US Foreign Exchange Interventions: Domestic Politics and International Factors

US Foreign Exchange Interventions: Domestic Politics and International Factors Prepared for Handbook of Global International Policy, edited by S. Nagel. Quan Li Assistant Professor Department of Political Science The Pennsylvania State University 107 Burrowes Building University Park, PA 16802-6200 Office: 814.865-6575 Fax: 814.863-8979 Email: quanli@psu.edu, qxl4@psu.edu I. Introduction On August 15, 1971, the United States suspended the convertibility of the dollar into gold. The dollar was then formally devalued twice, in 1972 and 1973, and began to float in 1973. The flexible exchange rate regime was coupled with successive capital decontrols across states, shrinking policy autonomy and the emergence of global financial integration (Frieden, 1991; Andrews, 1994; Cohen, 1996). The effects of domestic macroeconomic policies were transmitted across state borders and policy makers scrambled for new policy tools to handle the monstrous speculation and frequent capital movements that can nullify macroeconomic policy measures. Under these circumstances, foreign exchange rate policy interventions have become one of the most important policy instruments available for external adjustment and stabilizing the international monetary system (Frieden, 1991). In addition, foreign exchange interventions may also affect domestic monetary policy. 1 Recent econometric analysis (Dominguez and Frankel, 1993a, 1 There are two types of foreign exchange interventions: unsterilized and sterilized. In an unsterilized intervention, an intervention can have a real effect on the aggregate money supply. In this sense, it is part of macroeconomic monetary policy. Thus, it can influence interest rates, prices and the economy as a whole. In fact, the US interventions are usually operated in a sterilized manner. That is, in addition to the currency transaction, the authorities will sell government securities to offset the increases in reserve. Under the Bretton Woods system, interventions were undertaken when exchange rates moved out of the parity bands. After Bretton Woods, individual 1

1993b; Dominguez, 1993) has verified the effectiveness of foreign exchange market intervention as an independent policy instrument. Foreign exchange operations by a state can be either through buying and selling home currencies in the open market or through payments between central banks or international agencies such as the IMF. The subject of this paper is the open market currency operations by the United States Treasury or Federal Reserve. When the Treasury or Fed buys US dollars, it actually sells foreign currency and thus depletes reserves. If it sells US dollars, it then increases reserves. Figure 1a and 1b shows how US interventions against the two most important foreign currencies -- the Deutsche Mark and the Japanese Yen -- fluctuated between March 1973 and December 1994. Figure 1a and 1b about here The US intervention activities in the foreign exchange market have exhibited huge variations over time and across currencies. Interventions against the Deutsche Mark were concentrated around the 70s and late 80s; such activities against the Japanese Yen clustered in the late 80s. Across currencies, more interventions have occurred against the Deutsche Mark compared to those against the Japanese Yen. From March 1973 to December 1994, there were on average 3.6718 interventions against the Deutsche Mark per month and only 0.8969 intervention against the Japanese Yen. Apparently, this is related to the fact that the Deutsche Mark has been a more important reserve currency in the international monetary system. The question posed in this article is whether there exist general patterns in the bureaucratic intervention activities that underlie political and economic interactions among and within nation states. Are these patterns attributable to various systemic and statist explanations in the study of state foreign economic policy? Can variations in the intervention activities across different foreign currencies be countries are not obliged to correct the exchange rate misalignment unless they desire to. For further discussion, see Marston (1987) and Lewis (1993). In this paper, I follow the convention and do not differentiate between these two types of interventions. 2

explained by these different theoretical approaches? Is it possible to integrate systemic and statist explanations such that we get a better understanding of state foreign economic policies? In this paper, I seek to answer these questions within the context of the American foreign exchange rate policy intervention activities. My analysis examines American foreign exchange monthly interventions against the Deutsche Mark and the Japanese Yen respectively from March 1973 to December 1994. The focus is upon the complementarity and integration of systemic and statist explanations. 2 A combination of systemic and statist explanations will hopefully provide us with a better story about state foreign economic policy behaviors. II. Integrating Systemic and Statist Explanations American foreign economic policies have been studied through different theoretical approaches at systemic, statist and societal levels of analysis. 3 Despite their differences, scholars of foreign economic policy agree that state institutions are important, and that statist approaches can enhance systemic and societal explanations (Ikenberry, Lake and Mastanduno, 1988). In recent years, much scholarly work has been done to explore the domestic sources of American foreign economic policy behaviors, especially the 2 The societal approach is not employed in this study due to the properties of the issue area, which preconditions possible theoretical synthesis of the different approaches in the study of foreign economic policy making (Cohen, 1994: 151-4). American foreign exchange policymaking is very bureaucratized, and societal interests have a hard time directly accessing the decision making process (Destler and Henning, 1989; Henning, 1994). Most often, societal interests find expression in the policy outcomes only indirectly through the channel of Congress, and so their influence is not directly examined in this paper. One often cited example is the lobby activities of Caterpillar at the Treasury to try to persuade it into taking a more interventionist approach in the early 1980s (Destler and Henning, 1989). However, the Treasury was not very responsive to these societal pressures, and Caterpillar had to resort to Congress for a solution. 3 For a review of these three approaches, see Ikenberry, Lake and Mastanduno (1988). For discussions about possible synthesis and about other approaches, see Cohen (1994). 3

