The Effects of Exchange Rate Volatility on
Foreign Direct Investment
Master Thesis
at
the Faculty of Economics and Business
at
the University of Groningen, the Netherlands
Student number: S4099354
Name:
Verena Lang
Supervisor:
Prof. Dr. Lensink
Second Assessor: Kaat, PhD
Study Program:
MSc IFM
Date:
May 30, 2020
1
Introduction
The ending of the Bretton Woods Agreement of monetary management in the 1970s made exchange rate volatility a primary source of uncertainty. Nowadays, companies that do not operate in foreign markets are exposed due to foreign competition regardless and need to have a strategy for how to deal with such problems. Naturally, companies are even more affected when operating cross-border. Nevertheless, foreign direct investments (hereafter abbreviated by FDI) have been encouraged by globalization, international agreements and reduced trade barriers. Given those facts, understanding how exchange rate volatility and FDI affect each other has become an important question in fields such as business, economics, finance and management.Figure (1): Average FDI inflows and exchange rate volatility This figure shows the average FDI inflows and exchange rate volatility across all countries in the sample (AUS, AUT, BEL, BRA, CAN; CHE, CHL, CHN, CZE, DEU, DNK, ESP, EST, FIN, FRA, GBR, GRC, HUN, IND, IRL, ISL, ISR, ITA, JPN, KOR, LTU, LUX, LVA, MEX, NLD, NOR, NZL, POL, PRT, SVK, SVN, SWE, TUR, USA, ZAF) for each year between 2001 and 2012. Exchange rate volatility is lagged by one year. FDI is in total numbers and not standardized by a county’s GDP as used in further analysis. Other than that, the variables are defined as in Table (A3) in the appendix.
volatility. For developed countries, mixed results were found. Finally, the paper tries to answer why this difference exists and what the results found mean for companies. Specifically, the following three differences between developed and developing nations are being discussed: the magnitude of exchange rate variability, the degree of openness and additional risk. While exchange rate variability does not seem to explain the divergence, high levels of openness are found to influence the effect exchange rate variability has on FDI. Additional risks are proven to highly correlate with the development of a country. The main contribution of this paper is to analyze the opposing hypotheses found in literature and to provide further evidence that the effect of exchange rate variability on FDI inflow is different in developed countries and in emerging economies. Therefore, the paper uses a unique combination of sample countries and time period and uses FDI inflow determinants that have empirically been proven to be statistically valid factors. Prior empirical studies focus either on a set of developed or developing countries and seem to pick the control variables randomly. Moreover, this thesis focusses explicitly on FDI inflows and therefore generates more representative findings. The results obtained in this paper explain part of the inconclusiveness found in previous studies and cluster the available information, which is an important step in providing incentive for further research.
2
Literature review
and employment in the non-affected countries to avoid experiencing a decline. This diversification of risk is highly valued by firms and, therefore, leads to new investments. Another mentioned advantage of increasing the FDI in the affected market is that companies can produce directly in the final market and expose only the inputs and intermediary goods to exchange rates, not the final product. However, this strand of literature has often been called outdated due to the fact that having more production facilities comes with higher fixed costs, and additional costs like updating transportation routes would constantly appear. This might consume all potential financial advantages. Phillips and Ahmadi-Esfahani (2008) reviewed the existing literature and estimate that 50% of the papers concerning the impact of exchange rate volatility on FDI find a negative correlation, less than 15% find a positive relationship, and the remaining papers find no significant impact. As a result of the majority of theoretical and empirical literature finding a negative impact on inward FDI, the following hypothesis is suggested: Hypothesis 1: Exchange rate volatility has a negative relationship with inward FDI. However, it seems that the empirical literature can be further split into studies focusing on developed and studies analyzing developing countries. Dal Bianco and To Loan (2017) note that most papers using a country sample consisting of developing countries seem to provide support for the hysteresis approach, while studies focusing on developed economies confirm the production flexibility theory. The three main explanations are presented below.
argument seems to hold for most of the countries examined. This means that the magnitude of the volatility might be the reason for the difference between developed and developing countries.
Second, since developed countries generally are also more open, they attract more FDI. Saini and Singhania (2016) empirically find that a one percent increase in openness leads to 1.9 percent more FDI inflows. Being more open also means that economic, political and financial connections are established and that the country becomes more integrated into the overall word developments, which helps to minimize shocks (Hau, 2002). This network might help in times of crisis and, therefore, portrays a positive image towards potential investors.
Third, developing countries are associated with further risks like contract enforcement, protection of property rights or political corruption. Additional risks, such as exchange rate volatility, therefore weigh more. Brown (2008) finds that most developing countries rely mainly on commodities, which are volatile themselves and therefore increase the effect exchange rate volatility has. Udoh and Egwaikhide (2008) state that high corruption, expropriation, transparency and other risks are significantly higher in developing countries than in developed economies. All of those reasons imply that companies investing in developing countries are more affected by uncertainty and are even more hesitant to invest since they might not recover their sunk costs.
Hypothesis 2: Being a developing country negatively affects the relationship between exchange rate volatility and inward FDI.
3
Data
!" (%&'
(,*,+) =
%&'./0123 (,*,+
years. In other words, the volatility of the exchange rates between two countries for year t was calculated based on the fluctuations of bilateral daily exchange rates calculated over the past three years. 9:!63<(,*,+ = 1 " − 1 × @(,*,+− A(,*,+ B , / +CD (2) where @ (,*,+ = ! " @E+ @E+FD . (3) @(*+ is the natural logarithm of the first change of the daily exchange rate (@E) between country i and j, A is the mean of those observations and " describes the size of the population. The daily exchange rates between all countries are provided by Bloomberg.
In some models, the change of the bilateral exchange rate level is included as a proxy for exchange rate volatility. This difference is squared since volatility is supposed not to be below zero. Furthermore, since exchange rate volatility is based on the behavior of the exchange rate, the level is included too in some models.
Naturally, in case country i and j both use the euro, the value for the exchange rate volatility takes the value of zero. Countries joining the Eurozone within the chosen time period are treated as if their national currency were replaced by the euro when the euro was introduced, even though the national currency could still be used and the country hence would have a volatility different from zero.
relatively stable currencies, which are used for trading, those currencies are therefore highly available. Moreover, they can mitigate shocks more easily and have better means of hedging, which enables them to counteract fluctuations, as opposed to small countries which are not part of a currency block (Clark et al., 2004).
