THE IMPACT OF THE EXCHANGE RATE AND ITS VOLATILITY ON FOREIGN DIRECT INVESTMENT IN SUB-SAHARAN AFRICA Adil H. Elsharif-Suliman University of Texas-Pan American, Department of Economics and Finance 1201 West university drive, Edinburg, Texas 78541 E-mail: [email protected] ABSTRACT This paper examines the relationship between foreign direct investment (FDI) and exchange rates for low-income countries of sub-Saharan Africa, using a panel data approach and two-Stage Least Squares (2SLS) method. The result shows that depreciation of the real exchange rate attracts more FDI to sub-Saharan African countries, while an increase in real exchange rate volatility discourages FDI flows to these countries. The U.S. dollar peg system is not an attractive exchange rate regime for foreign investors in sub-Saharan African countries because of exchange rate volatility. The novelty of this study rest on the fact that, the use of different periods to measure the exchange rate volatility gives a plausible support to the conclusion that relative price has magnificent influence on FDI inflows. Introduction This paper examines the relationship between FDI and the real exchange rate and its volatility in 20 sub-Saharan African countries 1 , covering the period from 1980 to 2003. Most of the previous studies about the relationship between FDI and the real exchange rate focus on the industrial developed countries and developing countries, but not on African countries. Froot and Stein (1991) use international capital market with the information imperfections assumption to test the relationship between FDI and exchange rates for the U.S. Goldberg and Klein (1998) investigate the relationship between FDI and the real exchange rate in a regional set of South East Asian and Latin American countries and both the U.S. and Japan. Kiyota and Urata (2004) have empirically examined the relationship between FDI and the real exchange rate from Japan to its partner countries while considering regional and sectoral differences in FDI. They incorporate the impact of the failures of law of one price between different markets on real exchange rate volatility. To extend the investigation of the relationship between FDI and the real exchange rate and its volatility to sub-Saharan African countries, this research extends Froot and Stein’s approach in several ways. First, it examines the impact of the real exchange rate and its volatility on FDI inflows. Second, it examines the impact of the U.S. dollar peg by African countries on FDI inflows, by using a dummy variable to control for the host country’s fixed exchange rate to U.S. dollars. Third, it controls for some macroeconomic variables, including income growth, income level and oil price. Fourth, it uses a two-Stage Least Squares (2SLS) method to control for endogeneity of income growth. Endogeneity of income growth is repeatedly indicated by new classical endogenous growth model. The current literature repeatedly has pointed to the role of oil prices in explaining growth. Both Hamilton (1983) and Lee et al. (1995) find that oil price changes are highly significant in explaining economic growth. I use 2SLS regression, which 543

mainly focuses on the relationship between oil prices and GDP growth. After controlling for endogeneity of income growth, the relationships between FDI and the real exchange rate and its volatility are robust across variable definitions and model specifications, all the coefficients, especially real exchange rate volatility coefficients are robust and increase in sizes. The result shows that depreciation of the real exchange rate attracts more FDI to subSaharan African countries while an increase in real exchange rate volatility discourages FDI flows to these countries. The coefficients of real exchange rate volatility are robust and highly significant with an increase in the standard deviation periods used to measure the volatility of the exchange rate. Given that there are high correlations between the measurements of exchange rate volatility, the use of one of these measurements at a time shows a relative increase in Adjusted R-Square. With the increase in the period used to measure exchange rate volatility, there is an increase in the noise, which is manifested through the exchange rate. The use of different periods to measure the exchange rate volatility gives a plausible support to the result that relative price has magnificent influence on FDI inflows. The result also indicates that the U.S. dollar peg system has positive impacts on FDI inflows, but these impacts are not robust across different measures and model specifications. Therefore, the U.S. dollar peg system is not an attractive exchange rate regime for foreign investors in sub-Saharan African countries because real exchange rate volatility has discouraged FDI inflows. Where most of the countries in sub-Saharan Africa have a high percentage of their incoming foreign investments in primary commodities, all these findings, in this paper, point to relative price as a major factor that might influence FDI inflows. Based on these findings, I conclude that both flexibility and uncertainty of exchange rate (volatility) should be considered in designing a successful macroeconomic policy related to FDI. Data Analysis and Statistical Methods This section describes the key variables used in this paper as follows: FDI is defined as the net inflow percentage of GDP. The real exchange rate for each country in sub-Saharan Africa is calculated as follows: EX = Oex (Pus /Phome) where EX is the real exchange rate, Oex is the official bilateral exchange rate, Pus is the U.S. consumer price index (CPIus), and Phome is the host country’s consumer price index (CPIhome). The official exchange rate is calculated as local currency units relative to the U.S. dollar, so that a rise in the exchange rate is a depreciation of the host country currency. I use the spot oil price of West Texas Intermediate with the monthly frequency and a unit of dollars per barrel. These data are obtained from the Federal Reserve Bank of St. Louis. GDP per capita based on purchasing power parity (PPP) is used to control for the income level. The use of PPP helps to control for the distortions due to anomalies of real exchange rate (nontradable goods and service, tariff and taxes). Growth of GDP is used to control for market growth. Data for all of these variables, except oil prices, are from the World Development Indicators (WDI) provided by the World Bank. For USPeg, this research uses the classification table provided by Levy-Yeyati and Sturzenegger (2003) to classify the countries used. I compare the result from this classification table (de facto classification), where countries are classified according to some economic criteria, with the result from the classification provided by IMF (de jure classification) where the classification was announced by the countries themselves, both results are similar. 544

