Governments in developing countries have long realized that migrant remittances are a significant source of external finance. Remittances—which arise when migrants send money back home to their families—are not only an important lifeline for some of the poorest countries in the world, but also constitute a sizable share of gross domestic product (GDP) for emerging market countries. In countries such as El Salvador, Haiti, Honduras, and Jordan, inflows of remittances exceed 15% of GDP. In 2004, a total of 42 developing countries had inflows of remittances greater than 5% of GDP.Footnote ² The World Bank estimates that total recorded flows of remittances reached $318 billion in 2007; this is a staggering sum that dwarfs other external financial sources, such as official development assistance, bank lending, and private investment. Annual flows of remittances even exceed foreign direct investment (FDI) for the majority of developing countries. Central bankers, such as Governor Salameh of the Central Bank of Lebanon (quoted previously), have certainly taken notice and are deeply aware of the impact of remittances on their economies.Footnote ³

Remittances pose a challenge to our understanding of the influence of global finance on national policy choices in the developing world. Indeed, as a form of capital inflow, remittances have many unusual characteristics. Most strikingly, they are “unrequited”: they do not result in claims on assets, debt service obligations, or other contractual obligations (Brown Reference Brown2006; Kapur Reference Kapur, Maimbo and Ratha2005). In contrast to purchases of financial or productive assets, which can be liquidated and repatriated, remittances cannot be withdrawn from a country ex post. They therefore cannot be lumped together with other capital flows that arguably cause household insecurity or income volatility, such as FDI and portfolio flows (e.g., Ahlquist Reference Ahlquist2006; Garrett Reference Garrett1998; Scheve and Slaughter Reference Scheve and Slaughter2004), or with financial capital that can be withdrawn by investors in reaction to unfriendly government policies (Jensen Reference Jensen2006; Li and Resnick Reference Li and Resnick2003; Mosley Reference Mosley2000, Reference Mosley2003). Moreover, migrants tend to increase their remittances when their countries of origin experience economic difficulties. As a result, remittances smooth the incomes of families and shield policy makers from the vagaries of the global economy. In short, financial transfers from migrants are a form of insurance for developing countries against exogenous shocks (Kapur Reference Kapur, Maimbo and Ratha2005; Lopez-Cordova and Olmedo Reference Lopez-Cordova and Olmedo2006; Lucas and Stark Reference Lucas and Stark1985; Rapoport and Docquier Reference Rapoport and Docquier2005; Yang and Choi Reference Yang and Choi2007).

What are the implications for national policy making when cross-border financial transfers within families emerge as a prominent force in the global economy? The prominence of remittances has potentially profound implications for a variety of national policy choices.Footnote ⁴ This article focuses on exchange rate policy, which is arguably the most important macroeconomic policy domain for governments in developing countries (Cooper Reference Cooper1999).Footnote ⁵ Indeed, the exchange rate is the most important price in an open economy, because it affects the prices of all other goods and services. As with most economic policy choices, exchange rate policy entails important trade-offs (Bernhard and Leblang Reference Bernhard and Leblang1999; Broz Reference Broz2002; Cohen Reference Cohen, Higgott, Leaver and Ravenhill1993; Frieden Reference Frieden1991; Leblang Reference Leblang1999; Walter Reference Walter2008). Policy makers choose fixed rates to facilitate international trade and investment and provide an anchor for monetary policy, but they lose the ability to adjust monetary policy to changing domestic circumstances—an ability commonly dubbed “domestic monetary policy autonomy.” Policy makers select floating rates to retain the ability to adjust interest rates in reaction to exogenous shocks or economic downturns, but they incur costs in terms of increased uncertainty in international economic relationships and greater difficulty in anchoring expectations about inflation.

This article argues that remittances are an important influence on exchange rate policy making in the developing world, along with political institutions, interest groups, and other political economy considerations. Remittances mitigate the political costs of lost monetary policy autonomy because they react countercyclically to economic downturns and otherwise insulate policy makers from economic volatility. In essence, remittances have the capacity to substitute (albeit imperfectly) for domestic monetary policy autonomy in the developing world. Therefore, I expect inflows of remittances to be positively associated with the implementation of fixed exchange rates. I develop this argument using conventional macroeconomic models in unconventional ways. Using Robert Mundell's (Reference Mundell1961) optimum currency area framework, I argue that migrant remittances serve a function similar to that of cross-border government transfers (or other supraregional risk-sharing mechanisms) in allowing the domestic economy to adjust to a fixed exchange rate.

