Foreign direct investment and country-specific human capital.

Kim, Jinyoung ; Park, Jungsoo

I. INTRODUCTION

The past three decades have witnessed an unprecedented increase in foreign direct investment (FDI) in the world; annual FDI flows from and to Organisation for Economic Cooperation and Development (OECD) countries have increased more than 30-fold from 1982 to 2005, while the world trade merely quadrupled during the same period; the ratio of gross private capital flows to gross domestic product (GDP) has risen from 10.3% in 1990 to 32.4% in 2005; the ratio of FDI flows to GDP has climbed up from 2.2% in the early 1990s to 4.4% in 2005. (1) As FDI gained importance in the international movement of capital, a variety of theoretical and empirical studies have investigated the factors that determine FDI. (2) In this article, we offer an important but under-studied determinant of FDI: foreign-educated labor in FDI host countries.

Specifically, using international bilateral data on FDI and foreign education, we present empirical evidence on the role of country-specific foreign-educated labor in an FDI host country as a determinant of foreign direct investments.

Production involves the process of combining physical capital and the human capital of employees. Human capital includes the general skills of workers, but it also includes their knowledge of firm-specific technology, managerial skills specific to the organization, and efficient communication skills with co-workers. Consequently, the labor that possesses these firm-specific skills, which can be readily used by a firm, can be more productive in that firm than in other firms. When a firm invests in a foreign country through a subsidiary that shares the technology of the parent firm, the labor force that has acquired various types of human capital specific to the parent firm, and thus to the subsidiary, can be more productive in the foreign subsidiary. For instance, the local managers of a foreign subsidiary who are able to speak the same language of the managers of their parent firm, or who know thoroughly well how the parent firm and its subsidiaries are organized and operated, can be more productively utilized in the subsidiary firm. Since human capital specific to a firm and its foreign subsidiaries can be acquired through education provided in the country of the parent firm, the availability of workers in a potential FDI host country who studied in the parent-firm country can be an important deciding factor for a firm in investing abroad through its foreign subsidiary. (3)

Foreign training can provide various types of country-specific human capital that would increase labor productivity more in the host country of training than in other countries. Country-specific human capital includes language capital, which has been shown by numerous papers such as Chiswick and Miller (1993) and Lazear (1999) to have considerable importance in labor market performances. Knowledge on firm organization and social system is another example of country-specific human capital that can be acquired through foreign training. Foreign education can also enhance the productivity of students in firms of the host country, because it provides information about the ability and potential of students in a form that can be more easily processed by firms in the host country. Greater knowledge regarding employees allows firms to utilize manpower more efficiently and enhance workers' productivity, as pointed out by Becker (1993), who discusses low worker mobility across countries as evidence of the importance of country-specific training.

Earlier theoretical models in international trade, such as MacDougall (1960), attribute international capital movement to the difference in factor endowment measured by physical capital per worker. (4) Empirical findings, however, show that the actual capital flows fall short of the theoretical prediction that capital moves from developed economies to developing economies to take advantage of the higher rates of returns to capital. (5) According to Lucas (1990), assuming that human capital and physical capital are complementary in production, physical capital fails to flow into developing economies owing to the lack of human capital endowment in those economies. (6)

Kim and Park (2011) incorporate country-specific human capital into the standard model of FDI determination. They argue that the availability of country-specific foreign-educated human capital, not just human capital in general, is important in attracting country-specific FDI, because human capital specific to an FDI source country available in a host country can be more conducive to the operations of subsidiaries from the source country. More specifically, they develop a multi-country model of production where domestic and foreign firms utilize capital as well as two types of labor--domestically educated and foreign-educated labor. The former type of labor has human capital specific to the home country, and the latter has skills specific to the foreign country of study. They postulate that the two types of labor inputs are imperfect substitutes in production owing to country-specific skills, and that foreign-educated labor has the relative advantage in productivity for foreign subsidiaries owing to the existence of organizational capital (Prescott and Visscher 1980) or firm-specific technologies learned through foreign education.

The model shows that an increase in foreign-educated labor will unambiguously raise the demand for capital in foreign subsidiaries, and thus for foreign direct investment. According to this implication, it is likely that having more U.S.-educated managers in Korea, for example, would make it easier for U.S. firms to directly invest in Korea. Furthermore, the direct investment of U.S. firms in Korea will be influenced

only by the abundance of U.S.-educated Korean managers, and not by the presence of managers educated in countries other than the United States.

We find evidence to support this theory on the importance of foreign education in attracting FDI when we compare the educational backgrounds of chief executives of Korean domestic firms with those of foreign multinational firms operating in Korea. We collected data on the destination countries for tertiary education of the CEOs at the top 100 foreign multinational firms in sales as well as for those at the top 500 domestic firms for the years 2006 and 2007. (7) Table 1 reports the proportion of foreign-educated CEOs in foreign and domestic firms. The results in the table indicate that foreign firms are more likely to hire foreign-educated executives than domestic firms: 30.1% of the CEOs in foreign firms studied in foreign countries, while only 19.7% of CEOs in domestic firms did so. This difference is statistically significant by the t-test statistic for difference in group proportion at the 5% level. (8) Furthermore, we find that among the foreign-educated CEOs of foreign firms, the proportion of those whose host country of study is the same as the country of the multinational firm is 54.6%.

To further investigate the validity of our hypothesis, this paper empirically tests whether foreign-educated labor attracts FDI from the host country of education against bilateral panel data for 63 developed and developing countries over the period 1963-1998. Our empirical findings strongly support our predictions and show that foreign-educated labor accounts for a sizable portion of growth in FDI flows during the sample period.

This paper is organized as follows. Section II details the empirical methodology and the data set used in our empirical analysis. Section III shows that the number of foreign-educated students, which approximates the size of labor pool with country-specific human capital, has a positive effect on FDI inflow from the foreign country where the students were educated. This effect is robustly present when we control for factors such as transportation cost, market size, and growth in FDI host and source countries, trade volume, and dyad-specific idiosyncratic effects. Section IV concludes the article with policy implications and a discussion on the importance of foreign-educated labor in explaining observed time-series changes in FDI since 1980.

II. EMPIRICAL IMPLEMENTATION

We test our prediction on the relationship between foreign-educated labor and FDI against bilateral FDI and student flow panel data. We estimate a reduced-form regression model with FDI as the dependent variable. A set of explanatory variables includes the number of students abroad and other explanatory variables identified in the earlier studies on FDI.

