The European Journal of Comparative Economics

Vol. 5, n. 1, pp. 87-105

ISSN 1722-4667

Determinants of Economic Growth:

Empirical Evidence from Russian Regions
Svetlana Ledyaeva*
Department of Business and Economics, University of Joensuu
Centre for Markets in Transition, Helsinki School of Economics
Mikael Linden*
Department of Business and Economics, University of Joensuu
Abstract

A modification of Barro and Sala-i-Martin empirical framework of growth model is specified to

examine determinants of per capita growth in 74 Russian regions during period of 1996-2005. We
utilize both panel and cross-sectional data. Results imply that in general regional growth in 1996-2005
is explained by the initial level of region’s economic development, the 1998 financial crisis, domestic
investments, and exports. Growth convergence between poor and rich regions in Russia was not found
for the period studied.
JEL Classification: E22, F21, P27
Keywords: Russian regions, economic growth

1. Introduction
Empirical growth analysis was pioneered by Barro (1991) and Mankiw et al
(1992). A large empirical literature on the determinants of economic growth in
transition economies appeared in the 1990s and 2000s, including Fischer, Sahay and
Vegh (1998), Havrylyshyn, Izvorski and van Rooden (1998), Berg et al. (1999), and
Havrylyshyn and van Rooden (2000). The studies have identified a variety of
microeconomic, structural, and institutional factors of economic growth in transition
economies in general. A good description of empirical literature published in the
1990s is available in a survey by Havrylyshyn (2001).
For the Russian economy the question of determinants of economic growth
during transition remains an open question. There are a lot of variables which could
be included into the growth model specification taking into consideration the fact that
the “traditional” growth regressions literature is quite different from the more recent
literature explaining growth in transition economies. Papers aiming to shed light on
this are few. Berkowitz and DeJong (2003) found that regional difference in reform
policies and in the formation in new legal enterprises can help account for regional
differences in growth rates in Russia. They estimate growth regression by Ordinary
Least Squares (here and after OLS) and Two Stage Least Squares (2SLS) using
cross-sectional data for 48 Russian regions. Note that regional growth differences for
such economies as USA and China are comparable. Both countries occupy quite
large territories which consist of many regions: states in USA and provinces in
China.
Papyrakis and Gerlagh (2007) analyze empirically determinants of economic
growth in the United States using cross-sectional data on 49 states. Their dependent

*Contact information: Department of Business and Economics, P.O. Box 111, FIN-80101
Joensuu, University of Joensuu, Finland. Emails: [email protected],
[email protected].
Available online at http://eaces.liuc.it
88
EJCE, vol. 5, n. 1 (2008)

variable is growth rate of Gross State Product (GRP). The regressors are initial
income, natural resources, investment, schooling, openness and corruption. They
found that empirical data seem to support the absolute convergence hypothesis for
US states, but the data also show that natural resource abundance is a significant
negative determinant of growth.
Cai, Wang and Du (2002) analyze empirically determinants of economic
growth in Chinese provinces during the period 1978-1998. They estimate
specification using panel data by OLS and FGLS. The finding is that (1) there is an
evidence of conditional convergence in China’s growth, namely, per capita GDP in
the initiative year is negatively related to growth rates in following years, (2) labor
market distortion negatively impacts regional growth rates, and (3) many other
variables used in previous studies impact growth performance.
Dermurger (2000) utilizes the same empirical panel data framework as Cai,
Wang and Du (2002) to analyze panel data from a sample of 24 Chinese provinces
(excluding municipalities) throughout the 1985 to 1998 period. The estimation of a
growth model shows that, besides differences in terms of reforms and openness,
geographical location and infrastructure endowment did account significantly for
observed differences in growth performance across provinces. The significant and
negative coefficient associated to the logarithm of lagged GDP per capita indicates a
catch-up phenomenon among Chinese provinces.
This paper attempts to find some evidence on the determinants of economic
growth across Russian regions. As the background of empirical analysis of regional
determinants of economic growth in Russia is very small, our focus is on the
traditional factors of economic growth. Special emphasize is put on dynamic panel
data methods to control for endogeneity problems found in growth empirics. We use
also the Oaxaca-Blinder decomposition method to examine the extent to which
differences in growth rates between sub-samples of relatively poor and rich Russian
regions can be explained by differences in specified factors of economic growth.
According to neoclassical theory lower-income countries tend to grow faster than
higher-income countries. The Oaxaca-Blinder decomposition helped us to find
further evidence on the factors of convergence between lower-income and higher-
income regions in present day Russia.
The main results of our paper are the following. We found that conditional
convergence is relevant across the Russian regions during the transition period.
Domestic investment and export can be considered as important factors of economic
growth in Russia. The Oaxaca-Blinder analysis produced some evidence on the
relative magnitudes of different factors of convergence across Russian regions, e.g.
that initial GRP per capita plays an important role here along with domestic
investments.
The reminder of the paper is constructed as follows. Section 2 describes the
background theory for the empirical model. Section 3 describes the data and
variables. Section 4 gives the estimation methods. Section 5 reports the results with
some discussion. Some additional results are given Section 6, and Section 7
concludes.

2. Empirical model
Growth regression studies have been used to explain differences in economic
performance across nations and regions. Assuming diminishing returns to capital,
neoclassical growth theory predicts a convergent growth trend among nations or
regions, i.e. poor countries or regions tend to grow faster than rich ones (Mankiw,
Available online at http://eaces.liuc.it
 89
Determinants of Economic Growth: Empirical Evidence from Russian Regions

Romer, & Weil, 1992). Islam (1995) was first to propose a dynamic panel data
approach to modifying the Mankiw-Romer-Weil model.
Assume each region i has a following production function:
Yt = F (K t , Lt , Xt ) [1]

where Yt is the total production at time t, F(.) is a concave production function

with homogeneity degree of one, Kt is the stock of physical capital, Lt is the labour
force, and Xt is a vector of all other relevant production inputs. The properties of
production function allow us to write it in the labour intensive form, i.e.
[2]
Yt / Lt = F ( Kt / Lt , I , Xt / Lt ) ⇒ yt = f ( kt , xt ).
Time differentiation of Eq. 2) gives
dy dk t N dx
j ,t
+
= f1 ∑ j=2 fj [3]
t

dt dt
dt
Let y* denote the steady-state level of income per effective worker, and y be
let
t t
its actual value at any time t, where t is the period average as in Islam (1995).
Approximating around the steady state the pace of convergence is given by
∂ ln yt = λ(ln y * − ln y ) [4]
∂t t t

where λ is the speed of convergence. This equation implies for given

ln y * and ln y
t t −1

ln y = (1 − e − λ t ) ln y * + e −λt ln y [5]
t t t −1

hjkj
Because equation (5) holds at any time, it can be rewritten by subtracting one-
period lag, ln yt −1 , from both sides:
∆ ln y = (1 − e − λ t ) ln y * + (e −λt −1) ln y [6]
t t t −1

