Host country reforms and FDI inflows: How much difference do they make?

~ } Pergamon Worm Development Vol. 26, No. 7, pp. 1299-1314, 1998

© 1998 Elsevier Science Ltd All rights reserved. Printed in Great Britain

0305-750X/98 $19.00+0.00 PII: SO305-750X(98)O0049-7

Host Country Reforms and FDI Inflows: How Much Difference do they Make?

VICTOR M. GASTANAGA, JEFFREY B. NUGENT and BISTRA PASHAMOVA

University of Southern California, Los Angeles, U.S.A. Summary. - - This study examines the effects of various policies on foreign direct investment (FDI) flows from the perspective of the "eclectic theory" of international investment, and hence the advantages of foreign ownership, host country location, and internationalization. Host country policies can influence FDI flows primarily through their influence on the advantages of location in the host country. Pooled cross-section and time-series data for 49 less-developed countries (LDCs) over 1970-95 are used to examine the effects of several different types of policy/institutional variables, including corporate tax rates, tariff rates, the degree of openness to international capital flows, exchange rate distortions, contract enforcement, nationalization risk, bureaucratic delay and corruption. A multivariate analysis of the effects on FDI flows of each type of policy, with and without controls for other relevant determinants is conducted. Two different sources of FDI data are employed as well as two different measures of capital controls. Although there are methodological problems in estimation, and especially collinearity problems, and relatively severe availability constraints for some of the relevant measures, the results demonstrate the relevance and importance for FDI flows of many of the policy/institutional variables under study. They also indicate that the results from panel methods yield results that differ rather significantly from those obtained from pure cross-section analysis. Because of complementarity between the various elements of a healthy investment climate, the effects of some policy reforms estimated from models which do not control for changes in other policies may overstate the effectiveness of individual policy reforms. © 1998 Elsevier Science Ltd. All rights reserved

Key words - - foreign direct investment, corporate tax rates, capital controls, growth, exchange rate distortions, corruption, developing countries, contract enforceability

1. I N T R O D U C T I O N

Over the last 15 years, there has been an enormous change in host country attitudes to, and policies toward, private foreign investment in general and foreign direct investment (FDI) in particular. This has been accompanied, and in some cases even motivated, by studies attempting to demonstrate the net benefits over costs of these flows. Not only does FDI add to gross capital formation and the balance of payments without the risk associated with additional loan repayments, but also it is often said to increase competition and give rise to positive techno- logical externalities and spillovers, thereby raising dynamic efficiency. Any of these latter effects can justify efforts to encourage such flows (or at least to discriminate against such flows less than in the past). What has not yet been examined sufficiently is the effectiveness of the institutional reforms in investment regulations

and of other relevant policy changes in host less- developed countries (LDCs) in encouraging FDI flows. Frequently, various grounds for pessimism about the responsiveness of FDI to such reforms have been expressed) For example, with respect to tax rate reductions, it is often stated that taxes abroad (i.e., in host countries) should not matter because the tax laws of many developed countries (DCs) provide deductions from taxes in the home country for taxes paid by subsidiaries abroad. But, is such pessimism justified?

The purpose of this paper is to identify the relative importance of host country reforms on foreign and domestic investment after controlling for some other relevant determinants of observed changes in FDI flows across a sample of LDCs. Do the reforms increase FDI inflows appreciably or do they merely allow existing FDI to earn higher profits without increasing the level of FDI flows beyond the level which would occur in the absence of such reforms?

1299

1300 WORLD DEVELOPMENT

The vast majority of existing studies on the influence of policy-related variables on FDI flows have consisted of international cross-section studies. Such studies have been used to show the effects of such policy-related variables as tax rates, tariff rates, corruption and various other institutional variables on FDI. 2 Yet, despite attempts to separate out other influences, as with all cross-section studies, the results may well reflect other non-measured influences which vary across countries but not over time. For this reason, the results of such studies may not apply to the more relevant policy reform context of changes over time. While, in principle, the bias in the estimates of such effects could be in either direction, in practice it would seem highly likely that countries with low tax rates or other favorable policy variables also have a number of other favorable but unmeasured characteristics, implying that the estimates of such effects are likely to be biased upward. For this reason, it would seem important at least to supplement the cross-section studies with time-series estimates.

The main barrier to undertaking time-series studies is, of course, the absence of time-series of sufficient length and variation over time on the relevant policy variables to allow the effects of such changes on FDI to be accurately estimated. For example, in many countries corporate tax rates, tariff rates, regulations on capital flows, and various other institutional factors remain unchanged for decades for reasons of international treaty, political equilibrium, policy gridlock and other factors. Further difficulties in time-series estimation arise from the relatively recent origin of high quality and comparable data on FDI and from the need to introduce lags into the specification since the effects on FDI may not be realized immediately. Another problem with pure time-series analysis of FDI is that the variations over time may be rather volatile, reflecting many idiosyncratic influences particular to an individual country. As a result, existing time-series analyses of policy variables on FDI flows have largely been limited to a few developed countries)

In order to mitigate the problems arising from either pure cross-section or pure time-series analyses, in this study we use panel data to estimate the policy influences. Estimates with and without country-specific fixed effects are obtained. The panel data with a fixed-effects approach allows us to distinguish more system- atically between the effects of policy changes and other less variable elements of the investment climate on FDI over time as well as across countries. While some panel studies of policy

impacts on FDI do exist, once again they are largely confined to data for the United States across states and over time. 4

The effects of the following indicators of policy reform are examined: (a) corporate tax rates, (b) indicators of openness of capital flows, (c) the degree of distortion in exchange rates, (d) tariff rates, (e) contract enforcement, (f) nationalization risk, and (g) bureaucratic efficiency (or reduced corruption). The effects of such policies are analyzed in a multivariate regression analysis with panel methods.

The remainder of the paper is organized as follows: section 2 contains a brief review of the literature. Section 3 identifies the data sources and methods used in the study. Section 4 presents the results of multivariate analyses of the relation between the policy variables and FDI inflows. Our conclusions are given in section 5.

2. THEORY AND LITERATURE REVIEW

Although several different frameworks (neoclassical, monetary, transaction costs, portfolio theory) have evolved for analyzing the determinants and effects of FDI, 5 an exception- ally flexible and increasingly popular one is the "eclectic theory" of Dunning (1981a) and Dunning (1981b). According to this theory, FDI is determined by three sets of advantages which direct investment should have over the other institutional mechanisms available to the firm for satisfying the needs of its customers at home and abroad. The first such advantage is its ownership advantage in the host country, e.g., the advantage which the firm has over its rivals in terms of its brand name, patent or knowledge of technology or marketing. The second advantage is that of location, i.e., why it is important for the firm to operate in the host country (instead of in its home country or somewhere else). This advantage may derive from the host country's comparative advantage or its transaction cost advantage, including the absence of a tariff on the product if it is produced in the host country. Third is the internationalization advantage, i.e., why a "bundled" FDI approach is preferred to "un-bundled" product licensing, capital lending, or technical assistance.

