Trade, foreign direct investment, and growth: evidence from transition economies.
Nath, Hiranya K.
INTRODUCTION
The experiences of economic transition from a centrally planned to
a market-based system in Central and Eastern Europe (CEE) and the former
Soviet Union raise two pertinent questions. First, what triggered growth
that ended the 'transition recession' experienced by these
transition economies in the early 1990s? Second, what would sustain
growth in subsequent periods? This paper is primarily concerned with the
second question and examines the role of trade and foreign direct
investment (FDI) in growth in 13 transition economies of Central and
Eastern Europe, and the Baltic Region (CEEB). (1) Because these
countries have substantially liberalised international trade and have
attracted large FDI inflows in the last few years, it is important to
examine the significance of these factors in the growth of these
economies.
The volume of trade in these countries has increased: total exports
from and imports into these countries have more than quadrupled between
1990 and 2005. (2) FDI inflows into these 13 countries increased
steadily from less than a billion US dollar (USD) in 1990 to about 43
billion USD in 2005, or from 0.31% to about 6% of real GDP during the
period. There is, however, wide variation across the recipient
countries. For example, Czech Republic, Hungary, and Poland received
about 67% of total FDI inflows into the region. Six countries, Albania,
Estonia, Latvia, Lithuania, Macedonia, and Slovenia, together received
less than 10%.
Figure 1 displays trends in growth of per capita real GDP,
FDI-to-GDP ratio, and volume of trade (exports plus imports) as a share
of real GDP, also used as a measure of trade openness, all averaged
across the cross-section of 13 transition economies and expressed in
percentages between 1991 and 2005. As we can see, the average growth
rate was negative until 1993. Then it fluctuated and has been steadily
rising since 2001. The growth rates have averaged over 5% in the last 3
years of this period. The volume of trade appears to have a clear upward
trend. The FDI share has been increasing steadily with some slow down in
2003.
The transition economies of CEEB experienced a substantial decline
in output in the initial phase of transition, a phenomenon often
referred to as the transition recession. Fischer et al. (1996a) argue
that restrictive macroeconomic policies and restructuring of the economy
caused such decline in economic activities. However, the extent and the
speed of recovery varied across countries. There is a substantial amount
of literature that addresses various aspects of the transition recession
and attempts to identify the factors that triggered the recovery. Some
notable works include de Melo et al. (1996), Fischer et al. (1996a, b),
Sachs (1996), de Melo et al. (1997), Hernandez-Cata (1997), Havrylyshyn
et al. (1998), Berg et al. (1999), Polanec (2004), and Popov (2007).
These studies examine one or more of four different sets of variables to
understand the growth experiences of the early transition years. These
four categories of factors are macroeconomic variables, structural
reform variables, initial conditions, and institutional factors.
[FIGURE 1 OMITTED]
The recovery and growth since the transition recession was over
leave us with only a few data points, not enough to conduct any
meaningful time series analysis of the growth experiences of transition
countries. Pooling time series and cross-section data may, however,
provide a useful way of studying growth in those countries. (3) There
have been attempts in recent years to use a panel data approach to
evaluate the contribution of various factors to growth in transition
economies. For example, in a study very similar in spirit to the current
research, Cernat and Vranceanu (2002) use a panel data analysis of 10
CEE countries to assess the impact of globalisation on output
performance. Their results indicate that increased EU integration and
trade liberalisation are conducive to development. Furthermore,
increased FDI inflows seem to be associated with better output
performance. (4)
In this paper we examine empirically the role of FDI and trade in
the process of economic growth in 13 transition economies of the CEEB
region. The empirical work is motivated primarily by an extension of the
growth theory that includes trade and FDI as additional determinants of
growth. Using fixed effects panel data estimation methods applied to
data from 1991 to 2005, this paper examines the effects of trade and FDI
on growth after controlling for gross domestic investment (GDI) and
other macroeconomic variables such as inflation, fiscal balance, size of
the government, real money growth, the lending rate, and foreign
exchange reserves; and structural variables such as tariff revenue and
infrastructure reform index.
This paper improves upon some previous work on growth in transition
economies by explicitly addressing three methodological issues. First,
in order to deal with the problem of omitted variables, a very general
specification of the model including the largest possible number of
variables is estimated and F-tests are conducted to implement a
'general-to-specific' approach of selecting the most
parsimonious specification. Second, we conduct panel unit root tests to
determine the stochastic trend properties of the variables.
Nonstationary variables are included in the regression equation in their
stationary forms. Furthermore, we formally test for groupwise
heteroscedasticity and cross-sectional correlation. The test results
help us choose the appropriate estimation technique. Third, by including
the lagged dependent variable (LDV), we estimate a dynamic version of
the model to mitigate the problem of serial correlation.
Our analysis suggests that a significant positive effect of trade
on growth is a robust result for transition economies of the CEEB
region. Additionally, domestic investment is an important determinant of
growth. In general, FDI does not appear to have any significant effect
on growth. When we control for the effects of domestic investment and
trade on FDI, however, it is found to have a significant positive effect
on growth, but only after 1995. Among other findings, macroeconomic
stabilisation through fiscal and monetary policies as reflected in
fiscal balance, size of the government, and real money growth plays a
significant role. That the real lending rate turns out to be an
important determinant of growth underlines the importance of the
development of the financial sector in transition economies. These
results have important policy implications.
The rest of the paper is organised as follows. The next section
discusses theoretical background of the linkages between trade, FDI, and
growth. In the subsequent section, we describe the data and the
methodology. The subsequent section presents the empirical results and
analysis. In the last section, we summarise and include a few concluding
remarks.
LINKAGES BETWEEN TRADE, FDI, AND GROWTH: A THEORETICAL BACKGROUND
The importance of trade and FDI for the growth of developing
countries has been emphasised in both theoretical and empirical
literature. Apart from the traditional Ricardian argument of efficiency
gain from specialisation, there have been several other hypotheses put
forward to argue how trade may affect growth in developing countries. In
early works (eg Rosenstein-Rodan, 1943, Nurkse, 1953, Scitovsky, 1954,
Fleming, 1955, Hirschman, 1958), exports were seen as providing the big
push to break away from the vicious cycle of low-level equilibrium in
which developing countries are often caught. Later, exports were thought
to fill in the foreign exchange gap that prevented imports of machinery
needed to be competitive in the market (see McKinnon, 1964). More
recently, Coe and Helpman (1995) argue that trade enhances the spillover effects of foreign R&D on domestic productivity. Another strand of
the recent literature uses new growth theory framework to link trade
policy to growth. Externalities associated with liberal trade policies
are seen as leading to higher levels of GDP or higher growth. (5)
The importance of FDI for growth is emphasised for its role in
augmenting domestic capital stock and as a conduit for technology
transfer, two essential elements in the modern growth literature. (6)
Studies that use the new growth theory paradigm to examine the effects
of FDI on growth take two different routes. For example, extending a
hypothesis advanced by Jagdish Bhagwati (1973), Balasubramanyam et al.
(1996) were able to show that the growth-enhancing effects of FDI were
stronger in countries with a more liberal trade regime. They argue that
a liberal trade regime is likely to provide an appropriate environment
conducive to learning that must go along with the human capital and new
technology infused by FDI. Others (eg Borensztein et al., 1998) rely on
the absorptive capability of the recipient country in the form of stock
of human capital for technological progress that is assumed to take
place through a process of capital deepening in the form of new
varieties of capital goods introduced by FDI.
There are two dimensions to the hypothesis that FDI interacts with
trade to have a positive effect on growth. First, a more liberal trade
environment with export orientation attracts larger FDI inflows because
it not only allows foreign capital to take advantage of low cost of
labour in the host country but also provides access to a larger market.
Second, the neutrality of incentives associated with export orientation
allows exploitation of economies of scale, better capacity utilisation,
and a lower capital-output ratio, thus making foreign capital more
productive. Moreover, exports also promote technical innovation and
dynamic learning from abroad and thereby create a more favourable
environment for externalities and learning from technology spillovers
associated with FDI.
Some of the recent theoretical work (Helpman et al., 2004; Antras
and Helpman, 2004) has explored the relationship between trade and FDI.
Under certain conditions, trade and FDI have been shown to be
substitutes. As this line of research highlights the role of
within-sector productivity differences for determining the patterns of
international trade and FDI, it seems to have implications for growth in
countries receiving the benefits of trade and/or FDI. For the purpose of
our empirical study, the theoretical expositions of the linkages between
trade, FDI, and growth translate into an extended growth equation with
trade and FDI as additional variables alongside domestic investment.
DATA AND METHODOLOGY
Data
The main sources of data for this study are the United
Nations' Statistical Database, the Foreign Direct Investment
Database compiled by the United Nations Conference on Trade and
Development (UNCTAD), and the Transition Reports for various years
prepared by the European Bank for Reconstruction and Development (EBRD).
We obtain national accounts data on GDP per capita, gross fixed
investment, government consumption expenditures, exports and imports of
goods and services from the UN Statistical Database. These data are
available both in national currency and in USD; and both at current
prices and at 1990 constant prices. We use constant 1990 USD data. We
obtain the net FDI inflows data in current USD for CEEB countries from
the UNCTAD. (7) Our sample covers a period from 1990 to 2005. (8)
It may be noted that the national accounts data on the transition
economies have serious problems, which have been emphasised by Fischer
et al. (1996a) and others. The GDP data for years immediately after
transition are likely to overstate the decline of output and the
increases in prices because the pre-transition prices were used to
measure output, which was of extremely poor quality. Moreover,
statistical agencies had been collecting and publishing data on output
mainly from the state sector and, therefore, they may have underreported
the expansion of the private sector during the initial years of
transition.
We construct the following variables for the empirical analysis.
