Do civil and political repression really boost foreign direct investments?

Harms, Philipp ; Ursprung, Heinrich W.

I. INTRODUCTION

The globalization of the economy is an issue that continues to attract a great deal of attention in the political arena. The exchange of opinion unfortunately often does not follow civilized patterns but is articulated in street riots. The third ministerial conference of the World Trade Organization (WTO) in November-December 1999, for example, gave rise to the now-legendary Battle of Seattle, and the 55th Annual Meeting of the International Monetary Fund (IMF) and the World Bank Group (WB), which took place in September 2000 in Prague, was also accompanied by violent demonstrations.

The arguments of the demonstrating opponents of economic globalization, peaceful or violent, appear to follow a standard pattern. One of the groups demonstrating in Prague, for example, described its objectives as follows:

We will be exposing the links between the IMF/WB, the WTO and transnational corporations and the ways how they work to maximize private profits and limit the power of people to protect the environment, determine their economic destiny, and safeguard their human rights. ... Our goal is to give the proper name to what the policies of the IMF/WB really cause in the South as well as in the Central and Eastern Europe. We will be demanding an immediate suspension of these practices leading to environmental destruction, growing social inequality and poverty and curtailing of people's rights. (1)

In short, globalization is interpreted as a devious maneuver undertaken by multinational firms that, on the one hand, relocate production to undermine the tax and regulation policies of democratic nation states and, on the other hand, exploit the politically and economically repressed workers in third-world autocracies: "Essentially, the WTO, and the 'new' Global Economy, hurt the environment, exploit workers, and disregard civil society's concerns. The only beneficiaries of globalisation are the largest, richest, multinational corporations." (2)

It would be wrong to denigrate these statements as mere battle cries of street fighters because similar patterns of argumentation can be found in the extensive popular literature on globalization. The reproach that multinational enterprises have a special liking for autocratic countries in which workers are not allowed to organize themselves (with the result that the wage rates do not reflect their productivity) can be found, for example, in William Greider's 1998 bestseller One World, Ready or Not: The Manic Logic of Global Capitalism. Greider (1998, 38), in particular, argues against the hypothesis that foreign direct investment (FDI) may have a liberalizing effect in these countries:

The promise of a democratic evolution requires skepticism if the theory is being promoted by economic players who actually benefit from the opposite condition--the enterprises doing business in low-cost labor markets where the absence of democratic rights makes it much easier to suppress wages. A corporation that has made strategic investments based on the cost advantages offered by repressive societies can hardly be expected to advocate their abolition.

Greider understands, of course, that FDI decisions are influenced by balancing labor cost advantages against losses of labor productivity. However, he writes in this context:

The general presumption that low-cost workers in backward countries were crudely unproductive was simply not true. In fact, dollar for dollar, the cheaper workers often represented a better buy for employers than the more skillful workers who were replaced. Their productivity was lower but it also improved rapidly-much faster than their wages. In order to attract foreign capital, their governments often made certain this was the case. (Greider, 1998, 74)

Greider's statements echo a line of argument that has a long tradition in the political-economic literature analyzing the effect of democracy on economic growth. (3) Various authors have postulated that due to their greater ability to enact painful yet efficiency-enhancing policies, autocratic rulers provided a better environment for domestic and foreign investment than democratic governments (see, e.g., Huntington and Dominguez, 1975). Although autocratic regimes obviously also resorted to some redistribution to create loyalties and to consolidate their power, it was claimed that this redistribution usually came in the form of "measures which transfer rights over the control of labor from labor to capital" (Wintrobe, 1998, 161). As a result, regimes characterized by "bureaucratic authoritarianism" (O'Donnell, 1988) were supposed to foster domestic capital accumulation, foreign investment, and growth.

The benefits of authoritarian rule for private investors, however, are not uncontested, and a number of authors argue that the allegedly greater security of property rights offered by repressive regimes may be an illusion. Though an autocratic ruler may shield the owners of capital from the citizens' pressure for redistribution, there are few constraints that limit his or her own behavior, and though the business climate he or she currently offers to investors may be favorable, the investors face a considerable risk of policy reversals--driven either by the dictator's own distributional interests, the need to gather support by sudden turns to populism, or the violent transformation of the entire system by a revolution. This is the point brought forward by Olson (1993) and McGuire and Olson (1996), who contrast the lack of credibility of a short-sighted autocratic ruler with the longer time horizon of the average citizen and stress the importance of democratic checks and balances as a commitment device.

As Przeworski and Limongi (1993) point out, neither of the two positions is completely convincing. The view that dictatorships enhance growth by providing a better protection of property rights "fails to answer why an autonomous state would behave in the interests, long- or short-term ones, of anyone else" (Przeworski and Limongi, 1993, 57) and why an autocratic ruler should place the interests of capital above those of labor. On the other hand, Olson's (1993) more optimistic view of democracy's role in solving the time inconsistency problem does not explain why a country's citizens should have a longer time horizon than a dictator. Hence, what emerges as a conclusion of this literature is that theory does not deliver an unambiguous hypothesis about democracy's impact on investment and growth, and that empirical research has to show which of the forces described dominates.

Given the heated debate about the role of multinational firms in developing countries, it is surprising that there is almost no systematic analysis of the relationship between globalization and the respect for human rights. Do civil and political repression really boost FDI, and are multinational firms attracted by countries in which workers' rights are curtailed, or do they prefer societies in which political participation and civil liberties are guaranteed and in which an organized labor force is able to pursue its interests? In this article we try to answer these questions by systematically examining the hypothesis that multinational enterprises preferably invest in countries in which the working populations' civil and political rights are disregarded.

We find that indices of political and civil repression have a significant influence on foreign direct investment per capita and that this influence is negative. This result emerges both from a cross-section analysis that considers average values for the early to mid-1990s and from panel estimations that exploit the time series variation of the data. (4) Moreover, a greater degree of unionization among workers seems to attract rather than deter foreign investors. The results of our study thus not only support the view that the location decisions of multinational firms are influenced by the host country's political system, they also contradict the widespread perception that international investors are attracted by societies in which political rights are repressed and workers' representation is curtailed.

