Portal de Periódicos da CAPES

Exchange rates and US direct investment into Latin America

2008; Taylor & Francis; Volume: 17; Issue: 3 Linguagem: Inglês

10.1080/09638190802137083

1469-9559

Isabel Ruiz, Susan Pozo,

Global Financial Crisis and Policies

Abstract This paper analyzes the impact of exchange rate levels and exchange rate uncertainty on US foreign direct investment into Latin America. By decomposing exchange rate uncertainty into temporary (short-run) and permanent (long-run) components, we further explore whether the nature of uncertainty matters. Our empirical findings support the view that exchange rate uncertainty has a negative impact on US investment flows into Latin America. Moreover, it is the persistency in uncertainty rather than transitory uncertainty that mostly deters foreign investment. In contrast, investors do not appear to be affected by discrete movements in exchange rate levels. Keywords: foreign direct investmentexchange ratesCGARCHLatin AmericaJEL Classification: F31F23 Acknowledgements The authors gratefully acknowledge numerous helpful comments on an earlier version of the paper from the editor and an anonymous referee. The authors would also like to thank Mark Wheeler, Ana Maria Herrera, Carlos Vargas-Silva, Mark W. Frank and Hiranya Nath for valuable comments and suggestions. Ruiz acknowledges support provided by Sam Houston State University's Faculty Research Grant. Notes 1. The US ranks first as the source of FDI inflows to Latin America (62%), with Europe ranking second (31%) and Japan third (7%) (ECLAC 2000 ECLAC 2000 Foreign investment in Latin America and the Caribbean 2000 Report [Google Scholar]). 2. For a few exceptions with respect to Latin America, refer to Goldberg and Klein (1997 Goldberg, L. and Klein, M. . Foreign direct investment, trade and real exchange rate linkages in Southeast Asia and Latin America. National Bureau of Economic Analysis (NBER), Working Paper: 6444. [Google Scholar]) and Esquivel and Larrain (2002 Esquivel, G. and Larrain, F. . The impact of G-3 exchange rate volatility on developing countries. Research Papers for the Intergovernmental Group of Twenty-Four on International Monetary Affairs. United Nations and Harvard Center for International Development. [Google Scholar]). While these studies explore how exchange rate changes affect FDI, they do not address the impact of exchange rate uncertainty on FDI. 3. Besides analyzing the channels through which exchange rates affect FDI, some studies have also suggested that not only is the channel relevant, but also the type of FDI flow (see Blonigen 1997 Blonigen, B. 1997. Firm specific assets and the link between exchange rates and foreign direct investment. The American Economic Review, 87(3): 447–465. [Web of Science ®] , [Google Scholar]). 4. The OLI model, first developed by Dunning (1980 Dunning, J. 1980. Towards an eclectic theory of international production: some empirical tests. Journal of International Business Studies, 11(1): 9–31. [Crossref], [Web of Science ®] , [Google Scholar], 1988 Dunning, J. 1988. Explaining International Production, London and Boston: Unwin Hyman. [Google Scholar], 1995 Dunning, J. 1995. Reappraising the eclectic paradigm in an age of alliance capitalism. Journal of International Business Studies, 26(3): 461–491. [Crossref], [Web of Science ®] , [Google Scholar]), is also known as the ‘eclectic model’ of FDI because it is a blend of three different theories of foreign direct investment. ‘O’ stands for ownership advantages (firm specific advantages). ‘L’ reflects location advantages that influence the choice of investment location. This choice depends, among others, on factors that include the economic conditions of a country. Finally, ‘I’ reflects internalization advantages. Internalization (firm) advantages addresses the question of how to invest abroad. 5. We also included the SP500 as an alternative measure of US wealth to test for robustness. In addition, since FDI is most often thought to be a long-run investment, we proxied the cost of capital (the interest rate in the US) with the Triple A 10-year bond rate. Neither of these variables added significant effects to the results so they are not included in the final results. All data, except the nominal exchange rate and the US interest rates, are seasonally adjusted. See the appendix for additional detail on data procedures and sources. 6. We are thankful to an anonymous referee for this suggestion. For robustness, we broadened the dummy variable idea to account for intermediate exchange rate regimes. We checked for robustness in two different ways. First, we created two dummy variables. The first dummy (D 1) took a value of one for periods of time when the exchange rate was classified as fixed. The second dummy (D 2) took a value of one when the exchange rate was classified as intermediate. Second, we also constructed a categorical variable that takes a value of zero for times in which the exchange rate was classified as a flexible exchange rate, a value of one for intermediate exchange rate regimes and a value of two for fixed exchange rate systems (pegged to the US dollar). The results of both estimates are consistent with the original specification. While the overall results for all the variables also remained consistent, the only difference to note is that in specification (2), our measure of volatility turned out to be statistically significant. 7. It has been argued that, when dealing with panel data, panel unit root tests may have higher power than unit root tests based on individual time series (see Baltagi 2001 Baltagi, B. H. 2001. Econometric Analysis of Panel Data, 2nd ed., Chichester: Wiley. [Google Scholar] for more details). 