The growth rate of per capita real GDP is calculated as 100 times first
log differences of per capita real GDP and is used as the dependent
variable (GROWTH) in the growth equation. (9) Percentage share of
exports plus imports in GDP is taken as a measure of the trade variable
(TRADE). FDI inflow as a percentage share of GDP (in constant 1990 USD)
is taken as the FDI variable (FDI). Note that FDI current price series
has been converted into constant 1990 USD by using an implicit deflator calculated from the series on gross fixed investment. FDI inflows are
subtracted from gross fixed investment to calculate GDI. The percentage
share of GDI in GDP is taken as the domestic investment variable (GDI).
Additionally, data on CPI inflation, fiscal balance, nominal
exchange rate, employment growth, money growth, domestic credit growth,
lending rate, gross foreign exchange reserves, share of private sector
in GDP, share of industry in employment, tariff revenues, budgetary
subsidies and current transfers, and infrastructure reform index, the
variables that are deemed important for growth, are obtained from
various issues of EBRD's Transition Reports. (10) The Appendix
includes a description of the variables along with availability and
sources of the data.
The summary statistics of the variables of interest (GROWTH, GDI,
FDI, and TRADE) are presented in Table 1. Per capita real GDP in the
CEEB countries grew at an average rate of 1.63% during 1991-2005. The
average growth rate, however, varies widely across countries and so does
its variance over time. Among the CEEB countries, Poland has recorded
the highest average annual rate of per capita real GDP growth, 3.43%,
during this period, followed by Estonia, 3.08%. In Former Yugoslav
Republic (FYR) of Macedonia, the average annual growth rate has been
negative. On an average, these countries have invested 18% of their GDP
in building domestic stock of fixed capital during this period. Seven
countries, Albania, Czech Republic, Lithuania, Poland, Romania, Slovak
Republic, and Slovenia, have exceeded this average. FDI inflows have
accounted for about 4% of real GDP, on average. This share is about 6%
in the Czech Republic and above 8% in Hungary. Average trade volume
among these countries has been about 80% of GDP, with Estonia and Slovak
Republic over 100%. In most countries, the increase in this ratio over
the sample period has been substantial.
Methodology
We use panel data estimation techniques for our empirical analysis.
As discussed above, extension of basic growth theory suggests that
alongside domestic investment, trade, and FDI are important determinants
of growth. We therefore consider GDI, FDI, and TRADE to be the main
right-hand side variables in our growth equation. Although time
invariant initial conditions have been shown to be important for
subsequent growth in general (see, eg, Barro, 1991) and for transition
economies in particular (see de Melo et al., 1997, and Berg et al.,
1999), we leave them out in favour of country-specific fixed effects for
two reasons. First, previous studies (eg Berg et al., 1999) have shown
that more than one initial condition may be important for growth and
macroeconomic performance in transition economies. (11) Inclusion of too
many initial conditions may lead to imprecise estimation of the
coefficients. Moreover, there may be country-specific factors other than
initial conditions that contribute to variations in growth experiences
in transition economies. Therefore, our choice of a fixed effects model
is dictated by the desire for a parsimonious specification and a concern
for the omitted variable problem. Second, the objective of the study is
to examine the contribution of trade and FDI to growth in transition
economies, and the role of initial condition or relative importance of
different initial conditions for growth is not of particular interest.
However, for completeness and to facilitate comparison with previous
studies (eg de Melo et al., 1997; Polanec, 2004), we also examine, as a
part of the sensitivity analysis, whether the results of our empirical
analysis are robust enough to include the initial conditions in the
growth equation.
Although growth theory provides some guidance, growth in countries
that are going through economic and political transition could just be a
black box. Therefore, choosing appropriate control variables is a
difficult task. As shown by previous works, growth in transition
economies may well be affected by, in addition to initial conditions,
macro variables, structural reform variables, and institutional factors.
Based on suggestions from previous works and data availability, we
choose two categories of variables: macroeconomic variables and
structural reform variables. The first category includes CPI inflation
(INF), fiscal balance as percentage of GDP (FBAL), size of the
government as measured by the percentage share of government consumption
expenditures in GDP (GOV), nominal exchange rate (X), employment growth
(EMP), real money growth (MONEY), real domestic credit growth
(DOMCREDIT), real lending rate (LRATE), and gross foreign exchange
reserves as a percentage share of GDP (RES).
These variables either reflect the effects of macroeconomic
stabilisation policies or represent macroeconomic factors that
potentially affect growth. For example, like Berg et al. (1999), we use
inflation as a stabilisation proxy. Fiscal balance is expected to affect
growth through crowding out and government consumption expenditures
through a short-run aggregate demand stimulus. The nominal exchange rate
captures the effect of exchange rate targeting in stabilisation
policies. However, because these countries adopted different exchange
rate regimes and they made changes, some drastic, over the time it is
difficult to speculate on the effects of the exchange rate on growth.
(12) Employment growth is expected to affect growth through augmentation of the labour stock. Real money growth, domestic credit growth, and the
lending rate are assumed to capture real effects of monetary policy and
of developments in the financial sector. Gross foreign exchange reserves
are expected to contribute to growth by alleviating the foreign exchange
constraint for trade and investment.
The category of structural variables includes the share of private
sector in GDP (PVT), tariff revenue as a percentage of total imports
(TARIFF), budgetary subsidies and current transfers as a percentage
share of GDP (SUB), percentage share of industry in total employment
(INDEMP), and infrastructure reform index (INFRA). The first variable is
an indicator of the speed and extent of structural reform and is
expected to have a positive effect on growth through increased
efficiency. TARIFF measures the extent of trade liberalisation.
Budgetary subsidies to enterprises and households are expected to have a
positive effect on growth by encouraging investment, thus galvanising aggregate demand. The share of industry in total employment reflects the
relative size of the labour force engaged in electricity, power,
manufacturing, mining, and water, and a larger share in those crucial
sectors is assumed to contribute positively to growth through structural
change of the economy. The infrastructure reform index is expected to
capture the effects of improvements in transportation, communication,
and power generation on growth. (13) Country-specific fixed effects will
capture some of the important differences in institutions across the
transition economies. (14)
We estimate a pooled time-series cross-section regression of the
following form:
[g.sub.it] = [[mu].sub.i] + [beta]'[X.sub.it] +
[gamma]'[Z.sub.it] + [[epsilon].sub.it]
where [g.sub.it] is the annual growth rate of per capita real GDP
for country i in year t; [[mu].sub.i] is the country-specific fixed
effect; [X.sub.it] is the vector of variables of interest: GDI, FDI, and
TRADE; and [Z.sub.it] is the vector of control variables; i = 1, 2, ...
N indexes country and t = 1, 2, ... T indexes time.
Among various issues and concerns about this empirical methodology,
the following have been formally addressed. First, nonstationarity of
time-series data is often a cause for concern for meaningful analysis of
the data because it may lead to a spurious relationship. The
conventional univariate unit root tests suffer from lack of power when
the length of the sample period is short. The panel unit root tests,
which are relatively new techniques, supposedly alleviate the problem of
lack of power by combining data in time and cross-section dimensions.
We, therefore, conduct panel unit root tests on the variables of
interest as well as on all potential control variables. We use two most
commonly used test procedures suggested by Levin et al. (2002) and Im et
al. (2003), respectively. The first test assumes a common unit root
process for all cross-sectional units whereas the second assumes
different unit root processes for individual cross-sectional units. Both
methods have their advantages and disadvantages (see Baltagi (2002) for
a discussion).
Second, given the differences in growth experiences among
transition economies, one would expect tremendous variation of variables
in the model. Moreover, geographic contiguity, and similarity and links
between erstwhile political systems make it likely that there are some
common factors that affect these countries. We, therefore, formally test
for groupwise heteroscedasticity and cross-sectional correlation.
Following Greene (1997), we conduct simple Lagrange multiplier (LM)
Tests. For serial autocorrelation, however, we rely on pooled
Durbin-Watson (DW) test statistics. These tests also help us determine
the appropriate estimation method.
Third, although country fixed effects take care of time invariant
country-specific factors, the model may still suffer from an omitted
variable problem if some important time-variant control variables are
not included. Moreover, some of these variables may be correlated with
each other. Thus, while exclusion of relevant variables may lead to the
omitted variables problem, inclusion of them may give rise to the
problem of collinearity. To address these problems, we first estimate a
general model including all control variables listed above. The obvious
drawback of including many variables is that, given lack of degrees of
freedom, the coefficients are imprecisely estimated. If some variables
have negligible effects, excluding them would lead to more precise
estimates. Moreover, multicollinearity may show up in terms of
statistically insignificant individual coefficient with high R2.
Remedies of this problem include exclusion of variables that are
collinear with others. We therefore adopt a less stringent application
of Hendry (1995)'s general-to-specific approach. We then apply a
sequence of F-tests to reduce the model to more parsimonious
specifications admissible under our data set. We start with excluding a
single variable under each category of control variables, and then we
test for exclusion of an entire category of variables. This
general-to-specific approach would help us find the most parsimonious
specification of our model. (15)
EMPIRICAL RESULTS
Table 2 presents the results of the panel unit root tests conducted
on growth of per capita real GDP and all other variables that are
potential determinants of growth. Specifying the test equation under
both Levin-Lin-Chu and ImPesaran-Shin test procedures is a formidable
task. Because there are no clearcut guidelines, we conduct these tests
under two specifications: with only individual effects in the test
equation, and with both individual effects and linear time trends. As we
can see from the table, for GDI, FDI, X, DOMCREDIT, RES, and INFRA, the
results are mixed. While we can reject the null of unit root under some
specifications we cannot do so under others. Because in at least two out
of four specifications we do not find them to be unit root processes, we
assume that these variables are stationary. Only for SUB do the results
unequivocally indicate that it is a unit root process. Therefore, we use
the first difference, which is the stationary form, of SUB in our
estimation of the regression equation.
We then estimate a fixed effects panel regression equation for
GROWTH with the three variables of interest, GDI, FDI, and TRADE, and
all other controls as discussed above. The estimation results are used
to conduct tests for groupwise heteroscedasticity, and then for
cross-section correlation. The test results are reported in Table 3.
