International Interactions, 29:159–193, 2003 Copyright © 2003 Taylor & Francis

THE EFFECT OF SECURITY ALLIANCES ON EXCHANGERATE REGIME CHOICES QUAN LI Assistant Professor, Department of Political Science, The Pennsylvania State University, University Park, PA, USA Because an exchange-rate arrangement by nature involves more than one country and because it has various economic and political implications, it is affected inevitably by interstate political relations. Most previous research explains the exchangerate regime choice as a function of individual country attributes, ignoring the role of interstate political relations and the anchor-currency choice. In this paper, I examine how security alliances influence a country’s choices over the flexible-fixed regime and the anchor currency. Alliances increase the ex ante attractiveness of pegging to one’s ally, because security ties can reduce concerns over relative gains, motivate active collaboration by the anchor-currency ally to defend the regime, and signal to the currency market the durability of the regime. Hence, a country is biased toward pegging to its ally, relative to either pegging to a nonally or choosing the flexible regime. I test the argument for both the Bretton Woods and the postBretton Woods periods. I find that alliance ties affect both the anchor currency and the flexible-fixed regime choices, as expected. But these effects appear to function through the defense-pact alliance alone and are most pronounced for the developing countries. KEY WORDS: exchange-rate regime choices, security alliance, multinomial logit

The choice of an exchange-rate regime is one of the most important macroeconomic policies for a national government. The exchange-rate arrangement affects how governments manage monetary policy and capital flows. Under the well-known Mundell-Fleming conditions, in an open economy, a government that adopts the fixed regime must choose between capital mobility and monetary policy autonomy; the

Received for publication 11 April 2001. I thank Scott Bennett, Charles Eisenstein, Kenneth Oye, Eric Plutzer, Jim Ray, Thomas Willett, and two anonymous reviewers for helpful comments and suggestions. I also thank Alex Braithwaite, Monica Lombana, and Manu Krishnakumar for research assistance. An earlier version of the paper was presented at the Annual Meeting of the American Political Science Association, Washington D. C., 2000. Replication data are available from the author upon request. Address correspondence to Quan Li, Department of Political Science, Pennsylvania State University, 107 Burrowes Building, University Park, PA 16802, USA. E-mail: [email protected]

159

160

Q. LI

tradeoff does not hold under a floating regime. Hence, the two regimes have different welfare and distributional implications. Furthermore, the exchange-rate arrangement affects international monetary relations, involving the exercise of currency power, various distributional consequences, and cross-border externalities. Most previous studies explain the regime choice as a function of various economic and political attributes of a country. However, because an exchange-rate arrangement by nature involves more than one country and because it has various economic and political implications both domestically and internationally, interstate political relations inevitably influence the regime choices. Should the government adopt a floating regime, allowing the foreign exchange market to determine the exchange rate, or the fixed regime under which its currency is pegged to another? If the fixed regime is preferred, to which country’s currency should one peg its own? Decisions over these two issues, both of which influence a country’s exchange-rate arrangement, are theoretically and statistically interdependent. Yet by treating the regime choice as a single decision based on country characteristics alone, most previous research has neither taken into account this interdependence nor addressed the role of interstate political relations. The result is an incomplete understanding of the causal process. For example, even during the heyday of the Bretton Woods system, not all countries adopted the fixed regime, and not all countries pegged their currencies to the U.S. dollar. National attributes alone cannot fully explain the observed variations in these choices. A causal theory must, therefore, demonstrate how these two decisions are interdependent and how they are affected by international politics. In this paper, I link international political relations to national exchange-rate arrangements. I examine how security alliance relations influence exchange-rate arrangements in a context that takes into account decisions over both the flexible-fixed regime and the anchor currency. Alliance ties, which generally correlate positively with the quality of political relations, increase the ex ante attractiveness of the fixed regime to both the pegging and the anchor-currency countries. Such security relations can reduce concerns over relative gains, motivate collaborative defense of the exchange rate parity, and enhance the regime’s credibility in the currency market. Hence, all else equal, a country is more likely to peg to its military ally when choosing its exchange-rate arrangements. I test this argument on the effects of alliance ties for both the Bretton Woods and the post-Bretton Woods periods by applying a constrained multinomial logit model. This article is organized as follows. The first section discusses how the flexiblefixed regime choice and the anchor-currency choice are interrelated, while reviewing the relevant literature. The second develops a theoretical argument of how security alliances affect a country’s exchange-rate arrangement. The third discusses the research design for an empirical analysis of alliance effects. Section four presents the statistical findings. Section five discusses some sensitivity analysis, and the final section summarizes the findings and discusses implications. DUAL CHOICES IN NATIONAL EXCHANGE-RATE ARRANGEMENTS Most studies of the exchange-rate arrangement treat the problem as a binary choice

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

161

Figure 1. Country i’s Choice of the Exchange Rate Regime.

between the fixed and flexible regimes.1 However, the choice problem for any country i involves more than a single decision. If country i should choose the fixed regime, it also has to choose an anchor currency to which it pegs its own.2 Figure 1 presents a schematic version of this decision problem. In Figure 1, only the currencies of two countries j and k or a basket of their currencies l have the potential to serve as anchor currencies. Now to choose an optimal exchange-rate regime, country i may have to make two decisions instead of one. Each decision is treated in probabilistic terms due to the influence of random events. For the sake of illustration, country i chooses to fix its exchange rate with probability p1 or float it with probability p2. If it favors the fixed regime, i also pegs to j, k, or the currency basket l with probability pj, pk, or pl, respectively. The addition of an anchor-currency choice has two important implications for explaining a country’s exchange-rate arrangement. First, one must account for both the fixed-flexible regime choice and the anchor-currency choice because the two decisions are correlated. Second, one also has to consider not only the attributes of different countries but also the relations between country i and countries j and k. I will elaborate these two implications below in conjunction with a discussion of previous research. Most previous studies of the exchange-rate regime choice focus on one of the two decisions in Figure 1, i.e., the choice between the fixed and flexible regimes. The basic economic rationale is from the theory of optimum currency area (OCA). According to this theory, an optimum currency area should consist of diversified economies closely linked by trade and have high factor mobility such that the fixed exchange rate system or a currency union is economically optimal for member economies (Mundell, 1961; McKinnon, 1963). Although international monetary economists (e.g., Krugman, 1995; Cooper, 1999) consider the OCA theory inadequate for

162

Q. LI

explaining this important choice problem, a body of econometric work has established some empirical regularities (e.g., Heller, 1978; Holden et al., 1979; Bosco, 1987; Melvin, 1985; Edwards, 1996). The probabilities p1 and p2 of choosing the fixed and flexible regimes, as denoted in Figure 1, are suggested to be a function of country i’s attributes, such as its economic openness, market size, capital mobility, its credibility in inflation control, and its preferences regarding unemployment and inflation. The fixed regime is often found to be associated with countries that are small and open to trade, with a low degree of international financial integration, but have strong preferences for inflation control. In contrast, political economists explain this choice as the outcome of politics.3 Domestic distributive consequences of the fixed and flexible regimes lead to various conflicts of interest and contests over exchange-rate policymaking. Such conflicts occur between the camp of international traders and investors and the camp of import-competing producers and nontradable goods producers, or between the tradable and nontradable sectors, or among other competing interests (Frieden, 1991; Hefeker, 1997; Gowa, 1983). Party politics and domestic political institutions may also affect the policy preferences of governments in exchange-rate arrangements (Simmons, 1994; Bernhard and Leblang, 1999; Broz, 2002; Leblang, 1999). The existence of a hegemon arguably has a dominant influence over the whole international monetary system and consequently the choices states make over exchange-rate regimes (Kindleberger, 1986; Eichengreen, 1992, 1996; Cohen, 1997). Although these studies offer important insights regarding the choice between the fixed and flexible regimes, they fail to explain why a country that prefers the fixed regime pegs its currency to that of a particular other country. Country i’s attributes per se do not explain its anchor-currency choice. In light of the simple model in Figure 1, explaining the regime choice without accounting for the anchor-currency choice is conceptually inadequate, a point worth further elaboration below. On the one hand, the lack of an appropriate anchor currency prevents a country desiring the fixed regime from realizing its preference, forcing it to adopt the flexible regime. Using the terms in Figure 1, the probability of the fixed regime p1 can be high (because of country i’s attributes), whereas the probabilities pj, pk, and pl may be low (because of relational factors other than country i’s attributes). In this case, country i is unlikely to find an appropriate anchor currency (as implied by small pj, pk, and pl), forcing it not to adopt the fixed regime despite its strong preference (as implied by a high p1). Hence the choice between the flexible and fixed regimes is affected by the choice of an anchor currency.4 On the other hand, Figure 1 shows that country i’s eventual anchor-currency choice is also influenced by the size of probability p1 or p2. If probability p1 is low and p2 is high, country i is less likely to care about the choice of an anchor currency. Hence, to explain an exchange-rate arrangement, one must consider simultaneously the probabilities p1 and p2 alongside the probabilities pj, pk, and pl. The related statistical interdependence needs to be dealt with in statistical estimation, which I shall discuss in detail in the research design section. Conceptualizing the choice of an exchange-rate regime as in Figure 1 suggests another important implication. The anchor-currency choice necessarily involves interactions between country i and country j or k. In addition to country i’s attributes, which affect the fixed-flexible regime choice, and the attributes of countries j and k,

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

163

which help them attract other countries to peg to their currencies, the ties between country i and country j or k are particularly relevant. The choice among alternative anchors has distributional implications. For example, some Asian countries are imbedded in several commodity chains, some linked to an American market and others to a Japanese one. The choice between the yen and dollar pegs not only affect traders but also investors, in terms of the level of transaction costs. Such a choice is inherently a function of interstate relations. Not all analysts ignore this interactive aspect in an exchange-rate arrangement. Cooper (1975) and Hamada (1976) model the choices of alternative monetary regimes by two countries (in a two-economy world) as a battle-of-the-sexes game that has two nonequivalent Nash equilibria. In their analyses of such strategic interactions, however, Cooper and Hamada do not consider the problem of choosing an anchor currency, but rather investigate how two countries agree upon a set of monetary rules facilitating their economic exchanges.5 The voluminous literature on currency competition and substitution may also be relevant to the anchor-currency choice problem (see, for example, Cohen, 1997, 1998). According to one view, because of the benefits accruing to the country whose currency is widely used internationally, and because of the real danger of a foreign currency replacing the currency within a country, national currencies are in constant Darwinian struggle for survival. National currencies that enjoy superior credibility, market confidence, and wide commercial use are more likely to be chosen as anchor currencies. Although this argument is clearly relevant to an empirical evaluation of the determinants of the anchor-currency choice, it ignores the effects of the specific relations between the potential pegging and anchor-currency countries. The theory of optimum currency area (OCA) also touches upon interactions among countries, an often-overlooked aspect of the theory. In Mundell’s (1961) initial formulation, economic variables are not the only factors viewed as important in shaping a currency area; international cooperation among the central banks within a currency area is also deemed essential. More important, when the OCA does not coincide with but cuts across the political domains of sovereign states, the interference of political relations is most conspicuous.6 Heller (1978) and Moon (1982) are among the few scholars who examine empirically the issue of choosing an anchor currency. They both find that trade relations are an important determinant of the anchor-currency choice. Pertinent to this analysis, Moon (1982) examines how trade dependence affects the choice of the U.S. dollar as an anchor currency. He also addresses the effect of military alliances and finds that when a country is allied with the United States, it is more likely to peg to the U.S. dollar. But his study performs only univariate analyses on pegging to the dollar, without considering other potential anchor-currencies or controlling for the fixedflexible regime choice. Below, I lay out explicitly the causal mechanisms through which a state’s security alliances affect its exchange-rate arrangements. SECURITY ALLIANCES, ANCHOR-CURRENCY, AND FIXED-FLEXIBLE REGIME CHOICES Historically, countries that prefer the fixed regime can usually choose among sev-