While the first of these advantages depends rather strictly on the characteristics of the firm and its business, host country policies and institutions can play a prominent role in the second and third advantages. For example, with respect to the location advantage, a high tariff in the host country may contribute very substantially to

HOST COUNTRY REFORMS AND FDI 1301

the host country's location advantage for an import substituting industry. 6 On the other hand, a foreign firm interested in a suitable location for serving markets elsewhere in the world, might well be discouraged by high tariffs in the country, especially if they would apply to imported inputs. Other advantages may derive from host country efforts to develop its communications network, transport facilities or other forms of infrastructure. Depreciation of the host country's currency relative to that of the home country would be another factor raising the location advantage of the host country. The host country's corporate or other tax rates relative to those at home or in other countries abroad would be another factor affecting that locational advantage. The clarity of the country's laws, the extent and honesty of law enforcement, the efficiency of the bureaucracy and the absence of corruption would all be other transaction cost-reducing factors which could make host country location desirable. Some of these latter institutional characteristics of the host country can also affect the third or internationalization advantage. For example, for a country with a poorly developed or severely regulated licensing system and only very primi- tive legal and other mechanisms for selling intel- lectual and other property, the un-bundled FDI approach might be preferred over licensing and other institutional means at the foreign firm's disposal.

The fact that the theory is compatible with any theory of comparative advantage, be it of the Heckscher-Ohlin, classical or product cycle variants, as well as with any transaction cost explanation, makes it unusually flexible. Another advantage is that it can be applied at either the micro or macro levels. At the micro level, it allows one to relate individual firm, industry, and home and host country characteristics to one another. At the macro level, the eclectic theory can be used to explain the relative importance of various FDI-sending and/or receiving countries.

As mentioned above, research on the effects of policy variables on FDI, especially with respect to LDCs, is rather limited. While there has been much research on the general determinants of FDI in LDCs, 7 with surveys by Agarwal (1980) and Schneider and Frey (1985), 8 much of this literature focuses on very general factors like comparative labor costs, country size, the nature of the exchange rate regime and political factors including political instability (Green, 1972). Many of these studies include primarily DCs in the sample and most of those which include substantial numbers of LDCs, including the relatively more comprehensive one by Schneider

and Frey (1985), are estimated with pure cross- section (for each of three years in the late 1970s), giving rise to biases resulting from the failure to account for other non-measured factors which might be related to the economic and country risk factors.

Several more recent empirical studies of FDI determinants mention the potential importance of policy-related variables such as tax rates, foreign investment incentives and openness in the determination of FDI. Yet, in their empirical analysis, they fail to analyze them. 9 For example, Tsai (1994) notes the importance of qualitative factors such as political stability and incentive policies, but does not include them in the empirical analysis on the grounds that such variables are difficult to define and quantify. His study of FDI determinants, although using a more sophisticated simultaneous equation approach, 1° includes only very conventional economic variables such as market size and growth factors (measured by GDP per capita and GDP growth per capita), the per capita trade balance and the hourly wage rate in manufacturing. Although data for two different periods are used in this study, the model is estimated as a pure cross-section for each of the two periods, 1975-78 and 1983-86. Market size and growth have positive effects on FDI flows, although in the latter case only in the second period.

A distinctive approach is that of Lucas (1993) on the determinants of FDI flows in seven East and Southeast Asian economies. He develops an innovative theoretical model based on the derived demand for foreign capital of a profit- maximizing, multiple product monopolist. Two versions of the model are employed, a basic form containing only relative price and other endow- ment type variables, and an extended version which includes also location, market size, expectations formation, political risk and the regula- tory framework. The basic model is estimated in logarithmic and linear form separately for each country with time-series data for 1960-87. The results show that, for five of the seven countries studied, FDI flows have been responsive to the rental equivalent cost of capital and product price H and to dummy variables representing different regime types. The extended model (including the relative magnitudes of international reserves and strikes as proxies for political risk) is estimated by pooling the data for several countries of the sample, and allowing for country differences in both shift parameters and slope coefficients ~2 and once again the dummy variables for specific important events. For three of the countries, the regime change dummy


variables have little effect. In the other four countries, however, the dummies have significant effects, demonstrating that FDI rises with "good" political events and falls with "bad" events] 3 Clearly, even this otherwise useful and innovative study stops short of an evaluation of the effects of policies on FDI flows.

Riedel (1995) applies the same type of eclectic model employed in this study to intraregional trade and FDI among many of the same Asian countries used in the Lucas study. Yet, while trade patterns are analyzed in considerable detail, accounting for sectoral differences, relative prices, factor endowments and other considerations, the FDI portion of the study merely shows that FDI is closely related to trade patterns. For a single country (Colombia), Torrisi (1985) finds the effect of a dummy variable for participation in the Andean Common Market (perhaps the only integration scheme to attempt to enforce a rather strict common set of regulations on foreign investment) to be negative but not significant.

Among the more recent studies which have attempted to incorporate institutional factors such as host country risk and corruption in the determination of FDI are Wheeler and Mody (1992), Hines (1995) and Wei (1997). Wheeler and Mody use a composite measure of risk based on 13 indicators, including corruption, but do not find the hypothesized negative effect of risk on FDI. ~4 Hines uses only a corruption index, along with GDP growth and other factors, to show that, after 1977, FDI from the United States grew more rapidly in less corrupt countries than in more corrupt ones. Other policy-related factors are ignored. Wei (1997) examines two-year FDI flows from 14 OECD countries to 45 host countries and focuses on the effects of corruption after controlling for the marginal tax rate on foreign corporations, GDP per capita, population, the wage rate, distance between source and host countries, linguistic ties and political instability. For each of several specifications of the model, two different measures of corruption and the marginal tax rate are shown to have significant negative effects on FDI. Once again, Wei's estimates are based on pure cross- section data (and thus are subject to the same limitations mentioned in the Introduction).

3. DATA AND METHODS

Since the purpose of this paper is to trace the effects (if any) on FDI inflows of policy reforms taken by recipient LDCs in recent years, the study focuses on these, ignoring many commonly

analyzed economic variables such as wage rates. Nevertheless, although a wide variety of factors could be considered as reflecting trade and other policies, the focus of this study is on those more quantifiable elements which are also more changeable over time. Among these are exchange rate distortions, measured by the "black market premium" (BMP), the degree of openness to capital flows in general (DGOPEN), and to FDI inflows in particular (INFDI), corporate tax rates (TAX), tariff rates (TARIFF), nationalization risk (NATRISK), contract enforcement (CONTRACT) and bureaucratic delay (BURDELAY). The size of country effect is dealt with by defining the dependent variable in terms of FDI per units of GDP. Since FDI flows may also be induced by rapid growth of real GDP in the host country, both current, future (reflecting perfect rational expectations) 15 and lagged growth rates in real GDP (GROWTH, GROWTH+I and GROWTH_a, GROWTH_z, respectively) are also considered as determinants of FDI.

The data sources utilized for this study are as follows:

Two different sources of data on FDI are utilized: (a) for aggregate FDI inflows in millions of US dollars 1970-95: the International Monetary Fund's (IMF) Balance of Payments Statistics Yearbook and (b) FDI outflows from the United States in millions of US dollars (which permits a breakdown by sector of destination) for 1983-94 from the US Department of Commerce, Bureau of Economic Analysis. As mentioned above, in both cases the variable to be explained in the analysis is the F D I - G D P ratio. Time-series data on GDP (at current prices in millions of US dollars) are taken from United Nations MEDS (and updated with data from the CD-ROM version of the IMF's Inter- national Financial Statistics). In our subsequent analysis, the ratio of aggregate FDI to GDP from the IMF's Balance of Payments Yearbook and International Financial Statistics source is labeled "FDI" and that of US aggregate FDI to GDP "USFDI," and that in the manufacturing sector alone "USMAN."

