Journal of Macroeconomics 32 (2010) 858–871

Contents lists available at ScienceDirect

Journal of Macroeconomics

journal homepage: www.elsevier .com/ locate/ jmacro

Foreign direct investment, human capital and non-linearitiesin economic growth

Constantina Kottaridi a, Thanasis Stengos b,*

a Department of Economics, University of Peloponnese, School of Management and Economics, Greeceb Department of Economics, University of Guelph, Guelph, Ont., Canada N1G 2WI

a r t i c l e i n f o

Article history:Received 24 February 2009Accepted 17 January 2010Available online 4 February 2010

JEL classification:O47

Keywords:Cross-country growth regressionsFDIHuman capitalSemi-parametric model

0164-0704/$ - see front matter � 2010 Elsevier Incdoi:10.1016/j.jmacro.2010.01.004

* Corresponding author.E-mail addresses: ckottaridi@yahoo.com (C. Kott

1 This debate will be presented analytically later i2 For theoretical models please refer to Findlay (1

(2005).3 For surveys please refer to de Mello (1997), Kum

a b s t r a c t

This paper makes a contribution to the existing literature on the foreign direct investment(FDI) and economic growth nexus by contrasting past empirical evidence and conventionalwisdom and arriving at some interesting new results. By applying non-parametric meth-ods, and thus taking into account non-linear effects of initial income and human capitalon economic growth, we explore the FDI effect on growth in much greater detail than pre-vious studies. Our findings not only confirm the non-linear effect of human capital in thepresence of FDI inflows but also suggest that FDI inflows are growth enhancing in the mid-dle-income countries while there is a ‘two-regime’ FDI effect for high-income countries.This new finding appears to be independent of OECD country membership.

� 2010 Elsevier Inc. All rights reserved.

1. Introduction

The role of foreign direct investment (FDI) in the growth process has been (and still is) the subject of long and intensedebate.1 Although this debate has provided many potential insights into the relationship between FDI and growth, it is true thatmost of the existing theory provides contradicting predictions about their relationship.2 FDI is considered a vehicle throughwhich new ideas, advanced techniques, technology and skills are transferred across borders and provide substantial spillovereffects. Within the framework of new growth theories, that stress the effect of technological progress on long-run growth rates,FDI should be considered an important factor boosting growth. There is a body of literature that analyses the effect of FDI ongrowth and another concentrating on knowledge spillovers to domestic firms3. The existing empirical evidence also seems to becontradictory: firm-level studies of particular countries often conclude that FDI is not beneficial to growth and also fail to obtainpositive spillover effects to domestic enterprises. On the other hand, country-wide studies examining the effect of FDI inflows inthe growth process of countries usually provide positive results, especially in specific environments.

All the above are of immediate practical interest for developing and least developed countries (LDC), which often lack thenecessary background in terms of education, infrastructure, economic and political stability in order to be able to innovateand generate new discoveries and designs. In this vein, FDI and its agents, Multinationals Corporations (MNCs) may conceiv-ably help technological advancement domestically. On the other hand, developing countries and LDCs lack the necessary

. All rights reserved.

aridi), tstengos@uoguelph.ca (T. Stengos).n the literature review.978), Walz (1997), Markusen and Venables (1999), Baldwin et al. (2005), Dinopoulos and Segerstrom

ar and Siddharthan (1997) and Saggi (2000).

C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871 859

environment, hence they are not able to reap the benefits associated with FDI and as a consequence they are only used asplatforms for MNCs to promote their own benefit by establishing rent-seeking activities. Moreover, the presence of MNCsmay adversely affect domestic firms, given the market power of their proprietary assets such as technology, superior brandnames and aggressive marketing techniques and as a result, FDI may crowd-out domestic investment.

Our findings contradict the bulk of existing parametric evidence which suggests that a beneficial effect of FDI to economicgrowth exists only for countries with a minimum threshold of absorptive capacity, i.e. countries at higher levels of income. Incontrast to this we find that FDI inflows have a non-linear effect on growth. These new findings appear to be independent ofOECD country membership.

Although there are a number of studies that bring up the issue of non-linear effects of FDI on growth, they are imposingspecific restrictions as to the non-linearity on the grounds of human capital, level of development, financial development anddegree of openness to trade, by simply incorporating interaction terms in a linear regression framework, or splitting the sam-ple of countries into groups according to the above. Instead, in our study we impose no such prior restrictions on the poten-tial non-linearity of FDI on economic growth and we resort to non-parametric techniques, thus surpassing the existingcriticism on parametric econometric specifications.

Apart from this methodological contribution, this paper makes another two contributions to the relevant literature; first,we confirm that the recently established non-linear effects of human capital on growth still hold in the presence of FDI inflows;second, we allow for possibly non-linear human capital effects along with non-linear FDI effects on growth. Hence, we ‘‘test”for the joint effect and interaction of FDI and human capital on economic growth. In our analysis we use a wide range of coun-tries, both developed and developing in order to be able to distinguish potential differential effects between the two groups.

Our results have potentially interesting policy implications. The non-linearity appearing in the relationship indicates a dif-ferential impact of FDI on growth which does not necessarily hold on the basis of the countries’ human capital absorptivecapacity. Rather, this study suggests that the relationship is much more complex than that since the human capital itself ex-erts also a non-linear effect on economic growth. This may signal the need for a more specialized analysis and policy designwithin each country since (i) FDI may take place in very different sectors/industries among countries on the one hand and onthe other hand even if it is in the same sectors/industries it might exhibit different productivities (ii) the evidence is also con-sistent with Durlauf and Johnson (1995) pointing to a model in which countries pass through distinct phases of developmenttowards a unique steady state. That is, at a given time interval, countries display differences in their growth characteristics intheir transition to a high growth position (Galor, 2005) and this is reflected in the observed non-linearities in the data.

The rest of the paper is organized as follows: Section 2 discusses related literature; Section 3 presents the benchmarkmodel and data, Section 4 analyses the non-linear models and estimation. Section 5 then presents empirical results and fi-nally Section 6 concludes.

2. Literature review

As mentioned above, firm-level studies usually fail to reach positive growth or positive spillover effects into the host na-tion. Among those we find the influential study of Aitken and Harrison (1999) for Venezuela, Haddad and Harrison (1993) fora number of developing countries, Kokko (1994) for Mexico regarding industries where foreign affiliates exhibit higher pro-ductivity and a larger market share than the domestic firms. In other industries though, she finds positive effects betweenforeign presence and local productivity. Kokko et al. (1996) for Uruguay and Kathuria (2001) for India find similar results. Anaffirmative positive affect is suggested in Blomström (1986) for Mexico.

The literature is much richer in the macroeconomic context. Positive effects of FDI on growth or productivity spillovers areattributed to De Gregorio (1992) for 12 Latin–American countries, Blomström et al. (1992) for 78 developing countries,Blomström et al. (1994) for a sample of both developed and developing countries,4 Bende-Nabende and Ford (1998) for Taiwan,Zhang (2001) for the majority of East-Asian economies and Latin America and Baldwin et al. (2005) for nine OECD countries.