Dev is a measure for development. As stated in hypothesis 2, development seems to be linked
Additionally, other factors regarding the host country often cited as important determinants of inward FDI have been added. Those variables are infrastructure, GDP growth, financial development and inflation (e.g. Gharaibeh, 2015). Baker et al. (2012) explain that physical infrastructure enables business and attracts more foreign investment. GDP change reflects economic activity and should be favorable for FDI (Cevis and Camurdan, 2007). Moreover, Dellis (2018) argues that although MNCs are not locally financially constrained, they interact significantly with the domestic financial systems. This explains why financially developed economies tend to attract more FDI than countries with a weak financial system. Lastly, high inflation results in macroeconomic instability and might therefore drive off FDI. Table (1): Summary statistics of all main variables This Table presents statistical characteristics of all main variables used in further analysis. All non-binary variables except variability measures are winsorized at a 1% level and independent variables are lagged by one year. The fewer observations of ln(FDIInf) compared to FDIInf can be explained by the fact that the natural logarithm of zero is computed as a missing value. The variables are defined as in Table (A3) in the Appendix.
Variable Obs Mean Std. Dev. Min Max
FDIInf 12,482 0.0035769 0.0576335 -1.067132 2.852518 Ln(FDIInf) 8,476 -8.387198 2.544059 -16.55909 1.048202 Vola 17,151 9.42399 12.46702 0 952.33 Dev 17,157 28500.14 20763.8 740.1143 102518.1 GDP 17,157 1.11E+12 2.24E+12 9.29E+09 1.45E+13 urban 17,157 73.90326 13.88798 29.235 97.546 open 17,157 85.10282 49.4325 22.15427 320.5983 tax 17,157 41.2636 12.60842 15 76.7 dist 17,157 5930.892 5153.127 215.6626 19006.65 comlang 17,157 0.0768286 0.2663263 0 1 colony 17,157 0.0320569 0.1761563 0 1 skilldiff2 17,157 942.1303 1224.371 0.1774263 5666.603 union 17,157 0.0525733 0.2231866 0 1 GDPCh 17,157 2.812443 3.747693 -14.81416 14.23139 infra 17,157 90.95314 31.24057 7.854585 161.172 infl 17,157 3.218297 2.597556 -0.923494 14.71492 finDev 16,923 83.42185 42.27656 12.6737 202.691
main independent variable, Vola, has a low mean relative to the maximum, which indicates that there are some large outliers. These outliers are caused by either developing countries or small nations. An overview of the main variables by country is provided in the Appendix.
Although the time period and the sample of countries was chosen based on data availability, the summary statistics in Table (1) shows that there are still several missing values in some variables. The main dependent variable, FDIInf, was corrected using mirror data. UNCTAD states that mirror data does not always reflect the actual amount since countries use different measures but also uses this method to construct datasets for countries that do not offer any official data. Following this approach, missing values for FDI inflows were substituted with the outflow level of the partner country, if available. However, missing values for FDI and other variables still appeared.1 Due to those missing values, the number of observations differs per estimate in the following analysis.
4
Methodology
Multivariate regressions are conducted to examine the relationship between FDI and exchange rate volatility. The regression model used for this analysis is the following:where Vola is the main variable of interest and used to analyze hypothesis 1. The remaining variables are defined as in Table (A3) the Appendix. As explained in the literature section above, volatility is expected to be negatively related to FDI inflows. Hence, following the hysteresis approach, high exchange rate volatility is expected to refrain risk averse companies from investing due to the uncertainty of future profitability. Therefore, beta 1 is expected to be negative. Furthermore, an interaction term of Vola and Dev is included to investigate the moderating effects of development on the main relationship. The full regression model is defined as follows:
ln (%&')S,T,U = VW + VD× 9:!6(,*,+FD + VB× &5H(,+FD
+ VY× 9:!6(,*,+FD × &5H(,+ + VZ× 7&8(,+FD + V\× J4[6"(,+FD+ V]× :I5"(,+FD + V_× K6^(,+FD+ V`× OLMK(,*,+ + Vd× a:b!6"c(,*,++ VDW× a:!:"<(,*,+ + VDD× MNL!!OLee2(,*,+FD + VDB× J"L:"(,*,+FD + VDY× 7&8fℎ(,+FD + VDZ× L"e46(,+FD + VD\× L"e!(,+FD+ VD]× eL"&5H(,+FD + %&'S,T,UFD+ h+ + i(,* + j(,*,+, (7)
where the interaction term VY indicates how economic development plays into the main
equation. Moreover, by using different model specifications, the results become more representative.
Furthermore, panel data allows controlling for unobservable variables such as cultural factors and variables that change over time but not across entities. To account for a potential omitted variable bias, fixed-effects (FE) or random-effects (RE) can be included. To see which model is more appropriate, a Hausman specification test HRE (1978) is conducted. The null
Table (2): Regression analysis of the full sample This Table shows the regression of (1) Volatility and development on the non-logarithmic FDI inflows; (2) Volatility and development on the logarithmic FDI inflows; (3) 3-year volatility and the binary development measure on the logarithmic FDI inflows; (4) Change of the yearly exchange rate level and development on the logarithmic FDI inflows; (5) Mean exchange rate and development on the logarithmic FDI inflows; (6) Volatility and determinants associated with FDI on the logarithmic FDI inflows. For all models, panel random effects and time fixed effects have been applied. All independent variables except binary time-invariant variables are lagged by one year. Moreover, control variables that correlate more than 50% with other variables have not been included in the model. The variables are defined as in Table (A3) in the Appendix.