Equation (3) expresses the convergence process of growth rate over time. It
implies convergence in growth rates, conditional on the steady-state growth rate.
Equation (6) is a general feature of the neoclassical growth model, without relying on
the Mankiw-Romer-Weil approximation. That is, if the steady-state growth rates are
identical across countries, the actual growth rates must convergence.
In order to estimate the described scheme in panel data regressions we use the
empirical framework suggested by Barro and Sala-I-Martin (1995) adopted for panel
data (see, e.g., Soto 2000, Carcovic and Levine 2002, Laureti and Postiglione 2005).
This framework relates real per capita growth rate to initial levels of state variables,
such as the stock of physical capital and the stock of human capital, and to control
variables. Following the idea of Barro and Sala-I-Martin (1995), we assume that a
higher level of initial per capita GRP reflects a greater stock of physical capital per
capita. Following
Available online at http://eaces.liuc.it
90
EJCE, vol. 5, n. 1 (2008)

Soto (2000), we also assume that the initial stock of human capital is reflected in the
lagged value of per capita output in the short-run. The neoclassical growth model
predicts that, for given values of the control variables, an equiproportionate increase
in initial levels of state variables reduces the growth rate. Thus we can approximate
equation (6) with reference to Eq. 2) and Eq. 3)1
∆ ln yit = α1 ln yi , t −1 + α2 ln ki ,t + β 'ln xit [7]
where y is per capita GRP in region i (i=1, in period t (t=1996,
…,74)2 …,2005),
i ,t

yi , t −1 is (initial) per capita GRP in region in period t-1, α1 is a negative parameter

i
reflecting the convergence speed, α2 is a positive parameter giving the impact of
capital-labour ratio to per capita GRP growth rate, xi together with ki ,tis a row vector
,t
of control variables in region i during period t with associated parameters β .

3. Data and variable choice

We use five control variables which can be viewed as important factors in the
Russian economy’s regional development in the analyzed period. They are
represented in Table 1.

Table 1. Control variables*

Variable Description
Dummy_1998 Dummy variable for the year 1998 of major financial crisis in Russia
ln (I / N )i ,t Natural logarithm of per capita domestic investment
ln (Exp / N )i ,t Natural logarithm of per capita export
ln( R / N )i ,t Natural logarithm of resource index
ln(FDI/N)i,t Natural logarithm of per capita Foreign Direct Investment (here and
after FDI)
*) all variables are region i =1,…,74 in period t =1996,…,2005
for

First we include a dummy variable for the year 1998, to control for the major
financial crisis that occurred in Russia. The second variable is the natural logarithm
of
per capita domestic investment in physical capital, ln (I / N )i ,t , i.e. investment
originated from Russia, in million dollar in year 2000 prices 3. According to the
existing theory and most empirical findings we expect this to be positively related to
the dependent variable. Note that we do not use capital stock here as a variable.
Instead we have the change of stock, i.e. investment per capita, as source of
economic growth. The lagged stock effects operate via the lagged output per capita
variable.

1 Instead we could have assumed that production function in Eq. 2) is Cobb/Douglas type and log-
linearize it.
2 Actually there are 89 regions in Russia. We exclude from the analysis the autonomous territories,
which are included in other regions. These are Neneckij, Komi-Permyatckij, Hanty-Mansijskij,
Yamalo-Neneckij, Dolgano-Neneckij, Evenkijskij, Ust-Ordynskij and Aginskij Buryatskij, and
Koryakskij. Regions for which most data are missing, namely Ingushetiya, Chechnya, Kalmykiya,
Alaniya, Mari-el and Chukotka, are also excluded.
3 The transformation was done using the USA deflator, which is 100 for the year of 2000.
Available online at http://eaces.liuc.it
 91
Determinants of Economic Growth: Empirical Evidence from Russian Regions

The third variable is the natural logarithm of per capita export (from region i to
abroad), ln (Exp / N )i ,t , in million dollars at year-2000 prices. This variable was
included
to predict the positive contribution of the degree of openness of the regional
economy to economic growth. We add this variable to test the hypothesis of export-
led economic growth in Russia during transition. This proposition is also based on
recent finding of Awokuse (2007) that trade stimulates economic growth in three
transition economies, namely Bulgaria, Czech Republic and Poland.
The fourth variable, the natural logarithm of the resource index, ln( R / N )i , t (for
calculation details, see Appendix 1), was included because of the high dependence of
Russian economy on natural resources. In accordance with the aggregate production
function approach in the short run, the natural resources stock is positively related to
economic growth and is treated as an additional input. As we operate with a short
period of only 10 years (1996-2005) of the present transitory phase of the Russian
economy, we expect this variable to have some importance for the Russian regional
growth process. However historic experience across countries provides support to the
hypothesis that resource scarce economies often outperform resource-abundant
countries in terms of economic growth. A number of recent studies has described and
analysed the resource curse hypothesis (Gylfason, 2000, 2001ab, Leite and
Weidmann 1999, Papyrakis and Gerlagh 2004, Rodriguez and Sachs 1999, Sachs and
Warner 1995, 1997, 1999a). The conclusion is widely accepted: natural riches tend to
frustrate rather than promote economic growth. Thus it is important to investigate if
the resource abundance (in particular oil and gas abundance) retards or stimulates
regional growth in Russia.
The last variable included in our specification is natural logarithm of FDI per
capita in million dollars in year-2000 prices. Most theoretical and empirical findings
imply that FDI has a strong positive growth impact on the recipient economy (see,
e.g., theoretical studies of Dunning and Narula 1996, Borenstein, Gregorio and Lee
1998, Markusen and Venables 1999, and empirical studies of Balasubramanyam,
Salisu and Sapsford 1996, Bende -Nabende and Ford 1998, Soto 2000, Li and Liu
2005). For the Russian economy the question of aggregate FDI impact on economic
growth remains an open question.
The source of all data used is Russia’s regions yearbooks published yearly by
Goskomstat (Russian State Statistical Agency). We are aware that Russian statistics
suffer from a bad data problem. However Goskomstat is the only public source of
Russian statistics. We are also aware that the estimated period is a transition period
in Russia characterized for macro instability. This also may cloud the estimation
results. We carefully examined the data and found that the main source of instability
in it is financial crisis of 1998. Before and after this year the dynamics of variables is
more or less stable. Thus by introducing a dummy variable for the year of 1998 we
partly solve this problem.
Empirical panel data studies on growth are generally carried out for periods of
around 30 years, with five-year or three-year average observations (see e.g. Barro
and Lee 1994, Caselli, Esquivel and Lefort 1996, Carcovic and Levine 2002). Our
estimation time period is limited to 10 years (1996-2005) since 1) relatively short
transition period of the Russian economy (15 years), 2) capital inflows into Russia
have been registered by the state statistical authorities only since 1995, and 3) the
data for all the other variables are available only since 1996. Thus we follow Soto
(2000) and Laureti and Postiglione
Available online at http://eaces.liuc.it
92
EJCE, vol. 5, n. 1 (2008)