Marginal corporate tax rates (TAX) by year and country are taken from various publications of Price Waterhouse 16 for various years over 1969-95. In cases in which different tax rates apply to wholly-owned subsidiaries of foreign corporations than to other enterprises with foreign capital in them, a weighted average of the two marginal tax rates is used. Time-series data on the average tariff rates (TARIFF) are calculated from time-series data on import duties


from the IMF's Government Finance Statistics Yearbooks and on the value of imports (both in current prices and local currency) from the IMF's International Financial Statistics.17

Time-series data on exchange rate distortions as proxied by the "black market premium" (BMP) are taken from the data files of the World Bank, 1960-92 and updated with information from the Worm Currency Yearbooks. The index of the absence of corruption (CORRUPT) is taken from Mauro (1995). Time-series data on indicators of contract enforcement (CONTRACT), nationalization risk (NATRISK) and bureaucratic delay (BURDELAY), developed originally by Business Environmental Risk Intelligence (BERI), are obtained from World Bank sources.

Time-series indicators of the degree of openness to international capital flows, in general, (DGOPEN) and to inward FDI, in particular, (INFDI) are constructed from the various issues of Annual Report on Exchange Arrangements and Restrictions published by the International Monetary Fund. The latter index (INFDI) is constructed as follows: "0" if the restrictions are high, "1" if they are moderate, and "2" if they are low or non-existent. The former more general index (DGOPEN) is the sum of INFDI and similarly constructed indexes for outward FDI, inward portfolio, outward portfolio flows and the right of ownership of a bank account by non-residents. Since each component of the general index of capital openness ranged from "0" to "2," in principle, the aggregate index could vary from "0" to "10."

Finally, for the control variables, the data on real GDP from which the rates of growth of real GDP (GROWTH) can be calculated are taken from the IMF's International Financial Statistics CD-ROM while that on the relative price of oil (OILPRICE) is taken from the United Nations MEDS and updated with data from International Financial Statistics.

Descriptive statistics, the countries included in the sample and data sources for all variables used in the analysis are presented in Table 1. Note first that, while the data on FDI flows are gross flows in the sense that they do not involve FDI flows of host country enterprises into another country, negative gross flows occur when foreign firms sell off some of their assets in the host country. Second, note that the country coverage is more limited for USFDI, USMAN, TARIFF, CORRUPT, CONTRACT, NATRISK and BURDELAY than for other variables. Third, note that over the whole period and across countries the mean corporate tax rate is about 35%, the effective tariff rate a little less than

10%, and the indicators of openness to capital flows (DGOPEN and INFDI), and investor- friendly institutions (CONTRACT, NATRISK, and BURDELAY) all low relative to their maxima.

4. MULTIVARIATE ANALYSIS OF FDI INFLOWS ACROSS COUNTRIES AND OVER

TIME

In this section we use panel data for up to 49 LDCs on FDI inflows relative to GDP, each of the aforementioned policy variables and GDP growth rates, current, future and lagged, and the relative price of oil to assess the effects of the policy variables on FDI flows after controlling for growth rates and, for oil exporting countries, the relative price of oil. While primary use is made of the aggregate FDI to GDP ratios as the dependent variable (FDI), the US FDI flows to LDCs in all sectors (USFDI) and in manufacturing alone (USMAN) are also used.

As noted above, while several previous studies have examined the effects of some (but not all of these variables), they have largely been estimated with pure cross-section data, i.e., based on comparisons across countries rather than over time. Our analysis begins by replicating the appli- cation of the cross-section methods to estimating the effects of the policy variables under study here. Table 2 reports the results obtained from a cross- section of the 23 LDCs in the sample for which all the relevant policy variables are available. All variables are the averages for the country over 1970-95. As indicated, the dependent variable is the average FDI-GDP ratio based on aggregate FDI data from the IMF and different sets of explanatory variables (indicated in the rows of the table) are used. Due to relatively high correlation among some pairs of the explanatory variables, not all explanatory variables can be used at the same time. For example, especially high correlations exist among CONTRACT, BURDELAY and CORRUPT, implying that only one of these variables can be included in the specification at a time without introducing serious multi- collinearityY

Several aspects of these results deserve comment. First, the marginal corporate tax rate seems to have a highly nonlinear effect on FDI (the coefficient for the linear term TAX being positive but that of the TAX 2 being negative). According to the results of column (1), the net effect of the tax would become negative only after the tax rate exceeds 40% or in excess of 30% for the estimates in column (2). Second, since DGOPEN is based in part on INFDI, these


variables are closely related to one another. Since D G O P E N is slightly less closely related with the other explanatory variables, when it is used instead o f l N F D I , its effects are more highly significant on FDI. Both measures, however, are shown to have positive effects. Third, as suggested above, the effect o f G R O W T H is also positive and significant, indicating its relevance as a control variable in assessing the effects o f the policy variables. Fourth, the fact that TARIFF has a positive ( though not highly significant) effect on FDI would seem to

suggest that, for the sample as a whole, FDI flows to LDCs were more strongly motivated by tariff- j u m p i n g than by their export potential. Finally, all three measures of investment-fr iendly institutions, i.e., C O N T R A C T , B U R D E L A Y and C O R R U P T , all have positive and generally significant effects on FDI. The values of R 2 are reasonably high for cross-section analysis, and the relation is highly significant. By and large, with the exception o f those wi th respect to the measures I N F D I and D G O P E N (specially constructed in this study),

Table 1. Descriptive statistics

Variable Time period Countries Mean Std Dev. Min Max

FDI 1970-95 49 1.396 2.386 - 16.61l 15.870 USFDI 1983-94 33 1.137 6.039 - 23.618 72.174 USMAN 1983-94 33 0.201 0.771 - 3.772 5.518 TAX 1970-95 49 35.960 10.996 0.000 64.000 INFDI 1970-95 49 0.502 0.896 0.000 2.000 DGOPEN 1970-95 49 2.472 2.825 0.000 10.000 CORRUPT 1980-83 37 5.868 2.366 0.000 10.000 BMP 1970-95 49 30.711 197.435 -28.210 6714.000 GROWTH 1970-95 49 4.289 5.594 - 19.004 32.356 TARIFF 1972-95 44 0.091 0.054 0.000 0.330 CONTRACT 1972-95 22 2.052 0.428 0.250 3.300 BURDELAY 1972-95 22 1.683 0.485 0.380 3.230 NATRISK 1972-95 22 2.243 0.494 0.080 3.500 OILPRICE 1970-95 Time-series* 108.563 54.291 35.251 205.210