(1) (2) (3) (4) (5) (6)
VARIABLES FDIInf Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf)
Vola -1.03e-05 -0.000571 -0.000697 (9.64e-06) (0.00218) (0.00219) Vola3y -0.00165*** (0.000214) ERmeanCh2 -7.1e-09*** (4.23e-10) ERmean -8.52e-08 (1.26e-05)
Dev 3.38e-07** 1.41e-05*** 1.41e-05*** 1.41e-05*** (1.55e-07) (2.53e-06) (2.53e-06) (2.54e-06)
DevBi 1.0035*** (0.191) GDP -0** -0*** -0*** -0*** -0*** -0*** (0) (0) (0) (0) (0) (0) tax -0.00021*** -0.0195*** -0.0359*** -0.01951*** -0.01948*** -0.0230*** (7.59e-05) (0.00393) (0.00466) (0.00393) (0.00393) (0.00415) dist -3.71e-07** -0.00012*** -0.00011*** -0.00012*** -0.00012*** -0.00011*** (1.85e-07) (1.25e-05) (1.25e-05) (1.25e-05) (1.25e-05) (1.23e-05) comlang 0.000289 1.379*** 1.336*** 1.3781*** 1.3782*** 1.294*** (0.00465) (0.240) (0.234) (0.2403) (0.2403) (0.223) colony 0.0108 0.560 0.533 0.5586 0.5587 0.608* (0.0129) (0.389) (0.392) (0.3897) (0.3896) (0.368) skilldiff2 2.52e-06 -1.14e-05 2.80e-05 -2.21e-05 -1.18e-05 -1.86e-05 (1.90e-06) (4.14e-05) (4.17e-05) (4.14e-05) (4.18e-05) (4.14e-05) union 0.00196 1.221*** 1.275*** 1.2223*** 1.2220*** 1.278*** (0.00325) (0.237) (0.233) (0.237) (0.237) (0.224) GDPCh 0.000123 0.0111* 0.0124* 0.01113* 0.0112* 0.0112 (0.000107) (0.00654) (0.00750) (0.0065) (0.00656) (0.00751) urban 0.0202*** 0.0212*** (0.00454) (0.00445) infra 0.00239 0.00255 (0.00159) (0.00155) finDev -0.00189 0.000815 (0.00119) (0.00116) infl 0.000255** -0.00527 -0.00558 -0.00555 -0.00926 (0.000129) (0.0108) (0.0107) (0.0107) (0.0106) open 0.00922*** (0.00114) Constant 0.00751 -7.7901*** -9.200*** -7.7938*** -7.7921*** -9.975*** (0.00587) (0.230) (0.393) (0.2295) (0.2294) (0.416) Obs. 11,578 7,900 7,811 7,900 7,900 7,811 Panels 1,284 1,220 1,220 1,220 1,220 1,220
Year FE yes yes yes yes yes yes
Panel RE yes yes yes yes yes yes
effect of exchange rate volatility on the FDI inflows of developed countries. Overall, the exchange rate volatility of developing economies seems to have a more negative impact on FDI inflows than volatility has on developed economies, as suggested by hypothesis 2. The effects of all control variables broadly remain unchanged compared to Table (2). Table (3): Interaction terms This Table shows the regression of (1) Volatility, development and an interaction of volatility and development on the logarithmic FDI inflows; (2) Volatility, a binary development measure and the interaction of volatility and a binary development measure on the logarithmic FDI inflows; (3) 3-year volatility, development and the interaction of 3-year volatility and development on the logarithmic FDI inflows; (4) 3-year volatility, a binary development measure and the interaction of 3-year volatility and a binary development measure on the logarithmic FDI inflows; (5) Change of the yearly exchange rate level and development on the logarithmic FDI inflows. The x between two variables denotes interaction terms. For all models, panel random effects and time fixed effects have been applied. All independent variables except binary time-invariant variables are lagged by one year. Moreover, control variables that correlate more than 50% with other variables have not been included in the model. The variables are defined as in Table (A3) in the Appendix. (1) (2) (3) (4) (5)
VARIABLES FDIInf Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf)
Vola -0.00849** -0.0114** (0.00417) (0.00457) Vola3y -0.00219*** -0.00152*** (0.000553) (0.000147) ERmeanCh2 -2.92e-08* (1.64e-08)
Dev 5.53e-06** 8.04e-06*** 1.41e-05***
(2.68e-06) (2.56e-06) (2.53e-06)
(0.00118) (0.00122) (0.00118) (0.00122) union 1.229*** 1.283*** 1.219*** 1.279*** 1.221*** (0.229) (0.225) (0.229) (0.225) (0.237) GDPCh 0.00555 0.0108 0.00699 0.0124* 0.0112* (0.00655) (0.00745) (0.00659) (0.00751) (0.00654) finDev -0.000171 -0.000096 (0.00121) (0.00121) infl -0.00632 -0.00879 -0.00587 (0.0108) (0.0107) (0.0107) Constant -8.447*** -9.789*** -8.486*** -9.920*** -7.787*** (0.249) (0.412) (0.248) (0.406) (0.229) Observations 7,900 7,811 7,900 7,811 7,900 Panels 1,220 1,220 1,221 1,220 1,220
Year FE yes yes yes yes yes
Panel RE yes yes yes yes yes
Table (4): Regression analysis of emerging economies
This Table shows the regression of (1) Volatility and development on the non-logarithmic FDI inflows; (2) 3-year volatility and determinants associated with FDI on the logarithmic FDI inflows; (3) 3-year volatility and development on the logarithmic FDI inflows; (4) Mean exchange rate and development on the logarithmic FDI inflows; (5) Volatility, development and an interaction of volatility and development on the logarithmic FDI inflows; (6) Volatility and determinants associated with FDI on the logarithmic FDI inflows; (7) Change of the yearly exchange rate level and development on the logarithmic FDI inflows. For all models, panel random effects and time fixed effects have been applied. All independent variables except binary time-invariant variables are lagged by one year. Moreover, control variables that correlate more than 50% with other variables have not been included in the model. The variables are defined as in Table (A3) in the Appendix. (1) (2) (3) (4) (5) (6) (7)
VARIABLES FDIInf Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Vola -0.0103** -0.0299*** -0.0112** (0.00518) (0.00794) (0.00505) Vola3y -0.0018*** -0.0017*** (0.000135) (0.000135) ERmeanCh2 -8.4e-09*** (7.98e-10) ERmean 2.17e-05* (1.18e-05)
Dev 4.45e-05** 4.62e-05** 4.67e-05** 2.37e-05 3.85e-05** (1.95e-05) (1.95e-05) (1.95e-05) (1.95e-05) (1.82e-05)
