Post on 11-Jun-2020
transcript
FDI and ICT Innovation Effects on Productivity Growth:
A Comparison between Developing and Developed Countries
Sotiris K. Papaioannou*
Abstract
This paper investigates for possible innovation effects stemming from Foreign Direct
Investment (FDI) and Information and Communication Technologies (ICT) on productivity
growth. An augmented production function was estimated using a sample of developing
and developed countries in 1993-2001. A uniform positive and significant innovation effect
from FDI was established for all countries, while divergent results between developing and
developed countries were obtained for ICT and the interaction effects of FDI. These results
are robust to possible endogeneity and omitted variable problems. They also suggest that
the level of development matters in estimating such impacts.
JEL classification: F21; O30; O47
Keywords : FDI; ICT; Innovation; Productivity
(The paper is competing to the Young Economist Award)
* Ph.D. Student, Athens University of Economics and Business, 76 Patission Street, 10434 Athens, Greece e-mail: sopa@aueb.gr. **I am grateful to the Greek State’s Scholarship foundation for its financial support. 1
1. Introduction
The rapidly rising level of economic integration, stimulated by advances in
transportation as well as in information and communication technology, renders technology
adoption, coming from foreign developed countries, a matter of great importance for
economic growth and productivity improvement. Furthermore, economic theory suggests
that learning through international economic activity might be particularly important for all
countries, especially for those lagging behind the most developed ones.
Foreign Direct Investment (FDI) is considered, among others, an important channel
for technology diffusion. Multinational enterprises possess superior technology and
management techniques, some of which are captured by local firms when multinationals
locate in a particular economy. Other sources of positive FDI effects are forward and
backward linkages between multinational and domestic firms, as well as the host country’s
access to specialised intermediate inputs, which in turn raise the economy’s total factor
productivity. Furthermore, the new ‘information economy’ of the last decades is associated
with increased diffusion of Information and Communication Technologies (ICT) which is
expected to deliver higher productivity gains and enhanced growth1.
An emerging body of empirical literature is concerned with how FDI affects labor
productivity and economic growth in host economies. Most studies in this literature have
been conducted at the micro-level using firm-level or industry data and are usually limited
to manufacturing industry. The existing empirical evidence is mixed, depending on the type
of data examined (cross-sectional versus panel data), the level of development of the FDI
recipient country, the econometric analysis employed and the research design. Fewer
studies have been conducted at the macro or international level given the lack of long time-
1 See the review paper by Blomstrom and Kokko (1998) for the FDI spillover effects on labor productivity and growth. Also, see Daveri (2002) for a recent review on the ICT growth effects. 2
series data on FDI and relevant country or industry characteristics. On the other hand,
recent evidence on the growth contribution of ICT capital is also mixed with the so-called
‘productivity paradox’ remaining still unresolved. Thus, as richer data are becoming
available for longer periods and more countries, the macro-economic effects of technology
transfer through FDI and ICT become appealing.
The aim of this study is towards this direction: to search for innovation effects on
aggregate productivity growth generated by either FDI technology transfer or/and ICT
investment. A production function is employed to investigate these effects using a panel of
developing and developed countries in the period 1993-2001. The study of this period
attracts a special interest given the sizable increase in world FDI after the major political,
financial and economic reforms of the 1990s.
Evidence is provided for a positive and significant impact of FDI on productivity
growth in both developing and developed countries. A uniform positive and significant
innovation effect from FDI on growth was established for all countries, while divergent
results between developing and developed countries were obtained for ICT and the
interaction effects of FDI. These results are robust to possible endogeneity and omitted
variable problems. They also suggest that the level of development matters in estimating
such impacts.
The rest if this paper is organized as follows. Next section summarizes the results of
the relevant literature. Section 3 contains the econometric specification of the model. In
section 4 the data are described and some descriptive statistics are presented, while section
5 provides the empirical results. Finally, section 6 concludes.
3
2. Related Literature
An emerging body of literature considers how production externalities, arising from
FDI, affect the host economies. So far, the existing evidence on the impact of FDI, as a
mechanism of technology diffusion, is mixed. However, serious econometric problems
characterise most of the cross sectional studies, such as the endogeneity issue and the
omitted variables bias. In a relatively early study on some OECD developed countries,
Barrell and Pain (1997) suggest that there is evidence for significant spillovers and
increased export performance from the presence of inward FDI. In a related work,
Borensztein et al. (1998), using a panel of 69 developing countries in the 1970s and 1980s,
found a significant positive effect of FDI on growth only for countries with a minimum
threshold stock of human capital. These results suggest the importance of the absorptive
capacity of the host economies in assimilating the advanced technologies transferred,
usually from developed countries, a hypothesis thoroughly explored in relevant micro-
studies.
Hejazi and Safarian (1999) estimated that FDI is a dominant channel for R&D
diffusion in OECD countries with its importance being higher than that of trade. However,
de Mello (1999) using both time-series and panel data techniques in a number of OECD
and non-OECD countries, during the period 1970-1990, provided evidence that the extent
to which FDI is growth-enhancing depends on the complementarity or substitutability
between FDI and domestic investment. Furthermore, Balasubramanyam et al. (1999)
suggest that an important role is exerted by the size of the local market, the competitive
environment and the availability of human capital in order for FDI to promote economic
growth, while Elahee and Pagan (1999) find positive evidence for the role of FDI in East
Asian and Latin America countries, over the period 1985-1993.
4
Barthelemy and Demurger (2000), using panel data on 24 Chinese provinces in the
period 1985-1996, provide evidence for a positive and mutual relationship between FDI
and economic growth. Furthermore, they stress the importance of human capital for the
adoption of foreign technologies and economic growth. Haveman et al. (2001), using data
from 1970 to 1989 and 74 countries, find evidence for a positive effect of international
integration indicators, such as openness, membership in a trade block and FDI, on
economic growth. In addition, they suggest that these indicators are significantly correlated
and should be examined together in order for their estimated impacts to be robust. By
contrast, Zhang (2001), in a study of 11 East Asian and Latin America countries during the
period 1960-1997, finds that there is a strong variation in the growth enhancing impact of
FDI. According to his findings, FDI is more likely to boost economic growth in countries
with particular characteristics like liberalised trade regimes, improved education, large
export-oriented FDI and macroeconomic stability, e.g. Hong Kong, Indonesia, Singapore,
Taiwan and Mexico. Further evidence in favour of a positive growth FDI effect is provided
by Ram and Zhang (2002) using a cross section of 85 countries between the years 1990 and
1997, and Campos and Kinoshita (2002) utilising panel data of 25 transition economies in
the period 1990-1998.