A. Description of Variables Used

The data for annual foreign direct investment inflows and outflows are taken from OECD's International Direct Investment Statistics Yearbook (1999), covering 63 countries for the period 1980-1998. The countries are listed in Appendix A.

We take the number of students who studied abroad in country j as proxy for the size of labor educated in foreign country j. Our data on students abroad are taken from UNESCO's Statistical Yearbooks. The data contain annual bilateral flows of students studying abroad at the tertiary level. The data are available for 63 countries over the period 1963-1996, although for some countries the data are not available for all the years. (9)

One concern for using this variable as a measure for the size of labor educated in foreign country j is that some students do not return to their home countries after completion of study abroad. Based on the surveys of foreign students, Glaser (1979) reports that the return rates among the students are quite high in most large host countries, with the exception of the United States, and non-returning emigrant students typically maintain a close connection with their home countries. Moreover, many of these non-returning students show interest in working later in their home countries as managers for multinational firms. As a robustness check in the empirical section, we estimate the relationship between students abroad and FDI in the subsample that excludes the United States (Table 5). Since a 15-year-lagged value of students abroad is used as an explanatory variable in the baseline model of our empirical study (the reason for the time lag is given in the next section), we use the foreign student data over the period 1965-1983 to match with the FDI data for 1980-1998. This implies that the survey results in Glaser (1979) are relevant in characterizing the students in our data set.

Another related issue is that foreign-educated labor in our story may include foreign labor migrating from the FDI source country. One may argue that the effect of foreign-educated labor on FDI cannot be tested with our data on foreign students if the students abroad are inversely correlated with foreign labor so that changes in the number of students abroad cannot capture the changes in total foreign-educated labor. However, the OECD country sample from 1986 to 1995 demonstrates a positive correlation between a change in the foreign labor force and the flow of students abroad with a 15-year lag: the correlation coefficient is 0.4330. In addition, we estimate the effect of students abroad on FDI in the sample of countries with high shares of foreign labor in the total labor force to check whether our estimation results are still robust in those countries that are relatively open to foreign labor inflows (Table 5).

The third concern is that our measure is not a stock variable. In the long-run steady-state equilibrium, however, the flow variable of students abroad should have a one-to-one positive relationship with the stock variable. Our empirical specifications attempt to estimate the relation of the stock variable of students abroad and FDI in the steady-state equilibrium. Since a long-run relationship is better estimated with cross-country variations in the data, we report in the section for sensitivity analysis a panel regression result with between effects.

To control for the transaction cost factors in FDI, our explanatory variables include the distance between a source country and a host country, and three binary variables for whether the same language is used in two countries, whether they practice the same religion, and whether one country was once a colony of the other. We also include as explanatory variables what are identified as determinants of FDI in the literature: the ratio of per capita GDP in an FDI source country to that in a host country, the total GDP levels and the per-capita GDP growth rates in both countries, the GDP shares of domestic investment and government spending in an FDI host country, and the real exchange rate of an FDI host country. The summary statistics of all variables used in this study are provided in Table 2, and the data sources of all control variables are described in Appendix A.

B. Econometric Model Specification

Our baseline specification is a log-linear model:

(1) ln([FDI.sub.ijt]) = [d.sub.i] + [d.sub.j] + [d.sub.t] + [d.sub.i] x T + [d.sub.j] x T + [[beta].sub.0] ln([STDT.sub.ijt-15]) + [beta]'[X.sub.ijt] + [[epsilon].sub.ijt],

where [FDI.sub.ijt] is the real FD1 from country i to country j in year t as a share of country i's GDP, and [STDT.sub.ijt-15] represents the students abroad from country j who studied in country i in year t - 15 as a share of country j's population. FDI and STDT are paired in a reverse direction as is required in our theory: FDI from country i to country j, and students from country j to country i. Note also that the number of students abroad 15 years ago is paired with FDI of the current year to account for the time needed to acquire education and return home. This time lag will also help avoid the reverse causality and endogeneity problems. (10) [X.sub.ijt] is a vector of regressors, including the logarithms of all the explanatory variables discussed in the previous section; [d.sub.i] and [d.sub.j] are country-specific dummy variables for an FDI source country and a host country, respectively; and [d.sub.t] is a dummy variable for calendar years. In this specification, we also include a time trend variable (T) and its interaction terms with dummy variables [d.sub.i] and [d.sub.j] to allow for country-specific time trends. The error term [[epsilon].sub.ijt] is assumed to be distributed i.i.d. with zero mean.

As part of sensitivity analysis, we conducted regressions under various types of specifications, including models with the level of FDI as a dependent variable and dyad-specific random, fixed, and between effects.

III. EMPIRICAL FINDINGS

A. Results from Regression Analysis of FDI

This section reports the results obtained from the regression model in subsection "Econometric Model Specification" in Section II. In Table 3, the dependent variable is the ratio of FDI to GDP in an FDI source country. All variables are entered in logs, except binary variables. Model 1 presents our baseline ordinary least squares (OLS) estimates. In Models 2 and 3, we introduce additional regressors to the baseline model to further isolate the autonomous effect of STDT on FDI. In Model 4, we run the same baseline specification with the FDI level (FDIL) as the dependent variable instead of the GDP ratio of FDI, and the number of students (STDTN) as a regressor in place of the population share of students, all in logs.

Table 3 indicates that the effect of STDT on FDI is statistically significant and consistent with our theory: more foreign direct investment flows into a country with a larger pool of workers who studied in an FDI source country. STDT has a statistically significant effect in all regressions, regardless of the alternative sets of explanatory variables introduced.

Factors such as shorter distance, using the same language, and having the same religion may reduce the transaction costs in FDI and increase foreign direct investment flows. This prediction is confirmed by all the models in Table 3. The estimated effects of students abroad on FDI, thus, cannot be ascribed to geographical and cultural proximity of countries, which will generate a bias toward a positive association between students abroad and FDI. The results in Table 3 suggest that more FDI flows into those countries that used to be colonies of an FDI source country, although the effect is only marginally significant.

Various macroeconomic variables are included as explanatory variables to incorporate the earlier findings that show FDI decisions are affected by changes in macroeconomic conditions such as market size, market growth, exchange rates, and investment environment. (11) The per capita GDP ratio (RELPCGDP) is included as a regressor to proxy the (inverse of the) relative returns to capital. We expect that the more FDI invested, the higher (lower) the rate of returns to capital in the FDI host (source) country will be. Furthermore, the per capita income in the source country may proxy the demand for a clean environment, which can be a push factor for FDI to relocate polluting manufacturing facilities to other countries. In both effects, the per capita GDP ratio is expected to have an adverse effect on FDI. The results in Table 3 support this prediction, albeit the effect is only marginally significant.