I begin with an overview of remittances in the global economy, including trends, causes, and consequences. I also summarize the ample evidence of the countercyclicality of remittance flows. I then provide an empirical test of the hypothesis that remittances, along with interest group pressures, political institutions, and macroeconomic conditions, are important determinants of exchange rate regimes in the developing world. Using newly available World Bank data on annual remittances from 1982–2006 for up to 74 developing countries, I demonstrate that countries for which remittances constitute a substantial share of GDP are more likely to adopt fixed exchange rates. This finding is of particular significance given the recent ideological shift against fixed rates: it appears that remittances encourage policy makers to go against the tide. The findings are robust to multiple model specifications, including de facto and de jure measures of exchange rate policy. I also account for possible endogeneity by using migrant flows to wealthy countries as an instrumental variable for remittances. The article concludes with a discussion of the broader implications of remittances for the political economy of national policy making in a global economy.

REMITTANCES: DEFINITIONS, TRENDS, AND CONSEQUENCES

International financial transfers from migrants to family members in their home countries are known as remittances. A typical remittance transaction contains two parts: (1) the migrant contracts with an agent—a money service business such as Western Union, a bank, or an informal agent—and transmits the money to the agent via cash, check, credit card, or other debit instruction; and (2) the agent instructs its own affiliate in the receiving country to deliver the remittance to the beneficiary (Ratha Reference Ratha2005a).

Remittances have experienced strong growth over the past two decades. Recorded remittances to developing countries increased from $31.2 billion in 1990 to $160 billion in 2004, and to more than $300 billion in 2007. The rate of growth was highest for “lower middle income countries” (with approximate GDP per capita between $1,000 and $3,500), a category that includes countries such as El Salvador, Indonesia, and Tunisia. The growth in remittances is particularly striking in comparison to portfolio investment (private debt and equity), which declined by 20% between 1995 and 2004, and official development assistance, which increased by a modest 34% over the same period. The result of these trends is that remittances are second only to FDI as a source of external capital flows in the developing world (Figure 1). Indeed, remittances were larger than the total of all public and private capital inflows—including FDI, foreign aid, and private debt and equity investment—for 36 countries in 2004. Even in Mexico, which is known for attracting investment from U.S. corporations, inflows of remittances have been nearly equal to FDI inflows since 2003.Footnote ⁶

FIGURE 1.

Capital Inflows ($Millions) to Developing Countries, 2004

Source: World Bank (2006).

Migrants in the United States remitted nearly $39 billion to their countries of origin in 2004, making it the largest source country for remittances (World Bank 2006). The other significant source countries include many of the large continental European economies (Germany, France, Switzerland, and Italy) as well as the countries of the Gulf, including Saudi Arabia, Kuwait, and Oman.

It is a misconception that remittances flow only to very poor countries. Perhaps surprisingly, in 2004, France, Spain, and Belgium were among the 10 largest recipients of remittances. Among developing countries, more than 70% of total remittances accrue to those in the “middle income” bracket, including China, Honduras, and Peru. Nevertheless, for poor countries such as Mongolia, Nepal, and the Gambia, remittances frequently constitute more than 10% of GDP and thus are a critical lifeline for the resident population (Figure 2).

FIGURE 2.

Remittance Inflows (%GDP), to Selected Countries, 2004

Source: World Bank, World Development Indicators (various years).

THE POLITICAL ECONOMY OF EXCHANGE RATE REGIMES

The analytical heart of the literature on the political economy of exchange rates is the Mundell-Fleming model and its famous implication that countries must choose to forgo one of three policy goals: exchange rate stability, full capital mobility, or domestic monetary policy autonomy (Fleming Reference Fleming1962; Mundell Reference Mundell1960). In today's world of highly integrated financial markets, a discrepancy between the domestic and world interest rates causes capital to flow in the direction of the higher return. If the exchange rate is allowed to float, it will adjust accordingly—appreciating with capital inflows and depreciating with capital outflows. However, if the exchange rate is fixed, then the interest rate differential is quickly arbitraged away by the capital flows. The result is that the combination of mobile capital and a fixed exchange rate renders monetary policy ineffective as a policy tool.