combination of statist and societal approaches. However, most of these studies have focused on trade policy issues, and there have been few efforts combining systemic and statist explanations to account for state foreign economic policy. This study seeks to find out whether systemic and statist approaches are of complementary nature in explaining foreign economic policy behaviors. Such an assessment requires combining them within the same statistical model to control for possible spurious relationships. The statistical analysis can be meaningful only when we resolve the incompatibility between the assumptions of the theories at two different levels of analysis. Systemic theories view nation states as unitary actors, and foreign economic policies result from the choices of these unitary actors. On the other hand, statist approaches believe political power is distributed among different domestic actors and their conflict and compromises lead to certain foreign economic policy outcomes. The introduction of domestic politics demands some modification of the unitary actor assumption. Here, I assume that the Treasury is the sole representative of the United States in these foreign exchange policy intervention activities for reasons to be discussed next. The bureaucracy has to make a policy choice regarding whether to intervene in the foreign exchange market to realize their policy preferences under various constraints. This assumption makes it possible for us to examine the interaction of international and domestic factors. 4 4 As stipulated in the Gold Reserve Act of 1934, the Treasury has exclusive control over the use of its Exchange Stabilization Funds (ESF) for intervention. And yet the actual operations are carried out by the foreign exchange desk of the Federal Reserve Bank of New York serving as the Treasury s agent. Developments in the foreign exchange market have made the Treasury dependent on the Fed s cooperation in taking joint actions and using some of the Fed's resources in order to carry out effective interventions. However, based on the convention and precedent, the Treasury exercises greater authority over this matter and makes all final decisions regarding intervention activities. Therefore, throughout the paper, I will consider that the Treasury is the decision-maker in terms of all foreign exchange policy interventions and that the Federal Reserve takes only a secondary role, mainly responsible for carrying out the actual buying and selling instructions from the Treasury bureaucracy. This is not an unreasonable assumption given that the Treasury's constituency is the overall economy (Cohen, 1994: 55), i.e. it takes the national interest in a growing economy as its own interest (Henning, 1994). In the foreign economic policy 4

Given that the Treasury is the decision-maker in foreign exchange rate policy, systemic and statist explanations can be integrated from a policy choice perspective. The bureaucracy makes choices to realize policy preferences under various domestic and international constraints. As Ikenberry et al (1988) point out, foreign economic policy makers are different from domestic policy makers; they stand in between the domestic and international arenas. They are expected to pursue specific foreign policy and national interests, or follow the call of their own ideological. As a result, bureaucratic policy preferences are formed both in light of state relative capabilities and ideological partisanship. The realization of these preferences, however, is constrained by both the US obligations to international regimes and domestic institutional structures, especially congressional political control. The United States has been a critical stabilizer of the international monetary system. The Treasury bureaucracy, who represents the US at the International Monetary Fund and World Bank and negotiates about macroeconomic policy coordination issues, has performed the stabilization function (Destler and Henning, 1989; Henning, 1994:109; Cohen, 1994:52-3). The Treasury engages the United States in international obligations and has to commit financial resources to fulfill these obligations. On the other hand, the domestic institutional structure forces the bureaucracy to strike a balance between its own policy preferences and responsiveness to congressional pressures. As a result, both the fulfillment of international obligations and the realization of bureaucratic preferences depend upon the bureaucracy s policy autonomy. The bureaucracy is caught in a two-level game (Putnam, 1988). More specifically, the bureaucracy is constrained in its ability to realize its policy preferences and fulfill obligations to international regimes by domestic political institutions. There is a trade-off between maintaining bureaucratic policy autonomy and satisfying congressional preferences. This can only be shown in a model combining both systemic and statist factors. area, the Treasury enjoys sufficient authority delegated by Congress and the President to represent the US (Cohen, 1994: 51-3). 5

A. Systemic Approaches to Foreign Economic Policy Two prominent schools of thought represent systemic approaches to the study of state foreign economic policy. They are structural realism and neoliberal institutionalism. They share assumptions about the state as a unitary rational actor and the anarchic nature of the international system. They promise to predict the recurrent patterns of state behaviors. Foreign exchange rate policy has taken on a more prominent position in a state s foreign economic policy over time because of the adoption of the flexible exchange rate regime and the trend toward global financial integration (Frieden, 1991). From the point view of systemic level scholars, it has become an important policy instrument for promoting economic growth and national power in general (Helleiner, 1994; Webb, 1995). On the other hand, foreign exchange rate policy is one of those issue areas that witnesses strong influences from international regimes (Keohane, 1984). Therefore, it is appropriate for evaluating neorealism and neoliberalism in terms of their predictions about American foreign exchange rate policy behaviors. Structural realism (Waltz, 1979) posits that state behaviors are structurally determined by the characteristics of the international system; within a system of anarchy, the distribution of power among states and concerns over their relative capabilities determine state actions and their outcomes. The international system is characterized by anarchy, i.e. absence of common interstate authority. The direct implication of anarchy is that states rely on self-help and thus are extremely sensitive to the erosion of their own relative power and the accruing of relative gains to others (Waltz, 1979; Grieco, 1988). Empirical studies of relative gains have been sporadic, and the results are highly mixed. 5 With the decline of US influence in international finance and the rise of Germany and Japan as world economic powers, 5 In his case analysis of US vs. Japan industrial policies, Mastanduno (1991) finds that relative gains concerns do not always translate into policy outcomes. Iida (1995) does not discern significant relative gains concerns by either Japan or US in terms of foreign exchange rate policy coordination. 6