(2005) and utilize yearly data in our estimations. The summary statistics of all the
variables involved in baseline estimation is represented in Appendix 2.

4. Econometric Methods
The specified growth adjustment model (Eq. 7, above), is estimated with panel
data allowing for region and time specific components. That is

∆ ln yit = α1 ln yi,t-1 + α2 ln (I/N)it + β1 ln (Exp / N)it + β2 ln ( R / N)it +

[8]
β3 ln (FDI / N)it + β4D1998 + νi + τt + εit
where vi is a region-specific effect, τt is a period-specific effect common to all
regions, and εit is the error term.
The fixed effect (FE) and random effect (RE) estimators of the panel data
model with lagged dependent variable in the set of regressors produces biased
coefficient estimates with small samples. The basic problem of using least square
methods is that the lagged dependent variable is correlated with the error term, as the
dependent
variable lnyit is a function of vi , and it immediately follows that lnyi , t −1 is also a
function of vi (for details, see e.g. Baltagi, 2002, pp. 129-131)4.
In order to cope with the above mentioned problems estimators based on the
N→
General Method of Moments (GMM) are employed, which are consistent for
∞
with fixed T. We exploit the GMM-DIFF (GMM based on differenced variables)
procedure of Arellano and Bond (1991). This calls for first differencing and using
lags of the dependent and explanatory variables as instruments for the lagged
dependent and endogenous right hand side variables.
In general, the GMM estimator can be viewed as simultaneous estimation of a
system of equations, one for each year, using different instruments in each equation
and restricting the parameters to be equal across equations. First-differencing the
equations
removes the individual effects vi , thus eliminating a potential source of omitted
variables bias estimation, and removes the problems of series non-stationary. Note
also that one of the advantages of using a dynamic model is that both short-run and
long-run elasticities can be obtained.
As linear GMM estimators, the Arellano-Bond estimators have one- and two-
step variants. Bond (2002, p.9-10) pointed out that: “…a lot of applied work using
these GMM estimators has focused on results for the one-step estimator than the two-
step estimator. This is partly because simulation studies have suggested very modest
efficiency gains from using the two-step version, even in the presence of
considerable heteroskedasticity…. Simulation studies have shown that the
asymptotic standard errors
tend to be much too small, or the asymptotic t-ratios much too big, for the two-step
estimator, in sample size where the equivalent tests based on the one-step estimator
are quite accurate. Windmeijer (2000) provides a formal analysis of this issue, and
proposes a finite-sample correction for the asymptotic variance of the two-step GMM
estimator which is potentially very useful in this class of models.” In our study we
report two-step variants of the estimators (we also did one-step estimation, but as the
results are quite

4 However we estimated fixed effects and random effects models for robust-checking purposes (for

detail section 5).

Available online at http://eaces.liuc.it
 93
Determinants of Economic Growth: Empirical Evidence from Russian Regions

similar we report only the two-step robust estimations). They are obtained using a
finite-sample correction to the two-step covariance matrix derived by Windmeijer
(2005).
GMM estimation has the further advantage that it can treat the explanatory
variables as strictly exogenous, predetermined or endogenous. If we assume that
explanatory variables ( X nt ) are strictly exogenous (i.e. that E ( X nt ξns ) = 0 for all t, s =
1,
2, …, T) then the current and all lagged Xit are valid instruments for the lagged
dependent variable as a regressor. If X nt are assumed to be predetermined
( E(X ξ ) = 0 for all t ≤ s ), then only X i1 , X i 2 ,...X
nt nt
are valid instruments. And,
i , s−1
finally, Xit are allowed to be endogenous (E ( X nt ξnt ) = 0 for t < s) then only
if
X i1 , X i 2 ,...X are valid instruments. For further details, see e.g. Bond (2002) and
i , s−2
Baltagi (2002, p. 129-136).

5. Results

5.1. Base-line results

Estimation results for panel data are presented in Table 2 and 3. Note that we
estimated model in Eq. 8 first in pooled OLS and in regional FE/RE effect forms.
GMM-DIFF is first-differencing version of the model. The descriptive statistics of
the dependent and explanatory variables is represented in Appendix 2. In accordance
with descriptive statistics of the dependent variable several extreme outliers have
been detected and trimmed away from the sample. However in order to control for
possibly remaining outliers we also perform pooled Least Absolute Deviation (LAD)
estimation. BP –stands for Breusch-Pagan test for pooled of data, i.e. ignoring the
regional effects. For panel estimation we perform fixed effects model as Hausman
test firmly favours fixed effects model against random effects model. The correlation
matrix of the variables is represented in Appendix 3.