Regarding the impact of ICT on growth and productivity, the existing evidence
shows mixed results. Most recent estimates converge to the conclusion that, at least for
USA and high technology sectors, this effect is positive and significant. For the remaining
European and other world countries, the results are not conclusive. Schreyer (2000) and
Daveri (2002) examine the contribution of ICT on G7 and European countries,
respectively, and show that there do not exist powerful signs for beneficial effects on
productivity. In an empirical study of the period 1985-1993, Dewan and Kraemer (2000)
5
come to the conclusion that the developed countries enjoyed substantial gains and achieved
an increase in their output by the use of ICT. On the contrary, the developing countries do
not benefit from essential returns and this is justified by the lack of additional infrastructure
investments. Finally, Gust and Marquez (2004), in their study of the period 1992-1999,
show that productivity growth has slowed in a number of industrialized countries due to
regulations affecting labour market practices which have impeded the adoption of
information technology.
3. Econometric Specification
3.1. Production Function
To capture the effect of FDI on productivity growth, a production function is
specified with several types of inputs. The present study considers FDI as a special type of
knowledge and technology capital introduced in the production process. Consequently, the
regression analysis will be carried on by decomposing the overall effect of total capital to
that of its individual domestic and foreign components2.
In order to capture the FDI and ICT effects, the paradigm of Hall and Mairesse
(1995) will be followed in specifying an aggregate Cobb-Douglas production function,
which incorporates four inputs, domestic capital (K), labour (L), foreign capital (F) and
ICT capital:
Yit = A ect(Kit)α(Lit)β(Fit)γ(ICTit)δ euit (1)
where the subscripts of i and t denote country and year, respectively; Y measures gross
output of each country, A and c are constant parameters, while t is a time trend. Parameters
α, β, γ and δ are the elasticities of domestic capital, labor, foreign capital and ICT with
2 In the subsequent sections, the terms of FDI and foreign capital are used interchangeably. 6
respect to output and finally uit is the error term capturing unobserved variations between
countries and over time.
After taking logarithms and following the assumption of constant returns to scale,
the level of output per worker can be expressed as a function of domestic capital, foreign
capital and ICT to labour ratios:
ln( + 1 ln( ) + ∂ 2 ln( ) + 3 ln( ) + (2) ctAyit += ln) ∂ itk itf ∂ itict itu
where small case letters denote figures per worker, while the parameters 1, 2, 3
represent the elasticities α, γ, and δ, respectively. Since the goal is to estimate a growth
equation, the first differences of the above equation are taken to obtain the following
form:
∂ ∂ ∂
lnyit - lnyit-1 = c + 1(lnkit - lnkit-1) + 2 (lnfit - lnfit-1) + 3 (lnictit - lnictit-1) + (3) ∂ ∂ ∂ itε
The above formulation is further augmented by a number of other variables proposed by
the new growth theory (Mankiw et al., 1992). Thus, the lagged level of output per worker
(in its logarithmic scale) is introduced, to capture the catch-up effect among countries, as
suggested by Barro (1997). Human capital is, also, introduced, the importance of which
may be strong for economic growth, as Barro (1991) has already found for a cross section
of 98 countries in the period 1960-1985. Other control variables include the transparency
indicator, the government share of GDP and the openness of each country to international
trade, defined as the ratio of total imports and exports to GDP.
3.2. Fixed Effects or Random Effects?
When dealing with panel data, it is typical to view the unobserved factors affecting
the dependent variable as consisting of two types: those that are constant and those that
vary over time. Consequently, the following structure of the error term is specified:
7
= ηi + (4) itε itα
The term ηi is an unobserved time-invariant country effect, while is the idiosyncratic
error that varies independently across countries and time, assumed to be uncorrelated with
the other explanatory variables (Hsiao, 1986; Johnston and Dinardo, 1997). Therefore, the
cross country time-varying growth equation can be rewritten as:
itα
lnyit - lnyit-1 = c + ∂ 0( lnyit-1) + ∂ 'Χit + ηi + (5) itα
where Χit denotes the vector of explanatory variables included in (3) as well as all control
variables mentioned earlier, except the lagged per worker GDP, and ∂ the vector of the
corresponding parameters. Depending on the assumption about the correlation between the
cross-section effect ηi and the explanatory variables, two empirical models can be specified
based on whether the random effect or the fixed effect estimator is used. The former one is
more efficient but it is based on the assumption that the country effect, ηi, is not correlated
with the vector of the explanatory variables X. Furthermore, the latter one is more
consistent since it does not require the existence of this assumption but, on the other hand,
is less efficient due to loss of variation in the data by the imposition of country dummies.
Since the choice of any of these two techniques implies a consistency–efficiency trade off,
the best strategy followed by many researchers is to test whether the difference between the
random effect and the fixed effect estimates is significantly different from zero. After
applying the Hausman (1978) specification test3, the results indicate that the difference
between the random effect and fixed effect estimates is statistically significant indicating
that the best strategy would be the employment of a fixed effect estimator.
3 The Hausman statistic is distributed as a chi-square variable whose value reaches 351.98 (p-value: 0.00) when the initial hypothesis is that the difference in coefficient estimates is not systematic. 8
3.3. Endogeneity Issues
Although the basic motivation of most of the existing theoretical and empirical
work is the potential effect of FDI on economic growth, the association between GDP
growth and FDI does not mean that causality runs from one direction. The direction of
causation may run either way.
To correctly assess the empirical relationship between productivity growth, FDI,
ICT and other variables included in the vector of explanatory variables, X, the generalized
method of moments (GMM) estimator is used, as developed by Arellano and Bond (1991).
They propose to differentiate equation (5) which becomes:
(lnyit – lnyit-1) - (lnyit-1 – lnyit-2) = 0 (lnyit-1 – lnyit-2) +∂ '(Xit - Xit-1) + (αit -αit-1) (6) ∂
While differencing eliminates the country specific effect, a new bias is introduced
by the construction of the error term, αit–αit-1 which is correlated with the lagged dependent
variable, lnyit-1–lnyit-2. However, after accepting that the error term is not serially correlated
and that the set of explanatory variables, X is weakly exogenous, that is to say that the
explanatory variables are not correlated with future values of the error term, Arellano and
Bond (1991) propose the following moment conditions:
E[(lnyit-s–lnyit-s-1)*(αit–αit-1)]=0 for s≥ 2; t= 3,...,T (7)
E[Xit-s(αit – αit-1)] = 0 for s ≥ 2; t = 3,...,T (8)
Using these moment conditions, Arellano and Bond propose a GMM estimator which uses
the lagged values of some explanatory variables as instruments in a differenced regression
equation. These explanatory variables are treated as predetermined, in that it is supposed
that past values of the disturbance term have some impact on their future realizations.
