < .05; *** p < .01.
Table 4 indicates that cooperation increases the probability of arranging fulltime activities within the Activity Guarantee. But when the cooperative
contract between the PES and the municipality is excluded from the index
(see Model 2), the effect is statistically insignificant. The findings are very
robust to model specification.
123
.6
.4
0
.2
Pr(Full time activity)
.8
1
Essay III
0
1
2
3
4
5
Cooperation
Figure 1. Predicted Probabilities of Organizing Full-Time Activities within the
Activity Guarantee as the Level of Cooperation Changes (Continuous Control
Variables Held at Mean Values and Discrete Control Variables Held at Mode
Values)
Figure 1 displays the probability of arranging full-time activities as the level
of cooperation changes. In this graph, logit estimates from Model 1 have
been used. The probability varies between approximately .40 and .80
depending on how much the agencies collaborate. This is a large difference.
The effect is, of course, even stronger if the contract is employed as the indicator of cooperation, but weaker and insignificant at all conventional levels
if the contract is excluded from the model specification.
A few notes on control variables are appropriate. The probability of
arranging full-time activities within the Activity Guarantee increases when
the PES office prioritize clients with an especially difficult situation on the
labor market and when unemployment is high. Large organizations and
agencies located in large cities have a lower probability of arranging fulltime activities. Note that it seems that the control variables do not affect the
variables AG Activity and AG-Enrollment in the same manner.
The findings presented above clearly indicate that cooperation between
PES offices and municipal labor market administrations improves the PES
offices’ implementation of the Activity Guarantee. There are, however, two
methodological challenges to the results. First, there might be some important control variables that are left out from the model specification. The
capacity of the PES offices is probably the most difficult thing to assess in a
quantitative study. I have tried to control for as many variables as possible
124
When does cooperation improve public policy implementation?
and the problems are probably not that big. Although I recommend some
caution, it seems we can be quite confident in the conclusions.
Second, the time order between cooperation and implementation might be
reversed. For example, if a PES office arranges full-time activities for clients
in the Activity Guarantee, the need for external assistance may increase
and/or municipalities may be more interested in collaborating with the PES.
In turn, this will lead to more cooperation between the authorities. This is
certainly a possibility, although the assumed time order is probably more
likely. But this problem should not be ignored. One way to test causal order
is to check the timing. Cross-section data are limited in providing evidence
of time precedence, whereas panel data have advantages (Finkel 1995).
Some panel data are available for a smaller number of cases (137 agencies).
I have information about whether the PES arranged full-time activities for
the clients in the Activity Guarantee in 2001 and 2003. I also know if a
collaboration contract existed in 2001 and 2003.7 These data can be utilized
to estimate a two-wave cross-lagged effect model, which is a test of the time
order between variables. The basic idea is to predict each variable (AGEnrollment and AG Contract) in 2003 by its previous value in 2001 (the
lagged dependent variable), as well as the value of the other variable in
2001.8
The cross-lagged model cannot be regarded as a definite solution to the
time order problem, although it is a useable test. Data exist only for a
subsample of agencies and only some of the important dependent and independent variables are available. I also have to assume that the causal lags
are about two years. That is, signing a contract in 2001 affects implementation approximately in 2003. It is reasonable to assume that there is some
time lag, although two years is perhaps on the high side. To save space, a
table or a figure reporting the results is not presented in the article. But the
test indicates that having a collaborative contract in 2001 significantly (at the
.05 level) increases the propensity to organize full-time activities in 2003.
For instance, having a cooperation contract in 2001 increases the probability
of arranging full-time activities in 2003 by .24, given that full-time activities
were not arranged in 2001. On the other hand, the effect of arranging fulltime activities in 2001 on the propensity to have a collaborative contract in
2003 is not statistically significant (although there is a small positive
coefficient). This speaks clearly in favor of the time order assumed in the
analyses. That is, cooperation precedes implementation.
7
Data for 2001 come from a research project at the Institute for Labour Market Policy
Evaluation (IFAU); see Forslund, Fröberg, and Lindqvist (2004).
8
The cross-lagged model is compatible with the “Granger test” for causality employed in
time series analysis. Finkel (1995) describes the cross-lagged model and the underlying
assumptions.