Another line of research points to a differential impact between developed and developing countries, for example, DeMello (1999) for 15 OECD and 17 non-OECD countries for the period 1970–1990 and Xu (2000) for US FDI in 40 countriesfor the period 1966–1994.

Finally, there is an array of works that stress the positive role of FDI conditioned on adequate local factors,5 especiallyhuman capital. Borenztein et al. (1998) in their study of 69 developing economies for 1970–1989 concluded that the effectof FDI is dependent on the human capital stock. Bengoa and Sanchez-Robles (2003) reached the same conclusion for LatinAmerica based on economic stability and liberalized financial markets.

All relevant studies discussed above regarding the growth enhancing role of FDI based on local ‘‘absorptive capacity”, im-pose restrictions as to the type of non-linearity and are confined to parametric techniques by simply incorporating interac-tion terms in their regressions or by splitting the sample of countries into groups to test such a hypothesis.

Recently, two studies emerged to contradict the majority of macroeconomic evidence of a beneficial effect. Carkovic andLevine (2005) criticized existing empirical studies as not fully controlling for simultaneity bias, country-specific effects and

4 However, when they split their sample of developing countries into two groups based on their level of income per capita they found that FDI was notstatistically significant for lower income developing countries although it remained positive.

5 Balasubramanyam et al. (1996) examined a number of developing countries for 1970–1985 and concluded that FDI is enhancing for those that follow anexport oriented trade policy regime, Alfaro et al. (2003) found growth enhancing effects of FDI in economies with sufficiently developed financial markets.

860 C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871

the use of lagged dependent variables in their growth regressions. They use GMM techniques and they assess the FDI-growthrelationship for 72 countries covering the period 1960–1995. Their findings suggest that FDI does not exert an independentinfluence on economic growth. Durham (2004) also examined 80 countries between 1979 and 1998 using extreme boundanalysis and failed to achieve a robust positive effect.

Though a large portion of studies stresses the particular role of human capital for FDI to be beneficial to host countries, thecontribution of human capital to growth is controversial in its own. Whereas at the micro level there is consistent evidencethat education raises incomes significantly,6 evidence at the macro level has been mixed. Studies such as Barro (1991), Bilsand Klenow (2000), Mankiw et al. (1992) and others use enrollment rates for primary and secondary education and pointtoward a positive and significant effect. Benhabib and Spiegel (1994), Kyriacou (1991), Lau et al. (1991) and Pritchett(2001) on the other hand find an insignificant or even negative result for the stock of human capital, i.e. the total means yearsof schooling. Some authors (Barro, 1998; Barro and Sala-i-Martin, 1995) incorporate differentiated measures of human cap-ital not only by level of education but also by gender. Regarding the time dimension of growth, it is found that as the fre-quency of changes over which growth rates are calculated increases there is less evidence of a positive effect of humancapital accumulation on growth (Krueger and Lindahl, 2001; Islam, 1995).

The vast majority of existing studies to the empirics of economic growth have assumed that human capital exerts thesame effect on economic growth both across countries and across time and have assumed a (log) linear relationship. Moti-vated by theories emphasizing threshold externalities (Azariadis and Drazen, 1990) several researchers have questioned thisassumption. Durlauf and Johnson (1995) and Masanjala and Papageorgiou (2004) use the regression tree and the thresholdregression methodology to show the existence of multiple regimes.

Liu and Stengos (1999) allow for two non-linear components, one for the initial GDP level and the other for the secondaryenrollment rate. Kalaitzidakis et al. (2001) use semi-parametric techniques and find that there are substantial non-linearitiesin the growth-human capital nexus. Kourtellos (2003) also uses a semi-parametric smooth coefficient model to study a localgeneralization of the Solow model in the spirit of Durlauf et al. (2001). More recently, Mamuneas et al. (2006) estimate ageneral model of the economic growth process for 51 countries during 1971–1987 by allowing the contribution of both tra-ditional inputs and human capital to vary across both countries and time and find that the average output elasticity of hu-man capital varies substantially across countries and above that in some cases the estimate is negligible.

The aim of our study is to detect potential non-linearities in the FDI-economic growth relationship in the presence of hu-man capital as we attempt to check whether the non-linearities of the human capital effect found in the literature still holdsin the presence of FDI. In addition we also try to assess the joint effect of FDI and human capital given that the majority of therelated literature points to a positive sign.7

3. Benchmark model and data

Following the standard approach in the literature we allow for investment to be divided between its domestic and its for-eign direct component.

We let Y, Kd, Kf , H and L represent total output, domestic physical capital stock, foreign physical capital stock, human cap-ital stock and labor respectively and we also allow for A to be a technological parameter. Technology is assumed to grow expo-nentially at the rate g, or At ¼ A0egt . We define kd as the stock of domestic capital per effective unit of labor, kd ¼ Kd=AL and kf

as the stock of foreign capital per effective labor, kf ¼ Kf =AL, and y as the level of output per effective unit of labor, y ¼ Y=AL.Hence, in common with most recent contributions we employ panel data estimations and estimate the following model:

6 Com7 Tw8 Equ

the diredirect iborrow

yit ¼ a0 þ a1Dj þ a2 lnðIdit=YÞ þ a3 lnðnitÞ þ a4ðln xitÞ þ a5If

it=Y þ a6hit þ eit ð1Þ

where yit refers to the growth rate of income per capita during each period, Dj is a dummy variable for regions, nit is the pop-ulation growth, xit is per capita income at the beginning of each period, Id=Y is the domestic investment taking place in theeconomy, If =Y foreign direct investment and hit is human capital.

To check for the joint effects of FDI and human capital according to the majority of literature we re-estimate the abovemodel adding their interaction, hence the model becomes:

yit ¼ a0 þ a1Dj þ a2 lnðIdit=YÞ þ a3 lnðnitÞ þ a4 lnðxitÞ þ a5If

it=Y þ a6hit þ a7Ifit � hit þ eit ð2Þ

We follow the recent literature and we measure human capital as total mean years of schooling. The human capital datawere obtained from the Barro and Lee (2001) data base and they were used in Savvides and Stengos (2009).

Foreign direct investment is obtained by United Nations Cooperation on Trade and Development (UNTAD). FDI inflowscomprise capital provided (either directly or through other related enterprises) by a foreign direct investor to a FDIenterprise. FDI includes the three following components: equity capital, reinvested earnings and intra-company loans.8 Data

monly referred to as Mincerian wage regressions.o influential papers claim that such joint impact does not exist (they both employ parametric techniques) (Durham, 2004; Carkovic and Levine (2005)).ity capital is the foreign direct investor’s purchase of shares of an enterprise in a country other than that of its residence. Reinvested earnings comprisect investor’s share (in proportion to direct equity participation) of earnings not distributed as dividends by affiliates or earnings not remitted to thenvestor. Such retained profits by affiliates are reinvested. Intra-company loans or intra-company debt transactions refer to short- or long-terming and lending of funds between direct investors (parent enterprises) and affiliate enterprises.

