Transcript The Role of FDI in Eastern Europe and New Independent

The Role of FDI in Eastern Europe
and New Independent States: New
Channels for the Spillover Effect.
Irina Tytell
Ksenia Yudaeva
What can we expect from FDI in a
transition country?
Existing literature:
• Foreign firms are more efficient (Survey by
Lipsey (2004) and Lipsey and Sjoholm
(2005)
• Unclear, whether there are positive
spillovers (Konings (2001), Damijan et al
(2003), Javorcik and Spatareanu (2004)
• Backward linkages
Other possibilities:
• Education and institutions may have an effect
both on the type of FDI, and on spillovers
• On production function
– domestic firms become more similar or more different
from foreign
• Different FDI (exporting vs. competing with
imports) may have different effects. Exporting
FDI (Moran, 2005)
Four countries
• Poland: the richest of the 4 countries, with
relatively good quality of government and the
rule of law, but high costs of starting a business
and enforcing contracts.
• Romania: 3rd in terms of GDP per capital,
relatively good institutions (smallest costs of
starting business)
• Russia: 2nd rich, relatively bad institutions (but
lower costs of starting business and enforcing
contracts than in Poland)
• Ukraine: twice as poor as Poland, bad
institutions.
Data
•
•
•
•
Poland: Amadeus 1999-2004
Romania: Amadeus 1999-2004
Russia: Rosstat. 1998-2003.
Ukraine: Statistical office. No capital, and no costs data.
2000-2004.
• Foreign ownership data are cleaned for off-shores, NIS,
and other suspected cases.
• For Poland and Romania only the latest ownership info.
• In Russia there is information about natural resource
extracting sectors. We do not use it in regressions: list of
industries is the same for all countries.
• Russia: OKONH => NACE
FDI statistics
• In Poland and Romania the share of foreign
firms in production is above 20%.
• In Russia and Ukraine it is less that 5%
• Shares of FDI by industry in Poland and
Romania is correlated with the coefficient 0.3.
• Shares of FDI by industry in Russia and Ukraine
is correlated at 0.86%
• These shares aren’t so tightly correlated
between the two country groups.
Static specification
Methodology
• Cobb-Douglas production function:
ln VAi,t    L ln Li,t  K ln Ki,t   FDI DENSITYt ,s,r  i ,t
Keane (2005)
Industry: NACE (1.1) 2 digit.
• Proxy for spillovers:

FDI DENSITY 
t , s ,r
( L  FDI )

t , s ,r
L
Direct effect of foreign ownership
Dependent variable: Log Value Added
Russia
Ukraine
Poland
1.114***
1.143***
0.507***
[0.005]
[0.006]
[0.016]
-0.420***
0.035
-0.020
Log Employment X FDI
[0.024]
[0.029]
[0.027]
0.052***
0.298***
Log Fixed Capital
[0.003]
[0.010]
0.273***
0.041**
Log Fixed Capital X FDI
[0.016]
[0.018]
0.532***
0.210
0.089
FDI
[0.097]
[0.170]
[0.132]
Year Dummies
yes
yes
yes
Sector Dummies
yes
yes
yes
Region Dummies
yes
yes
yes
N
62305
38558
9435
R2
0.742
0.625
0.646
* significant at 10%; ** significant at 5%; *** significant at 1%; robust standard errors in brackets
Log Employment
Romania
0.672***
[0.007]
-0.073***
[0.012]
0.282***
[0.004]
0.047***
[0.007]
0.328***
[0.039]
yes
yes
yes
47114
0.765
Foreign companies have higher TFP, and higher capital intensity. Difference in
TFP is much smaller in Poland than in other countries. Difference in capital
intensity is much smaller in Poland and Romania than in Russia.