The Mundell-Fleming conditions imply that governments face a trade-off between credibility and flexibility (Bearce Reference Bearce2007; Bernhard, Broz, and Clark Reference Bernhard, Broz and Clark2002; Frankel Reference Frankel1999). Credibility arises from the fixed exchange rate, which decreases transaction costs for investors, traders, and other groups with ties to the global economy (Frieden Reference Frieden2002). Reducing or eliminating exchange rate volatility can facilitate international borrowing and stabilize the real value of debts denominated in foreign currencies (Calvo and Reinhart Reference Calvo and Reinhart2002; Walter Reference Walter2008). A fixed rate also leads to monetary stability by tying the hands of monetary policy makers (Giavazzi and Pagano Reference Giavazzi and Pagano1988). Businesses and the public at large moderate their wage and price expectations because they believe the primary goal of monetary policy is to maintain exchange rate parity (Canavan and Tommasi Reference Canavan and Tommasi1997; Keefer and Stasavage Reference Keefer and Stasavage2002). Countries with high inflation are therefore especially interested in the credibility-enhancing features of a fixed exchange rate. However, at the most fundamental level, a fixed exchange rate requires the government to subordinate domestic concerns—whether political or economic—in favor of international concerns (Frieden Reference Frieden2006; Simmons Reference Simmons1994). Often this implies that the government must sacrifice short-term economic growth and employment levels to preserve the exchange rate. Moreover, rigidly fixed exchange rates may be prone to speculative attack, thereby undermining the currency stability they were designed to provide.Footnote ¹⁷

In contrast, flexibility is associated with floating exchange rates, which provide monetary policy makers with the capacity to adjust interest rates to changing domestic economic circumstances. Under flexible rates, policy makers can ease monetary policy to offset an economic downturn, thereby stabilizing employment and output. Moreover, the exchange rate can adjust to counteract current account imbalances. This flexibility comes at the cost of lower monetary policy credibility, because in the absence of a transparent target for the exchange rate, the public is unsure of policy makers’ commitments to maintaining stable prices.

A political economy model of exchange rate regime determination can be assembled largely around these trade-offs. The basic model starts with the presumption that political leaders respond to domestic (and sometimes international) political pressures from interest groups, and that these pressures are broadly mediated through and constrained by political institutions. Because actual lobbying efforts in favor of or against exchange rate policy are rare, I follow existing work in asserting a link between sectoral size (or the magnitude of a particular economic activity such as exports) and political influence over economic policy outcomes (e.g., Brooks Reference Brooks2004; Chwieroth Reference Chwieroth2007; Copelovitch n.d.; Frieden, Ghezzi, and Stein 2001).Footnote ¹⁸

The institutional influences on exchange rate policy can be culled from the emerging literature on exchange rate regime determination. Existing scholarship argues that the degree of democracy is positively associated with floating exchange rates because leaders in democratic countries face pressures from constituents to use monetary policy for domestic adjustment purposes (Bernhard and Leblang Reference Bernhard and Leblang1999; Leblang Reference Leblang1999). Broz (Reference Broz2002) further argues that democracies, which benefit from greater political transparency than non-democracies, can guard the credibility of their monetary policy-making process without tying their hands with a fixed exchange rate. Other scholars examine the relative political costs of enduring the often painful domestic adjustments required to maintain a fixed exchange rate, which are arguably lower in stable governments and those with small numbers of veto players (Edwards Reference Edwards, Ito and Krueger1999; Keefer and Stasavage Reference Keefer and Stasavage2002; Simmons Reference Simmons1994). Finally, focusing on developed democracies, studies such as those by Clark (Reference Clark2002), Clark and Hallerberg (Reference Clark and Hallerberg2000), and Hallerberg (Reference Hallerberg2002) examine the trade-off between fiscal and monetary policy discretion within the Mundell-Fleming framework. They note that fixed exchange rates enhance the power of fiscal policy when capital is fully mobile. Governments are therefore more likely to adopt fixed exchange rates when fiscal policy, rather than monetary policy, is the more effective tool for electoral gain, as in Organization for Economic Co-operation and Development (OECD) multiparty coalition states where targeted spending can be rewarded by voters (Hallerberg Reference Hallerberg2002).