Although the test strongly rejects the null hypothesis of no
heteroscedasticity across countries, there is little evidence of
cross-section correlation. Based on these results, we decide to use a
feasible generalised least square (FGLS) method with cross-section
weights that mitigate the problems arising from cross-section
heterogeneity. It corrects for cross-sectional heterogeneity by using
estimated cross-section residual variances as weights to transform the
variables.
The test results for the general-to-specific approach of model
selection are presented in Table 4. Based on these results we decide to
include INF, FBAL, GOV, MONEY, LRATE, RES, TARIFF, and INFRA as control
variables in our panel regression model. In Table 4, we also report the
test result that indicates that including the country-specific fixed
effects is appropriate.
In Table 5, we present the regression results. (16) Column 1
includes coefficient estimates along with standard errors and other
relevant statistics estimates from the FGLS method, which we will call
the one-stage/single-stage method in order to distinguish it from its
alternative. Note that the standard errors are estimated using
White's heteroskedasticity consistent variance-covariance estimates
that are robust to general heteroskedasticity. Column 2 presents
estimates obtained from the two-stage estimation process. Intuitively,
GDI, FDI, and TRADE may affect each other. (17) Therefore, we estimate
an equation for each of these three variables, using another two as
regressors in the first stage, obtain the residual and use it as a
regressor in our growth equation in the second stage. For example, we
regress GDI on FDI and TRADE, FDI on GDI and TRADE, and TRADE on GDI and
FDI and extract the residuals, which are then included as explanatory variables in the growth equation. Thus, GDI now represents the residual
variation in domestic investment after controlling for the effects of
FDI and TRADE. Similarly, FDI and TRADE reflect residual variations in
FDI and trade, respectively, after controlling for the effects of the
remaining two variables of interest.
The results indicate that among the variables of interest, trade
has a significant positive effect on per capita real GDP growth, and
this result is robust under alternative estimation methods. The
single-stage FGLS estimate indicates that a 1% point increase in TRADE
increases per capita real GDP growth rate by about 0.06% point, whereas
two-stage estimate indicates a slightly larger effect, 0.068. Domestic
investment also has significant positive effect on the per capita growth
rate. A 1% point increase in GDI leads to about a 0.11% point increase
in per capita GDP growth rate in single-stage estimate, whereas the
effect is larger, 0.141, when the two-stage estimation method is used.
Although the effect of FDI on per capita growth is positive, it is
statistically not significant. It may be noted that GDI and FDI have a
significant negative relationship, as revealed by the first-stage
estimates, which may be suggestive of crowding out as a result of FDI in
transition economies. GDI and TRADE, and TRADE and FDI are found to have
significant positive relationships indicating complementary roles
between them.
Among the control variables, significant positive effects of fiscal
balance and real money growth and significant negative effects of size
of the government and real lending rate are robust across
specifications. The significant effect of fiscal balance accords well
with the previous study by Berg et al. (1999), and highlights the
importance of macroeconomic stabilisation for growth of the transition
economies of the CEEB region. Contrary to our expectation of a positive
effect of GOV through its effect on aggregate demand, the size of the
government has significant negative effect on growth. This may reflect
inefficiency associated with large government. Although inflation
appears to have a negative effect, it is not statistically significant.
The significant positive effect of real money growth may have
highlighted the aggregate demand stimulus of money supply growth.
Furthermore, that real lending rate has significant negative effect
suggests that tighter credit market conditions adversely affect growth.
The significant positive impact of tariff is, however, counterintuitive.
One might suspect that there is collinearity between TRADE and TARIFF,
but exclusion of TRADE does not render the coefficient negative nor
makes it statistically insignificant. The result may just reflect better
enforcement of tariff laws.
We report the pooled DW test statistics for all three methods and
they indicate that the null of no serial correlation is rejected at a 5%
significance level. We therefore estimate a dynamic version of the
equation including the LDV. As LDV is correlated with country-specific
fixed factors, it renders estimates of the coefficients biased and
inconsistent. Note that only if T [right arrow] [infinity], the least
squares estimates will be consistent for the dynamic error panel model.
Some researchers, for example Islam (1995), favour least squares
estimates for moderate size T if N is relatively large, arguing that the
bias may not be large in those cases. (18) The trade coefficient is
statistically significant at a 10% level when single-stage FGLS is used.
The two-stage estimate is, although positive, not statistically
significant. Both GDI and FDI have negative signs under single-stage
FGLS, and neither is statistically significant. Under two-stage
estimation, the coefficient estimate of GDI becomes positive but remains
statistically insignificant. Even long-run effects of these variables,
calculated by multiplying the estimated coefficients by 1/(1-[??]) where
[??] is the estimated coefficient of the LDV, are smaller than those in
the static model. Note that the earlier results about the effects of the
control variables are robust to this dynamic specification of the model
except that GOV and TARIFF are no longer significant. The DW statistics
in the LDV models suggest that the issue of autocorrelation is resolved.
Sensitivity analysis
We conduct three different sensitivity exercises. First, because
the relative importance of initial conditions and reform measures is one
of the central topics in the growth literature on the transition
economies, we will examine whether our results with regard to the
effects of trade, FDI, and GDI on growth hold when we explicitly
introduce initial conditions and reform measures in our regression
models. We experiment with three different sets of initial conditions.
We first use the logarithm of per capita real GDP in 1990 for the
CEEB countries as the initial conditions. Some studies (de Melo et al.,
1996; Fischer et al., 1996a, b) use this variable as the only initial
condition. The neoclassical growth model predicts that countries with
higher initial per capita income will experience slower growth compared
to countries with lower initial per capita income. However, as de Melo
et al. (1997) argue, in addition to initial per capita income, there may
be a host of initial conditions representing initial level of
development, resources and growth, initial economic distortions, and
institutional characteristics that are important for growth in the
transition economies. Following their suggestions, we consider 11
initial conditions: per capita real GDP in 1990 (Y1990); the average
annual growth rate between 1985 and 1989 (PRGR); urbanisation in 1990
(URBAN); a dummy variable for richness in terms of natural resources
(RICH); a categorical variable for whether the country was an
independent state, part of a federal state, or a newly created country
(STATE); black market exchange rate premium (BLCMKT); extent of
overindustrialisation in 1990 (INDIST); a dummy variable for whether the
country is neighbouring a thriving market economy (LOCAT); repressed inflation (REPR); trade dependence (TDEP); and the time under central
planning (MARME). (19) For details on these conditions, see de Melo et
al. (1997). Finally, we also consider a set of eight initial conditions
as suggested by Polanec (2004). In addition to the last six initial
conditions of de Melo et al. (1997), this set also includes price
liberalisation index (PLI) and trade liberalisation index (TLI) in 1990
published by EBRD.
As for reform measures, following Polanec (2004) we use the
year-to-year change in an unweighted average of EBRD transition
indicators (DREFORM). There are eight indicators that cover large- and
small-scale privatisation, enterprise restructuring, price
liberalisation, trade and forex system, competition policy, banking
reform and interest rate liberalisation, securities markets and nonbank
financial institutions, and overall infrastructure reform. The values of
these indicators range from 1 to 4 and are based on subjective judgments
of country economists at the EBRD. Furthermore, as Polanec (2004) points
out, by using these indicators we are assuming that the effects of
reforms on growth are the same at various stages of reform, which may be
highly unlikely.
As discussed by de Melo et al. (1997) and Polanec (2004), these
initial conditions may be highly correlated and, therefore, inclusion of
all these time-invariant conditions may introduce the problem of
multicollinearity. In order to reduce the dimensionality of the set of
initial conditions and to find an appropriate common interpretation, we
resort to the method of principal components. For the set of 11 initial
conditions, the first two components account for about 60% of
variability in initial conditions. The most important cluster has high
positive factor loadings for TDEP, BLCMKT, MARME, URBAN, and REPR, and
has high negative factor loading for STATE. Except for URBAN, this
cluster looks very similar to PRIN1 in de Melo et al. (1997), which they
interpret as a measure of macroeconomic distortions. The second most
important cluster has high positive factor loading for REPR, RICH,
INDIST, BLCMKT, and high negative factor loading for LOCAT and Y1990. In
this case, the similarity with PRIN2 in de Melo et al. (1997) ends in
high positive factor loadings for INDIST. Given that we consider only a
subset of the countries in their sample, this is not surprising. We
include these two clusters (IC_cluster1 and IC_cluster2) in one of the
specifications of the panel regression model with initial conditions.
For the next set of eight initial conditions that we consider, the
first principal component explains about 45 % of variability. The most
important cluster of these conditions, has high positive factor loadings
for BLCMKT, TDEP, MARME, and REPR, and has negative factor loadings for
both PLI and TLI. Although the factor loadings are smaller in value than
those obtained by Polanec (2004), they are qualitatively very similar.
We include this cluster (IC_cluster) along with the initial per capita
real GDP in an alternative specification of our panel regression model.
Note that a table (Table A1) presenting the factor loadings for all the
principal components discussed above is included in the Appendix.
The results from two-stage estimation of these alternative
specifications with initial conditions and reform measures are presented
in Table 6. Columns 1-3 present the results when only the initial
conditions are included in panel regression model. As we can see,
significant negative effect of per capita real GDP in 1990 is a robust
result. IC_cluster1 and IC_cluster2 do not have any significant impact
on growth. When the cluster of eight initial conditions is included
along with 1990 per capita real GDP, however, it appears to have
significant positive effect on growth. (20) The fact that there are some
similarities between IC_cluster1 and IC_cluster in terms of factor
loadings seems to suggest that this positive significant effect may have
been driven by price and trade liberalisation in 1990. In all cases, GDI
has a significant positive effect on growth. The estimated coefficient
of FDI is not statistically significant under any of the specifications.