No Productivity Spillovers
Panel A: OLS on year, sector, region means
Dependent variable: Log Value Added
Log Employment
Log Fixed Capital
FDI DENSITY
N
R2
Russia
Ukraine
Poland
Romania
0.932***
[0.019]
0.191***
[0.013]
0.779***
[0.196]
5995
0.722
1.018***
[0.035]
0.960***
[0.352]
2163
0.357
0.318***
[0.025]
0.335***
[0.017]
-0.073
[0.062]
1320
0.636
0.624***
[0.020]
0.325***
[0.014]
-0.124***
[0.046]
3755
0.764
Russia
Ukraine
Poland
Romania
0.784***
[0.017]
0.009
[0.010]
-0.027
[0.094]
yes
20289
60429
0.148
0.945***
[0.018]
-
0.392***
[0.063]
0.191***
[0.055]
0.148
[0.096]
yes
3300
7483
0.126
0.604***
[0.015]
0.196***
[0.007]
0.023
[0.055]
yes
8497
35252
0.550
Panel B: Panel estimation with firm fixed effects
Dependent variable: Log Value Added
Log Employment
Log Fixed Capital
FDI DENSITY
Year and Firm Fixed Effects
N groups
N observations
R2
0.068
[0.273]
yes
13548
37628
0.320
Exception: Exporting FDI in Russia
Dependent variable: Log Value Added
Log Employment
Log Fixed Capital
FDI DENSITY
EXPORT-ORIENTED FDI DENSITY
Year and Firm Fixed Effects
N groups
N observations
R2
0.785***
[0.017]
0.009
[0.010]
-0.120
[0.106]
0.386*
[0.210]
yes
20289
60429
0.148
Production Function Spillovers?
Panel C: Panel estimation with heterogeneous production functions
Dependent variable: Log Value Added
Russia
Ukraine
Poland
0.201**
[0.099]
-0.163**
[0.067]
-0.142
[0.303]
-
-0.356**
[0.143]
0.213*
[0.110]
-0.052
[0.056]
0.008
[0.029]
Log Employment X Sector
yes
yes
yes
yes
Log Fixed Capital X Sector
yes
-
yes
yes
FDI DENSITY X Sector
yes
yes
yes
yes
Log Employment X FDI DENSITY
Log Fixed Capital X FDI DENSITY
Romania
Year and Firm Fixed Effects
yes
yes
yes
yes
N groups
20289
13490
3300
8497
N observations
60429
37424
7483
35252
R2
0.153
0.327
0.170
0.556
* significant at 10%; ** significant at 5%; *** significant at 1%; robust st. errors in brackets; p-values in parentheses
In Russia domestic firms, competing with foreigners are less capital intensive, in
Poland, where there are more FDI, domestic firms, competing with foreign, are
more capital intensive.
Role of Education: direct effect is
stronger in low education regions
Panel A: Bottom 1/3 of regions by share of people with secondary education
Dependent variable: Log Value Added
Russia
Ukraine
1.135*** 1.221***
[0.011]
[0.012]
-0.504*** 0.008
Log Employment X FDI
[0.061]
[0.065]
0.039*** Log Fixed Capital
[0.006]
0.292*** Log Fixed Capital X FDI
[0.044]
0.857*** 0.301
FDI
[0.260]
[0.389]
Panel B: Top 1/3 of regions by share of people with secondary education
Log Employment
Dependent variable: Log Value Added
Log Employment
Log Employment X FDI
Log Fixed Capital
Log Fixed Capital X FDI
FDI
Russia
Ukraine
1.086***
[0.011]
-0.387***
[0.048]
0.063***
[0.007]
0.241***
[0.029]
0.585***
[0.166]
1.032***
[0.012]
-0.018
[0.051]
0.719**
[0.329]
Poland Romania
0.476***
[0.027]
-0.005
[0.053]
0.322***
[0.017]
-0.004
[0.034]
0.505**
[0.223]
0.652***
[0.010]
-0.099***
[0.018]
0.284***
[0.007]
0.071***
[0.012]
0.352***
[0.063]
Poland Romania
0.535***
[0.026]
-0.053
[0.037]
0.277***
[0.016]
0.089***
[0.026]
-0.166
[0.177]
0.661***
[0.012]
-0.048**
[0.022]
0.295***
[0.008]
0.028**
[0.013]
0.297***
[0.064]
Role of Education: spillovers concentrated in
high education regions