The consistency of the GMM estimator is based on the validity of the instruments
used in the differenced regression and the absence of second order serial correlation in the 9
error term. For this reason, Arrelano and Bond (1991) propose two specification tests. The
first one is a Sargan test of over-identifying restrictions which tests for the validity of the
instruments used in the regression. The second one is a test which examines for second-
order serial correlation4. Failure to reject the null hypotheses of both tests gives support to
the above model.
4. Data and Descriptive Statistics
A panel of 43, developed and developing, countries, in the period 1993-2001, was
constructed for this empirical application. The required data were taken from a variety of
sources. GDP data were taken from Penn World Tables (Heston et al., 2002) and the World
Bank (2003) database. Capital stock data were estimated using the perpetual inventory
method and gross investment figures from IMF (2003)5. The initial values of the capital
stock series were taken from Penn World Tables (Heston and Summers, 1991).
Since data on total fixed investment of a particular country includes FDI, the capital
stock series constructed as above would be correlated with the FDI series. Furthermore,
physical capital is a stock variable, so it would not be correct to include FDI as a flow
variable. Subsequently, a procedure was followed to break down the capital series into its
domestic and foreign component. More specifically, the value of foreign capital was
approximated by utilizing the share of FDI stock to GDP, as published by UNCTAD
(2003a). Domestic capital was then obtained by subtracting the value of foreign capital
stock from that of total capital stock.
4 By construction, the differenced error term is first order correlated, but this does not imply that so does the original error term. 5 In countries for which no initial estimate is given, the capital stock variable is calculated as the sum of gross investments that have been realized until previous year minus their accumulated depreciation. The depreciation for each year is calculated using the Winfrey mortality function. 10
Data on ICT capital are not existent. Instead, data on ICT spending are provided by
the World Information Technology and Services Alliance (WITSA, 2002) which can be
used to capture the ICT effects. The ICT spending data comprise of household
consumption, public consumption and private investments. Furthermore, the data regarding
the number of workers were taken from the International Labor Organization (ILO, 2003).
Human capital was approximated by male secondary enrollment rates obtained from
World Bank (2003), while the government share of GDP is taken from Penn World Tables
(Heston et al., 2002). The transparency index and the openness indicator were taken from
Transparency International Organization (2004) and Penn World Tables (Heston et al.,
2002), respectively. All value variables are expressed in purchasing power parity (PPP) in
order to make the data compatible across countries
It should be made clear that the number of countries included was determined by
the availability of data on both FDI and ICT, which are the variables this paper focuses on.
With this in mind, first a description of the data is made and then follows the regression
analysis. Tables 1-3 contain descriptive statistics for the variables and group of countries
under investigation. According to Table 1, developing countries have increased their share
on world FDI, at the expense of the most developed ones, with the exception of some East-
Asian countries, which have witnessed an important decrease, due to the 1997 financial
crisis. A special case worth mentioning is that of China, being now the first FDI recipient
economy in the world. Overall, inward FDI seems to have gained more importance as an
investment mechanism, since its percentage share on Gross Fixed Capital Formation has
increased from 4.4 % in the period 1991-1996, to 12.8% in 20016, as reported by UNCTAD
(2003b).
6 Similar conclusions are drawn for outward FDI, since its share has increased from 5% in the period 1991-1996 to 11.3% in 2001. 11
Table 2 contains descriptive statistics, for all variables that will be employed in the
econometric analysis. As it is evident, the developing countries exhibit higher FDI, ICT and
capital growth rates, while their GDP per worker growth is similar to that of the developed
countries. Apparently, a longer time period might be necessary in order for the developing
countries to reap the benefits of their investments. Finally, it appears that the developing
countries exhibit higher government presence in their economies and seem to adopt less
liberalized trade policies, which may have negatively affected their economic growth.
5. Regression Analysis and results
5.1. Initial Results
Regressions are performed on a pooled cross-section time-series data set consisting
of 43 countries in a nine year period (1993-2001). Annual labour productivity growth is
regressed on a number of explanatory variables suggested by growth theory. Equation (5) is
estimated using the fixed effects methodology, the results of which are presented in tables
4-6.
The baseline regression in each table (column 1) includes the lagged level of output
per worker (YL), two forms of capital inputs: domestic and foreign capital growth per
worker (GKD and GKF), and a proxy for the human capital variable (SCHOOL). As it is
evident from table 4, the elasticity of both forms of capital is highly significant with that of
domestic being much higher as expected, the catch-up effect is negative and significant
implying income convergence, while the impact of schooling on growth is negative but
insignificant. Next column reports results based on the initial regression with the addition
of the growth rate of ICT spending per worker (GICT)7 and three control variables: an
7 Because of lack of ICT capital data, the ICT spending variable was used instead. In this case, the coefficient does not measure elasticity, but the return to productivity of ICT spending. 12
indicator referring to the level of transparency of the corresponding country (TI), the degree
of openness measured by the share of total trade to GDP (OPEN) and the government share
of GDP (GOVSH).
It is interesting to notice in column 2 the positive and significant, at the 10% level,
ICT growth effect, something that has long been disputed in the empirical literature. The
inclusion of the other three variables did not affect much the previous estimates, while
openness and the level of transparency seem to exert a significant impact on growth. The
explanatory power of the model improves (R2 = 0.30) and remains satisfactory for this type
of analysis. The government share of GDP variable shows no significant correlation with
economic growth as found in other empirical studies.
Columns 3-7 report the estimation results after introducing interaction terms of
foreign capital growth with domestic capital growth, ICT spending growth, openness of
trade and human capital. It can be noted that the previous parameter estimates remain
robust across the alternative regressions, with the exception of foreign capital, the
magnitude of which is increasing substantially. More discussion on the interaction terms is
given in a separate section that follows.
5.2. Differences between Developing and Developed Countries
A Chow test for the equality of coefficient estimates between the developing and
developed countries was rejected. For this reason, separate regressions for the two sub-
samples (developing and developed countries) were performed, the results of which are
presented in tables 5 and 6, respectively.
Concerning the developing countries (table 5), we can easily notice that the
domestic capital effect is of equal importance, as compared to the full panel of countries.
13
Furthermore, the foreign capital impact is large and positive in most of the regressions.