*OILPRICE is set equal to zero for countries that are not oil exporters. See the text for definitions. Sample Countries: Argentina, Bahrain, Barbados, Bolivia, Botswana, Brazil, Chile, China, Colombia, Costa Rica, Cote d'Ivoire, Cyprus, Dominican Republic, Ecuador, Egypt, El Salvador, Fiji, Gabon, Guatemala, Haiti, Honduras, India, Jamaica, Jordan, Kenya, (South) Korea, Malaysia, Mauritius, Mexico, Morocco, Nigeria, Oman, Pakistan, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Saudi Arabia, Senegal, Singapore, South Africa, Thailand, Trinidad and Tobago, Turkey, Uruguay, Venezuela, Zambia and Zimbabwe Definition of Variables FDI: Total Inward FDI flows from the 1MF's Balance of Payments Statistics (BOP) Yearbook and International Financial Statistics (IFS) Yearbook as a percentage of GDP. USFDI: Total US FDI from the US Department of Commerce Bureau of Economic Analysis (BEA) as a percentage of GDP. USMAN: Manufacturing US FDI from the US Department of Commerce Bureau of Economic Analysis (BEA) as a percentage of GDP. TAX: Corporate tax rate from Price Waterhouse's country books. INFDI: Index of the degree of openness to inward FDI constructed from the IMF's Annual Report on Exchange Rate Arrangements and Restrictions, which is "0" if restrictions are high, "1" if moderate, "2" if low or non-existent. DGOPEN: Index of the degree of general openness to capital flows constructed from the IMF's Annual Report on Exchange Rate Arrangements and Restrictions, which ranges from "0" if restrictions are high to "10" if low or non-existent. CORRUPT: Index of absence of corruption from Mauro (1995), which ranges from "0" (most corrupt) to "10" (least corrupt). BMP: Black Market Premium from the World Bank. GROWTH: Rate of growth of real GDP, calculated using real GDP from the UN's MEDS and from the IMF's IFS. TARIFF: Tariff revenue as a fraction of the value of imports, in domestic currency. Tariff revenue is from the IMF's Government Financial Statistics (GFS) Yearbook, imports from the IMF's IFS. CONTRACT: Contract Enforcement index from Business Environmental Risk Intelligence (BER1), which ranges from "0" if enforcement is poor to "4" if good. BURDELAY: Bureaucratic Delay index from BER1, which ranges from "0" if delay is high to "4" if low. NATRISK: Nationalization Risk index from BERI, which ranges from "0" if risk is high to "4" if low. OILPRICE: Dummy variable for oil exporter multiplied by an index of real oil prices.


and the relatively significant nonlinear effects o f TAX, these results are rather consistent with those obtained in other cross-section studies.

Tables 3 and 4 present the results using the fullest possible panel for the case in which the F D I - G D P ratios are calculated from aggregate IMF FDI data. Table 3 contains the results without fixed effects and Table 4 the corresponding results with fixed effects. Tables 5 -8 contain the corresponding results for the case in which the F D I - G D P ratios are those calculated on the basis of US FDI data (both aggregate, USFDI and manufacturing only, U SMA N) . Since preliminary estimates without lagged dependent variables indicated first-order autocorrelation of the resid- uals, lagged de9Pendent variables are included in all specifications.

The pooled panel results for seven different specifications of the RHS variables when aggregate F D I - G D P ratios calculated from the IMF data are used are given in Table 3. Each specification has been estimated with the largest possible number of observations. Given the similarity of their definitions, I N F D I and D G O P E N have been used alternatively (never together). Since as in the cross-section estimates

of Table 2, the coefficient of B M P is never statistically significant at even the 10% level, B M P is included in only one of the specifications. Other differences in specification from one column to another pertain to the inclusion or not of GROWTH,_1 (in addition to GROWTH,+1) and the inclusion or not of C O N T R A C T and C O R R U P T which are fairly closely related. The column entries are arranged from left to right according to sample size. The entries of the first two columns are obtained from the full sample of 49 countries. The next two columns require data on TARIFF which are available for only 44 countries in the sample, and finally the entries in columns (5)-(7) require data on C O N T R A C T which like the closely related N A T R I S K and B U R D E L A Y are available for only 22 sample countries.

We concentrate on the policy effects. Once again, the coefficients of the marginal tax rate variables (TAX and TAX 2) indicate that the effect of tax rates on FDI are highly nonlinear. While in this case the coefficients are not statistically significant, due to strong collinearity between these two variables (r = 0.96), the standard errors are undoubtedly overestimated. Note also that

Table 2. Cross-section OLS estimation: IFS FDI data a

Independent variable (1) (2) (3) (4)

Intercept - 21.177" - 16.403" - 13.079" * * - 14.717" * * ( - 2.619) ( - 2.073) ( - 1.778) ( - 1.957)

TAX 0.765" 0.345 0.350 0.5148 (2.285) (1.577) (1.035) (1.560)

TAX 2 - 0.009 * - 0.006 - 0.004 - 0.006 ( - 2.320) ( - 1.601) ( - 1.026) ( - 1.609)

INFDI 0.769 (1.296)

DGOPEN 0.182" 0.467* (2.475) (2.281)

GROWTH 0.608* 0.328"** 0.461" 0.718"* (2.836) (1.715) (2.158) (3.771)

CONTRA CT 2.254* (2.232)

TARIFF 11.001 9.775 (1.589) (1.399)

BURDELA Y 1.768" * * (1.818)

CORRUPT 0.318" 0.337* (2.095) (2.425)

R z 0.6530 0.6335 0.5470 0.4984 Adj R 2 0.5229 0.4961 0.4138 0.3870 F-test: p-value 0.0045 0.0067 0.0125 0.0110 Obs 23 23 23 23

aT-Statistics reported in parentheses. *Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.


the magnitudes of the respective coefficients are reasonably robust to various specifications. The specific values obtained for the tax rate parameters show that the net effects become negative only at tax rates above the range 20%, considerably lower than was indicated by the cross- section results of Table 2. Once again, the coefficients of INFDI or DGOPEN are positive and fairly significant in all specifications. Although the use in all columns except (5) of the variable CORRUPT (representing the freedom from corruption) reveals a positive influence, these coefficients are generally not statistically significant. 2° Notably, in these pooled time-series cross-section results, the effects of TARIFF become negative, though generally not significant. Even in this context, CONTRACT and CORRUPT are quite highly correlated, and to a lesser extent so too is TARIFF. When CONTRACT but not CORRUPT is included in the specification as in column (5), its coefficient is positive and significant. Likewise, with

CORRUPT but not INFDI, TARIFF and CONTRACT as in column (6), the effect of CORRUPT is positive and significant. When all of them are included at the same time, the significance of all their effects is reduced. The growth variables, whether future, or lagged or both (as measures of the expected rate of growth) have positive and generally significant effects on the F D I - G D P ratios. Finally, as expected, the effect of the lagged dependent variable in each specification is positive and highly significant.

Table 4 presents the corresponding results for the (IMF-based) aggregate F D I - G D P ratios when fixed effects are included. Once again, seven different specifications are estimated, beginning in columns (1)-(3) with equations estimated for the full sample of 49 LDCs. With TARIFF added to the specification in columns (4) and (5), the sample is reduced to 44 countries and finally in columns (6) and (7), because B U R D E L A Y and NATRISK are included, these estimating equations can only be

Table 3. Pooled OLS estimation: IFS FDI data

Independent variable (1) (2) (3) (4) (5) (6) (7)

Intercept -0.711 -0.704 -0.450 -0.868 -0.814 -0.777 -1.316 (-0.847) (-0.837) (-0.513) (-0.981) (-0.794) (-0.787) (-1.247)

FDI_~ 0.744** 0.743** 0.755** 0.739** 0.831"* 0.841"* 0.8243** (28.358) (27.916) (26.889) (25.792) (28.686) (32 .648) (27.825)

TAX 0.039 0.039 0.023 0.035 0.014 0.031 0.032 (0.880) ( 0 . 8 7 1 ) (0.513) (0.782) (0.287) (0.637) (0.637)

TAX 2 -0.001 -0.001 -0.000 -0.000 -0.000 -0.000 -0.000 (-0.978) (-0.970) (-0.480) (-0.667) (-0.278) (-0.782) (-0.654)

BMP -0.000 (-0.314)

DGOPEN 0.070 * * (3.126)

INFDI 0.163" 0.162" 0.152"** 0.161" (1.931) ( 1 . 9 0 6 ) (1.769) (1.890)

CORRUPT 0.036 0.037 0.045 0.043 0.054* 0.030 (1.300) ( 1 . 3 2 0 ) (1.581) (1.541) (2.299) (1.007)