the foreign economic policy behaviors of the US should reflect such concerns over relative increase in power of these two countries. There are competing hypotheses regarding the impact of relative gains concerns. For a pure structural realist, the concern over relative gains is constant, perennial and thus always present. Foreign exchange rate policy is an external economic adjustment strategy that can be employed strategically by a state to its own advantage. Gilpin (1987: 163) points out that the mercantilist bias for export surplus in modern economies is clearly reflected in the foreign exchange policy interventions of modern industrial powers. Competitive currency devaluations in the early period of this century is an often cited example (Gowa, 1983). Under a flexible exchange rate regime, dollar appreciation reduces US export competitiveness and thus affects its economic power; more importantly, it increases the trade competitiveness of the corresponding country. With an unconditional concern over relative gains, the US is expected to intervene whenever the dollar appreciates. Critics of this unconditional concern over relative gains (Keohane, 1993; Snidal, 1991) agree that states are concerned about relative gains, but that such concerns tend to be conditional, depending on the issue, number of players, their perceptions, and the power differential between them. 6 A state is concerned about relative gains when its competitor is catching up. Thus, it is hypothesized here that when the dollar appreciates, accompanied by an increase in the economic growth differential between other industrial powers and the US, the United States will intervene in the foreign exchange market. Intervention, as an expression of relative gains concerns, is conditional upon the growth differential between its economic competitors and US. 6 Foreign exchange market interventions are usually targeted at the bilateral exchange rate involving two states. According to (Keohane, 1993: 276), this qualifies the foreign exchange policy behaviors as a test for the relative gains concerns. 7

The distribution of power among states also has important implications for the distribution of economic adjustment costs among them (Gilpin, 1987). From the perspective of a structural realist, this has been a perennial theme for bargaining among states. Foreign exchange intervention operations demand the allocation of financial resources and thus impose burdens of adjustment on the states. The key issue is who is to adjust or to intervene more actively in the foreign exchange market? From the point of view of structural realism, the distribution of power determines the behaviors of states (Waltz, 1979). Therefore, the asymmetry in the power distribution among states predicts that the weaker state is to shoulder the burden of economic adjustment. Under a flexible exchange rate regime, an important determinant of such bargaining power is the relative level of trade dependence between states (Webb, 1995: 47-9). The less trade dependent state can afford to hold out longer under unfavorable foreign exchange market conditions. A state that is more trade dependent intervenes more actively in the foreign exchange market. Structural realism has been criticized on the grounds that it has neglected the impact of international regimes upon state behaviors (Krasner, 1983; Haggard and Simmons, 1987). Keohane (1984) suggests that international regimes--rules, norms, and decision-making procedures--can affect the behaviors of nation states. States or their decision-makers have a demand for mutually beneficial exchanges and thus set up international regimes among themselves. International regimes can reduce the relevant uncertainty and transaction costs by increasing contacts among states and structuring their mutual expectations. States and their decision-makers adopt a different incentive structure due to the shadow of the future. But, the extent to which international regimes influence state policy behaviors depends on the issue area (Keohane and Nye, 1977; Cohen, 1994). In the foreign exchange policy area, international regimes do exist. In the post Bretton Woods System period, a new international monetary regime emerged around early 1976 (Keohane and Nye, 1977: 84). It is codified in the IMF Executive Board decisions: (1) Countries should not manipulate exchange rates in order to prevent balance of payments adjustment or to gain unfair competitive advantages to others; (2) countries should intervene to 8

counter disorderly market conditions; (3) countries should take into account the exchange rate interests of others. 7 The United States agreed to these rules and had them approved by Congress. These principles have been strengthened over time through an important institutional channel, the seven power summit and its declarations and agreements (Putnam and Bayne, 1987; Funabashi, 1988). Two hypotheses are generated to assess the importance of these international regimes with respect to the US foreign exchange market interventions. First, to correct disorderly market conditions, the Treasury has to devote precious resources for the purpose of stabilizing the foreign exchange market. Drastic exchange rate volatility increases market uncertainty and affects the efficient allocation of resources. Under the influence of the international regime, the higher the volatility, the more actively the Treasury will intervene in the market. Second, if international regimes have an even stronger impact, the bureaucracy ought not to consistently intervene for the sole purpose of correcting US balance of payments deficits, and yet other states balance of payments difficulty should have the opposite effect. It is expected that the dollar appreciation together with current account surplus will lead the Treasury to intervene so as to help other countries, and that the dollar appreciation together with current account deficit will not lead the Treasury to intervene in order to gain an advantage in trade competitiveness through currency devaluation. This hypothesis tests whether US is strongly committed to the maintenance of international regimes in the international monetary system. B. Statist Approaches to Foreign Economic Policy 7 IMF Executive Board Decision no. 5392-(77/63), adopted 4/1977, as cited in Dominguez (1993). 9