I use the fixed effects model (FEM), where the intercept is cross-section (group) specific, assuming that the effects of omitted countries-specific variables are constant over time and correlate with the repressors. Two major types of omitted variables are important to control for, the country time-invariant variables and the period country-invariant variables. I estimate GLS estimators which use estimated cross-section residual variances (CSW), assuming that the residuals are cross-section heteroskedastic and contemporaneously uncorrelated. I estimate GLS estimators which use estimated cross-section residual covariance matrix (SUR), assuming that the residuals are both cross-section heteroskedastic and contemporaneously correlated. Although the model residuals are assumed to be normally distributed and homogeneous, the model might have autocorrelation because of time-lagged temporal effects, or it may contain variables that do not vary within the groups. This could lead to country-specific (group-wise) heteroskedasticity bias that would further contaminate the estimation. I also check all of these biases by using two residual tests, as suggested by Mollick et al. (2006). The first test is the tstatistic test, which is associated with the lagged residual within a standard Lagrange Multiplier test on the residuals of the panel data regression. The second test is the LM-NR2 statistic test. This test is driven from the value of N and R2 computed in second auxiliary regression of the panel data; this statistic test follows a chi-squared distribution with degrees of freedom equal to the number of estimated parameters in the auxiliary regression. The LM-NR2 statistic is calculated under the null hypothesis of no serial correlation up to lag order 1, which is reasonable for annual data. All of these tests help us to check for both heteroskedasticity and serial correlations that may originate from the measured variables in the regression. Because our model is subjected to endogeneity of the income, where the disturbance term of GDP growth is correlated with the disturbance term of the dependant variable, violating ordinary least squares (OLS) regression’s assumption of recursivity. To test for endogeneity of income growth, I use Durbin-Wu-Hausman tests, and I use 2SLS to correct for endogeneity of income growth. Model and Methodology To model the relationship between FDI and real exchange rate, the model is based on Froot and Stein (1991). In addition, I control for exchange rate volatility, U.S. dollar peg system, and some macroeconomic variables. Panel data from 20 sub-Saharan African countries were used in this research. The benchmark model specification is built on the following equation: ln

β 3 ln

FDI it GDPit

=

α 0 + α i TRENDit

INCOME_LEVELit +

β4

+

β1 ln

EXit +

β 2 ln

EX_VOLit

INCOME_GROWTHit + β 5 USPeg + μ it

(i = 1,2……,20) and (t = 1,2………24)

(1)

where i denotes country and t denotes time, EXit is the real exchange rate between the ith country’s currency and the U.S. dollar, EX_VOLit is the volatility of the real exchange rate calculated with three different standard deviation periods, TRENDit is the trend, INCOME_LEVELit stands for income level, and INCOME_GROWTHit stands for income growth, USPeg is a dummy variable to control for fixing a host country’s exchange rate to the 545

U.S. dollar (where α i > 0, β 1 > 0, β 2 ?, β 3 > 0 β 4 > 0, β 5 ? ). For a description of the variables in equation (1), refer to Table 1. I compute the basic statistics and correlations for the variables used in the analysis. The mean values and standard deviations of the variables show the large variation in these variables across the sample countries. It is clear that the variability is not only across years but also across countries. I observed strong correlations between all measurements of volatility (SD_12M_VOL, SD_24M_VOL, and SD_36M_VOL) and the real exchange rate, see Table 2. I enter each of these volatility measurements into the regression equation separately. Table 1: Summary Statistics Variable Name

Mean

Std. Dev.