TARIFF -2.090*** -1.637 - 1.628 -1.069 (-1.878) (-1.466) (-1.433) (-0.882)

CONTRACT 0.350* 0.267*** (2.338) (1.680)

GROWTH,+1 0.025* 0.025* 0.030* 0.030* 0.033** 0.020*** (2.113) ( 2 . 0 9 5 ) (2.532) (2.508) (2.648) (1.603)

GROWTH, ~ 0.030* 0.029* (2.454) (2.185)

R 2 0.5835 0.5765 0.6303 0.6346 0.7979 0.7570 AdjR 2 0.5798 0.5721 0.6258 0.6301 0.7944 0.7540 F-test: p-value 0.0001 0.0001 0.0001 0.0001 0.000l 0.0001 Obs. 696 694 579 579 418 482

0.146"** (1.637)

0.031" (2.443)

0.8016 0.7972 0.0001 414

aT-Statistics reported in parentheses. *Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.


estimated with a more limited sample of 22 countries. In order to distinguish between the effect of the reduction in sample size and the inclusion of the additional variable, column (6) presents the results with the same specification as (5). The F-test results given at the bot tom of the table show that, collectively, the fixed effects are highly significant, suggesting that unmeasured cross country differences play a large role in explaining the observed intercountry and inter temporal variation in FDI flows. As in Table 3, the effects of growth expectations (proxied by GROWTHt+I and G R O W T H t _ O are positive and significant as are those of the lagged dependent variables (FDI_I) .

Again, our focus is on the effects of the policy variables. Once these fixed effects (country dummy variables) are introduced, even the linear tax rate terms are negative and generally significant. The magnitude of the coefficients indicates that a 10% increase in the marginal corporate tax rate would lower the F D I - G D P ratio by about 0.2. This is a much more substantial negative effect than obtained from pure cross- section results. While the estimated effects of the B M P are once again statistically insignificant,

both the direction and magnitude of the effects of T A R I F F are rather sensitive to the specification. In particular, the effects are as before positive (but not significant) in the results for the sample of 44 countries reported in columns (4) and (5). Yet once other prominent elements of the investment climate, such as B U R D E L A Y and N A T R I S K , are accounted for as in columns (6) and (7), the coefficients for T A R I F F become negative and fairly significant.

Another important result of these specifications with the 22-country sample is that the effect of N A T R I S K is positive and significant, clearly showing that the absence of nationalization risk has a significant positive influence on the F D I - G D P ratios. Finally, note that from the more reliable estimates of the effects of O I L P R I C E in columns (1)-(3) of the table, it would seem that the effect of O I L P R I C E on FDI is negative and significant. While a rise in oil price might be expected to raise the growth rate of G D P and thereby FDI, once one controls for the growth rate (as is done in these tables), a rise in the real price of oil seems to make oil-exporting countries less dependent on, and less interested in, foreign investment. It also has the effect

Table 4. Fixed effects estimation: IFS FDI data

Independent (1) (2) (3) (4) (5) (6) (7) (8) variable

FDI_~ 0.400** 0.4218"* 0.403** 0.316"* 0.316"* 0.425* 0.404** 0.416"* (13.05) h (13.57) (13.27) (9.23) (9.21) (9.74) (8.62) (9.26)

TAX -0.021"** -0.028* -0.021"** -0.014 -0.014 -0.019"** -0.021"** -0.028*** (-51.81) ( - 2.24) ( - 1.86) ( - 1.24) ( - 1.20) ( - 1.82) ( - 1.87) ( - 1.82)

B M P -- 0.000 - 0.000 (--0.05) (-0.12)

DGOPEN 0.077* 0.078* 0.059*** 0.063*** 0.041 0.037 0.046 (2.39) (2.43) (1.73) (1.83) (1.27) (1.10) (1.40)

INFDI 0.209* (2.01)

TARIFF 1.749 1.773 -2.666 -3.425*** -3.313"** (0.90) (0.91) ( - 1.44) ( - 1.71) ( - 1.66)

BURDELAY -0.503* -0.520* (-2.01) (-2.09)

NATRISK 0.511 * 0.500* (2.26) (2.24)

OILPRICE -0.005* -0.006* -0.005* 0.003 -0.004 -0.003 (-2.11) (-2.16) (-2.10) (0.94) (-1.51) (-1.23)

GROWTH,÷~ 0.041"* 0.041"* 0.041"* 0.022* 0.024* 0.026* 0.027* 0.028* (3.86) (3.88) (3.87) (1.93) (2.07) (2.20) (2.20) (2.36)

GROWTH,_~ 0.033** 0.034** (2.62) (2.36)

R 2 0.5555 0.5663 0.5600 0.6538 0.6542 0.8462 0.8494 0.8524 F-test: p-value a 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 No. cross-sections 49 49 49 44 44 22 22 22 Time-series length 22 22 22 19 19 19 19 19

aReported is the F-test for fixed effects. bt-Statistics in parentheses. *Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.


of tilting investment activities toward the oil sector, from which foreign investors are often excluded.

Tables 5 and 6 present the corresponding results for the pooled panel without fixed effects for the F D I - G D P ratios for the case in which the FDI flows are from the US alone. Table 5 reports the results for the F D I - G D P ratios for US FDI flows destined for the manufacturing sector alone while Table 6 reports for those destined for all sectors. By comparing the sample sizes for these tables with the corresponding ones for Table 3, it can be seen that the use of the US data greatly limits the number of observations. For this reason, we decided that the introduction of TARIFF, CONTRACT, NATRISK, and B U R D E L A Y would be unwar- ranted since this would require substantial further reductions in sample size. Although the tax rate variables are somewhat more significant in Table 6 than in Table 5 (indicating that FDI flows may be more sensitive to tax rates in other

sectors :1 than in manufacturing), they are generally not significantly different than zero. On the other hand, the sensitivity of the US-sourced F D I - G D P ratios to INFDI or DGOPEN is at least as great in these tables as for those from all sources in Table 3. In this case, however, the F D I - G D P ratios appear to be more sensitive to these variations in DGOPEN when FDI to other sectors is included (as in Table 6) than for FDI to manufacturing alone (Table 5) but less sensitive to variations in INFDI. Since DGOPEN is a broader index than INFDI (which pertains exclusively to inward flows of FDI), this would seem to indicate that controls on capital flows other than inward FDI have greater effects in other sectors than in manufacturing. 22

Tables 7 and 8 present corresponding versions of the estimates of the models for F D I - G D P ratios for FDI coming only from the United States when fixed effects are allowed for. As was the case in Table 4 above (compared to Table 3), once fixed effects are allowed, the linear tax rate

Table 5. Pooled OLS estimation: BEA manufacturing FDI data


Intercept 0.016 -0.199 -0.197 -0.044 -0.276 - 1.267 -0.610 (0.060) a (-0.714) (-0.473) (-0.121) (-0.762) (0.940) (-2.156)

USMAN_I 0,110 0.081 0.078 0.107 0.079 0.096 0.098 (1.230) (0.897) (0.784) (1.191) (0.861) (1.061) (1.085)

TAX - 0.003 0.000 0.000 0.003 0.006 0.035 (-0.416) (0.057) (-0.010) (0.186) (0.324) (0.573)

TAX: 0.000 0.000 -0.001 (-0.374) (-0.328) (-0.713)

INFDI 0.362** 0.367** 0.182 0.210 (2.844) (2.805) (1.148) (1.382)

DGOPEN 0.097** 0.112"* 0.096** (3.352) (3.470) (3.244)