The real GDP level of an FDI host country (GDP2) is expected to have the effect of scale economies and a positive association with FDI. This variable, however, shows an insignificant effect on FDI. A faster-growing economy may attract more FDI because of the potential market growth in the future. (12) The real GDP growth rates of an FDI source and host country (GROW1 and GROW2) generally show the expected effects on FDI, albeit the effects are not strong.

Our regressions include the GDP share of domestic investment in an FDI host country (I) as an explanatory variable. In Kim and Park (2011), both FDI and domestic investment are endogenously determined and negatively correlated, since domestic and foreign firms compete to produce the same product. On the other hand, earlier studies have suggested that a greater share of domestic investment in a host country may reflect an atmosphere favorable toward private enterprise, implying a positive correlation between domestic investment and FDI. Table 3 shows that the effect of domestic investment is positive but insignificant, possibly because of these two competing effects. The GDP share of government spending in an FDI host country (G)is entered as a regressor to control for the involvement of government in the economy, which may shape the environment for private business and foreign direct investment. (13) In Table 3, this variable is generally shown to have a negative effect on FDI, implying that weak government intervention attracts foreign direct investment.

As a proxy for international competitiveness, we introduce the real exchange rate of an FDI host country (EXCHANGE) as an explanatory variable. Theoretically, overvaluation of the real exchange rate may lead to a reduction in FDI if FDI and trade are substitutes. On the other hand, a fall in the prices of intermediate good imports for foreign firms may induce more FDI. The results in Table 3 strongly support the former theory. (14)

A strong relationship in trade between two countries may facilitate the flow of FDI as well as the flow of students between the countries, resulting in a spurious correlation between FDI and STDT, without the causal influence of STDT on FDI as provided in our theory. To control for the effect of trade, we introduce the volume of exports from an FDI source country to a host country as a regressor in Model 2, although trade may be an endogenous variable in FDI regression. (15) The result in Model 2 shows that trade has a significantly positive association with FDI, consistent with our expectation. Surprisingly, however, we observe a strong and significant effect of STDT on FDI, even when trade is introduced into the regression model. The estimated effect of STDT cannot therefore be ascribed to a higher trade volume between two countries, and ultimately to the various factors that influence their trade relationship.

Higher tariff in an FDI host country may reduce imports and instead increase FDI to the host country, as the relative cost of imports rises with higher tariff. However, an increase in tariff would discourage FDI, because a higher tariff would raise the prices for imported intermediate goods used by multinational firms. Moreover, a higher tariff can be associated with lower FDI, as the tariff rate reflects the degree of economic and political openness of a country. Another measure for the degree of economic and political openness would be the number of tourists in an FDI host country. The tariff rate and the number of tourists in an FDI host country are introduced as additional regressors in Model 3 to further isolate the independent effect of STDT on FDI. As expected, the tourist variable shown in Model 3 has a significant and positive association with FDI, while the tariff rate has an adverse, but insignificant effect on FDI.

Human capital in an FDI host country is considered an important factor for attracting FDI, especially the FDI flows to developing economies, as discussed in Section I. An increase in secondary school enrollment may raise the efficiency units of domestic unskilled labor, raising the input demands as well as production levels of domestic and foreign firms in an FDI host country, and hence increasing FDI. (16) On the other hand, an increase in tertiary school enrollment may raise the efficiency units of domestically educated skilled labor, and therefore, as predicted by our theory, an FDI host country will receive less FDI with an increase in domestically educated labor. The results in Model 3 show that more students at the secondary level in an FDI host country would attract more FDI, which is consistent with our prediction. The effects of the level of tertiary school students in FDI host and source countries on FDI flows are also supportive of our theory.

B. Additional Sensitivity Tests

Dyad-Specific Effects. Although we included many variables as regressors in our regressions in Table 3 to control for dyad-specific characteristics in our data, there may potentially be other important dyad-specific factors that are missing in our analysis. In Table 4, we introduce various forms of dyad-specific effects in our baseline specification. We report only the estimated coefficients associated with the number of students in all parts of this table to save space.

Table 4 starts with Model 1, which does not include country-specific constants or dyad-specific effects. However, Model 1 as well as the other three models includes source-country-specific and host-country-specific time trends and calendar-year dummies, as in our baseline specification. Models 2, 3, and 4 report the regression results from the models with dyad-specific random effects, fixed effects, and between effects, respectively. In all models except Model 3, the effects of STDT are still statistically significant and consistently positive. We note that the coefficient associated with STDT in Model 3 is insignificant, which may be because there is little within-dyad variations in both the FDI-to-GDP ratio and the number of students relative to population. This implies that most of the variations in our data take place between dyads, and our estimation of the STDT effect on FDI is largely based on the between-dyads variations, which is more desirable for the estimation of the long-term relationship between STDT and FDI. We also note that our estimation is based on variations across pairs of countries, not across countries, and the results in Model 4 are not simply due to country-level differences.

To examine whether the between-dyads effect found in part A is strongly driven by a few specific pairs of countries in the sample (the United States and China, for example), and also to address other issues arising from regional differences, we provide sensitivity analysis involving regional subsamples in Table 5, where we both exclude and include various regions from the analysis. Overall, we still obtain qualitatively similar results (see the following subsection on "Regional Subsamples").

One concern in our model specification is that there is no theoretical validation for a particular lag between STDT and FDI, although we have tried various lags for sensitivity. It is also possible that the effect of STDT on FDI may last for several years to warrant a dynamic panel-data model. The between-effects model can offer an estimate for the aggregate effect of the numbers of students with various lags on the dyad-specific aggregate of FDI, as the between-effects regression is performed in accordance with the dyad-specific means of the dependent variable and the regressors. Note that in the between-effects model, our dependent variable is the average value of ln(FDI) over the period 1980-1998, while the student variable is averaged over the period 1965-1983, which helps avoid the problem of endogeneity in the student variable.