Maximum

Minimum

Observations

Cross sections

FDI EX Trend SD _12M_Vol SD _24M_Vol SD _36M_Vol Oil Price INCOME_LEVEL INCOME_GROWTH US Peg Dummy

1.3383 395.8495 12.2877 292.3374 512.3358 719.5851 4.0117 2507.1480 0.0296 0.4733

2.2922 740.2903 6.5922 2338.0980 3638.4190 4812.7090 2.1739 2579.2340 0.0494 0.4999

14.9066 5547.8710 24.0000 38826.7100 45063.4100 58378.9100 10.2705 11420.2400 0.1945 1.0000

-6.8978 2.4013 1.0000 0.0254 0.0590 0.0784 1.5358 489.4582 -0.1510 0.0000

431 431 431 431 431 431 431 431 431 431

20 20 20 20 20 20 20 20 20 20

FDI: EX:

Foreign Direct Investment, net inflows % of GDP. Real exchange rate calculated as local currency units relative to the U.S. dollar, so that a rise in exchange rate is depreciation of host country currency. Real exchange rate volatility, standard deviation for 12 months. Real exchange rate volatility, standard deviation for 24 months. Real exchange rate volatility, standard deviation for 36 months.

EX_12M_VOL: EX_24M_VOL: EX_36M_VOL: Oil Price:

Spot oil price: West Texas Intermediate—frequency is monthly, units is dollars per barrel, covers the period between 1980 and 2003, converted to standard deviation for 12 months. INCOME_LEVEL: GDP per capita, PPP (constant 2000 US$). Gross domestic product per capita converted to international dollars PPP. INCOME_GROWTH: Growth of GDP (constant 2000 US$) calculated as percentage change. Dollar figures for GDP, converted from domestic currencies using 2000 official exchange rates.

Table 2: Correlation Matrix FDI EX TREND SD_12_M SD_24_M SD_36_M OIL PRICE INCOME_LEVEL INCOME_GROWTH USPEG

FDI

EX

TREND

SD_12_M

SD_24_M

SD_36_M

OIL PRICE

INCOME_LEVEL

INCOME_GROWTH

USPEG

1.0000 0.1498 0.1796 -0.1080 -0.0961 -0.0822 0.0568 0.2415 0.2482 0.0387

1.0000 0.0841 0.6981 0.6965 0.6957 0.0105 -0.4215 -0.0799 -0.2311

1.0000 -0.0513 -0.0440 -0.0381 -0.1115 0.0698 -0.0131 0.1951

1.0000 0.9760 0.9625 0.0362 -0.5207 -0.2076 -0.2425

1.0000 0.9859 0.0144 -0.5170 -0.1925 -0.2563

1.0000 0.0039 -0.5201 -0.1751 -0.2708

1.0000 0.0304 0.0036 0.0118

1.0000 0.2263 0.1384

1.0000 -0.0243

1.0000

Regarding the relationship between FDI and real exchange rate level, Blonigen (1997), Cushman (1985) provide evidence that a depreciation in a host country’s real exchange rate will lead to an increase in FDI inflows to that country. Based on the definition of the real exchange 546