125
Essay III
Effects of cooperation: The youth policy
In 2003, the Activity Guarantee was a more complex task than the youth
policy. Cooperation is therefore expected to be a less successful strategy
when it comes to youth. The findings are reported in Table 5. The table
shows estimates from OLS regression models. Robust standard errors are
reported, due to heteroscedasticity.
The regression coefficients of the cooperation indicators are very close to
zero regardless of how cooperation is operationalized in Models 1–3. The
effect is also statistically insignificant. The only reasonable conclusion is
thus that cooperation does not affect implementation. Instead, traditional factors highlighted in the literature appear to account for the variation in performance. More clients per staff member and a larger share of clients with an
especially difficult situation on the labor market make implementation
worse. An increase of financial resources and a challenging local labor
market situation result in a larger share of young unemployed clients enrolled in labor market programs.
To test the robustness of the findings, numerous diagnostic tests have
been carried out. I have also specified the model in many ways using various
sets of control variables. The results are robust and there is no indication
whatsoever that collaboration improves the implementation of the youth
policy. I have no data to examine time order but it is reasonable to assume
that cooperation is causal prior to implementation since the analysis in the
previous section provided some evidence of this.
To summarize, cooperation improves the implementation of the Activity
Guarantee but not the youth policy. Is this really a consequence of task
complexity or does something else explain the findings? The research setting
implies that many potential explanations are taken into account: for example,
the policy area, the actors involved, and the local context. This means that
the explanation has to do with the policies. A hypothesis competing with
complexity is the agencies’ interests in the two policies. The municipalities
may, for some reason, pay less attention to the youth or the agencies’
objectives concerning youth may diverge to a large extent. But quantitative
evidence indicates that the target groups of both policies receive equally and
very high priority from both authorities (see Table 1 the Introduction to this
thesis). Thus, task complexity is a more promising explanation.
126
When does cooperation improve public policy implementation?
Table 5. OLS Regression Analysis with Y-Enrollment as Dependent Variable
(Robust Standard Errors in Parentheses)
1
Cooperation, five items
2
–0.004
(0.008)
Cooperation, three items
–0.004
(0.011)
Youth contract
Rehabilitation
Long-term unemployed clients
Non-Nordic clients
Unemployment
Local population
Socialist government
Priority of youth clients
Clients per staff member
Financial resources
Organizational size
Constant
R²
3
–0.037
(0.081)
–0.599 **
(0.235)
–0.516 ***
(0.170)
0.019 ***
(0.004)
–0.000
(0.000)
0.014
(0.016)
0.004
(0.006)
–0.023 ***
(0.006)
0.044 *
(0.026)
–0.018
(0.013)
0.758 ***
(0.092)
0.397
Standard error of regression
0.110
Number of observations
211
–0.124
(0.097)
–0.554 **
(0.229)
–0.560 ***
(0.175)
0.018 ***
(0.004)
–0.000
(0.000)
0.010
(0.015)
0.003
(0.005)
–0.023 ***
(0.006)
0.048 *
(0.026)
–0.018
(0.013)
0.766 **
(0.090)
0.405
0.110
222
–0.004
(0.018)
–0.033
(0.080)
–0.598 ***
(0.222)
–0.529 ***
(0.171)
0.019 ***
(0.004)
–0.000
(0.000)
0.012
(0.016)
0.004
(0.006)
–0.023 ***
(0.006)
0.046 *
(0.026)
–0.022 *
(0.012)
0.762 **
(0.092)
0.407
0.109
216
* p < .10; ** p < .05; *** p < .01.
127
Essay III
Conclusion
Interorganizational cooperation is an important factor to consider in policy
implementation. Contacts between various public authorities are inescapable
in contemporary democratic states. Practitioners and scholars are therefore
interested in understanding how interorganizational relationships impinge on
the delivery of political decisions. But when does cooperation actually
improve implementation? Research thus far has not provided a satisfying
answer to this question. Part of the answer is offered in this study: the impact
of cooperation increases with task complexity.
Collaborative efforts of Public Employment Service (PES) offices and
municipal labor market administrations in Sweden have been examined. The
findings indicate that cooperation improves implementation of measures
directed towards persons who have been unemployed for a very long time.