TRADE has a statistically significant positive effect on growth under
all specifications.
Columns 4-7 of Table 6 present the results for specifications that
include reform measures in addition to initial conditions. As the
infrastructure reform index is part of the overall reform indicator, we
now exclude this variable. Except that IC_cluster1 is now significant at
the 10 % level, the effects of initial conditions remain qualitatively
unaltered. As before, both GDI and TRADE have positive effects on
growth. The estimated coefficients are, however, not statistically
significant when Y1990 is included as the only initial condition. FDI
does not seem to matter for growth. The reforms variable has a
significant negative effect. Although it seems to suggest that an
increase in the change in reform measures, that is, an acceleration in
reform, hurts growth, a plausible interpretation of this result is
difficult to obtain without further scrutiny. Thus, positive, and often
significant, effects of trade and domestic investment, and insignificant
effects of FDI on growth are robust to the inclusion of initial
conditions.
Second, we exclude those years when most transition economies in
the CEEB region experienced negative growth. By 1995 the transition
recession largely ended in the region except in Macedonia. Therefore, we
re-estimate the model for the period 1995-2005. The results are
presented in columns 2-3 of Table 7. As we can see, the effect of GDI is
similar in magnitude as before, though it is now significant at the 5%
level. TRADE is significant at a 1% level and the magnitude of its
effect is larger. Under the two-stage estimation, these effects are even
larger in magnitude and stronger in statistical significance. The most
interesting result is that although the FDI coefficient is positive and
not statistically significant under the single-stage FGLS method, it is
not only positive but also highly significant under the two-stage
estimation method. The effects of fiscal balance, size of the
government, real money growth, and real lending rate are still
significant and have the same signs as before. However, foreign exchange
reserves now have a significant positive effect but the effect of tariff
is no longer statistically significant. Infrastructure, on the other
hand, has a statistically significant negative effect, which is
puzzling. It may be correlated with one of the variables of interest.
Only an estimation of the model without TRADE makes the estimated
coefficient of INFRA positive, though statistically insignificant. Thus,
trade and infrastructure index may be correlated.
Third, since N is small in our case, we estimate the dynamic
version of the model using the generalised method of moments as
suggested by Arellano and Bond (1991). This method exploits the
orthogonality conditions that exist between lagged values of the
dependent variable and the disturbances to introduce the lagged values
as instruments. We estimate the model in differences, with lags of the
dependent variable from lag 2 and above, and all explanatory variables
as instruments. (21) The results are reported in columns 3-4 of Table 7.
The trade coefficient is statistically significant at a 10 % level of
significance under the single-stage method, and the magnitude of the
estimated coefficient is comparable to the one under single-stage FGLS.
Estimated coefficients for both GDI and FDI are negative but not
statistically significant. The estimated coefficients of the control
variables have the same signs as the FGLS estimates except for RES,
which now becomes negative but statistically not significant. Also, only
for fiscal balance and real lending rate are the coefficients
statistically significant. The AR(1) coefficient is positive and
statistically significant at the 10% level. (22)
To summarise, our results indicate that the significant positive
effect of trade on growth is a robust empirical result for transition
economies of the CEEB region. Domestic investment, too, appears to be an
important determinant of growth. FDI does not have any significant
effect on growth when we consider the entire sample period. However,
when we control for the effects of domestic investment and trade on FDI,
it appears to have significant positive effects on growth after 1995.
Among other findings, macroeconomic stabilisation through fiscal and
monetary policies as reflected in fiscal balance, size of the
government, and real money growth play a significant role. That real
lending rate turns out to be an important determinant of growth
underlines the importance of the development of the financial sector in
transition economies.
These results have important policy implications for the transition
economies of the CEEB region. They are even more significant as most of
these countries have recently joined the European Union. With free
mobility of factors of production and liberal trade policies these
countries are expected to achieve high growth.
CONCLUDING REMARKS
This paper examines the effects of trade and FDI on growth using
data for 13 transition economies in the CEEB region. An extension of
traditional growth theory that includes trade and FDI as additional
determinants of growth provides the motivation for this study, which
tries to understand growth and its sustainability in the transition
economies. The transition countries of the CEEB region have witnessed a
substantial increase in trade and FDI during the first decade of their
transition from plan to market. Applying fixed effects panel estimation
methods to a data set for 1991-2005, this paper finds that a significant
positive effect of trade on growth is a robust result for these
transition economies. Domestic investment appears to be an important
determinant of growth. In general, FDI does not have any significant
impact on growth in transition economies. However, when we control for
the effects of domestic investment and trade on FDI, it appears to be a
significant determinant of growth for the period after 1995. Among other
findings, macroeconomic stabilisation through fiscal and monetary
policies as reflected in fiscal balance, size of the government, and
real money growth play a significant role. That real lending rate turns
out to be an important determinant of growth underlines the importance
of the development of the financial sector in transition economies.
APPENDIX: DESCRIPTION OF THE VARIABLES AND DATA
GROWTH: Growth rate of per capita real GDP. Calculated as 100 times
first log differences of per capita real GDP (1990 USD), available from
UN Statistical Database. Available for the entire period 1991-2005 for
all countries.
GDI: Real gross domestic investment as a percentage of real GDP.
Data on real gross fixed capital formation are obtained from the UN
Statistical Database and real net FDI inflows are subtracted and the
percentage shares in GDP are calculated. Data are available for all
years: 1991-2005 for all countries.
FDI: Real foreign direct investment as a percentage of real GDP.
Data on net FDI inflows at current USD are obtained from the UNCTAD.
They are converted into constant dollars by applying an implicit
deflator for gross fixed investment. Percentage shares in real GDP are
then calculated. Data are available since 1992 for Albania, Croatia,
Estonia, Latvia, Lithuania, Macedonia, and Slovenia. For Czech Republic
and Slovak Republic, data are available only since 1993.
INF: CPI inflation. Percentage change in annual average consumer
price index (CPI). Available from the EBRD's Transition Reports.
For Estonia, Latvia, Lithuania, and Macedonia, data are available only
from 1992.
FBAL: Fiscal balance. Government budget balance as a percentage of
GDP. Available from the EBRD's Transition Reports. For Estonia and
Latvia, data are available only from 1994; for Lithuania from 1993; and
for Macedonia, Romania, and Slovak Republic from 1992.
GOV: Size of the government. Real government consumption
expenditures as a percentage share in real GDP (1990 USD). Calculated
from the data obtained from the UN Statistical Database. Data are
available for all years: 1991-2005 for all countries.
X: Nominal exchange rate. Natural log of nominal exchange rate as
reported by the EBRD's Transition Reports. Data are available for
all years: 1991-2005 for all countries.
EMP: Employment growth. Percentage growth of employment as reported
in the EBRD's Transition Reports. Data are available for Estonia,
Latvia, Lithuania, and Macedonia since 1992.
MONEY: Real money growth. Percentage changes in broad measures of
money (M2) are available from the EBRD's Transition reports. The
real money growth rate is calculated by subtracting CPI inflation. Data
are available for Bulgaria, Lithuania, and Slovenia from 1992; for
Albania, Czech republic, and Slovak republic from 1993; for Croatia,
Estonia, and Latvia from 1994; and for Macedonia from 1996.
DOMCREDIT: Real domestic credit growth. Percentage changes in
outstanding bank credits available from the EBRD's Transition
Reports. Data are available for Bulgaria and Slovenia from 1992; for
Croatia, Czech republic, Latvia, Macedonia, and Slovak Republic from
1994; for Estonia from 1995; and for Lithuania from 1996.
LRATE: Real lending rate. Data on nominal lending rates (per annum)
are available from the EBRD's Transition Reports. We subtract CPI
inflation to obtain real lending rates. Data are available for Croatia,
Czech republic, Lithuania, Macedonia, and Slovenia from 1992; for
Latvia, Romania, and Slovak Republic from 1993; and for Estonia from
1994.
RES: Gross foreign exchange reserves as a percentage share of GDP.
Gross foreign exchange reserves excluding gold as a percentage of GDP
are available from the EBRD's Transition Reports. Data are
available for Croatia and Lithuania from 1992; for Estonia, Latvia,
Romania, and Slovak Republic from 1993; and for Macedonia from 1995.
PVT: Share of private sector in GDP. Private sector value-added as
a percentage of GDP, available from the EBRD's Transition Reports.
Available for all countries for the entire period: 1991-2005.
TARIFF: Tariff revenue as a percentage of total imports. All
revenues from international trade as a percentage of value of imports of
merchandise goods. Reported in the EBRD's Transition Reports. Data
are available for Albania, Poland, and Slovak republic from 1992; and
for Czech Republic, Estonia, Latvia, and Lithuania from 1993.
SUB: Budgetary subsidies and current transfers as a percentage
share of GDP. Budgetary transfers to enterprises and households,
excluding social transfers. Reported in EBRD's Transition Reports.
Data are available for Albania and Slovak republic from 1992; for the
Czech Republic and Lithuania from 1993; for Estonia and Latvia from
1994; for Croatia from 1996; and for Macedonia from 1997.
INDEMP: Percentage share of industry in total employment. Share of
employment in electricity, power, manufacturing, mining and water in
total employment in the economy. Reported in the EBRD's Transition
Reports. Data are available for Latvia, Lithuania, and Macedonia from
1992; for Slovenia from 1993; and for Albania from 1994.
INFRA: Infrastructure reform index. EBRD index of infrastructure
reform that covers electric power, railways, roads, telecommunications,
and water, and waste water reforms. Reported in the EBRD's
Transition Reports. Available for the entire period 1991-2005 for all
countries.