Only in column 6 the foreign capital coefficient was negative, a fact that can be attributed
to the presence of its interaction with schooling, the variation of which is very small in the
sample.8 The importance of ICT is also positive, but no conclusive inference can be drawn
since its significance is small. Furthermore, all the other control variables (transparency,
openness indicator and government share) exert a positive and mostly significant impact on
productivity growth.
With respect to the developed countries (table 6), a first reading of the results
reveals that the coefficient of domestic capital intensity is positive and significant and its
magnitude is greater than the one obtained from the entire panel. Similarly, the foreign
capital impact is positive and significant, in most of the alternative regressions, while its
magnitude is also increasing as more interaction terms are added in the baseline regression.
Furthermore, ICT shows a small negative but insignificant impact, the openness indicator a
positive and significant effect, while the government share coefficient is negative and
insignificant.
5.3. Discussion of the Results
The estimation results, obtained from the panel data analysis described above,
suggest that the accumulation of FDI contributes positively and significantly to the
productivity growth of countries, irrespective of their level of development. Overall, the
results indicate the rising importance of foreign capital, relative to that of domestic capital,
for economic growth. This is further justified by comparing the estimated coefficients of
foreign and domestic capital to their relative shares in total capital. As it is evident from
8 It should be reminded that a high degree of correlation between FDI and its interaction with schooling was observed in the data as shown in Table 3. 14
table 9, the effect of foreign capital on productivity growth is relatively high if is taken into
account its low share in total capital. It deserves to mention that, in the group of developing
countries, its impact on productivity growth is higher compared to that of domestic capital,
which is partially justified by its relatively higher share in total capital, indicating the
higher efficiency gains rising from the employment of FDI in these countries.
Regarding the innovation effects from ICT, the results provide some preliminary
evidence with regards to the importance of the ‘new economy’ for growth in the developing
countries. This fact opposes the main finding of Dewan and Kraemer (2000), supporting
that, mainly, the developed countries have benefited from the use of ICT. The present
evidence about the innovation effect from ICT is therefore inconclusive, and is in
accordance with many other studies in the literature that have failed to explain the
productivity paradox.
Among the other significant contributors to growth, the trade openness is
mentioned, which exerts a positive and significant impact on growth, confirming previous
similar evidence (Haveman et al., 2001). On the contrary, human capital, proxied by
schooling, had no influence on growth. However, the existing evidence is not conclusive
about the significance of human capital on growth. In an earlier cross country study,
Benhabib and Spiegel (1994) did not find any significant impact when human capital
entered the growth equation as a separate input. One of the difficulties in estimating growth
regressions with panel data is the measurement of human capital. The lack of long annual
time series data leads to the use of less appropriate proxies as the one used in this paper,
15
which may not capture properly the effect of education9. More discussion on this issue is
given in the following section.
5.4. Complementarity of Foreign Capital
One of the most arguable issues in the FDI-growth nexus is whether FDI-related
capital complements human capital or/and domestic investment. To investigate these
issues, interaction terms between foreign capital growth and domestic capital growth, ICT
spending growth, openness or schooling, were included in the regressions reported in
columns 3-7 of tables 4-6. When introducing interaction terms in a regression, collinearity
may result among them. In this case, foreign capital was found to be highly correlated only
when interacted with schooling10. A deeper analysis of their correlation indicates that this
was due to the small variability of schooling, the effect of which was insignificant in most
of the regressions. No serious correlation problems were created from the presence of the
other interaction terms.
As it is evident from the results in column 3, of tables 4-6, FDI interacts negatively
with domestic capital in developing countries, but positively in the developed ones. The
interaction effect, however, is significant only in the former ones as well as in the entire
sample. This finding could mean that foreign capital, embodying higher technological
advancement, cannot produce productivity gains, by complementing domestic capital, due
to the low absorptive capacity of the less developed host economy and, probably, due to
other institutional and cultural reasons. By contrast, in the case of a developed recipient
country, the technology gap is not that large to constrain complementarity with domestic
9 According to Barro and Lee (1994), the only measure of human capital that is most significantly correlated with growth is average years of male secondary schooling. However, this variable could not be used in this study as it is available on a five-year basis. 10 See correlation matrix in table 3. 16
capital. This implies that FDI may contribute to growth not only because it adds to
domestic capital, but also due to higher efficiency gains. Another possible explanation is
that foreign capital may not be deemed to be so much efficient or productive than the one
already employed in host economies.
These results are in contrast to the major findings of de Mello (1999) suggesting
that the degree of substitutability between domestic capital and FDI is higher in
technologically advanced economies. However, the period investigated in this paper (90’s)
is different, in many aspects, from the period (1970-1990) examined by de Mello.
Furthermore, de Mello (1997), in a survey paper, supports the view that the period
prevailed by complementarity may be short-lived. According to this aspect, a
Schumpeterian view of FDI innovative investment, which emphasizes creative destruction,
may overlook the scope for complementarity between FDI and domestic investment.
Assuming this hypothesis is valid, then the less technologically advanced countries may
promote the incorporation of additional more modern technologies that are also
complementary to FDI related capital and, subsequently, substitute for older, domestically
employed, technologies.
Regarding the interaction effect of foreign capital with ICT, it turns out that a
degree of substitutability characterizes the entire sample of countries. This effect is found
statistically significant in the entire panel and the sample of developing countries. If the
argument made earlier about technology imported by foreign countries not being so
technologically advanced is true, then it can be claimed that a high employment of both
FDI and ICT could lead to overinvestment and inefficiencies in the production process.
Furthermore, the interaction term between FDI accumulation and the openness
indicator is positive and significant only in developed countries. This finding indicates the
17
importance of trade liberalization for productivity gains to realize from FDI. Also, it can
possibly be argued that foreign capital is export enhancing in the case of developed
countries, while for the developing ones, foreign capital either crowds out domestic firms,
by the increase of competition, or increases imports at the expense of local producers. As,
Helpman and Krugman (1985) note, the association between outward FDI and exports in
technological leaders is mirrored by the link between imports and inward FDI in
technological followers.
These results are in line with theory which predicts that the economy with the
greater amount of human capital specializes in the production of goods and services that
cannot be produced anywhere else, so that technologically advanced economies increase
exports to FDI host countries. Similar studies show that FDI is more growth-enhancing in
countries that pursue export promotion than in those promoting import substitution
(Bhagwati, 1978), while a positive impact of outward FDI is found by Lin (1995) on both
exports of the home country to the recipient economy and imports of the host country from
the home economy.