CORRUPT 0.094* 0.088*** (1.963) (1.890)

BMP 0.000 (0.447)

GROWTH.I -0.011 (-0.678)

GROWTH 0.025 * * * 0.026* * * 0.025 0.029 * * * 0.025 * ** 0.042* 0.039* (1.771) (1.877) (1.496) (1.899) (1.723) (2.467) (2.381)

GROWTH_ 1 0.001 (0.094)

GROWTH_z 0.017 0.020 0.020 0.017 0.019 0.027 0.024 (1.206) (1.422) (1.261) (1.164) (1.301) (1.586) (1.469)

DW stat 1.920 1.899 1.865 1.927 1.898 1.954 1.944 R e 0.154 0.174 0.207 0.159 0,175 0.193 0.185 Adj R 2 0.119 0.140 0.160 0.109 0.127 0.143 0.149 F-test: p-value 0.0010 0.0003 0.0005 0.0038 0.0015 0.0009 0.0002 No. obs 127 127 109 127 127 121 121

at-Statistics in parentheses. *Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.


terms (TAX) have negative and frequently (but not always) significant effects. Presumably for reasons given above, the magnitude of these effects is greater when FDI is to all sectors (Table 8) than when it is to manufacturing alone (Table 7). Yet because for many of the countries for which low tax rates have an important effect (again, banking and finance) have low tax rates throughout the period, it is more difficult to distinguish these effects from those of the unobserved country differences (via fixed effects) making for lower levels of significance of these coefficients. Because of the greatly reduced number of over t ime observations in the case of US data, it is even more difficult to distinguish the effects of the nearly fixed I N F D I and D G O P E N variables than it was when the I M F data were used.

With the exceptions noted, however, the results with FDI coming from the Uni ted States (Tables 5 and 6), are broadly consistent with those obtained from FDI coming from all

countries (Table 3) and those obtained from Tables 7 and 8 are consistent with those obtained from Table 4.

5. C O N C L U S I O N S

The above results are not without shortcomings. First, certain other types of policies such as sales and property taxes or other policy distortions such as the variance of tariff rates across commodities, have not been considered. Second, for several other relevant policy/institutional variables (especially CORRUPT) , the number of observations is very small, suggesting that the results reflect largely intercountry differences rather than over-t ime reforms, which were the major objective of the study. Third, for other policy/institutional measures, such as CON- TRACT, B U R D E L A Y and N A T R I S K , the neces- sary data are available for an unfortunately small sample of countries. Fourth, the measures of

Table 6. Pooled OLS estimation: BEA total FDI data


Intercept

USFDI- 1

TAX

TAX 2

INFD1

DGOPEN

CORRUPT

BMP

GROWTH+I

GROWTH

GROWTH _

GROWTH 2

DW stat R 2 Adj R 2 F-test: p-value No.obs

2.253 -1.321 1.042 (1.093) ~ (-0.660) (0.341)

-0.131 -0.318"* -0.347** (-1.432) (-3.436) (-3.490)

-0.031 0.019 -0.022 (-0.574) (0.369) (-0.293)

0.203 (0.215)

1.053"* 1.112"* (4.869) (4.610)

-0.004 (-0.545)

-0.273* --0.256* -0.430** - 2.383) ( - 2.444) ( - 3.245)

0.245* 0.152 0.220*** (2.079) (1.396) (1.649)

2.026 2.085 2.124 0.081 0.225 0.284 0.045 0.195 0.244 0.0538 0.0001 0.0001 133 133 115

0.339 -2.159 11.890 0.314 (0.121) (-0.821) (1.128) (0.152)

-0.145 -0.320** -0.154 -0.123 ( - 1.560) ( - 3.427) ( - 1.610) ( - 1.309)

0.094 0.075 -0.488 (0.653) (0.575) ( - 1.010)

- 0.002 - 0.001 0.004 ( - 0.933) ( - 0.469) (0.785)

0.046 -0.726 -0.168 (0.048) (-0.605) (-0.146)

1.038"* (4.716)

0.018 (0.176)

0.343 0.194 (0.962) (0.559)

0.074 (0.644)

-0.297* -0.259* -0.378** -0.357** ( - 2.446) ( - 2.423) ( - 2.819) ( - 2.695)

0.227*** 0.148 0.277* 0.313" (1.899) (1.339) (1.999) (2.332)

2.023 2.085 2.023 2.024 0.090 0.227 0.119 0.099 0.039 0.184 0.067 0.061 0.0990 0.0001 0.0318 0.0260

133 133 127 127

~t-Statistics in parentheses. *Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.


even the most available and relevant policy variables are far from ideal. For example, they reflect crude weighting procedures which could prove unrealistic and introduce biases of unknown direction on the estimates. Moreover, the tax rates might more ideally be measured in effective rate terms taking into consideration rules about depreciation rates, investment credits and the average cost of capital following the procedure suggested by Auerbach (1990) and/or tax rules in the country of origin as was done by Shah and Slemrod (1990) in their study of USFDI in Mexico. Fifth, even the longest available time-series contain at a maximum only 26 observations and the largest number of countries for which such series are available turns out to be 49. Sixth, the control variables included growth rates, and the real price of oil but not certain other variables found to have significant influences on FDI in other studies, such as political variables, exchange rates, wage rates, and social and political instability, or firm-level characteristics which are relevant to the eclectic model of FDI utilized above. 23 Finally, all the above relationships treat the variables on the right hand side as exogenous or predetermined. In the case of the growth variables, however,

growth could be thought of as determined jointly with FDI as has been allowed for in some of the studies reviewed above.

Nevertheless, despite these shortcomings, all of which could be overcome with rather straight- forward extensions of this study as more and better data become available, certain rather important conclusions emerge.

First, by way of confirming some results of previous studies, 24 there are three significant findings: (a) Despite some rhetoric to the contrary, exchange rate distortions in host countries do not seem to exert a significant detri- mental influence on FDI. (b) The expected rate of growth (as measured by actual growth rates, future, current or past) is a highly significant determinant of F D I - G D P ratios. Hence, to the extent that LDCs can raise their overall growth rates through a variety of efforts not directly associated with FDI, indirectly these policies would also affect FDI. (c) Although there is insufficient variation in the available measures of corruption over time, the pooled regression results confirm the findings of Wei (1997) and others that corruption has a deleterious effect on FDI, and indeed that a doubling in the corruption index has an effect on FDI inflows that is

Table 7. Fixed effects estimation: BEA manufacturing FDI data


USMAN_t -0.376** -0.375** -0.381"* -0.380** -0.285** -0.282** -0.289** (-4.02) b (-4.02) (-3.74) (-3.75) (-3.41) (-3.30) (-3.40)

TAX -0.049*** -0.047*** -0.053* -0.051"** -0.041"** 0.139 -0.042*** ( - 1.91) ( - 1.81) ( - 1.95) ( - 1.86) ( - 1.83) (0.78) ( - 1.89)

TAX 2 - 0.002 (-1.03)

INFDI 0.277 0.273 (0.75) (0.76)

DGOPEN O. 150 O. 146 (0.94) (0.94)

BMP 0.000 0.000 (0.41) (0.41)

GROWTH+I -0.021 -0.021 -0.013 -0.013 -0.017 -0.018 -0.017 ( - 1.49) ( - 1.52) ( - 0.66) (-0.69) ( - 1.25) ( - 1.34) ( - 1.30)

GROWTH 0.013 0.013 0.007 0.007 0.016 0.016 0.017 (0.99) (0.97) (0.39) (0.38) (1.23) (1.30) (1.35)

GROWTH_ 1 0.007 (0.52)

R 2 0.523 0.524 0.564 0.565 0.503 0.506 0.501 F-test: P-value a 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 No. cross-sections 23 23 20 20 27 27 27 Time-series length 6 6 6 6 6 6 6



approximately equal to a percent increase in the corporate tax rate. Yet, because CORRUPT tends to be correlated with other elements of the investment climate, once one includes these other elements into the analysis as we have done here, the effects attributed to CORRUPT are diminished.