C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871 861

on FDI flows are presented on net bases (capital transactions’ credits less debits between direct investors and their foreignaffiliates). Net decreases in assets or net increases in liabilities are recorded as credits (with a positive sign), while net increasesin assets or net decreases in liabilities are recorded as debits (with a negative sign). Hence, FDI flows with a negative sign indi-cate that at least one of the three components of FDI is negative and not offset by positive amounts of the remaining compo-nents. These are called reverse investment or disinvestment.9 All other data we have used regarding GDP per capita, GDP percapita growth, gross fixed capital formation measuring domestic investment are in constant 2000 US$ and are obtained from theWorld Development Indicators (WDI) of the World Bank.10 Population data are also taken form the WDI.

The sample we are using covers the period 1970–2004. In order to lessen the impact that year to year fluctuations in outputhave on our results, we take five-period averages according to most empirical growth literature (Maasoumi et al. 2007; Dur-lauf et al. 2008; Henderson et al. 2009) hence we end up with seven time points. We incorporate a wide range of developedand developing countries, in particular 25 OECD countries and twenty non-OECD countries from all over the world, represent-ing all regions. The selection of developing countries was based on the availability of the data especially with regards to thehuman capital variable. A full list of the countries and the regions they belong to may be found in Appendix A.

4. Non-linear model specification and estimation

Parametric estimates, such as those in Eq. (1), assume a unique response coefficient for human capital and FDI in growthregressions. Some works, however, have indicated that this assumption is not warranted. Azariadis and Drazen (1990), Dur-lauf and Johnson (1995) and Murphy et al. (1989) point to the possibility of threshold effects in the growth process, the for-mer focusing on thresholds in human capital. Alternatively, the growth experience is a non-linear function of human capital.While non-linearities in the convergence process have been extensively discussed in the literature11, remarkably little hasappeared in connection with human capital and nil to the best of our knowledge regarding the effect of FDI on economic growth.

We hereby use a semi-parametric partially linear regression (PLR) specification of the growth regression function. In con-trast to a standard linear parametric formulation, a semi-parametric PLR specification is an adequate representation for thedata. A particular version of the PLR model would be to impose a more stringent additive separable structure on the un-known components. In that case the semi-parametric partially linear additive regression (PLAR) specification of the modelin (1) can be written as follows:

9 For10 Nev

from th11 See12 We

these si

yit ¼ aþ XTitcþ

Xp

s�1

gsðZitÞ þ uit ð3Þ

Where Xit is a vector of dimension q, c is a q � 1 vector of unknown parameters, and the gs(�)-functions are unknownfunctions of each individual Z. In the context of Eq. (3) Xit = fDj; I

kit=Y;nitg and p = 3, where Z1it refers to initial income xit ,

Z2it refers to human capital and Z3it to FDI. Linton and Nielsen (1995), Fan et al. (1998) and Fan and Li (2003) use marginalintegration to estimate the components of the PLAR model in equation (3). In our case we estimated such a specificationusing marginal integration as in Fan and Li (2003), but the above specification (3) was rejected in favor of the more generalform given by12:

yit ¼ XTitcþ gðZitÞ þ uit ð4Þ

Where Xit is a vector of dimension q, c is a q � 1 vector of unknown parameters, Zit is a vector of dimension p and gð�Þ is anunknown function. As in the case of (3), in the context of Eq. (4), Xit ¼ fDj; I

kit=Y;nitg and Zit ¼ fZ1it; . . . ; Zpitg, with p = 3

where Z1it refers to initial income xit , Z2it refers to human capital and Z3it to FDI. The main difference in (4) is that wenow allow for a full interaction among the Z’s as we no longer impose additive separability as in (3). Robinson (1988)provided a way of obtaining a

ffiffiffinp

-consistent estimator of the parameter vector c by concentrating out the influence ofthe nuisance variables, the Zs. This is accomplished by conditioning them through kernel methods and estimating theconditional expectations EðyitjZitÞ and EðXitjZitÞ. In the second stage of a two-step estimation procedure, the kernel esti-mates of EðyitjZitÞ and EðXit jZitÞ are used to estimate c. The semi-parametric PLR model is estimated using a densityweighted approach to avoid the use of a trimming parameter and the bandwidth is chosen with cross-validation, see Liand Racine (2007).

Using the estimates of c we can redefine the dependent variable in (4) and directly estimate by kernel methods theunknown function gð�Þ. This function includes all interactions between the Z’s. We will only concentrate on the estimatesof (4), where the graphical representation of the different components will take place at the mean value of the otherremaining components. The bandwidth parameter used in the non-parametric kernel estimation will be obtained bycross-validation.

more detailed information please refer to the UNCTAD World Investment Report 2005: Transnational Corporations and the Internationalization of R&D.ertheless, we have used other data sources for these variables for robustness purposes. We have also experimented with the corresponding variablese Penn World Tables 6.2 where the variables are given in 2000 PPP US$. Results were similar.for example Quah (1996).test the adequacy of the specification of the PLAR model in (3) by adding simple interactions of the Z’s in the linear part. We strongly rejected (3) as

mple interactions between the different Z’s were highly significant. For a formal test of additivity in a non-parametric setting, see Sperlich et al. (2002).

Table 1LSDV regressions of equation (1) and GMM estimation of total sample, 5-year averages Dependent variable: GDP per capita growth, cross-section weightedpanel.

(1)All

(2)OECD

(3)Non-OECD

(4)High-income

(5)Middle-income

(6)Low-income

GMM all

x �0.626*** �1.408*** �0.584 �1.850*** �1.392*** �0.112 �0.7542***

�4.823 �4.878 �1.595 �8.216 �8.354 �0.120 �3.0472n 0.093 0.249 �2.276*** 0.424 �2.251*** 5.342 �0.0024

0.251 0.732 �3.725 1.429 �2.849 1.039 �0.0054Id/Y 2.010*** 1.887*** 1.349** 1.254*** 1.154*** 3.799* 1.1026***

7.024 5.011 2.233 4.381 3.745 1.822 5.8408If /Y 0.027 �0.009 0.254*** 0.029 0.478*** 0.975 �0.0071

0.686 �0.575 3.016 1.157 3.622 1.384 �0.2602h 0.122 0.263*** 0.027 0.242** �0.357 1.360* 0.4915***

1.127 3.457 0.045 2.417 �1.619 1.780 6.4679Africa �0.373 3.683*** 11.467*** �22.028*** 2.2292

�0.416 2.809 10.082 �4.438 1.4650America 1.355 7.923** 14.493*** 1.1598

1.511 2.453 5.566 0.6362Asia 1.356* 7.749** 5.279*** 15.533*** 13.428*** �17.410*** 1.8351

1.878 2.481 4.126 5.989 10.209 �2.879 1.1818EU 1.077 7.963** 14.670*** 1.2816

1.105 2.485 5.769 0.7727Latin America �0.411 6.325** 3.340** 12.247*** 0.3869

�0.421 2.075 2.406 7.240 0.2636Oceania 0.861 7.505** 13.792*** 1.8896

0.835 2.305 5.464 1.0128Other-Eur 1.608 8.594*** 15.481*** 1.7664

1.738 2.642 6.124 0.922

Obs 309 173 136 166 115 28 84R-squared 0.260 0.282 0.371 0.491 0.497 0.288 0.837Adjusted R-squared 0.233 0.238 0.337 0.459 0.464 0.224 0.812SE of regression 2.312 2.056 2.521 2.002 2.051 1.984 2.452Sargan test (p-value) 0.430

x, n, Id/Y are in logs, regressions with white heteroscedastic standard errors, t-statistics are shown in parentheses.* Significant at 10% level.** Significant at 5% level.*** Significant at 1% level.