rate in this research, as units of host country currency per one U.S. dollar, an increase in the coefficient between FDI and the real exchange rate means depreciation in the host country currency. In this case, I expect to see a positive statistical relationship between FDI and real exchange rate level. Volatility’s coefficient can be affected by the way volatility is specified and measured. Cushman (1985) uses different measurements of volatility, based on surprise of the currency and standard deviation. He finds a positive relationship between FDI and real exchange rate volatility. Dell'ariccia (1999) uses different variables to measure exchange rate volatility: the standard deviation of the first difference of the logarithmic exchange rate and the sum of the squares of the forward errors. He indicates that estimated results are consistent with the previous studies, confirming a negative effect of volatility on trade. Baek and Okawa (2001) use a log of the nominal exchange rate to calculate exchange rate volatility. They indicate that the use of the logarithm transformation on volatility improves the property of the data. Geweke (1986) and Nelson (1991) indicate that volatility should be specified in a logarithmic form in order to avoid non-negativity and produce an uncorrelated residual. To measure the impact of oil price variability on FDI, it is not the level of oil price but rather the change in oil price that is the most effective measure used in prior literature. Hamilton (1983) finds that oil price change has a negative correlation with real U.S. GNP growth from 1948 to 1980. Lee et al. (1995) construct an oil price-shock variable that reflects the unanticipated and time-varying conditional variance of oil price change. This variable is found to be highly significant in explaining economic growth across different sample periods. Peter (1996) measures volatility with the standard deviation of daily prices and uses industrial production growth as a proxy for economic growth. He finds that both oil price changes and oil price volatility explain a reasonable fluctuation in industrial production. In this paper I concern mainly with the variability of oil prices that may influence FDI or interact with the income growth to spill over to FDI inflows, therefore, the standard deviation of oil price over a one-year period is used to measure oil price variability. Discussion of the Results Benchmark Model I use the fixed effects method to estimate equation 1. I apply the cross-section weights technique (CSW), assuming the presence of cross-section heteroskedasticity and contemporaneously uncorrelated residuals, and I apply the seemingly unrelated regression (SUR) weighted least squares technique, assuming both cross-section heteroskedasticity and contemporaneous correlation. It seems reasonable to expect that some of sub-Saharan African countries might behave differently than others. At the same time, all of these countries might be similarly affected by certain factors such as historical or ideological backgrounds, which may influence all of the countries during the post-independence period. I base the result on comparing the estimated results of these two techniques. Though both techniques come up with similar results, coefficient values under the SUR technique are larger. Exchange rate volatilities are measured by standard deviations for 12, 24, and 36 months. Table 3 reports the results for the benchmark model, using Equation 1. I find that, under both CSW and SUR techniques, all of the estimated coefficients for the real exchange rate and its volatility are statistically significant and have correct signs, but 547

estimated coefficients for the TREND are insignificant. The trend variable is used to control for an upward or downward trend in FDI flows to sub-Saharan African countries. The trend coefficients may be swept away by countries’ cross-sectional effects; there is overwhelming evidence in the current literature that FDI flows to these countries are unstable. Adjusted R2 values show some improvement in value between different measures of volatility. Given that there are high correlations between the measurements of exchange rate volatility, the use of one of these measurements one at a time shows a relative increase in Adjusted R-Square. The use of exchange rate volatilities measured by standard deviations for 12, 24, and 36 months, in Table 3, is working like nonnested models because neither measure is a special case of the other. Therefore, based on Adjusted R2 it is clear that with the increase in the period used to measure exchange rate volatility, there is an increase in the noise which is manifested through the exchange rate. Two tests of serial correlation were run, a t-test and an LM-NR2 test; only the t-test indicates the presence of serial correction under all measures of volatility. To address the serial correlation in the t-tests with a larger sample, it is possible to find a serial correlation even though its effect is very slight. The results in Table 3 prove that real exchange rate depreciation will increase FDI inflow, and real exchange rate volatility will decrease FDI inflows. One interesting finding is that the coefficients of real exchange rate volatility are increasing in size and significance with an increase in the standard deviation periods used to measure the volatility of the exchange rate. The results indicate that in the long-run, i.e., two- and three-year periods, exchange rate volatility has a larger impact on FDI inflows than short-run volatility, i.e., oneyear standard deviation volatility. Regarding the results for real exchange rate level, all of the real exchange coefficients are significant. Table 3: Benchmark Model ln

FDI it GDPit

=

α0

+

α i TRENDit

+

β 1 ln

EXit +

β 2 ln

EX_VOLit +

β 3 ln INCOME_LEVELit

+ β 4 ln INCOME_GROWTHit + β 5 USPeg + μ it ( i = 1,2……,20) and ( t= 1,2………24) Notes: Annual data from 1980 to 2003 for 20 countries (N). FEM stands for fixed intercept. Use both cross-sectional weights and seemingly unrelated regression. For the Fixed Effects Model, the White heteroskedasticity covariance matrix is employed. T-statistic is reported in parentheses. The asterisks *, **, and *** indicate significance levels of 10%, 5%, and 1%, respectively. Volatility Model