Table A1: Factor loadings of the first two principal components of 11
initial conditions and of the first principal component of eight
initial conditions
Initial Principal Principal Principal
condition component 1 component 2 component 1
variables (11 initial (11 initial (8 initial
conditions) conditions) conditions)
1 2 3
Y1990 0.234 -0.237 --
STATE -0.359 0.182 --
PRGR 0.240 0.011 --
RICH -0.193 0.438 --
TDEP 0.438 0.111 0.465
BLCMKT 0.402 0.303 0.511
INDIST -0.056 0.384 0.036
URBAN 0.318 0.032 --
LOCAT 0.162 -0.523 0.045
MARME 0.383 -0.010 0.437
REPR 0.301 0.439 0.407
PLI -0.218
TLT -0.339
Note: The eigenvalues corresponding to the first two principal
components of 11 initial conditions are 4.574 and 1.941, respectively,
and these two components together explain 59.23% of variability. The
eigenvalue corresponding to the first principal component of eight
initial conditions is 3.562 and this component explains 44.52% of the
variability.
Acknowledgements
Earlier versions of this paper were presented at the 51st Annual
North American Meetings of the Regional Science Association
International in Seattle, 11-13 November 2004 and at the 80th Annual
Conference of the Western Economic Association International in San
Francisco, 4-8 July 2005. I thank Josef Brada, Don Bumpass, Don Freeman,
loan Voicu, and two anonymous referees for their insightful comments,
and to Gabi Eissa, Dhrubajyoti Nath, and Rhitwik Patowary for their
excellent research assistance. Usual disclaimer applies.
REFERENCES
Anderson, TW and Hsiao, C. 1981: Estimation of dynamic models with
error components. Journal of the American Statistical Association 76:
598-606.
Antras, P and Helpman, E. 2004: Global sourcing. Journal of
Political Economy 112(3): 552-580.
Arellano, M. 1989: A note on the Anderson-Hsiao estimator for panel
data. Economics Letters 31: 337-341.
Arellano, M and 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.
Balasubramanyam, VN, Salisu, M and Sapsford, D. 1996: Foreign
direct investment and growth in EP and IS countries. Economic Journal
106: 92-105.
Baltagi, BH. 2002: Econometric analysis of panel data, 2nd Edition.
John Wiley & Sons Ltd.: Chichester, UK.
Barro, RJ. 1991: Economic growth in a cross section of countries.
Quarterly Journal of Economics 106(2): 407-443.
Berg, A, Borensztein, E, Sahay, R and Zettelmeyer, J. 1999: The
evolution of output in transition economies: Explaining the differences.
IMF Working Paper (WP/99/73). International Monetary Fund: Washington,
DC.
Bhagwati, JN. 1973: The theory of immiserizing growth: further
application. In: Connolly, MB and Swoboda, AK (eds) International Trade
and Money. University of Toronto Press: Toronto.
Borensztein, E, De Gregorio, J and Lee, J-W. 1998: How does foreign
direct investment affect economic growth? Journal of International
Economics 45: 115-135.
Campos, NF and Kinoshita, Y. 2003: Why does FDI go where it goes?
New evidence from transition economies. IMF Working Paper (WP/03/228).
International Monetary Fund: Washington, DC.
Cernat, L and Vranceanu, R. 2002: Globalisation and development:
New evidence from central and Eastern Europe. Comparative Economic
Studies XLIV(4): 119-136.
Coe, DT and Helpman, E. 1995: International R&D spillovers.
European Economic Review 39: 859-887.
de Melo, M, Denizer, C and Gelb, A. 1996: From plan to market:
Patterns of transition. World Bank Policy Research Working Paper 1564.
World Bank: Washington, DC.
de Melo, M, Denizer, C, Gelb, A and Tenev, S. 1997: Circumstance
and choice: The role of initial conditions and policies in transition
economies. World Bank Policy Research Working Paper 1866. World Bank:
Washington, DC.
European Bank for Reconstruction and Development (EBRD) Transition
Reports. Various issues between 1994-2006. EBRD: London.
Fischer, S, Sahay, R and Vegh, C. 1996a: Stabilization and growth
in transition economies: The early experience. Journal of Economic
Perspective 10(2):45-66.
Fischer, S, Sahay, R and Vegh, C. 1996b: Economies in transition:
The beginnings of growth. American Economic Review 86(2); Papers and
Proceedings of the Hundredth and Eighth Annual Meeting of the American
Economic Association San Francisco, CA, January 5-7, pp. 229-233.
Fleming, JM. 1955: External economies and the doctrine of balanced
growth. Economic Journal 65: 241-256.
Greene, WH. 1997: Econometric analysis, 3rd Edition. Prentice-Hall:
New Jersey.
Grogan, L and Moers, L. 2001: Growth empirics with institutional
measures for transition countries. Economic Systems 25: 323-344.
Grossman, GM and Helpman, E. 1992: Innovation and growth:
Technological competition in the global economy. MIT Press: Boston, MA.
Havrylyshyn, O, Izvorski, I and Rooden, R van. 1998: Recovery and
growth in transition economies 1990-97: A stylized regression analysis.
IMF Working Paper (WP/98/141). International Monetary Fund: Washington,
DC.
Helpman, E, Melitz, MJ and Yeaple, SR. 2004: Export versus FDI with
heterogeneous firms. American Economic Review 94(1): 300-316.
Hendry, DF. 1995: Dynamic econometrics. Oxford University Press:
Oxford, New York. Hernandez-Cata, E. 1997: Growth and liberalization during transition from plan to market. IMF Staff Papers 44(4): 405-429.
Heston, A, Summers, R and Aten, B. 2006: Penn world table Version
6.2. Center for International Comparisons of Production, Income and
Prices at the University of Pennsylvania: Philadelphia, PA.
Hirschman, AO. 1958: The Strategy of Economic Development. Yale
University Press: New Haven, CT.
Im, KS, Pesaran, MH and Shin, Y. 2003: Testing for unit roots in
heterogeneous panels. Journal of Econometrics 115: 53-74.
Islam, N. 1995: Growth empirics: A panel data approach. Quarterly
Journal of Economics 110(4): 1127-1170.
Kutan, AM and Vuksic, G. 2007: Foreign direct investment and export
performance: Empirical evidence. Comparative Economic Studies 49:
430-445.
Lee, M and Tcha, MJ. 2004: The color of money: The effects of
foreign direct investment on economic growth in transition economies.
Review of World Economies 140(2): 211-229.
Levin, A, Lin, CF and Chu, C. 2002: Unit root tests in panel data:
Asymptotic and finite-sample properties. Journal of Econometrics 108:
1-24.
McKinnon, RI. 1964: Foreign exchange constraints in economic
development and efficient aid allocation. Economic Journal 74: 388-409.
Nurkse, R. 1953: Problems of capital formation in underdeveloped countries. Oxford University Press: Oxford.
Polanec, S. 2004: Convergence at last? Evidence from transition
countries. Eastern European Economics 42(4): 55-80.
Popov, V. 2007: Shock therapy versus gradualism reconsidered:
Lessons from transition economies after 15 years of reforms. Comparative
Economic Studies 49: 1-31.
Rosenstein-Rodan, PN. 1943: Problems of industrialization of
Eastern and South-Eastern Europe. Economics Journal 53: 202-211.
Sachs, JD. 1996: The transition at mid decade. American Economic
Review 86(2); Papers and Proceedings of the Hundredth and Eighth Annual
Meeting of the American Economic Association San Francisco, CA, 5-7
January, pp. 229-233.
Saggi, K. 2002: Trade, foreign direct investment and international
technology transfer: A survey. World Bank Research Observer 17(2):
191-235.
Scitovsky, T. 1954: Two concepts of external economies. Journal of
Political Economy 62: 151-153.
Sohinger, J. 2005: Growth and convergence in European transition
economies: The impact of foreign direct investment. Eastern European
Economics 43(2): 73-94.
UNCTAD. Statistics on foreign direct investment
http://stats.unctad.org/fdi/eng/ReportFolders/ Review/explorerp.asp.
United Nations. UN statistical database
http://unstats.un.org/unsd/snaama/selectionbasicFast.asp.
(1) These countries are Albania, Bulgaria, Croatia, Czech Republic,
Estonia, Hungary, Latvia, Lithuania, Macedonia FYR, Poland, Romania,
Slovak Republic, and Slovenia. Ideally one would like to include all
transition economies in this investigation. But for some of the
countries in the former Soviet Union, reliable data are not available
for a significant part of the sample period considered in this paper.
(2) The numbers discussed and reported in this paragraph and the
next are based on the author's calculation.
(3) To our knowledge, Islam (1995) is the first study to implement
panel data approach to cross-country growth data.
(4) In a related study, drawing on the insights provided by a
production function with a low elasticity of substitution between
capital and labour, for short-run growth dynamics in the transition
economies, Lee and Tcha (2004) empirically show that the marginal
contribution of FDI to growth is greater than that of domestic
investment. In another study, Sohinger (2005) shows, in a less formal
way, that FDI, with its growth-enhancing effects, has played a
significant role in setting the transition economies in the CEEB region
onto the path of convergence with their more affluent neighbours.
(5) See Grossman and Helpman (1992) for a comprehensive discussion
of a class of such models.
(6) In the literature, the role of FDI in transferring technology
has received much attention and spurred intense debate. For a recent
survey, see Saggi (2002).
(7) FDI inflows in the recipient economy 'comprise capital
provided (either directly or through other related enterprises) by a
foreign direct investor to an enterprise resident in the economy. FDI
flows are recorded on a net basis (capital account credits less debits
between direct investors and their foreign affiliates) in a particular
year' (UNCTAD).
(8) Although transition began in 1989 in most countries, data are
either not available or too noisy for this initial year of the process.
When we calculate growth rates of per capita real GDP, we lose one
year's data. Therefore, we use the sample period 1991-2005 in our
estimation.
(9) There have been studies that use per capita real GDP, mostly in
logarithms, as the dependent variable. For example, see Berg et al.