862 C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871

5. Empirical results

We begin our discussion with the parametric estimates which are given in Tables 1 and 2. We split the sample betweenOECD and non-OECD countries in order to detect possible membership-based differentiation. We have also split the sampleamong countries classified by the World Bank as high-income, middle-income and low-income.13

Table 1 presents results of model 1. The coefficient estimates for initial GDP per capita and domestic investment are of theanticipated sigh and significance and are robust to the alternative sub-samples used. Coefficient estimates of populationgrowth are most of times of the wrong sign (positive) though non-significant, while in two sub-samples they emerge withthe correct sign and significance (non-OECD and middle-income countries). The estimates of FDI turn out to be positive andinsignificant for the whole sample, however surprisingly, they are negative yet insignificant for the OECD and positive butinsignificant for high-income economies. On the contrary, they are positive and significant for the non-OECD sample and forthe middle-income countries. For low-income countries the coefficient estimate is as expected, positive but non-significant.The estimates for human capital follow a more expectable pattern: they are positive and highly significant for the OECD andhigh-income economies and significant for the low-income ones.14 For comparison purposes, the last column of Table 1 pre-sents results obtained through GMM estimation of the entire sample. Results are qualitatively the same with the simpler LSDVregression, with the exception that human capital is now very significant. FDI turns out non-significant again.15

Turning to Table 2 where the estimations include the interaction effect between FDI and human capital, again, coefficientestimates with regard to initial GDP per capita and domestic investment are as before and actually of very close magnitude.Nevertheless, we find some differences for FDI and human capital. Human capital gains significance now for the total sample

13 We have also separated the middle-income countries between those of upper–middle income and lower–middle income. We then have grouped upper–middle income with high income and lower–middle income with low-income.

14 It is not expectable for the low-income countries; however, one might think that even lower human capital bases are beneficial for these countries’ growth.15 We have used as instruments past values of all variables incorporated in the model to account for possible endogeneity according to Brock and Durlauf’s

(2001) discussion of the fact that most variables in a growth regression are endogenous because of the fact that growth theories are open–ended. The Sargantest of over identifying conditions indicated that the set of instruments is valid.

Table 2LSDV regressions of equation (2) and GMM estimation of total sample – 5 year averages Dependent variable: GDP per capita growth, cross-section weightedpanel.

(1)All

(2)OECD

(3)non-OECD

(4)High-income

(5)Middle-income

(6)Low-income

GMM all

x �0.659*** �1.255*** �0.361 �1.801*** �1.385*** �0.005 �0.663***

�6.686 �4.488 �0.821 �7.492 �8.367 �0.007 �2.572n 0.141 0.072 �2.101*** 0.413 �2.313*** 2.058 0.072

0.389 0.190 �4.513 1.262 �3.033 0.353 0.158Id/Y 1.920*** 1.932*** 1.208** 1.317*** 1.159*** 3.179 1.404***

6.759 4.922 2.090 4.106 3.636 1.545 3.358If/Y 0.324 �0.569** 1.030*** �0.292* 0.828** 7.930** �0.259

1.199 �2.237 2.577 �1.804 2.060 2.584 �0.902h 0.168** 0.134* �0.032 0.098 �0.294 2.139** 0.416***

2.041 1.868 �0.157 1.417 �1.380 2.297 3.803If/Y

*h �0.032 0.057** �0.102** 0.032* �0.052 �2.456** 0.027�1.153 2.058 �2.088 1.828 �1.054 �2.639 0.910

Africa �0.301 1.976 11.015*** �20.631*** 1.266�0.414 1.220 9.710 �3.477 0.672

America 1.460** 7.703** 15.100*** 0.1251.908 2.322 5.016 0.058

Asia 1.386** 7.497** 3.813** 15.974*** 12.994*** �15.162*** 0.7342.269 2.316 2.473 5.213 9.893 �3.091 0.366

EU 1.287 7.410** 14.963*** 0.1611.514 2.277 5.189 0.078

Latin America �0.367 6.141** 2.019 11.869*** �0.561�0.436 1.977 1.352 7.575 �0.306

Oceania 0.973 7.208** 14.122*** 0.8031.104 2.177 4.882 0.360

Other-Eur 1.793** 8.120** 15.831*** 0.6462.323 2.448 5.462 0.283

Obs 309 173 136 166 115 28 84R-squared 0.270 0.295 0.428 0.441 0.505 0.593 0.798Adjusted R-squared 0.241 0.247 0.392 0.405 0.468 0.451 0.763SE of regression 2.309 1.997 2.471 1.992 2.055 2.427 2.422Sargan test (p-value) 0.463

x, n, Id/Y are in logs, regressions with white heteroscedastic standard errors, t-statistics are shown in parentheses.* Significant at 10% level.** Significant at 5% level.*** Significant at 1% level.

C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871 863

whereas the interaction term turns out to be negative but insignificant. What we can observe overall is that we get oppositesigns between the FDI and the interaction term. That is to say, when FDI is positive and significant, the interaction is negativeand significant and vice versa. For the OECD sample, FDI comes out negative and highly significant, human capital positiveand significant and their interaction positive and significant. Exactly the opposite happens for the non-OECD sample thoughthe human capital is negative now but not significantly different from zero. It is noteworthy that FDI is again highlysignificant and positive for the middle-income countries as in Table 1 while human capital and the interaction term arenon-significant. It comes as a surprise that for the low-income economies both FDI and human capital are very significantwhile their interaction though significant emerges with a negative sign and highly significant. The last column presentsresults from GMM estimation of the whole sample. Again, we see that results are qualitatively the same with the simpleLSDV regression.

All in all, results are quite mixed and non-definitive as pointed elsewhere in the literature. Contrary to the majority of stud-ies, we do not obtain beneficial effects of FDI for the developed countries (OECD and high-income samples) whist we do so forthe middle-income group and non-OECD economies. Also, it appears that in some groups FDI and human capital do not workwell together, rather they are detrimental for economic growth whilst in other groups results pinpoint reinforcing effects ofFDI when combined with human capital and negative effects of FDI alone. These results are indicative of the complicated rela-tionships of both FDI and growth and human capital and growth as has been explained in the previous section.

In this regard, our parametric evidence contrasts a number of studies that pinpoint to FDI-led growth in developed butnot developing or less developed countries; nevertheless it is in accordance with the bulk of mixed empirical evidence asdiscussed in the literature review and in line with the two more recent parametric contributions discussed in previous sec-tions (Durham, 2004; Carkovic and Levine, 2005); It is also clearly very difficult to detect non-linearities in FDI with the para-metric specification.