EX EX_VOL TREND INCOME_LEVEL INCOME_GROWTH USPeg

2

ADJ R N T RESIDUAL-TESTS T-TEST 2 LM-NR

SD_12M_ Vol 1 2 FEM FEM CS-Weight Sur Weight

SD_24M_ Vol 3 4 FEM FEM CS-Weight Sur Weight

SD_36M_ Vol 5 6 FEM FEM CS-Weight Sur Weight

0.5024 (4.0754)*** -0.0560 (-1.9322)* 0.0128 (1.5160) 1.1492 (3.8107)*** 2.9256 (2.9369)*** 0.2441 (1.1679)

0.6111 (5.4368)*** -0.1225 (-4.6893)*** 0.0009 (0.1328) 1.3320 (7.6635)*** 2.3599 (3.0329)*** 0.5253 (3.6245)***

0.5375 (4.0114)** -0.0880 (-2.5403)*** 0.0111 (1.2649) 1.1077 (3.6530)*** 3.2502 (3.1478)*** 0.3388 (1.6249)

0.6702 (6.0117)*** -0.1412 (-4.8335)*** 0.0027 (0.4138) 1.2122 (6.7153)*** 2.6378 (3.2839)*** 0.5511 (3.6625)***

0.5089 (4.1519)** * -0.1222 (-3.7812)* ** 0.0132 (1.5534) 1.0966 (3.6424)** * 3.9488 (3.9145)** * 0.3464 (1.7269)*

0.6359 (6.8283)*** -0.1508 (-5.9745)*** 0.0067 (1.1184) 1.2368 (6.8894)*** 4.1949 (5.8576)*** 0.4854 (3.4458)***

0.5448 20 24

0.4537 20 24

0.5519 20 24

0.4602 20 24

0.5632 20 24

0.4667 20 24

(7.6206)***

(8.5088)***

(7.1899)***

(7.8712)***

(7.0015)***

(7.7049)***

(4.4517)

(2.0107)

(4.2961)

(1.9591)

(4.2495)

(1.7185)

548

I add the oil price variable to the benchmark model. The result shows that there are small improvements in the coefficients of exchange rate volatilities, the coefficients for the trend start to change in values and significance after the oil price variable is added, USPeg coefficients also change in values and significance. It is clear that most of the coefficients in the model are sensitive to the introduction of oil price variable. Then I remove income growth and income level while keeping oil price with remaining variables in the benchmark model. The result shows that the coefficients of the real exchange rate and the coefficients of the exchange rate volatilities increase significantly. The results from these two tests indicate very clearly that oil price is interacting with income growth when it is included as an independent variable in the model. When income level and income growth are dropped from the model, oil price captures less variation in FDI, (The tables which contains the results for these tests are available upon request). Using the ceteris paribus interpretation of the multiple regression analysis, to explain the results; in the first test, I include oil price along with the same variables used in the benchmark model. In the second test, I include oil price but remove income level and income growth from the variables used in benchmark model. If oil price influences FDI indirectly and not only through GDP growth, the coefficients of oil price should be increase in second test. In other words, given the positive relationship between the GDP growth variable and the oil price variable; the inclusion of the oil price with GDP growth in the same model should underestimate the effect of oil price and not overestimate it; the coefficients of oil price in the second test, where GDP growth is removed, should increase over the coefficients of oil price in the first test, where both variables are included. Comparing the results in these two tests, the coefficients of oil price decrease instead of increase in the second test. Therefore, the results from these two tests indicate very clearly that including the oil price violates the ceteris paribus interpretation of the multiple regressions. Two Stage Least Squares To examine for endogeneity of income growth, the Durbin-Wu-Hausman (DWH) test is applied. The DWH test suggested by Davidson and MacKinnon (1993). The result of this test indicates that there is a correlation of error between income growth and FDI over the period 1980–2003. Table 4 reports the results for the 2SLS test. The results are acceptable and show improvements over the regular panel regression results. All of the coefficients increase in size and significance when 2SLS is used. The 2SLS estimate of income growth is almost twice as large as the OLS estimate, but its T-value is smaller than the T-value for the OLS estimate. The loss in significance level of 2SLS estimate is the price to pay when we have an endogenous variable, where the standard error of the estimate for 2SLS is larger than the standard error of the OLS estimate, which affects the significance of the tests; this indicates that income growth is endogenous. Though the U.S. dollar peg variable tests yield signs similar to previous tests, the overall findings indicate that the U.S. dollar peg system is not an attractive system for potential investors in sub-Saharan African countries. The exchange rate volatility has two aspects 2 : flexibility and uncertainty. The U.S. dollar peg system may reduce flexibility by controlling the nominal price of exchange rate adjustments, creating a favorable FDI environment, but the exchange rate uncertainty may lead to overvaluation of the host currencies and reduce the attractiveness of host countries to foreign investors. 549