Groups in society that benefit from stable currency relations with other countries, such as exporters and certain investors, can be expected to use their political influence to press for exchange rate stability (Frieden Reference Frieden1991). However, a clear mapping of sectoral interests is not always possible for a large sample of countries. Frieden and his colleagues argue that exporters and import-competers both value currency depreciation and therefore oppose a rigidly fixed exchange rate, whereas foreign investors and creditors value the stability of a fixed rate (Blomberg, Frieden, and Stein Reference Blomberg, Frieden and Stein2005; Frieden Reference Frieden2002; Frieden, Ghezzi, and Stein 2001). However, when international trade occurs between developing countries with limited capacities to hedge exchange risk, manufacturers and other exporters might prefer the currency stability and lower transaction costs afforded by a fixed exchange rate. Likewise, there are no clear partisan divides over exchange rate policy in a contemporary cross-national context. Although right governments have traditionally been in favor of price stability and the interests of the financial community, whereas left governments have favored full employment and income equality, today the mapping of those interests onto exchange rate policy is not straightforward. Left governments, for example, might be torn between an autonomous monetary policy to respond to economic downturns (under a floating exchange rate) and a possible expansion of export sector employment (under a fixed exchange rate). Likewise, right governments might prefer floating exchange rates if alternative mechanisms are available for ensuring stable prices, or if there are additional benefits to financial interests that result from a floating currency.Footnote ¹⁹

REMITTANCES AND EXCHANGE RATE REGIME CHOICE

An important consideration in the political economy of exchange rate regimes is the influence of capital mobility.Footnote ²⁰ The disparate studies discussed previously conceive of capital mobility as the sensitivity of capital flows to domestic rates of return (e.g., Goodman and Pauly Reference Goodman and Pauly1993). Scholars generally measure capital mobility as a policy choice: if governments impose no restrictions on capital flows, then capital is assumed to be responsive to differential rates of return (e.g., Oatley Reference Oatley1999). In empirical studies of exchange rate regimes, an index of financial policy openness from Quinn (Reference Quinn1997) or Chinn and Ito (Reference Chinn and Ito2006), or a simple dichotomous variable based on IMF surveys of capital controls, is frequently the only included measure of a country's relationship with international financial markets.Footnote ²¹ The standard argument is that financial closure allows governments to reap the benefits of fixed exchange rates without sacrificing domestic monetary policy autonomy (Bernhard and Leblang Reference Bernhard and Leblang1999; Broz Reference Broz2002; Leblang Reference Leblang1997, Reference Leblang1999). Financial openness, in contrast, makes the adoption of fixed exchange rates less attractive and therefore less likely.Footnote ²²

For developing countries, the international financial consequences of immigration must also enter the equation. Introducing remittances into the political economy model of exchange rates does not imply an abandonment of the Mundell-Fleming conditions. Indeed, mobile capital will respond to differential rates of return, even in countries that are heavily dependent on remittances. However, I argue that such countries will be less concerned about forgoing domestic monetary policy autonomy. Consider the impact of an increase in remittances during a recession in the receiving country. Households use the funds to bolster their consumption of food and basic necessities, and to maintain existing small businesses and other investments. Such spending and investment has a multiplier effect on the economy, triggering additional investment and consumer spending. In short, remittances—when sufficiently large in relation to the economy—constitute an automatic stabilizer that performs a function similar to that of countercyclical monetary policy. As such, remittances stand apart from other capital flows in that they do not exacerbate the trade-off between fixed exchange rates and domestic monetary policy autonomy. To be clear, remittances are not a panacea for economic instability: they are unlikely to prevent recessions or to respond with enough force to allow a country to sustain a fixed rate in the face of a massive speculative attack. The argument is simply that remittance inflows make it less costly for countries to adopt fixed rates.

Robert Mundell's (Reference Mundell1960, Reference Mundell1961) analyses of optimum currency areas (OCAs) provide a useful perspective on the importance of remittances in the determination of exchange rate policy. The OCA framework, elaborated by McKinnon (Reference McKinnon1963) and others, argues that countries that choose to share a common currency should respond similarly to economic shocks, such as sudden changes in the prices of commodities. The logic is straightforward: a single currency implies a single monetary policy. The same logic applies to countries with fixed exchange rates: a country that fixes its currency to the U.S. dollar essentially imports U.S. monetary policy. If economic conditions vary substantially across different regions of the currency area, a single monetary policy will prove woefully inadequate in stabilizing the economy. However, because asymmetric shocks are always possible, even in the most economically homogeneous of currency unions, countries must somehow adjust their own domestic economies to fit the prevailing monetary policy. The OCA literature has focused on two adjustment mechanisms: (1) labor mobility within the union should be high enough to allow workers in adversely affected regions to relocate to more favorable employment environments; and (2) the currency union itself should have a system of “risk sharing”—usually defined as public transfers from a supraregional authority—to respond to local shocks, just as the U.S. federal government sends emergency funds to states in times of crisis.