As noted previously, parametric estimates assume a unique response coefficient for FDI and human capital in growthregressions. We proceed to test the main parametric specification using the linearity test of Li and Wang (1998), which isa bootstrap version of the test for the functional form of Zheng (1996). The linear specifications of the first column of Table1 and 2 respectively were rejected with p-values 0.001 and 0.023, respectively.

Table 3Semiparametric model – Dependent variable: GDP per capita growth.

n �0.665(�1.167)

Id/Y 1.311***

(3.586)Africa �0.861

(1.174)Asia 1.479**

(2.300)EU 0.420

(1.097)Latin America �0.584

(1.242)OceaniaOECD 0.842

(0.931)

x (25th quartile) �2.145**

(2.173)x (50th quartile) �0.524

(2.036)x (75th quartile) �0.051**

(1.613)If/Y (25th quartile) 0.002

(1.400)If/Y (50th quartile) 0.253**

(2.081)If/Y (75th quartile) 1.198**

(2.231)h (25th quartile) 0.035*

(1.648)h (50th quartile) 0.442**

(2.055)h (75th quartile) 2.617**

(2.326)

Robust t-statistics are shown in parentheses.* Significant at 10% level.

** Significant at 5% level.*** Significant at 1% level.

864 C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871

The estimation of the semi-parametric PLR specification allow us to obtain graphical representations of the non-paramet-ric components: initial GDP per capita, FDI and human capital. Before proceeding with these graphical representations wenote that, following Fan and Li (1996), we also tested the PLR specification (4) for the model that applies to the whole sampleagainst a more general non-parametric model that would allow for a general non-parametric regression function under thealternative. We failed to reject the null of the semi-parametric model with a p-value of 0.35. Table 3 below presents thesemi-parametric estimates of the variables that enter the linear part of model (4), that is, all the variables other than initialincome, human capital and FDI that enter the non-linear part. A comparison with the parametric estimates in Tables 1 and 2shows that although the two sets of estimates differ, they are qualitatively similar. The lower panel of Table 3 presents thederivatives of the non-parametric components (slope coefficients) at the 25th, 50th and 75th quartile along with robustt-statistics. These slope estimates were obtained as derivative estimates from local linear estimation of the non-parametricfunction of equation (4). The slope estimates in most cases are statistically significant and are quite different from each otherat different sample points refuting the idea of a constant slope coefficient that underlies parametric estimation.

Our semi-parametric results of the PLR model are presented in Figs. 1–5. Fig. 1 shows a representative semi-parametric fitfor initial income. The horizontal axis shows the logarithm of initial income per capita and the vertical axis the growth rateg1ðZ1itÞ in standardized form along with 95% confidence intervals. The graph is drawn at the mean values of the other twovariables that enter the unknown function g(�), namely FDI and human capital.16

This is consistent with recent empirical evidence (see Durlauf and Johnson, 1995; Liu and Stengos, 1999; Pack and Page,1994; Quah, 1996) on convergence. The curvature of the graph implies that, on average, middle-income countries experiencethe highest growth rates; it appears that very high-income countries get the lowest growth rates.

Fig. 2 shows the estimate of the non-parametric component of human capital. The horizontal axis shows total mean yearsof schooling and the vertical axis g2ðZ2itÞ in standardized form. Again the graph is shown at the mean values of the otherremaining variables, initial GDP per capita and FDI and the confidence bands are computed using 199 bootstraps. Previousstudies suggesting a non-linear relationship to economic growth are confirmed: there are clearly thresholds in the effect ofhuman capital. It is evident that for mean total years of schooling falling between 5 and 11, there is no effect on growth,

16 The confidence band is computed using 199 bootstrap values.

8.8

9.2

9.6

10

10.4

10.8

4 6 8 10 12 14 16log(h)

g2(Z2)

Fig. 3. GDP growth and human capital – high-income countries.

-0.6-0.20.20.61

1.41.82.22.63

5 6 7 8 9 10 11

log(x)

g1(Z1)

Fig. 1. GDP growth and initial GDP.

-0.50

0.51

1.52

2.53

3.5

0 2 4 6 8 10 12 14 16log(h)

g2(Z2)

Fig. 2. GDP growth and human capital.

C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871 865

while at low and very high levels it is beneficial. This result is in line with Kalaitzidakis et al. (2001) where they found that formean years of schooling between 0.9 and 4.4 the relationship is positive with a slight differentiation in what we find here,i.e., that the positive effect holds between approximately 1 and 5 mean years of schooling. The effect of human capital turnsout positive again for above 11 years of schooling.

The above results indicate that for low human capital countries, increases in schooling promote their growth. For middlehuman capital countries, it appears that the effect is non-productive in line with Kalaitzidakis et al. (2001). This neutral re-sult is counter to the widely held belief that higher education translates into higher wage and productivity benefits in themarket place. A plausible explanation may involve rigidities and unproductive practices in the labor markets that inhibithigher educational levels to translate into higher wage employment. Kalaitzidakis et al. (2001) found a similar result forthese levels of female education, raising an issue on the discriminatory practices of labor markets. Regarding higher second-ary education (involving average schooling years of 11 and above) the argument may rest on Barro (1998) and Romer’s(1990) suggestion that human capital contributes to growth by facilitating the absorption of new technologies. This effectis more likely to take place into these levels as a result of their complementarities with new technologies.

10.4

10.8

11.2

11.6

12

12.4

0 1 2 3 4 5 6 7 8 9 10log(x)

g2(Z2)

Fig. 4. GDP growth and human capital – middle-income countries.

866 C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871

To enrich our insight on the relationship between human capital and economic growth, we estimate results for the high-income countries versus the middle-income ones.17

The pattern is quite differentiated here; it is apparent that in high-income countries human capital is growth enhancingafter 8 years of schooling whilst lower levels of human capital are counter-productive. This could be because higher-incomecountries are characterized by technology-intensive industries where a higher educational level is required in order for com-plementarities to take place .

For middle-income countries, it emerges that there is a range of mean years of schooling, between 2.5 and 6 that exertsnegative effects on growth whilst for lower and higher human capital than that we get higher economic growth. As Pritchett(2001) suggests, in some lower income/human capital countries investment in education may be directed towards unpro-ductive activities and it might be the case that there are inefficiencies both in the educational system and the labor marketsin this sub-sample of countries.

Figs. 5–7 are the main figures for our paper. They show the estimate of the non-parametric component for FDI share,drawn at the mean values of initial GDP per capita and human capital. The horizontal axis shows the FDI share and the ver-tical axis g3ðZ3itÞ in standardized form. Again the confidence bands are computed using 199 bootstraps. Fig. 5 illustrates thetotal sample case.

The relationship between FDI and output growth appears to be non-linear and, in particular, there appear to be twoapproximate intervals (‘regimes’) of FDI where the effect on growth is linear but with different slopes. The first interval isfor FDI shares up to 8 where the effect is positive but at low rates while for higher FDI shares the effect is much stronger.