Finally, the accumulation of FDI interacts positively with schooling only in the case
of developing countries, but this finding is insignificant. In general, as mentioned earlier,
the growth impact of schooling itself, as well as its interaction with foreign capital, was
found to be quite poor. These findings can be attributed either to measurement error or
inappropriate variable selection to capture the human capital effect. Similar evidence is
provided by Alfaro et al. (2004) showing that the interaction term of FDI with schooling is
negative. Previous findings, provided by Borensztein et al. (1998) suggest that FDI
promotes productivity growth only when the host country owns a sufficient stock of human
capital. It is likely that most of the countries under consideration have managed to hold a
18
minimum threshold stock of human capital so that an improvement of productivity is an
outcome exclusively owed to an increase of foreign capital. As it is evident from table 8,
male secondary enrollment rates display high percentages, so that FDI alone can become an
important vehicle for economic growth, holding human capital rates fixed. On the contrary,
when holding FDI fixed, a human capital increase is not a sufficient condition for a
productivity improvement.
5.5. Robustness Check
Another possible source of biased results is the case of simultaneity where a
correlation between the regressors and the error term exists. In this case, a positive
correlation between productivity and FDI is, in principle, just as likely to mean that foreign
capital is attracted to high-productivity countries as it is to mean that foreign capital raises
host country’s productivity. Furthermore, the empirical findings of Haddad and Harrison
(1993) and Aitken et al. (1997) give support to this argument. Possible variables that are
expected to be endogenous are those of foreign and domestic capital per worker, ICT
spending per worker and the interaction terms of foreign capital, as productivity shocks are
likely to affect them. To examine this hypothesis, the Arellano and Bond (1991) panel data
estimator is used, by using as instruments the lagged level of the dependent variables as
well as those of the explanatory variables mentioned above.
The first column of table 7 reports the results based on the one-step estimator, in
which the error term is assumed to be independent and homoskedastic across countries and
over time11. As it is evident, most parameters of interest retain their sign and significance
indicating that, even when controlling for the case of endogeneity, the conclusions
11 The estimates of this table are based on the entire panel of countries. An attempt was made to perform the estimator in the sub-samples of developing and developed countries, but due to the large number of lagged variables required, the model could not be safely estimated. 19
emanating from the initially described model are valid. However, as mentioned above, the
consistency of the GMM estimator is based on the validity of the instruments used in the
differenced regression (equation 6) and the absence of second order serial correlation in the
error term. As we can see from table 7, the reported Sargan test and the test which
examines for second-order serial correlation fail to reject their null hypotheses implying
that the instruments used are valid and that the error term does not exhibit second-order
serial correlation. Overall, these tests give further support to the estimated model and its
implications.
It should be noted that the error term assumption of independency and
homoskedasticity across countries and over time, is not fully realistic. For this reason,
Arellano and Bond (1991) propose a two-step estimator. This estimator results after
relaxing the assumptions of independence and homoskedasticity and constructing a
variance-covariance matrix obtained by the residuals of the first step. However, as shown
by Arellano and Bond (1991) and Blundell and Bond (1998), the asymptotic standard errors
of the two-step estimator are biased downwards, while the one-step estimator is
asymptotically inefficient relative to the two-step estimator. Consequently, the coefficient
estimates of the two-step estimator are asymptotically more efficient but the asymptotic
inference is more reliable in the case of the one-step estimator. As we can see from the
results reported in the second column of table 7, even when considering the two-step
estimator, the results do not differentiate with respect to the parameters of interest.
6. Conclusion
This paper investigates for possible innovation effects on productivity growth,
generated by the adoption of FDI, together with any impacts stemming from the
20
employment of ICT. Such effects were estimated by specifying an extended aggregate
production function and using a sample of 43 countries over the period 1993-2001. The
model was estimated by applying the fixed effect estimator for panel data and the Arellano-
Bond formula to correct for endogeneity problems.
A positive and significant impact of foreign capital is established in all groups, the
effect being larger among developing countries. Positive, yet not always significant, ICT
effects were found in the entire sample and among the developing countries. Other
interesting results include the strong substitutability between foreign and domestic capital
in the developing countries as opposed to weak complementarity observed in the developed
ones; the substitutability between foreign capital and ICT in all countries and, finally, the
positive interaction of foreign capital with openness in the developed countries. These
results provide further support to the hypothesis that FDI plays a crucial role in explaining
productivity growth in all countries, but more emphatically among the developing ones.
Possible weaknesses of this study include the, relatively, low number of years and
countries under examination, as well as the use of proxy variables for ICT and human
capital, mainly attributed to the lack of appropriate data. However, this paper is one of the
few panel data studies to compare developing and developed countries with respect to the
productivity effects stemming from the combined use of FDI and ICT. Despite the
weaknesses, the present study can stimulate future research on the above issues as more
data become available for an increased number of countries and years.
21
References
Aitken, B., Hanson, G., Harrison, A., 1997, Spillovers, foreign investment and export
behavior, Journal of International Economics 43 (1), 103-132.
Alfaro, L., Chanda, A., Ozcan, S. K., Sayek, S., 2004, FDI and economic growth: the role
of local financial markets, Journal of International Economics 64 (1), 89-112.
Arellano, M., Bond, S., 1991, Some tests of specification or panel data: Monte Carlo
evidence and an application to employment equations, The Review of Economic Studies 58
(2), 277-297.
Balasubramanyam, V. N., Salisu, M., Sapsford, D., 1999, Foreign direct investment as an
engine of growth, Journal of International Trade and Economic Development 8 (1), 27-40.
Barrell, R., Pain, N., 1997, Foreign direct investment, technological change and economic
growth within Europe, The Economic Journal 107, 1770-1785.
Barro, R., 1991, Economic growth in a cross section of countries, Quarterly Journal of
Economics 106 (2), 407-433.
Barro, R., 1997, Determinants of economic growth: A cross country empirical study MIT
Press, Cambridge, MA.
Barro, R., Lee, J.W., 1994, Sources of economic growth, Carnegie Rochester Conference
Series on Public Policy 40, 1-46.
Barthelemy, J.C., Demurger, S., 2000, Foreign direct investment and economic growth:
Theory and application to China, Review of Development Economics 4 (2), 140-155.
Benhabib, J., Spiegel, M., 1994, The role of human capital in economic development.
Evidence from aggregate and cross-country data, Journal of Monetary Economics 34 (2),
143-173.
22
Bhagwati, J.N., 1978, Anatomy and consequences of exchange rate regimes, Studies in
International Relations 1(10) NBER, New York.
Blomstrom, M., Kokko, A., 1998, Multinational corporations and spillovers, Journal of
Economic Surveys 12 (3), 247-277.