Systemic approaches to the study of a country's foreign economic policy have been criticized as inadequate explanations for their intentional disregard for domestic politics. Both structural realists and neoliberal institutionalists (Waltz, 1979; Keohane, 1993: 294) acknowledge such inadequacy, but argue that systemic explanations are a necessary first cut before we move to other levels of analysis. Statist approaches explain state foreign economic policy outcomes by examining both domestic institutions and the policy preferences of policy makers (Ikenberry et al., 1988; Simmons, 1994). Their explanatory power is argued to be independent of systemic factors. Foreign exchange policy interventions are at the discretion of a bureaucracy, which is constrained by domestic political structures. Previous research (Funabashi, 1988; Destler and Henning, 1989; Frankel, 1990; Henning, 1994) has explored the influence of domestic institutions and bureaucracies with regard to foreign exchange policy interventions. The ability of policy makers to realize their own policy preferences depends to a large extent on the institutional arrangements of domestic politics. One such institutional approach studies the centralization and diffusion of power within the state, especially the relationship between the legislative and the executive (Ikenberry, 1988: 227). On foreign economic policy issues, economic interdependence and the bureaucratization of policy making have increased the salience of the issues and made the politicians more watchful of the bureaucrats behaviors (Putnam and Bayne, 1987: 18). Since the 1980s, there has been a large body of empirical literature studying the political control of the bureaucracy in American politics, and yet no similar study has been undertaken in terms of foreign exchange rate policy. 8 I argue that the study of American foreign economic policy behaviors can benefit from the empirical analysis of the political control of bureaucracy. American foreign exchange rate policy is formulated and implemented by a bureaucracy---the Treasury. Yet, there is a principal-agent type of institutional relationship between the bureaucracy and Congress. According to the Constitution, Congress is empowered "to coin money, regulate the value thereof and of 8 For a good review, see Wood and Waterman (1994). 10

foreign coin". This implies that Congress controls foreign exchange rate policy. Over time, Congress has delegated that power to the bureaucracy, expecting that the agent strive for outcomes desired by the principal. The problem, however, is that the agent can develop its own distinctive interests such that the principal must find a way to ensure that the agent acts in the principal's interests (Moe, 1984). Studies of the political control of bureaucracy suggest that Congress can influence the bureaucracy and its policy output through appointments, budget, oversight, and legislation (Wood and Waterman, 1994; Ringquist, 1995). In the foreign exchange rate policy area, appointment and budget are not effective policy channels for Congress. Rather, Congress exercises control over the bureaucratic policy output mainly through oversight and legislation. Since the bureaucracy enjoys an information advantage over the principal, the information asymmetry between them increases the possibility of bureaucratic shirking (Wood and Waterman, 1994: 23; McCubbins, 1985). Since foreign exchange rate policy is of high technical sophistication (Destler and Henning, 1989), Congress must reduce such information asymmetry to avoid bureaucratic shirking. Oversight hearings are a low cost and readily available tool for such a purpose. In addition, oversight hearings serve as a signaling device for Congress to convey its interests to the bureaucracy. In the general sphere of foreign economic policy, Congress has in fact made more frequent use of oversight hearings to affect the bureaucracy (Cohen, 1994: 96-7). Treasury officials are called upon to report and explain their foreign exchange policy actions before various committees of both houses, including finance, banking, foreign relations, commerce, and the joint economic committees. It is hypothesized that if oversight hearings in foreign exchange rate policy are an effective policy instrument for political control, the bureaucracy responds with significantly different intervention activity in the foreign exchange market. Congress is reluctant to intervene in foreign exchange rate policy since it is an area of low political visibility and high technical sophistication (Destler and Henning, 1989; Helleiner, 1994). In addition, many of those oversight hearings serve the purpose of reducing information asymmetry between Congress and the bureaucracy. Thus, Congress usually does not give specific instructions about the intervention operation, which is carried out at the discretion of the bureaucracy. 11