The rest of the article is structured as follows. The next section presents our data set and performs a cross-section regression using average data for the early and mid1990s. We then exploit the time-series dimension of our data set by performing a fixed-effects panel regression. Finally, we apply a two-step procedure, first identifying the factors that cause the time-series variation in FDI and then checking whether measures of democracy influence the time-invariant country fixed effects. The last section summarizes and concludes.

II. DEMOCRACY AND FDI: AN EMPIRICAL ANALYSIS

Cross-Section Analysis

To investigate whether there is a significant effect of a country's political system on the volume of FDI inflows, we start by considering the relationship between average FDI and a measure of political repression. Our sample consists of 62 developing and emerging-market countries and covers the period between 1989 and 1997. (5)

Annual data on FDI inflows for a large number of developing countries are assembled by the World Bank and published in Global Development Finance (World Bank, 2000). To control for country size, we divide the total volume of FDI by population size. The mean of this ratio for the years 1989 through 1997 is our dependent variable AVFDIPC--the average level of per capita FDI.

To characterize a country's political regime we use indices developed by Raymond Gastil and published on an annual basis by Freedom House (2000). The first of these indices measures the extent of political rights, that is, the people's ability to "participate freely in the political process" (Freedom House, 2000). It ranges from one to seven, with a value of seven reflecting a maximum of political repression. By taking the average of this index for the years 1989 through 1997 we compute the regressor AVPOL (average political repression).

The second index provided by Freedom House measures the extent of civil liberties, that is, the "freedom to develop views, institutions, and personal autonomy apart from the state" (Freedom House, 2000). Like the political rights index, this measure is defined on the interval between one and seven, and the average value for the years 1989 through 1997 is the variable AVCIVIL (average civil repression).

Finally, Freedom House uses the mean of the two measures to classify countries as free, partly free, or not free. The average of this composite index for the years 1989 through 1997 is our regressor AVREPRESS. (6)

The scatterplot in Figure 1 provides a first impression of the correlation between the composite index of political and civil repression (AVREPRESS) and average FDI per capita between 1989 and 1997. Although the data points in this plot are affected by various other factors which we will control for in the following more systematic analysis, there clearly seems to be a negative relationship; that is, FDI per capita is lower for the more repressed countries. This impression is confirmed by the result of a univariate regression given in the first column of Table 1: The coefficient of AVREPRESS is negative and the t statistic indicates a highly significant effect of this variable on average FDI per capita.

To be sure that this result is not just due to the omission of other determinants of FDI, we introduce a number of control variables that we believe to have a significant effect on AVFDIPC. (7) First, we use the illiteracy rate in 1990 (ILLIT) to account for the notion that a higher level of education raises the productivity of capital and thus increases a country's attractiveness for foreign investors. (8) Our second control variable is the dummy OIL, which assumes the value of one whenever a country was a net exporter of oil throughout the 1990s and zero otherwise. Because we conjecture that resource-abundant countries offer higher returns to foreign investors, we expect the coefficient of this variable to be positive. To account for the notion that macroeconomic mismanagement lowers aggregate productivity and deters foreign investors, we also include the average inflation rate between 1989 and 1997 (AVINFLA), which we expect to have a negative influence on FDI. Moreover, because we conjecture that trade ba rriers lower productive efficiency by reducing competition and raising transaction costs, we use two different variables that measure the countries' openness to international trade. Our first measure, AVTRADE, is computed by dividing the sum of a country's exports and imports by its GNP. This measure has the advantage of being available for a large number of countries on an annual basis. Moreover, it reflects administrative as well as natural barriers to trade. As an alternative, we also use the more direct measure of administrative trade openness provided by Gwartney et al. (2001). This trade openness indicator (TOI) reflects the extent of tariff protection, restrictions to capital movements, and other distortions and is defined on a scale from zero to ten, with a higher value reflecting a greater degree of trade openness. Unlike AVTRADE, the indicator AVTOI is not affected by cyclical factors and provides a more precise notion of a country's trade policy. However, it is not available on an annual basis; we therefore used the average of the observations for 1990-92 and 1995-97. We expect that higher values of AVTRADE and AVTOI are correlated with more FDI per capita. (9)

We also use a variable that measures the absence of "political risk" (AVNORISK). This variable is based on the International Country Risk Guide assessment of the likelihood of expropriation, exchange controls, and contract repudiation by a host country's government (see Political Risk Services, 1997 and previous issues). As detailed in Harms (2000), AVNORISK may take values between 0 and 30, with a higher value reflecting a lower degree of political risk. (10) Our last control variable captures the quality of the business environment by measuring corruption in government, the quality of the bureaucracy, and a country's law-and-order tradition. Like AVNORISK, this index is the sum of three subindices published in the International Country Risk Guide. Each individual subindex ranges from zero to six, with a value of six reflecting a minimum of corruption, an efficient and transparent bureaucracy, and a mature law-and-order tradition, respectively. Computing the average of this sum for the years 1989 through 199 7 yields the variable AVBUSCL. A number of recent contributions (Wei, 2000; Smarzynska and Wei, 2000) have demonstrated that, at least at the firm level, greater corruption lowers the incentive to invest in a given country. Hence, we expect AVBUSCL to have a positive influence on foreign direct investment. (11)

Using the set of variables just described we estimate the following equation by ordinary least squares

(1) [AVFDIPC.sub.i] = [[beta].sub.0] + [[beta].sub.1][Repression.sub.i]

+ [[beta].sub.2][ILLIT.sub.i] + [[beta].sub.3] [OIL.sub.i]

+ [[beta].sub.4][AVNORISK.sub.i]

+ [[beta].sub.5][AVINFLA.sub.i]

+ [[beta].sub.6][Tradeopenness.sub.i]

+ [[beta].sub.7][AVBUSCL.sub.i] + [[member of].sub.i]

where [Repression.sub.i] stands for one of the indices [AVREPRESS.sub.i], [AVPOL.sub.i], or [AVCIVIL.sub.i]; and [Tradeopenness.sub.i], for one of the measures [AVTRADE.sub.i] or [AVTOI.sub.i]. As shown in the second column of Table 1, the negative effect of our repression index survives the inclusion of the control variables described above. The sign of the coefficient is still negative, and although the t statistic has decreased, it is still above the critical value for a 5% level of significance.