The implementation of measures directed towards youth is not affected. The
striking difference between these policies, it is argued, is that the program
for clients with an especially difficult situation on the labor market (the
Activity Guarantee) is much more complex. Three questions emerge as a
consequence of the analysis: How reliable are the findings? Could they be
generalized? And what are the implications?
The empirical results have been subjected to numerous checks and the
findings appear to be very robust. It seems that we can be quite confident of
the results, although more research is always valuable. A couple of minor
caveats should be reported. Data are cross-sectional, which implies some
restriction when it comes to causal statements. The analysis is also limited to
a couple of measures of implementation performance and to two policies.
It is always difficult to generalize the results from a study of a certain
policy area to other contexts. But the present research setting provides rather
good opportunities. I have argued that the case comes near to being a
“critical case”. Labor market policies require a lot of resources, such as
information, to be implemented effectively. The PES offices and the
municipalities hold important resources, there are clear connections between
the authorities, and they share the overall goal of reducing unemployment.
Thus, if we do not find positive effects of cooperation on the implementation
of both complex and less complex policies within this policy area, we are not
likely to find positive effects when the demand for resources is less obvious
and the ties between the actors are less clear.
The main conclusion of this study is that interorganizational cooperation
is a reasonable strategy in order to make the implementation of political
ideas better—but only under certain circumstances. This conclusion is
important for central discourses in political science, such as implementation
research and research on multi-level governance. There is a general positive
attitude towards cooperation. The findings suggest that it is appropriate to be
more careful. Decision-makers in the public sector should not always stress
128
When does cooperation improve public policy implementation?
the importance of cooperation, but rather reflect on the necessity to
collaborate as much as possible across organizational boundaries. Many
public concerns could probably be handled quite well without extensive
interorganizational cooperation. This does not, of course, mean that interorganizational cooperation should always be completely avoided.
The idea of solving public problems by means of partnerships of actors is
popular both in “real life politics” and among academic scholars. For
example, several researchers taking the “governance” discourse as a point of
departure praise the partnership model.9 This model, based on the idea that
several local actors should collaborate in order to improve the delivery of
public policy, has had a pervasive breakthrough in recent years. The findings
of this study suggest that we should not be too eager to praise the partnership
model. Cooperation might, of course, be valuable for other reasons than
those investigated here. But for the sake of clarity, it is important to develop
and test theories about when interorganizational cooperation improves public
policy implementation or other important aspects of society. This is better
than stipulating that collaboration is a “good thing” in general.
9
See, for example, Pierre (2000) for an overview of the governance literature.
129
Essay III
References
Alter, Catherine, and Jerald Hage. 1993. Organzations working together.
Newbury Park, CA: Sage.
Alexander, Ernest R. 1995. How organizations act together. Interorganizational coordination in theory and practice. Luxembourg: Gordon
and Breach Publishers.
Bardach, Eugene. 1998. Getting agencies to work together. The practice and
theory of managerial craftsmanship. Washington D.C.: Brooking Institution Press.
Brehm, John, and Scott Gates. 1997. Working, shirking, and sabotage.
Bureaucratic response to a democratic public. Ann Arbor: Univ. Michigan
Press.
Bäck, Henry. 2003. Party politics and the common good in Swedish local
government. Scandinavian Political Studies 26(2):93–123.
Carling, Kenneth, and Laura Larsson. 2005. Does early intervention help the
unemployed youth? Labour Economics 12:301–19.
Daley, Dorothy M., and David F. Layton. 2004. Policy implementation and
the environmental protection agency: What factors influence remediation at
superfund sites? Policy Studies Journal 32(3):375–92.
deLeon, Peter. 1999. The missing link revisited: Contemporary implementation research. Policy Studies Review 16(3/4):311–38.
Finkel, Steven E. 1995. Causal analysis with panel data. Thousand Oaks:
Sage.
Forslund, Anders, Daniela Fröberg, and Linus Lindqvist. 2004. The Swedish
activity guarantee. OECD social, employment and migration Working
Paper 16, OECD, Paris, France.
Gill, Jeff, and Kenneth J. Meier. 2001. Ralph’s pretty-good grocery versus
Ralph’s super market: Separating excellent agencies from the good ones.