VolaDev 2.4e-06*** (6.14e-07) GDP -0*** -0*** -0*** -0*** -0*** -0*** -0*** (0) (0) (0) (0) (0) (0) (0) tax -0.0380** -0.0574*** -0.0363** -0.0366** -0.0380** -0.0586*** -0.0383** (0.0164) (0.0198) (0.0164) (0.0164) (0.0165) (0.0197) (0.0163) dist -6.38e-06 -5.36e-05 -7.06e-06 -5.23e-06 -4.05e-06 -5.26e-05 -1.33e-05 (4.10e-05) (3.94e-05) (4.12e-05) (4.11e-05) (4.14e-05) (3.94e-05) (4.07e-05) comlang -0.0906 0.274 -0.0498 -0.028 -0.0917 0.236 -0.0229 (0.747) (0.712) (0.746) (0.745) (0.752) (0.709) (0.739) colony 2.415*** 2.018** 2.403*** 2.397*** 2.408*** 2.026** 2.369*** (0.782) (0.798) (0.789) (0.790) (0.789) (0.789) (0.763) skilldiff2 -4.27e-05 -5.55e-05 -4.97e-05 -5.69e-05 -5.45e-05 -4.70e-05 -5.48e-05 (7.01e-05) (7.06e-05) (6.98e-05) (7.19e-05) (6.98e-05) (7.08e-05) (7.13e-05) open -0.00633 0.000844 -0.00754 -0.0072 -0.00825 0.00185 (0.00530) (0.00692) (0.00531) (0.0053) (0.00527) (0.00687) union 2.318 2.613 2.338 2.383 2.334 2.586 2.384 (2.196) (2.137) (2.220) (2.222) (2.202) (2.109) (2.164) GDPCh 0.0146 0.0265* 0.0190 0.0196 0.00824 0.0219 0.0192 (0.0154) (0.0159) (0.0156) (0.0155) (0.0151) (0.0158) (0.0155) infl 0.0266 0.0249 0.0215 0.0213 0.0393** 0.0309 0.0123 (0.0191) (0.0187) (0.0188) (0.0188) (0.0194) (0.0192) (0.0184) finDev 0.00647 0.00644 (0.00565) (0.00564) infra 0.0203*** 0.0199*** (0.00559) (0.00569) urban 0.00710 0.00752 (0.00939) (0.00943) Constant -8.357*** -8.963*** -8.456*** -8.514*** -8.107*** -8.886*** -8.636*** (0.790) (1.079) (0.783) (0.782) (0.795) (1.074) (0.773) Obs. 1,639 1,639 1,639 1,639 1,639 1,639 1,639 Panels 232 232 232 232 232 232 232
Year FE yes yes yes yes yes yes yes Panel RE yes yes yes yes yes yes yes
Table (5): Regression analysis of developed economies
This Table shows the regression of (1) Volatility and development on the non-logarithmic FDI inflows; (2) 3-year volatility and determinants associated with FDI on the logarithmic FDI inflows; (3) 3-year volatility and development on the logarithmic FDI inflows; (4) Mean exchange rate and development on the logarithmic FDI inflows; (5) Volatility, development and an interaction of volatility and development on the logarithmic FDI inflows; (6) Volatility and determinants associated with FDI on the logarithmic FDI inflows; (7) Change of the yearly exchange rate level and development on the logarithmic FDI inflows. For all models, panel random effects and time fixed effects have been applied. All independent variables except binary time-invariant variables are lagged by one year. Moreover, control variables that correlate more than 50% with other variables have not been included in the model. The variables are defined as in Table (A3) in the Appendix. (1) (2) (3) (4) (5) (6) (7)
VARIABLES FDIInf Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Ln(FDIInf) Vola 0.000294 -0.00424 0.000271 (0.00150) (0.00448) (0.00152) Vola3y -0.00173* -0.00163* (0.00102) (0.000988) ERmeanCh2 6.99e-07 (3.21e-06) ERmean -0.00025 (0.000263)
Dev 8.0e-06*** 8.1e-06*** 8.1e-06*** 6.42e-06** 1.3e-05*** (2.90e-06) (2.90e-06) (2.91e-06) (3.00e-06) (3.00e-06)
VolaDev 1.94e-07 (1.75e-07) GDP -0*** -0*** -0*** -0*** -0*** -0*** -0*** (0) (0) (0) (0) (0) (0) (0) tax -0.0201*** -0.0243*** -0.0202*** -0.0201*** -0.0200*** -0.0241*** -0.0236*** (0.00444) (0.00489) (0.00444) (0.00444) (0.00444) (0.00490) (0.0045) dist -0.0001*** -0.0001*** -0.0001*** -0.0001*** -0.0001*** -0.0001*** -0.0001*** (1.32e-05) (1.31e-05) (1.32e-05) (1.34e-05) (1.33e-05) (1.31e-05) (1.34e-05) comlang 1.482*** 1.413*** 1.477*** 1.467*** 1.491*** 1.420*** 1.524*** (0.224) (0.226) (0.224) (0.2242) (0.224) (0.226) (0.239) colony 0.421 0.418 0.426 0.4217 0.417 0.412 0.3179 (0.404) (0.403) (0.404) (0.404) (0.405) (0.403) (0.433) skilldiff2 -5.53e-06 4.86e-06 -6.11e-06 -3.83e-06 -3.45e-06 5.54e-06 -3.73e-05 (5.06e-05) (5.08e-05) (5.05e-05) (5.07e-05) (5.03e-05) (5.09e-05) (5.02e-05) open 0.00921*** 0.0102*** 0.00919*** 0.00921*** 0.00929*** 0.0103*** (0.00125) (0.00131) (0.00125) (0.00125) (0.00125) (0.00130) union 1.090*** 1.154*** 1.087*** 1.0932*** 1.092*** 1.156*** 1.0913*** (0.229) (0.225) (0.228) (0.2282) (0.229) (0.225) (0.237) GDPCh 0.00883 0.0140 0.00904 0.00894 0.00805 0.0137 0.01437* (0.00746) (0.00901) (0.00745) (0.00745) (0.00748) (0.00901) (0.00740) infl 0.0131 0.0164 0.0137 0.01372 0.0124 0.0158 0.0133 (0.0137) (0.0140) (0.0138) (0.0137) (0.0137) (0.0139) (0.0137) finDev 0.0005 0.000479 (0.00139) (0.00139) infra -0.00355* -0.00354* (0.00183) (0.00183) urban 0.0149*** 0.0147*** (0.00560) (0.00560) Constant -8.542*** -9.180*** -8.519*** -8.542*** -8.513*** -9.188*** -7.557*** (0.294) (0.545) (0.293) (0.294) (0.294) (0.546) (0.296) Obs. 6,262 6,172 6,261 6,262 6,262 6,172 6,262 Panels 988 988 988 988 989 988 988
Year FE yes yes yes yes yes yes yes Panel RE yes yes yes yes yes yes yes
The first possible explanation, as discussed by Hausmann et al. (2004), is that exchange rates are about three times more volatile in developing countries than they are in developed economies. To test whether high volatility has a greater impact on FDI than low volatility does, new dummies are generated at the median, the 75th empirical quantile and the 90th empirical quantile. The dummy and the interaction term of the dummy with Vola are added as additional variables. In general, all dummies for high volatility have negative effects of FDI inflows. Since only the first interaction term is significant, the results in Table (A4) in the Appendix indicate that high versus low volatility does not explain the difference between developed and developing countries. In other words, the effect of Vola does not seem to depend on the level of Vola.