To further explore the FDI-growth relationship and for comparison with the parametric results purposes, we proceeded tosplit again the full sample into groups of countries as we have done for the human capital case. Then, we plotted the fittedvalues of Fig. 5 according to these groups.

As with the human capital case, the picture is again differentiated here; we basically observe two regimes, a counter-pro-ductive and a positive one. It appears that only in high FDI shares can the economies reap benefits from foreign productionwhilst for low and middle shares the effect is detrimental. Though this may seem puzzling at a first glance, it can be explainedon the grounds that these countries have experienced high growth rates in the past when growing their economies and theyhave reached a level where further growth is mitigated. Taking into consideration that high-income countries have advancedtechnology and a competitive private sector, our results may indicate substitution effects between domestic and foreign invest-ments, i.e., crowding out effects. This non-parametric result complies with the parametric evidence provided in this study.

Regarding the middle-income economies, we depict a two ‘regime’ result as we did for the entire group however withreverse slopes.18 The figure implies that as disinvestment decreases these countries get more pronounced increases ingrowth than when FDI increases. It is noteworthy though that FDI is always beneficial for this group of countries which isin accordance with our parametric evidence.

The above results come in contrast with conventional wisdom that FDI is growth enhancing in higher-income countriesbased on the rationale that these have the necessary absorptive capacity to reap the benefits of technology and innovationtransferred through FDI. Following this rationale, the conventional wisdom and most parametric results, indicate either a nilor a negative effect of FDI on economic growth for less advanced countries. Non-parametric evidence however suggests thatless developed countries can benefit as well. On top of that, for middle and high-income countries, evidence implies that theeffect of FDI on growth varies at different levels of FDI itself.

It appears that the way FDI affects growth differs across and within countries. The relationship seems to be complex andthe impact varies according to a country’s level of FDI. Following the spirit of the cross-country regression literature (Durlaufand Johnson, 1995; Masanjala and Papageorgiou, 2004; Kourtellos, 2003) parameter heterogeneity may exist in the sensethat the effect of a change in a particular variable is not the same. Different coefficient estimates appear per country for

17 We are not able to test for the low-income countries due to limited number of observations.18 Actually, a closer look at the first ‘regime’ reveals two sub-regimes; up to 10.5 the decrease in the growth rate is much higher than the next interval before

the beneficial impacts begins.

2.4

2.8

3.2

3.6

4

4.4

-6 -2 2 6 10 14 18log(If/Y )

g3(Z3)

Fig. 5. GDP growth and FDI.

11.4

11.6

11.8

12

12.2

12.4

12.6

12.8

0 2 4 6 8 10 12 14 16 18log(If/Y )

g3(Z3)

Fig. 6. GDP growth and FDI – high-income countries.

1313.514

14.515

15.516

16.517

17.5

-4 -2 0 2 4 6 8log(If/Y )

g3(Z3)

Fig. 7. GDP growth and FDI – middle-income countries.

C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871 867

the effect of initial income and human capital on growth and so does FDI. In other words, there exists a different FDI-growthnexus in different countries.

There is an important caveat to our results however. Given the limitations of our data we were not able to explore theissue of endogeneity within the context of the semi-parametric PLR model that we use. Most of the existing econometricliterature relies on sieve methods to carry out estimation as in Ai and Chen (2003) as well as using specific models, suchas the smooth coefficient model of Cai et al. (2006). In our case we do not have enough data to be able to carry out suchan analysis. What we have done is to use lagged values of FDI in place of current values to see whether we obtain similargraphical representation of the effect of FDI on growth and overall we did. This is only a check of whether the correlationbetween current FDI and growth is of a similar nature as that of past FDI and growth.19

How could one explain the above results? The FDI-growth relationship is definitely a complex one and one should bear inmind that the composition of FDI is very different across countries: FDI is directed to different industries and activities hence,the different characteristics of these may be responsible for the different coefficient estimates that emanates in the analysis.

19 For space limitations the graph is not included but it is available from the authors.

868 C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871

Foreign direct investors select a location based on its comparative or competitive characteristics like for example cheaperresources, expertise in a particular area and so on. As these investors enter the country, bringing in new capital and advancedmethods (in production, management, marketing etc.) a country’s resources may be now better exploited on one hand andspillover effects may take place from foreign investors to indigenous ones on the other. For high-income countries, it is verylikely that congestion effects might take place which will be to the detriment of the indigenous firms, hence a crowding outeffect may be in place for a range of FDI. Levels of FDI higher than that can take place in a wider range of activities hence bemore spread and not create any congestion effects. Another plausible explanation may involve rigidities and distortions inmarkets that do not allow FDI to translate into higher productivity and hence income growth which may be overcome athigher levels of FDI.

6. Conclusions

In this paper we study the influence of FDI on the process of economic growth among a wide set of OECD and non-OECDcountries by allowing the impact to differ both across each country and also across each time period. We apply non-para-metric techniques taking into account the previously documented non-linear effects of initial income and human capitalon economic growth. To this end, we verify that initial income and human capital have a non-linear effect on economicgrowth as suggested earlier in the literature even in the presence of foreign investments.

Our basic results are in contrast to the bulk of the existing parametric evidence which suggests that a beneficial effect ofFDI to economic growth exists only for countries with a minimum threshold of absorptive capacity, i.e. countries at higherlevels of income. Contrary to this, we find that FDI inflows have a non-linear effect on growth, they are growth enhancing indeveloping countries and there is a ‘two-regime’ FDI effect for high-income economies.

Our paper parallels the widely discussed issue of non-linearities in convergence. For our purposes, we collected data fromthe WDI of the World Bank, UNCTAD (2005), Barro and Lee (2001) for 25 OECD and 20 non-OECD countries over the period1970–2004. The countries were selected based on their availability of human capital data.

The evidence may be of particular policy interest and it suggests that the FDI-growth nexus is a complex one. Whilst theprevious section lays down some possible explanations for the differential FDI impact, our results certainly open up the floorfor more thorough research, into the issue both in empirical and theoretical basis.

Appendix A. Sample countries

OECD country

Region Non-OECD country Region

Australia

Oceania Brazil Latin America Austria EU Chile Latin America Belgium EU China Asia Canada Canada Colombia Latin America Denmark EU Cote D’Ivoire Africa Finland EU Egypt Africa France EU India Asia Germany EU Indonesia Asia Greece EU Malaysia Asia Iceland Other Europe Morocco Africa Ireland EU Nigeria Africa Italy EU Pakistan Asia Japan Asia Panama Latin America Korea, Rep Asia Paraguay Latin America Mexico Latin America Peru Latin America Netherlands EU Singapore Asia New Zealand Oceania Thailand Asia Norway Other Europe Tunisia Africa Portugal EU Spain EU Sweden EU Switzerland Other Europe Turkey Asia United Kingdom EU United States America

Total = 25

Total = 20

C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871 869

Appendix B. References on the FDI-growth relationship

B.1. Macro studies

Authors

Sample Technique Conclusion

De Gregorio (1992)

12 Latin–American countries(1950–1985)

Panel data

Positive and significant correlationbetween FDI and growth Productivity of FDI larger than thedomestic investment