Congressional interest in bureaucratic policy output is constituency based and reelection oriented (Schattsneider, 1935; Weingast and Moran, 1983; Weingast, 1984). When societal pressures become heightened and other means of political control are not effective, Congress attempts to institutionalize its interest through direct legislation to bind bureaucratic behaviors. Of course, direct legislation, as a means of influence, is much harder to wield and much less flexible, compared to oversight hearings. But once adopted, it is much more powerful and influential in affecting bureaucratic behaviors. Congress attempted to pass legislation regarding foreign exchange rate policy as early as 1983; seven bills were proposed in both Houses in the year of 1985 (Destler and Henning, 1989:110). Finally, in 1988, these attempts culminated into a subtitle under the Omnibus Trade Act. Two hypotheses are generated to assess the impact of congressional legislation upon the bureaucratic policy behaviors. Specifically, the Exchange Rate and International Economic Policy Coordination Act of 1988, included in Subtitle A under Title III of PL100-418 Omnibus Trade & Competitiveness Act of 1988, authorizes international negotiations by the Treasury on exchange rate and economic policies and requires regular reports from Treasury about foreign exchange intervention activities. It is hypothesized that this piece of legislation elevates the prominence of the issue and sends a signal of greater Congressional control to the executive bureaucracy. If the bureaucracy is responsive, the enactment of the act should indicate qualitatively different bureaucratic behaviors and is expected to exert a long run influence. That legislation by itself does not specify actual intervention activities. On the other hand, Congress is more explicit about the operation of intervention when it comes to the trade-related dimension of foreign exchange rate policy. It not only requires that the Treasury Department designates an officer of multilateral development bank procurement to promote US company exports and their access to multilateral development bank procurement opportunities, but also exhorts the Treasury to 12

be more sensitive to the US trade balance. 9 This indicates that when it comes to the trade related aspect of foreign exchange rate policy, Congress expects more active bureaucratic interventions than before the law was enacted. Since dollar appreciation has been specifically pointed out as a problem during the midand late 80s, a bureaucracy responsive to the legislation is expected to be more interventionist when the dollar appreciates. The ideological dispositions of foreign exchange policy makers constitute a more autonomous dimension of the bureaucratic policy preference structure. Policy decisions are often informed and guided by the political beliefs and attitudes of the policy makers (Krasner, 1978). Ideological partisanship serves as a good indicator of these unobserved political beliefs and attitudes of the policy makers and reflects their differences in policy preferences (Quinn and Shapiro, 1991). Democrats are in favor of a Keynesian type of government intervention in the national economy whereas Republicans are more likely to support laissez faire, neoclassical policy packages. Partisanship matters in terms of policy output. With respect to foreign exchange rate policy, the same theoretical expectations apply; Democrats intervene more than Republicans (Odell, 1982). There exists some qualitative empirical evidence regarding the policy decisions of the Treasury officials and Reagan himself during his first term (Destler and Henning, 1989; Funabashi, 1988). III. Operationalization, Data and Methodology The dependent variable in my study is direct US exchange rate policy interventions in the foreign exchange open market. Most foreign exchange operations conducted by the Treasury have been concentrated in the US Dollar, the Deutsche Mark, the Japanese Yen and the Swiss Franc. In this paper, I will focus on the buying and selling of dollars ($) against the Deutsche Mark (DM) and the Yen in open 9 See the Multilateral Development Banks Procurement Act of 1988, Subtitle C under Title III of the Omnibus Act of 1988. 13

market intervention operations. In most studies on the effects of interventions by economists, intervention is measured by the amount of currency buying or selling. The purpose of this paper is to specify the conditions under which the Treasury decides to go into the market to intervene, and not the amount of intervention. Therefore, the underlying concept of foreign exchange market intervention is best measured as an event count of the number of interventions in the foreign exchange market over a specific period. The dependent variable is measured by the number of interventions per month by the United States against the Mark and the Yen respectively from March 1973 to December 1994. 10 Data on the dependent series are obtained by counting the number of days of intervention in the respective market from the daily intervention data provided by the Federal Reserve Board of Governors. The working assumption here is that the bureaucracy considers a day of intervention as one intervention event. This assumption may not be totally unrealistic from a market operational point of view. The same measure is also used in Klein (1993). To test the hypothesis that states have an unconditional concern for relative gains, I use a dichotomous measure of dollar appreciation. It is coded one if the difference between the monthly average exchange rates of Mark/Dollar or Yen/Dollar for two consecutive months is greater than zero, and is coded zero otherwise. It tests the stronger version of the relative gains hypothesis, i.e. the relative gains concerns are always present. The expected sign of this variable is positive. The hypothesis about conditional relative gains concerns is tested by using an interaction term of Japan (or Germany) and US industrial production growth differential and the dollar appreciation dummy 10 These dates are the starting and ending dates of the daily intervention data provided by the Federal Reserve Board of Governors. In addition, this measure of intervention captures only those events by Treasury or Fed using American funds as in the Exchange Stabilization Fund (ESF) or Fed account. It does not include those actions undertaken by the US authorities representing the foreign central banks (as using their funds in their accounts with Fed). 14