(1999) and Cernat and Vranceanu (2002). Polanec (2004) argues in favour
of using growth rate of average labour productivity. There are others
(eg Fischer et al., 1996a, b; Sachs, 1996; de Melo et al., 1997) who use
the growth rate of aggregate real GDP as the dependent variable. There
are some concerns, however, about the use of the growth rate of per
capita real GDP measured in 1990 constant USD. For example, the
differences in the movements of exchange rates over time across
countries may introduce some systematic bias in the estimation of the
coefficients in our regression model when we use the growth rate of per
capita real GDP measured in constant USD instead of constant national
currency as the dependent variable. Furthermore, because of the
differences in domestic prices across countries, there may have been
important differences even with growth rate of per capita real GDP
measured in international purchasing power parity (PPP) dollar. Ideally,
one would like to use per capita real GDP in PPP dollar, but data for
all relevant variables measured in PPP dollars and for the sample period
considered are not readily available. To have a sense of the extent of
possible biases, we examine growth of per capita real GDP in 1990
constant national currency and growth of per capita real GDP in
international PPP dollar. We observe that growth rates of per capita
real GDP in both US dollars and in national currency track each other
very closely, and for most countries they are perfectly correlated. The
correlation coefficient is the lowest with a value of 0.98 for Bulgaria.
Thus, the bias introduced by differences in movements of exchange rate
should be negligible. We further obtain growth rates of per capita real
GDP measured in PPP dollar from Heston et al. (2006) for 1991-2004 (for
some countries data are not available for all the years), plot them
alongside growth rates of per capita real GDP in 1990 USD, and calculate
the correlation coefficients. The correlation ranges between 0.825 (for
Bulgaria) and 0.992 (for Croatia). For Albania, Bulgaria, Hungary, and
Lithuania the correlation coefficients are less than 0.90. Most of the
deviations in these two sets of growth rates are in the early years of
transition. Interested readers can obtain the data and graphs from the
author. As mentioned above, these deviations are likely to introduce
some biases in coefficient estimates. However, the results do not seem
to change qualitatively. The growth rates of per capita real GDP are
almost perfectly correlated with the growth rates of aggregate real GDP
for all countries.
(10) Data on other variables that may affect growth are also
reported in the Transition Reports. They are not included in our set of
potentially relevant variables for one of two reasons: (i) incomplete
data with data missing for a significant part of our sample period; (ii)
they represent the same aspects of the economy as the ones that are
included.
(11) However, they have argued that the effects of these initial
conditions taper off as time passes. This is another reason why they may
be excluded in investigating growth over an extended period of time.
(12) For a discussion on exchange rate regime, stabilisation, and
growth in transition economies, see Fischer et al. (1996a).
(13) Intuitively some of the variables are expected to affect one
another and, therefore, to be correlated. Our general-to-specific
approach of model selection should eliminate the possible collinearity
among the variables.
(14) Grogan and Moers (2001) present a cross-section analysis of 25
transition economies to show that institutions are important for growth
and FDI.
(15) Note that we do not apply the general-to-specific approach to
our variables of interest. Therefore, even in the most parsimonious
specification, a multicollinearity problem may arise if two or more of
these variables are collinear.
(16) From the data description in the Appendix, it is clear that we
have missing data for some variables used in this study. That is, we
have to use an unbalanced sample. In the case of missing values for the
variables, we use the largest sample possible in each cross-section. An
observation will be excluded from the estimation of our regression model
if any of the explanatory or dependent variables for that cross-section
are unavailable in that period.
(17) There is some evidence of mutual relationship among GDI, FDI,
and TRADE. For example, Campos and Kinoshita (2003) find that trade has
a positive effect on FDI in transition economies. Kutan and Vuksic
(2007) further investigate the effects of FDI on the export performance
of 12 CEE countries and find that while FDI has increased exports by
increasing supply capacity through augmentation of the physical capital
stock in all countries in their sample, it has helped exports through
FDI-specific effects such as technology transfer, higher productivity,
and information about export markets, only among the new members of the
European Union.
(18) See Baltagi (2002 pp. 129-30) for a discussion. Many
alternatives for getting around the problems associated with dynamic
specification of fixed effects model have been suggested. Notable works
include Anderson and Hsiao (1981), Arellano (1989), and Arellano and
Bond (1991).
(19) Except for initial per capita real GDP, for other initial
condition variables we use the same acronyms as de Melo et al. (1997).
(20) An experiment with the sample period 1991-1995 reveals that
IC_cluster1 has significant negative, IC_cluster2 has significant
positive, and IC_cluster has significant negative effects on growth.
Although the result for IC_cluster1 accords well with de Melo et al.
(1997), the result for IC_cluster2 does not seem to conform to these
results. This may be due to some important interactions the IC-cluster2
may have with the control variables included in our regression model.
The result for the IC_cluster, on the other hand, accords well with the
result of Polanec (2004) for the period 1990-1994. The estimated
coefficient of Y1990 is negative and statistically significant under all
specifications and estimation methods. We do not report the results to
save space. Interested readers can obtain the results of this
experiment, and also the one-stage estimation results for specifications
in Table 6 from the author.
(21) Under the assumption that the regressors are strictly
exogenous. See Baltagi (2002, pp. 139).
(22) We do conduct some additional experiments, the results of
which are not reported. For example, although we choose our model based
on formal tests for exclusion of control variables, some of the included
variables may intuitively affect each other or may affect one of our
variables of interest. For example, real money growth may have an effect
on real lending rate through its effects in the money market. Similarly,
tariff that reflects trade liberalisation policy may have an impact on
the trade variable. Gross reserves may also affect trade by alleviating
the foreign exchange constraint. Size of the government may have a
negative impact on fiscal balance. Finally, infrastructure may have a
positive impact on FDI and trade. Therefore, in a series of experiments,
we exclude one control variable at a time from the benchmark equation
and re-estimate the model. The main conclusions are that GDI and TRADE
are significant determinants of growth. Under none of these alternative
specifications, the estimated coefficient for FDI turns out to be
significant when full sample is used. Because the main results do not
change, we do not report the results, and thereby save space.
HIRANYA K. NATH
Department of Economics and International Business, Sam Houston
State University, Huntsville, TX 77341-2118, USA. E-mail:
eco_hkn@shsu.edu
Table 1: Summary statistics of the variables of interest: 1991-2005
Country Per capita real GDI-to-GDP
GDP growth rate ratio
1 2
Albania
Mean 2.94 18.72
Standard dev 11.62 7.26
(Max, Min) (12.69, -32.23) (29.21, 4.37)
Bulgaria
Mean 1.31 10.06
Standard dev 5.51 2.71
(Max, Min) (6.20, -8.86) (15.66, 4.94)
Croatia
Mean 0.36 16.97
Standard dev 8.96 3.3
(Max, Min) (7.35, -24.37) (22.97, 12.83)
Czech Republic
Mean 1.43 22.55
Standard dev 4.39 3.46
(Max, Min) (5.93, -12.37) (28.55, 17.08)
Estonia
Mean 3.08 14.4
Standard dev 8.75 2.67
(Max, Min) (11.70, -22.16) (19.63, 7.73)
Hungary
Mean 2.11 17.18
Standard dev 4.68 5.16
(Max, Min) (5.33, -12.48) (25.21, 7.05)
Latvia
Mean 0.68 14.53
Standard dev 13.69 6.28
(Max, Min) (10.66, -41.61) (26.40, 7.09)
Lithuania
Mean 0.71 18.3
Standard dev 10.34 2.95
(Max, Min) (10.21, -23.68) (22.52, 14.19)
Macedonia, FYR
Mean -0.04 14.00
Standard dev 3.88 3.97
(Max, Min) (4.08, -9.10) (19.23, 1.80)
Poland
Mean 3.43 22.29
Standard dev 3.5 2.96
(Max, Min) (6.90, -7.63) (27.10, 17.86)
Romania
Mean 1.18 20.64
Standard dev 6.51 2.34
