Foreign Ownership, Survival and Growth
Dynamics in Turkish Manufacturing
Erol Taymaz, Middle East Technical University, Ankara
Yesim Üçdoğruk, Dokuz Eylül University, Izmir
Kamil Yılmaz, Koç University, Istanbul
REF Workshop on Productivity ● 30 November 2012 ● Istanbul
Questions

Are foreign firms different at the time of
entry?



Do foreign firms survive longer?



Internal and external funds
Knowledge
More productive, larger, etc
Footlose
Do foreign firms make life difficult for
domestic firms?


Competition
Spillovers
Questions



What are the determinants of firm
(employment) growth?
Do foreign firms grow faster than domestic
ones?
How does the presence of foreign firms affect
growth prospects of domestic firms
Data

Annual Survey of Manufacturing Industries




All private establishments employing 10+ people
1983-2001 period
About 10,000 establishments per year
Annual Industry and Service Statistics Database



Data for establishments in the service sectors along
with the ones in the industry.
2003-2009 period
Simplified survey questionnaires; data series are no
longer comparable with 2001 and before.
Share of Foreign Firms in Manufacturing
35
Share of foreign firms (%)
30
25
20
15
10
5
0
1985
1990
1995
Employment
2000
Value added
2005
Number of firms
2010
Survival functions
1.0
Survival rate (%)
0.8
0.6
0.4
0.2
0.0
0
2
4
6
8
10
12
Age
Domestic
Foreign
Domestic LSE
14
16
Model
Cox proportional hazard model
hij(t) = hj(t)exp(Xijtβ)