The OCA criteria are rarely realized in practice, especially for developing countries that anchor their currencies to the Euro, the U.S. dollar, or other developed country currencies. Shocks to developed and developing economies are likely to be asymmetric, and labor mobility is rarely high enough to be an effective short-term stabilizer. On the issue of risk sharing, however, many developing countries depend on remittances to offset economic downturns. Remittances are not fiscal transfers per se, because no central government has the power to direct them to countries in need. Yet, they do enable countries to cede some of the risks of forgone monetary policy autonomy to migrants, who in turn remit funds to their families in countercyclical fashion.Footnote ²³

The previous discussion sets the stage for an empirical analysis of exchange rate policy in the developing world. The existing literature has emphasized the political and economic factors that determine how policy makers reconcile the trade-off between credibility and flexibility, but has neglected the role of remittances in tilting the balance in favor of fixed exchange rates. To be clear, remittances are not dispositive for policy makers: they ease the political costs of tying their hands with a fixed rate, but other political economy factors will weigh heavily in a policy maker's decision calculus. A complete political economy model must therefore incorporate not only a range of political and institutional variables that determine how policy makers address the trade-offs of exchange rate regime choice, but also the role of remittances as a determinant of the severity of those trade-offs.

EMPIRICAL ANALYSIS

To assess the political economy of exchange rate regimes in the developing world, I assembled a time-series cross-sectional data set with annual observations on up to 74 developing countries from 1982 to 2006.Footnote ²⁴ The hypothesis to be tested is that remittance inflows increase the probability that a country will choose to fix its exchange rate, controlling for a variety of political, economic, and institutional mechanisms. The dependent variable is the de facto exchange rate regime, coded as a four-category ordinal variable based on data from Reinhart and Rogoff (Reference Reinhart and Rogoff2004). Higher values indicate greater degrees of exchange rate flexibility.Footnote ²⁵ Unlike de jure classifications based on official government policy, these de facto measures of exchange rate regimes are derived from a combination of foreign reserve activity, parallel market exchange rates, and extensive country chronologies (Reinhart and Rogoff Reference Reinhart and Rogoff2004). They therefore capture the actual operation of the exchange rate regime over time. In the robustness section, I employ an alternative measure of the dependent variable based on the IMF's de jure classification.

The percentage of countries in the world with de facto fixed exchange rates has remained relatively stable since 1980, hovering around 45%. However, there has been a steady decline in the number of countries with de jure fixed exchange rates, arguably reflecting a shifting “climate of ideas” in favor of floating exchange rates in the developing world (Collins Reference Collins1996).Footnote ²⁶ Between 1980 and 1995, the percentage of countries with fixed exchange rates fell dramatically from 70% to less than 30% (Figure 3). The adoption of the Euro starting in 1999 reversed the overall trend, but for developing countries fixed rates remain far less popular today than in the 1970s and 1980s. This downward trend is addressed in the empirical analysis of de jure policy in the “Robustness” section.

FIGURE 3.

Trends in de facto and de jure Fixed Exchange Rates, 1980–2005

Source: IMF (multiple years) and Reinhart and Rogoff (2004). Includes regimes coded as 1 (see text for discussion).

Data on the key explanatory variable, inward remittances as a share of GDP, are newly available from the World Bank's World Development Indicators (various years).Footnote ²⁷ I use these data with a degree of caution. World Bank researchers are able to estimate only the officially recorded inward remittances for each country-year, not the flows through unofficial channels, such as the hawala system and other informal value transfer systems. As discussed previously, recorded flows have risen substantially in recent times, and a portion of this increase may be attributable to a shift from unofficial to official transmission channels, rather than an increase in remittances per se. The World Bank attempts to mitigate this problem by using estimates from its own country desks or from national central banks when official balance-of-payments statistics are missing or of questionable construction. Nevertheless, unofficial flows remain outside the scope of the data set. I return to this issue in the “Robustness” section.