Finally, to make sure that the results are not affected by the functional form of the model, I repeat all previous tests by using the logarithm of income growth. The results from these tests are available upon request. The results show correct signs for all of the variables in the model except GDP growth, but most coefficients turn insignificant. The negative and small numbers in the income growth series may cause these changes. To insure against the sensitivity of the result for model changes, I run three separate analyses, first dropping the year 2003 from the data, then the year 2003 and 2002, and finally, the years 2003, 2002 and 2001; the results are similar. I drop the country with upper-middle income, leaving only the countries with low-income; the results are robust.

FDI it GDPit

Table 4: Model (2SLS _ Estimate) = α 0 + α i TRENDit + β 1 ln EXit + β 2 ln EX_VOLit + β 3 ln INCOME_LEVELit + β 4 ln INCOME_GROWTHit + β 5 USPeg + μ it ( i = 1,2……,20) and ( t= 1,2………24)

Notes: Annual data from 1980 to 2003 for 20 countries (N). FEM stands for fixed intercept. Use both cross-sectional weights and seemingly unrelated regression. For the Fixed Effects Model, the White heteroskedasticity covariance matrix is employed. T-statistic is reported in parentheses. The asterisks *, **, and *** indicate significance levels of 10%, 5%, and 1%, respectively. Volatility Model

EX EX_VOL TREND INCOME_LEVEL INCOME_GROWTH USPeg

ADJ R2 N T DW RESIDUAL-TESTS T-TEST LM-NR2

SD_12M_Vol 1 2 FEM FEM CS-Weight Sur Weight

SD_24M_Vol 3 4 FEM FEM CS-Weight Sur Weight

SD_36M_Vol 5 6 FEM FEM Sur Weight CS-Weight

0.6321 (4.0060)*** -0.0890 (-1.8682)* 0.0145 (1.4629) 0.9310 (2.8680)*** 5.9227 (1.9386)* 0.2242 (1.0039)

0.6998 (5.2911)*** -0.1733 (-4.4962)*** 0.0064 (0.9066) 1.2838 (7.5104)*** 4.8767 (2.2304)** 0.3773 (2.3717)**

0.7038 (4.0881)*** -0.1537 (-2.7378)**) 0.0114 (1.1170) 0.8635 (2.6673)*** 6.4019 (2.0659)** 0.3298 (1.4677)

0.7280 (5.4701)*** -0.2290 (-5.1293)*** 0.0087 (1.2170) 1.1301 (6.3453)*** 5.9501 (2.6614)*** 0.4322 (2.6954)***

0.6596 (4.0676)*** -0.1676 (-3.0098)*** 0.0143 (1.4391) 0.8033 (2.5200)** 7.1201 (2.3620)** 0.3502 (1.5855)

0.6936 (5.5628)*** -0.2088 (-4.8000)*** 0.0138 (2.0040)** 1.0641 (5.7938)*** 7.2250 (3.2516)*** 0.4223 (2.7613)***

0.5323 20 24 1.1697

0.4492 20 24 1.2889

0.5407 20 24 1.1544

0.4562 20 24 1.2859

0.5440 20 24 1.1788

0.4593 20 24 1.3191

(8.0274)***

(8.7036)***

(7.7528)***

(8.2624)***

(7.6797)***

(8.4149)***

(4.9231)

(2.1306)

(4.9143)

(2.0689)

(4.9878)

(1.9220)