164

Q. LI

eral competing anchor currencies. Under the Bretton Woods system, the British pound and the French franc were alternatives to the U.S. dollar in terms of serving as anchor currencies. After 1973, the list of alternatives expanded for many countries and included the Australian dollar, the German deutsche mark, the Indian rupee, the Italian lira, the Spanish peseta, the Russian ruble, the South African rand, and even the Singapore dollar. What are the economic incentives of a country in choosing its anchor currency? A country often prefers an anchor currency that offers a stable store of value and a widely accepted medium of exchange. More important, by choosing an anchor currency, a state selects a nominal exchange rate anchor. The motivation is to maintain real exchange rate stability and a desirable valuation path. Furthermore, pegging also leads to exchange rate stability that greatly reduces the exchange rate risk, facilitating international trade and investment. However, it happens that for some countries an anchor currency country that is optimal from an economic perspective is unfriendly politically. Such instances have some welfare-reducing implications for both the anchor-currency choice and the flexible-fixed regime choice. In the end, the country may either choose to peg to some basket of currencies or, as discussed in the previous section, be forced to adopt the flexible regime. In this paper, I argue that security ties can influence these choices at the margin, helping a country avoid picking a second-best option due to political considerations. Security alliances add to the ex ante attractiveness of the fixed regime to both the pegging and the anchor-currency countries. I elaborate the causal mechanism below. States ally because they intend to adjust power distribution (Waltz, 1979), react to threats (Walt, 1987), trade off security and autonomy (Morrow, 1991), pursue an optimal portfolio of security risk and return (Conybeare, 1992), or because they share similar preferences (Bueno de Mesquita, 1981; Smith, 1995; Fearon, 1997). Whatever the reason, military allies have to align their security interests, at least tentatively. Moreover, because alliance generates coordination costs and institutionalized commitments that enhance its credibility (Morrow, 1994; Fearon, 1997), and because allies do honor their commitments most of the time (Leeds, Long, and Mitchell, 2000), alliance ties may strengthen common security interests. Hence, alliance ties frequently motivate international monetary cooperation that is essential for the maintenance of the fixed exchange-rate regime. Oye (1985) and Kirshner (1995) both provide illustrative examples of monetary cooperation among allies in their analyses of the Tripartite Monetary Agreement among Britain, France, and the United States, as well as the joint efforts by the United States and Britain in defending the currency of their Chinese ally in the late 1930s.7 A country may expect less enmity and more active collaboration from its anchor-currency ally in defending the exchange parity. All else equal, it is such an expectation that makes pegging to one’s ally more appealing.8 The anchor-currency country has both economic and political incentives to defend the pegging of its ally. It can expect to derive positive economic benefits and security externalities from the arrangement. Studies of alliance effects on bilateral trade flows argue that allies tend to trade more with each other because the resulting efficiency gains turn into military resources that benefit allies, generating positive security externalities (Gowa, 1994; Mansfield and Bronson, 1997). This logic is consistent with the argument here regarding pegging between countries. A stable fixed regime signifi-

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

165

cantly reduces exchange rate risks and hence increases both bilateral trade and investments between the pegging and anchor-currency countries. These economic benefits in trade and investments can be translated into improved military power for both countries. If such improvement in military power occurs for a pair of potential adversaries, both countries would be concerned about the relative gains. Hence, the pegging country is more likely to refrain from a pegging relationship while the anchor country is less likely to help defend the fixed regime. But, for a pair of security allies, concerns over relative gains are much less significant while relatively compatible security interests motivate them to turn economic gains from pegging into mutually beneficial military resources. Thus, an alliance relationship can propel one country to peg to its ally anchor while inducing the other country to help defend its pegging ally. The anchor-currency country can also reap additional positive security externalities from the pegging arrangement. Kirshner (1995) offers an illustrative discussion of the British experience. During World War II, Britain experienced declining reserves. Relying on countries that continued to peg to the British pound, Britain was able to maintain confidence in its currency and expand its sterling liabilities dramatically to support its war efforts.9 The expectation of plausible future support by the pegging country offers an incentive for the anchor-currency country to intervene to maintain the exchange parity of its ally. The expectation that good political relations motivate concerted efforts at currency defense has another important implication. As models of currency crises suggest, the collapse of the fixed exchange-rate regime results from a loss of credibility of the government in the eyes of the currency market in terms of the government’s willingness and ability to defend the regime (Krugman, 1979). The loss of credibility also can be driven purely by self-fulfilling expectations, just like a bank run can be caused by a rumor (Obstfeld, 1994). Alliance ties, as they inform market observers about the quality of political relations between two countries, provide a signal to the currency market about the willingness of the anchor-currency country to come to the aid of its pegging ally in defending the regime.10 Hence, alliance ties help to stabilize market expectations about the durability of the fixed regime. Just as pegging to a country of low inflation rate enhances one’s own credibility in inflation control, pegging to a security ally helps to acquire additional credibility for the fixed regime. Frequently, a small open economy desiring the fixed regime may have little choice of to whom to peg except for one or two large countries, such as Britain in the 1930s or the United States after World War II. In such instances, bad political relations with these large countries may steer the small country away from pegging and toward the flexible regime instead, because pegging provides the anchor-currency country with an additional instrument of statecraft that might be used against the pegging country. Conversely, if the large country is an ally, the small economy does not need to avoid pegging as it expects less future political manipulation. Hence, the absence of alliance ties, when reflecting bad political relations, may cause a country favoring the fixed regime to choose the flexible regime instead. We can, therefore, conclude that a security alliance induces a country to peg to the currency of its military ally. However, alliances are heterogeneous and do not embody the same level of compatible security interests or commitments (Singer and Small, 1966; Moul, 1988). Some alliance ties may be less likely than others in bring-

166

Q. LI

ing about pegging that involves high stakes and strong commitments. A neutrality agreement merely specifies the avoidance of engagement in a coalition aggressive toward either party of the agreement. An entente pact requires only consultation or cooperation in a time of crisis. In contrast, a defense-pact alliance is much more demanding, asking for a coalition and coordinated military efforts in terms of the willingness to help each other militarily. Therefore, defense-pact allies share more compatible security interests, stronger security commitments to and greater influence over each other and, consequently, can expect of each other more credible and friendly actions. The greater the commitment, the more important a country is to the political purposes of its anchor-currency ally, the less likely the anchor-currency country is to adopt policies that are harmful to the interests of the pegging country, and the more likely it is to make efforts defending the fixed regime. Therefore, if an alliance does affect the anchor-currency choice, we should find the effect to be strongest between the defense-pact allies, relative to the neutrality or entente pact. RESEARCH DESIGN This section discusses the research design for testing the effects of security alliances on the dual exchange-rate regime choices. Because the choices are affected by the interactions between the pegging and anchor-currency countries as well as their national attributes, I identify pairs of countries that represent potential pegger-potential anchor dyads. The unit of analysis is the potential pegger-potential anchor dyad year. The spatial domain includes all pairs consisting of any possible pegging country and its potential anchor-currency countries. For this study, the relevant potential anchor-currency countries include major financial powers and each possible pegging country’s largest trading partners in the past. As a result, the potential anchor countries include both those that have the potential to serve as anchor countries but never did so, such as Japan, and all those countries whose currencies have once served as the anchors for some other countries, including the U.S., U.K., France, Germany, Japan, Italy, Australia, India, Spain, South Africa, and Singapore. Russia is excluded due to missing data. More specifically, the U.S., U.K., France, and Germany are always included in the sample as potential anchors, but Japan, Italy, Australia, India, Spain, South Africa or Singapore are only included as a potential anchor when it has served as the largest trading partner for the potential pegging country in a dyad. This criterion ensures against artificially increasing the sample size and causing spurious findings.11 To examine the robustness of results, I perform separate analyses with all countries and then with the non-OECD developing countries alone as possible pegging countries. The temporal domain includes years from 1966 to 1992, restricted by data availability for exchange-rate regime choices and security alliances. Statistical analysis is also performed separately for the Bretton Woods sample period (1966–1973) and the post-Bretton Woods sample period (1974–1992) because the collapse of the Bretton Woods system brought about different exchangerate regime options, a detail discussed below. Dependent Variable The dependent variable measures the anchor-currency and the flexible-fixed re-

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

167

gime choices. Data on the dependent variable are collected from the International Monetary Fund Annual Reports on Exchange Arrangements and Exchange Controls and the International Financial Statistics Yearbook.12 The coding categories used by the IMF changed somewhat over time, most notably after the breakdown of the Bretton Woods system. Separate models are thus estimated for the two periods before and after 1973. Between 1966 and 1973, 89 countries ever pegged their currencies to the U.S. dollar, 6 countries to the British pound, 3 to the French franc, and 11 countries ever maintained the flexible exchange-rate regime. (Note that the sum of the number of countries can be greater than the total number of countries in a year because countries change their exchange-rate arrangements over time. The same caveat applies below for the post-Bretton Woods period.) For this period, the dependent variable is coded 1 if a possible pegging country pegs to any potential anchor-currency country in the dyad in a year, 2 if it has the flexible regime in the year, and 0 otherwise (i.e., if it has the fixed regime but does not peg to the potential anchor-currency country in the dyad). Between 1973 and 1992, the exchange-rate arrangements were more complicated. The IMF provides three broad categories: (1) pegging, (2) limited flexibility, and (3) more flexibility. There are specific types under each category and they also change somewhat over time, as documented in the IMF publications listed above. Pegging. For some years in this period, 38 countries ever pegged to the dollar, 8 countries to the pound, and 12 countries to the franc. Other currencies also emerged as occasional or long lasting anchors, such as the Australian dollar (2 countries), German deutsche mark (1 country), Indian rupee (1 country), Italian lira (1 country), Spanish peseta (1 country), Russian ruble (5 countries), South African rand (3 countries), and Singapore dollar (1 country). Many countries that desired the fixed regime and yet found the existing single anchor currencies unsuitable chose to peg to the SDR (special drawing rights) basket (22 countries) or some other composite of currencies of their own choice (50 countries). Limited flexibility. Some countries maintained exchange rates of limited flexibility, either in terms of the U.S. dollar (9 countries) or under the European cooperative exchange arrangements (9 countries). The European cooperative exchange arrangements refer to the cooperative arrangement for multicurrency intervention (the “snake”) for the early years and the European Monetary System for the later years. As many argue (e.g., McKinnon, 1993; Eichengreen, 1996), under these European cooperative arrangements, the German deutsche mark served essentially as the anchor currency. More flexibility. Three types of the flexible regime were ever used by 64 countries in this period. In the first type, the exchange rate is adjusted relatively frequently according to a set of indicators. The second is one of other managed floating. The third is one of independent floating. For the post-Bretton Woods period, the dependent variable is coded 1 if a possible pegging country pegs to a potential anchor-currency country in the dyad in a year; when the possible pegging country adopts a regime of limited flexibility, it also is treated as pegging to the United States or Germany. Despite being a category of