does in developing ones. However, due to a lack of feasibility, this hypothesis is left for further research.
openness or the additional risk could be the drivers of this effect. Regarding the openness, companies face more challenges operating in the market and might need more resources to set up the facility, but also to keep operating due to trade barriers. Also, the additional risks result in costs as more attention needs to be paid to security, changing market conditions and political influence. Consequently, many companies decide not to invest more into already established markets or even leave the market. Moreover, developing markets are not as tapped and first-time investments can easily be delayed, as the company waits for better conditions or opportunities in other countries.
Contrary, developed countries are not as affected by exchange rate variability. In other words, companies decide to invest in developed companies even if the exchange rate is high. One reason might be, that overall, investments in developed countries are higher and that companies do not want their investments to fail, therefore, they would rather take the losses instead of divesting. This also goes with the fact that developed countries tend to have better measures in place and return to a stabilized situation much faster than developing economies can and rough times are seen only as temporary.
6.1 Industry-specific determinants of exposure
A company’s industry gives insights into the expected exchange rate exposure. For instance, the amount of sunk costs or the time span of its projects determines the flexibility.
Acting in industries with high sunk costs, such as the energy sector, requires high initial investments. Consequently, those companies have more to lose in case they have to leave the market again, since setting uo the infrastructure is the biggest investment. Based on this argument, Dixit and Pindyck (1994) developed the hysteresis approach. Campa (1993) builds on that and shows that industries facing high sunk costs are more risk averse and reluctant to invest in countries with highly volatile exchange rates due to uncertainty avoidance.
MNCs to overcome financial constraints that handicap local firms. This is explained by the access to a variety of markets, which results in lower cost of capital.
Moreover, having monopoly power might protect companies from the damages volatile exchange rates cause. Few competitors and high market concentration usually encourage companies to ask for high markups. This additional income enables the company to slim its margin and, therefore, increase output. Campa and Goldberg (1999) prove this empirically but also states that in some countries, companies with monopoly power rather decrease the amount of goods sold and that the elasticity of consumer demand rises with volatility. The percentage of imported resources needed for production determines the exchange rate risk a company faces, according to Campa and Goldberg (1999). They find that importing a large share from a country using another currency, subsequently leads to more unsecure profits. Additionally, it depends on whether the foreign production facility serves the local market or whether the finalized products are exported to other countries. In case the products stay in the market, only the cheaper inputs are subject to exchange rate volatility. Under these circumstances, exchange rate volatility should increase FDI (Dal Bianco and To Loan, 2017) while exporting the product leads to further uncertainties.
7
Conclusion and future research
This paper tests the hypotheses that exchange rate volatility has a negative effect on FDI and that the relationship becomes more significant when emerging economies are involved. These arguments are based on an extensive literature review, which shows that there is still a lot of ambiguity. Although the clear majority of prior theoretical and empirical literature argues in favor of a negative relationship, some papers have found no significant effect or even a positive correlation. To test the hypotheses, multivariate regressions are applied. Next, the sample countries are separated into developed and developing countries, as proposed by literature. This differentiation proves to be important and explains part of the ambiguity found in literature.The overall rather negative relation between FDI and exchange rate variability found across the full sample is not surprising given the fact that currency unions would otherwise not have emerged. The result implicates that countries with high exchange rate volatility get deprived of FDI inflows into their country. The main arguments supporting this finding are that FDI per definition requires an initial investment and that factors potentially endangering those investments drive off FDI.
As a next step, interaction terms between exchange rate volatility and development are used to test the hypothesis that the distinction between developed and developing countries provides further insights. The results indicate that such a differentiation might be important, which is why the sample was later split into developed and developing countries.
economies or the fact that volatility is seen as an additional source of uncertainty. The magnitude of exchange rate volatility does not seem to explain the difference between developed and developing economies. For very high degrees of openness, a significant interaction term is found, indicating that the openness of developed countries is a potential moderator. The high correlation of additional risks and development could have led to multicollinearity issues, which is why no regression was conducted. However, this confirms the results found in prior research and might indicate that the difference in uncertainty other than exchange rate volatility determines the relationship between FDI and exchange rate volatility. The analysis was redone using the inverted hyperbolic sine distribution to account for the fact that by using the logarithm of negative FDI inflows or those that have the value of zero, those observations drop out of the sample. Although the results are not statistically significant, the analysis also suggests a negative trend, which supports the results found. Still, it is possible that the results are valid only for a restricted subsample, in which the FDI inflows have positive values. An additional factor requiring further analysis is the differentiation between long- and short-term volatility, since it seems important for the investment decision of developed countries. The results clearly show that especially countries historically known for their unpredictable exchange rates are deprived of FDI since companies prefer certain profits. Countries that show high volatility in the short-term are not as affected.
Future research could also analyze FDI distinguished in its many forms. For instance, since some countries require companies to enter via joint ventures with a locally established firm, the amount of FDI inflows might not reflect the effect of the exchange rate properly.