Stratified by 2-digit industries
Estimated for


New domestic, new foreign and all new
firms
Log likelihood ratio test
Descriptive statistics
(mean values)
Label
Description
1983-2001 dataset
New firms
llab
Log size (employment)
lab*
Size (employment)
lgr
Growth rate of employment
exit
Exitor
entry
Entrant
fdi
Foreign firm
lage
Log age
entrelsize Relative entry size
fdiqs
Sectoral share of foreign firms
fdiqr
Regional share of foreign firms
N
Number of observations
n
Number of growth variables
* Geometric average
3.580
35.86
0.023
0.108
0.164
0.025
1.406
-0.330
0.113
0.177
114906
89374
New foreign
firms
4.799
121.36
0.059
0.042
0.129
1.000
1.500
0.425
0.202
0.219
2881
2424
2003-2009 dataset
New firms
3.480
32.45
0.014
0.359
0.267
0.023
0.900
-0.723
0.116
0.204
33749
14889
New foreign
firms
4.234
68.97
0.166
0.079
0.101
1.000
1.150
-0.078
0.209
0.259
762
512
Descriptive statistics
(mean values)
Label
Description
1983-2001 dataset
New firms New foreign
2003-2009 dataset
New firms
firms
New foreign
firms
fdisupp_q
Share of foreign suppliers
0.052
0.077
0.056
0.072
fdibuy_q
Share of foreign buyers
0.039
0.070
0.054
0.082
cfdiqs
Change in fdiqs
0.006
0.008
0.002
0.006
cfdiqr
Change in fdiqr
0.003
0.010
-0.008
0.012
lentrate
Sectoral entry rate
0.057
0.049
0.027
0.018
sectgr
Sectoral output growth rate
0.075
0.074
0.038
0.050
sectgrpr
Sectoral price growth rate
0.484
0.499
0.050
0.042
hhi
Herfindahl index
0.051
0.072
0.072
0.069
mes
Minimum efficient scale
3.477
3.730
3.253
3.502
N
Number of observations
114906
2881
33749
762
n
Number of growth variables
89374
2424
14889
512
Results – Survival (1983-2001)
Models
Variables
fdi
fdiqs
fdiqr
fdisupp_q
fdibuy_q
cfdiqs
cfdiqr
lentrate
sectgr
sectgrpr
1
2
3
4
-0.778***
[0.098]
0.250**
[0.112]
0.119
[0.081]
0.135
[0.473]
-1.036***
[0.371]
-0.778***
[0.098]
0.208*
[0.115]
0.153*
[0.091]
0.177
[0.474]
-1.006***
[0.371]
0.371
[0.258]
-0.189
[0.189]
-0.720***
[0.099]
0.128
[0.119]
0.110
[0.091]
-0.107
[0.474]
0.030
[0.383]
0.430*
[0.249]
-0.157
[0.189]
-0.033
[0.215]
-0.208***
[0.074]
-0.289**
[0.118]
0.241**
[0.102]
-0.107
[0.126]
0.302***
[0.084]
0.146
[0.507]
0.311
[0.405]
0.808***
[0.215]
-0.100
[0.079]
-0.276**
[0.125]
5
0.241**
[0.102]
-0.126
[0.128]
0.334***
[0.095]
0.161
[0.508]
0.321
[0.405]
0.192
[0.279]
-0.144
[0.185]
0.808***
[0.216]
-0.105
[0.079]
-0.277**
[0.125]
6
7
0.285***
[0.102]
-0.112
[0.128]
0.282***
[0.096]
0.108
[0.509]
0.273
[0.404]
0.154
[0.280]
-0.116
[0.187]
0.709***
[0.220]
-0.101
[0.079]
-0.283**
[0.126]
0.270**
[0.118]
0.001
[0.149]
0.165
[0.110]
0.373
[0.601]
-0.459
[0.478]
0.272
[0.316]
-0.461**
[0.222]
0.754***
[0.281]
-0.193*
[0.099]
-0.242
[0.158]
Results – Survival (1983-2001)
Variables
1
2
3
0.670***
[0.184]
-0.307***
[0.035]
87956
-68859
87956
-68857
87956
-68801
hhi
mes
relsize
lw
pmargin
subinput
suboutput
Models
4
0.514**
[0.202]
-0.516***
[0.039]
-0.678***
[0.019]
-0.281***
[0.022]
-0.041
[0.066]
0.095
[0.102]
0.283***
[0.046]
5
0.520**
[0.202]
-0.517***
[0.039]
-0.678***
[0.019]
-0.281***
[0.022]
-0.042
[0.066]
0.094
[0.102]
0.283***
[0.046]
6
0.497**
[0.203]
-0.467***
[0.039]
-0.657***
[0.019]
-0.252***
[0.022]
0.016
[0.066]
0.096
[0.101]
0.246***
[0.046]
-0.079***
[0.007]
84014
-59084
84014
-59084
84012
-59029
kl
ttrans
rddum
N
Log likelihood
7
0.149
[0.259]
-0.386***
[0.052]
-0.661***
[0.023]
-0.274***
[0.032]
-0.050
[0.078]
0.036
[0.110]
0.236***
[0.049]
-0.092***
[0.008]
0.487***
[0.179]
-0.148***
[0.045]
56921
-37086
Results - Survival (2003-2009)
Models
Variables
fdi
fdiqs
fdiqr
fdisupp_q
fdibuy_q
cfdiqs
cfdiqr
lentrate
sectgr
sectgrpr
1
-1.138***
[0.152]
-0.056
[0.096]
-0.365***
[0.094]
21.059***
[6.991]
-7.576**
[3.536]
2
-1.140***
[0.152]
-0.073
[0.100]
-0.070
[0.148]
21.150***
[6.997]
-7.697**
[3.554]
0.029
[0.254]
-0.514***
[0.182]
3
-1.108***
[0.152]
0.113
[0.107]
0.032
[0.150]
15.090**
[7.054]
-6.156*
[3.591]
0.223
[0.262]
-0.534***
[0.184]
0.180
[0.241]
0.045
[0.060]
-0.134
[0.138]
4
-0.460***
[0.148]
0.131
[0.125]
0.042
[0.118]
25.283***
[9.040]
-5.674
[4.643]
1.255***
[0.321]
-0.094
[0.068]
-0.364**
[0.155]
5
-0.462***
[0.148]
0.078
[0.130]
0.013
[0.186]
26.106***
[9.060]
-6.508
[4.687]
0.546*
[0.320]
0.052
[0.230]
1.229***
[0.321]
-0.090
[0.068]
-0.378**
[0.155]
6
-0.395**
[0.157]
0.047
[0.180]
-0.001
[0.229]
42.750***
[13.199]
-7.849
[6.437]
0.731*
[0.415]
0.236
[0.305]
1.627***
[0.503]
-0.058
[0.100]
-0.446**
[0.213]
7
-0.351**
[0.155]
0.069
[0.182]
-0.046
[0.233]
43.049***
[13.179]
-8.141
[6.471]
0.773*
[0.420]
0.173
[0.309]
1.597***
[0.501]
-0.047
[0.098]
-0.489**
[0.211]
Results - Survival (2003-2009)
Models
Variables
1
2
3
-0.457***
[0.153]
-0.182***
[0.015]
4
-0.862***
[0.195]
-0.323***
[0.022]
-0.511***
[0.013]
0.012
[0.027]
-0.000***
[0.000]
0.100
[0.138]
0.392***
[0.037]
5
-0.874***
[0.195]
-0.327***
[0.022]
-0.512***
[0.013]
0.012
[0.027]
-0.000***
[0.000]
0.101
[0.138]
0.391***
[0.037]
6
-0.875***
[0.270]
-0.374***
[0.033]
-0.575***
[0.018]
0.066*
[0.037]
-0.000***
[0.000]
0.270
[0.177]
0.192***
[0.053]
-0.045***
[0.011]
18230
18223
18223
16305
16305
12829
7
-0.863***
[0.270]
-0.363***
[0.033]
-0.566***
[0.018]
0.015
[0.038]
-0.000***
[0.000]
0.135
[0.187]
0.238***
[0.054]
-0.059***
[0.011]
0.345***
[0.048]
-0.241
[0.263]
12829
-31656
-31640
-31590
-22654
-22653
-12796
-12777
hhi
mes
relsize
lw
pmargin
subinput
suboutput
kl
ttrans
rddum
N
Log likelihood
Results of the Cox proportional
hazards model