With the exception of AVINFLA, the control variables in the second column of Table 1 have the expected sign, but none of them seem to have a significant influence on FDI per capita. Though collinearity among the regressors may have contributed to this result, the third column of Table 1 demonstrates that the fit of the model can be substantially improved if we use the natural log of AVFDIPC as the dependent variable. (12) In this semilog specification, almost all control variables have a significant influence, and the t statistic of AVREPRESS increases. (13)

It is, of course, possible that our results suffer from endogeneity. Starting with Lipset (1959), a large literature discusses the effect of a country's prosperity on its political regime, (14) and it is conceivable that our results are driven by a positive influence of FDI on democratization. However, we believe that if direct investment affects a country's political regime, this process takes much more time than investors' reactions to changing political circumstances. Hence, the fact that our panel covers a rather short time interval should reduce problems of endogeneity. To check this presumption, we used the 1989 value of the composite measure of political and civil repression instead of AVREPRESS. It turned out that this did not substantially alter our results. (15)

Moreover, it is apparent from the fourth and the fifth columns of Table 1 that the two elements of the composite index differ substantially in their effect. AVCIVIL, which reflects the repression of civil liberties has both a higher coefficient (in absolute value), and its t statistic is higher than the corresponding value for AVPOL, which suggests that international investors put a greater weight on the guarantee of civil liberties than on the extent of political participation. Because the "existence of free businesses and cooperatives" is a criterion for the civil repression index, these results might indicate that investors are attracted by regimes that offer a high degree of economic freedom, while other features like a free press, personal social freedoms, and so on are of secondary importance. To isolate the effect of purely economic freedom, we use the index of economic freedom assembled by Gwartney et al. (2000) as an additional regressor. This measure is defined on a zero-to-ten scale, with a higher value reflecting a greater degree of freedom; because it is not available on an annual basis, we used the average of the values for 1990 and 1997 to compute AVECEREE. The first column in Table 2 shows that when we include AVECEREE, the coefficient and the t statistic of the civil repression index do indeed decrease in absolute value. (16) Nevertheless, the significance levels of both variables are still low enough to suggest that although economic freedom is a prerequisite for foreign direct investment, foreign investors also shun countries in which other aspects of civil liberties are suppressed.

The third through sixth columns of Table 1 and the first column of Table 2 show that in the estimation of the semilog version of equation (1), the average inflation rate has a significantly positive effect on average per capita FDI. This result points at a major shortcoming of the approach we followed so far: Using averages over a longer time period neglects the fact that in some countries the regressors have changed dramatically during the period under consideration. A striking example is Argentina, whose inflation rate decreased from 3,080% in 1989 to less than 1% in 1997. Despite this formidable improvement, Argentina's average inflation rate for the entire time period is still very high (623%). On the other hand, Argentina has attracted huge volumes of FDI in the mid-1990s. (17) Without knowing the evolution of the time series, one might naively interpret these numbers as evidence of a positive relationship between inflation and FDI. Of course, this is just a consequence of using averages. This discussion emphasizes the importance of accounting for the time-series variation of the variables. In the following section we exploit both the cross-section and the time-series dimension of our data.

However, before doing this, it is worthwhile to consider another variable that plays an important role in the reasoning of globalization opponents. As Greider (1998) emphasizes, the postulated positive effect of political repression on FDI is due to a curtailment of workers' rights and representation. To verify whether this line of argument is supported by the empirical evidence, we add a measure of trade union density (UNION) to the semilog version of equation (1). (18) The second column of Table 2 shows that the coefficient of UNION has a positive sign but with a rather low t statistic. This result is not surprising because a high degree of union membership does not automatically reflect a strong representation of workers' interests. In a repressive regime, unions are used to control workers and to enforce the government's decisions. To account for this effect, we generated two interactive measures of effective unionization: UNIONEFF1 is computed by dividing UNION by the index of civil repression AVCIVIL. ( 19) UNIONEFF2 results from calculating UNION. (8-AVCIVIL). Both measures are admittedly crude, but it is striking that (as shown by the third and fourth columns in Table 2) both have a significant positive effect on FDI per capita. Hence, our cross-section analysis does not support the notion that a repressed work force is helpful to attract foreign investors. In fact, the opposite seems to be the case.

Panel Data Estimation

We now turn to a specification of the empirical model that also takes into account the time series variation of the data. Table 3 presents the result of estimating the following equation.

(2) [FDIPC.sub.it] = [[alpha].sub.i] + [[lambda].sub.t] + [[beta].sub.1][Repression.sub.it] + [[beta].sub.2][NORISK.sub.it] + [[beta].sub.3][INFLA.sub.it] + [[beta].sub.4][TRADE.sub.it] + [[beta].sub.5][BUSCL.sub.it] + [[epsilon].sub.it]

In (2), [FDIPC.sub.it] is per capita FDI in country i in the year t. (20) [[alpha].sub.i] is a country-specific fixed effect and [[lambda].sub.t] is a dummy that is used to capture time-varying factors that affect all countries. [Repression.sub.it] stands for one of the indices [REPRESS.sub.it], [POL.sub.it], or [CIVIL.sub.it]. Finally, [NORISK.sub.it], [INFLA.sub.it], [TRADE.sub.it], and [BUSCL.sub.it] reflect the absence of political risk, the inflation rate, trade openness, and the business climate for country i in the year t. (21) We omit the illiteracy rate (ILLIT) and the oil dummy (OIL) in equation (2) because the fixed effects approach requires that there is some within-group variation of the regressors. (22) Moreover, we only use TRADE as a measure of trade openness because the alternative TOI is not available on an annual basis.