Blomström et al. (1992)

78 LDCs (1960–1985) Cross-country Positive and significant correlationbetween FDI and growth Impact is larger in countries withhigher-income levels

Blomström et al. (1994)

144 Industries Mexico (1975) Cross-section Positive association between thetechnology inputs of foreignaffiliates and local competitors’investment and output growth

Borenztein et al. (1995/98)

69 LDCs (1970–1989) Cross-country paneldata

FDI enhances growth by helpingtechnological diffusion

Balasubramanyam et al.(1996)

46 LDCs (1970–1990)

Cross-country FDI enhances growth only incountries that implement outwardoriented policies

Bende-Nabende and Ford(1998)

Taiwan

Time series FDI exerts positive effects onoutput

De Mello (1999)

OECD countries and othernon-OECD countries (1970–1990)

Panel data and timeseries

FDI fosters growth if a grade ofcomplementarity betweendomestic investment and FDIexists

UNCTAD (2005)

142 Countries (1970–1995) Panel data FDI influences the next periodgrowth if human capital is present

Xu (1999)

US FDI to 40 countries (20 DCsand 20 LDCs) (1966–1994)

Panel data

Technology transfers from FDIcontributes to productivity growthto developed but not lesserdeveloped countries due to lack ofadequate capital

Zhang (2001)

East-Asian economies andLatin America

Panel data

Key advantage created by FDI torecipient countries is technologytransfer and spillover efficiencybut depends on recipientcountries’ absorptive capabilities

Carkovic and Levine (2005)

72 Countries (1960–1995) Cross-section paneldata

FDI does not exert an independentinfluence on economic growth

Bengoa and Sanchez-Robles(2003)

Latin America

Panel data FDI contributes to growth basedon economic stability andliberalized financial markets.

Alfaro et al. (2003)

– Calibration exercises Growth enhancing effects of FDI ineconomies with sufficientlydeveloped financial markets.

Durham (2004)

80 Countries (1979–1998) Panel data FDI does not have an unmitigated,positive effect on economicgrowth

Baldwin et al. (2005)

9 OECD countries and 7industries

Cross-section

FDI encourages long-run growthvia technology spillovers

870 C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871

B.2. Firm-level studies

Authors

Sample Technique Conclusion

Blomström(1986)

145 industries Mexico changebetween 1970–1975

Cross-section

Mexican industries dominated by foreign firms tend to bemore efficient

Aitken andHarrison(1999)

43,010 plants Venezuela(1976–1979)

Panel data

Positive relationship between foreign participation equityand plant performance but only for small plants andincreases in foreign equity negatively affect theproductivity of wholly domestically-owned firms in thesame industry

Haddad andHarrison(1993)

All enterprises with >10employees or sales revenue>100,000 dirhams Morocco(1985–1989)

Panel data

Firms with some foreign ownership exhibit higher levels ofoverall multi-factor productivity, but the rate of growth ofproductivity is higher for their wholly domestically-ownedcounterparts.

Kokko (1994)

216 Industries Mexico (1970) Cross-section

The industries where large productivity gaps and largeforeign shares occur simultaneously allow the foreignaffiliates to crowd-out local competitors

Kokko et al.(1996)

159 Manufacturing plantsUruguay (1990)

Cross-section

No signs of spillovers from FDI to Uruguayanmanufacturing plants

Kathuria(2001)

368 firms India 14 years(1975–1976 to 1988–1989)

Panel data

Positive spillovers from the presence of foreign-ownedfirms but the nature and type of spillovers vary dependingupon the industries

References

Ai, C., Chen, X., 2003. Efficient estimation of models with conditional moment restrictions conditioning unknown functions. Econometrica 71, 1795–1843.Aitken, B., Harrison, A., 1999. Do domestic firms benefit from direct foreign investment? Evidence from Venezuela. American Economic Review 89 (3), 605–

618.Alfaro, L., Areendam, Ch., Kelemli-Ozcan, S., Selim, S., 2003. FDI and economic growth: The role of financial markets. Journal of International Economics 61

(1), 512–533.Azariadis, C., Drazen, A., 1990. Threshold externalities in economic development. Quarterly Journal of Economics 105, 501–526.Balasubramanyam, V.N., Salisu, M., Sapsford, D., 1996. Foreign direct investment and growth in EP and Is countries. The Economic Journal 106, 92–105.Baldwin, R., Braconier, H., Forslid, R., 2005. Multinationals, endogenous growth and technological spillovers: Theory and evidence. Review of International

Economics 13 (5), 945–963.Barro, R.J., 1991. Economic growth in a cross section of countries. Quarterly Journal of Economics 106, 407–433.Barro, R.J., 1998. Human Capital and Growth in Cross-Country Regressions. Harvard University Press (Manuscript).Barro, R.J., Lee, J.W., 2001. International data on educational attainment: Updates and implications. Oxford Economic Papers 53, 541–563.Barro, R., Sala-i-Martin, X., 1995. Economic Growth. McGraw-Hill, New York.Bende-Nabende, A., Ford, J.L., 1998. FDI, policy adjustments and endogenous growth: Multiplier effects from a small dynamic model for Taiwan, 1959–

1995. World Development 26 (7), 1315–1330.Bengoa, M., Sanchez-Robles, B., 2003. Foreign direct investment, economic freedom and growth: New evidence from Latin America. European Journal of

Political Economy 19 (3), 529–545.Benhabib, J., Spiegel, M., 1994. The role of human capital in economic development: Evidence from aggregate cross-country data. Journal of Monetary

Economics 34, 143–174.Bils, M., Klenow, P.J., 2000. Does schooling cause growth? American Economic Review 90, 1160–1183.Blomström, M., 1986. Foreign investment and productive efficiency: The case of Mexico. Journal of Industrial Economics 35 (1), 97–110.Blomström, M., Lipsey, R., Zejan, M., 1992. What explains developing country growth? NBER Working Paper No. 4132. Cambridge, Mass.Blomström, M., Kokko, A., Zejan, M., 1994. Host country competition and technology transfer by multinationals. Weltwirtschaftliches Archiv 130 (3), 521–

533.Borenztein, E., De Gregorio, J., Lee, J.-W., 1998. How does foreign direct investment affect economic growth? Journal of International Economics 45, 115–

135.Brock, W., Durlauf, S., 2001. Growth empirics and reality. World Bank Economic Review 15 (2), 229–272.Cai, Z., Das, M., Xiong, H., Wu, X., 2006. Functional coefficient instrumental variable models. Journal of Econometrics 133, 189–209.Carkovic, M., Levine, R., 2005. Does foreign direct investment accelerate economic growth? In: Moran, T.H., Graham, E.M., Blomström, M. (Eds.), Does

Foreign Investment Promote Development. Institute for International Economics/Center for Global Development, Washington DC, US.De Gregorio, J., 1992. Economic growth in Latin America. Journal of Development Economics 39, 59–84.De Mello, L.R., 1999. Foreign Direct Investment led growth: Evidence from time series and panel data. Oxford Economic Papers 51, 133–151.De Mello, L.R., 1997. Foreign direct investment in developing countries and growth: A selective survey. The Journal of Development Studies 34 (1), 1–34.Dinopoulos, E., Segerstom, P., 2005. Multinational Firs and Economic Growth, mimeo. Available from: <http://www.cerge.cuni.cz/pdf/events/papers/