168

Q. LI

limited flexibility, the pegging country still imposes on itself a commitment. More important, since the category involves specific anchor currencies (i.e., the dollar or deutsche mark), the effects of alliances should be expected. The dependent variable is coded 2 if the potential pegging country in a dyad has the flexible regime, as defined by the category of more flexibility. It is coded 0 if the possible pegging country has the fixed regime but does not peg to the potential anchor-currency country in the dyad in a year. This includes that the country fixes to the SDR or some other currency composite. The coding scheme requires some further clarification. First, the flexible regime is included as a category of the dependent variable because, as noted, the probability of choosing the flexible regime affects and is affected by whether a country pegs to any potential anchor. Without including this option, one truncates the sample to only countries with the fixed regime and fails to account for the statistical interdependence between the flexible-fixed regime and the anchor-currency choices. Second, the flexible regime is singled out as a separate category because some factors (the possible pegging country’s attributes) affect its choice between the fixed and flexible regimes but do not affect its choice among competing anchor currencies. Moreover, the effects of alliance ties should be stronger in affecting its choice among potential anchor currencies than the flexible-fixed regime choice. But an alliance tie should still be relevant to the fixed-flexible regime choice, as it enhances the attractiveness of the fixed regime. Third, it is worth noting that the dependent variable coding for both periods reduces some wide variety of the exchange-rate arrangement to analytically tractable categories, especially for the post-Bretton Woods period. As discussed earlier, the specific types under each of the three general categories have changed somewhat over time in the IMF coding. But the three broad types remain the same over time. For the purpose to assess the effect of security ties on the choices of pegging and floating, it is appropriate to collapse the categories as above. Independent Variables Alliance Variables A dichotomous variable Alliance measures the presence or absence of any type of alliance ties. It equals 1 if a possible pegging country is allied with a potential anchor-currency country in the dyad in a year through defense, neutrality, or entente pacts according to the COW project (Singer and Small, 1968), and 0 otherwise. It is expected that a country is more likely to peg to its security ally, relative to pegging to a nonally or selecting the flexible regime. As discussed earlier, we expect that the effects of alliances should be greater for the defense-pact alliance and unclear for the neutrality or entente pacts, because heterogeneous alliance agreements demand different levels of commitment and only the defense-pact alliance requires active military participation and coordination. To assess this expectation, the variable Alliance is replaced by two other variables in a second model specification. Defense Pact equals 1 if a possible pegging country has a defense pact alliance with a potential anchor country in the dyad in a year, and 0 otherwise; Entente Neutrality equals 1 if a possible pegging country has a neutrality or entente pact alliance with a potential anchor country in the dyad in a year and 0 otherwise. Alliance data are from the EUGene software (Bennett and Stam, 2000).

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

169

Alliance is not the only factor that influences the exchange-rate regime choices. To assess accurately the effects of alliance, I also include a number of control variables based on previous theoretical explanations. Trade Ties A country’s trade patterns are commonly found to influence its exchangerate arrangements. Countries with geographically concentrated trade are more likely to adopt the fixed regime (Heller, 1978; Bosco, 1987; Holden, et al., 1979). More important, if a country’s trade is concentrated toward one trade partner, the country ought to find it more beneficial and is more likely to peg to that particular partner (Heller, 1978; Moon, 1982). Many countries have large trade volume with other countries, inducing them to peg to the latter’s currencies. Without controlling for trade ties, alliance effects will be overestimated. The measure of trade ties follows that of Heller (1978) and Moon (1982). Bilateral /Total Trade is the percentage ratio of the bilateral trade between a possible pegging country and a potential anchor country in the dyad over the total trade of the possible pegging country. The stronger the trade ties are, the more likely a country pegs to its trading partner, and the less likely it chooses the flexible regime. Data are from the IMF Direction of Trade. Credibility of Potential Anchor-Currency Country Whether a potential anchorcurrency country maintains its credibility in price stability affects whether its currency is or continues to be chosen as an anchor. First, an important benefit of pegging is that it brings credibility from the anchor to the pegging country in inflation control—an important economic policy objective (e.g., Willett and Mullen, 1982; Edwards, 1996). If the anchor country is unable to keep inflation low, the credibility benefit to the pegging country decreases, reducing the likelihood of the particular anchor to be chosen. Second, the inability of the anchor-currency country to control inflation hinders market confidence and is likely to induce speculative attacks against the fixed regime, forcing its collapse. Anchor Inflation is the GDP-deflator based inflation rate of the potential anchorcurrency country that measures its credibility in inflation control.13 The higher the anchor inflation rate, the less likely another country pegs to this potential anchor currency, and the more likely it pegs to another currency or chooses the flexible regime. Data are from the World Bank’s 1999 World Development Indicators. Probability of Choosing Fixed Regime Based on Pegger’s Attributes As discussed earlier, the fixed-flexible regime choice and the anchor-currency choice are interdependent both theoretically and statistically. Hence, to explain a country’s exchange-rate arrangements, it is important to control for the probability the country adopts the fixed or flexible regime based on its attributes. Following previous studies, I estimate the probability of choosing the fixed regime relative to the flexible regime as a function of national economic and political attributes. Specifically, I estimate a probit model of the choice between the fixed and flexible regimes at the country level from 1966 to 1995. The dichotomous dependent variable is coded 1 if a country chooses the fixed regime in a year and 0 otherwise. The independent variables include market size, economic development, trade openness, inflation differential against the world, economic growth, capital control, and democracy level, following previous empirical work (Heller, 1978; Bosco, 1987; Melvin, 1985; Edwards, 1996; Broz, 2002; Leblang, 1999). Size and democracy are expected to

170

Q. LI

reduce the probability of the fixed regime, while economic development, trade openness, inflation differential against the world, economic growth, and capital control are expected to increase the probability of the fixed regime. The linear predictor from the probit model is saved and transformed, according to the inverse Mill’s ratio formula,14 into a new variable—the hazard of being selected into the fixed regime (Pegger’s Attributes).15 Details of the variables in the probit and model results are included in the appendix. Policy Inertia Public policies are difficult to change as they are subject to path dependence and considerable inertia. This applies to exchange-rate arrangements as well (Leblang, 1999). I use the lagged dependent variable (Prior Choice) to capture policy inertia. The lagged dependent variable also controls for the possible effect of temporal dependence and variables not in the model. Methodology Recall that the unit of analysis is the potential pegger-potential anchor dyad year. For any particular observation, the dependent variable is 1 if a country pegs to a potential anchor-currency country in the dyad in a year, 2 if it has the flexible regime in the year, and 0 otherwise (i.e., if it has the fixed regime but does not peg to the potential anchor-currency country in the dyad in the year). The purpose here is to assess the effect of alliance on the probability that a potential pegging country chooses the other country in the dyad as its currency anchor, relative to that it chooses the flexible regime or that it chooses the fixed regime but does not pick the other country in the dyad as the anchor. With this trichotomous dependent variable, I use constrained multinomial logit for estimation. The baseline category is that one country fixes its exchange-rate regime and pegs to the other country in the dyad. The probability of choosing the baseline category is compared with two other categories, either that one country chooses the flexible regime or that it chooses the fixed regime but does not peg to the other country in the dyad. A negative (positive) coefficient indicates that the independent variable raises (reduces) the probability of choosing the baseline category. It is worth noting that I expect alliance variables to be negative. Because the potential pegging country’s attributes only affect its flexible-fixed regime choice and not the anchor-currency choice, the effect of country attributes is constrained in estimation to only influencing the choice between the flexible regime and the fixed regime with pegging in the dyad. In other words, it is constrained to have no effect on the choice between the fixed regime with pegging in the dyad and the fixed regime without pegging in the dyad. The data structure is pooled time series cross sectional, where each cross section is defined by a dyad between a possible pegging country and a potential anchorcurrency country. Models of such data structure are likely to have autocorrelated and heteroskedastic disturbances. To deal with these potential problems, I estimate Huber/White robust standard errors adjusted for clustering over the dyad. This robust estimator produces consistent standard error estimates in the presence of heteroskedastic error variance; it is also robust against serial correlation within the dyad (Wiggins, 1999). Furthermore, the lagged dependent variable also soaks up temporal dependence in the data. Because exchange-rate arrangements may influ-

171

-2.7676*** (0.1840) 2.6071*** (0.1596) 3352 462 0.38

-0.0336*** (0.0060) 0.1427*** (0.0220)

-0.7810*** (0.2182)

-0.0370*** (0.0131) 0.1434*** (0.0222) -0.0783*** (0.0105) -0.7088*** (0.1748) 1.3157*** (0.2352) 3352

-0.9054*** (0.2805)

Flexible Regime

-2.7066*** (0.1846) 2.5104*** (0.1596) 3352 517 0.39

-1.1931*** (0.2442) -0.0162 (0.4505) -0.0310*** (0.0062) 0.1496*** (0.0223) -0.9658*** (0.2910) -0.6806 (0.7301) -0.0361*** (0.0134) 0.1503*** (0.0225) -0.0773*** (0.0101) -0.6856*** (0.1767) 1.2254*** (0.2382) 3352

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

N Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed & No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

-2.6620*** (0.1950) 2.7723*** (0.1738) 2963 405 0.38

-0.0346*** (0.0062) 0.1202*** (0.0218)

-1.0381*** (0.2555)

Fixed and No Pegging

-0.0420*** (0.0125) 0.1205*** (0.0221) -0.0672*** (0.0099) -0.7973*** (0.1925) 1.5453*** (0.2490) 2963

-1.4099*** (0.4054)

Flexible Regime

Model 3

-2.5431*** (0.1926) 2.5737*** (0.1705) 2963 482 0.39

-2.1578*** (0.3510) 0.2083 (0.6192) -0.0274*** (0.0068) 0.1298*** (0.0224)

Fixed and No Pegging

-2.0117*** (0.4976) -0.3189 (0.8372) -0.0365*** (0.0132) 0.1300*** (0.0227) -0.0666*** (0.0098) -0.7425*** (0.1871) 1.3608*** (0.2453) 2963

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Table 1.1 Constrained Multinomial Logit Model of Anchor-Currency and Fixed-Flexible Regime Choices, 1966–1973 (fixed and pegging as baseline category)

172

-2.0635*** (0.1068) 3.7233*** (0.1941) 10945 943 0.53

-0.0467*** (0.0070) 0.0212*** (0.0058)

-0.6005** (0.2944)

-0.0341*** (0.0071) 0.0215*** (0.0058) -0.2882*** (0.0382) 0.9308*** (0.0566) 1.3341*** (0.2295) 10945

-0.2771 (0.2787)

Flexible Regime

-2.0340*** (0.1072) 3.6807*** (0.1928) 10945 971 0.53

-0.9204*** (0.3336) 0.4450 (0.5807) -0.0434*** (0.0072) 0.0203*** (0.0059) -0.4100* (0.2991) 0.4308 (0.6102) -0.0323*** (0.0073) 0.0206*** (0.0059) -0.2850*** (0.0384) 0.9473*** (0.0574) 1.2960*** (0.2273) 10945