The first column in Table 3 shows that changes in the absence of political risk as measured by our variable NORISK have a significant effect on FDI per capita. Moreover, INFLA has the correct sign but is not significant. The poor performance of the inflation rate (INFLA) is mainly due to the inclusion of Zaire, which went through a dramatic hyperinflation in the mid-nineties. The second column of Table 3 shows that if we drop Zaire from the sample, the coefficient and the t tatistic of INFLA increase in absolute value. Changes in trade openness and business climate as measured by TRADE and BUSCL have no significant effect on FDI per capita, but the time dummies for the years 1994 onward (not reported in the table) have positive and highly significant coefficients. This seems to capture the overall emerging markets frenzy that characterized the mid1990s and that was rather triggered by changing perceptions of international investors than by changing conditions in individual host countries.

The composite index of political and civil repression (REPRESS) still has the correct sign, but the t statistic is just slightly above the critical value for a 10% level of significance. Foreign investors seem to have honored the democratization that took place in many emerging markets in the early 1990s, but we cannot confidently reject the hypothesis that there is no real relationship between the two magnitudes. In particular, if--apart from Zaire--we drop the former socialist countries Hungary, Poland, and Romania where REPRESS decreased substantially during the 1990s, the t statistic drops to a value of -0.90 (see the third column in Table 3). When we use the individual measures of political and civil repression, POL and CIVIL, it turns out that, for the entire sample (minus Zaire), POL has a strong effect, and the coefficient of CIVIL is much lower in absolute value and not significantly different from zero. This confirms our conjecture that the result in the second column of Table 3 is predominantly driven by the substantial transformation of the political regimes in the three former socialist countries and explains why the coefficient of our democracy measure becomes insignificant once we drop these countries from the sample. For some countries the measure of freedom varied substantially during the period under consideration, but there are many cases where this variation was negligible or completely absent. The most striking example is China, which witnessed a huge increase of FDI inflows in the 1990s but where the repression index did not change at all and remained at a value of seven throughout this period. Hence, while time-varying regressors caused trouble in the preceding subsection, our results are now blurred by a lack of time-series variation. (23) The following section presents an approach that is able to cope with this dilemma.

A Two-Step Procedure

The fixed-effects approach adopted in the preceding section is a powerful remedy against omitted variable bias because the coefficients of the country dummies capture all country-specific FDI determinants that do not change over time. As we have seen, however, this advantage does not come without costs, because it requires that all regressors exhibit substantial variation over time. Moreover, while revealing how changes of a regressor affect changes of FDI in a given country, the results of a fixed-effects regression do not really explain why levels of FDI differ across countries.

To solve this problem, we will now follow the two-step procedure applied, for example, by Dickens and Katz (1987) and Olson et al. (2000). To account for the time-series variation of our regressors, we start by estimating equation (2) but omitting the repression index and our measure of the business climate. Apart from a vector of coefficients this yields an estimate of the fixed effect [[alpha].sub.i] for each country. These fixed effects reflect the influence of time-invariant factors on the volume of FDI per capita in country i.

In a second step, we estimate the equation

(3) [[alpha].sub.i] = [[beta].sub.0] + [[beta].sub.1][Repression.sub.i] + [[beta].sub.2][ILLIT.sub.i] + [[beta].sub.3][OIL.sub.i] + [[beta].sub.4][AVBUSCL.sub.i] + [[epsilon].sub.i].

The results of this regression are given in Table 4. Both the measure of political and civil repression AVREPRESS and the illiteracy rate ILLIT have the expected sign and are significant, whereas the oil dummy OIL and the measure of the business climate AVBUSCL do not exert a significant influence on direct investment. (24) Similar results emerge when we use our measures of political and civil repression, AVPOL and AVCIVIL, instead of the composite index. As in Table 1, the measure of civil repression has both a larger coefficient (in absolute value) and a higher t statistic than the index of political repression. Hence, ceteris paribus, a greater extent of repression, in particular a restriction of civil liberties, seems to reduce rather than increase a country's attractiveness for foreign investors.

The fourth column in Table 4 indicates that at least our first measure of effective unionization has a significantly positive effect on foreign direct investment per capita. As in the cross-section analysis, this suggests that instead of deterring multinational firms, a stronger representation of workers' interests induces more foreign direct investment.

III. SUMMARY AND CONCLUSIONS

The often-heard claim that multinational enterprises, in choosing host countries for their FDI, have a preference for undemocratic regimes that deny their citizens basic human rights and suppress worker representation is not supported by our empirical investigation. On the contrary, individual freedom rather appears to attract FDI. International economic integration may thus generate more economic advantages for countries respecting civil and political rights than for repressive autocracies.

On one hand, this conclusion is very much in harmony with the existing inquiries into the other two main lines of attack that the opponents of globalization usually resort to--namely, the claims that economic integration destroys the environment and debilitates democratic public decision-making processes. Surveying the extensive theoretical and empirical literature on international environmental economics and on welfare state policies in integrated economies, one is led to conclude that the globaphobic world view is simply not supported by comprehensive scholarly investigations into these complex issues. (25)

On the other hand, our result that more liberal regimes attract a larger volume of foreign direct investment is much sharper than the rather inconclusive evidence provided by the empirical literature on democracy and growth. (26) To explain this difference, one has to take into account that most empirical studies on the democracy-growth relationship use data sets that cover the time between the early 1960s and the mid-1980s. Our investigation, instead, refers to a much more recent period, in which technologies have become increasingly sophisticated and in which an educated labor force has become more valuable as a factor of production than unskilled workers. As a consequence, the labor-capital antagonism of the Cold War era has been replaced by a much more symbiotic relationship between physical and human capital; instead of proposing harmful redistribution, skilled domestic workers may actually share the interests of multinational firms and represent these interests in the domestic political process. This ma y be the reason why the protection against workers' distributional interests allegedly offered by autocratic rulers has lost much of its appeal and why foreign investors give a greater weight to the accountability and predictability of democratic governments.

APPENDIX

DEFINITIONS AND SOURCES

AVX: Average value of variable X between 1989 and 1997 (unless indicated otherwise).

FDIPC: Foreign direct investment in U.S. dollars per capita. Source: World Bank (2000) and IMF (2000).