Public Administration Review 61(1):9–17.
Goggin, Malcolm L. 1986. The “too few cases/too many variables” problem
in implementation research. Western Political Quarterly 39(2):328–47.
Goggin, Malcolm L., Ann O’M. Bowman, James P. Lester, and Laurence J.
O’Toole, Jr. 1990. Implementation theory and practice. Toward a third
generation. New York: HarperCollins.
Gustafsson, Agne. 1999. Kommunal självstyrelse. Stockholm, Sweden: SNS.
130
When does cooperation improve public policy implementation?
Hjern, Benny, and David O. Porter. 1981. Implementation structures: a new
unit of administrative analysis. Organization Studies 2(3):211–27.
Hooghe, Liesbet, and Gary Marks. 2003. Unraveling the central state, but
how? Types of multi-level governance. American Political Science Review
97(2):233–43.
Hudson, Bob, Brian Hardy, Melanie Henwood, and Gerald Wistow. 1999. In
pursuit of inter-agency collaboration in the public sector. Public
Management 1(2):235–60.
Jennings, Edward T., Jr. 1994. Building bridges in the intergovernmental
arena: Coordinating employment and training programs in the American
states. Public Administration Review 54(1):52–60.
Jennings, Edward T., Jr., and Jo Ann G. Ewalt. 1998. Interorganizational
coordination, administrative consolidation, and policy performance. Public
Administration Review 58(5):417–28.
Jones, Norah, Paul Thomas, and Les Rudd. 2004. Collaborating for mental
health services in Wales: A process evaluation. Public Administration
82(1):109–21.
Keiser, Lael R., and Joe Soss. 1998. With good cause: Bureaucratic
discretion and the politics of child support enforcement. American Journal
of Political Science 42(4):1133–56.
Korpi, Walter. 2006. Power resources and employer-centered approaches in
explanations of welfare states and varieties of capitalism. Protagonists,
consenters, and antagonists. World Politics 58:167–206.
Lundin, Martin. 2007. Explaining cooperation: How resource
interdependence, goal congruence, and trust affect joint actions in policy
implementation. Journal of Public Administration Research and Theory;
doi:10.1093/jpart/mul025.
Lundin, Martin, and Per Skedinger. 2006. Decentralisation of active labour
market policy: The case of Swedish local employment service committees.
Journal of Public Economics 90(4-5):775–98.
Martin, John P., and David Grubb. 2001. What works and for whom: A
review of OECD countries’ experiences with active labour market policies.
Swedish Economic Policy Review 8(2):9–56.
Meier, Kenneth J., and Deborah R. McFarlane. 1996. Statutory coherence
and policy implementation: The case of family planning. Journal of Public
Policy 5(3):281–98.
131
Essay III
Norton, Alan. 1994. International handbook of local and regional
government. A comparative analysis of advanced democracies. Aldershot:
Edward Elgar.
Nyberg, Sten, and Per Skedinger. 1998. Arbetsförmedlingarna – mål och
drivkrafter. Ds 1998:16. Stockholm: Ministry of Finance.
O’Toole, Laurence J. Jr. 1983. Interorganizational co-operation and the
implementation of labour market training policies: Sweden and the Federal
Republic of Germany. Organization Studies 4(2):129–50.
———. 2003. Interorganizational relations in implementation. In Handbook
of public administration, ed. J. Pierre and B. G. Peters, 234–44. London:
Sage.
———. 2004. The theory-practice issue in policy implementation research.
Public Administration 82(29):309–29.
O’Toole, Laurence J., Jr., and Robert S. Montjoy. 1984. Interorganizational
policy implementation: A theoretical perspective. Public Administration
Review 44(6):491–503.
Peters, B. Guy, and Jon Pierre. 2001. Developments in intergovernmental
relations: toward multi-level governance. Policy & Politics 29(2):131–5.
Pierre, Jon (ed.). 2000. Debating governance. Oxford, UK: Oxford Univ.
Press.
Pressman, Jeffery L., and Aaron Wildavsky. 1984. Implementation. How
great expectations in Washington are dashed in Oakland; or, why it's
amazing that federal programs work at all, this being a saga of the
economic development administration as told by two sympathetic observers
who seek to build morals on a foundation of ruined hopes. 2nd ed. Los
Angeles: Univ. California Press.