Lastly, quantifying the moderating and strengthening effects discussed for industry- and company-specific determinants could be analyzed in more depth. The paper shows that factors like the cost of the initial investment, the time-span of a project, foreign subsidiaries, monopoly power and a company’s sourcing and selling decisions also have to be considered when classifying the impact exchange rate volatility has on companies. This is because exchange rate volatility affects a company’s break-even point, profitability margin and volume of production and sales. So far, those issues have mainly been discussed in theoretical literature but could provide important managerial implications as companies could estimate their exposure more precisely and, therefore, react more appropriately. Moreover, this paper leads to important policy implications. It shows that developing countries are more easily deprived of FDI when exchange rate volatility increases. This finding could be important for institutions trying to facilitate the development in emerging economies. It also proves that emerging economies could attract more FDI inflows by reducing its volatility, opening its borders or reducing other risks. The direct way would be to stabilize the volatile currency, but also indirect channels such as opening the economy or reducing other risks seem to have similar effects. Those measures have often been proposed to induce economic development and this paper proposes that they also influence each other. By concentrating on openness or reducing risks like corruption, the effects of exchange rate volatility would be reduced without specifically targeting this issue. This would lead to more FDI inflow, hence help the country to become developed and, therefore, attract more investments.
Given the nature of exchange rate variability and the fact that a company’s FDI decisions are based on multiple factors, just like the availability of natural resources or nontransferable knowledge, it is unlikely that the two factors of exchange rate volatility and development can solely explain a firm’s business decisions. However, this paper proved that both factors are important determinants.
9
Appendix
Table (A1): Summary statistics by country This table presents the mean values of the most important variables used in the analysis separately for each country. The countries are the OECD member states and the BRICS countries expect Russia. The BRICS have been added in order to get a more representative sample of emerging countries. The variables are defined as in Table (A3) in the Appendix.Table (A2): Correlation matrix
The Table presents the correlations of all the variables with each other. The variables are defined as in Table (A3) in the Appendix.
FDIInf Ln(FDIInf) Vola Vola3y ERmean ERmeanCh2 Dev DevBi GDP urban open
Continued Table (A2): Correlation matrix
tax dist comlang colony skilldiff2 union GDPCh infra infl finDev
Table (A3): Definition of variables
This table presents the variables used, their definition and calculation. Moreover, the source is presented in order to make replication possible.
Variable - short Variable - long Calculation/Definition Source FDIInf FDI inflows FDI inflows into country i from country j
in year t divided by country i’s GDP
UNCTAD Vola Short-term volatility
in percent Volatility is measured using the annualized standard deviation of day to day logarithmic historical price changes of the exchange rate between the currencies of country I and j within year t Bloomberg Vola3y Long-term volatility in percent Volatility is measured using the annualized standard deviation of day to day logarithmic historical price changes of the exchange rate between the currencies of country I and j over the years t, t-1 and t-2. Bloomberg REER Real effective exchange rate A country’s currency with constant trade-weights to a basket of other countries’ currencies and relative trade balances (2005=100) OECD ERmeanCh2 Change of the
exchange rate Squared difference of the yearly mean of the real exchange rate from t-1 to t. Bloomberg ERmean Exchange rate Mean of daily bilateral exchange rate
levels between two countries I and j in year t
Bloomberg
Dev Development GDP divided by population of country I in
year t UNCTAD
DevBi Development
(binary) Value of one if the country i is considered a developed country
(according to the definitions provided by UNCTAD)
UNCTAD
GDP GDP GDP of country I in year t World Bank urban Urban population Percentage of population living in urban
areas in country I in year t
OECD open Trade openness The sum of exports and imports of goods
and services measured as a share of gross domestic product of country I in year t
World Bank
tax Corporate tax rate Amount of taxes and mandatory
contributions payable by businesses after accounting for allowable deductions and exemptions in country I in year t OECD and official government statements of the respective country dist Distance Distance between the most populated
cities of country I and j Cepii Mayer, T. & Zignago, S. (2011) comlang Common language Value of 1 if country I and j share a
colony Colonial
relationship Value of 1 if country I and j ever had a colonial relationship Cepii skilldiff2 Squared skill
difference Squared skill difference of country I and j measured as percentage of tertiary level enrollment rate in year t
OECD
union Trade agreement or
union Value of 1 if country i and j are in a trade block Cepii GDPCh GDP Change Change of GDP of country I between year
t-1 and t in percent
World Bank infra Infrastructure Mobile cellular subscriptions per 100
people registered in country I in year t OECD infl Inflation Change of inflation in country I between
year t-1 and t in percent World Bank finDev Financial
development Financial resources provided to the private sector by financial corporations, measured as a percentage of GDP of country I in year t
World Bank
comcur Common currency Value of 1 if country I and j use the same currency in year t
Cepii
Table (A4): Dummies for different levels of volatility This Table shows the regression of (1) Volatility, the dummy of volatility being higher than the mean and the interaction between the two; (2) Volatility, the dummy of volatility being higher than the 75 percent quantile and the interaction between the two; (3) Volatility, the dummy of volatility being higher than the 90 percent quantile and the interaction between the two. For all models, panel random effects and time fixed effects have been applied. All independent variables except binary time-invariant variables are lagged by one year. Moreover, control variables that correlate more than 50% with other variables have not been included in the model. The variables are defined as in Table (A3) in the Appendix. (1) (2) (3)
VARIABLES Ln(FDIInf) Ln(FDIInf) Ln(FDIInf)
Vola 0.0138* -0.00727 -0.00452 (0.00738) (0.00533) (0.00442) Vola50 -0.109* (0.0599) VolaVola50 -0.0136* (0.00760) Vola75 -0.148** (0.0685) VolaVola75 0.00759 (0.00556) Vola90 -0.411*** (0.0857) VolaVola90 0.00625 (0.00442) GDP -0*** -0*** -0*** (0) (0) (0) tax -0.0232*** -0.0232*** -0.0229*** (0.00412) (0.00414) (0.00414) dist -0.000103*** -0.000104*** -0.000103***
(1.23e-05) (1.24e-05) (1.24e-05)
comlang 1.268*** 1.283*** 1.273***
(0.222) (0.222) (0.222)
colony 0.612* 0.624* 0.627*
(0.367) (0.367) (0.366)
skilldiff2 -1.44e-05 -1.89e-05 -1.70e-05
(4.14e-05) (4.14e-05) (4.14e-05)
open 0.00924*** 0.00919*** 0.00942*** (0.00114) (0.00114) (0.00114) union 1.279*** 1.272*** 1.276*** (0.224) (0.223) (0.223) GDPCh 0.0125* 0.0117 0.0114 (0.00753) (0.00752) (0.00754) infl -0.00755 -0.00787 -0.00165 (0.0105) (0.0105) (0.0105) finDev 0.000894 0.000843 0.000797 (0.00116) (0.00116) (0.00116) infra 0.00255* 0.00254 0.00213 (0.00155) (0.00155) (0.00155) urban 0.0211*** 0.0212*** 0.0216*** (0.00443) (0.00444) (0.00446) Constant -9.970*** -9.929*** -10.00*** (0.417) (0.418) (0.418) Observations 7,810 7,810 7,810 Panels 1,220 1,220 1,220
Year FE yes yes yes
Panel RE yes yes yes
Table (A5): Dummies for different levels of openness
This Table shows the regression of (1) Volatility, the dummy of openness being higher than the mean and the interaction between the two; (2) Volatility, the dummy of openness being higher than the 75 percent quantile and the interaction between the two; (3) Volatility, the dummy of openness being higher than the 90 percent quantile and the interaction between the two. For all models, panel random effects and time fixed effects have been applied. All independent variables except binary time-invariant variables are lagged by one year. Moreover, control variables that correlate more than 50% with other variables have not been included in the model. The variables are defined as in Table (A3) in the Appendix.