CIVIL: Gastil index of civil liberties. Scale: 1-7 (maximal repression). Source: Freedom House (2000) and previous issues.

POL: Gastil index of political rights. Scale: 1-7 (maximal repression). Source: Freedom House (2000) and previous issues.

REPRESS: Gastil composite index of political rights and civil liberties (REPRESS = [CIVIL + POL]/2). Source: Freedom House (2000) and previous issues.

ILLIT: Illiteracy rate in 1990 in percent. Source: UNESCO (1999).

OIL: Dummy variable: 1, if oil exports throughout 1990s greater than imports; 0 otherwise. Source: United Nations (1998).

NORISK: Measure for likelihood of expropriations, exchange controls, and default on government contracts. Scale: 0-30 (minimal risk). Source: Political Risk Services (1997) and previous issues (for a given year, the value in the December issue of the previous year is used).

BUSCL: Measure for corruption, quality of bureaucracy, and law-and-order tradition (annual averages). Scale: 0-18 (optimal business climate). Source: Political Risk Services (1999).

INFLA: Annual inflation rate in percent. Source: IMF (2000).

TRADE: (Exports + imports)/GNP. Source: World Bank (2000).

AVTO1: Trade openness indicator, reflecting tariff rates, black market exchange premium, restrictions on capital movements, and deviations of the actual size of the trade sector from the expected size (average of values for 1990-92 and 1995-97). Scale: 0-10 (maximal openness). Source: Gwartney et al. (2001).

AVECFREE: Index of economic freedom (average of values for 1990 and 1997). Scale: 0-10 (maximal economic freedom). Source: Gwartney et al. (2000).

UNION: Percentage share of unionized workers in nonagricultural sectors in 1995. Source: ILO (1997).

UNIONEFF1: UNION/AVCIVIL.

UNIONEFF2: UNION. (8-AVCIVIL).

COUNTRIES IN THE SAMPLE (27)

Algeria, Argentina, Bangladesh, * Bolivia, Botswana, Brazil, Burkina Faso, * Cameroon, Chile, China, Costa Rica, Cote d'Ivoire, Democratic Republic of Congo (Zaire), * Dominican Republic, Ecuador, Egypt, El Salvador, Ethiopia, Gabon, Ghana, Guatemala, Haiti, * Honduras, Hungary, India, Indonesia, Jamaica, * Jordan, Kenya, Republic of Korea, Madagascar, * Malawi, * Malaysia, Mali, Mexico, Morocco, Mozambique, * Nicaragua, Niger, * Nigeria, Pakistan, Panama, Paraguay, Peru, Philippines, Poland, Romania, Senegal, Sierra Leone, * Sri Lanka, * Sudan, * Syria, * Tanzania, Thailand, Togo, * Tunisia, Turkey, Uganda, Uruguay, Venezuela, Zambia, Zimbabwe.

TABLE A1

Descriptive Statistics and Correlations

 AVREPRESS AVPOL AVCIVIL ILLIT OIL AVNORISK AVTOI

Mean 4.09 4.05 4.13 33.91 0.25 18.99 5.44
SD 1.43 1.66 1.25 24.18 0.44 3.74 1.44
AVREPRESS 1.00 0.99 0.98 0.54 0.24 -0.47 -0.44
AVPOL 1.00 0.93 0.54 0.24 -0.44 -0.38
AVCIVIL 1.00 0.53 0.24 -0.49 -0.49
ILLIT 1.00 -0.13 -0.67 -0.62
OIL 1.00 0.29 0.01
AVNORISK 1.00 0.56
AVINFLA -0.12 -0.14 -0.09 -0.14 -0.08 0.05 -0.06
AVTRADE -0.17 -0.11 -0.23 -0.23 0.00 0.22 0.53
AVBUSCL -0.31 -0.28 -0.33 -0.47 0.07 0.67 0.23
AVECFREE -0.57 -0.56 -0.57 -0.60 0.02 0.61 0.84
UNION 0.04 0.03 0.05 -0.30 0.17 0.17 -0.03
UNIONEFF1 -0.41 -0.38 -0.42 -0.46 -0.01 0.31 0.08
UNIONEFF2 -0.42 -0.40 -0.44 -0.44 -0.03 0.27 0.06
 AVINFLA AVTRADE AVBUSCL AVECFREE UNION UNIONEFF1

Mean 102.64 0.75 8.86 5.66 16.66 4.86
SD 311.57 0.42 2.21 1.16 11.97 4.17
AVINFLA 1.00 -0.09 0.08 -0.07 0.14 0.11
AVTRADE 1.00 0.14 0.28 -0.03 0.02
AVBUSCL 1.00 0.18 0.43 0.55
AVECFREE 1.00 -0.16 0.06
UNION 1.00 0.82
UNIONEFF1 1.00

 UNIONEFF2

Mean 68.58
SD 54.22
AVINFLA 0.17
AVTRADE 0.04
AVBUSCL 0.48
AVECFREE 0.02
UNION 0.80
UNIONEFF1 0.97

Notes: Mean of dependent variable (AVFDIPC): 26.88. Standard deviation:
44.98. Descriptive statistics and correlations refer to the reduced
sample (without Madagascar and Zaire) that is used for the regression
in the second column of Table 1.

