1 THE HIGH POLITICS OF IMF LENDING By STROM C. THACKER* Introduction degree of scrutiny given role of In CONSIDERING ternational Monetary Fund (IMF, or Fund) in international economy, we know little about causes underlying of IMF's be havior.1 During 1980s, IMF became a "lender of last resort" for many developing country governments that had been cut off from pri vate credit markets faced destabilizing imbalances of payments. After return private capital began voluntarily what were called emerging markets in early 1990s, anticipated erosion of Fund's role in developing world did not materialize. Faced with re current payments' imbalances, pressures for currency devaluation, macroeconomic instability associated with crises in Latin America, Asia, Russia, developing nations have turned with increasing IMF frequency credits stabilization plans. Despite growing body of research on IMF's critical role in international finance, we still have few explanations of only patchy empirical data on why IMF approves loans some countries but not ors. As Fund delves furr in management of balance of payments cur rency crises around world, both oretical practical impera tives dictate that we more specify fully test more systematically *A preliminary version of this paper was at presented 1997 annual meeting of American Po litical Science Association in Washingn, D.C. I would like thank Tim McKeown, Patrick Con way, William Bernhard, Mary Matws, Dane Rowls, participants in faculty research seminar in Department of International Relations at Bosn University, anonymous World Politics referees for ir insightful comments suggestions. Yvonne Ochoa Jaya Badiga pro vided able research assistance. The usual disclaimers 1 apply. The literature on IMF is extensive. For useful surveys, see Graham Bird, "The International Monetary Fund Developing Countries: A Review of Evidence Policy Options," Inter national Organization 50, no. 3 (1996); idem, IMF Lending Developing Countries: Issues Evidence (London: Roudedge, 1995); Sebastian Edwards, "The International Monetary Fund Devel oping Countries: A Critical Evaluation," NBER no. Working Paper, 2909 (1989); Tony Killick, IMF Pro grams in Developing Countries: Design Impact (London: Roudedge, 1995); John Williamson, ed., The Lending Practices of InternationalMonetary FundXWashingn, D.C: Institute for International Economics, 1982); idem, IMF Conditionality (Washingn, D.C: Institute for International Eco nomics, 1983). World Politics 52 (Ocber 1999), 38-75
2 HIGH POLITICS OF IMF LENDING 39 competing explanations of IMF behavior. What facrs influence IMF's decisions lend? Are se decisions based on technical eco nomic criteria, or do y reflect preferences of Fund's more powerful members? What are those preferences, how do y affect IMF's relationship with its developing country clients? In or words, matter? does politics matter? More importantly, how does The literature on international institutions multilateralism politics suggests that operation of multilateral economic organizations like IMF will assume growing importance in a posgemonic international order.2 This body of ory raises several important ques tions: To what extent do multilateral institutions modify interests or constrain behavior of ir member states? Can more power fill states use se as organizations effective instruments of national or are foreign policy, such pressures diluted or transformed in passing through multilateral channels? Finally, what underlying inter ests drive behavior of large powers within multilateral in stitutions, how do y do so? On practical side, Gallarotti has shown that poorly managed international not can organization only be ineffective but also can destabilize international system.3 The debates surrounding proposed increases in Fund resources recent loan packages negotiated with South Korea, Indonesia, Russia, Brazil demonstrate growing popular recognition of se kinds of problems, but scholarly research has yet address se questions adequately. Economists have made inroads in isolating economic bases of IMF lending, but y are first point out that ir models remain incomplete. Researchers have found statistically significant results for 2 For effective treatments of se related issues, see Robert O. Keohane, After Hegemony: Co operation Discord in World Political Economy (Princen: Princen University Press, 1984); idem, International Institutions State Power: Essays in International Relations Theory (Boulder, Colo.: Westview, 1989); idem, "Multilateralism: An Agenda for Research," International Journal 45, no. 1 (1990); John Gerard Ruggie, ed., Multilateralism Matters: The Theory Praxis an of Institutional Form (New York: Columbia University Press, 1993); Stephen D. Krasner, ed., International Regimes (Ithaca, N.Y.: Cornell University Press, 1983); Kenneth A. Oye, ed., Cooperation under Anarchy (Princen: Princen University Press, 1986). International institutions multilateralism are not necessarily equivalent. The IMF fits s Ruggie definition in Multilateralism Matters of multilateral organization as "defined by such generalized deci sion-making rules as or voting consensus procedures" (p. 14). On IMF see decision-making procedures, Frank Southard, "The Evolution of International Monetary Fund," Princen Essays in Interna tional Finance, no. 135 (1979); Frederick K. Lister, Decision-Making Strategies for International Organizations, vol. 20, World Affairs (Denver, Colo.: Graduate School of International Studies, Uni versity of Denver, 1984). 3 Giulio M. Gallarotti, "The Limits of International Organization: Systematic Failure in Man agement of International Relations," International Organization 45, no. 2 (1991).