Foreign firms are more likely to survive than domestic
firms, but the difference between domestic and foreign
firms could be explained to a large extent by their firmspecific characteristics.
Once firm-specific characteristics are controlled for, it is
ambiguous if foreign firms can survive more or if they
are footloose.
It seems that foreign firms have advantages over
domestic firms not because they are foreign, but they
are multinational.
Foreign spillover variables have usually weak and
ambiguous impact on survival probabilities. This is partly
due to the limitation of spillover data.
Results of the Cox proportional
hazards model



Negative correlation between sectoral share of foreign firms
and survival probability,
but this correlation could be caused by other sector-specific
factors (level of concentration, sectoral growth rates, etc.).
Regarding horizontal spillovers,


there is a weak evidence that imply that change in the sectoral share of
foreign firms (the dynamic effect of the existence of foreign firms) has a
negative impact on survival of domestic firms.
Domestic firms feel the competitive pressures while foreign firms are
increasing their market share (through growth or entry), but the level of
foreign share itself does not matter much.
Average size of new firms by cohort
(entrants after 1992)
0.1
0.0
0
1
2
3
4
5
6
7
8
9
10
Relative size (%)
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
Age
-0.7
1992
1993
1994
1995
1996
1997
1998
1999
2000
Average size of new firms by cohort
(2003-2008)
0.0
0
1
2
3
4
5
6
-0.2
Relative size (%)
-0.4
-0.6
-0.8
-1.0
-1.2
-1.4
Age
2003
2004
2005
2006
2007
2008
7
Average size of new FDI firms by
cohort (entrants after 1992)
1.2
1.0
Relative size (%)
0.8
0.6
0.4
0.2
0.0
0
1
2
3
4
-0.2
5
6
7
8
9
10
Age
1992
1993
1994
1995
1996
1997
1998
1999
2000
Average size of new FDI firms by
cohort (2003-2009 dataset)
1.2
1.0
0.8
Relative size (%)
0.6
0.4
0.2
0.0
0
1
2
3
4
5
6
-0.2
-0.4
-0.6
-0.8
-1.0
Age
2003
2004
2005
2006
2007
2008
7
Results