(1) (2) (3)
VARIABLES Ln(FDIInf) Ln(FDIInf) Ln(FDIInf)
Vola -0.000669 -0.000812 -0.00104 (0.00224) (0.00231) (0.00247) open 0.00943*** 0.0101*** 0.00909*** (0.00124) (0.00127) (0.00125) open50 -0.0287 (0.0973) Volaopen50 -0.000597 (0.00554) open75 -0.170* (0.102) Volaopen75 0.00339 (0.00774) open90 -0.195 (0.127) Volaopen90 0.0310*** (0.0113) GDP -0*** -0*** -0*** (0) (0) (0) tax -0.0228*** -0.0226*** -0.0233*** (0.00416) (0.00418) (0.00415) dist -0.000108*** -0.000108*** -0.000111***
(1.24e-05) (1.24e-05) (1.24e-05)
comlang 1.291*** 1.296*** 1.307***
(0.223) (0.223) (0.222)
colony 0.608* 0.608* 0.603
(0.368) (0.369) (0.371)
skilldiff2 -1.93e-05 -2.40e-05 -1.65e-05
(4.16e-05) (4.16e-05) (4.14e-05)
union 1.279*** 1.274*** 1.277*** (0.224) (0.225) (0.225) GDPCh 0.0113 0.0104 0.0126* (0.00751) (0.00756) (0.00752) finDev 0.000845 0.000789 0.000737 (0.00116) (0.00116) (0.00116) infra 0.00251 0.00297* 0.00255 (0.00156) (0.00156) (0.00156) urban 0.0211*** 0.0201*** 0.0214*** (0.00447) (0.00450) (0.00445) infl -0.00888 -0.00920 -0.00818 (0.0106) (0.0106) (0.0106) Constant -9.978*** -9.973*** -9.949*** (0.418) (0.417) (0.417) Observations 7,810 7,810 7,810 Panels 1,220 1,220 1,220
Year FE yes yes yes
Panel RE yes yes yes
Table (A6): Inverted hyperbolic sine distribution of the FDI inflows into developed countries
This Table shows the regression of (1) Volatility and development on the non-transformed FDI inflows; (2) 3-year volatility and determinants associated with FDI on the transformed FDI inflows; (3) 3-year volatility and development on the transformed FDI inflows; (4) mean exchange rate and development on the transformed FDI inflows; (5) Volatility, development and an interaction of volatility and development on the transformed FDI inflows; (6) Volatility and determinants associated with FDI on the transformed FDI inflows; (7) change of the yearly exchange rate level and development on the transformed FDI inflows. For all models, panel random effects and time fixed effects have been applied. All independent variables except binary time-invariant variables are lagged by one year. Moreover, control variables that correlate more than 50% with other variables have not been included in the model. The variables are defined as in Table (A3) in the Appendix. (1) (2) (3) (4) (5) (6) (7)
VARIABLES FDIInf IHS_FDIInf IHS_FDIInf IHS_FDIInf IHS_FDIInf IHS_FDIInf IHS_FDIInf Vola 5.05e-07 -0.000194 -8.01e-06 (1.22e-05) (0.000175) (1.08e-05)
Vola3y -2.05e-06 -1.03e-06
(6.50e-06) (6.32e-06)
ERmeanCh2 -9.50e-10
(2.59e-08)
ERmean -4.57e-06
(3.17e-06)
Dev 2.44e-07* 2.23e-07** 2.24e-07** 1.66e-07* 3.39e-07** (1.29e-07) (1.04e-07) (1.04e-07) (9.67e-08) (1.39e-07)
VolaDev 7.24e-09
(6.73e-09)
GDP 0* 0** 0** 0** 0** 0** -0**
(0) (0) (0) (0) (0) (0) (0)
tax -1.63e-05 -3.17e-05 -1.41e-05 -1.14e-05 -8.73e-06 -3.14e-05 -0.00013** (3.33e-05) (3.98e-05) (2.55e-05) (2.59e-05) (2.51e-05) (3.97e-05) (5.36e-05) dist -1.03e-07 -1.53e-07 -1.02e-07 -7.10e-08 -1.39e-07 -1.51e-07 -4.06e-07** (1.59e-07) (1.42e-07) (1.33e-07) (1.42e-07) (1.22e-07) (1.42e-07) (1.63e-07) comlang 0.000224 0.000954 0.00121 0.000982 0.00146 0.000953 0.00146 (0.00480) (0.00432) (0.00428) (0.00434) (0.00429) (0.00431) (0.00458) colony 0.0147 0.0122 0.0126 0.0126 0.0125 0.0122 0.0108 (0.0141) (0.0127) (0.0126) (0.0126) (0.0127) (0.0127) (0.0131) skilldiff2 2.19e-06 1.70e-06 1.46e-06 1.50e-06 1.52e-06 1.70e-06 2.39e-06 (2.24e-06) (1.82e-06) (1.71e-06) (1.73e-06) (1.75e-06) (1.82e-06) (1.96e-06) open 0.0002*** 0.00021*** 0.00017*** 0.00017*** 0.00017*** 0.00021*** (7.44e-05) (7.17e-05) (6.02e-05) (6.05e-05) (6.14e-05) (7.18e-05) union 0.00101 0.00250 0.00150 0.00155 0.00159 0.00249 0.00200 (0.00307) (0.00277) (0.00285) (0.00284) (0.00289) (0.00277) (0.00314) GDPCh -0.000148 -7.08e-05 -0.000124 -0.000124 -0.000155 -7.08e-05 4.83e-05 (0.000152) (0.000126) (0.000123) (0.000123) (0.000140) (0.000126) (0.000100) infl 0.000147 1.01e-05 0.000150 0.000149 0.000106 1.16e-05 0.000317** (0.000117) (9.95e-05) (0.000101) (0.000101) (8.97e-05) (9.85e-05) (0.000143)