variable. The industrial growth differential is calculated as the difference of the industrial production index between Germany (or Japan) and US; it is lagged two months since such information is usually available to the bureaucracy with a two-month lag. It tests conditional relative gains concern, and the expected sign is positive. To test the hypothesis of asymmetric adjustment costs, I use the relative trade dependence ratio between the US and Germany (or Japan), i.e. US trade dependence level over Germany (or Japan) trade dependence level. The measure used for trade dependence is export value (reported monthly) over GNP (reported quarterly), both in current dollars. The measures are again lagged two months. The higher the US trade dependence ratio relative to Germany or Japan, the more adjustment costs the US will share, the more active the US will be in its interventions. For the two hypotheses regarding international regimes, exchange rate volatility is measured as the absolute value of the percentage change in the monthly exchange rate (Deutsche Mark/Dollar or Yen/Dollar). The greater the volatility, the more disorderly the market conditions are, the more active the bureaucracy should intervene in the market under the constraint of international regimes. Greater volatility leads one to expect more active interventions. The second hypothesis about international regimes is tested by including an interaction term of the two month lagged US trade balance variable and the dollar appreciation dichotomous variable. The interaction term is expected to be positively related to the number of monthly interventions against the respective foreign currency. The greater the trade surplus, along with dollar appreciation, the more active the bureaucracy should be in curbing the deteriorating impact on other countries balance of payments situation; the greater the trade deficit, when accompanied by dollar appreciation, the less interventionist the bureaucracy should be in manipulating the exchange rate to improve its own balance of trade situation. 15

The domestic institutions hypotheses are tested by the following measures. For the impact of oversight hearings, I use the number of relevant congressional oversight hearings, lagged for one month. Here, only those congressional hearings are considered in which chief Treasury officials testify with specific relevance to US foreign exchange rate policy or intervention activities regarding the relevant currency or country. Though Congress does not give specific instruction about the direction of action in those oversight hearings, the greater the pressure Congress exerts through hearings, the more responsive the bureaucracy will be. The Exchange Rates and International Economic Policy Coordination Act of 1988 was approved by the President on August 23, 1988. To test whether it impacts behaviors of the Treasury in the foreign exchange market, a binary variable representing the law is coded as one since September of 1988 and zero before that. There is no specific expectation about the degree of intervention, but the bureaucracy is expected to behave differently afterwards. On the other hand, to test the influence of the Multilateral Development Banks Procurement Act of 1988 and the Omnibus Trade and Competitive Act in general with respect to trade related issues, an interaction term of the law dummy variable and the dollar appreciation dummy variable is included in the model. Under relevant circumstances, the bureaucracy is expected to intervene more actively. To assess the importance of ideological partisanship, I use the presidential administration party affiliation. Months with a Democrat President in office are coded one whereas those under a Republican President are coded zero, with the expected sign being positive. This measure implies that bureaucratic policy preferences should be more Keynesian oriented and that the bureaucracy is prone to intervene in the market under a Democrat President. Under a Republican President, the bureaucracy believes in more of a laissez faire, hands-off approach. This is reasonable if we accept the assumption that when the President selects his cabinet, he picks those who are close to him in policy preferences. One possible problem is that this does not reflect the variations within one administration with different Secretaries of the Treasury. On 16

the other hand, it may be more realistic because the President does and is able to keep his cabinet in line with his own policy preferences through personnel changes; his appointment power enables him to do so. Since the dependent variable is the number of foreign exchange interventions undertaken by the United States within each month from March 1973 to December 1994, it is discrete event count. For event count models, OLS regressions are not appropriate because the dependent variable is not normally distributed (King, 1989a; 1989b; 1989c) Another commonly used statistical technique for event count analysis is Poisson regression. It is also not suitable for my analysis here since Poisson regression assumes that events are independent of each other and occur at a constant rate (δ 2 = 1). These assumptions do not hold for US foreign exchange interventions because the dependent variable is possibly plagued with underdispersion or overdispersion. If an intervention has successfully stabilized the market, the probability for a follow-up intervention is decreased. This is a case of underdispersion (δ 2 < 1). However, market speculators sometimes test the commitment of a government in terms of intervention and stabilization. In such cases, government intervention can lead to more speculation activities and higher market volatility, thus making another intervention necessary for the sake of intervention effectiveness. This is the case of overdispersion (δ 2 > 1). King (1989b; 1989c) has developed a maximum likelihood estimator for discrete event count variables, in which a dispersion parameter is estimated that tells us what type of dispersion is with the underlying process. This Generalized Event Count (GEC) estimation routine, contained in the statistical package COUNT developed by King, is used to estimate the event count models of the US interventions against the Deutsche Mark and the Japanese Yen respectively. 11 11 Because interventions against the Deutsche Mark and the Yen are modeled separately, a potential estimation problem arises. Intuitively, due to decreasing exchange and capital controls across borders, foreign currency markets are tightly linked technically, and speculations in the $/DM and $/Yen markets may well be interrelated. If this is true, then intervention activities may also be related across currencies. Statistically, this implies that the error terms of these two separate models are correlated, and the price for estimating them individually is a loss of efficiency. It would be desirable to model them as a Seemingly Unrelated GEC system. But, such an estimator is not available in 17