[FIGURE 1 OMITTED]

TABLE 1

Results of the Cross-Section Analysis

 AVFDIPC In(AVFDIPC)

Constant 81.35 *** 5.56 1.00 0.51
 4.79 (0.11) (0.85) (0.45)
AVREPRESS -13.44 *** -7.76 ** -0.32 ***
 (-3.96) (-2.16) (-2.83)
AVPOL -0.25 **
 (-2.48)
AVCIVIL

ILLIT -0.37 -0.03 *** -0.03 ***
 (-1.63) (-3.63) (-3.62)
OIL 17.15 1.18 *** 1.12 ***
 (1.20) (3.34) (3.20)
AVNORISK 0.98 0.09* 0.10 *
 (0.44) (1.77) (1.95)
AVINFLA 0.006 0.001 * 0.001 *
 (0.37) (1.84) (1.82)
AVTRADE 25.77 1.01 *** 1.06 ***
 (1.27) (3.60) (3.71)
AVTOI

AVBUSCL 2.56 0.05 0.05
 (0.82) (0.72) (0.71)
Number of observations 62 60 58 58
[R.sup.2] 0.18 0.34 0.69 0.68

 In(AVFDIPC)

Constant 1.54 -0.73
 (1.23) (-0.45)
AVREPRESS -0.26 *
 (-1.82)
AVPOL

AVCIVIL -0.39 ***
 (-3.11)
ILLIT -0.03 *** -0.02 *
 (-3.75) (-1.77)
OIL 1.23 *** 1.10 ***
 (3.40) (2.93)
AVNORISK 0.09 0.03
 (1.59) (0.37)
AVINFLA 0.001 * 0.001 *
 (1.91) (1.88)
AVTRADE 0.94 ***
 (3.41)
AVTOI 0.50 ***
 (3.06)
AVBUSCL 0.05 0.11
 (0.68) (1.15)
Number of observations 58 47
[R.sup.2] 0.69 0.61

Note: t statistics are based on White's (1980) heteroskedasticity-
consistent covariance-matrix.

***, **, *: significance levels of 1%, 5%, 10%.
TABLE 2

Cross-Section Results

 In(AVFDIPC)

Constant -0.69 0.27 -1.37 -1.46
 (-0.38) (0.21) (-1.09) (-1.14)
AVCIVIL -0.31 * -0.40 **
 (-1.86) (-2.63)
ILLIT -0.02 ** -0.02 * -0.02 ** -0.02 **
 (-2.47) (-1.87) (-2.63) (-2.68)
OIL 1.29 *** 1.20 *** 0.93 *** 0.94 ***
 (3.85) (5.16) (4.18) (4.34)
AVNORISK -0.01 0.16 *** 0.19 *** 0.19 ***
 (-0.12) (2.80) (3.04) (2.98)
AVINFLA 0.001 ** 0.001 0.001 0.001
 (2.03) (1.44) (1.59) (1.44)
AVTRADE 0.73 ** 1.00 *** 1.07 *** 1.04 ***
 (2.56) (3.17) (3.64) (3.44)
AVBUSCL 0.16 -0.02 -0.08 -0.06
 (1.62) (-0.24) (-0.96) (-0.77)
AVECFREE 0.45 *
 (1.75)
UNION 0.02 *
 (1.70)
UNIONEFF1 0.09 ***
 (3.11)
UNIONEFF2 0.006 ***
 (2.81)
Number of observations 54 46 46 46
[R.sup.2] 0.66 0.70 0.69 0.69

Note: t statistics are based on White's (1980) heteroskedasticity-
consistent covariance-matrix.

***, **, *: significance levels of 1%, 5%, 10%.
TABLE 3

Results of the Fixed-Effects Regressions

 FDIPC
 Without
 Zaire,
 Hungary,
 Without Poland,
 Entire Sample Zaire Romania

REPRESS -3.54 * -3.69 * -1.91
 (-1.71 (-1.76) (-0.90)
POL

CIVIL

NORISK 1.50 ** 1.18 * 1.11
 (2.32) (1.76) (1.61)
INFLA -0.001 -0.006 *** -0.007 ***
 (-1.10) (-2.62) (-2.73)
TRADE -10.30 -12.59 -9.72
 (-0.79) (-0.92) (-0.73)
BUSCL -0.17 0.10 -0.12
 (-0.16) (0.10) (-0.12)
Time dummies yes yes yes
Number of observations 558 549 522
[R.sup.2] 0.60 0.60 0.62

 FDIPC



 Without Without
 Zaire Zaire

REPRESS

POL -3.75 **
 (-2.03)
CIVIL -1.22
 (-0.66)
NORISK 1.25 * 0.96
 (1.84) (1.52)
INFLA -0.007 *** -0.006 **
 (-2.68) (-2.52)
TRADE -13.33 -10.89
 (-0.97) (-0.80)
BUSCL 0.03 0.32
 (0.03) (0.31)
Time dummies yes yes
Number of observations 549 549
[R.sup.2] 0.60 0.60

Note: t statistics are based on White's (1980) heteroskedasticity-
consistent covariance-matrix.

***, **, *,: significance levels of 1%, 5%, 10%.
TABLE 4

Determinants of Country-Specific Fixed Effects

 [[alpha].sub.i] (Country Fixed Effect)

Constant 29.20 20.54 39.65
 (0.84) (0.61) (1.06)
AVREPRESS -7.63 **
 (-2.41)
AVPOL -5.60 **
 (-2.13)
AVCIVIL -9.68 **
 (-2.50)
ILLIT -0.52 ** -0.55 ** -0.51 **
 (-2.30) (-2.28) (-2.46)
OIL 13.04 11.77 13.72
 (0.87) (0.78) (0.92)
AVBUSCL 2.37 2.55 2.12
 (0.72) (0.77) (0.65)
UNIONEFF1

UNIONEFF2

Number of observations 60 60 60
[R.sup.2] 0.23 0.22 0.24

 [[alpha].sub.i]
 (Country Fixed Effect)

Constant 13.00 7.43
 (0.29) (0.17)
AVREPRESS

AVPOL

AVCIVIL

ILLIT -0.78 ** -0.79 **
 (-2.62) (-2.63)
OIL 11.94 12.29
 (0.68) (0.71)
AVBUSCL 0.95 1.63
 (0.20) (0.37)
UNIONEFF1 3.51 *
 (1.72)
UNIONEFF2 0.24
 (1.64)
Number of observations 48 48
[R.sup.2] 0.25 0.24

Note: t statistics are based on White's (1980) heteroskedasticity-
heteroskedasticity-consistent covariance-matrix.

***, **, *: significance levels of 1%, 5%, 10%.

(1.) See http://southcom.homestead.com/prague.html.

(2.) See www.s11.org/s11-dynamic.html and www.s11.org/wef.html.