Second, with respect to the effects of capital controls for which special measures have been developed here, even after allowing for fixed effects, positive and significant effects are detected in the case of aggregate FDI (as shown in Table 4). The fact that the effects of the overall index (DGOPEN) are frequently more highly significant than INFDI (which is one component of DGOPEN) indicates, that the influence of capital controls on inward FDI is by no means limited to the controls on inward FDI flows themselves. Hence, cosmetic types of capital account liberalization may be insufficient to exert strong influences over FDI. 25

Third, and perhaps the most interesting finding is that, once one controls for fixed effects, corporate tax rates exert a significantly negative and linear influence on FDI flows. The estimates of the absolute magnitude of the effect

on the F D I - G D P ratio of a 1% increase in the corporate tax rate vary from a low of -0 .025 in the case of the aggregate FDI (from the I M F source) to a high of - 0 . 0 5 in the case of USFDL From the pooled results, it can be seen that there is a nonlinear relationship between tax rates and FDI, the linear term being positive and the quadratic term negative. Using the estimates of columns (4) and (5) of Table 3 as the basis, it can be seen that the net effect becomes negative at corporate tax rates of some 25-35%, which are rather moderate for L D C standards. One should bear in mind that India and several sub-Saharan African countries still have corporate tax rates of over 60%. LDCs wanting to attract more FDI, therefore, may have to consider lowering their corporate tax rates to become more competit ive with other countries.

Fourth, estimates of the effects of tariff rates (TARIFF) on FDI are especially sensitive to the estimation method and specification. This is in part because tariff rates are correlated with other potential determinants of FDI. Al though the pure cross-section results indicate the effect of TARIFF on FDI to be positive, in a time-series context, and especially after one has controlled

Table 8. Fixed effects estimation: BEA total FD1 data


USFDI_I -0.590** -0.590** -0.604** -0.604** -0.614"* -0.613"* -0.611"* (-6.24) b (-6.24) (-5.93) (-5.93) (-8.24) (-8.22) (-8.23)

TAX -0.272 -0.265 -0.241 -0.232 0.539 0.595 -0.234 ( - 1.35) ( - 1.30) ( - 1.08) ( - 1.03) (0.39) (0.43) ( - 1.39)

TAX 2 -0.010 -0.011 (-0.54) (-0.60)

INFDI 0.225 0.283 (0.07) (0.09)

DGOPEN 0.309 0.346 (0.23) (0.25)

BMP 0.005 0.005 (0.52) (0.52)

GROWTH+I 0.057 0.056 0.191 0.190 0.052 0.073 0.077 (0.50) (0.49) (1.09) (1.08) (0.51) (0.72) (0.76)

GROWTH 0.105 (1.06)

GROWTH_I -0.252* -0.253* -0.356* -0.357* -0.197" -0.180"** -0.177"** ( - 2.21) ( - 2.22) ( - 2.39) ( - 2.40) ( - 2.03) ( - 1.89) ( - 1.86)

R 2 0.419 0.419 0.437 0.437 0.5261 0.522 0.521 F-test: P-value a 0.0001 0.0440 0.0010 0.1340 0.0001 0.0001 0.0001 No. cross-sections 23 23 20 20 27 27 27 Time-series length 6 6 6 6 6 6 6



for o ther re la ted influences, the net effect of TARIFF on F D I - G D P rat ios becomes negative. Whereas cross-sectionally and in the early par t of the period, tar i ff- jumping seemed to be an impor tan t motive for FDI, over t ime in individual countr ies t rade l iberal izat ion has become the more impor t an t motive for FDI.

Fifth, several of the different measures of inst i tut ional characterist ics re levant to FD I flows, such as CONTRACT, BURDELAY, NATRISK, are shown to have significant effects on FDI.

Finally, the substant ial differences be tween the cross-section results and those with pane l data wi thout and especially with fixed effects suggest tha t it may be misleading to base one ' s es t imates abou t the effects of policy variables on FDI exclusively on cross-section analysis. Un- measu red differences across countr ies seem to play a relatively large role in explaining varia- t ions in FDI flows across countr ies and over t ime even af ter control l ing for a variety of o ther variables.

N O T E S

l. See, e.g., Amirahmadi and Wu (1994).

2. Among such studies are those of Hines and Rice (1994), Slemrod (1990a), Hines and Hubbard (1990) and Wei (1997).

3. See for example, Hartman (1984), Slemrod (1990a) and Slemrod (1990b) for the United States, Boskin and Gale (1987) for a few developed countries (DCs).

4. See, e.g., Cummins and Hubbard (1994) and Hines (1997).

5. One macro-monetary approach is outlined in Fry (1995, chapter 12) and others in Froot (1992). An exploitation thesis is developed in Hyrner and Rowthorne (1970), the role of the Product Cycle in Vernon (1966), and a portfolio analysis in Rugman (1977).

6. In other words, "tariff-jumping" can be an important motive for FDI.

7. Research on FDI in the United States and other developed countries will not be reviewed here. Gener- ally, such studies have been based on theories of the firm, industrial organization and taxation. An overview of the different industrial organization approaches which focus on ownership-specific assets and location- specific advantages (Dunning, 1985; Wilson, 1993) and the "tax capitalization" approach of Hartman (Hartman, 1984; Hartman, 1985), Feldstein and Jun (1987), Boskin and Gale (1987) can be found in Cummins and Hubbard (1994).

8. Another useful survey, though one not confined or even concentrating primarily on LDCs, is that of Lizondo (1990).

9. Another example of a study which emphasizes the importance of investment incentives and political risk in the determination of FDI is Amirahrnadi and Wu (1994).

10. FDI flows and the rate of economic growth are jointly determined.

11. In the model the product price is measured by the export price index because of the export orientation of most FDI in these economies. The rental cost of capital is measured as r =pk*ir/(1-tx), where pk is the

fixed capital formation deflator in host country, ir the real interest rate and tx---corporate tax profit rate.

12. Homogeneity tests are performed to justify the choice of countries in each pool.

13. Another shortcoming is that the data employed do not cover the late 1980s at which time more countries experimented with policy reforms.

14. An interesting finding of Wheeler and Mody (1992) is that the importance of several of the determinants varies by level of development. Quality of host country infrastructure and the rate of economic growth are the most important for the lowest income countries, but tax rates and the existing stock of FDI more important for somewhat higher income countries.

15. Naturally, the use of future growth in these regressions could suggest reversed causality, i.e., FDI in year t cause growth in year t+ 1. Possible bias in the estimates of the growth terms on the FDI-GDP ratios is by no means of central importance to the study. What we want in such a control is the measure(s) of growth (expectations) which seems to be most closely related to FDI which, once controlled for, allows us to best isolate the effects of the policy variables on FDI.

16. Doing Business in... , Corporate Tax Rates in 60 Countries, Corporate Tax Rates in 80 Countries, Corporate Tax Rates Throughout the World.