Blundell, R., Bond, S., 1998, Initial conditions and moment restrictions in dynamic panel
data models, Journal of Econometrics 87 (1), 115-143.
Borensztein, E., Gregorio, J., Lee, J.W., 1998, How does foreign direct investment affect
economic growth?, Journal of International Economics 45 (1), 115-135.
Campos, N., Kinoshita, Y., 2002, Foreign direct investment as technology transferred:
Some panel evidence from the transition economies, Manchester School 70 (3), 398-419.
Daveri, F., 2002, The new economy in Europe: 1992-2001, Oxford Review of Economic
Policy 18 (3), 345-362.
Dewan, S., Kraemer, K., 2000, Information technology and productivity: Evidence from
country-level data, Management Science 46 (4), 548-562.
Elahee, M.N., Pagan, J.A., 1999, Foreign direct investment and economic growth in East
Asia and Latin America, Journal of Emerging Markets 4 (1), 59-67.
Gust, C., Marquez, J., 2004, International comparisons of productivity growth: The role of
information technology and regulatory practices, Labour Economics 11 (1), 33-58.
Haddad, M., Harrison, A., 1993, Are there positive spillovers from direct foreign
investment? Evidence from panel data for Morocco, Journal of Development Economics 42
(1), 51-74.
Hall, B.H., Mairesse, J., 1995, Exploring the relationship between R&D and productivity in
French manufacturing firms, Journal of Econometrics 65 (1), 263-293.
Hausman, J., 1978, Specification tests in econometrics, Econometrica 46 (6), 1251-1271.
23
Haveman, J., Lei, V., Netz, J., 2001, International integration and growth: A survey and
empirical investigation, Review of Development Economics 5 (2), 289-311.
Hejazi, W., Safarian, A., 1999, Trade, foreign direct investment and R&D spillovers,
Journal of International Business Studies 30 (3), 491-511.
Helpman, E., P., Krugman, R., 1985, Market structure and foreign trade MIT Press,
Cambridge, MA.
Heston, Α., Summers, R., 1991, The Penn World Table (Mark 5) Version 5.6: An expanded
set of international comparisons 1950-1988, Quarterly Journal of Economics 106 (2), 327-
368.
Heston, Α., Summers, R., Aten, B., 2002, Penn World Table: Version 6.1 Center for
International Comparisons, Pennsylvania.
Hsiao, C., 1986, Analysis of Panel Data Cambridge University Press, Cambridge.
ILO, 2003, Labor Statistics, Available at: http://www.ilo.org.
IMF, 2003, International Financial Statistics, Available at: http://www.imfstatistics.org.
Johnston, J., Dinardo, J., 1997, Econometric methods McGraw-Hill, New York.
Lin, A. L., 1995, Trade effects of foreign direct investment: Evidence for Taiwan with four
ASEAN countries, Weltwirtschafisliches 131, 737-747.
Mankiw, G., Romer, D., Weil, D., 1992, A contribution to the empirics of economic
growth, Quarterly Journal of Economics 107 (2), 407-437.
de Mello, L.R., 1997, Foreign direct investment in developing countries and growth: A
selective survey, Journal of Development Studies 34 (1), 1-34.
de Mello, L.R., 1999, Foreign direct investment-led growth: Evidence from time series and
panel data, Oxford Economic Papers 51 (1), 133-151.
24
Ram, R., Zhang, K., 2002, Foreign direct investment and economic growth: Evidence from
cross-country data for the 1990’s, Economic Development and Cultural Change 51 (1),
205-215.
Schreyer, P., 2000, The contribution of information and communication technology to
output growth: A study of the G7 countries OECD, Paris.
Transparency International Organization, 2004, Corruption perceptions index, Available at:
http://www.transparency.org.
UNCTAD, 2003a, FDI statistics, Available at: http://www.unctad.org.
UNCTAD, 2003b, World Investment Report, New York.
WITSA, 2002, Digital Planet, Arlington-USA.
World bank, 2003, World Development Indicators, Available at: www.worldbank.org.
Zhang, K., 2001, Does foreign investment promote economic growth? Evidence from East
Asia and Latin America, Contemporary Economic Policy 19 (2), 175-185.
25
Table 1: FDI Inflow Shares
DEVELOPING COUNTRIES DEVELOPED COUNTRIES 1990 2003 1990 2003 China 1.68 9.56 United States 23.29** 5.32 Mexico 1.27 1.92 Un. Kingdom 14.65 2.59 Malaysia 1.26 0.44 France 7.51 8.4 Thailand 1.24 0.32 Spain 6.72 4.58 Argentina 0.88 0.08 Netherlands 5.06 3.51 Indonesia 0.53 -0.10 Australia 3.91 1.41 Brazil 0.48 1.81 Belgium 3.87 5.26 Egypt 0.35 0.04 Canada 3.65 1.17 Turkey 0.33 0.10 Italy 3.08 2.93 Chile 0.32 0.53 Singapore 2.68 2.03 Philippines 0.26 0.05 Switzerland 2.64 2.17 Colombia 0.24 0.31 Hong Kong 1.58 2.42 Venezuela 0.22 0.45 Germany 1.42 2.3 Hungary 0.15 0.44 Portugal 1.26 0.17 India 0.11 0.76 Sweden 0.95 0.58 Poland 0.04 0.75 Japan 0.84 1.13 Romania 0.00 0.27 New Zealand 0.83 0.36 S. Africa -0.04 0.13 Denmark 0.54 0.46 Greece 0.48 0.01 Norway 0.48 0.42 Korea 0.38 0.67 Finland 0.38 0.49 Austria 0.31 1.22 Ireland 0.3 4.55 Israel 0.04 0.67 TOTAL 9.31 17.86 TOTAL 86.86 54.82
* The FDI inflows are calculated as a percentage of world inflows. ** Countries are sorted by descending order according to their 1990 shares.