Flexible Regime

Model 6 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

N Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 5

All Potential Pegging Country–Potential Anchor Dyads

-2.1064*** (0.1148) 3.7334*** (0.1982) 9967 932 0.52

-0.0464*** (0.0072) 0.0214*** (0.0055)

-1.0818*** (0.2903)

Fixed and No Pegging

-0.0311*** (0.0071) 0.0216*** (0.0055) -0.3373*** (0.0395) 0.8489*** (0.0586) 1.4694*** (0.2309) 9967

-0.6656*** (0.2780)

Flexible Regime

Model 7

-2.0404*** (0.1159) 3.6344*** (0.1940) 9967 1020 0.53

-2.0188*** (0.3670) 0.4292 (0.5883) -0.0388*** (0.0075) 0.0205*** (0.0057)

Fixed and No Pegging

-0.8723*** (0.3099) 0.2170 (0.6134) -0.0279*** (0.0077) 0.0207*** (0.0057) -0.3345*** (0.0397) 0.8801*** (0.0601) 1.3999*** (0.2251) 9967

Flexible Regime

Model 8

Developing Potential Pegging Country–Potential Anchor Dyads

Table 1.2 Constrained Multinomial Logit Model of Anchor-Currency and Fixed-Flexible Regime Choices, 1974–1992 (fixed and pegging as baseline category)

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

173

ence economies and politics of both countries in a dyad, the dependent variable may affect the independent variables. Following the conventional practice, I control for the possible simultaneity bias by lagging all independent variables by one year. RESULTS Parameter Estimates for Alliance Variables Tables 1.1 and 1.2 present the constrained multinomial logit estimates for the Bretton Woods and the post-Bretton Woods periods, respectively. Each table includes two model specifications with alternative alliance variables. Each table also includes results for two different samples, with all countries or non-OECD developing countries acting as the possible pegging countries. Therefore, each table presents results for four models. Each model produces two sets of estimates reflecting, respectively, the effects of independent variables on the likelihood that a potential pegging country chooses the fixed regime but does not peg to the other country in the dyad relative to that it pegs to the other country in the dyad, and the effects of those variables on the likelihood that a potential pegging country chooses the flexible regime relative to that it pegs to the other country in the dyad. For both choices, the reference category is the likelihood that the potential pegging country pegs to the other country in the dyad. Hence, recall that a statistically significant negative (positive) coefficient indicates that the independent variable increases (decreases) the probability of pegging in a dyad, relative to either the fixed regime but no pegging or the flexible regime choice. Since all hypotheses are directional, the one-tail test is applied. In both tables, the Wald tests for the models are statistically significant at 1 percent level. The model goodness of fit, measured by the pseudo R2, appears to be reasonably good across all models as well. Below, I focus on discussing statistical results on alliance variables in both tables. The general alliance variable (Alliance) is statistically significant for both choices in both Models 1 and 3 in Table 1.1. During the Bretton Woods system, a potential pegging country is more likely to peg to the currency of its ally when choosing among competing potential anchor currencies; it is also more likely to choose pegging to its ally over the flexible regime. The results are similar for the post-Bretton Woods period in Table 1.2, except for the choice between pegging and the flexible regime. In Models 5 and 7, a country is still more likely to peg to its ally than a nonally when choosing among competing potential anchor currencies. However, alliance now makes it more likely to choose pegging over the flexible regime only within the developing nations sample. In the alternative model specification, two variables Defense Pact and Entente Neutrality allow us to distinguish the effects of different alliance types. In Table 1.1, the defense pact alliance variable is statistically significant for both choices in Models 2 and 4. During the Bretton Woods system, a potential pegging country is more likely to peg to its defense-pact ally when choosing among competing potential anchor currencies; it is also more likely to choose pegging over the flexible regime due to such alliance ties. These effects also hold for the developing countries. In the post-Bretton Woods period (Table 1.2), the results for the defense pact alliance variable are consistent with those in Table 1.1. In Model 6, a potential pegging country is still more likely to peg to its defense-pact ally in choosing among competing anchor currencies. The

174

Relative Risk 1.08

Relative Risk 1.14

Relative Risk 1.48

Relative Risk 2.16

All Potential Pegging Country–Potential Anchor Dyads: Bretton Woods Period (1966–1973) Relative to Fixed but No Pegging Relative to Flexible Regime No Defense Defense ∆ in Prob. of Relative No Defense Defense ∆ in Prob.of Pact Pact Pegging Risk Pact Pact Pegging 0.0356 0.1086 +0.073 3.05 0.8924 0.9647 +0.0723

Developing Potential Pegging Country–Potential Anchor Dyads: Bretton Woods Period (1966–1973) Relative to Fixed but No Pegging Relative to Flexible Regime No Defense Defense ∆ in Prob. of Relative No Defense Defense ∆ in Prob.of Pact Pact Pegging Risk Pact Pact Pegging 0.0359 0.2437 +0.2078 6.79 0.8576 0.9812 +0.1236

All Potential Pegging Country–Potential Anchor Dyads: Post Bretton Woods Period (1974–1992) Relative to Fixed but No Pegging Relative to Flexible Regime No Defense Defense ∆ in Prob. of Relative No Defense Defense ∆ in Prob.of Pact Pact Pegging Risk Pact Pact Pegging 0.0227 0.0541 +0.0314 2.38 0.096 0.1416 +0.0456

Developing Potential Pegging Country–Potential Anchor Dyads: Post Bretton Woods Period (1974–1992) Relative to Fixed but No Pegging Relative to Flexible Regime No Defense Defense ∆ in Prob. of Relative No Defense Defense ∆ in Prob.of Pact Pact Pegging Risk Pact Pact Pegging 0.0224 0.1412 +0.1178 6.30 0.1173 0.2537 +0.1364

Probability of Pegging

Probability of Pegging

Probability of Pegging

Probability of Pegging

Table 1.3 Substantive Effects of Defense-Pact Alliance on Anchor Currency and Fixed-Flexible Regime Choices (Fixed and Pegging as Baseline Category)

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

175

defense-pact alliance also affects the choice between pegging and the flexible regime. As Model 8 indicates, these effects also hold for the developing countries. Entente Neutrality is not statistically significant in either table, for either the Bretton Woods or the post-Bretton Woods period, for either all the potential pegging countries or the developing potential pegging countries alone. An alliance of entente or neutrality pact does not increase the probability that a country pegs to such an ally, either relative to the choice among competing anchor currencies or the choice between the flexible and fixed regimes. In general, alliance ties affect both the choice of anchor currency and that between the flexible and fixed regimes. Moreover, its effect appears to function through the defense-pact alliance only. The effects also hold for the developing countries. Substantive Significance of Defense-Pact Alliance Effect Even though the defense-pact alliance variable is statistically significant in various models, its parameter estimates are effects on the log odds of the dependent variable at a certain outcome category. They do not provide an intuitive representation of the substantive significance of this variable. Based on the parameter estimates and various scenarios in Tables 1.1 and 1.2, Table 1.3 illustrates the size of effect of the defense pact alliance on the probability that a potential pegging country pegs to the currency of the other country in a dyad, relative to the probability that it fixes but does not peg to the other country in the dyad, or relative to the probability that the potential pegging country chooses the flexible regime. In computing the probability that the dependent variable takes on a value of 1 (that is, one country pegs to the other in the dyad), the continuous independent variables are held constant at their mean level. The lagged dependent variable takes on a value of 0 (that is, one country holds the fixed regime but does not peg to the other in the dyad) while comparing the probability of fixed with pegging in the dyad relative to fixed without pegging; it takes on a value of 2 (that is, the potential pegging country in a dyad has the flexible regime) while comparing the probability of fixed with pegging relative to the flexible regime. The defense pact alliance variable varies from 0 (no such alliance tie between two countries in a dyad) to 1 (such alliance tie exists). Hence, Table 1.3 includes the probability of pegging by a country to the currency of the other in a dyad relative to the probability that it fixes but does not peg to the other country in the dyad or relative to the probability that it adopts the flexible regime, with alliance and without alliance, the change in the probability of pegging without alliance to with alliance, and the relative risk between these two alliance scenarios. Here the relative risk refers to the probability of pegging to one’s defense pact ally in a dyad divided by the probability of pegging to a nonally in the dyad. This statistic is especially useful when the size of the absolute probability is small. Hence, our discussion focuses on the relative risk scores under various scenarios. As Table 1.3 shows in the first four columns in terms of the choice between pegging and fixed but no pegging, during the Bretton Woods period and for the whole sample, a potential pegging country is about three times more likely to peg to the other country in the dyad when they have a defense pact alliance than when they do not have such a tie (given a relative risk score of 3.05). In contrast, a developing potential pegging country is close to seven times more likely to peg to the other

176

Q. LI

country in the dyad, in the presence of such an alliance than in its absence. These patterns remain similar in the post-Bretton Woods period. The relative risk scores are 2.38 and 6.30 for all potential pegging countries and the developing pegging countries, respectively. As Table 1.3 shows in the last four columns (contrasting pegging and the flexible regime), during the Bretton Woods period and for the whole sample, a potential pegging country is only slightly more likely to peg to the other country when they share a defense pact alliance in the dyad than when they do not (given a relative risk score 1.08). The security tie has only a slightly larger effect for the developing potential pegging country (given a relative risk score 1.14). Such effect, however, is noticeably larger in the post-Bretton Woods period, particularly so for the developing countries. A developing potential pegging country is more than twice as likely to peg to the other country in the dyad when they share a defense pact alliance in the dyad than when they do not (given a risk score of 2.16). In summary, the substantive effect of the defense pact alliance on pegging is quite large, relative to the scenario that one country has the fixed regime but does not peg to the other country in the dyad. In contrast, relative to the scenario of the potential pegging country adopting the flexible regime, the effect of the defense pact alliance on pegging is substantively small in the Bretton Woods period, but appears to be growing in the post-Bretton Woods period. It is also worth noting that these effects are much more pronounced for the developing countries. Parameter Estimates for Control Variables The control variables in Tables 1.1 and 1.2 also produce some interesting findings. Trade ties (Bilateral/Total Trade) are in the expected direction and statistically significant in all models. That is, for both the all country sample and the developing country sample, and for both periods, trade ties increase the probability that a country pegs to the currency of its close trade partner, either in the choice among competing anchor currencies or in choosing between pegging and the flexible regime. This finding is consistent with those of Heller (1978) and Moon (1982). The credibility of the anchor currency country over price stability (Anchor Inflation) is statistically significant in all models for both periods (in both Tables 1.1 and 1.2) and in the expected direction. Both during and after the Bretton Woods system, high (or low) inflation of a potential anchor-currency country reduces (or raises) the probability that its currency is used as an anchor, and raises (or reduces) the probability that a potential pegging country adopts the flexible regime. These effects remain robust for the developing countries. As expected, the probability that a potential pegging country chooses the fixed regime over the flexible regime based on its attributes (Pegger’s Attributes) has a statistically significant effect over the choice between pegging and the flexible regime. For both the Bretton Woods and the post-Bretton Woods periods (in both Tables 1.1 and 1.2), the more a country prefers the fixed regime (based on its various national attributes), the less likely it chooses the flexible regime, or vice versa. The effects remain robust for the sample of developing countries. Policy inertia (Prior Choice) is statistically significant in all models in both Tables 1.1 and 1.2, with interesting variations. Consistent across all models for both periods

