The Effects of Common Currencies on Trade

Transcription

1 The Effects of Common Currencies on Trade Countries select particular exchange rate arrangements for a variety of reasons. The ability to conduct an independent monetary policy is often cited as the main advantage of having a floating exchange rate regime. Countries sometimes tie their exchange rate to that of a larger country to benefit from the relative stability of the foreign currency. A more prevalent rationale for adopting fixed exchange rates or even common-currency arrangements, however, is the notion that exchange rate volatility introduces uncertainty into cross-border transactions, reducing the volume of trade that would otherwise take place. 1-1

2 The Effects of Common Currencies on Trade Fixing the exchange rate eliminates uncertainty, and so promotes trade. Adopting a neighbor s currency as one s own is a credible commitment to exchange rate stability and has the extra advantage of eliminating transactions costs; Both effects may give extra encouragement to trade. Studies often find that the effect of exchange rate volatility on trade is positive (Brada and Mendez, 1988) or statistically insignificant (Belanger, Gutierrez, and Raynauld, 1992). Recent theoretical work also suggests that the negative association between exchange rate volatility and trade may not hold in the context of general equilibrium or for all forms of risk aversion (Bacchetta and van Wincoop, 2000, and De Grauwe, 1988). 1-2

3 The Effects of Common Currencies on Trade In a drastic departure from past empirical studies that fail to find a significant link between exchange rate stability and trade, a recent set of papers by Andrew Rose and his colleagues find extremely large positive effects of common currencies on the volume of trade. The most dramatic, and widely cited, of his findings is that two countries sharing the same currency trade three times as much as they would with different currencies (Rose, 2000, p.7). Frankel and Rose (2002) have combined estimates of the trade-creating effects of common currencies with evidence of a link between trade and income, concluding that some countries could increase their per capita income by 20 percent over 20 years by dollarizing or adopting the euro. 1-3

4 The Effects of Common Currencies on Trade Rose s estimates of the trade-creating effects of adopting common currencies are obtained using a gravity model. Rose s model also includes several variables that are intended to capture trading partners cultural and historical links and membership in regional trading blocs. The key feature of Rose s gravity model is the inclusion of two monetary variables: a dummy variable to indicate whether trading partners use the same currency and a measure of the volatility of the bilateral exchange rate. 1-4

5 The Effects of Common Currencies on Trade In this paper/course, we re-examine the trade-creating effects of common currencies, replacing Rose s equation with a more general functional form. Rather than controlling for cultural, geographic, and historical factors by introducing particular variables, we allow for general fixed effects that are specific to each trading pair. Preview Empirical Model The gravity equation used by Rose and his colleagues Fixed effects model Data and Estimation Results 1-5

6 Competing Gravity Models Rose and his colleagues estimate gravity models, such as LTRADE ijt =β 0 +β 1 LRGDP ijt +β 2 LRGDPPC ijt +θ 1 LDIST ij +θ 2 BORDER ij +λ 1 COMLANG ij +λ 2 COMCOL ij +λ 3 COLONIAL ij +ωfta ijt +τt ijt +γcu ijt +δvolatd jt +ε ijt, (1) where LTRADE ijt is the log value of bilateral trade (exports plus imports) between country i and j at time t, LRGDP ijt is the log product of real GDP, LRGDPPC ijt is the log product of per capita incomes of the exporting and importing countries, LDIST ij is the great-circle distance between the trading partners (in logs), and BORDER ij is a dummy variable that is unity if the two countries share a common border. 1-6

7 Competing Gravity Models LTRADE ijt =β 0 +β 1 LRGDP ijt +β 2 LRGDPPC ijt +θ 1 LDIST ij +θ 2 BORDER ij +λ 1 COMLANG ij +λ 2 COMCOL ij +λ 3 COLONIAL ij +ωfta ijt +τt ijt +γcu ijt +δvolatd jt +ε ijt, (1) To control for time-invariant cultural and historical factors, Rose and his colleagues typically include several dummy variables. These dummy variables take the value of 1, for example, when the partners have a common language, COMLANG ij ; were colonies of the same colonizer,comcol ij ; or if one was a colony of the other, COLONIAL ij. To control for the trade-creating effects of regional integration, the dummy FTA ijt is equal to 1 when the partners are members of the same free trade area or other form of regional integration regime. The model also includes a vector of time dummies, T ijt. 1-7

8 Competing Gravity Models LTRADE ijt =β 0 +β 1 LRGDP ijt +β 2 LRGDPPC ijt +θ 1 LDIST ij +θ 2 BORDER ij +λ 1 COMLANG ij +λ 2 COMCOL ij +λ 3 COLONIAL ij +ωfta ijt +τt ijt +γcu ijt +δvolatd jt +ε ijt, (1) The variables of most present interest are the two monetary variables: CU ijt, a dummy that is equal to 1 if i and j use the same currency, and VOLATD jt, which is a measure of the volatility of the exchange rate between the currencies of i and j. The latter is the standard deviation of the first difference in the log of the bilateral exchange rate between i and j for the five years prior to the year of the observation. 1-8

9 Competing Gravity Models We apply an alternative (more general) specification that uses trading pair-specific fixed effects to control for timeinvariant geographic, cultural, and historical factors. Rather than controlling for these factors with a list of particular variables, as Rose does, we use fixed effects that are specific to each of the trading pairs. The general advantage of this fixed-effects approach is that it avoids estimation bias that can arise because of misspecified or omitted time invariant factors that are correlated with bilateral trade and some right-hand-side variables. 1-9

10 Competing Gravity Models We use least squares with dummy variables (LSDV) to estimate the equation LTRADE ijt =β 0 +β ij +β 1 LRGDP ijt +β 2 LRGDPPC ijt +ωfta ijt +τt ijt +γcu ijt + δvolatd jt +ε ijt. (2) β ij, the trading pair-specific component of the intercept, incorporates all time-invariant factors. These include the distance and contiguity variables, as well as the time-invariant cultural and historical factors (LDIST ij, BORDER ij, COMLANG ij, COMNAT ij, COMCOL ij, COLONIAL ij ). The fixed-effects term also controls for factors that are not included in (1). The trading-pair fixed effect is estimated using a separate dummy variable for each trading pair in the data set. 1-10

11 Competing Gravity Models The main benefit of the fixed-effects approach is that it addresses the possibility of omitted-variables bias by controlling for all factors that are fixed over the sample period, not only those included in the estimation. The United States and Panama or the United States and Liberia use the U.S. dollar. There are obvious historical reasons why they would (i) be more likely than others to share a currency and (ii) trade more than would be otherwise predicted. Because model (1) does not control for factors such as these, which are correlated with both trade and the likelihood of sharing a currency, it provides biased estimates. It is not feasible to create variables that capture the unique historical relationships between these countries or between any pair of countries. 1-11

12 Data and Estimation Results Our (unbalanced) panel data set was constructed using data from Gosh and Yamarik (2004). The data set consists of six annual observations for 186 developing and developed countries. The annual observations are for 1970, 1975, 1980, 1985, 1990, and 1995 and are representative of international trade. 1-12

13 Regression Results for Equation (1) Variable Coefficient Std. Error t-statistic Prob. C LRGDP LRGDPPC LDIST BORDER COMLANG COMCOL COLONIAL FTAIJ VOLATD CU Period fixed (dummy variables) Adjusted R-squared

14 Gravity Model with Trading-Pair Fixed Effects Coefficient Std. Error t-statistic Prob. C LRGDP LRGDPPC FTAIJ VOLATD CU Effects Specification Cross-section fixed (dummy variables) Period fixed (dummy variables) Adjusted R-squared