Panel A: Bottom 1/3 of regions by share of people with secondary education
Dependent variable: Log Value Added
Russia
Ukraine
-0.230
[0.327]
0.051
[0.153]
0.006
[0.089]
-2.912
[2.440]
0.750
[0.509]
-0.352
[0.490]
0.016
[0.120]
0.060
[0.068]
0.114
[0.331]
0.027
[0.101]
-0.043
[0.055]
Log Employment X Sector
yes
yes
yes
yes
Log Fixed Capital X Sector
yes
-
yes
yes
FDI DENSITY
Log Employment X FDI DENSITY
Log Fixed Capital X FDI DENSITY
Poland Romania
Panel B: Top 1/3 of regions by share of people with secondary education
Dependent variable: Log Value Added
Russia
Ukraine
-0.277
[0.676]
0.412***
[0.149]
-0.244***
[0.093]
4.148*
[2.405]
-0.964*
[0.506]
-
0.026
[0.473]
-0.410**
[0.172]
0.300***
[0.116]
-0.105
[0.263]
-0.033
[0.086]
0.046
[0.044]
Log Employment X Sector
yes
yes
yes
yes
Log Fixed Capital X Sector
yes
-
yes
yes
FDI DENSITY
Log Employment X FDI DENSITY
Log Fixed Capital X FDI DENSITY
Poland Romania
Corruption in Russia: Productivity
differences are larger if corruption low
Dependent variable: Log Value Added
Log Employment
Log Employment X FDI
Log Fixed Capital
Log Fixed Capital X FDI
FDI
Year Dummies
Sector Dummies
Region Dummies
N
R2
1/3 with
high
corruption
1.093***
[0.012]
-0.516***
[0.073]
0.058***
[0.007]
0.375***
[0.045]
0.110
[0.256]
yes
yes
yes
14799
0.712
1/3 with
low
corruption
1.122***
[0.013]
-0.336***
[0.053]
0.048***
[0.008]
0.194***
[0.034]
0.817***
[0.196]
yes
yes
yes
13709
0.757
Corruption: Spillovers are concentrated
in low corruption regions
1/3 with
high
corruption
-0.341
[0.745]
0.098
[0.169]
-0.049
[0.103]
1/3 with
low
corruption
-0.874*
[0.456]
0.468**
[0.216]
-0.185
[0.114]
Log Employment X Sector
yes
yes
Log Fixed Capital X Sector
yes
yes
Year and Firm Fixed Effects
N groups
N observations
R2
yes
4743
14346
0.200
yes
4476
13138
0.126
Dependent variable: Log Value Added
FDI DENSITY
Log Employment X FDI DENSITY
Log Fixed Capital X FDI DENSITY
Countries with good institutions and
high FDI inflows (Poland):
• Foreign companies are more productive
than domestic, but difference is not large,
with the exception of the regions with low
education level
• Domestic companies in the sectors with
high FDI Density are more capital
intensive than other domestic companies.
This effect is observed only in the regions
with high education level
• No productivity spillovers
Countries with bad institutions and
low FDI inflows (Russia):
• Foreign companies are more productive than the
domestic ones, and difference in productivity is
almost 70%.
• Domestic companies in the sectors with high FDI
Density are less capital intensive and more labor
intensive than other domestic companies. This
effect is mainly observed in the regions with high
education level and/or low corruption level.
• No productivity spillovers with the exception of
exporting FDI.
Dynamic Specifications
Methodology:
Specification: autoregressive panel data with fixed
effect:
TFPi,t 1  TFPi,t   FDI TFPt ,s,r   FDI DENSITYt ,s,r  i,t
ei,t  i  vi.,t
Blundell and Bond (1998) GMM: a system of
equations in levels and first differences, which
uses lagged first differences and lagged levels
of endogenous variables as instruments
Specification
• Dependent variable: TFP is the residual
from Cobb-Douglas production functions
estimator
• Additional controls: FDI TFP.