As expected, entrants start small: the entry size is much
lower than the average size so that relative size at the time
of entry (age 0) is negative for all cohorts.
However, the relative size of new firms increase rapidly over
time, and converges towards sector average. The rapid
increase in the average size can be explained by two factors:
exit of small firms, and growth of new firms.
The first process (exit) has been studied in detail in the
preceding section, and it was found that small firms are
more likely to exit. If small firms exit, than the average size
of remaining firms will increase even if they do not grow at
all.
Results



The second process, new firm growth, will be studied in
detail in the following subsection.
The pattern of growth of new foreign firms: As it is obvious,
the entry size of foreign firms is much higher than the entry
size of domestic firms especially in the 1984-2001 period.
The average relative size of new foreign firms also tends to
increase over time, as observed in the case of domestic
firms.
Average size of exitors by survival
duration (1983-2001 dataset)
0.0
0
1
2
3
4
5
6
7
8
9
10
-0.1
Relative size (%)
-0.2
-0.3
-0.4
-0.5
-0.6
-0.7
-0.8
Age
1
2
3
4
5
6
7
8
9
10
Average size of exitors by survival
duration (2003-2009 dataset)
0.0
0
1
2
3
4
5
-0.2
Relative size (%)
-0.4
-0.6
-0.8
-1.0
-1.2
-1.4
-1.6
Age
1
2
3
4
5
6
6
Average size of survivors by age
(1983-2001 dataset)
0.2
0.1
Relative size (%)
0.0
0
1
2
3
4
5
6
7
8
9
10
-0.1
-0.2
-0.3
-0.4
Age
-0.5
2000
1999
1998
1997
1996
1995
1994
1993
1992
Average size of survivors by age
(2003-2009 dataset)
0.4
0.2
Relative size (%)
0.0
0
1
2
3
4
5
6
-0.2
-0.4
-0.6
-0.8
Age
2008
2007
2006
2005
2004
2003
7
Average size of foreign survivors by
age (1983-2001 dataset)
1.2
1.0
Relative size (%)
0.8
0.6
0.4
0.2
0.0
0
1
2
3
4
5
6
7
8
9
10
-0.2
-0.4
Age
2000
1999
1998
1997
1996
1995
1994
1993
1992
Average size of foreign survivors by
age (2003-2009 dataset)
1.5
Relative size (%)
1.0
0.5
0.0
0
1
2
3
4
5
6
-0.5
-1.0
Age
2008
2007
2006
2005
2004
2003
7
Results

Relative size of domestic exitors:



Exitors do not grow much after entry, the smaller the entry
size is, the shorter the duration of survival will be,
Exitors tend to become even smaller in a few years
preceding their exit.
The visual description of the exit process of domestic
firms provide evidence that a part of the increase in
the relative size of new firms can be explained by the
exit process because smaller firms tend to exit first.
Results



The relative size of domestic survivors: Surviving new
firms grow really quite fast and reach sector average
in about 5-6 years.
These figures provide visual evidence on the
differences between growth patterns of new
domestic and foreign firms.
We will use regression analysis to test if there is any
statistically significant difference between growth
rates of domestic and foreign firms, and to check if
spillovers from foreign firms have any affect on the
growth rates of domestic firms.
Employment Growth
gi,t1  log Lt1  log Lt   0 log Lt  Xit 
log Lt1   0 log Lt  Xit 



Unobserved firm-specific effects and the endogeneity of
the lagged dependent variable (log(Lt)) in the model
 GMM-system method
Main limitation: attrition bias because some firms exit
from the market through a non-random process.
Heckman selection model: takes into account the selection
(attrition) bias.