Salonen, Tapio, and Rickard Ulmestig. 2004. Nedersta trappsteget. En
studie om kommunal aktivering. Växjö, Sweden: Växjö Univ.
Sannerstedt, Anders. 2001. Implementering – Hur politiska beslut genomförs
i praktiken. In Politik som organisation. Förvaltningspolitikens grundproblem, ed. B. Rothstein, 18–48. Stockholm, Sweden: SNS Förlag.
Schermerhorn, John R. Jr. 1975. Determinants of interorganizational
cooperation. Academy of Management Journal 18(4):846–56.
Smith, Ken G., Stephen J. Carroll, and Susan J. Ashford. 1995. Intra- and
interorganizational cooperation: Toward a research agenda. The Academy
of Management Journal 38(1):7-23.
132
When does cooperation improve public policy implementation?
Swedish Association of Local Authorities. 1999. Kommunerna i arbetsmarknadspolitiken: en enkätundersökning. Stockholm, Sweden: Swedish
Association of Local Authorities (Kommunförbundet).
Van de Ven, Andrew H. 1976. On the nature, formation, and maintenance of
relations among organizations. Academy of Management Review 1:24–36.
Van de Ven, Andrew H., and Gordon Walker. 1984. The dynamics of interorganizational coordination. Administrative Science Quarterly 29(4):598–
621.
Weiss, Janet A. 1987. Pathways to cooperation among public agencies.
Journal of policy analysis and management 7(1):94–117.
Winter, Soren. 2003a. Implementation: Introduction. In Handbook of public
administration, ed. J. Pierre and B. G. Peters, 205–11. London: Sage.
Winter, Soren. 2003b. Implementation perspectives: Status and
reconsideration. In Handbook of public administration, ed. J. Pierre and B.
G. Peters, 212–22. London: Sage.
White, Steven. 2005. Cooperation costs, governance choice and alliance
evolution. Journal of Management Studies 42(7):1383–1412.
133
Essay III
Appendix: Description of variables
AG Activity
The PES office arranges full-time activities for all (or almost all) clients
participating in the Activity Guarantee. Dummy variable: 0 = no (.29); 1 =
yes (.71). Based on questionnaire data.
AG-Enrollment
Proportion of persons registered for a sequence of 835 days or more at the
PES office enrolled in the Activity Guarantee; average of four different
points of measurement (15th of February, May, August and December 2003).
Based on register data from the National Labor Market Administration.
Mean: 0.49, sd: 0.16, min: 0.00, max: 0.97.
Y-Enrollment
Young persons aged 20 to 25 with at least 110 days of open unemployment
in sequence, divided by the total number of young people registered at the
PES for at least 110 days; average of four different points of measurement
(15th of February, May, August and December 2003). Based on register data
from the National Labor Market Administration. Mean: 0.66, sd: 0.14, min:
0.10, max: 0.95.
Cooperation, five items
Index of cooperation. Includes cooperative groups at manager level,
cooperative groups at case worker level, daily communication at case worker
level, cooperative contract concerning the youth policy, and cooperative
contract concerning the Activity Guarantee. Based on questionnaire data.
Mean: 3.50, sd: 1.36, min: 0.00, max: 5.00.
Cooperation, three items
Index of cooperation. Includes cooperative groups at manager level,
cooperative groups at caseworker level, and daily communication at
caseworker level. Based on questionnaire data. Mean: 1.96, sd: 0.96, min:
0.00, max: 3.00.
AG Contract
The PES office and the municipality have signed a cooperation concerning
the Activity Guarantee. Dummy variable: 0 = no (.20); 1 = yes (.80). Based
on questionnaire data.
Youth Contract
The PES office and the municipality have signed a cooperation contract
about youth programs. Dummy variable: 0 = no (.23); 1 = yes (.77). Based
on questionnaire data.
134
When does cooperation improve public policy implementation?
Rehabilitation
PES office responsible for vocational rehabilitation. Dummy variable: 0 = no
(.96); 1 = yes (.04). Data obtained from the Internet homepage of AMV,
www.ams.se.