8. We further estimated all the regressions including lagged uncertainty. That is, instead of matching uncertainty in the current quarter with FDI in the current quarter, we use uncertainty in the previous quarter as the regressor in an equation estimating FDI in the present quarter. The estimation results, not reported here but available from the authors, did not differ substantively from the results we present here. 9. The measures of uncertainty included in the expressions provided in Table 2 are generated regressors. We should note that it is important to account for this for the reported standard errors. However, we note this as a limitation in our study. We would normally apply a two-stage regression form of estimation. However, we are unable to correct for the generated regressors problem because the frequency of the data in the GARCH model does not match the frequency of the data in the FDI equation. 10. Kiyota and Urata (2004 Kiyota, K. and Urata, S. 2004. Exchange rate, exchange rate uncertainty and foreign direct investment. World Economy, 27(10): 1501–1536. [Crossref] , [Google Scholar]) find that for Japanese investors, the US dollar pegged exchange rate system has a mixed impact on FDI flowing to a sample of developing countries. They find that, for some industries, the impact of a pegged currency is negative while for others it is positive. 11. In March 1991, the Argentinean Congress passed the ‘Convertibility Law’ fixing the peso's exchange rate at par with the US dollar. The Convertibility Law required the Argentinean peso to be fully backed with dollar reserves. Moreover, under convertibility, the owner of a peso had the right of freely converting pesos into dollars at the fixed rate of one for one. 12. Most empirical work finds that GARCH (1,1) adequately represents the conditional variance (see Bollerslev et al. 1992 Bollerslev, T., Chou, R. and Kroner, K. 1992. ARCH modeling in finance: a review of the theory and empirical evidence. Journal of Econometrics, 52(1–2): 5–59. [Crossref], [Web of Science ®] , [Google Scholar]). In cases where the GARCH (1,1) does not fit the series well, ARCH(1) is often adequate. 13. This result, although unusual, is sometimes the case for developing countries' exchange rates (see Speight and McMillan 2001 Speight, A. and McMillan, D. G. 2001. Volatility spillovers in East European black market exchange rates. Journal of International Money and Finance, 20(3): 367–378. [Crossref] , [Google Scholar]). We conducted a formal test of the null hypothesis of integration in variance for the Brazilian real exchange rate on the basis of a Wald test of the restriction α1 + β1 = 1. The null could not be rejected at the 1% level of significance. Integration in variance is often the result of structural breaks in the unconditional variance that produce a clustering of large and small deviations and is reflected in extreme GARCH persistence (see Speight and McMillan 2001 Speight, A. and McMillan, D. G. 2001. Volatility spillovers in East European black market exchange rates. Journal of International Money and Finance, 20(3): 367–378. [Crossref] , [Google Scholar] and Lamoreaux and Lastrapes 1990 Lamoreaux, C. and Lastrapes, W. 1990. Persistence in variance, structural change and the Garch model. Journal of Business and Economic Statistics, 8(2): 225–234. [Google Scholar]). 14. As a robustness test we also used the last monthly observation of uncertainty (within each quarter) as a measure of uncertainty for that quarter. 15. To explore the possibility of a nonlinear relationship, the squared terms of these uncertainty proxies were also used in the model. These results, not reported here, were almost identical to the reported results in Table 2. 16. The original model defines the permanent component as a unit root process (ρ = 1). However, Engle and Lee (1999 Engle, R. and Lee, G. 1999. “A Long-Run and Short-Run Component Model of Stock Return Volatility”. In Cointegration, Causality and Forecasting: A Festschrift in Honour of Clive W. J. Granger, Edited by: Engle, R. and White, H. 475–497. Oxford: Oxford University Press. [Google Scholar]) extend the model to a more general specification in which they allow the permanent component to be a non-unit root process. 17. See Engle and Lee (1999 Engle, R. and Lee, G. 1999. “A Long-Run and Short-Run Component Model of Stock Return Volatility”. In Cointegration, Causality and Forecasting: A Festschrift in Honour of Clive W. J. Granger, Edited by: Engle, R. and White, H. 475–497. Oxford: Oxford University Press. [Google Scholar]) for a more detailed explanation of the stationarity and non-negativity conditions. Note that the component model reduces to the GARCH(1,1) if either α1 = β1 = 0 or ρ = φ = 0. 18. Plots of the GARCH estimates and of the permanent and temporary components for each of the countries are available from the authors upon request. 19. Calderon and Schmidt-Hebbel (2003 Calderon, C. and Schmidt-Hebbel, K. 2003. Macroeconomic policies and performance in Latin America. Journal of International Money and Finance, 22(7): 895–923. [Crossref], [Web of Science ®] , [Google Scholar]) show evidence that portfolio equity and debt flows do not impact growth while FDI is the only major category of capital inflows that is relevant for long-term growth in Latin America. Furthermore, Borensztein et al. (1999) find that FDI flows into developing countries contributed to economic growth in a proportion greater than domestic investment. 20. We note one limitation of our study. The exchange rate uncertainty variables that we incorporate into our FDI equations are generated regressors. One normally should correct the standard errors of generated regressors to more accurately assess their significance in the regression equation. As we noted before, we are unable to correct for the generated regressor problem because the frequency of the data in the GARCH model does not match the frequency of the data in the FDI equation.