IV. Findings The generalized event count model (GEC) is applied to Treasury interventions against the Deutsche Mark and the Japanese Yen using three different model specifications. Table 1 presents the findings of the systemic explanations for interventions against the two currencies respectively. Table 2 contains the estimates of the statist model. Finally, Table 3 presents the results from the combined systemic and statist model. Across all six models in the three tables, the gamma parameter is statistically significant and positive, rejecting the null hypothesis that the dispersion is constant (δ 2 = 1). It shows the presence of overdispersion and the necessity of using a GEC model to analyze the underlying event occurrence process. The result that market speculators do test the authority's commitment to intervention effectiveness is in line with those observations made by Treasury officials in the reports to Congress. In addition, it indicates that controlling for such overdispersion is critical to the tests of our hypotheses without getting spurious relationships. Table 1 about here For the systemic model, both structural realism and neoliberal institutionalism receive some empirical support. In the Dollar/DM case, the dollar appreciation dummy and relative trade dependence ratio are both statistically significant and in the expected direction as structural realism suggests in the unconditional relative gains concerns and asymmetric adjustment cost hypotheses. On the other hand, the interaction term of US trade balance and dollar appreciation is also significant and positive, indicating that international regimes do induce more active bureaucratic actions in the creation of collective goods in the international monetary system. The conditional relative gains concerns and exchange rate volatility the current literature. Given that the potential problem is a loss of efficiency, it should only decrease the chances of finding significant results in estimating separate GEC models. 18

hypotheses, however, are not statistically supported. Neither systemic theory by itself provides the best explanations for the Treasury interventions against the Deutsche Mark. In the Dollar/Yen model, the asymmetric adjustment cost variable is significant and in the expected direction, yet none of the relative gains variables is statistically significant. Exchange rate volatility comes out significant with an expected sign whereas the interaction of US trade balance and the dollar appreciation is not. Compared to the Dollar/DM model, the Dollar/Yen model also provides limited support for both structural realism and neoliberalism. But, such support is lent to different sets of hypotheses in the two models. The implications of these differences are to be discussed in the next section. Table 2 about here For the statist model, law and ideological partisanship both have statistically significant impacts upon the bureaucratic intervention operations against the Deutsche Mark. The Exchange Rate and International Economic Policy Coordination Act of 1988 affected the bureaucratic incentive toward market interventions. Although the direction of its impact is not specified, its sign turns out to be negative, indicating fewer bureaucratic interventions against the Deutsche Mark after 1988. In addition, under a Democrat President, the bureaucracy tends to be more active in intervention operations than under a Republican President. Congressional oversight hearings are statistically insignificant in terms of their influence over the bureaucratic intervention operations against the Mark. The interaction term of the dollar appreciation and law, gauging the impact of congressional preference with respect to trade issues, turns out to be insignificant although very close to the 10% significance level. For the Dollar/Yen model, oversight hearings, law and the interaction term of law and the dollar appreciation are all statistically significant. Oversight hearings and the interaction term of law and the dollar appreciation are in the expected direction. Without a specific expected direction, law turns out to have positive influence over bureaucratic interventions against the Yen. Domestic institutions have 19

noticeable influence upon the probability of bureaucratic interventions against the Japanese Yen. The ideological partisanship does not exhibit such impact. The results in terms of the direction of law and the impact of partisanship and oversight hearings are different from those in the Dollar/Mark model. A comparison of these two models seems to reveal a trade-off relationship as to how close the bureaucracy can follow its ideological partisanship and the calling of Congress. Table 3 about here Table 3 includes findings about the combined Dollar/DM and Dollar/Yen models. An important question the combined models try to answer is whether the systemic and statist explanations are complementary in nature. Statistically, the log-likelihood ratio (LR) test (Pindyck and Rubinfeld, 1991:240) evaluates the joint hypothesis that the additional set of parameter estimates in the combined model are zero, compared to the baseline model -- the systemic or the statist model. The hypothesis is rejected with statistical significance for both combined models. Neither the statist nor the systemic model washes out the impact of the other one, showing their complementary nature; both provide relevant explanatory variables, and the omission of either may well lead to a specification error. The combined Dollar/DM model shows that the significant results in the individual models are robust; most of the previously significant variables remain significant. Even after controlling for the impact of other intervening variables, relative gains concerns still come out significant. The statist model s explanatory variables exhibit the strong influences of domestic politics. More interestingly, some previously insignificant variables--the conditional relative gains concern variable and the interaction term of law and dollar appreciation--become significant. The combined Dollar/Yen model retains most previously significant variables and shows greater precision and smaller standard errors for their parameter estimates. Yet, the systemic variables seem to be less robust, compared to the domestic politics variables. The previously significant relative trade dependence ratio now becomes insignificant, and the dollar appreciation dummy variable turns out to be statistically 20