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

177

(both tables), a country that has already pegged to an anchor currency is likely to continue to do so, relative to the possibility that it chooses the fixed regime but does not peg to the other country in the dyad. However, in terms of the likelihood that a potential pegging country chooses the flexible regime relative to the choice of pegging to the other country in the dyad, the negative coefficients in Table 1.1 and the positive coefficients in Table 1.2 suggest that countries are more likely to continue with the fixed regime instead of the flexible regime under the Bretton Woods, but they are more likely to stick to the flexible regime in the post-Bretton Woods period. These findings suggest that models, without controlling for such path dependence, are likely to be under-specified and that the breakdown of the Bretton Woods marks a system-wide shift in the nature of policy inertia. SENSITIVITY ANALYSIS In this section, I evaluate the robustness of the results in Tables 1.1 and 1.2 under several scenarios: excluding the prior choice variable and controlling for the regional effect, the Cold War, the U.S. effect, or colonial ties between countries. These scenarios arguably may affect the results of alliance variables in Tables 1.1 and 1.2, hence deserving some sensitivity analysis. Again, models are estimated for both all potential pegging countries and just the developing ones, for the Bretton Woods and post-Bretton Woods periods. To save space, the statistical results are presented in the Appendix for reference. For the purpose of this paper, I focus on the effect of the defense pact variable in the Appendix tables. As an aside, the effects of other control variables in the robustness tests remain broadly consistent with those in Tables 1.1 and 1.2. In Appendix Tables 2.1 and 2.2, the prior choice variable, measured by the lagged dependent variable, is excluded. Because the prior regime choice is relevant theoretically, its exclusion can lead to exaggerated effects of other variables. But its inclusion may leave less variance for other variables to explain and thus, is worth examination. Results for the defense pact alliance in these appendix tables remain the same in terms of hypothesis testing as those in Tables 1.1 and 1.2. The effects are generally larger in magnitude without including the lagged dependent variable. In Appendix Tables 3.1 and 3.2, the regional effect, measured by the number of countries with the flexible regime in the relevant region, is included. I assess whether it affects the results of the defense pact variable. As the Appendix tables show, while the regional effect variable is statistically significant and positive, the hypothesis testing results for the defense pact variable remain identical to those in Tables 1.1 and 1.2, except for the choice between pegging and the flexible regime for all countries in the post-Bretton Woods period. In Appendix Table 4.1, a dummy variable for the Cold War is included in the models for the post-Bretton Woods period. The end of the Cold War led to a systemwide shift in the power distribution in international politics, potentially affecting alliance patterns. As the results in Appendix Table 4.1 show, the defense pact alliance remains quite robust in hypothesis testing, except for the choice between pegging and the flexible regime for all countries. In fact, comparing the magnitudes of the variable with those in Table 1.2, the inclusion of the Cold War dummy actually renders the estimates of the defense pact alliance larger. Hence, the effect of the defense

178

Q. LI

pact alliance in Table 1.2 is not an artifact of the Cold War. In Appendix Tables 5.1 and 5.2, a U.S. dummy, coded one for any dyad that involves the United States and zero otherwise, is included. The United States was at the center of both the Bretton Woods system and a network of international alliances during the first period. After the Bretton Woods system, the U.S. continued to occupy an extremely influential position in the international monetary system. One might argue that the results for the defense pact variable could be an artifact of the dominant influence of the U.S. and its alliance patterns. On the other hand, the inclusion of the U.S. dummy actually reduces the scope of the question, as our interest is to detect the effect of alliance ties of all countries, including those of the United States. Hence, this robustness test of the defense pact alliance to the U.S. influence is quite strong. As the appendix tables show, the U.S. dummy is statistically significant and negative for all models. As we expect, the results for the defense pact alliance appear weaker than those in Tables 1.1 and 1.2 in terms of hypothesis testing, but they remain broadly consistent with those in Tables 1.1 and 1.2 except for the choice between pegging and the flexible regime for all countries only. In Appendix Tables 6.1 and 6.2, the colonial ties between the potential anchor and the potential pegging countries are controlled for, as one may argue that a former colony may be more likely to peg to its former colonizer(s) given special their economic ties. The variable is coding as 1 if a colonial tie ever existed between two countries in a dyad in the samples. Data are from Hensel (1999). As the Appendix tables show, the colonial ties variable is statistically significant and positive for all models in both tables. Given colonial experience, a country is actually less likely to peg to its former colonizer and more likely to adopt the floating regime. This is consistent with the fact that colonies often fight wars with their colonizers, generating enmity. The hypothesis testing results for the defense pact variable remain consistent with those in Tables 1.1 and 1.2, except for the choice between pegging and the flexible regime for all countries in the post-Bretton Woods period. In sum, the only statistically insignificant result for the defense pact alliance variable in the analyses above is its effect on the choice between pegging and the flexible regime for all countries only. This result is consistent with the insignificant substantive effect of the defense pact alliance variable in Tables 1.1 and 1.2 over the choice between pegging and the flexible regime for all countries. (Recall the small relative risk scores 1.08 and 1.14 for the two periods in Table 1.3, discussed in the previous section). Overall, the sensitivity analysis indicates that results in Tables 1.1 and 1.2 are robust.16 CONCLUSION The premise of the paper is that an exchange-rate arrangement by nature involves more than one country and has economic and political implications. Hence, two issues arise. First, an exchange-rate regime may concern two choices—the decision between the flexible and fixed regimes and the decision over the anchor currency. Second, although these two choices are fundamentally motivated by economic considerations, international politics also plays a role. In an effort to explore the linkage between international politics and international economics, I examine how security

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

179

alliances affect the dual choices in an exchange-rate arrangement. Alliance ties increase the ex ante attractiveness of pegging to one’s ally, because security ties reduce concerns over relative gains, motivate active collaboration on the part of the anchorcurrency ally to defend the fixed regime, and provide a signal to the currency market about the durability of the exchange parity. Hence, a country is biased toward pegging to its ally, relative to either pegging to a nonally or choosing the flexible regime. Heterogeneity in the level of commitment and compatibility of interests between allies further leads to the expectation that the defense pact alliance should have a larger effect over these exchange regime choices than the neutrality and entente alliances. The argument is tested in empirical analyses for both the Bretton Woods and postBretton Woods periods, for all potential pegging countries and the developing ones alone. The empirical test is limited to the period till 1992 due to the availability of alliance data, hence missing the controversy over the fixed-flexible regime choice in much of the 1990s. Some argue that a whole different dynamic seems to be at work since 1992 and that capital flows of the 1990s have transformed the choices of nations. While future tests should expand the temporal domain, there is no clear reason why my findings about the effects of alliance should weaken or disappear. I believe that the patterns I find are robust since international politics still has a tremendous impact over international economics even in the new digital economy. The findings suggest that the defense-pact alliance shifts the potential pegging country toward choosing the fixed regime over the flexible regime and toward pegging to the currency of its ally rather than that of a nonally. These effects are stronger for the developing potential pegging countries. In contrast, entente or neutrality pact alliances have no statistically significant effects over a country’s exchange-rate arrangements. These findings stand up to various statistical controls and alternative model specifications. The uncovering of the defense-pact alliance effects results from a new conceptualization of a country’s exchange-rate regime choice and from the use of appropriate statistical techniques. In this new conceptualization, the exchange-rate regime choice consists of two interdependent decisions: the choice between the flexible and fixed regimes and the choice over the anchor currency. The inclusion of the anchor currency choice explicitly points to the relevance of interstate political relations, producing a model and findings beyond the pegging country’s attributes alone. The empirical analyses are consistent with this conceptualization of interdependent decisions. The constrained multinomial logit model generates a nuance estimation of differentiated effects with respect to the two decisions. Without the conceptualization of two interdependent decisions, security alliances would appear irrelevant to a country’s exchange-rate arrangement, as implied by the country attributes-based models. Although the defense pact alliance effects are shown to be statistically significant and substantively important for the exchange-rate arrangement in this analysis, will they remain relevant when economic globalization continues to progress rapidly? At least two consequences of economic globalization may affect how international politics interacts with international economics. First, accompanying the growing integration of goods and capital markets is the possibility of currency union, such as the European Monetary Union (EMU), that unifies interstate exchange-rate arrangements through a single currency. Participation in a monetary union makes the choice over

180

Q. LI

exchange-rate arrangements a nonissue. Second, because globalization gives salience to economic issues and facilitates the reduction of military conflict, security concerns may matter less today than during the Cold War and, consequently, related exchange rate manipulations may also become rarer. Declining security concerns may lead a country to become a less responsive ally in economic affairs, reducing the benefits of pegging to the ally. Although these two scenarios appear to be plausible and could potentially weaken the alliance effects over exchange-rate regime choices, both the prospect and history of economic globalization suggest caution. Monetary integration is still limited to a small number of countries. As long as these states are not integrated politically, alliance ties among them affect the sustainability of the monetary union. In addition, monetary union, as an extremely strong form of interstate cooperation, is unlikely to spread throughout an anarchical international system of sovereign states. Finally, history tells us that the trend of economic globalization is not irreversible. In many ways, current economic globalization resembles the nineteenth century trade expansion and capital market integration, but that first wave of globalization was later disrupted by two world wars and a global economic depression. Until international politics abandons the organizing principle of an anarchical system of sovereign states, security relations between countries will remain relevant to international economics. NOTES 1. A few studies (e.g., Melvin, 1985; Bernhard and Leblang, 1999) treat the problem as a trichotomous choice empirically. It is worth noting that treating the exchange-rate regime choice as a binary problem greatly reduces the wide variety of exchange-rate arrangements. For example, the fixed regime can cover a lot of quite different arrangements, with some larger or smaller intervention band, explicit or discretionary intervention criteria, with or without a currency board or a formal relationship between the anchor and member currencies. Similarly, the flexible regime may also have various alternative arrangements. The simplifying assumption of a binary choice between the flexible and fixed regimes, which is adopted in most previous studies as well as partly in this analysis, renders the complex problem analytically tractable. 2. In fact the problem can be further complicated, as discussed in note 1. For the purpose of this analysis, however, further complication is not necessary. 3. See Cohen (1998) and Eichengreen (1995) for reviews on the effects of politics. 4. New Zealand is a useful illustration. Like other small open economies, New Zealand preferred the fixed regime. However, its trade was divided almost equally among three major trading partners that were also its potential anchor-currency countries (the U.S., U.K., and Australia). The trade advantage of fixing to a particular currency appeared low, and New Zealand eventually chose the flexible regime. I thank Thomas Willett for suggesting this example. 5. Applying the battle of sexes game helps to illustrate that the regime choice concerns problems in international cooperation. However, the simply game does not offer an analytical solution to the regime choice problem itself because both Cooper and Hamada do not consider a structural model of economic behavioral relations in the simple 2x2 game. Moreover, as Cooper (1975, 1999) points out, a comprehensive game theoretic analysis of negotiations over international monetary system faces insurmountable difficulties because of disagreements over distribution of gains, differences in preferences, and differences in contextual conditions. 6. Focusing on the European monetary cooperation, many recent studies (e.g., Eichengreen and Frieden 1994; Sandholtz, 1993) examine why the European countries gave up their autonomy in favor of a monetary union. The EMU case resembles, but is more demanding than, choosing an anchor currency. In addition, the empirical scope of EMU is much more limited than that in this analysis.