Regional Subsamples. To seek economic favors from a particular developing country, such as provision of incentives for FDI inflows or reduction of regulations on FDI, a developed country may selectively admit more students from the developing country, which would later have a favorable policy toward the developed country. A positive relation between STDT and FDI may arise because of this, which is beyond the employment effect proposed in our model. Since this type of political consideration is expected to play a more significant role between a developed country and a developing country, we estimate our baseline regression model in Model 1 of Table 5 with a subsample that includes only the FDI flows between the developed "North" countries. We find that the effect of STDT is much more pronounced in this regression than in a regression with the whole sample. This suggests that we cannot attribute the positive link between STDT and FDI in our empirical findings to the political consideration discussed above.

Table 5 also reports the results with several alternative sets of subsamples: "North to South" FDI flows in Model 2, and "South to North" flows in Model 3. A rather surprising finding is that the effect of STDT is more pronounced for FDI flows between two developed countries than for those from developed to developing countries. One possible reason is that the quality of education acquired by students from North countries is higher than that acquired by students from "South" countries, so that the foreign-educated labor in North countries attracts more FDI. (17) We see an insignificant effect of STDT on FDI for the "South to North" subsample, probably because of the small number of countries included and the short time series for each country.

In order to investigate whether our results have been mainly influenced by the inclusion of China and the United States, which are the largest sender of students abroad and the largest source of foreign direct investment, respectively, we exclude all observations involved with the two countries in Model 4. The U.S. observations are excluded, with the additional concern of the high non-returning rate among foreign students, as discussed in Section II. In Model 5, we address the issue of foreign labor in the domestic labor market discussed in Section II, by performing a regression on a subsample with FDI host countries having high foreign labor force shares. (18) Even in the regressions with these subsamples, the effect of STDT on FDI is quantitatively robust.

County-Specific Foreign Students. A rise in the number of foreign-educated students who studied in one country may have an influence on FDI inflow from other countries of foreign study. In Table 6, we include the total number of students from country j who studied in foreign countries other than country i (STDTNREST) as a regressor in addition to STDTN, using our baseline specification in Model 4 of Table 3. For comparison, this baseline specification is rerun with a smaller sample of observations that have non-missing values of STDTNREST (see Model 2 in Table 6). Model 1 shows that STDTNREST has a negative impact on the FDI inflow from country i to country j, while the effect of STDTN is significant and positive. The negative cross-effect is again confirmed in Model 3, where the FDI inflows from countries other than country i (FDILREST) are negatively associated with STDTN; a rise in a country-specific STDTN reduces FDI inflows from the rest of the countries.

This negative association may be because there is competition amongst the FDI source countries. Although a rise in country-specific foreign-educated labor will make a favorable environment for all FDI inflows to the host country, it will be most favorable for FDI from the country where the workers were educated. Competition may then crowd out the opportunity for investment from the rest of the countries. However, Model 1 in Table 7 indicates that the net impact of STDTN on the total FDI inflow from all countries (TOTFDIL) is positive, albeit insignificant. To further investigate the net impact of STDTN, we use the FDI stock obtained from Lane and Milesi-Ferretti (2001) as the dependent variable in Model 2 in Table 7. This result also shows a positive effect.

As an additional test to check the validity of our prediction on the FDI effect of students abroad, an alternative specification to Model 1 in Table 7 was estimated with total portfolio investments to an FDI host country as the dependent variable, instead of total FDI. (19) Interestingly, the coefficient associated with STDTN in this regression was insignificant with the t-statistic of -0.06. This finding lends potent support to our theory that country-specific human capital attracts FDI, not necessarily other types of international capital flows.

Alternative Time Lags between FDI and STDT. In Table 3, we report the results when the time lag between FDI and the number of students is 15 years. To test the sensitivity of our results to different lags of STDT, we use alternative specifications, where [STDT.sub.t-s] (s = 5, 6 ..., or 25) is introduced as a regressor in place of our baseline-model variable [STDT.sub.t-15]. In each specification, the effect of the number of students abroad with the respective alternative lag is found to be significantly pronounced. (20) We also use a specification where all the STDT variables lagged from 10 years to 20 years are simultaneously included. The sum of all the estimated coefficients associated with the lagged variables of STDT is found to be positive and statistically significant. Moreover, the magnitude of the sum is very close to and statistically not different from the estimated effect of STDT with a 15-year lag in model 1 of Table 3. These results indicate the robustness of our estimated effect of STDT on FDI.

IV. CONCLUDING REMARKS

We used bilateral FDI and foreign student data for 63 developed and developing countries over the period 1963-1998 to test our proposition that an increase in country-specific foreign-educated labor will raise FDI inflow from the foreign country where the labor was educated.

Despite the limitations of our data, the empirical evidence in this paper strongly confirms our proposition under various alternative specifications, controlling for the determinants recognized in the literature. Our results also indicate that country-specific foreign-educated labor only attracts FDI from the host country of foreign education. In fact, the FDI inflows from other countries are crowded out. The net effect of foreign-educated labor on total FDI inflow, however, is positive. Consistent with our theory, we have evidence that foreign-educated labor attracts FDI, but not necessarily other types of foreign capital.

The estimated effect of students abroad on FDI, presented in Table 3, is not only statistically but also quantitatively significant. According to our simple calibration exercise, the change in ln(STDT) can explain approximately 14.7% of the actual change in ln(FDI) from 1980 to 1998 (see Appendix B for the calculation method).

Needless to say, we have left a number of issues unaddressed in this paper. Regarding our empirical analysis, a more extensive survey of data on the non-return rates of students abroad and bilateral flows of foreign labor would provide us with a more accurate measure of the foreign-educated labor in our empirical implementation. We leave the study of these issues to future work.

ABBREVIATIONS

FDI: Foreign Direct Investment

FDIL: Foreign Direct Investment Level

GDP: Gross Domestic Product

OECD: Organisation for Economic Cooperation and Development

OLS: Ordinary Least Squares

doi: 10.1111/j.1465-7295.2012.00478.x

APPENDIX A

The 63 countries included in this study are Algeria, Argentina, Australia, Austria, Belgium-Luxembourg, Brazil, Bulgaria, Canada, Chile, China, Colombia, Costa Rica, Czech Republic, Czechoslovakia, Denmark, Egypt, Finland, France, Germany, Greece, Hong Kong, Hungary, Iceland, India, Indonesia, Iran, Ireland, Israel, Italy, Japan, Korea, Kuwait, Libya, Malaysia, Mexico, Morocco, The Netherlands, Netherlands Antilles, New Zealand, Norway, Panama, Philippines, Poland, Portugal, Romania, Russia, Saudi Arabia, Singapore, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland, Taiwan, Thailand, Turkey, Ukraine, United Arab Emirates, United Kingdom, United States, USSR, and Venezuela.