A cursory overview of the data suggests that remittances are highly correlated with exchange rate regime outcomes. Using the sample in Model 1 (described as follows), the mean level of remittances for countries with fixed exchange rates is 7.9% of GDP, whereas the mean for countries with floating rates is 3.5% of GDP.Footnote ²⁸

I first construct a model (Model 1) that adds remittances along with key policy indicators, macroeconomic conditions, and national institutional characteristics. The size of the economy (GDP, logged) and exports as a share of GDP (lagged one period) capture the conventional argument that smaller and more open economies are more likely to benefit from a fixed exchange rate. Exports are also an interest group indicator with the expectation that firms that are dependent on external demand for their revenues are likely to prefer the stability of a fixed rate. Based on prior scholarship, capital account openness should be negatively associated with the adoption of fixed rates. However, the OCA framework suggests that countries with more open capital accounts should be more likely to adopt fixed exchange rates because high levels of financial integration can generate strong domestic support for stable cross-border financial relationships. The model includes the “KAOPEN” index of capital account openness from Chinn and Ito (Reference Chinn and Ito2006). It is based on the binary coding of restrictions in the IMF's Annual Report on Exchange Arrangements and Exchange Restrictions and focuses on four dimensions of restrictions: the existence of multiple exchange rates, restrictions on the current and capital accounts (where the latter are measured as the proportion of the past five years without controls), and requirements to surrender export proceeds.Footnote ²⁹ The index has a mean of zero and ranges from −2.66 (full capital controls) to 2.66 (complete liberalization).

Model 1 includes a measure of democracy based on the Polity IV database (Marshall and Jaggers Reference Marshall and Jaggers2007). The variable ranges from -10 (most autocratic) to 10 (most democratic). Following Broz (Reference Broz2002) and Leblang (Reference Leblang1999), I expect this variable to be positively associated with floating. The rate of inflation (lagged one period) is included with the expectation that high inflation countries choose a fixed rate as an anchor for monetary policy.Footnote ³⁰ The level of foreign currency reserves (as a ratio of months of imports) reflects the resources available to the central bank to intervene in foreign exchange markets.Footnote ³¹ Also included are the current account balance as a share of GDP and terms-of-trade volatility.Footnote ³² Policy makers in countries with current account imbalances and volatile trading patterns face incentives to allow the currency to float. Finally, the model includes the level of economic development (GDP per capita) and a dummy variable that takes a value of 1 after 1998 for any country that has joined the European Union (EU) during the sample period. This coding scheme accounts for the external pressure to maintain a stable parity with the Euro as a prerequisite to joining the EU and ultimately the Eurozone.Footnote ³³

Model 2 adds more refined interest group and institutional indicators. First, it includes Henisz's (Reference Henisz2002) political constraints measure. The construction of this variable begins by identifying the number of effective branches of government—including the executive, the legislative body or bodies, the judiciary, and any other subnational units—with veto power over policy change. This initial measure is modified to reflect whether these veto points are controlled by different political parties, and the degree of preference heterogeneity within each branch. Higher values represent “stronger,” or less constrained, governments. The theoretical expectation for the impact of political constraints is ambiguous (von Hagen and Zhou 2006). Weak governments might choose a fixed rate to fend off political pressures for expansionary monetary policy; however, the maintenance of a fixed rate might require a strong (i.e., relatively unconstrained) government to subordinate domestic concerns in favor of stable monetary relations with other countries.

Also included in Model 2 are a measure of government instability and the share of manufacturing output in GDP. Government instability is measured as the percentage of the previous five years in which the country experienced an “adverse shift in the pattern of governance,” including a major shift toward authoritarianism, a revolution in the political elite, contested dissolution of federal states, or the collapse of central authority [Political Instability Task Force (PITF), various years].Footnote ³⁴ It provides another indicator of the ability of the government to maintain a fixed exchange rate. However, as Edwards (Reference Edwards1996) notes, greater instability increases the costs of abandoning a peg and therefore reduces the ex ante probability that a peg will be chosen, whereas instability makes decision makers less concerned about the costs of reneging on an exchange rate commitment in the future. The manufacturing indicator provides a more fine-grained interest group indicator alongside the more general measure of a country's export dependence. Frieden, Ghezzi, and Stein (2001) find that large manufacturing sectors are associated with floating exchange rates, but it is possible that this finding is limited to the high inflation Latin American countries in which fixed exchange rates were historically associated with an anticompetitive appreciation of the real exchange rate.