Selection and growth equations
Growth eq incorporates selectivity-bias correction term
Need exogenous variables that affect the selection but not growth
Results – Growth (1983-2001)
Variables
llab
1
2
0.471***
[0.009]
0.694***
[0.006]
0.316***
[0.032]
-0.055
[0.035]
0.035
[0.030]
-0.384***
[0.094]
0.151
[0.111]
fdi
fdiqs
fdiqr
fdisupp_q
fdibuy_q
cfdiqs
cfdiqr
lentrate
sectgr
sectgrpr
-0.105**
[0.053]
0.057
[0.043]
-0.269**
[0.127]
0.141
[0.158]
0.025
[0.033]
-0.007
[0.042]
0.194***
[0.039]
0.054***
[0.009]
0.043***
[0.014]
0.092***
[0.027]
0.071***
[0.008]
0.151***
[0.012]
Models
3
0.694***
[0.007]
0.276***
[0.034]
-0.074**
[0.038]
0.043
[0.032]
-0.393***
[0.095]
0.186*
[0.113]
0.058**
[0.029]
-0.099***
[0.035]
0.124***
[0.031]
0.071***
[0.008]
0.155***
[0.013]
4
5
0.667***
[0.006]
0.229***
[0.034]
-0.071*
[0.037]
0.093***
[0.032]
-0.552***
[0.091]
0.201*
[0.111]
0.048*
[0.029]
0.013
[0.036]
0.194***
[0.031]
0.046***
[0.008]
0.060***
[0.013]
0.630***
[0.007]
0.131***
[0.041]
-0.055
[0.038]
0.146***
[0.034]
-0.642***
[0.095]
0.137
[0.113]
0.062*
[0.032]
-0.101**
[0.045]
0.194***
[0.042]
0.047***
[0.009]
0.108***
[0.015]
Results – Growth (1983-2001)
Variables
hhi
mes
1
2
-0.295***
[0.090]
0.378***
[0.014]
-0.335***
[0.063]
0.192***
[0.010]
-0.070***
[0.004]
0.086***
[0.006]
0.046***
[0.011]
0.132***
[0.020]
-0.128***
[0.019]
70425
68350
lage
lw
pmargin
subinput
suboutput
Models
3
4
5
-0.318***
[0.064]
0.197***
[0.010]
-0.070***
[0.004]
0.086***
[0.006]
0.049***
[0.011]
0.106***
[0.022]
-0.131***
[0.020]
-0.317***
[0.062]
0.181***
[0.011]
-0.051***
[0.004]
0.023***
[0.006]
0.013
[0.011]
0.078***
[0.022]
-0.141***
[0.019]
0.089***
[0.003]
68350
68348
-0.480***
[0.071]
0.194***
[0.012]
0.030***
[0.005]
0.125***
[0.008]
0.026*
[0.014]
0.040
[0.027]
-0.121***
[0.023]
0.089***
[0.004]
0.056*
[0.033]
0.034***
[0.006]
47856
kl
ttrans
rddum
N
Results – Growth (2003-2009)
Variables
llab
1
2
0.741***
[0.035]
0.791***
[0.028]
0.804***
[0.219]
-0.055
[0.072]
-0.195***
[0.071]
-0.387
[0.350]
-0.282
[0.175]
fdi
fdiqs
fdiqr
fdisupp_q
fdibuy_q
cfdiqs
cfdiqr
lentrate
sectgr
sectgrpr
0.267***
[0.067]
0.024
[0.070]
-0.014
[0.307]
-0.224
[0.168]
-0.038
[0.080]
0.049
[0.175]
0.072
[0.159]
0.065**
[0.033]
0.176***
[0.062]
-1.487***
[0.219]
-0.036
[0.034]
0.203***
[0.072]
Models
3
0.795***
[0.028]
0.806***
[0.221]
-0.067
[0.077]
-0.170**
[0.072]
-0.384
[0.351]
-0.270
[0.175]
0.058
[0.079]
-0.356**
[0.168]
-1.477***
[0.219]
-0.038
[0.034]
0.205***
[0.072]
4
5
0.835***
[0.028]
0.385**
[0.182]
-0.018
[0.071]
-0.176**
[0.072]
-0.924***
[0.332]
-0.041
[0.161]
0.079
[0.077]
-0.351**
[0.173]
-1.313***
[0.213]
0.046
[0.034]
0.172**
[0.072]
0.831***
[0.027]
0.201
[0.152]
-0.009
[0.065]
-0.148**
[0.069]
-0.866***
[0.304]
-0.050
[0.150]
0.064
[0.074]
-0.387**
[0.169]
-1.251***
[0.209]
0.056*
[0.033]
0.187***
[0.070]
Results – Growth (2003-2009)
Variables
hhi
mes
1
2
-0.265***
[0.097]
0.089***
[0.023]
-0.077
[0.082]
0.018
[0.017]
-0.457***
[0.026]
0.277***
[0.052]
0.510***
[0.052]
0.113
[0.092]
0.047
[0.033]
8617
8437
lage
lw
pmargin
subinput
suboutput
Models
3
4
5
-0.070
[0.082]
0.016
[0.017]
-0.466***
[0.026]
0.274***
[0.053]
0.505***
[0.052]
0.117
[0.092]
0.045
[0.033]
-0.112
[0.076]
0.001
[0.017]
-0.465***
[0.027]
0.210***
[0.051]
0.405***
[0.057]
0.116
[0.100]
0.081**
[0.034]
0.050***
[0.009]
8437
7450
-0.123*
[0.074]
0.006
[0.016]
-0.454***
[0.026]
0.218***
[0.049]
0.401***
[0.056]
0.125
[0.096]
0.080**
[0.034]
0.052***
[0.009]
0.038
[0.023]
0.015
[0.060]
7450
kl
ttrans
rddum
N
Results – Growth Model