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

181

7. As Kirshner (1995, pp. 61–62) paraphrased an economist’s opinion, “Allied support of the Chinese currency, first through silver purchases and then through direct exchange stabilization support, was ‘invaluable.’ Without such support, the Chinese currency would have collapsed, and China would have probably been knocked out of the war before Pearl Harbor.” See Kirshner (1995) for details on this and other cases. 8. It is worth noting that the choices over a country’s exchange-rate regime are typically within the jurisdiction of the government rather than the (independent) central bank (Henning, 1994), further facilitating the linkage between security and economic issues. 9. From 1938 to 1948, Britain’s gold and dollar reserves dropped from 864 million pounds to 369 million pounds while its sterling liabilities expanded from 760 million pounds to 2,700 million pounds (Kirshner, 1995, p. 144). Britain was able to expand its sterling liabilities for war financing in the presence of declining reserves because the pegging countries supported by holding British currency and sterling securities. 10. Sobel (1999) makes a similar information argument that property rights and democratic institutions reduce uncertainty and inform the international capital market about the equilibrium path of a country’s economy. 11. Australia, India, Spain, Italy, South Africa, and Singapore as potential anchors count only 8.5 percent of the post-Bretton Woods sample and about 7 percent of the Bretton Woods sample. 12. The IMF coding scheme sometimes generates differences between the self reported choice and the de facto practice. The IMF has noticed this weakness and reported since 1998 or 1999 “corrected” data that recognizes the difference between the self-reported de jure exchange regimes of nations and the de facto choices they have made (as reflected in variation in actual exchange rates and reserves). Some preliminary alternative data have been used in some studies. See, for example, Levy-Yeyati and Sturzenegger (2002). While such measurement errors may be random and do not affect the estimates, future analysis based on alternative data coding is desirable for robustness check. 13. I also examine the effect of an alternative measure, the inflation rate differential between the potential anchor-currency country and the possible pegging country in a dyad. Nations would choose as anchor a country with a similar level of inflation in order to reduce the need for foreign exchange intervention because nominal movements mirror real movements when inflation rates are similar. The statistical results for alliance variables remain robust. For robustness check, I also examined another alternative measure: the inflation differential between the potential anchor-currency country and the world average. The inferences remain unchanged. Heller (1978) uses a similar measure in his analysis. 14. Estimation is conducted using STATA7. The Mills ratio formula is mills = exp(-0.5*phat2)/ (sqrt(2*_pi)*normprob(phat)), where phat is the linear preditor from the probit model. See: http:// www.stata.com/support/faqs/stat/mills.html for details. 15. This corresponds to the two-step Heckman (1979) selection model. For an example of applications of the Heckman selection model in international relations, see Reed (2000). 16. A further robustness test is to estimate the models based on samples with the U.S. as the only potential anchor. Statistical results from this test turn out to be consistent with those in Tables 1.1 and 1.2. While selecting the United States as the only potential anchor rules out possible spurious findings due to the selection of potential anchors, the test is narrow and inferences are simply not generalizable.

REFERENCE Bennett, D. Scott and Allan C. Stam (2000). “EUGene: A Conceptual Manual.” International Interactions, Vol. 26, pp. 179–204. Bernhard, William and David Leblang (1999). “Democratic Institutions and Exchange Rate Commitments.” International Organization, Vol. 53 No. 1, pp. 71–97. Bosco, L. (1987). “Determinants of the Exchange Rate Regimes in LDCs: Some Empirical Evidence.” Economic Notes, Vol. 16, No. 1, pp. 119–143. Broz, Lawrence (2002). “Political System Transparency and Monetary Committment Regimes.” International Organization, Vol. 56, No. 4, pp. 863–889. Bueno de Mesquita, Bruce (1981). The War Trap, New Haven, CT: Yale University Press.

182

Q. LI

Cohen, Benjamin (1997). “The Political Economy of Currency Regions,” in Edward D. Mansfield, and Helen V. Milner, eds., The Political Economy of Regionalism, New York: Columbia University Press. pp. 51–76. ——— (1998). The Geography of Money, Ithaca, NY: Cornell University Press. Conybeare, John A. C. (1992). “A Portfolio Diversification Model of Alliances: The Triple Alliance and Triple Entente, 1879–1914.” Journal of Conflict Resolution, Vol. 36, No. 1, pp. 53–85. Cooper, Richard N. (1975). “Prolegomena to the Choice of an International Monetary System.” International Organization, Vol. 29, No. 1, pp. 63–97. ——— (1999). “Exchange rate choices,” Harvard University unpublished manuscript. Edwards, Sebastian (1996). “The Determinants of Choice Between Fixed and Flexible Exchange Rate Regimes.” National Bureau of Economic Research Working paper 5756. Eichengreen, Barry (1992). Gold Fetters: The Gold Standard and the Great Depression, New York: Oxford University Press. ——— (1995). “The Endogeneity of Exchange Rate Regimes,” in Peter B. Kenen, ed., Understanding Interdependence, Princeton, NJ: Princeton University Press, pp. 1–33. ——— (1996). Globalizing Capital: A History of the International Monetary System, Princeton, NJ: Princeton University Press. Eichengreen, Barry, and Jeffrey Frieden (1994). The Political Economy of European Monetary Unification, Boulder, CO: Westview Press. Fearon, James D. (1997). “Signaling Foreign Policy Interests: Tying Hands Versus Sinking Costs.” Journal of Conflict Resolution, Vol. 41, No. 1, pp. 68–90. Frieden, Jeffrey (1991). “Invested Interests: The Politics of National Economic Policies in a World of Global Finance.” International Organization, Vol. 45, No. 4, pp. 425–451. Gowa, Joanne (1983). Closing the Gold Window: Domestic Politics and the End of Bretton Woods, Ithaca, NY: Cornell University Press. ——— (1994). Allies, Adversaries, and International Trade, Princeton, NJ: Princeton University Press. Hamada, Koichi (1976). “A Strategic Analysis of Monetary Interdependence.” Journal of Political Economy, Vol. 84, August, pp. 677–700. Heckman, James J. (1979). “Sample Selection Bias As a Specification Error.” Econometrica, Vol. 47, No. 1, pp. 153–162. Hefeker, Carsten (1997). “Interest Groups and Monetary Integration: the Political Economy of Exchange Regime Choice,” in Thomas Willett, ed., The Political Economy of Global Interdependence, Boulder, CO: Westview Press. Heller, R. (1978). “Determinants of Exchange Rate Practices.” Journal of Money, Credit, and Banking, Vol. 10, August, pp. 308–321. Henning, C. Randall (1994). Currencies and Politics in the United States, Germany and Japan, Washington, DC: Institute for International Economics. Hensel, Paul (1999). “ICOW Colonial History Data Set.” Florida State University, Version 0.1, http://garnet.acns.fsu.edu/~phensel/icowdata.html#colonies. Holden, P., M. Holden, and E. Suss (1979). “The Determinants of Exchange Rate Flexibility: An Empirical Investigation.” Review of Economics and Statistics, Vol. 61, pp. 327–333. International Monetary Fund (Various years). Annual Reports on Exchange Arrangements and Exchange Controls, Washington DC: International Monetary Fund. ——— (Various years). Direction of Trade, Washington D. C.: International Monetary Fund. ——— (Various years). International Financial Statistics Yearbook, Washington DC: International Monetary Fund. Kindleberger, Charles P. (1986). “International Public Goods Without International Government.” American Economic Review, Vol 76, No. 1, pp. 1–13. Kirshner, Jonathan (1995). Currency and Coercion: The Political Economy of Monetary Power,

SECURITY ALLIANCES AND EXCHANGE-RATE REGIME CHOICES

183

Princeton, NJ: Princeton University Press. Krugman, Paul (1979). “A Model of Balance-of-Payments Crises.” Journal of Money, Credit and Banking, Vol. 11, pp. 311–325. Krugman, Paul R. (1995). “What Do We Need to Know About The International Monetary System?” in Peter B. Kenen, ed., Understanding Interdependence, Princeton, NJ: Princeton University Press, pp. 509–530. Leblang, David (1999). “Domestic Political Institutions and Exchange Rate Commitments in the Developing World.” International Studies Quarterly, Vol. 43, pp. 599–620. Leeds, Brett Ashley, Andrew G. Long, and Sara McLaughlin Mitchell (2000). “Reevaluating Alliance Reliability: Specific Threats, Specific Promises.” Journal of Conflict Resolution, Vol. 44, pp. 686–699. Levy Yeyati, Eduardo and Federico Sturzenegger (2002). “Classifying Exchange Rate Regimes: Deeds vs. Words,” unpublished manuscript at http://www.utdt.edu/~ely/ papers.html. Mansfield, Edward, and Rachel Bronson (1997). “Alliances, Preferential Trading Arrangements, and International Trade.” American Political Science Review, Vol. 91, No. 1, pp. 94–107. McKinnon, Ronald (1993). “Rules of the Game: International Money in Historical Perspective.” Journal of Economic Literature, Vol. 31, No. 1, pp. 1–44. Melvin, Michael (1985). “The Choice of an Exchange Rate System and Macroeconomic Stability.” Journal of Money, Credit and Banking, Vol. 17, No. 4, pp. 467–478. Moon, Bruce E. (1982). “Exchange Rate System, Policy Distortions, and the Maintenance of Trade Dependence.” International Organization, Vol. 36, No. 4, pp. 715–739. Morrow, James D. (1991). “Alliances and Asymmetry: An Alternative to the Capability Aggregation Model of Alliances.” American Journal of Political Science, Vol. 35, No. 4, pp. 904–933. ——— (1994). “Alliance, Credibility, and Peacetime Costs.” Journal of Conflict Resolution, Vol. 38, No. 2, pp. 270–294. Moul, William Brian (1988). “Balances of Power and the Escalation of War of Serious Disputes Among the European Great Powers, 1815–1939: Some Evidence.” American Journal of Political Science, Vol. 32, No. 2, pp. 241–275. Mundell, R. A. (1961). “A Theory of Optimum Currency Areas.” American Economic Review, Vol. 51, pp. 657–665. Obstfeld, Maurice (1994). “The Logic of Currency Crises.” Banque de France—Cahiers économiques et monétaires, Vol. 43, pp. 189–213. Oye, Kenneth A. (1985). “The Sterling-Dollar-Franc Triangle: Monetary Diplomacy 1929– 1937.” World Politics, Vol. 38, No. 1, pp. 173–199. Reed, William (2000). “A Unified Statistical Model of Conflict Onset and Escalation.” American Journal of Political Science, Vol. 44, No. 1, pp. 84–93. Sandholtz, Wayne (1993). “Choosing Union: Monetary Politics and Maastricht.” International Organization, Vol. 47, No. 1, pp. 1–39. Simmons, Beth (1994). Who Adjusts? Domestic Sources of Foreign Economic Policy During the Interwar Years, Princeton, NJ: Princeton University Press. Singer, J. David, and Melvin Small (1966). “Formal Alliances, 1815–1939: A Quantitative Description.” Journal of Peace Research, Vol. 3, No. 1, pp. 1–32. Smith, Alastair (1995). “Alliance Formation and War.” International Studies Quarterly, Vol. 39, No. 4, pp. 405–425. Sobel, Andrew C. (1999). State Institutions, Private Incentives, Global Capital. Ann Arbor, MI: The University of Michigan Press. Walt, Stephen M. (1987). The Origins of Alliances, Ithaca, NY: Cornell University Press. Waltz, Kenneth N. (1979). Theory of International Politics, New York: McGraw-Hill, Inc.