Data on language, religion, and colony are from Encyclopedia Britannica. Data on distance are from the Internet map search engine at www.indo.com/distance. Data on per capita GDP, total GDP, GDP shares of domestic investment and government spending, and real exchange rates are obtained from Penn World Tables (Mark 6). The bilateral trade flow data for 63 countries over 1976-1999 are taken from World Bank Institute's Trade and Production Database. Data on tariff rates are available in Trends in Average Tariff Rates for Developing and Industrial Countries 1980-99 of World Bank Institute. Data on tourists and school enrollment rates are from World Development Indicators (2001). Corporate income tax data are from Devereux, Griffith, and Klemm (2002), and total portfolio investments are from International Financial Statistics, published by IMF (2001).

APPENDIX B

In our data, the actual sample mean values of In(FDI) were -6.001 in 1980 and -4.919 in 1998, while the corresponding sample mean values of ln(STDT) were -5.964 and -5.303, respectively. Using the estimated coefficients in Model 1 of Table 3, and assuming that all variables other than FDI and STDT are constant at the sample mean values of the variables in 1980, we calculate the predicted change in the mean value of In(FDI) from 1980 to 1998 explained by the change in the sample mean value of In(STDT) from 1965 to 1983. The predicted change in the mean value of In(FDI) from 1980 to 1998 is 0.240(-5.303 + 5.964) = 0.159, whereas the actual change in the sample mean value of ln(FDI) during the same period was 1.081. This calculation suggests that the change in ln(STDT) can explain 14.7% (= 0.159/1.081) of the actual change in ln(FDI) during this period.

REFERENCES

Barrell, R., and N. Pain. "An Econometric Analysis of U.S. Foreign Direct Investment." Review of Economics and Statistics, 78, 1996, 200-07.

Becker, G. S. Human Capital. Chicago: University of Chicago Press, 1993.

Benhabib, J., and M. Spiegel. "The Role of Human Capital in Economic Development: Evidence from Aggregate Cross-Country Data." Journal of Monetary Economics, 34, 1994, 143-73.

Blonigen, B. A. "Firm-Specific Assets and the Link between Exchange Rates and Foreign Direct Investments." American Economic Review, 87(3), 1997, 447-65.

Brainard, S. L. "A Simple Theory of Multinational Corporations and Trade with a Trade-off between Proximity and Concentration." NBER Working Paper No. 4269, 1993.

Caves, R.E. "Exchange-Rate Movements and Foreign Direct Investment in the United States," in The Internationalization of U.S. Markets, edited by D. B. Audretsch and M. P. Claudon. New York: New York University Press, 1989, 199-228.

Chiswick, B. R., and P. W. Miller. "Language in the Immigrant Labor Market," in Immigration, Language and Ethnicity: Canada and the United States, edited by B. R. Chiswick. Washington DC: AEI Press, 1993.

Cushman, D. O. "Real Exchange Rate Risk, Expectations, and the Level of Direct Investment." Review of Economics and Statistics, 32, 1985, 297-308.

Devereux, M. P., R. Griffith, and A. Klemm. "Corporate Income Tax Reforms and International Tax Competition." Economic Policy, 17(35), 2002, 450-95.

Dunning, J. H. International Production and the Multinational Enterprise. London: George Allen & Unwin, 1981.

Edwards, S. "Capital Flows, Foreign Direct Investment, and Debt-Equity Swaps in Developing Countries." NBER Working Paper No. 3497, 1990.

Federation of Korean Industries. Archives of Korean Business Figures (Hankook Je-gye In-sa Rok). Seoul: Federation of Korean Industries, 2008.

Froot, K. A., and J. C. Stein. "Exchange Rates and Foreign Direct Investment: An Imperfect Capital Markets Approach." Quarterly Journal of Economics, 106(4), 1991, 1191-217.

Glaser, W. A. The Brain Drain: Emigration and Return. The United Nations Institute for Training and Research, Report No. 22. Elmsford, NY: Pergamon Press, 1979.

Goldberg, L. S., and C. D. Kolstad. "Foreign Direct Investment, Exchange-Rate Variability and Demand Uncertainty." International Economic Review, 36, 1995, 855-73.

Helpman, E., and P. Krugman. Market Structure and International Trade. Cambridge, MA: MIT Press, 1985.

Hines Jr, J. R. "Altered States: Taxes and the Location of Foreign Direct Investment in the U.S." American Economic Review, 86, 1996, 1076-94.

Horstmann, I., and J. R. Markusen. "Endogenous Market Structures in International Trade." Journal of International Economics, 20, 1992, 225-47.

Institute of International Education. Open Doors: Report on International Educational Exchange. New York: IEE, 2001.

International Monetary Fund. International Financial Statistics. Washington, DC: IMF, 2001.

Jones, R. W. "International Capital Movements and the Theory of Tariffs and Trade." Quarterly Journal of Economics, 81, 1967, 1-38.

Kemp, M. C. "The Gain from International Trade and Investment: A Neo-Heckscher-Ohlin Approach." American Economic Review, 56, 1966. 788-809.

Kim, J. "'Economic Analysis of Foreign Education and Students Abroad." Journal of Development Economics, 56, 1998, 337-65.

Kim, J., and J. Park. "'Foreign-Educated Labor and Foreign Direct Investment: Theory and Evidence." Working Paper, Discussion Paper Series, Institute of Economic Research, South Korea, 2011.

Labs, J. "How Gillette Grooms Global Talent." Personnel Journal, 72, 1993, 64-70.

Lane, P., and G. M. Milesi-Ferretti. "The External Wealth of Nations: Measures of Foreign Assets and Liabilities for Industrial and Developing Countries." Journal of International Economics, 55(2), 2001, 263-94.

Lazear, E. P. "Culture and Language." Journal of Political Economy, 179(6), 1999, S95-126.

Lucas, R. "Why Doesn't Capital Flow from Rich to Poor Countries?" American Economic Review, 80, 1990, 92-96.

MacDougall, G. D. A. "The Benefits and Costs of Private Investment from Abroad: A Theoretical Approach." Economic Record, 36, 1960, 13-35.

Markusen, J. R. "Multinationals, Multi-Plant Economies, and the Gains from Trade." Journal of International Economics, 16, 1984, 205-26.