(3.) See Przeworski and Limongi (1993) for an excellent survey.

(4.) In a recent study focusing on foreign aid, Alesina and Dollar (2000) find no significant effect of freedom on FDI as a share of gross national product (GNP). However, this result may be driven by the fact that both foreign investment and aggregate income are affected by democracy, whereas the ratio of the two magnitudes may not depend on the political system.

(5.) Our choice of this period was determined by limited data availability and by our desire to gather a balanced panel for a large number of developing countries.

(6.) This label accounts for the fact that a higher value of the composite index reflects a greater degree of political and civil repression.

(7.) Our choice of control variables is inspired by a number of earlier studies on the determinants of FDI, namely, Schneider and Frey (1985), Wheeler and Mody (1992), and Jun and Singh (1996). Moreover, we tried to find a reasonable balance between the number of regressors and the size of our sample, neglecting potential determinants of FDI whenever reliable data were only available for a small number of countries. A detailed description of all variables and their sources is given in the appendix.

(8.) Because the illiteracy rate as a measure of a country's human capital stock exhibits a strong negative correlation with per capita income, this variable also accounts for the fact that richer economies are better locations for market-seeking FDI. Note that illiteracy rates are not available for Zaire and Madagascar, which shrinks our sample size to 60 countries.

(9.) Note, however, that a less liberal trade regime may also raise the volume of FDI by encouraging tariff-hopping (see Jun and Singh [1996]).

(10.) Because existing investment barriers are also captured by AVNORISK, this variable also accounts for the fact that many repressive states severely restricted the inflow of foreign capital until the early 1990s.

(11.) Knack and Keefer (1995) and numerous subsequent studies use individual components of our measures AVNORISK and AVBUSCL to construct measures that represent the security of property rights and the quality of the institutional environment. Replacing our two separate variables AVNORISK and AVBUSCL by a composite measure that reflects the risks of expropriation and contract repudiation as well as the extent of corruption and the rule of law does not substantially alter our results.

(12.) Table A1 in the appendix shows that the partial correlations among the regressors are not very large. Moreover, the estimated coefficients do not vary by much if we remove some randomly selected observations from our sample. Hence, we do not believe that our results are seriously affected by collinearity, and rather than producing biased estimates by omitting variables that are correlated with our measures of political and civil repression, we decided to include such variables in our regressions. The superiority of the semilog specification is confirmed by an appropriate Box-Cox test. Note that because the values of AVFDIPC for Cameroon and Gabon are negative, the natural log cannot be computed for these countries, and our sample size shrinks to 58. However, the improved fit of the model is not due to the exclusion of these data points.

(13.) The sixth column of Table 1 shows that the coefficient and the t statistic of AVREPRESS decrease in absolute value if we use AVTOI instead of AVTRADE as a measure of trade openness. This is due to the fact that the Gwartney et al. (2001) index covers only 49 countries of our sample, neglecting many African and some Eastern European countries. In this subsample, the measure of political and civil repression exhibits considerably less variation, which results in a lower t statistic.

(14.) For more recent contributions on this question see, for example, Huber et al. (1993) and Londregan and Poole (1996).

(15.) More specifically, the coefficient of the repression variable in this regression was -0.28 and the t statistic -2.61.

(16.) Note, however, that these results are not directly comparable to those in Table 1 (fifth column) because the Gwartney et al. (2000) index of economic freedom covers only a subset of the countries in our sample. Moreover, because the index of economic freedom depends on criteria like past inflation rates and the absence of expropriation risk (which we capture by separate regressors), there is a high risk of collinearity. To account for this problem, we also ran regressions in which we dropped the variables AVNORISK, AVINFLA, AVTRADE, and AVBUSCL. In these regressions (not reported here), the coefficients and t statistics of AVCIVIL and AVECFREE were -0.37 (-2.31) and 0.37 (1.97), respectively.

(17.) The value of AVFDIPC for Argentina is US $116 per capita.

(18.) UNION is the percentage share of union members among workers in a country's nonagricultural sector in 1995, as given by the International Labour Office (ILO, 1997). Note that, due to limited data availability, the inclusion of this variable in our regression reduces our sample to 46 observations.

(19.) We chose AVCIVIL to compute effective unionization since the civil repression index explicitly depends on the existence of free trade unions.

(20.) As demonstrated in the preceding subsection, a semilog version of this equation would be preferable. However, this alternative is not feasible because a large number of observations on annual FDI is negative.

(21.) The restriction of imposing a common intercept on all countries is clearly rejected by an appropriate F-test. Moreover, the result of a Hausman (1978) test prevents us from using the random effects model.

(22.) As detailed in Hsiao (1986) and Baltagi (1995), the fixed-effects approach amounts to regressing deviations from the within-group mean of the dependent variable on deviations from the within-group means of the regressors. However, the oil dummy evidently does not change over time, and since UNESCO provides illiteracy rates on a five-year basis only, there are simply not enough data points to include ILLIT in the regression.

(23.) We believe that the low t statistic of the business climate variable BUSCL is also due to a lack of time-series variation. Many countries witnessed massive improvements of their political risk ratings during the 1990s, but the level of corruption, the quality of the bureaucracy, and the rule of law turned out to be much more persistent, and values of BUSCL hardly changed over time.

(24.) Table 1 has shown that OIL has a significantly positive effect if a semilog specification is used. Hence, the low t statistic of the dummy in Table 4 should be interpreted with caution.

(25.) See, for example, Schulze and Ursprung (1999) and (2001), respectively.

(26.) Scully (1992), Przeworski and Limongi (1993), Benabou (1996), Brunetti (1997), Drazen (2000), and Haan and Storm (2000) provide recent surveys of this literature.

(27.) For countries that are marked with an asterisk, no unionization data are available.

REFERENCES

Alesina, A., and D. Dollar. "Who Gives Foreign Aid to Whom and Why?" Journal of Economic Growth, 2000, 33-63.

Baltagi, B. Econometric Analysis of Panel Data. New York: Wiley, 1995.