Given the ordinal nature of the dependent variable, I estimate the models using ordered probit with standard errors clustered on country. A lagged dependent variable is included to account for the temporal sluggishness of exchange rate policy.Footnote ³⁵ Summary statistics for all variables are presented in Table 1. Table 2 presents the regression results. The sample for Model 1 consists of 992 country-year observations with 73 developing countries; the sample size is reduced to 824 observations and 70 countries for Model 2 due to the limited availability of the additional covariates.Footnote ³⁶

TABLE 1.

Summary Data

TABLE 2.

Ordered Probit Results (de facto and de jure Exchange Rate Regimes)

Note: Ordered probit coefficients; standard errors (clustered on country) in parentheses.

*p ≤ .10; **p ≤ .05; ***p ≤ .01.

The results from Models 1 and 2 support the hypothesis that inward remittances are associated with fixed exchange rate regimes in developing countries. The coefficient for remittances is negative and statistically significant. (Recall that lower values of the dependent variable imply greater degrees of exchange rate fixity.) This result is robust to the inclusion of political, institutional, and OCA-related macroeconomic variables. In both models, inflation is negatively signed and significant, reflecting policy makers’ desires to provide a nominal anchor for monetary policy when the domestic price level is unstable. In addition, the coefficient for democracy is positive and significant, which supports the idea that democratically elected leaders are vulnerable to popular pressures to conduct an autonomous monetary policy under a floating exchange rate. Not surprisingly, the lagged dependent variable is highly significant, reflecting the temporal sluggishness of exchange rate policy.

The results from Model 2, which contains additional covariates to capture a range of economic and institutional determinants of exchange rate policy, demonstrate that remittances are an important influence on policy making. Note that this is a global analysis of 70 developing countries from 1982 to 2006. In addition to providing evidence of the impact of remittances on exchange rate politics, the findings support certain existing arguments in the literature and challenge others. Capital account openness is negatively signed and significant, indicating an association between financial openness and fixed exchange rates. This is largely in line with Mundell-Fleming expectations; however, simpler measures of capital controls have been shown to be positively associated with floating exchange rates in prior scholarship (Broz Reference Broz2002; Leblang Reference Leblang1999). Although exports as a share of GDP is not significant, manufacturing production is significant and negatively signed. This finding is not consistent with Frieden, Ghezzi, and Stein's (2001) findings for Latin America, but it is theoretically consistent with the notion that manufacturers in the developing world desire stable currency relations with their foreign consumers. The level of reserves is also significant in Model 2, indicating an association between fixed exchange rates and ample supplies of foreign currency. The remaining covariates, including political constraints and government instability, are not significant.

Because the substantive interpretation of ordered probit coefficients is not straightforward, I provide simulations in Figure 4 using estimates from Model 2.Footnote ³⁷ The solid line demonstrates how the probability of fixing the exchange rate changes as remittances increase while the other variables are held at their means. The dotted lines represent 95% confidence intervals. I limit the range of remittances (the X axis) to 0% to 20%, although a few countries in the sample have remittances in excess of this level.Footnote ³⁸ When remittances increase from 0% to 10% of GDP, the probability of fixing the exchange rate increases from 6% to 12%. For countries with remittances at the high end of the sample range, the probability of fixing exceeds 20%. These findings are substantial, especially considering that the model includes a lagged dependent variable that may suppress the impact of the other independent variables (Achen Reference Achen2000).

FIGURE 4.

Predicted Probability of Fixing the Exchange Rate by Level of Remittances

Note: Results based on Model 2. All other variables held at their means. Dotted lines represent 95 % confidence intervals. Simulations conducted using CLARIFY (Tomz, Wittenberg and King 2003).