Markusen, J. R., and K. E. Maskus. "General-Equilibrium Approaches to the Multinational Firm: A Review of Theory and Evidence." NBER Working Paper, No. 8334, 2001.

Markusen, J. R., and A.J. Venables. "The Theory of Endowment, Intra-Industry and Multinational Trade." Journal of International Economics, 52, 2000, 209-34.

Noorbakhsh, F., A. Paloni, and A. Youssef. "Human Capital and FDI Inflows to Developing Countries: New Empirical Evidence." World Development, 29(9), 2001, 1593-610.

OECD. International Direct Investment Statistics Yearbook 1999. Paris: OECD Publishing, 1999.

Prescott, E., and M. Visscher. "Organizational Capital." Journal of Political Economy, 88(3), 1980, 446-61.

Root, F. R., and A. A. Ahmed. "Empirical Determinants of Manufacturing Direct Foreign Investment in Developing Economies." Economic Development and Cultural Change, 27, 1979, 751-67.

Wheeler, D., and A. Mody. "International Investment Location Decisions: The Case of U.S. Firms." Journal of International Economics, 33, 1992, 57-76.

Williamson, O. Markets and Hierarchies: Analysis and AntiTrust Implications. New York: Free Press, 1975.

World Bank. World Development Indicators, various years.

Zhang, K., and J. R. Markusen. "Vertical Multinationals and Host-Country Characteristics." Journal of Development Economics, 59, 1999, 233-52.

(1.) World Development Indicators 2007, World Bank.

(2.) See Edwards (1990) and Markusen and Maskus (2001) for literature surveys on the determinants of FDI.

(3.) Parent firms often train local labor hired in subsidiaries at their own expense in order to develop firm-specific skills. For the example of Gillette Corporation, see Laabs (1993). However, some skill training, such as foreign language education, is general enough to be acquired through foreign education paid in part by students.

(4.) With the industrial organization approach to trade, such theoretical models as Helpman and Krugman (1985) predict that a multinational firm will extend its business to countries that differ significantly in relative endowments. but not to very similar countries. Other trade models, for example, Kemp (1966) and Jones (1967), consider technology differences as a major factor in capital movements.

(5.) To explain the larger flows of FDI between developed economies than between developed and developing economies. "horizontal" models such as Markusen (1984), Horstmann and Markusen (1992), Brainard (1993), and Markusen and Venables (2000) argue that, given the moderate to high trade costs and plant-level as well as firm-level scale economies, multinational firm activity will arise between similar countries.

(6.) Zhang and Markusen (1999) also provide a model where the availability of skilled labor in the host country influences the volume of FDI inflows. Empirical evidence for the effect of human capital on FDI has been scarce with mixed results. Root and Ahmed (1979) do not find human capital to be a significant factor of FDI inflows for 58 developing countries, whereas Noorbakhsh, Paloni, and Youssef (2001) show that human capital has become a significant determinant of FDI inflows to developing countries in more recent years. Benhabib and Spiegel (1994) provide indirect empirical support for the role of human capital in FDI, by showing that human capital induces greater accumulation of physical capital stock.

(7.) Looking up company websites and who's-who information on the Internet, we were able to determine the countries of study for 92 CEOs of foreign firms. For the CEOs of domestic firms, we used data from the Federation of Korean Industries (2008). From the 500 domestic firms, we excluded those with sales greater than the largest foreign firm in the sample. Furthermore, for both the foreign and domestic firm samples, we excluded those CEOs with non-Korean names. The sample sizes after the selection are 73 and 218 CEOs for foreign and domestic firms, respectively, and the mean of sales is 0.4 trillion Korean won for foreign firms and 1.05 trillion Korean won for domestic firms.

(8.) We performed a t-test with the null hypothesis that the group proportions of both samples are the same and the alternative hypothesis that the share of foreign-educated CEOs is greater for foreign firms. We were able to reject the null hypothesis at the 5% level (p value = .0322).

(9.) Kim (1998) reports the descriptive analysis of the data in earlier years, and the effect of foreign education on economic growth using the data.

(10.) Instead of the 15-year gap, we also used gaps of different years (ranging from 5 to 25 years) as alternatives, which yielded qualitatively the same result regarding the effect of STDT on FDI (see Section IV).

(11.) Refer to Wheeler and Mody (1992) and Barrell and Pain (1996) for a discussion of these variables.

(12.) Williamson (1975), Dunning (1981). and Edwards (1990) show that there is a tendency for multinational firms to choose host countries with larger market size or greater growth potentials.

(13.) Government policies, country-specific incentives, and political variables have also been suggested to play a role in attracting FDI in Edwards (1990), Wheeler and Mody (1992), and Hines (1996).

(14.) See Cushman (1985), Caves (1989). Froot and Stein (1991). Goldberg and Kolstad (1995), and Blonigen (1997) for discussion on the real exchange rate effect on FDI. The real exchange rate index in our study is the price level of GDP, relative to U.S. prices. An increase in the index implies rising overvaluation in the real exchange rate.

(15.) This is one reason why we do not include the trade variable in our baseline specification.

(16.) See Proposition 3 of Kim and Park (2011) for proof.

(17.) In the 2000-2001 academic year, only 29% of all foreign students from Africa who studied in the United States were enrolled in graduate programs, while 40% of students from Europe were enrolled in graduate programs, as shown in Open Doors, Institute of International Education (2001). According to the UNESCO Statistical Yearbook, the United States is the largest host country of foreign students, attracting more than 30% of foreign students worldwide in 1996, while France is the second largest, with 12% of foreign students.

(18.) We select FDI host countries whose average share of foreign labor in total labor force over the period 1986-1995 is more than 5%, according to World Development Indicators (2001). These countries are Sweden, France, Austria, Germany, Belgium-Luxembourg. the United States, Switzerland, Canada, and Australia.

(19.) The regression result is not reported to save space. Bilateral portfolio investment flows were not available in International Financial Statistics (IFS), 2001.

(20.) The results are not shown to save space. Interestingly, the coefficient on STDT rises in general as the year lag gets greater, peaks in the specification with the 22-year lag (coefficient of 0.3042), and falls as the lag increases beyond 22 years. This may reflect the fact that those students abroad who have acquired It)reign education very recently or long ago do not contribute to the present pool of labor that is conducive to current foreign direct investment.

JINYOUNG KIM and JUNGSOO PARK *

* Jinyoung Kim acknowledges financial support through the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2010-330-B00093). Jungsoo Park was supported by Sogang University Research Grant. We thank the editor, a referee, and participants in seminars at several universities for helpful comments. All errors are exclusively the responsibility of the authors.