184

Q. LI

Wiggins, Vince (1999). “Comparing CTGLS with Regress Cluster ( ),” STATA Corporation, URL: www.stata.omc/support/faqs/stata/xtgls_rob.html. Willett, Thomas D., and John Mullen (1982). “The Effects of Alternative International Monetary Systems on Macroeconomic Discipline and Inflation Biases,” in R. E. Lombra and W. E. Witte, eds., Political Economy of International Domestic Monetary Relations, Ames: Iowa State University Press. World Bank. (1999). The 1999 World Development Indicators CD-ROM, Washington, DC: World Bank.

CONTRIBUTOR Quan Li is an Assistant Professor of Political Science at the Pennsylvania State University, University Park. His research appears in International Organization, International Studies Quarterly, Journal of Politics, British Journal of Political Science, Comparative Political Studies, Journal of Peace Research, Political Research Quarterly, and elsewhere.

Appendix: Probit Estimates of Determinants of Fixed Regime Choice Based on Country Attributes (A Country-Level Analysis, 1966–1995) Parameter Estimates Size -0.2805*** Economic Development 0.0001** Trade Openness 0.0025 Inflation Credibility -0.0001 Economic Growth Rate -0.0039 Capital Control 0.3067 Democracy Level -0.0430*** Constant 6.6673*** Pseudo R2 0.28 Observations 2,485 • Yearly dummy variables not reported. • significant at 10%; ** significant at 5%; *** significant at 1%.

Robust Standard Errors (0.0004) (0.0316) (0.4509) (0.2020) (0.6498) (0.1728) (0.0001) (0.0000)

Variable Description and Data Sources: • Size: Real GDP (PPP), Penn World Table Mark 5.6. • Economic Development: Real GDP (PPP) Per Capita, Penn World Table Mark 5.6. • (Exports+Imports)/GDP, 1999 World Development Indicators. • Inflation Differential (against world average), computed using GDP deflator-based inflation rates from 1999 World Development Indicators. • Economic Growth Rate: Annual percentage Change in Real GDP PPP, 1999 World Development Indicators. • Capital Control: capital account restrictions dummy, IMF Annual Reports on Exchange Arrangements and Controls. • Democracy Level: Democracy score–autocracy score, Polity III.

185

1.9177*** (0.1880) 3396 163 0.18

-0.0413*** (0.0069) 0.0758*** (0.0117)

-1.1336*** (0.2526)

-0.0380*** (0.0114) 0.0773*** (0.0117) -0.0887*** (0.0117) 1.3797*** (0.2253) 3396

-1.2333*** (0.2786)

Flexible Regime

1.8005*** (0.1821) 3396 197 0.19

-1.7548*** (0.2958) -0.0298 (0.4723) -0.0367*** (0.0070) 0.0870*** (0.0124) -1.3572*** (0.2864) -0.8938 (0.6389) -0.0359*** (0.0117) 0.0885*** (0.0124) -0.0870*** (0.0113) 1.2635*** (0.2198) 3396

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

N Wald Test (χ2) Pseudo R2

Constant

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

2.0730*** (0.2046) 3004 143 0.19

-0.0419*** (0.0072) 0.0603*** (0.0115)

-1.3180*** (0.2854)

Fixed and No Pegging

-0.0445*** (0.0112) 0.0614*** (0.0115) -0.0791*** (0.0113) 1.5389*** (0.2412) 3004

-1.7400*** (0.3786)

Flexible Regime

Model 3

1.8726*** (0.1935) 3004 198 0.22

-2.8363*** (0.4665) 0.1575 (0.5770) -0.0306*** (0.0075) 0.0715*** (0.0123)

Fixed and No Pegging

-2.3789*** (0.4766) -0.5332 (0.6908) -0.0360*** (0.0119) 0.0726*** (0.0123) -0.0789*** (0.0112) 1.3538*** (0.2314) 3004

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 2.1 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1966–1973 (Fixed and Pegging as Baseline Category; Prior Choice Excluded from Model)

186

3.0721*** (0.2262) 11017 141 0.14

-0.0558*** (0.0084) 0.0242*** (0.0060)

-1.0306*** (0.2940)

-0.0288*** (0.0069) 0.0239*** (0.0060) -0.4571*** (0.0586) 2.9328*** (0.2538) 11017

-0.2964 (0.2873)

Flexible Regime

3.0283*** (0.2234) 11017 155 0.14

-1.5474*** (0.3404) 0.4686 (0.6267) -0.0500*** (0.0085) 0.0222*** (0.0060) -0.4772* (0.3092) 0.6162 (0.6502) -0.0260*** (0.0072) 0.0219*** (0.0061) -0.4404*** (0.0585) 2.8705*** (0.2497) 11017

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

N Wald Test (χ2) Pseudo R2

Constant

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

3.0589*** (0.2318) 10039 148 0.15

-0.0567*** (0.0086) 0.0248*** (0.0057)

-1.4456*** (0.2986)

Fixed and No Pegging

-0.0266*** (0.0067) 0.0246*** (0.0058) -0.4996*** (0.0574) 2.9312*** (0.2574) 10039

-0.6331** (0.2891)

Flexible Regime

Model 3

2.9657*** (0.2265) 10039 190 0.16

-2.7011*** (0.4167) 0.4343 (0.6310) -0.0456*** (0.0091) 0.0227*** (0.0060)

Fixed and No Pegging

-0.8686*** (0.3167) 0.4391 (0.6548) -0.0221*** (0.0072) 0.0225*** (0.0061) -0.4792*** (0.0564) 2.8312*** (0.2505) 10039

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 2.2 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1974–1992 (Fixed and Pegging as Baseline Category; Prior Choice Excluded from Model)

187

0.4124*** (0.0791) -3.1745*** (0.2155) 2.7019*** (0.1729) 3352 721 0.40

-0.0336*** (0.0064) 0.0995*** (0.0212)

-0.9145*** (0.2238)

-0.0391*** (0.0144) 0.1003*** (0.0214) -0.0735*** (0.0091) 0.3518*** (0.0972) -1.2846*** (0.3304) 1.5293*** (0.2694) 3352

-1.0910*** (0.3018)

Flexible Regime

0.4078*** (0.0789) -3.1144*** (0.2181) 2.6097*** (0.1728) 3352 798 0.41

-1.2980*** (0.2501) -0.1901 (0.4788) -0.0311*** (0.0065) 0.1054*** (0.0217) -1.1653*** (0.3099) -0.8200 (0.7805) -0.0383*** (0.0148) 0.1062*** (0.0219) -0.0727*** (0.0088) 0.3479*** (0.0963) -1.2515*** (0.3328) 1.4433*** (0.2722) 3352

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

Observations Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Regional Effect

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

0.3396*** (0.0800) -3.0005*** (0.2276) 2.8249*** (0.1800) 2963 646 0.39

-0.0344*** (0.0063) 0.0862*** (0.0199)

-1.2161*** (0.2631)

Fixed and No Pegging

-0.0439*** (0.0133) 0.0866*** (0.0202) -0.0647*** (0.0084) 0.2838*** (0.1059) -1.2384*** (0.3438) 1.7033*** (0.2803) 2963

-1.6277*** (0.4331)

Flexible Regime

Model 3

0.3753*** (0.0856) -2.9139*** (0.2305) 2.6136*** (0.1765) 2963 667 0.41

-2.4918*** (0.3990) 0.0795 (0.6522) -0.0262*** (0.0069) 0.0920*** (0.0213)

Fixed and No Pegging

-2.4345*** (0.5230) -0.4190 (0.8953) -0.0374*** (0.0144) 0.0924*** (0.0216) -0.0643*** (0.0084) 0.3146*** (0.1094) -1.2117*** (0.3476) 1.5131*** (0.2812) 2963

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 3.1 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1966–1973 (Fixed and Pegging as Baseline Category; Regional Effect Included in Model)

188

0.0803*** (0.0215) -2.1624*** (0.1135) 3.1219*** (0.2630) 10945 1089 0.54

-0.0478*** (0.0074) 0.0361*** (0.0078)

-0.5743** (0.3002)

-0.0375*** (0.0080) 0.0364*** (0.0078) -0.2429*** (0.0388) 0.1331*** (0.0231) 0.7752*** (0.0743) 0.1911 (0.2979) 10945

-0.1855 (0.2977)

Flexible Regime

0.0826*** (0.0211) -2.1319*** (0.1146) 3.0567*** (0.2584) 10945 1110 0.54

-0.9068*** (0.3361) 0.5175 (0.5856) -0.0445*** (0.0076) 0.0358*** (0.0081) -0.3120 (0.3245) 0.5568 (0.6046) -0.0361*** (0.0083) 0.0360*** (0.0081) -0.2400*** (0.0392) 0.1347*** (0.0230) 0.7919*** (0.0753) 0.1394 (0.2925) 10945

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

Observations Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Regional Effect

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

0.0939*** (0.0211) -2.2294*** (0.1194) 3.0518*** (0.2588) 9967 1104 0.53

-0.0476*** (0.0076) 0.0373*** (0.0076)

-1.0549*** (0.2979)

Fixed and No Pegging

-0.0349*** (0.0082) 0.0375*** (0.0076) -0.2831*** (0.0397) 0.1526*** (0.0238) 0.6620*** (0.0779) 0.1669 (0.3067) 9967

-0.6268** (0.3083)

Flexible Regime

Model 3

0.1014*** (0.0205) -2.1635*** (0.1220) 2.8881*** (0.2518) 9967 1148 0.54

-2.0169*** (0.3780) 0.5200 (0.5957) -0.0398*** (0.0079) 0.0373*** (0.0082)

Fixed and No Pegging

-0.8558*** (0.3638) 0.3621 (0.6124) -0.0320*** (0.0090) 0.0375*** (0.0082) -0.2832*** (0.0404) 0.1560*** (0.0238) 0.6939*** (0.0800) 0.0772 (0.3002) 9967

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 3.2 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1974-1992 (Fixed and Pegging as Baseline Category; Reigonal Effect Included in Model)

189

-0.2386 (0.1850) -2.0731*** (0.1075) 3.9148*** (0.2903) 10945 956 0.54

-0.0469*** (0.0071) 0.0245*** (0.0066)

-0.5660** (0.2958)

-0.0331*** (0.0072) 0.0247*** (0.0066) -0.2394*** (0.0361) -1.2265*** (0.2003) 0.9449*** (0.0584) 2.1571*** (0.3113) 10945

-0.1874 (0.2775)