050916_t.pdf>.Durham, J.B., 2004. Absorptive capacity and the effects of foreign direct investment and foreign portfolio investment on economic growth. European

Economic Review 48 (2), 285–306.Durlauf, S.N., Johnson, P., 1995. Multiple regimes and cross country growth behavior. Journal of Applied Econometrics 10, 365–384.Durlauf, S.N., Kourtellos, A., Minkin, A., 2001. The local Solow growth model. European Economic Review 45, 928–940.Durlauf, S.N., Kourtellos, A., Tan, C., 2008. Are any Growth Theories Robust? Economic Journal 118, 329–346.Fan, J., Hardle, W., Mammen, E., 1998. Direct estimation of low dimensional components in additive models. Annals of Statistics 26, 943–971.Fan, Y., Li, Q., 2003. A new kernel-based method for estimating additive partially linear models. Statistica Sinica 13, 739–762.Fan, Y., Li, Q., 1996. Consistent model specification tests: Omitted variables, parametric and semiparanetric functional forms. Econometrica 64, 865–890.Findlay, R., 1978. Relative backwardness, direct foreign investment, and the transfer of technology: A simple dynamic model. Quarterly Journal of

Economics 92, 1–16.

C. Kottaridi, T. Stengos / Journal of Macroeconomics 32 (2010) 858–871 871

Galor, O., 2005. The transition from stagnation to growth: Unified growth theory. In: Aghion, P., Durlauf, S. (Eds.), Handbook of Economic Growth.Amsterdam, North Holland.

Haddad, M., Harrison, A., 1993. Are there positive spillovers from direct foreign investment? Evidence from panel data for Morocco. Journal of DevelopmentEconomics 42, 51–74.

Henderson, D.J., Papageorgiou, C., Parmeter, C.F., 2009. Are any growth theories linear? Why we should care about what the evidence tells us. MunichPersonal RePec Archive Paper No. 8767. Available from: <http://mpra.ub.uni-muenchen.de/8767/>.

Islam, N., 1995. Growth empirics: A panel data approach. Quarterly Journal of Economics 10, 1127–1170.Kalaitzidakis, P., Mammumeas, T., Savvides, A., Stengos, T., 2001. Measures of human capital and non-linearities in economic growth. Journal of Economic

Growth 6, 229–254.Kathuria, V., 2001. Foreign firms, technology transfer and knowledge spillovers to Indian manufacturing firms: A stochastic frontier analysis. Applied

Economics 33, 625–642.Kokko, A., 1994. Technology, market characteristics, and spillovers. Journal of Development Economics 43, 279–293.Kokko, A., Tansini, R., Zejan, M., 1996. Local technological capability and spillovers from FDI in the Uruguayan manufacturing sector. Journal of Development

Studies 34, 602–611.Kourtellos, A., 2003. Modeling Parameter Heterogeneity in cross-Country Growth Regression Models. mimeo, Department of Economics, University of

Cyprus.Krueger, A.B., Lindahl, M., 2001. Education for growth: Why and for whom? Journal of Economic Literature 39, 1101–1136.Kumar, N., Siddharthan, N.S., 1997. Technology, market structure and internationalization: Issues and policies for developing countries. Routledge and UNU

Press, London and New York.Kyriacou, G.A., 1991. Level and growth effects of human capital. C.V. Starr Center Economic Research Reports, 91-26. New York University.Lau, L.J., Jamison, D.T., Louat, F.F., 1991. Education and productivity in developing countries: An aggregate production function approach. Working Paper No.

612, World Bank.Li, Q., Wang, S., 1998. A simple consistent bootstrap test for a parametric regression function. Journal of Econometrics 87, 45–165.Li, Q., Racine, J.S., 2007. Nonparametric Econometrics. Princeton University Press, Princeton, NJ.Linton, O.B., Nielsen, J.P., 1995. A kernel method of estimating structural nonparametric regression based on marginal integration. Biometrica 82, 93–100.Liu, Z., Stengos, T., 1999. Non-linearities in cross-country growth regressions: A semiparametric approach. Journal of Applied Econometrics 14, 527–538.Maasoumi, E.J., Racine, J.S., Stengos, T., 2007. Growth and convergence: A profile of distribution dynamics and mobility. Journal of Econometrics 136, 483–

508.Mamuneas, T., Savvides, A., Stengos, T., 2006. Economic development and the return to human capital: A smooth coefficient semiparametric approach.

Journal of Applied Econometrics 21, 111–132.Mankiw, N., Romer, D., Weil, D.N., 1992. A contribution to the empirics of economic growth. Quarterly Journal of Economics 107, 407–437.Markusen, J.R., Venables, A., 1999. Foreign direct investment as a catalyst for industrial development. European Economic Review 43 (2), 335–356.Masanjala, W.H., Papageorgiou, C., 2004. The Solow model with CES technology: Nonlinearities and parameter heterogeneity. Journal of Applied

Econometrics 19, 171–201.Murphy, K., Shleifer, A., Vishny, R., 1989. Industrialization and the big push. Journal of Political Economy 97, 1003–1026.Pack, H., Page, J., 1994. Accumulation, exports, and growth in the high performing Asian economies. Carnegie-Rochester Conference Series on Public Policy

40, 199–236.Pritchett, L., 2001. Where has all the education gone? World Bank Economic Review 15, 367–391.Quah, D.T., 1996. Empirics for economic growth and convergence. European Economic Review 40, 1353–1375.Robinson, P., 1988. Root N-consistent semiparametric regression. Econometrica 56, 931–954.Romer, P.M., 1990. Human capital and growth: Theory and evidence. Carnegie-Rochester Conference Series on Public Policy 32, 251–268.Saggi, K., 2000 Trade Foreign Direct Investment, and International Technology Transfer: A Survey. Issued as WT/WGTI/W/88, dated 19 September 2000.

World Trade Organization, Geneva.Savvides, A., Stengos, T., 2009. Human Capital and Economic Growth. Stanford University Press, Stanford, California.Sperlich, S., Tjostheim, D., Young, L., 2002. Nonparametric estimation and testing of interaction in additive models. Econometric Theory 18 (2), 197–251.UNCTAD, 2005. World Investment Report: Transnational Corporations and the Internationalization of R&D. United Nations Cooperation on Trade and

Development.Walz, U., 1997. Innovation, foreign direct investment and growth. Economica 64, 63–79.Xu, B., 2000. Multinational enterprises, technology diffusion, and host country productivity growth. Journal of Development Economics 62, 477–493.Zhang, K., 2001. Does foreign direct investment promote growth? Evidence from East Asia and Latin America. Contemporary Economic Policy 19 (2), 175–

185.Zheng, J.X., 1996. A consistent test of functional form via nonparametric estimation techniques. Journal of Econometrics 75, 263–289.