Table 2. Panel data estimation for regional growth. Yearly observations 1996 - 2005. Dependent variable:
GRP per capita growth rate (t-values in parenthesis)

Variable Pooled OLS Pooled LAD3) Fixed effects

Constant 0.053 (0.40) -0.07 (-0.42)
Initial income -0.55 (-22.40)* -0.38 (-11.30)* -0.70 (-25.00)*
ln(I/N) 0.4 (21.80)* 0.27 (10.10)* 0.60 (27.40)*
3
ln(Exp/N) 0.0 (4.50)* 0.034 (3.40)* -0.006 (-0.38)
4
ln(R/N) -0.0003 (-0.88) -0.0005 (-1.15) 0.001 (0.82)
ln(FDI/N) -0.0002 (-0.24) -0.0004 (-0.32) -0.0005 (-0.48)
D1998 -0.22 (-6.70)* -0.36 (-8.30)* -0.14 (-4.60)*
Number of obs. 583 583 583
Adjusted R2 0.6 0.73
4
BP test 1) 45.70* (p-value: 0.0)
Hausman test 2) 1695.70* (p-value:
0.0)
1) The null hypothesis: the pooled OLS model is adequate against the random effects alternative.
2) The null hypothesis: the random effects model is consistent against the fixed effects model.
3) LAD – Least Absolute Deviation estimation. *) Statistically significant at 5% level
Available online at http://eaces.liuc.it
94
EJCE, vol. 5, n. 1 (2008)

We report here the GMM-DIFF results under assumptions that all explanatory
variables are endogenous. We use the levels of explanatory variables as instruments
and utilize standard expanding instruments starting with two year lags. Two statistics
evaluate the validity of the instruments used. The Sargan statistic of over-identifying
restrictions tests the hypothesis that the instruments are not correlated with the
residuals. The hypothesis is essential for the consistency of the estimators. The
Arellano-Bond methodology assumes also that there is no second order
autocorrelation in the first difference errors. However first order autocorrelation is
acceptable. Arellano and Bond (1991) suggest tests for this.
Table 3. GMM panel data estimation for regional growth. Yearly observations 1998 -2005. Dependant
variable: GRP growth rate (asympt. t-values in parenthesis)

Variable GMM two-step GMM two-step

with time dummies
Constant 0.067 (23.90)* 0.14 (13.90)*
Initial income -0.71 (27.10)* -1.05 (-4.20)*
ln(I/N) 0.34 (19.70)* 0.06 (5.72)*
ln(Exp/N) 0.044 (4.68)* 0.00 (0.22)
1
ln(R/N) 0.003 (6.02)* -0.001 (-1.12)
ln(FDI/N) 0.001 (1.59) -0.001 (-1.61)
D1998 -0.56 (-28.64)* -0.53 (-36.62)*
D1999 -0.59 (-38.03)*
D2000 0.017 (1.30)
D2001 -0.012 (-1.13)
D2002 0.038 (3.82)*
D2003 0.076 (7.72)*
D2004 0.12 (9.51)*
D2005 0.14 (13.88)*
Number of obs. 499 499
Test for residual AR(1) -5.57* (p-value: 0.00) 0.58 (p-value: 0.56)
Test for residual AR(2) 0.01 (p -value: 0.99) -2.75* (p-value:
0.01)
Sargan test for instrument validity 60.10* (p-value: 0.01) 39.1 (p-value: 0.33)
0
*) Statistically significant at 5% level

From the results we conclude that they do not differ much between panel and
GMM-DIFF estimations. The magnitude of coefficient for domestic investment
variable is considerably lower with time dummies. The convergence parameter is
larger (in module terms) for the results with time dummies. However the coefficient
of export variable is positive and significant for the results without regional effects
and time dummies.
Though time dummies are highly significant in GMM-DIFF regressions we
will focus on the results without time dummies. We still control for the year of 1998
when the major financial crisis during transition happened in Russia. Note also that
the misspecification test for residual AR(2) structure in the first difference errors
indicate that GMM-DIFF results are more reliable for the case without time
dummies.
The calculated parameter estimates for α1 are negative which indicates
conditional conver-gence, i.e. higher growth in response to lower starting GRP per
capita when the other explanatory variables are held constant. In particular the result
Available online at http://eaces.liuc.it
 95
Determinants of Economic Growth: Empirical Evidence from Russian Regions

predicts that a 1% increase in the initial level of GRP per capita leads to 0. 71%
decrease in GRP per capita growth.
Using equation (6), i.e. α1 = e−λt −1 (for our case t = 1 ), we can calculate the
speed of convergence. For α1 = −0.71 it equals to 1.24%. Thus 1.24% of the GRP per
capita gap is closed every year. This is smaller than “2% rule”. In general the
conclusion is that conditional convergence in Russian regions is slower than that
expected from growth theory. Dummy variable for financial crisis in 1998 is
negatively related to economic growth in Russian regions. In particular, if the other
explanatory variables are held constant the financial crisis of 1998 decreases GRP
per capita growth in Russia by 0. 56%.
The results imply that the most important factor of economic growth in Russian
regions during 1996-2005 was domestic investment, ln (I / N ) . This is a typical
result in
the theoretical and empirical literature. 1% increase in the level of domestic
investment leads to 0. 34% increase in GRP per capita growth.
No other variables show any evident statistical relationship with the dependent
variable. There is some small evidence that export is positively related to economic
growth (however the magnitude of the coefficient is very small). However the result
is not stable.
Foreign direct investment seems not to be important for Russian economic
development in the analysed period. The result may be due to their small amounts,
and to the problem of “bad” statistics and/or the impossibility to use FDI variable
with considerable time lags (4-5 years). The latter fact arises due to theoretical
proposition that FDI effects economic growth with considerable time lag. We
included FDI variable with lags of 1 and 2 in the model but the estimated coefficients
were not statistically significant (not reported). The insignificance of foreign direct
investment may be explained also by its disadvantageous industrial distribution
across industries and regions. FDI in Russia are mostly concentrated in Moscow
region, and in market oriented industries (food industry. trade and catering).
Natural resources themselves do not necessarily enhance economic growth in
the short-run. But still domestic investment in resource industries may be quite
productive, especially if it is associated with exports. Thus resources may positively
influence economic growth through investment variable. However further research is
needed to prove this proposition.

5.2. Sensitivity analysis of baseline results

First we construct a dataset with 5 years intervals (averages of period 1996-
2000 and 2001-2005), i.e. there are only two observations along the time dimension.
The panel data results are presented in Table 4. In general the results confirm our
base-line findings.
However the convergence parameter is considerably smaller now. This is quite
expected as in our baseline estimation it reflects the dependence of GRP per capita
growth rate in current year on GRP per capita level in previous year. Here the
convergence parameter reflects the dependence of average GRP per capita growth
rate in five-year period on GRP per capita in the year that precede this period. In
general we conclude that a 1% increase in the initial per capita income leads to 0.20-
0.22 % decrease in the average GRP per capita growth in the coming five years.