Concluding Remarks The results show that both real exchange rate and its volatility (price) have influenced the FDI inflows. The depreciation of the host country’s real exchange rate increases FDI inflows, and the increase of the host country real exchange rate volatility decreases FDI inflows. The results are robust and consistent among different terms used to measure volatility. The volatility of one-year, two- and three-year standard deviation periods have significant and consistent impacts on FDI inflows. The coefficients of exchange rate volatility (noises) increase in size and significance with an increase in the standard deviation periods used to measure the volatility of the exchange rate. Different volatility indexes (measures) have shown the importance of the real exchange rate volatility (prices) in explaining FDI inflows. Most of the sub-Saharan African 550

countries peg their currencies at a fixed rate to the U.S. dollar. This research finds that the impacts of these pegs on FDI inflows are inconsistent across different model specifications. The policy of maintaining stable nominal exchange rates under high exchange rate volatility may lead to a loss in price competitiveness and a reduction in FDI inflows. The use of the pegged exchange rate as an incentive to attract FDI involves a trade-off between uncertainty and flexibility, and as a result, price competitiveness will influence FDI inflows.

REFERENCES Ardeni, Pier Giorgio, 1989, Does the Law of One Price Really Hold for Commodity Prices?, American Journal of Agricultural Economics 71, 661-669. Baek, In-Mee, and Tamami Okawa, 2001, Foreign Exchange Rates and Japanese Foreign Direct Investment in Asia, Journal of Economics and Business 53, 69-84. Blonigen, Bruce A., 1997, Firm-Specific Assets and the Link between Exchange Rates and Foreign Direct Investment, American Economic Review 87, 447-465. Cushman, David O., 1985, Real Exchange Rate Risk, Expectations, and the Level of Direct Investment, Review of Economics & Statistics 67, 297-308. Davidson, Russell , and James MacKinnon, 1993. Estimation and Inference in Econometrics (Oxford University Press, New York). Dell'ariccia, Giovanni, 1999, Exchange Rate Fluctuations and Trade Flows: Evidence from the European Union, IMF Staff Papers 46, 315-344. Froot, Kenneth A., and Jeremy C. Stein, 1991, Exchange Rates and Foreign Direct Investment: An Imperfect Capital Markets Approach, Quarterly Journal of Economics 106, 11911217. Geweke, J., 1986, Modeling Persistence of Conditional Variances: A comment, Econometric Reviews 5, 57-61. Goldberg, Linda S., and Michael Klein, 1998, Foreign Direct Investment, Trade and Real Exchange Rate Linkages in Developing Countries, "in R.Glick (ed), Managing Capital Flows and Exchange Rates: Perspectives from Pacific Basin", Cambridge University Press 73-100. Hamilton, James D., 1983, Oil and the Macroeconomy since World War II, The Journal of Political Economy 91. Kiyota, Kozo, and Shujiro Urata, 2004, Exchange Rate, Exchange Rate Volatility and Foreign Direct Investment, World Economy 27, 1501-1536.

551

Lee, Kiseok, Shawn Ni, and Ronald Ratti, 1995, Oil Shocks and the Macroeconomy: The Role of Price Variability, The Energy Journal 16, 39-56. Levy-Yeyati, Eduardo, and Federico Sturzenegger, 2003, Classifying Exchange Rate Regimes: Deeds vs. Words, European Economic Review 49, 1603-1635. Mollick, Andre Varella, René Alberto Ramos Durán, and Esteban Silva Ochoa, 2006, Infrastructure and FDI Inflows into Mexico: A Panel Data Approach, Global Economy Journal 6, 1-25. Nelson, Daniel B., 1991, Conditional Heteroskedasticity in Asset Returns: A New Approach, Econometrica 59, 347-370. Peter, Ferderer J., 1996, Oil Price Volatility and the Macroeconomy, Journal of Macroeconomics 18, 1-26. Footnotes 1

Burundi, Burkina Faso, Botswana, Central African Republic, Cote D’Ivoire, Cameroon, Congo Republic, Congo Democratic Republic, Ethiopia, Ghana, Gambia The, Mauritius, Malawi, Nigeria, Rwanda, Swaziland, Togo, Uganda, South Africa and Zimbabwe. 2 See Ardeni (1989) for more information about the impact of real exchange rate volatility.

552