26
Table 2: Descriptive Statistics of all variables
ENTIRE PANEL DEVELOPING COUNTRIES
DEVELOPED COUNTRIES
Variable Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. GY 0.02 0.05 0.02 0.07 0.02 0.02 GKD 0.06 0.10 0.09 0.15 0.05 0.03 GKF 0.11 0.20 0.15 0.26 0.11 0.16 GICT 0.14 0.23 0.20 0.30 0.10 0.16 YL 10.44 0.57 9.94 0.61 10.76 0.18 SCHOOL 0.94 0.26 0.70 0.17 1.10 0.18 GOVSH 12.74 6.86 17.34 6.23 9.35 5.13 OPEN 73.01 55.09 59.84 36.40 82.69 63.89 TI 6.02 2.45 3.72 1.37 7.60 1.65 GKF*GKD 0.01 0.03 0.01 0.05 0.01 0.01 GKF*GICT 0.02 0.09 0.04 0.14 0.02 0.05 GKF*OPEN 9.20 18.66 8.61 14.03 9.58 21.09 GKF*SCHOOL 0.10 0.17 0.08 0.18 0.11 0.16 GY = Growth Rate of Output per Worker, GKD = Growth Rate of Domestic Capital per Worker, GKF = Growth Rate of Foreign Capital per Worker, GICT = Growth Rate of ICT Spending per Worker, YL = Lagged Level of ln(GDP per Worker), SCHOOL= Secondary Schooling (Male), GOVSH = Government Share of GDP, OPEN = Openness of Trade, TI = Transparency Index, GKF*GKD = Interaction term of Foreign and Domestic Capital, GKF*GICT = Interaction term of Foreign Capital and ICT, GKF*OPEN = Interaction term of Foreign Capital and Openness Indicator, GKF*SCHOOL = Interaction term of Foreign Capital and Secondary Schooling.
27
Table 3: Correlation Matrix
GY GKD GKF GICT YL SCHOOL GOVSH OPEN TI GKF* GKD
GKF* GICT
GKF* OPEN
GKF* SCHOOL
GY 1.00
GKD 0.43 1.00
GKF 0.17 0.02 1.00
GICT 0.14 0.00 0.34 1.00
YL -0.07 -0.09 0.07 -0.03 1.00
SCHOOL -0.02 -0.33 0.02 -0.04 0.50 1.00
GOVSH 0.02 0.35 0.12 0.21 -0.19 -0.52 1.00
OPEN 0.08 -0.11 -0.02 -0.04 0.00 0.04 -0.13 1.00
TI 0.03 -0.32 0.04 -0.07 0.58 0.73 -0.49 0.24 1.00
GKF*GKD -0.18 0.29 0.30 0.00 0.28 -0.16 0.30 -0.12 -0.14 1.00
GKF*GICT -0.16 -0.09 0.55 0.43 -0.09 -0.12 0.23 -0.07 -0.14 0.08 1.00
GKF*OPEN 0.08 -0.09 0.81 0.24 0.02 0.08 0.04 0.27 0.13 0.18 0.42 1.00
GKF*SCHOOL 0.14 -0.05 0.95 0.28 0.10 0.16 0.01 0.00 0.15 0.18 0.51 0.82 1.00 * See table 2 for the definitions of variables. ** The correlation matrix is calculated over a sample consisting of 235 observations which covers the entire set of variables.
28
Table 4: Entire Panel: Fixed Effect Panel Data Estimates
Dependent Variable: Growth Rate of Output per Worker
Independent Variables
(1)** (2)
(3)
(4)
(5) (6) (7)
C 1.630 (3.85)
2.256 (4.89)
1.81 (4.30)
1.904 (4.28)
2.11 (4.49)
2.214 (4.79)
1.317 (3.32)
YL -0.151 (-3.75)
-0.223 (-4.97)
-0.176 (-4.30)
-0.189 (-4.40)
-0.207 (-4.53)
-0.220 (-4.91)
-0.131 (-3.39)
GKD 0.178 (5.63)
0.166 (5.31)
0.189 (6.64)
0.136 (4.49)
0.166 (5.31)
0.168 (5.37)
0.164 (6.16)
GKF 0. 044 (2.81)
0.043 (2.31)
0.078 (4.40)
0.088 (4.35)
0.093 (2.53)
0.133 (1.85)
0.291 (4.66)
GICT 0.028 (1.78)
0.012 (0.82)
0.046 (2.99)
0.028 (1.81)
0.025 (1.57)
0.023 (1.65)
SCHOOL -0.047 (-1.07)
-0.071 (-1.59)
-0.086 (-2.15)
-0.046 (-1.10)
-0.077 (-1.74)
-0.061 (-1.37)
-0.049 (-1.29)
TI 0.011 (1.93)
0.007 (1.38)
0.013 (2.26)
0.01 (1.62)
0.011 (1.93)
0.008 (1.60)
OPEN 0.0009 (2.39)
0.0009 (2.82)
0. 0006 (1.77)
0.0009 (2.48)
0.0009 (2.55)
0.0008 (2.59)
GOVSH 0.001 (0.72)
0.0005 (0.42)
0.0005 (0.43)
0.001 (0.74)
0.001 (0.75)
0.00009 (0.08)
GKF*GKD -0.755 (-6.65)
-0.837 (-7.81)
GKF*GICT -0.271 (-4.63)
-0.270 (-5.27)
GKF*OPEN -0.0008 (-1.58)
-0.0001 (-0.23)
GKF*SCHOOL -0.097 (-1.29)
-0.172 (-2.54)
Obs. 250 235 235 235 235 235 235
R2 0.25 0.304 0 439 0.376 0.313 0.310 0. 539
F stat. 16.90 10.05 15.94 12.30 9.28 9.15 17.54 * See table 2 for the definitions of variables. ** The t-statistics are reported in parentheses.
29
Table 5: Developing Countries: Fixed Effect Panel Data Estimates
Dependent Variable: Growth Rate of Output per Worker
Independent Variables
(1)** (2)
(3)
(4)
(5)
(6) (7)
C 1.699 (2.24)
2.861 (3.73)
2.187 (3.14)
2.447 (3.24)
2.61 (3.39)
2.86 (3.71)
1.575 (2.32)
YL -0.166 (-2.21)
-0.324 (-4.12)
-0.246 (-3.41)
-0.280 (-3.62)
-0.296 (-3.72)
-0.322 (-4.07)
-0.180 (-2.57)
GKD 0.172 (3.36)
0.181 (3.88)
0.198 (4.76)
0.154 (3.34)
0.177 (3.84)
0.178 (3.75)
0.168 (4.13)
GKF 0.050 (1.55)
0.072 (1.71)
0.122 (3.12)
0.105 (2.47)
0.178 (2.24)
-0.056 (-0.27)
0.261 (1.29)
GICT 0.036 (1.25)
0.016 (0.62)
0.048 (1.71)
0.028 (1.00)
0.038 (1.31)
0.023 (0.89)
SCHOOL -0.070 (-0.76)
-0.173 (-1.87)
-0.204 (-2.49)
-0.115 (-1.25)
-0.179 (-1.96)
-0.209 (-1.90)
-0.140 (-1.53)
TI 0.048 (3.27)
0.034 (2.56)
0.05 (3.54)
0.043 (2.88)
0.047 (3.22)
0.032 (2.55)
OPEN 0.002 (2.24)
0.002 (3.02)
0.001 (1.75)
0.001 (1.95)
0.002 (2.29)
0.001 (2.08)
GOVSH 0.011 (2.74)
0.008 (2.27)
0.009 (2.28)
0.012 (2.88)
0.011 (2.64)
0.006 (1.80)
GKF*GKD -0.697 (-4.06)
-0.722 (-4.19)
GKF*GICT -0.236 (-2.41)
-0.247 (-2.82)
GKF*OPEN -0.002 (-1.57)
-0.001 (-1.17)
GKF*SCHOOL 0.186 (0.62)
-0.050 (-0.19)
Obs. 92 81 81 81 81 81 81
R2 0. 266 0.502 0.618 0.550 0.524 0.505 0.686
F stat. 6.35 6.94 9.73 7.36 6.61 6.14 9.30 * See table 2 for the definitions of variables. **The t-statistics are reported in parentheses.