Available online at http://eaces.liuc.it

96
EJCE, vol. 5, n. 1 (2008)

Table 4. Panel data estimation for regional growth. Data in 5 years intervals (1996-2000 and 2001-2005).
Dependant variable: GRP growth rate. Initial incomes: years 1995 and 2000

Variable Pooled OLS Pooled LAD1)

Constant -0.43 (-2.09)* -0.68 (-2.91)*
Initial income -0.20 (-11.26)* -0.22 (-9.51)*
ln(I/N) 0.10 (3.50)* 0.09 (2.62)*
ln(Exp/N) 0.07 (5.60)* 0.05 (3.97)*
ln(R/N) -0.00 (-0.05) 0.0005 (0.85)
ln(FDI/N) 0.001 (0.45) 0.003 (0.97)
Dummy for 2nd period 0.20 (8.77)* 0.20 (7.53)*
Number of obs. 116 116
Adjusted R2 0.78
BP test 2) 8.28* (0.004)
1) LAD – Least Absolute Deviation estimation.
2) The null hypothesis: the pooled OLS model is adequate against the random
effects alternative. * Statistically significant at 5% level

As next step we estimate cross-section regressions for two sub-periods:

averages of periods 1996-1999 and 2000-2005. Here our question is whether the
growth determinants have changed from the “low growth years” to “the high growth
years”. The results are presented in Table 5.
Table 5. Cross-sectional results. The dependent variable is GRP per capita growth as average of period
1996-1999 and 2000-2005

Period 1996-1999 Period 2000-2005

Variable Initial income: year 1996 Initial income: year 2000
OLS LAD OLS LAD
Constant -0.35 (- -0.50 (-2.80)* 0.33 (0.90) 0.46 (0.80)
2.40)*
Initial income -0.23(- -0.22(-15.00)* 0.05 (0.62) 0.08 (0.65)
20.80)*
ln(I/N) 0.14 (6.40)* 0.11 (4.30)* -0.045 (-0.85) -0.053 (-0.61)
ln(Exp/N) 0.03 (4.80)* 0.038 (3.40)* 0.04 (2.22)* 0.035 (0.96)
ln(R/N) -0.00 (0.05) 0.001 (0.84) -0.00 (-0.10) -0.003 (-0.22)
ln(FDI/N) 0.002 (1.80)* 0.003 (0.90) 0.005 (1.40) 0.005 (0.50)
Number of obs. 59 59 55 55
Adjusted R2 0.94 0.20
Bera-Jarque 6.70 0.11
Normality test for
(p- (p-value:0.95)
residuals
value:0.04)
Heterogeneity test 37.50 30.00
for residuals (p-value:0.01) (p-value:0.07)
*) Statistically significant at 5% level

The results imply that in recent years there has been a significant structural
break in Russian economic development. Our specification works very poorly in the
second period – “the high growth years”. Adjusted R2 decreases from 0.94 to 0.20 in
the second period. Among the variables considered only export has a positive
influence, albeit quite small, on economic growth in Russia. This effect is
statistically significant at
Available online at http://eaces.liuc.it
 97
Determinants of Economic Growth: Empirical Evidence from Russian Regions

5% level. In particular, a 1% growth in export per capita increases GRP per capita
economic growth by 0.05-0.07 %. All the other variables are not statistically
significant.

6. Convergence between Russian regions. High-income regions versus

low-income regions

6.1. Regressions results

In order to find evidence for convergence between Russian regions we divide
Russian regions into two sub-samples on the basis of the average GRP per capita for
the regions in the period of 1996-2005. The first sub-sample consists of regions with
above-average GRP per capita value, and second sub-sample corresponds to lower-
income regions. By splitting the sample into rich and poor regions enables us to see if
the convergence speed differs between rich and poor Russian regions. Moreover
Oaxaca – Blinder decomposition presented below helps to shed light on the factors
that retard convergence between these two groups of regions.
Taking into account the fact that the Russian economy relies significantly on
natural resources, the division into rich and poor regions may be highly influenced by
resource availability in the regions, and so the main factor in absorptive capacity may
be resource availability. To check this we calculated the average resource index for
both groups and found that it does not differ significantly between the groups.
However we note that our dataset is not complete, i.e. we removed outliers and the
data is not available for some regions.
The estimation results for different regions are represented in Table 6.

Table 6. Pooled OLS and GMM-DIFF estimation results for the sub-samples of high income and low
income regions. Yearly observations 1996-2005 and 1998-2005.Dependant variable: GRP growth rate.

High-income regions Low-income regions

Pooled OLS GMM-DIFF Pooled OLS GMM-DIFF
Constant 0.2 (0.90) 0.087 (11.69)* -0.22 (-1.2) -0.06 (30.7)*
3
Initial income -0.48 (- -0.69 (21.85)* -0.62 (-20.00)* -0.73 (34.08)*
10.00)*
ln(I/N) 0.4 (10.00)* 0.24 (6.61)* 0.45 (20.50)* 0.36 (45.25)*
1
ln(Exp/N) 0.0 (1.40) 0.09 (4.81)* 0.04 (3.80)* 0.028 (4.03)*
3
ln(R/N) -0.001 (-1.30) 0.0005 (0.49) -0.001 (-0.50) 0.002 (12.23)*
ln(FDI/N) -0.03 (-2.60)* -0.017 (-1.61) 0.001 (0.20) 0.001 (6.38)*
D1998 -0.25 (-3.70)* -0.61 (-18.42)* -0.19 (-5.20)* -0.54 (-37.68)*
Number of obs. 176 152 401 341
Adjusted R2 0.5 0.69
5
BP test 1) 5.80* 35.40*
(p-value 0.02) (p-value: 0.00)
Test for -3.27* -4.63*
residual AR 1 (p-value:0.00) (p-value:0.00)
Test for -0.60 0.31
residual AR 2 (p-value:0.55) (p-value:0.75)
Sargan test for 19.74 45.77
instrument validity (p-value: 0.99) (p-value: 0.13)
1) The null hypothesis: the pooled OLS model is adequate against the random
effects alternative. *) Statistically significant at 5% level
Available online at http://eaces.liuc.it
98
EJCE, vol. 5, n. 1 (2008)

The results confirm the importance of domestic investment for economic

growth and conditional convergence in Russian regions. In general the results do not
differ much between low income and high income regions. The main difference is
that in low income regions FDI has positive effect on economic growth while in high
income regions it has negative effect. The some evidence is found that convergence
among the low income regions is faster than in high income regions. The initial
income variable coefficients estimates are larger is absolute terms for the low income
group compared to the high income group.