Kim: Professor, Department of Economics, Korea University, Anam-dong, Seongbuk-gu, Seoul 136-701, South Korea. Phone 82-2-3290-2202, Fax 82-2-928-4948, E-mail jinykim@korea.ac.kr

Park: Professor. School of Economics, Sogang University, Shinsu-dong, Mapo-gu, Seoul 121-742, South Korea. Phone 82-2-705-8697, Fax 82-2-704-8599, E-mail jspark@sogang.ac.kr

TABLE 1
Educational Backgrounds of CEOs in Foreign and Domestic Firms in Korea

 Foreign Firms Domestic Firms

 Number Percentage Number Percentage
 of CEOs of CEOs

Total 73 218
Educated in Korea 51 69.86 175 80.28
Foreign educated 22 30.14 43 19.72
Countries matched 12 54.55
Not matched 10 45.45

Notes: Among the 100 CEOs of the top 100 foreign multinational firms
in sales that operated in Korea in 2006-2007, we identified the host
countries of tertiary education for 92 CEOs but excluded from our
sample 19 CEOs with non-Korean names. For the sample of CEOs at
domestic firms, we first selected the top 500 domestic firms, and then
excluded those with sales bigger than the largest foreign firm in our
foreign-firm sample. For consistency, we also excluded CEOs of
domestic firms with non-Korean names.

TABLE 2
Descriptions and Summary Statistics of Variables

Variables Descriptions

[STDTN.sub.ij] Number of students from country j who were enrolled
 at institutions of higher education in country i
[STDT.sub.ij] [STDTN.sub.ij] as a share of the population in
 country j
[FDIL.sub.ij] Real FDI from country i to country j
 (1991 U.S. $ billions)
[FDI.sub.ij] [FDIL.sub.ij] as a share of country i's GDP
DISTANCE Distance between the capitals of FDI source and host
 countries (100 km)
LANGUAGE = I if the most popular languages in both countries
 are the same
RELIGION = I if the most popular religions in both countries
 are the same
COLONY = I if the FDI host country used to be a colony of
 the source country
GDP1 (GDP2) Real GDP of an FDI source (host) country
 (1991 U.S. $)
POP2 Population size (1,000)
RELPCGDP Ratio of per capita real GDP of an FDI host country
 to that of a source country
GROW] (GROW2) Real GDP growth rate of an FDI source (host) country
I Domestic investment rate of an FDI host country
G GDP share of government spending in an FDI host
 country
EXCHANGE Real exchange rate of an FDI host country
[TRADE.sub.ij] Real exports of country i to country j
 (1991 U.S. $ millions) as share of country i's GDP
TARIFF Average tariff rate of an FDI host country
TOURIST Total annual number of tourists in an FDI host
 country
TERI (TER2) Tertiary school enrollment rate in an FDI source
 (host) country
SECI (SEC2) Secondary school enrollment rate in an FDI source
 (host) country
STDTNREST Total number of students from country j who studied
 in foreign countries other than country i
FDILREST FDI inflows from countries other than country i
TOTFDIL Total FDI inflow from all source countries in a
 given year
FDISTOCK Stock of total FDI (according to Lane and Milesi-
 Ferretti, 2001)

Variables M SD

[STDTN.sub.ij] 669.3 1,898
[STDT.sub.ij] 0.0439 0.1646
[FDIL.sub.ij] 4.817 17.72
[FDI.sub.ij] 0.0267 0.0975
DISTANCE 58.45 48.02
LANGUAGE 0.2054 0.4040
RELIGION 0.3283 0.4696
COLONY 0.0225 0.1482
GDP1 (GDP2) 1.31e+9 1.75e+9
 (8.10e+8) (1.39e+9)
POP2 84,935 188,486
RELPCGDP 0.9711 1.269
GROW] (GROW2) 0.0395 0.1267
 (0.0578) (0.3334)
I 21.71 6.764
G 17.35 5.354
EXCHANGE 82.37 36.58
[TRADE.sub.ij] 3.927 8.656
TARIFF 13.24 12.38
TOURIST 1.18e+7 1.36e+7
TERI (TER2) 35.10 14.68
 (28.62) (17.19)
SECI (SEC2) 93.48 15.74
 (80.10) (23.76)
STDTNREST 5,973 6,239
FDILREST 67.80 156.2
TOTFDIL 7,012 14,924
FDISTOCK 58,173 112,094

TABLE 3
FDI Regressions

 (1) (2)

 Trade Volume
 Baseline Added

 Coef. SE Coef. SE

In(STDT) 0.240 *** (0.018) 0.167 *** (0.018)
ln(STDTN)
ln(DISTANCE) -0.553 *** (0.031) -0.158 *** (0.041)
LANGUAGE 0.140 ** (0.056) 0.081 (0.054)
RELIGION 0.321 *** (0.047) 0.249 *** (0.046)
COLONY 0.263 * (0.148) 0.143 (0.145)
ln(RELPCGDP) 0.887 * (0.537) 0.378 (0.524)
ln(GDP2) -0.118 (0.789) -0.394 (0.785)
In(GROW1) 0.056 * (0.033) 0.060 * (0.032)
In(GROW2) 0.015 (0.029) 0.013 (0.029)
In(I) 0.138 (0.290) 0.140 (0.297)
In(G) -0.606 ** (0.298) -0.416 (0.293)
In(EXCHANGE) 0.703 *** (0.171) 1.027 *** (0.202)
ln(TRADE) 0.537 *** (0.037)
In(TARIFF)
In(TOURIST)
In(TERI)
In(SEC1)
In(TER2)
ln(SEC2)
ln(GDPI)
In(POP2)
Observations 4,761 4,674
Adjusted [R.sup.2] 0.7137 0.7272

 (3) (4)