significant. Yet, exchange rate volatility is still statistically significant. As in the Dollar/Mark model, the domestic institutions and ideological partisanship variables are very robust. These findings shed light on bureaucratic policy behaviors in the murky world of international finance. Since both combined models are shown through the log-likelihood ratio (LR) tests and the generally decreasing magnitude of the standard errors of the parameter estimates to perform better than each individual model, I will focus on the combined models in discussing the implications of these findings in the next section. V. Conclusion The foreign exchange market now has the largest daily turnover among all markets including goods and services, and stocks and bonds, which is a consensus among both scholars and practitioners (Cohen, 1994: 29; Luca, 1995: 2). Foreign exchange intervention operations are an important policy instrument for state to regulate the market and stabilize international monetary system. The above findings help to answer questions about whether there exist any general patterns in the bureaucratic foreign exchange rate policy behaviors and whether these patterns are subject to explanations from both systemic and statist approaches. Some nonobvious findings concern whether the bureaucracy behaves differently in the interventions over the two currencies and why. Theoretical implications may also be inferred to help understand the absolute gains versus relative gains debate and demonstrate the merits of incorporating systemic and statist approaches. I start the discussion with reference to the neorealist and neoliberal debate. The global economy is claimed to be an issue area that may be immune from relative gains concerns (Lipson, 1984; Keohane and Nye, 1977). But, the results here indicate that even in a field where international coordination and cooperation are most prominent, nation states still exhibit behaviors driven by relative gains concerns. The intensity of such concerns is certainly related to the strength or power of the relevant competitor. This 21

may partly explain the fact that bureaucratic interventions tend to focus more against the Deutsche Mark than the Japanese Yen. Due to the critical role the Deutsche Mark played in the European Monetary System (EMS) during the floating exchange rate period, it is reasonable to find that the Untied States exhibit both conditional and unconditional relative gains concerns with respect to the Dollar/DM operations. Contrary to the uproar about Japan in the 1980s, the bureaucracy does not seem to share the common perception about a threatening Japanese Yen. On the other hand, the pessimistic realist prediction that the nation state always behaves in a manner that is consistent with relative gains concerns may have to be qualified. International regimes can correct the purely myopic self-interested behaviors of nation states. As in the Dollar/Yen interventions, the US does commit resources to deal with disorderly exchange rate volatility in a consistent manner; in the Dollar/DM operations, the Treasury makes intervention decisions that are aimed at helping countries in balance of payments difficulty. Although it may be argued that international regimes are no more than institutions of interest that reflect only the will of the dominant powers, the findings here do lend some support for the conclusion that nation states, even though caring about relative gains, may still act in a long-run oriented manner that may facilitate international cooperation. That international regimes matter does not require nation states to be altruistic, only that they should not be purely myopic. According to my current results, bureaucratic interventions in the foreign exchange market do exhibit patterns that support explanations from both neorealism and neoliberalism. Neither has shown superiority in terms of the empirical evidence presented here. What is not answered in this article regarding the neorealist versus neoliberalist debate is the specific conditions under which the bureaucracy switches from following relative gains concerns to honoring its obligations prescribed by relevant international regimes. This study has garnered strong evidence regarding the usefulness of the statist approach in explaining bureaucratic interventions in the foreign exchange market. Domestic politics alone can affect bureaucratic 22

intervention decisions. Across both foreign currencies, domestic institutions have been shown to have significant impacts upon bureaucratic behaviors. Direct legislation is an effective policy tool in order for Congress to extract responsiveness from the Treasury in all its intervention operations. As a response to the legislative instructions, the bureaucracy decreased its absolute level of intervention activities in the Dollar/DM operations after September of 1988 and targeted the allocation of resources to trade-related dollar appreciation episodes. The same period was characterized by Congressional pressures in passing trade protectionist bills and increasing oversight hearings. The focus of Congressional attention was upon Japan, and the bureaucratic response was to increase interventions against the Yen. The incorporation of the statist with the systemic explanations points to the most interesting theoretical implication in this article. Ikenberry et al. (1988) suggest that the foreign economic policy making bureaucracy enjoys relatively greater policy autonomy vis-à-vis its domestic economic policy making counterpart, due to its special institutional functions. The analysis here seems to suggest that actual bureaucratic policy autonomy may be very small even if it actually exists. In the context of foreign exchange policy interventions, the degree of domestic political control over the bureaucracy explains the extent to which the bureaucracy realizes its policy preference embedded in national interest and ideological partisanship, and its commitments to international regimes. The weaker the Congressional influences are, the more policy autonomy the bureaucracy enjoys to realize its own preferences and to fulfill its international obligations. In the Deutsche Mark intervention activities, Congressional influence is exercised only through the use of direct legislation, a highly effective but extremely unwieldy policy instrument. But, in the Dollar/Yen operations, Congress is shown to have more direct influence. It certainly makes good use of direct legislation, and is also able to affect bureaucratic behaviors through a more flexible and less costly policy tool -- oversight hearings. The level of congressional influence differs according to the currency and the foreign country involved. As a result, in the Deutsche Mark operations, the bureaucracy is better able to follow its own policy preferences in terms of national interest and ideological partisanship. It is also able to pursue the much more restrictive international regime commitment, which requires it to have great policy autonomy to exercise its discretion. 23