6.2. Oaxaca-Blinder decomposition of economic growth difference

We use the Oaxaca-Blinder decomposition approach (see e.g. Wei 2005,
Blinder 1973, Oaxaca 1973) to examine the contribution of control factors to the
difference in GRP per capita growth between the two sub-samples. Originally
Oaxaca-Blinder decomposition was proposed to analyze the contribution of factors to
the wage gender difference. Here we adopt it for the macroeconomic analysis, i.e. in
our study GRP per capita difference substitutes wage difference. By estimating a
growth regression for the whole sample we cannot see the contribution of control
variables to the regional specific convergence processes.
As predicted by neoclassical growth theory, the poor countries (here regions)
tend to grow faster than richer ones. However in Russian regions for the analysed
period this proposition is not true (see Table 7). The result also motivates use of the
Oaxaca– Blinder method in analysing the factors that retard convergence across
Russia.

Table 7. Growth rate difference between low income and high income regions
Mean of lower-income regions growth rates in the period of 1997-2005 (1) 0.056
Mean of higher-income regions growth rates in the period of 1997-2005 (2) 0.081
Difference (1-2) -0.025

As long as the expected means of the error terms in the regressions for high and
low income regions are both zeros, the total estimated difference in average GRP per
capita growth between the sub-samples can be represented by
ˆ ˆ
ln ( yit / y ) − ln ( y / y ) = β ` lnX li lnX hi . [9]
i , t −1 li it i , t −1 hi l − β `h
where ˆ ˆ represent, respectively the estimated 5

β `h and β `l on pooled OLS -

estimation coefficients of regressions for higher-income and lower-income regions sub-
samples (including constant). lnX hi and lnXli represent the averages of modeled factors
of economic growth for the two sub-samples. The total estimated difference or gap
can be further decomposed into the following three components:
ˆ + ˆ ˆ + ˆ ˆ
ln ( yit / yi , t −1 )li − ln ( yit / yi , t −1 )hi = β `h (lnX li − lnX hi ) (β`l − β `h )lnX hi (β`l − β `h )(lnX li − lnX hi )
[10]
=E+C+CE

5 Oaxaca-Blinder decomposition was originally derived for classical OLS regression (see eg Yun,

2004). The GMM approach allows in theory for decomposition but some practical problems remain.

Available online at http://eaces.liuc.it

 99
Determinants of Economic Growth: Empirical Evidence from Russian Regions

The first component on the right-hand side (E) is the portion of the gap due to
the difference in structural and control factors. The coefficient component C is
attributable to differences unexplained by these factors. CE is the interaction factor
between these two components. Note that method also generates detailed
decomposition results for individual regressors, i.e. specified factors of economic
growth.

6.3. Difference in growth rates between high income and low income
Russian regions: Factors of convergence
Table 8 reports the (predicted) difference decomposition of growth rates
between lower-income and higher-income regions derived from estimations. As the
results are based on pooled panel OLS estimation, the conclusions are preliminary
and approximate. However since the relative importance of specified factors is
similar in all estimations in Table 6. the inferences drawn can be useful.
Table 8. Predicted growth rates and decomposition of growth rates differences between lower-income and
higher-income regions. 1997-2004

Mean predictions and predicted gap

Mean prediction for lower-income regions 0.065

Mean prediction for higher-income regions 0.080
Predicted gap -0.015
Detailed linear decompositions

Total Factors Coefficients Interaction

Constant -0.30 -0.30 (-0.90)
lny
i , t −1 1.20 0.25 (7.00)* 0.85 (2.70)* 0.10 (2.60)*
ln (I / N )i ,t -0.78 -0.25 (-6.70)* -0.48 (-1.30) -0.05 (-1.30)
ln (Exp / N )i ,t -0.038 -0.04 (-1.20) 0.001 (0.10) 0.002 (0.06)
ln(R/N)i ,t -0.01 0.001 (0.70) -0.01 (-0.60) -0.001 (-0.40)
ln(FDI/N)i.t -0.13 0.13 (2.30)* -0.13 (-2.50)* -0.13 (-2.30)*
D 0.012 0.002 (0.90) 0.01 (1.2) -0.001(-0.12)
1998

Note: z - statistics in parentheses; * denote 5% significance level.

Variances/standard errors of components are computed as in Jann (2005).

Although the mean predictions do not differ significantly across the sub-
samples there is no evidence of the convergence process between high income and
low income Russian regions. On the contrary higher-income regions tend to grow
faster than lower-income regions. Nonetheless, the results in Table 8 are of some
interest as they may help to shed light on the factors that retard convergence of poor
regions to rich regions in Russia.
The difference in structural and control factors reveal that lower initial GRP
per capita in poor regions stimulates convergence of poor regions to rich regions, i.e.
the

Available online at http://eaces.liuc.it

100
EJCE, vol. 5, n. 1 (2008)

initial GRP per capita difference (lnyli , t −1 − lnyhi , t −1 ) , which is obviously negative in the
analyzed period, decreases the gap of GRP per capita growth between rich and poor
regions by 0.25. The result accords with the convergence proposition of neoclassical
growth theory.
Smaller amounts of domestic investment in poor regions in comparison with
rich regions retard convergence as expected, i.e. the decreasing domestic investment
difference (ln(I/N)li ,t − ln(I/N)hi ,t ) , decreases the gap of GRP per capita growth
between rich and poor regions by 0.25. However, the result for FDI is opposite, i.e.
smaller amounts of FDI in poor regions in comparison with rich regions enhance
convergence. This means that the FDI difference (ln(FDI/N)li , t − ln(FDI/N)hi ,t ) ,
which is negative, multiplied by negative coefficient for FDI variable for higher-
ˆ
income regions β `h gives a positive contribution of 0.13 to the gap of GRP per
capita growth between rich and poor regions.
Coefficients decomposition part shows the unexplained difference in growth effects. They
operate mainly via the initial income variable, positively influencing
ˆ ˆ
convergence process, i.e. ( β `l lnX hi , which are negative, decreases the
−β `h ) and
gap
of GRP per capita growth between rich and poor regions by 0.85. 6 The opposite result is
ˆ
obtained for the FDI variable. However the result for FDI variable is not reliable as β FDI
,l is not significant. The interaction decomposition result shows that initial income
variable is the only significant and reliable one. In general we conclude here that
only initial income and domestic investment can be considered as important factors
of convergence across Russian regions.