17. Because of the importance of tax exemptions on imports in many countries for both the public sector and investors or exporters favored by incentives legis- tlation, these rates are typically considerably lower than the nominal tariff rates.

18. Other variables included in Table 1, namely NATRISK and BMP, are also so highly correlated with these and other variables that they are excluded altogether from the pure cross-section analysis.

19. Since the estimated coefficients of the policy-type variables (available upon request) obtained for comparable specifications but without the lagged dependent variable are virtually identical to those presented in these tables, it is clear that the use of lagged dependent variables has no influence on the main objective of the paper.


20. AS noted in Table 1, however, since CORRUPT is available for only three consecutive years, it is essen- tially a single point, greatly limiting its usefulness in the present time-series context.

21. With the importance of off-shore banking and finance in tax havens (like some Caribbean islands and Bahrain), it is obvious that banking and finance are sectors in which location is especially sensitive to tax rate differences.

22. Although of little consequence to the main theme of the paper, the effects of the growth and lagged dependent variables are also rather different between Tables 5 and 6. Those for Table 5 are more similar to those in Table 3.

23. Cummins and Hubbard (1994) use such an approach with micro-level data on US firms.

24. See especially Shah and Slemrod (1990) and Wheeler and Mody (1992).

25. Since one would expect exchange rate distortions to be associated with capital controls, the fact that the measures of freedom from capital controls INFDI and DGOPEN have significant positive effects on FDI whereas the effects of the exchange rate distortions (BMP) have insignificant and even sometimes positive influences on FDI poses something of a puzzle. This is especially puzzling in view of studies showing that FDI is positively associated with economic growth rates and BMP negatively related to growth. This puzzle is clearly deserving of further attention.

R E F E R E N C E S

Agarwal, Jamuna P. (1980) Determinants of foreign direct investment: A survey. Weltwirtschaftliches Arch& 116, 739-773.

Amirahmadi, Hooshang and Weiping, Wu (1994) Foreign direct investment in developing countries, Journal of Developing Areas 28, 167-190.

Auerbach, Alan, "The cost of capital and investment in developing countries," Tax Incentives Evaluation Research Project Discussion Paper 1 (Washington, DC: World Bank, 1990).

Boskin, Michael J., and William G. Gale, "New results on the effects of tax policy on the international location of investment," in M. Feldstein (Ed.), The Effects of Taxation on Capital Accumulation (Chicago, IL: University of Chicago Press, 1987), pp. 201-213.

Cummins, Jason G., and R. Glenn Hubbard, "The tax sensitivity of foreign direct investment: Evidence from firm-level panel data," National Bureau of Economic Research Working Paper 4703 (Cambridge, MA: NBER, 1994).

Dunning, John D. (1981) Explaining the international direct investment position of countries toward a dynamic or developmental approach. Weltwirtschafi- liches Arch& 117, 30-64.

Dunning, John D., International Production and the Multinational Enterprise (London: Allen & Unwin, 1981b).

Dunning, John D. (Ed.), Multinational Enterprises, Economic Structure and International Competitiveness (Chichester, UK: Wiley, 1985).

Feldstein, Martin S., and Joosung Joo, "The effect of tax rules on nonresidential fixed investment: Some preliminary evidence from the 1980s," in Martin S. Feldstein (Ed.), The Effects of Taxation on Capital Formation (Chicago, IL: University of Chicago Press, 1987), pp. 101-162.

Froot, Kenneth, Foreign Direct Investment (Chicago, IL" University of Chicago Press, 1992).

Fry, Maxwell J., Money, Interest, and Banking in Economic Development, Second Edition (Baltimore, MD: Johns Hopkins University Press, 1995).

Green, Robert T., Political Instability as a Determinant of U.S. Foreign Investment (Austin, TX: Bureau of Business Research, Graduate School of Business, University of Texas, 1972).

Hartman, David G. (1984) Tax policy and foreign direct investment in the United States. National Tax Journal 37(4), 475-488.

Hartman, David G. (1985) Tax policy and foreign direct investment. Journal of Public Economics 26(1), 107-121.

Hines, James R. Jr, "Forbidden payment: Foreign bribery and American business after 1977," NBER Research Working Paper 5266 (Cambridge, MA: National Bureau of Economic Research, 1995).

Hines, J. R. Jr. (1997) Altered states: Taxes and the location of foreign direct investment in America. American Economic Review g6(5), 1076-1094.

Hines, James R., and Glenn Hubbard, "Coming home to America: dividend repatriations by U.S. multi- nationals," in Assaf Razin and Joel Slemrod (Eds.) Taxation in the Global Economy (Chicago, IL: University of Chicago Press, 1990), pp. 161-200.

Hines, James R. Jr. and Rice, E. M. (1994) Fiscal paradise: Foreign tax havens and American business. Quarterly Journal of Economics 109(1), 149-182.

Hymer, Stephen, and Robert Rowthorne, "Multi- national corporations and international oligopoly: The non-American challenge," in Charles P. Kindle- berger (Ed.) The International Corporation: A Symposium (Cambridge, MA: MIT Press, 1970).

Lizondo, J. Saul, "Foreign direct investment," IMF Working Paper No. 63 (Washington, DC: IMF, 1990).

Lucas, Robert E. B. (1993) On the determinants of direct foreign investment: Evidence from East and Southeast Asia. World Development 21(3), 391-406.

Mauro, Paolo (1995) Corruption and growth. Quarterly Journal of Economics 110, 681-712.

Riedel, James (1995) Intra-Asian trade and foreign direct investment. Asian Development Review XX, 111-146.


Rugman, Alan M. (1977) Risk, direct investment and international diversification. Weltwirtschaftliches Archiv 113, 487-500.

Schneider, Friedrich and Frey, Bruno S. (1985) Economic and political determinants of foreign direct investment. World Development 13(2), 161-175.

Shah, Anwar, and Joel Slemrod, "Do taxes matter for foreign direct investment," Anwar Shah (Ed.) Fiscal Incentives for Investment and Innovation (New York: Oxford University Press, 1990), pp. 481-501.

Slemrod, Joel, "Tax effects on foreign direct investment in the United States: Evidence from a cross country comparison," in Assaf Razin and Joel Slemrod (Eds.) Taxation in the Global Economy (Chicago, IL: University of Chicago Press, 1990a), pp. 79-117.

Slemrod, Joel, "The impact of the tax reform act of 1986 on foreign direct investment to and from the United States," in Joel Slemrod (Ed.), Do Taxes Matter? The Impact of the Tax Reform Act of 1986 (Cambridge, MA: MIT Press, 1990b).

Torrisi, C. R. (1995) The determinants of direct foreign investment in a small LDC. Journal of Economic Development 10, 29-45.

Pan-Long, Tsai (1994) Determinants of foreign direct investment and its impact on economic growth. Journal of Economic Development (June) 19, 137-163.

Vernon, Raymond (1996) International investment and international trade in the product cycle. Quarterly Journal of Economics 80, 190-207.

Wei, Shang-Jin, "How taxing is corruption on foreign investment into East Asia?" Mimeo (Cambridge, MA: Harvard University, Kennedy School, 1997).

Wheeler, David and Mody , A~shoka (1992) Inter- national investment location decisions: The case of U.S. firms. Journal of International Economics 33, 57-76.

Wilson, G. Peter, "The role of taxes in location and sourcing decisions," Alberto Giovannini, R. Glenn Hubbard and Joel B. Slemrod (Eds.) Studies in Inter- national Taxation (Chicago, IL: University of Chicago Press, 1993).

Documents

Host country reforms and FDI inflows: How much difference do they make?