Long-term unemployed clients
Share of clients at the PES office unemployed for six months or more in
2003. Based on register data from the National Labor Market
Administration. Mean: 0.11, sd: 0.05, min: 0.00, max: 0.30.
Non-Nordic clients
Proportion of clients at the PES office without Nordic citizenship in 2003.
Based on register data from the National Labor Market Administration.
Mean: 0.07, sd: 0.07, min: 0.01, max: 0.62.
Unemployment
Local unemployment rate in percent, including participants in measures, in
the municipality where the PES office is located (April 2003). Based on
official municipal statistics from KFAKTA03. Mean: 5.47, sd: 2.31, min:
1.90, max: 19.60.
Local population
Number of inhabitants in the municipality (in 1,000’s of persons) where the
PES office is located (December 2002). Based on official municipal statistics
from KFAKTA03. Mean: 95.56, sd: 192.14, min: 2.61, max: 758.15.
Socialist government
The chairman of the municipal executive board represents the Social
Democrats. Dummy variable: 0 = no (.33); 1 = yes (.67). Based on official
municipal statistics from KFAKTA03.
Priority of difficult clients
The PES manager’s rank of the objective “ensuring that there are labor
market programs for groups of unemployed with severe problems in the
labor market” among 13 objectives (scale 1-13, where 13 means highest
priority). Based on questionnaire data. Mean: 11.16, sd: 2.46, min: 2.00,
max: 13.00.
Priority of youth clients
The PES manager’s rank of the objective “ensuring that there are labor
market programs for young people under 25” among 13 objectives (scale 113, where 13 means highest priority). Based on questionnaire data. Mean:
12.21, sd: 1.85, min: 4.00, max: 13.00.
Clients per staff member
The average number of clients (measured in 10 clients) per week per
employee at the PES office. Based on register data from the National Labor
Market Administration. Mean: 4.96, sd: 1.74, min: 0.26, max: 12.44.
135
Essay III
Financial resources
The amount of financial resources at the PES office reserved for benefits to
clients participating in active measures (in 1,000 Swedish crowns) per week
divided by the number of clients per week. Based on questionnaire data.
Mean: 0.66, sd: 0.35, min: 0.00, max: 3.01.
Organizational size
The logarithm of the number of employees at the PES office. Based on
register data from the National Labor Market Administration. Mean: 2.63,
sd: 0.78, min: 0.00, max: 4.65.
136
Publication series published by the Institute for Labour
Market Policy Evaluation (IFAU) – latest issues
Rapporter/Reports
2007:1
Lundin Daniela ”Subventionerade anställningar för unga – en uppföljning av
allmänt anställningsstöd för 20–24-åringar”
2007:2
Lundin Daniela, Eva Mörk & Björn Öckert ”Maxtaxan inom barnomsorgen –
påverkar den hur mycket föräldrar arbetar?”
2007:3
Bergemann Annette & Gerard van den Berg ”Effekterna av aktiv arbetsmarknadspolitik för kvinnor i Europa – en översikt”
2007:4
Junestav Malin ”Socialförsäkringssystemet och arbetsmarknaden – politiska
idéer, sociala normer och institutionell förändring – en historik”
2007:5
Andersson Christian ”Lärartäthet, lärarkvalitet och arbetsmarknaden för
lärare”
2007:6
Larsson Laura & Caroline Runeson ”Effekten av sänkt sjukpenning för
arbetslösa”
2007:7
Stenberg Anders ”Hur påverkar gymnasialt komvux löneinkomster och vidare studier?
2007:8
”Forslund Anders & Kerstin Johansson ”Lediga jobb, arbetssökande och
anställningar – den svenska matchningsfunktionen”
2007:9
Kennerberg Louise ”Hur förändras kvinnors och mäns arbetssituation när de
får barn?”