7. Conclusions and policy implications

We examined empirically determinants of short-run economic growth in
Russian regions in the transition period (1996-2004). We used modified Barro Sala-I-
Martin empirical framework for the neoclassical production function model.
Advanced Arellano-Bond estimation method was applied for dynamic panel data.
The results suggested that initial conditions (conditional convergence), domestic
investment and exports are the most important ones for stimulating economic growth
in Russia. Of the other specified control variables, natural resource availability
surprisingly not contributed significantly to short-run economic growth in Russian
regions, although the Russian economy is traditionally considered to rely heavily on
natural resources. The same result was found when we replaced the natural resources
variable with the oil variable. A possible explanation is that natural resources
(especially oil resources) influence short-run economic growth, not directly, but
through the domestic investment and export variables.
For robust checking purposes we split our sample in two parts along the time
dimension, e.g. 1996-1999 and 2000-2005 to analyze empirically whether the growth
determinants have changed from the “low growth years” to “the high growth years”.
We found that while our specification performs very well for the first period, it
presents

6
Note that ˆ ˆ
(β `l −β `h )lnX hi corresponds to growth differences unexplainable by structural
factors,
i.e difference due to (unobserved) group differences.
Available online at http://eaces.liuc.it
 101

Determinants of Economic Growth: Empirical Evidence from Russian Regions

the second period poorly. In the first period initial income, domestic investment, and
export variables related significantly to economic growth in Russia while for the
second period only export variable was statistically significant and positive. The
explanatory power of the specification falls significantly in the second period as well.
We also divided the sample into two sub-samples along cross-sectional dimension
– high income regions and low income regions. The main difference between the sub
samples’ estimates is that in lower-income regions FDI has positive effect on
economic growth while in high-income regions it has negative effect. Growth
convergence between poor and rich regions in Russia was not found for the period
studied. Some results for the Oaxaca-Blinder (OB) decomposition of growth rate
differences between higher-income and lower-income regions were provided. OB
analysis produced some evidence on the relative magnitudes of different factors of
convergence across Russian regions, e.g. that initial GRP per capita plays a major
role here along with domestic investments. In general we conclude that lower initial
GRP per capita and faster convergence speed contributes positively to the gap of
GRP per capita growth between rich and poor regions while smaller domestic
investment in poor regions contribute negatively to this gap.
Our results in general go along with previous empirical studies on the
determinants of economic growth. Findings of conditional convergence, positive
influence of domestic investment and export are quite usual in the economic growth
literature, both theoretical and empirical. However it is a interesting result that in
recent years the traditional specification performs very poorly and only export
remains statistically significant. This sheds light on possible source of economic
growth in Russia in recent years – oil and gas export. There is already a hot
discussion among economists and politicians that recent signs of economic growth in
oil and gas abundant Russia are mostly due to favourable oil prices in the world
market and therefore oil and gas export. Our results give some support for such a
position.
Our results have some important policy implications. First, export-led growth
seems to be important in Russia during transition. Thus enhancing export Russian
authorities boost the economic growth in the country. However Russian export
mostly consists of oil and gas. This is especially evident in recent years. This option
must be only temporary, and Russian policy-makers must find other inner sources for
economic growth in the country. Second, FDI seems not to be important as a growth
determinant in Russia. This challenges policy makers to improve federal (as well as
regional) strategy towards foreign investors. Finally, in order to enhance economic
growth in poor Russian regions the authorities must stimulate domestic investment in
these regions.
Available online at http://eaces.liuc.it
102
EJCE, vol. 5, n. 1 (2008)

Acknowledgement
The authors would like to thank the editor and two anonymous referees for the
helpful suggestions and comments. Any remaining errors are the responsibility of the
authors.

Appendix 1
The Resource Index was calculated using the following formula of integrated coefficient:
1 m F
j ,it
Resource indexit = ∑ 100*
F
m j =1 jt

.
F
where i=1.….74 in period t=1997.….2003. j ,it is the actual resource indicator j
for a region i in period t. Fjt is the sample mean of the indicator in period t (in our case
n
1
the mean value for Russian regions, which is jt
F
= ∑ Fijt . where n is the number of
n i=1
Russian regions involved in the computation(74)). m is the number of indicators
included in the index computation (adopted from Ndikumana. 2000). Indicators.
included in the computation of the resource index are presented in Table A1.1.

Table A1.1. Indicators included in the Resource Index

Indicator

Electricity production per capita. kilowatt - hour

Oil digging including gas condensate ¨per capita. thousands of tones

Natural gas digging per capita. millions cubic meters

Coal digging per capita. thousands of tones

Black metals production per capita. thousands of tones

Available online at http://eaces.liuc.it

 103
Determinants of Economic Growth: Empirical Evidence from Russian Regions

Appendix 2
Table A2.1. Summary statistics of the dependent and explanatory variables used in baseline estimation:
unbalanced data set for 74 Russian regions along the time period of 1996-2005

Variable Mean Median Minimum Maximum

ln(GRP/N) -6.482 -6.520 -8.609 -3.871
∆ln(GRP/N) 0.063 0.151 -1.421 1.526
ln(I/N) -8.244 -8.272 -10.341 -5.386
ln(FDI/N) -7.987 -5.495 -55.262 1.854
ln(NR/N) -21.407 -5.086 -48.354 3.553
ln(EXP/N) -1.433 -1.274 -6.397 2.105
Variable Std. Dev. C.V. Skewness Ex. kurtosis
ln(GRP/N) 0.629 0.0971 0.472 1.038
∆ln(GRP/N) 0.3689 5.812 0.161 2.339
ln(I/N) 0.7262 0.088 0.428 1.129
ln(FDI/N) 10.774 1.348 -3.948 14.448
ln(R/N) 22.844 1.067 -0.294 -1.852
ln(EXP/N) 1.372 0.957 -0.458 0.217

Appendix 3

Table A3.1. Correlation matrix of variables used in baseline estimation. Unbalanced data set for 74
Russian regions along the time period of 1996-2005

∆ln(GRP/N) Ln(I/N) ln(FDI/N) Ln(NR/N) Ln(EXP/N) D1998

ln(GRP/N) 0.334* 0.888* 0.330* 0.073 0.662* 0.057
∆ln(GRP/N) 1 0.333* 0.025 0.008 0.179* 0.490*
ln(I/N) 1 0.329* 0.092* 0.614* 0.106*
ln(FDI/N) 1 -0.035 0.313* 0.075
ln(R/N) 1 0.102* 0.027
ln(EXP/N) 1 0.082*
* The level of significance is 0.05.