 Extra Regressors Dependent
 Added Variable = FDI Level

 Coef. SE Coef. SE

In(STDT) 0.248 *** (0.024)
ln(STDTN) 0.240 *** (0.018)
ln(DISTANCE) -0.597 *** (0.042) -0.559 *** (0.031)
LANGUAGE 0.151 ** (0.076) 0.142 ** (0.056)
RELIGION 0.310 *** (0.064) 0.326 *** (0.047)
COLONY 0.174 (0.197) 0.264 * (0.148)
ln(RELPCGDP) 1.308 (1.250) 0.461 (1.091)
ln(GDP2) -1.651 (1.802) 0.323 (1.263)
In(GROW1) -0.012 (0.052) 0.070 ** (0.034)
In(GROW2) 0.018 (0.044) 0.015 (0.029)
In(I) 0.316 (0.529) 0.114 (0.292)
In(G) -0.201 (0.432) -0.569 * (0.301)
In(EXCHANGE) 0.694 ** (0.347) 0.676 *** (0.170)
ln(TRADE)
In(TARIFF) -0.042 (0.234)
In(TOURIST) 0.562 ** (0.268)
In(TERI) 1.466 *** (0.811)
In(SEC1) 0.424 (0.589)
In(TER2) -0.806 (0.542)
ln(SEC2) 0.417 (0.574)
ln(GDPI) 1.152 (1.100)
In(POP2) -0.593 (1.873)
Observations 2,765 4,765
Adjusted [R.sup.2] 0.7168 0.7334

Notes: The rows show the estimated coefficient and robust standard
error (in parentheses) for each independent variable. *** Significant
at 1%; ** significant at 5%; * significant at 10%.

The dependent variable in all models except Model 4 is the log of the
GDP share of FDI. Model 4 uses the logs of FDI level (FDIL') as the
dependent variable and student number (STDTN) as a regressor, which
allows the logs of GDP] and POP2 to be included as additional
regressors. All models include as regressors source-country-specific
constants and time trends, host-country-specific constants and time
trends, and calendar-year dummies. Domestic human capital variables
(TERI, TER2, SECT, and SEC2) included in Model 3 are lagged 5 years
instead of 15 years owing to the limitation in data availability.

TABLE 4
Sensitivity Analysis: Dyad-Specific Effects

 (1) (2)

 No Dyad-Specific Random Effects
 Effects

 Coef. SE Coef. SE

ln(STDT) 0.348 *** (0.017) 0.323 *** (0.022)

Observations 4,761 4,761
Adjusted [R.sup.2] 0.6434 0.5695

 (3) (4)

 Fixed Effects Between Effects

 Coef. SE Coef. SE

ln(STDT) -0.013 (0.035) 0.213 *** (0.038)

Observations 4,761 4,761
Adjusted [R.sup.2] 0.3407 0.7887

Notes: The rows show the estimated coefficient and robust standard
error (in parentheses) for each independent variable. The coefficient
estimates associated with other regressors besides STDT in all models
are not reported to save space. The dependent variable is the log of
the GDP share of FDI. All models include as regressors source-
country-specific and host country-specific time trends, and calendar-
year dummies. Model 1 excludes dyad-specific effects. Models 2, 3, and
4 employ dyad-specific random, fixed, and between effects,
respectively. These models show estimates based on Model 1 in Table 3.

*** Significant at 1%.

TABLE 5
Sensitivity Analysis: Regional Subsample

 (1) (2)

 North [right arrow] North [right arrow]
 North South

 Coef. SE Coef. SE

ln(STDT) 0.353 *** (0.026) 0.163 *** (0.029)
Observations 2,790 1,622
Adjusted [R.sup.2] 0.7451 0.6961

 (3) (4)

 South [right arrow] W/O China,
 North United States

 Coef. SE Coef. SE

ln(STDT) 0.021 (0.089) 0.223 *** (0.020)
Observations 287 3,982
Adjusted [R.sup.2] 0.6066 0.6786

 (5)

 High Foreign
 Labor Share

 Coef. SE

ln(STDT) 0.261 *** (0.047)
Observations 1,390
Adjusted [R.sup.2] 0.7289

Notes. The rows show the estimated coefficient and robust standard
error (in parentheses) for each independent variable. The coefficient
estimates associated with other regressors besides STDT in all models
are not reported to save space. The dependent variable is the log of
the GDP share of FDI. All models include as regressors source-
country-specific constants and time trends, host-country-specific
constants and time trends, and calendar-year dummies. The countries
included in Model 5 are Australia, Austria, Belgium-Luxembourg,
Canada, France, Germany, Sweden, Switzerland, and the United States.
These models show the estimates based on Model 1 in Table 3.

*** Significant at 1%.

TABLE 6
Sensitivity Analysis: Country-Specific Foreign Students

 (1) (2)

 Students from the Rest Model 4 of Table 3 Rerun
Dependent of the Countries Added with the Smaller Sample
Variable =
 In(FDIL) In(FDIL)

 Coef. SE Coef. SE

ln(STDTN) 0.252 *** (0.021) 0.258 *** (0.021)
In(STDTNREST) 0.154 *** (0.055)
Observations 3,692 3,692
Adjusted [R.sup.2] 0.7307 0.7301

 (3)

 FDI from the Rest of the
Dependent Countries as the Dep. Var.
Variable =
 ln(FDILREST)

 Coef. SE

ln(STDTN) -0.043 *** (0.009)
In(STDTNREST)
Observations 3,692
Adjusted [R.sup.2] 0.8625

Notes: The rows show estimated coefficient and robust standard error
(in parentheses) for each independent variable. The coefficient
estimates associated with other regressors besides STDT in all models
are not reported to save space. The dependent variables are the logs
of FDI level in Models 1 and 2 and of FDt from the rest of the
countries in Model 3. All models include as regressors source-
country-specific constants and time trends, host-country-specific
constants and time trends, and calendar-year dummies. Models 1 and 2
show estimates based on model 4 in Table 3. Model 3 includes only
ln(GDP2), ln(POP2), ln(GROW2), In(I), In(G), and In(EXCHANGE) as
additional regressors.

*** Significant at 1%.

TABLE 7
Sensitivity Analysis: Aggregate FDI

 (1) (2)
Dependent In(TOTFDIL) In(FDISTOCK)
Variable =
 Coef. SE Coef. SE

ln(STDTN) 0.005 (0.006) 0.002 (0.002)
Observations 4,349 4,485
Adjusted [R.sup.2] 0.911 0.9931

Notes: The rows show estimated coefficient and robust standard error
(in parentheses) for each independent variable. The coefficient
estimates associated with other regressors besides STDT in all models
are not reported to save space. The dependent variables in Models 1
and 2 are the logs of aggregate FDI from all countries and of FIJI
stock, respectively. Both models include as regressors source-
country-specific constants and time trends, host-country-specific
constants and time trends, and calendar-year dummies. Both models
include only ln(GDP2), ln(POP2), ln(GROW2), In(I), In(G), and
ln(EXCHANGE) as additional regressors.