30
Table 6: Developed Countries: Fixed Effect Panel Data Estimates
Dependent Variable: Growth Rate of Output per Worker Independent Variables
(1)** (2)
(3)
(4)
(5)
(6) (7)
C 1.204 (2.99)
2.124 (3.45)
2.182 (3.50)
2.163 (3.54)
2.343 (3.85)
2.279 (3.60)
2.785 (4.45)
YL -0.108 (-2.84)
-0.197 (-3.27)
-0.202 (-3.32)
-0.201 (-3.36)
-0.219 (-3.68)
-0.213 (-3.42)
-0.265 (-4.30)
GKD 0.267 (3.19)
0.327 (3.56)
0.306 (3.11)
0.301 (3.25)
0.333 (3.70)
0.342 (3.67)
0.302 (3.09)
GKF 0.032 (2.84)
0.031 (2.54)
0.022 (1.22)
0.051 (2.97)
-0.026 (-1.03)
0.126 (1.34)
0.152 (1.52)
GICT -0.013 (-0.99)
-0.012 (-0.90)
-0.0008 (-0.06)
-0.022 (-1.65)
-0.013 (-1.03)
-0.006 (-0.41)
SCHOOL -0.033 (-1.02)
-0.019 (-0.59)
-0.018 (-0.55)
-0.016 (-0.51)
-0.014 (-0.44)
-0.006 (-0.18)
0.015 (0.45)
TI -0.005 (-1.41)
-0.005 (-1.45)
-0.005 (-1.45)
-0.003 (-0.99)
-0.005 (-1.41)
-0.003 (-1.00)
OPEN 0.0009 (2.23)
0.0009 (2.26)
0.0009 (2.23)
0.0009 (2.41)
0.001 (2.43)
0.001 (2.91)
GOVSH -0.001 (-1.56)
-0.0012 (-1.59)
-0.001 (-1.51)
-0.001 (-1.50)
-0.001 (-1.55)
-0.001 (-1.46)
GKF*GKD 0.247 (0.63)
0.271 (0.71)
GKF*GICT -0.123 (-1.65)
-0.174 (-2.34)
GKF*OPEN 0.0008 (2.51)
0.0009 (3.01)
GKF*SCHOOL -0.087 (-1.02)
-0.156 (-1.81)
Obs. 158 154 154 154 154 154 154
R2 0.161 0.228 0.230 0.245 0.266 0.235 0.313
F stat. 6.23 4.47 4.00 4.33 4.85 4.09 4.44 * See table 2 for the definitions of variables.
** The t-statistics are reported in parentheses.
31
Table 7: Robustness checks: Arellano-Bond Estimates
Dependent Variable: Growth Rate of Output per Worker
Independent Variables
One-Step Results
Two-Step Results
C -0.001 (-0.38)**
-0.002 (-1.56)
YL -0.300 (-3.17)
-0.195 (-2.18)
GKD 0.204 (6.43)
0.215 (5.60)
GKF 0.302 (5.06)
0.288 (2.63)
GICT 0.043 (3.09)
0.041 (5.15)
SCHOOL 0.042 (0.83)
0.071 (1.79)
TI -0.003 (-0.62)
0.002 (0.41)
OPEN 0.001 (3.20)
0.001 (5.02)
GOVSH 0.0002 (0.21)
-0.0001 (-0.32)
GKF*GKD -1.493 (-9.92)
-1.399 (-5.39)
GKF*GICT -0.330 (-5.12)
-0.389 (-7.01)
GKF*OPEN 0.0002 (0.52)
0. 0005 (1.13)
GKF*SCHOOL -0.182 (-2.80)
-0.197 (-2.15)
Obs. 124 124 Wald stat. 373.74 11481.19 Sargan test (p-value)*** 0.370 1.000 Autocovariance test of order 2 (p-value)**** 0.249 0.389
* See table 2 for the definitions of variables. ** The z statistics are reported in parentheses.
*** The null hypothesis is that the instruments used are not correlated with the residuals. **** The null hypothesis is that the errors in the first-differenced regression exhibit no second order serial correlation.
32
Table 8: Joint Effects of FDI and Schooling
ENTIRE PANEL
SCHOOLING<0.93 SCHOOLING>0.93 FDI PER WORKER
GROWTH<0.13 0.002 0.016
FDI PER WORKER GROWTH>0.13
0.037 0.023
DEVELOPING COUNTRIES SCHOOLING<0.70 SCHOOLING>0.70
FDI PER WORKER GROWTH<0.15
0.005 -0.008
FDI PER WORKER GROWTH>0.15
0.058 0.03
DEVELOPED COUNTRIES SCHOOLING<1.10 SCHOOLING>1.10
FDI PER WORKER GROWTH<0.11
0.017 0.012
FDI PER WORKER GROWTH>0.11
0.016 0.022
* Headings of rows and columns display average FDI and schooling rates. Numbers in cells are average productivity rates of the samples belonging in each case.
33
Table 9: Comparative Analysis of Foreign Versus Domestic Capital
ENTIRE PANEL
DEVELOPING COUNTRIES
DEVELOPED COUNTRIES
SHARE OF FOREIGN CAPITAL* 0.05 0.10 0.02 SHARE OF DOMESTIC CAPITAL* 0.95 0.90 0.98 RATIO OF FOREIGN TO DOMESTIC CAPITAL 0.26 0.59 0.02 FOREIGN CAPITAL COEFFICIENT** 0.29 0.26 0.15 DOMESTIC CAPITAL COEFFICIENT** 0.16 0.17 0.30 * Shares are calculated over total capital (domestic capital + foreign capital). ** The coefficient estimates are taken from columns with the highest R2 value.
34