(Max, Min) (8.57, -13.68) (23.76, 15.24)
Slovak Republic
Mean 1.94 21.68
Standard dev 5.98 4.86
(Max, Min) (5.93, -16.16) (28.20, 9.56)
Slovenia
Mean 2.12 23.20
Standard dev 4.20 4.37
(Max, Min) (5.11, -9.74) (29.32, 16.000)
Full sample
Mean 1.63 18.04
Standard dev 7.58 5.77
(Max, Min) (12.69, -41.61) (29.32, 1.80)
Country FDI-to-GDP Trade-to-GDP
ratio ratio
3 4
Albania
Mean 2.14 42.74
Standard dev 0.87 15.78
(Max, Min) (3.77, 0.69) (95.30, 34.70)
Bulgaria
Mean 4.65 86.52
Standard dev 4.19 10.57
(Max, Min) (14.56, 0.31) (108.40, 68.80)
Croatia
Mean 3.6 65.44
Standard dev 2.32 8.11
(Max, Min) (7.17, 0.13) (78.20, 39.20)
Czech Republic
Mean 5.95 94.17
Standard dev 3.74 18.39
(Max, Min) (12.35, 1.51) (124.80, 63.10)
Estonia
Mean 4.94 112.71
Standard dev 3.47 9.92
(Max, Min) (15.54, 1.21) (131.30, 93.10)
Hungary
Mean 8.17 90.18
Standard dev 2.61 28.28
(Max, Min) (14.38, 3.29) (128.80, 45.40)
Latvia
Mean 3.46 72.91
Standard dev 1.81 8.1
(Max, Min) (7.14, 0.49) (87.30, 58.00)
Lithuania
Mean 2.82 96.45
Standard dev 2.13 19.43
(Max, Min) (7.77, 0.12) (157.70, 71.00)
Macedonia, FYR
Mean 2.13 80.16
Standard dev 2.91 11.7
(Max, Min) (11.64, 0.00) (103.80, 58.80)
Poland
Mean 3.76 47.15
Standard dev 1.94 9.9
(Max, Min) (7.39, 0.38) (67.10, 34.50)
Romania
Mean 3.63 54.41
Standard dev 3.14 10.79
(Max, Min) (11.32, 0.15) (71.10, 33.40)
Slovak Republic
Mean 3.76 112.56
Standard dev 3.83 18.19
(Max, Min) (14.84, 0.92) (140.10, 86.10)
Slovenia
Mean 1.86 90.83
Standard dev 2.1 11.71
(Max, Min) (9.03, 0.56) (108.70, 63.10)
63.1095)
Full sample
Mean 3.93 80.43
Standard dev 3.31 26.56
(Max, Min) (15.54, 0.00) (157.70, 33.40)
Table 2: Panel unit test results
Variables Levin-Lin-Chu Test
Only individual Individual effects
effects in test and linear trends
equation in test equation
1 2
Growth of per capita -11.52 *** -4.91 ***
real GDP(GROWTH)
GDI-to-GDP ratio (GDI) 0.31 -2.60 ***
FDI-to-GDP ratio (FDI) -2.24 ** -2.90 ***
(Exports+Imports)-to-GDP -1.84 ** -14.03 ***
ratio (TRADE)
Inflation (INF) -77.01 *** -154.97 ***
Fiscal balance (FBAL) -4.20 *** -7.88 ***
Size of the government (GOV) -4.42 *** -2.73 ***
Nominal exchange rate (X) -22.96 *** 8.19
Employment growth (EMP) -4.95 *** -5.00 ***
Real money growth (MONEY) -66.01 *** -0.90
Real domestic credit 3.89 -51.88 ***
growth (DOMCREDIT)
Real lending rate (LRATE) -216.00 *** -231.52 ***
Gross reserves-to-GDP -2.00 ** -2.66 ***
ratio(RES)
Tariff revenue-to- imports -5.72 *** -8.33 ***
ratio (TARIFF)
Share of industry in total -4.12 *** -5.68 ***
employment (INDEMP)
Share of private sector -13.84 *** -3.01 ***
in GDP (PVT)
Budgetary subsidies- to- 0.21 1.10
GDP ratio (SUB)
Infrastructure reform -7.20 *** -1.43 *
index (INFRA)
Variables Im-Pesaran-Shin Test
Only individual Individual effects
effects in test and linear trends
equation in test equation
3 4
Growth of per capita -8.29 *** -3.58 ***
real GDP(GROWTH)
GDI-to-GDP ratio (GDI) -1.70 ** -3.45 ***
FDI-to-GDP ratio (FDI) -0.88 -1.58 *
(Exports+Imports)-to-GDP -2.54 *** -10.30 ***
ratio (TRADE)
Inflation (INF) -57.71 *** -105.69 ***
Fiscal balance (FBAL) -3.62 *** -4.18 ***
Size of the government (GOV) -3.73 *** -5.05 ***
Nominal exchange rate (X) -25.98 *** -0.69
Employment growth (EMP) -2.82 *** -3.39 ***
Real money growth (MONEY) -22.75 *** -9.85 ***
Real domestic credit -6.16 *** -17.26 ***
growth (DOMCREDIT)
Real lending rate (LRATE) -130.19 *** -95.22 ***
Gross reserves-to-GDP 0.68 -1.81 **
ratio(RES)
Tariff revenue-to- imports -3.90 *** -4.26 ***
ratio (TARIFF)
Share of industry in total -1.96 ** -2.44 ***
employment (INDEMP)
Share of private sector -10.30 *** 2.59
in GDP (PVT)
Budgetary subsidies- to- 0.48 0.47
GDP ratio (SUB)
Infrastructure reform -1.14 0.3
index (INFRA)
Note: For each test, the lag lengths are chosen on the basis of
Schwarz Information Criterion (SIC). *** Significant at 1% level;
** significant at 5% level; * significant at 10% level.
Table 3: LM tests for groupwise heteroscedasticity and cross-sectional
correlation
Null hypothesis Estimated test Degrees of 5% critical
statistic freedom value
1 2 3
There is no cross-sectional 167.04 12 21.03
heteroscedosticity
There is no cross-sectional 94.39 78 99.62
correlation
Note: The first test result is based on the variance-covariance
matrix of the estimated residuals obtained from the pooled LS
estimation. The second test result is based on the correlation
matrix of the estimated residuals obtained from a feasible GLS
estimation that uses estimated cross-section variances as weights
for various observations.
Table 4: F-test results for exclusion of control variables and fixed
effects
Category of Variable F-statistics
variables
1 2
Macroeconomic Inflation (INF) 4.15 **
variables Fiscal balance (FBAL) 7.18 **
Size of the government (GOV) 16.45 ***
Exchange rate (X) 0.31
Employment growth (EMP) 1.4
Real money growth (MONEY) 2.76 *
Real domestic credit growth 2.32
(DOMCREDIT)
Real lending rate (LRATE) 63.66 ***
Gross reserves-to-GDP ratio 7.06 **
(RES)
All macro variables 14.20 ***
Structural Tariff revenue-to-imports 10.40 ***
variables ratio (TARIFF)
Share of industry in total 0.04
employment (INDEMP)
Share of private sector in 0.14
GDP (PVT)
Budgetary subsidies- to- 1.55
GDP ratio (SUB)
Infrastructure reform index 14.59 ***
(INFRA)
All structural variables 4.26 ***
Fixed effects 5.96 ***
Category of Variable Degrees P-value
variables of freedom
1 3 4
Macroeconomic Inflation (INF) (1,118) 0.04
variables Fiscal balance (FBAL) (1,118) 0.01
Size of the government (GOV) (1,118) 0.00
Exchange rate (X) (1,118) 0.58
Employment growth (EMP) (1,118) 0.24
Real money growth (MONEY) (1,118) 0.09
Real domestic credit growth (1,118) 0.13
(DOMCREDIT)
Real lending rate (LRATE) (1,118) 0.00
Gross reserves-to-GDP ratio (1,118) 0.01
(RES)
All macro variables (9,118) 0.00
Structural Tariff revenue-to-imports (1,118) 0.00
variables ratio (TARIFF)
Share of industry in total (1,118) 0.84
employment (INDEMP)
Share of private sector in (1,118) 0.71
GDP (PVT)
Budgetary subsidies- to- (1,118) 0.21
GDP ratio (SUB)
Infrastructure reform index (1,118) 0.00
(INFRA)
All structural variables (5,118) 0.00
Fixed effects (12,118) 0.00
*** Significant at 1% level; ** significant at 5% level;
* significant at 10% level.
Table 5: Trade, FDI, and per capita real GDP growth: fixed effects
panel estimates for 13 CEEB transition economies
Feasible generalized least square
estimates
Independent variables
One-stage Two-stage
1 2
1-year Lagged per capita
real GDP growth rate
Gross domestic investment-to 0.111 *** (0.027) 0.141 *** (0.038)
-GDP ratio (GDI)
FDI-to-GDP ratio (FDI) 0.007 (0.094) 0.008 (0.116)
(Exports+Imports)-to-GDP 0.060 ** (0.025) 0.069 ** (0.026)
ratio (TRADE)
Inflation (INF) -0.017 (0.012) -0.017 (0.012)
Fiscal balance (FBAL) 0.315 *** (0.087) 0.315 *** (0.087)
Size of the government (GOV) -0.281 ** (0.121) -0.281 ** (0.121)
Real money growth (MONEY) 0.042 ** (0.016) 0.042 ** (0.016)
Real lending rate (LRATE) -0.032 *** (0.008) -0.032 *** (0.008)
Gross reserves-to-GDP ratio 0.062 (0.053) 0.062 (0.053)
(RES)
Tariff revenue-to-imports 0.262 ** (0.115) 0.262 ** (0.115)
ratio (TARIFF)
Infrastructure reform index -0.480 (1.067) -0.480 (1.067)
(INFRA)
[R.sup.2] 0.645 0.645
Adjusted [R.sup.2] 0.588 0.588
D-W statistics 1.263 1.263
No of observations 166 166
Feasible generalized least square
estimates
One-stage Two-stage
3 4
1-year Lagged per capita 0.259 *** (0.043) 0.259 *** (0.043)
real GDP growth rate
Gross domestic investment-to -0.002 (0.053) 0.044 (0.051)
-GDP ratio (GDI)
FDI-to-GDP ratio (FDI) -0.124 (0.098) -0.108 (0.113)
(Exports+Imports)-to-GDP 0.033 * (0.018) 0.032 (0.022)
ratio (TRADE)
Inflation (INF) -0.016 (0.010) -0.016 (0.010)
Fiscal balance (FBAL) 0.222 ** (0.100) 0.222 ** (0.100)
Size of the government (GOV) -0.089 (0.134) -0.089 (0.134)
Real money growth (MONEY) 0.038 ** (0.018) 0.038 ** (0.018)
Real lending rate (LRATE) -0.033 *** (0.007) -0.033*** (0.007)
Gross reserves-to-GDP ratio 0.058 (0.045) 0.058 (0.045)
(RES)
Tariff revenue-to-imports 0.119 (0.125) 0.119 (0.125)
ratio (TARIFF)
Infrastructure reform index -0.467 (0.954) -0.467 (0.954)
(INFRA)
[R.sup.2] 0.672 0.672
Adjusted [R.sup.2] 0.616 0.616
D-W statistics 1.807 1.807
No of observations 165 165
Note: The numbers in parentheses are the White robust cross-section
standard errors. For the results reported in columns 2 and 4,
we estimate
GDI = C - 0.699 x FDI + 0.063 x TRADE; FDI = C + 0.037 x TRADE -
0.285 x GDI and TRADE = C + 0.471 x GDI + 1.441 x FDI
(0.088) (0.011) (0.018) (0.074) (0.204) (0.307)
in the first stage. The estimated standard errors are in parentheses.
The estimated fixed effects are not shown above and included in
equation as C. We obtain the residuals from these equations and
use them as regressors in the growth equation in the second stage.
Sample period: 1991-2005. Dependent variable: Growth rate of per
capita real GDP.