Foreign firms have higher growth rates than domestic
firms, even after controlling for a number of firmspecific variables, including unobserved firm-specific
factors.
The presence of foreign firms has a detrimental impact
on the growth rate of domestic firms that



either use more inputs from foreign-dominated sectors,
or operate in regions where the share of foreign firms in
regional output increase
Unexpected result: Foreign presence in supplier
industries has also a negative impact on survival of user
firms.
Impact of economic crises on
growth and survival
2.0
1.5
1.0
0.5
0.0
1990
1992
1994
1996
1998
2000
2002
-0.5
-1.0
-1.5
-2.0
Survival
Growth
2004
2006
2008
2010
Impact of economic crises on
survival (by ownership)
2.0
1.5
1.0
0.5
0.0
1990
1992
1994
1996
1998
2000
2002
-0.5
-1.0
-1.5
-2.0
Domestic
Foreign
2004
2006
2008
2010
Impact of economic crises on
Growth (by ownership)
2.0
1.5
1.0
0.5
0.0
1990
1992
1994
1996
1998
2000
2002
-0.5
-1.0
-1.5
-2.0
Domestic
Foreign
2004
2006
2008
2010
Conclusions




There are significant differences between
entry characteristics of foreign and domestic
plants.
Entry-level differences persist after entry, and
foreign plants are more likely to survive.
Although foreign plants are less likely to exit,
neither foreign ownership itself nor foreign
presence in the market matter for survival.
For growth, however, foreign ownership
matters
Thanks …