Flexible Regime

-0.2439 (0.1851) -2.0427*** (0.1080) 3.8754*** (0.2892) 10945 980 0.54

-0.8896*** (0.3350) 0.4913 (0.5860) -0.0436*** (0.0072) 0.0238*** (0.0067) -0.3067 (0.2973) 0.4918 (0.5990) -0.0317*** (0.0075) 0.0240*** (0.0067) -0.2360*** (0.0362) -1.2335*** (0.2015) 0.9608*** (0.0591) 2.1258*** (0.3104) 10945

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

Observations Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Cold War

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

-0.3541* (0.2062) -2.1170*** (0.1148) 4.0364*** (0.3210) 9967 958 0.53

-0.0464*** (0.0072) 0.0242*** (0.0062)

-1.0446*** (0.2920)

Fixed and No Pegging

-0.0299*** (0.0073) 0.0244*** (0.0062) -0.2833*** (0.0380) -1.3129*** (0.2169) 0.8488*** (0.0605) 2.3730*** (0.3380) 9967

-0.5798** (0.2798)

Flexible Regime

Model 3

-0.3380 (0.2065) -2.0498*** (0.1159) 3.9231*** (0.3172) 9967 1033 0.53

-1.9833*** (0.3706) 0.4687 (0.5940) -0.0390*** (0.0075) 0.0234*** (0.0063)

Fixed and No Pegging

-0.7510*** (0.3143) 0.2551 (0.6009) -0.0274*** (0.0079) 0.0236*** (0.0063) -0.2796*** (0.0381) -1.3092*** (0.2187) 0.8793*** (0.0617) 2.3012*** (0.3335) 9967

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 4.1 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1974–1992 (Fixed and Pegging as Baseline Category; Cold War Dummy Included in Model)

190

-1.9174*** (0.1290) 3.7263*** (0.2852) 3352 933 0.49

-0.0093 (0.0069) 0.0518*** (0.0148) -3.7394*** (0.3196)

-0.0868 (0.1972)

-0.0074 (0.0114) 0.0527*** (0.0151) -3.1881*** (0.2913) -0.0780*** (0.0101) -0.4574*** (0.1594) 2.5016*** (0.3202) 3352

-0.1552 (0.2300)

Flexible Regime

-1.8981*** (0.1286) 3.6723*** (0.2851) 3352 1012 0.50

-0.3746** (0.2205) 0.5182 (0.4891) -0.0074 (0.0072) 0.0544*** (0.0147) -3.7064*** (0.3217) -0.1715 (0.2442) -0.0242 (0.7529) -0.0068 (0.0118) 0.0552*** (0.0150) -3.1836*** (0.2938) -0.0770*** (0.0098) -0.4510*** (0.1619) 2.4541*** (0.3233) 3352

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

Observations Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Pegger’s Attributes

U.S.

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

-1.9596*** (0.1292) 3.7658*** (0.3010) 2963 754 0.48

-0.0098 (0.0067) 0.0454*** (0.0149) -3.5954*** (0.3411)

-0.4170** (0.2504)

Fixed and No Pegging

-0.0127 (0.0099) 0.0460*** (0.0152) -3.0764*** (0.3366) -0.0681*** (0.0099) -0.5893*** (0.1725) 2.5851*** (0.3417) 2963

-0.6297* (0.3843)

Flexible Regime

Model 3

-1.9314*** (0.1224) 3.6314*** (0.2973) 2963 801 0.49

-1.3858*** (0.3555) 0.7153 (0.6418) -0.0031 (0.0079) 0.0505*** (0.0152) -3.5195*** (0.3469)

Fixed and No Pegging

-1.0878** (0.5041) 0.2852 (0.8571) -0.0077 (0.0114) 0.0511*** (0.0154) -3.0685*** (0.3426) -0.0676*** (0.0098) -0.5685*** (0.1640) 2.4617*** (0.3378) 2963

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 5.1 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1966–1973 (Fixed and Pegging as Baseline Category; U.S. Dummy Included in Model)

191

-1.9711*** (0.1105) 4.1628*** (0.2706) 10945 1317 0.55

-0.0322*** (0.0070) 0.0059** (0.0024) -1.7594*** (0.3034)

-0.1769 (0.2711)

-0.0184** (0.0076) 0.0062** (0.0024) -1.5393*** (0.3115) -0.2882*** (0.0379) 0.9666*** (0.0632) 1.7620*** (0.2904) 10945

0.1083 (0.2569)

Flexible Regime

-1.9590*** (0.1093) 4.1297*** (0.2711) 10945 1327 0.55

-0.4096* (0.3091) 0.7089 (0.5670) -0.0301*** (0.0071) 0.0055** (0.0023) -1.7260*** (0.3064) 0.0012 (0.2742) 0.7560 (0.5880) -0.0170** (0.0077) 0.0058** (0.0024) -1.5373*** (0.3116) -0.2851*** (0.0380) 0.9708*** (0.0630) 1.7334*** (0.2899) 10945

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

Observations Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Pegger’s Attributes

U.S.

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

-1.9853*** (0.1200) 4.1539*** (0.2762) 9967 1339 0.54

-0.0284*** (0.0069) 0.0068*** (0.0026) -1.8869*** (0.3224)

-0.7286*** (0.2644)

Fixed and No Pegging

-0.0140* (0.0078) 0.0071*** (0.0026) -1.5777*** (0.3313) -0.3365*** (0.0394) 0.8992*** (0.0670) 1.8764*** (0.2943) 9967

-0.2666 (0.2582)

Flexible Regime

Model 3

-1.9467*** (0.1190) 4.0606*** (0.2747) 9967 1356 0.54

-1.5726*** (0.3374) 0.5919 (0.5610) -0.0224*** (0.0073) 0.0065** (0.0026) -1.8091*** (0.3278)

Fixed and No Pegging

-0.3847* (0.2878) 0.4421 (0.5832) -0.0121 (0.0082) 0.0068*** (0.0026) -1.5660*** (0.3361) -0.3311*** (0.0392) 0.9158*** (0.0672) 1.8113*** (0.2902) 9967

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 5.2 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1974–1992 (Fixed and Pegging as Baseline Category; U.S. Dummy Included in Model)

192

-0.0611*** (0.0139) 0.1390*** (0.0217) 0.0000 (0.0000) 3.9907*** (1.2312) -2.5920*** (0.1728) 2.5836*** (0.1604) 3352 488 0.41

-0.5893*** (0.2510)

-0.0510*** (0.0178) 0.1398*** (0.0219) -0.0784*** (0.0104) 3.0874** (1.2097) -0.7069*** (0.1722) 1.3205*** (0.2332) 3352

-0.7521*** (0.2859)

Flexible Regime

-0.8508*** (0.2822) -0.1481 (0.5030) -0.0585*** (0.0137) 0.1427*** (0.0217) 0.0000 (0.0000) 3.8347*** (1.2414) -2.5663*** (0.1741) 2.5328*** (0.1686) 3352 526 0.41 -0.7298*** (0.2980) -0.7872 (0.7977) -0.0507*** (0.0184) 0.1435*** (0.0219) -0.0774*** (0.0101) 3.0541** (1.2102) -0.6982*** (0.1770) 1.2706*** (0.2443) 3352

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

Observations Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Colonial Ties

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

-0.0717*** (0.0198) 0.1216*** (0.0219) 0.0000 (0.0000) 4.3408*** (1.3910) -2.4677*** (0.1862) 2.7372*** (0.1790) 2963 408 0.41

-0.7052** (0.3407)

Fixed and No Pegging

-0.0660*** (0.0198) 0.1220*** (0.0221) -0.0677*** (0.0100) 3.4819*** (1.3119) -0.7740*** (0.1884) 1.5382*** (0.2473) 2963

-1.0690*** (0.4341)

Flexible Regime

Model 3

-1.6035*** (0.4567) 0.1069 (0.7322) -0.0626*** (0.0185) 0.1261*** (0.0220) 0.0000 (0.0000) 3.8964*** (1.4046) -2.4107*** (0.1827) 2.6164*** (0.1895) 2963 464 0.41

Fixed and No Pegging

-1.4893*** (0.5723) -0.3846 (0.9795) -0.0596*** (0.0214) 0.1264*** (0.0222) -0.0671*** (0.0098) 3.1303** (1.3837) -0.7362*** (0.1887) 1.4208*** (0.2621) 2963

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 6.1 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1966–1973 (Fixed and Pegging as Baseline Category; Colonial Ties Included in Model)

193

-0.0628*** (0.0091) 0.0184*** (0.0055) 0.0000 (0.0000) 4.9107*** (1.0191) -2.0046*** (0.1067) 3.6345*** (0.1827) 10945 1010 0.54

-0.2845 (0.3125)

-0.0410*** (0.0081) 0.0187*** (0.0055) -0.2901*** (0.0383) 4.6835*** (1.0604) 0.9537*** (0.0591) 1.2317*** (0.2195) 10945

-0.0238 (0.2943)

Flexible Regime

-0.5328* (0.3624) 0.4636 (0.5451) -0.0598*** (0.0093) 0.0179*** (0.0054) 0.0000 (0.0000) 4.8241*** (1.0187) -1.9858*** (0.1068) 3.6055*** (0.1829) 10945 1026 0.54 -0.1069 (0.3206) 0.4226 (0.5877) -0.0397*** (0.0084) 0.0182*** (0.0054) -0.2871*** (0.0383) 4.6342*** (1.0600) 0.9642*** (0.0597) 1.2045*** (0.2186) 10945

Flexible Regime

Model 2 Fixed and No Pegging

Robust standard errors in parentheses, adjusted for clustering over dyad. * significant at 10%; ** significant at 5%; *** significant at 1%

Observations Wald Test (χ2) Pseudo R2

Constant

Prior Choice

Colonial Ties

Pegger’s Attributes

Anchor Inflation

Trade Ties

Entent Neutrality

Defense Pact

Alliance

Fixed and No Pegging

Model 1

All Potential Pegging Country–Potential Anchor Dyads

-0.0677*** (0.0100) 0.0183*** (0.0053) 0.0000 (0.0000) 5.0976*** (1.0297) -2.0353*** (0.1136) 3.6588*** (0.1870) 9967 973 0.53

-0.7363*** (0.3109)

Fixed and No Pegging

-0.0397*** (0.0085) 0.0186*** (0.0053) -0.3407*** (0.0397) 4.6312*** (1.0730) 0.8740*** (0.0605) 1.3896*** (0.2187) 9967

-0.3917* (0.2968)

Flexible Regime

Model 3

-1.5710*** (0.4015) 0.4256 (0.5378) -0.0600*** (0.0102) 0.0179*** (0.0053) 0.0000 (0.0000) 4.8768*** (1.0267) -1.9922*** (0.1141) 3.5880*** (0.1860) 9967 1024 0.53

Fixed and No Pegging

-0.5137* (0.3377) 0.1713 (0.5757) -0.0374*** (0.0091) 0.0182*** (0.0053) -0.3370*** (0.0397) 4.5237*** (1.0708) 0.8945*** (0.0618) 1.3386*** (0.2159) 9967

Flexible Regime

Model 4

Developing Potential Pegging Country–Potential Anchor Dyads

Appendix Table 6.2 Constrained Multinomial Logit Model of Anchor Currency and Fixed-Flexible Regime Choices, 1974–1992 (Fixed and Pegging as Baseline Category; Colonial Ties Included in Model)