*** Significant at 1% level; ** significant at 5% level;
* significant at 10% Level.
Table 6: Trade, FDI, and per capita real GDP growth: fixed effects
panel estimates for 13 CEEB transition economies
Independent variables 1 2
1990 Per capita real GDP -1.087 *** (0.339)
(Y1990)
Cluster of 8 initial
conditions (IC cluster)
Cluster of 11 initial 0.379 (0.291)
conditions 1 (IC cluster 1)
Cluster of 11 initial -0.055 (0.363)
conditions 2 (IC-cluster 2)
Gross domestic investment- 0.115 *** (0.039) 0.195 *** (0.047)
to-GDP ratio (GDI)
FDI-to-GDP ratio (FDI) -0.020 (0.118) 0.007 (0.140)
(Exports+Imports)-to-GDP 0.038 * (0.021) 0.073 *** (0.025)
ratio (TRADE)
Inflation (INF -0.024 (0.015) -0.026 * (0.014)
Fiscal balance (FBAL) 0.120 ** (0.081) 0.081 (0.070)
Size of the government (GOV) -0.254 *** (0.062) -0.112 (0.073)
Real money growth (MONEY) 0.044 *** (0.016) 0.042 ** (0.016)
Real lending rate (LRATE) -0.040 *** (0.009) -0.041 *** (0.009)
Gross reserves-to-GDP 0.009 (0.021) 0.018 (0.032)
ratio (RES)
Tariff revenue-to-imports 0.012 (0.077) 0.175** (0.071)
ratio (TARIFF)
Infrastructure reform index -0.192 (0.318) -0.136 (0.309)
(INFRA)
Change in average transition
indicators (DREFORM)
[R.sup.2] 0.501 0.537
Adjusted [R.sup.2] 0.461 0.497
D-W statistics 1.007 1.102
No of observations 166 166
Independent variables 3 4
1990 Per capita real GDP -0.269115
(Y1990)
Cluster of 8 initial 0.248 *** (0.069)
conditions (IC cluster)
Cluster of 11 initial
conditions 1 (IC cluster 1)
Cluster of 11 initial
conditions 2 (IC-cluster 2)
Gross domestic investment- 0.179 *** (0.044) 0.088 ** (0.037)
to-GDP ratio (GDI)
FDI-to-GDP ratio (FDI) -0.015 (0.130) -0.015 (0.095)
(Exports+Imports)-to-GDP 0.062 *** (0.022) 0.053 *** (0.017)
ratio (TRADE)
Inflation (INF -0.024 * (0.014) -0.004 (0.013)
Fiscal balance (FBAL) 0.132 * (0.071) 0.342 *** (0.079)
Size of the government (GOV) -0.109 (0.070) -0.204 ** (0.103)
Real money growth (MONEY) 0.046 *** (0.015) 0.055 *** (0.019)
Real lending rate (LRATE) -0.042 *** (0.009) -0.032 *** (0.007)
Gross reserves-to-GDP 0.035 (0.026) 0.048 * (0.028)
ratio (RES)
Tariff revenue-to-imports 0.122 ** (0.058) 0.283 *** (0.098)
ratio (TARIFF)
Infrastructure reform index -0.179 (0.310)
(INFRA)
Change in average transition -5.604 *** (1.809)
indicators (DREFORM)
[R.sup.2] 0.551 0.669
Adjusted [R.sup.2] 0.512 0.616
D-W statistics 1.12 1.461
No of observations 166 166
Independent variables 5 6
1990 Per capita real GDP -1.398
(Y1990)
Cluster of 8 initial
conditions (IC cluster)
Cluster of 11 initial 0.492 * (0.252)
conditions 1 (IC cluster 1)
Cluster of 11 initial -0.255 (0.306)
conditions 2 (IC-cluster 2)
Gross domestic investment- 0.057 (0.045) 0.116 ** (0.048)
to-GDP ratio (GDI)
FDI-to-GDP ratio (FDI) -0.045 (0.100) -0.024 (0.116)
(Exports+Imports)-to-GDP 0.024 (0.016) 0.057 *** (0.021)
ratio (TRADE)
Inflation (INF -0.017 (0.014) -0.013 (0.015)
Fiscal balance (FBAL) 0.215 *** (0.068) 0.114 * (0.060)
Size of the government (GOV) -0.227 *** (0.059) -0.112 (0.070)
Real money growth (MONEY) 0.054 *** (0.017) 0.060 *** (0.019)
Real lending rate (LRATE) -0.042 *** (0.008) -0.043 *** (0.008)
Gross reserves-to-GDP 0.004 (0.017) 0.014 (0.020)
ratio (RES)
Tariff revenue-to-imports 0.013 (0.067) 0.175 *** (0.055)
ratio (TARIFF)
Infrastructure reform index
(INFRA)
Change in average transition -4.516 * (2.509) -5.392 ** (2.181)
indicators (DREFORM)
[R.sup.2] 0.522 0.603
Adjusted [R.sup.2] 0.484 0.569
D-W statistics 1.091 1.210
No of observations 166 166
Independent variables 7
1990 Per capita real GDP -0.737 * (0.394)
(Y1990)
Cluster of 8 initial 0.245 *** (0.064)
conditions (IC cluster)
Cluster of 11 initial
conditions 1 (IC cluster 1)
Cluster of 11 initial
conditions 2 (IC-cluster 2)
Gross domestic investment- 0.109 ** (0.043)
to-GDP ratio (GDI)
FDI-to-GDP ratio (FDI) -0.057 (0.107)
(Exports+Imports)-to-GDP 0.044 ** (0.018)
ratio (TRADE)
Inflation (INF -0.015 (0.014)
Fiscal balance (FBAL) 0.151 ** (0.061)
Size of the government (GOV) -0.099 (0.066)
Real money growth (MONEY) 0.061 *** (0.017)
Real lending rate (LRATE) -0.045 *** (0.010)
Gross reserves-to-GDP 0.034 * (0.018)
ratio (RES)
Tariff revenue-to-imports 0.114 ** (0.050)
ratio (TARIFF)
Infrastructure reform index
(INFRA)
Change in average transition -5.042 ** (2.046)
indicators (DREFORM)
[R.sup.2] 0.602
Adjusted [R.sup.2] 0.568
D-W statistics 1.231
No of observations 166
Note: The numbers in parentheses are the White robust cross-section
standard errors. IC cluster is calculated using the factor loadings
of the first principal component of eight initial conditions suggested
by Polanec (2004). IC cluster1 and IC cluster2 are calculated using
the factor loadings of the first two principal components of 11
initial conditions suggested by de Melo et al. (1997).
Sensitivity analysis results with initial conditions and reform
measures. Dependent variable: Growth rate of per capita real GDP.
(Two-stage feasible generalized least square estimates).
*** Significant at 1% level; ** significant at 5% level;
* significant at 10% level.
Table 7: Trade, FDI, and per capita GDP growth: fixed effects panel
estimates for 13 CEEB transition economies
Independent Feasible generalised least square
variables estimates 1995-2005
One-stage Two-stage
1 2
1-year lagged per capita
real GDP growth rate
Gross domestic investment-to 0.111 ** (0.043) 0.205 *** (0.068)
-GDP ratio (GDI)
FDI-to-GDP ratio (FDI) 0.006 (0.068) 0.506 *** (0.150)
(Exports+Imports)-to-GDP 0.076 *** (0.012) 0.112 *** (0.019)
ratio (TRADE)
Inflation (INF) -0.008 (0.008) -0.008 (0.008)
Fiscal balance (FBAL) 0.346 *** (0.061) 0.346 *** (0.061)
Size of the government (GOV) -0.325 *** (0.101) -0.325 *** (0.101)
Real money growth (MONEY) 0.046 *** (0.014) 0.046 *** (0.014)
Real lending rate (LRATE) -0.029 *** (0.003) -0.029 *** (0.003
Gross reserves-to-GDP 0.101 *** (0.036) 0.101 *** (0.036)
ratio (RES)
Tariff revenue-to-imports 0.163 (0.113) 0.163 (0.113)
ratio (TARIFF)
Infrastructure reform -2.501 *** (0.625) -3.126
index (INFRA)
[R.sup.2] 0.794 0.794
Adjusted [R.sup.2] 0.753 0.753
D-W statistics 1.938 1.938
No of observations 140 140
Independent Panel generalised method of moments
variables 1991-2005
One-stage Two-stage
3 4
1-year lagged per capita 0.262 * (0.138) 0.262 * (0.138)
real GDP growth rate
Gross domestic investment-to -0.158 (0.134) -0.112 (0.127)
-GDP ratio (GDI)
FDI-to-GDP ratio (FDI) -0.231 (0.159) -0.082 (0.156)
(Exports+Imports)-to-GDP 0.059 * (0.033) 0.049 (0.038)
ratio (TRADE)
Inflation (INF) -0.022 (0.016) -0.022 (0.016)
Fiscal balance (FBAL) 0.188 * (0.105) 0.188 * (0.105)
Size of the government (GOV) -0.230 (0.280) -0.230 (0.280)
Real money growth (MONEY) 0.041 (0.027) 0.041 (0.027)
Real lending rate (LRATE) -0.039 *** (0.008) -0.039 *** (0.008)
Gross reserves-to-GDP -0.007 * (0.97) -0.007 (0.097)08)
ratio (RES)
Tariff revenue-to-imports 0.197 (0.248) 0.197 (0.248)
ratio (TARIFF)
Infrastructure reform -0.210 (1.463) -0.210 (1.463)
index (INFRA)
[R.sup.2]
Adjusted [R.sup.2]
D-W statistics
No of observations 152 152
Note: The numbers in parentheses are the White robust cross-section
standard errors.
Sensitivity analysis results. Dependent variable: Growth rate of per
capita real GDP.
*** Significant at 1% level; ** significant at 5% level; * significant
at 10% level.