Robustness

There are a number of additional variables whose inclusion in the model could be theoretically justified. The following variables were added to Model 2 as robustness checks; none altered the statistical significance of remittances.Footnote ³⁹ As expected, a measure of partisanship (coded as a left or center-right dummy variable) was not significant.Footnote ⁴⁰ The inclusion of a measure of central bank independence reduced the sample size to just 35 countries, but it was in fact significant and negative.Footnote ⁴¹ It is possible that policy makers in developing countries are more likely to adopt fixed rates to “tie the hands” of central bankers who might not share their monetary policy preferences (O'Mahony Reference O'Mahony2007). It is also possible that central bank independence and fixed rates are imperfectly credible institutions with the same goal, and therefore may complement each other (Bodea n.d.). Some scholars argue that countries that want to stabilize the real value of their foreign debt service payments will prefer fixed exchange rates (e.g., Shambaugh Reference Shambaugh2004; von Hagen and Zhou Reference von Hagen and Zhou2006; Walter Reference Walter2008). A measure of total external debt, however, was not significant. Finally, foreign aid could condition the choice of exchange rate regime if policy makers believed that it was a reliable source of foreign exchange, especially in times of economic downturn. To test this hypothesis, I included a measure of foreign aid as a percentage of GDP. Not surprisingly, it was not significant. Foreign aid is not a reliable capital inflow for most countries, and it is frequently tied to policy adjustments and other conditions. It is therefore not surprising that it does not have the same impact on exchange rate regime choices as remittances.

In addition, I tested the robustness of the findings by using the IMF's de jure exchange rate regime classification as the dependent variable.Footnote ⁴² Since the end of the Bretton Woods monetary system, the IMF has required member countries to make official announcements of their exchange rate regimes. Article IV, Section 2 of the IMF's Articles of Agreement grants the IMF the responsibility for exercising “firm surveillance” over the exchange rate policies of members, which it has used to publish its Annual Report on Exchange Rate Arrangements and Exchange Restrictions. If remittances mitigated the opportunity costs of fixing the exchange rate, then they might also affect government pronouncements about exchange rate regimes.Footnote ⁴³ The results are included in Table 2, Model 3. This model includes an annual measure of the percentage of countries in the world under fixed rates as a way of capturing the “climate of ideas” regarding exchange rate policy (Collins Reference Collins1996; see also Broz Reference Broz2002, Frieden, Ghezzi, and Stein Reference Frieden, Ghezzi, Stein, Frieden and Stein2001, and Simmons Reference Simmons1994). This measure also captures the trend away from de jure fixed exchange rates for developing countries (Figure 3). As shown in Table 2, Model 3, the coefficient for remittances remains statistically significant and negatively signed.

Because the ordered probit model is limited in its ability to account for cross-country heterogeneity,Footnote ⁴⁴ I transformed the dependent variable into a binary measure and estimated a logit model with country fixed effects.Footnote ⁴⁵ This conditional logit model accounts for unobserved cross-country variation, including inter alia the degree of correlation between the economic cycles of the remitting and receiving countries, the cultural motivations for remitting, and other time-invariant characteristics of countries.Footnote ⁴⁶ It should be noted that exchange rate regimes and remittance levels as a share of GDP are relatively slow to change over time for many countries, and therefore, the fixed effects model provides a particularly strenuous test. Nevertheless, the coefficient for remittances remains negative and significant, although the sample size is reduced to 28 countries (434 observations) because of the fixed effects estimator.Footnote ⁴⁷ Results are included in Table 3, Model 4.Footnote ⁴⁸

TABLE 3.

Conditional Logit and Instrumental Variable Probit Results

Note: Dependent variables exchange rate regime 0 = fixed; 1 = floating. Standard errors (clustered on country) in parentheses. Model 4 contains country fixed effects. Model 5 uses a measure of annual emigration to 15 advanced countries as an instrument for remittances; second stage results shown.

**p ≤ .05; ***p ≤ .01.

Finally, as mentioned previously, it is not controversial to state that remittances data suffer from measurement error. The goal of the empirical models discussed is to subject the data to rigorous analysis and ensure that any inherent biases in favor of the argument are adequately addressed. Nevertheless, it is important to acknowledge the limitations of the analyses because the remittances data only reflect the information that governments are able to record. This prompts the question: is a country's ability to track and record remittances associated with its exchange rate policy?

It is highly unlikely that better recording capacity is associated with the adoption of fixed exchange rates. Indeed, the opposite case is more likely to hold. A floating exchange rate requires that the central bank conduct an independent monetary policy, which is a highly information-intensive process. Under a floating exchange rate, central banks require detailed models of the economy, frequent financial updates from financial institutions, reliable indicators of the domestic price level and money supply, and sufficient expertise (by way of governors, economists, and financial analysts) to make appropriate decisions about monetary policy. These are the types of characteristics that are likely to be associated with the ability to track inflows of remittances through the banking sector and through less formal channels. If this is true, the measurement error in the preceding analyses should make finding a positive association between fixed exchange rates and remittance inflows less likely, rather than more likely.