Benabou, R. "Inequality and Growth," in NBER Macroeconomics Annual 1996, edited by B. S. Bernanke and J. J. Rotemberg. Cambridge: MIT Press, 1996, 11-74.

Brunetti, A. "Political Variables in Cross Country Growth Analysis." Journal of Economic Surveys, 11, 1997, 163-90.

Dickens, W T, and L. Katz. "Inter-Industry Wage Differences and Industry Characteristics," in Unemployment and the Structure of Labour Markets, edited by K. Lang and J. S. Leonard. Oxford: Basil Blackwell, 1987, 48-89.

Drazen, A. Political Economy in Macroeconomics. Princeton: Princeton University Press, 2000.

Freedom House. Freedom in the World. New York: Freedom House, 2000.

Greider, W One World, Ready or Not: The Manic Logic of Global Capitalism. New York: Touchstone, 1998.

Gwartney, J., R. Lawson, and D. Samida. Economic Freedom of the World: 2000 Annual Report. Vancouver, BC: Fraser Institute, 2000.

Gwartney, J., R. Lawson, W Park, and C. Skipton. Economic Freedom of the World: 2001 Annual Report. Vancouver, BC: Fraser Institute, 2001.

Haan, J. de, and J. E. Sturm. "On the Relationship between Economic Freedom and Economic Growth." European Journal of Political Economy, 16, 2000, 215-41.

Harms, P. International Investment, Political Risk; and Growth. Boston: Kluwer, 2000.

Hausman, J. "Specification Tests in Econometrics." Econometrics, 46, 1978, 1251-71.

Hsiao, C. Analysis of Panel Data. New Haven: Yale University Press, 1986.

Huber, E., D. Rueschemeyer, and J. Stephens. "The Impact of Economic Development on Democracy." Journal of Economic Perspectives, 7, 1993, 71-85.

Huntington, S. P., and J. I. Dominguez. "Political Development," in Handbook of Political Science 3, edited by F. I. Greenstein and N. W Polsby. Reading, PA: Addison-Wesley, 1975, 1-114.

International Labour Office. World Labour Report-- Industrial Relations, Democracy and Social Stability. Geneva: ILO, 1997.

International Monetary Fund. International Financial Statistics Yearbook 2000. Washington, DC: IMF, 2000.

Jun, K. W, and H. Singh. "The Determinants of Foreign Direct Investment in Developing Countries." Transnational Corporations, 5, 1996, 67-105.

Knack, S., and P. Keefer. "Institutions and Economic Performance: Cross-Country Tests Using Alternative Institutional Measures." Economics and Politics, 7,1995, 201-27.

Lipset, S. M. "Some Social Requisites of Democracy: Economic Development and Political Legitimacy." American Political Science Review, 53, 1959, 69-105.

Londregan, J. B., and K. T Poole. "Does High Income Promote Democracy?" World Politics, 49, 1996, 1-30.

McGuire, M. C., and M. Olson. "The Economics of Autocracy and Majority Rule: The Invisible Hand and the Use of Force." Journal of Economic Literature, 34, 1996, 72-96.

O'Donnell, G. Bureaucratic Authoritarianism. Argentina, 1966-73, in Comparative Perspective. Berkeley: University of California Press, 1988.

Olson, M. "Dictatorship, Democracy, and Development." American Political Science Review, 87, 1993, 567-76.

Olson, M., N. Sarna, and A. Swamy. "Governance and Growth: A Simple Hypothesis Explaining Cross-Country Differences in Productivity Growth." Public Choice, 102, 2000, 341-64.

Political Risk Services. International Country Risk Guide (December). East Syracuse, NY: Political Risk Services, 1997.

-----. IRIS 3 File of ICRG Data. East Syracuse, NY: Political Risk Services, 1999.

Przeworski, A., and F. Limongi. "Political Regimes and Economic Growth." Journal of Economic Perspectives, 7, 1993, 51-69.

Schneider, F, and B. S. Frey. "Economic and Political Determinants of Foreign Direct Investment." World Development, 13, 1985, 161-75.

Schulze, G., and H. W. Ursprung. "Globalization of the Economy and the Nation State." World Economy, 22, 1999, 295-352.

-----, eds. International Environmental Economics: A Survey of the Issues. Oxford: Oxford University Press, 2001,

Scully, G. W Constitutional Environments and Economic Growth. Princeton: Princeton University Press, 1992.

Smarzynska, B., and S.-J. Wei, "Corruption and Composition of Foreign Direct Investment: Firm-Level Evidence." NBER Working Paper No. 7969, 2000.

UNESCO. UNESCO Statistical Yearbook 1999. Paris: UNESCO, 1999.

United Nations. United Nations Statistical Yearbook 1998. New York: United Nations, 1998.

Wei, S.-J. "How Taxing Is Corruption on International Investors?" Review of Economics and Statistics, 82, 2000, 1-11.

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

White, H. "A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity." Econometrica, 48, 1980, 817-38.

Wintrobe, R. The Political Economy of Dictatorship. Cambridge: Cambridge University Press, 1998.

World Bank. Global Development Finance 2000. Washington, DC: World Bank, 2000.

RELATED ARTICLE: ABBREVIATIONS

FDI: Foreign Direct Investment

GNP: Gross National Product

ILO: International Labor Office

IMF: International Monetary Fund

TOI: Trade Openness Indicator

WB: World Bank

WTO: World Trade Organization

HEINRICH W. URSPRUNG *

* We are indebted to Heike Henckel for valuable research assistance and to Erich Weede, Christoph Eisenring, Gunther Schulze, and two anonymous referees of this journal for very helpful comments.

Harms: Assistant Professor, University of Konstanz, Department of Economics, Box D138, D-78457 Konstanz, Germany. Phone +49(0)7531-88-3711, Fax +49(0)7531-88-3130, E-mail philipp.harms@unikonstanz.de

Ursprung: Professor, University of Konstanz, Department of Economics, Box D138, D-78457 Konstanz, Germany. Phone +49(0)7531-88-2332, Fax +49(0)7531-88-3130, E-mail heinrich.ursprung@unikonstanz.de