2007:10 Nordin Martin ”Invandrares avkastning på utbildning i Sverige”
2007:11 Johansson Mats & Katarina Katz ”Underutnyttjad utbildning och lönegapet
mellan kvinnor och män”
2007:12 Gartell Marie, Ann-Christin Jans & Helena Persson ”Utbildningens betydelse
för flöden på arbetsmarknaden”
2007:13 Grönqvist Hans & Olof Åslund ”Familjestorlekens effekter på barns utbildning och arbetsliv”
2007:14 Lindqvist Linus ”Uppföljning av plusjobb”
2007:15 Sibbmark Kristina ”Avidentifierade jobbansökningar – erfarenheter från ett
försök i Göteborgs stad”
2007:16 Hesselius Patrik & Malin Persson ”Incitamentseffekter och Försäkringskassans kostnader av kollektivavtalade sjukförsäkringar”
2007:17 Eriksson Stefan & Jonas Lagerström ”Diskriminering i anställningsprocessen: resultat från en Internetbaserad sökkanal”
2007:18 Eriksson Robert, Oskar Nordström Skans, Anna Sjögren & Olof Åslund
”Ungdomars och invandrades inträde på arbetsmarknaden 1985–2003”
2007:19 Agerström Jens, Rickard Carlsson & Dan-Olof Rooth ”Etnicitet och övervikt:
implicita arbetsrelaterade fördomar i Sverige”
2007:20 Bennmarker Helge, Kenneth Carling & Anders Forslund ”Vem blir långtidsarbetslös?”
Working Papers
2007:1
de Luna Xavier & Per Johansson “Matching estimators for the effect of a
treatment on survival times”
2007:2
Lundin Daniela, Eva Mörk & Björn Öckert “Do reduced child care prices
make parents work more?”
2007:3
Bergemann Annette & Gerard van den Berg “Active labor market policy
effects for women in Europe – a survey”
2007:4
Andersson Christian “Teacher density and student achievement in Swedish
compulsory schools”
2007:5
Andersson Christian & Nina Waldenström “Teacher supply and the market
for teachers”
2007:6
Andersson Christian & Nina Waldenström “Teacher certification and student
achievement in Swedish compulsory schools”
2007:7
van den Berg Gerard, Maarten Lindeboom & Marta López ”Inequality in
individual mortality and economic conditions earlier in life”
2007:8
Larsson Laura & Caroline Runeson “Moral hazard among the sick and
unemployed: evidence from a Swedish social insurance reform”
2007:9
Stenberg Anders “Does adult education at upper secondary level influence
annual wage earnings?”
2007:10 van den Berg Gerard “An economic analysis of exclusion restrictions for
instrumental variable estimation”
2007:11 Forslund Anders & Kerstin Johansson “Random and stock-flow models of
labour market matching – Swedish evidence”
2007:12 Nordin Martin “Immigrants’ return to schooling in Sweden”
2007:13 Johansson Mats & Katarina Katz “Wage differences between women and
men in Sweden – the impact of skill mismatch”
2007:14 Gartell Marie, Ann-Christin Jans & Helena Persson “The importance of
education for the reallocation of labor: evidence from Swedish linked
employer-employee data 1986–2002”
2007:15 Åslund Olof & Hans Grönqvist “Family size and child outcomes: Is there
really no trade-off?”
2007:16 Hesselius Patrik & Malin Persson “Incentive and spill-over effects of supplementary sickness compensation”
2007:17 Engström Per & Patrik Hesselius “The information method – theory and
application”
2007:18 Engström Per, Patrik Hesselius & Malin Persson “Excess use of Temporary
Parental Benefit”
2007:19 Eriksson Stefan & Jonas Lagerström “Detecting discrimination in the hiring
process: evidence from an Internet-based search channel”
2007:20 Agerström Jens, Rickard Carlsson & Dan-Olof Rooth “Ethnicity and obesity: evidence of implicit work performance stereotypes in Sweden”
2007:21 Uusitalo Roope & Jouko Verho “The effect of unemployment benefits on reemployment rates: evidence from the Finnish UI-benefit reform”
Dissertation Series
2007:1
Lundin Martin “The conditions for multi-level governance: implementation,
politics, and cooperation in Swedish active labor market policy”
2006:1
Hägglund Pathric “Natural and classical experiments in Swedish labour
market policy”
2006:2
Savvidou Eleni “Technology, human capital and labor demand”
2006:3
Söderström Martin “Evaluating institutional changes in education and wage
policy”
2006:4
Lagerström Jonas “Discrimination, sickness absence, and labor market
policy”
2006:5
Johansson Kerstin “ Empirical essays on labor-force participation, matching,
and trade”