Available online at http://eaces.liuc.it

104
EJCE, vol. 5, n. 1 (2008)

References
Arellano M., Bond S. (1991). ‘Some tests of specification for panel data. Monte Carlo
evidence and an application to employment equations’, Review of Economic Studies, 58,
277– 297
Awokuse T. (2007). ‘Causality between exports. Imports, and economic growth: Evidence
from transition economies’, Economic Letters, 94, 389-395
Baltagi B. (2002), Econometric Analysis of Panel Data, New York, Wiley
Balyasubramanyam N., Salisu M., Sapsford D. (1996), ‘Foreign direct investment and
growth in EP and IS countries’, The Economic Journal, 106, 92-105
Barro R. (1991), ‘Economic growth in a cross section of countries’, Quarterly Journal of
Economics, 106, 407-442
Barro R., Lee J. (1994), ‘Sources of Economic Growth’, Carnegie-Rochester Conference on
Public Policy, 40, 56-79
Barro R., Sala-I-Martin X. (1995), Economic growth, New York, McGrow-Hill. Inc.
Bende-Nabende A., Ford L. (1998), ‘FDI Policy adjustment and endogenous growth:
multiplier effects from a small dynamic model for Taiwan: 1959-1995’, World
Development 26, 7, 1315-1330
Berkowitz D., DeJong D. (2003), ‘Policy reform and growth in post-Soviet Russia’,
European Economic Review, 47, 337 – 352
Blinder A. (1973), ‘Wage discrimination: reduced form and structural estimates’, Journal of
Human Resources, 8, 4, 436-455
Bond S. (2002), ‘Dynamic Panel Data Models: A Guide to Micro Data methods and
Practice’, Cemmap Working paper No 02/9, Department of Economics, Institute for Fiscal
Studies, London, Wired at http://cemmap.ifs.org.uk/wps/cwp0209.pdf
Borenstein E., Gregorio J., Lee J. (1998), ‘How does foreign direct investment affect
economic growth?’, Journal of International Economics, 45, 115-135
Cai F., Wang D., Du Y. (2002), ‘Regional disparity and economic growth in China: The
impact of labor market distortions’, China Economic Review, 13, 197-212
Carcovic M., Levine R. (2002), Does foreign Direct Investment accelerate Economic
growth?, University of Minnesota Working paper
Caselli F., Esquivel G., Lefort F. (1996), ‘Reopening the convergence debate: A new look at
cross-country growth empirics’, Journal of Economic Growth, 1, 1-23
Demurger S. (2001),’Infrastructure Development and Economic Growth: An Explanation for
Regional Disparities in China’, Journal of Comparative Economics, 29, 95-117
Dunning J., Narula R. (1996), Foreign Direct Investment and Government, London, Routledge
Fischer S., Sahay R., Vegh C. (1998), ‘From transition to market: evidence and growth
prospects’, IMF Working Paper 98/52
Gylfason T. (2000),’ Resources. agriculture. and economic growth in economies in
transition’, Kyklos, 53, 545–580
Gylfason T. (2001a), ‘Natural resources. education. and economic development’, European
Economic Review, 45, 847–859
Gylfason T. (2001b), ‘Nature, power and growth’, Scottish Journal of Political Economy, 48,
558–588
Havrylyshyn O. (2001), ‘Recovery and growth in transition: a decade of evidence’, IMF Staff
Papers, 48, 53-87
Available online at http://eaces.liuc.it
 105

Determinants of Economic Growth: Empirical Evidence from Russian Regions

Havrylyshyn O., Izvorski I., van Rooden R. (1998), ‘Recovery and growth in transition
economies 1990-97: a stylized regression analysis’, IMF Working paper 98/141
Havrylyshyn O., van Rooden R. (2000), ‘Institutions matter in transition but so do policies’,
IMF working paper 00/70
Islam N. (1995), ‘Growth empirics: a panel data approach’, Quarterly Journal of Economics,
110, 1127- 1170
Jann B. (2005), Standard Errors for the Blinder–Oaxaca Decomposition, Wired at
http://repec.org/dsug2005/oaxacasehandout.pdf.
Laureti L,. Postiglione P. (2005), ‘The effects of capital inflows on the economic growth in
the MED area’, Journal of Policy Modelling, 27, 839-851
Leite C., Weidmann J., (1999), ‘Does mother nature corrupt? Natural resources corruption
and economic growth’, IMF Working Paper No. 99/85
Li X., Liu X. (2004), ‘Foreign Direct Investment and economic growth: An increasingly
endogenous relationship’, World Development, 33, 3, 393-407
Mankiw N., Romer D., Weil D. (1992),’A contribution to the empirics of economic growth’,
Quarterly Journal of Economics, 107, 407-437
Markusen J., Venables A. (1999), ‘Foreign direct investment as a catalyst for industrial
Development’, European Economic Review, 43, 335-356
Ndikumana L. (2000), ‘Financial determinants of domestic investment in Sub-Saharan Africa:
Evidence from panel data’, World Development, 28, 2, 381-400
Oaxaca L. (1973), ‘Male-Female wage differentials in Urban labour markets’, International
Economic Review, 14, 693-709
Papyrakis E., Gerlagh R. (2004), ‘The resource curse hypothesis and its transmission
channels’, Journal of Comparative Economics, 32, 181–193
Papyrakis E., Gerlagh R. (2007), ‘Resource abundance and economic growth in the United
States’, European Economic Review, 51, 1011-1039
Rodriguez F., Sachs J. (1999),’ Why do resource-abundant economies grow more slowly?’,
Journal of Economic Growth, 4, 277–303
Russia’s regions yearbook (2004), Moscow, Goskomstat
Sachs J., Warner A. (1995), ‘Natural resource abundance and economic growth’, NBER
Working Paper No. 5398, National Bureau of Economic Research, Cambridge, MA
Sachs J., Warner A. (1997), ‘Fundamental sources of long-run growt’, American Economic
Review, 87, 184–188
Sachs J., Warner A. (1999),’ The big push, natural resource booms and growth’, Journal of
Development Economics, 59, 43–76
Soto M. (2000), ‘Capital flows and growth in developing countries: recent empirical
Evidence’, OECD Development Centre Working paper No. 160, 07/2000, Organization for
Economic Cooperation and Development, Paris, Wired at http://miranda.sourceoecd.org
Wei W. (2005), ‘China and India: Any difference in their FDI performance?’, Journal of
Asian Economics, 16, 719-736
Windmeijer F. (2000), ‘A finite sample correction for the variance of linear two-step GMM
estimators’, IFS Working Paper No 19/00, Institute for Fiscal Studies, London
Windmeijer F. (2005), ‘A finite sample correction for the variance of linear two-step GMM
estimators’, Journal of Econometrics, 26, 25-51
Yun M. (2004), ‘Decomposition differences in the first moment’, Economic Letters, 82, 275-280

Available online at http://eaces.liuc.it