finDev 3.86e-05** 3.86e-05**
(1.65e-05) (1.65e-05)
infra -4.73e-06 -4.72e-06
(2.77e-05) (2.77e-05) urban 0.00024*** 0.00024*** (7.78e-05) (7.75e-05) Constant -0.0172* -0.0336*** -0.0155** -0.0157** -0.0142** -0.0336*** 0.00358 (0.00916) (0.0116) (0.00720) (0.00728) (0.00681) (0.0116) (0.00483) Obs. 9,413 9,260 9,412 9,412 9,413 9,261 9,412 Panels 1,039 1,038 1,038 1,038 1,039 1,039 1,038 Year FE yes yes yes yes yes yes yes Panel RE yes yes yes yes yes yes yes
Table (A7): Inverted hyperbolic sine distribution of the FDI inflows into developed countries
This Table shows the regression of (1) Volatility and development on the non-transformed FDI inflows; (2) 3-year volatility and determinants associated with FDI on the transformed FDI inflows; (3) 3-year volatility and development on the transformed FDI inflows; (4) mean exchange rate and development on the transformed FDI inflows; (5) Volatility, development and an interaction of volatility and development on the transformed FDI inflows; (6) Volatility and determinants associated with FDI on the transformed FDI inflows; (7) change of the yearly exchange rate level and development on the transformed FDI inflows. For all models, panel random effects and time fixed effects have been applied. All independent variables except binary time-invariant variables are lagged by one year. Moreover, control variables that correlate more than 50% with other variables have not been included in the model. The variables are defined as in Table (A3) in the Appendix. (1) (2) (3) (4) (5) (6) (7)
VARIABLES FDIInf IHS_FDIInf IHS_FDIInf IHS_FDIInf IHS_FDIInf IHS_FDIInf IHS_FDIInf Vola -1.58e-07 -2.94e-06** -2.93e-07 (1.40e-06) (1.39e-06) (1.38e-06)
Vola3y -7.19e-08 -2.56e-08
(1.13e-07) (9.24e-08)
ERmeanCh2 -0
(0)
ERmean 7.92e-09
(7.49e-09)
Dev 3.70e-08** 3.70e-08** 3.73e-08** 3.01e-08* 3.20e-08** (1.68e-08) (1.66e-08) (1.67e-08) (1.57e-08) (1.51e-08)
VolaDev 7.01e-10**
(3.57e-10)
GDP -0* -0* -0* -0* -0* -0* -0*
(0) (0) (0) (0) (0) (0) (0)
tax -1.22e-05 -1.67e-05 -1.22e-05 -1.21e-05 -1.20e-05 -1.68e-05 -1.38e-05 (1.03e-05) (1.68e-05) (1.02e-05) (1.02e-05) (1.03e-05) (1.68e-05) (1.01e-05) dist -4.39e-09 -1.73e-08 -4.33e-09 -3.74e-09 -4.25e-09 -1.73e-08 -8.31e-09 (2.36e-08) (2.05e-08) (2.35e-08) (2.35e-08) (2.37e-08) (2.06e-08) (2.27e-08) comlang 0.000689 0.000740 0.000690 0.000697 0.000697 0.000738 0.000703 (0.000492) (0.000484) (0.000491) (0.000492) (0.000493) (0.000484) (0.000492) colony 0.000607 0.000538 0.000607 0.000605 0.000604 0.000539 0.000587 (0.00129) (0.00131) (0.00129) (0.00129) (0.00129) (0.00131) (0.00127) skilldiff2 -8.74e-08** -9.23e-08* -8.76e-08** -9.06e-08** -9.05e-08** -9.18e-08* -9.27e-08** (4.34e-08) (5.03e-08) (4.30e-08) (4.41e-08) (4.41e-08) (5.07e-08) (4.40e-08) open -4.49e-06 -2.20e-06 -4.52e-06 -4.47e-06 -5.10e-06 -2.14e-06 (3.54e-06) (6.45e-06) (3.44e-06) (3.44e-06) (3.69e-06) (6.40e-06) union 0.00552 0.00567 0.00552 0.00554 0.00553 0.00567 0.00555 (0.00401) (0.00404) (0.00401) (0.00401) (0.00401) (0.00404) (0.00397) GDPCh 2.43e-05** 2.15e-05** 2.45e-05** 2.45e-05** 2.34e-05** 2.12e-05* 2.41e-05** (1.09e-05) (1.05e-05) (1.02e-05) (1.01e-05) (1.07e-05) (1.14e-05) (1.00e-05) infl 3.29e-05** 3.26e-05* 3.28e-05** 3.28e-05** 3.46e-05** 3.28e-05* 2.74e-05* (1.57e-05) (1.77e-05) (1.57e-05) (1.57e-05) (1.56e-05) (1.80e-05) (1.44e-05)
finDev 4.89e-06 4.86e-06
(5.81e-06) (5.79e-06)
infra 8.82e-06 8.81e-06
(7.84e-06) (7.83e-06)
urban 8.64e-07 8.51e-07
(8.04e-06) (8.02e-06)
Constant 0.000769* 0.000569 0.000766* 0.000753* 0.000813* 0.000576 0.000698 (0.000444) (0.000927) (0.000437) (0.000435) (0.000442) (0.000915) (0.000432) Obs 2,165 2,165 2,165 2,165 2,165 2,165 2,165
Panels 245 245 245 245 245 245 245
Year FE yes yes yes yes yes yes yes Panel RE yes yes yes yes yes yes yes