– Peri and Urban (2004))
Dynamic effect of FDI density on
productivity
Panel A: Basic specification
Dependent variable: TFP
Lagged TFP
Lagged FDI TFP
Lagged FDI DENSITY
Russia Ukraine Poland Romania
0.712***
[0.067]
0.254***
[0.059]
-0.250
[0.437]
0.983***
[0.080]
0.002
[0.030]
0.082
[0.466]
0.285
[0.335]
0.006
[0.128]
-0.669
[1.445]
0.606***
[0.064]
0.088***
[0.030]
-0.526***
[0.175]
Panel B: Specification with threshold effects
Dependent variable: TFP
Lagged TFP
Lagged FDI TFP
Lagged FDI DENSITY
Lagged FDI DENSITY over 50%
Russia Ukraine Poland Romania
0.717***
[0.066]
0.248***
[0.057]
0.246
[0.359]
-1.404
[1.080]
0.981***
[0.081]
-0.002
[0.030]
-0.323
[0.585]
0.586
[0.528]
0.361
[0.285]
-0.022
[0.110]
-0.338
[1.166]
-0.040
[0.669]
0.598***
[0.065]
0.108***
[0.033]
-1.052***
[0.308]
0.686***
[0.235]
• More productive foreign firms produce
larger productivity spillovers
• The effect of density of foreign firms is
negative in 3 out of 4 countries
– In Romania it is less negative in the sectors
with more than 50% foreign share
• In Russia the effect of FDI Density is
positive, but insignificant, and gets
negative and insignificant if foreign share
is larger than 50%
Dynamic effects of FDI Density on
K/L Ratio
Panel A: Basic specification
Dependent variable: Log K/L
Lagged Log K/L
Lagged FDI TFP
Lagged FDI DENSITY
Russia Ukraine Poland Romania
0.940***
[0.038]
-0.002
[0.004]
-0.079
[0.150]
-
0.939***
[0.159]
0.311*
[0.161]
-3.488**
[1.664]
0.753***
[0.028]
0.075*
[0.044]
-0.596**
[0.272]
Panel B: Specification with threshold effects
Dependent variable: Log K/L
Lagged Log K/L
Lagged FDI TFP
Lagged FDI DENSITY
Lagged FDI DENSITY over 50%
Russia Ukraine Poland Romania
0.935***
[0.038]
-0.003
[0.004]
0.129
[0.205]
-0.731**
[0.365]
-
0.920***
[0.172]
0.288*
[0.163]
-3.344*
[1.922]
0.238
[1.686]
0.752***
[0.029]
0.129**
[0.053]
-1.638***
[0.587]
1.123**
[0.474]
• In Poland and Romania the effect of FDI
TFP is positive and significant, but FDI
share is negative and significant, unless
foreign share is really high.
• Previous finding that Polish firms,
competing with foreign companies, are
more capital intensive can be due to the
effect of FDI quality.
• In Russia: no or negative effect of both
FDI TFP and FDI DENCITY.
Role of Education: productivity of
Romanian firms
Effect on productivity: In high educated regions positive effect of FDI
productivity, and insignificant spillovers. In Low education regions the first
effect is insignificant, and the second effect is negative.
Dependent variable: TFP
Lagged TFP
Lagged FDI TFP
Lagged FDI DENSITY
loweducation
bottom
1/3
0.620***
[0.093]
-2.944
[0.048]
-0.702***
[0.224]
higheducation
top 1/3
0.623***
[0.100]
0.093*
[0.049]
-0.245
[0.339]
Role of education: K/L ratio of
Romanian firms.
In highly educated regions foreign firms productivity has positive effect, and
density is insignificant. In low educated regions foreign firms productivity is
insignificant, and density is negative
Dependent variable: Log K/L
Lagged Log K/L
Lagged FDI TFP
Lagged FDI DENSITY
loweducation
bottom
1/3
0.713***
[0.056]
0.078
[0.066]
-0.829**
[0.352]
higheducation
top 1/3
0.739***
[0.046]
0.121*
[0.073]
-0.612
[0.553]
Summary of findings from dynamic
specifications
• More productive FDI lead to both productivity
and production function spillovers, which we
observed earlier.
• Both effects are stronger if education level is
high.
• After controlling for FDI productivity, FDI density
has no or negative effect on domestic firms
productivity and K/L ratio. But negative effect is
subject to a threshold.
Conclusions
• Differences in production function spillovers between 2
groups of countries, which may result from differences in
the type of FDI, which go to these countries
– Vertical in Poland and Romania
– Horizontal in Russia
• No productivity spillovers in the static specification with
the exception of exporting FDI. In dynamic specification
increased density of FDI produced negative effect on
host country firms productivity.
• High education and low corruption enhance production
function spillovers and positive dynamic effect from very
productive foreign companies.