3 40 WORLD POLITICS impact of a number of economic variables on IMF lending, but low overall explanary power of econometric models reviewed by Bird suggests that "re are probably a range of or non-economic facrs which still need be delineated."4 One likely source of noneco nomic facrs is politics, but scientists have not yet been able demonstrate systematic impact of variables on IMF lend ing.5 Several case studies offer suggestive, but not generalizable, evi dence of bases of IMF lending. Fewer studies have attempted construct a systematic explanation of Fund be havior. This paper attempts fill some of those gaps in literature proposes answers those questions by developing testing sta tistically a explanation of IMF lending patterns. The IMF^s stated decision-making procedures prohibit consider ation of facrs. Loans are made strictly on basis of monetarist "Financial Programming" model a "Doctrine of Eco nomic Neutrality" that is blind such facrs as international politics nature of developing country regimes.6 The Fund may impose strict lending requirements, but it applies m fairly all countries. The meetings of IMF executive board, which approves all Fund pro grams, are highly secretive.7 The specific considerations that determine board's deliberations decisions are refore not available in public domain. For its part, Fund staff publicly maintains posi tion of economic neutrality, but evidence presented in numerous case studies leaves open possibility that facrs play an impor tant role. There are at least three reasons suspect that politics matters in IMF. First, several studies have found extremely low rates of borrower compliance with Fund conditionally, yet IMF continues lend many of se problem debrs even after earlier programs have been canceled for noncompliance.8 Finch, a former IMF staff member, sug gested in late 1980s that economic facrs could not explain se 4 Bird (fn. 1,1995), Dane Rowls, "Political Economic Determinants of IMF Conditional Credit Arrangements: " (Manuscript, Norman Paterson School of International Affairs, Carlen University, Ot tawa, Ont., 1995). 6 Jacques J. Polak, "Monetary Analysis of Income Formation Payments Problems," IMF Staff no. Papers, 6 (1957), cited in Edwards (fn. 1); idem, "The Changing Nature of IMF Conditional ity," Princen Essays in International Finance, no. 184 (1991); Richard Swedberg, "The Doctrine of Economic Neutrality of IMF World Bank," Journal of Peace Research 23 no. 4 (1986). 7 R. S. Eckaus, "How IMF Lives with Its Conditionality," Policy Sciences 19 (Ocber 1986). 8 Southard (fn. 2), 13; Edwards (fn. 1); C David Finch, "The IMF: The Record Prospects," Princen Essays in International Finance, no. 175 (1989); John Spraos, "IMF Conditionality: In effectual, Inefficient, Mistargeted," Princen Essays in International Finance, no. 166 (1986).
4 HIGH POLITICS OF IMF LENDING 41 patterns: "Because decisions were no longer based on compatibility with repayment terms, lending was guided increasingly by preferences of leading industrial countries."9 Second, each country's on representative Fund's executive board is appointed by his or her home government in (Treasury, case of United States). Thus it should come as no surprise that positions of those representatives within Fund reflect interests of national govern ments y serve.10 As Smith puts it, "The IMF is itself a insti tution. It is managed by ly appointed individuals from member nations, interests of its members influence its deci sions."11 Although staff is less directly linked national govern ments, executive board must approve all proposed programs. The familiarity of Fund staff members with preferences of executive board discourages m from submitting loan packages that board is likely ve.12 Finally, weighted voting decision-making of procedures Fund also leave room for politics. As of April 1995, U.S. controlled percent of voting power in IMF, followed by Germany Japan with 5.5 percent each, France United Kingdom with 5.0 percent each.13 An evolving system of special majorities has helped U.S. maintain its influence beyond that dictated by its gradually de creasing voting share.14 An 85 percent majority is required for most important Fund decisions, giving U.S. alone, or groups of countries ve ger, power. The U.S. can also push through favored programs, which it not might be able do based on its votes alone. Al though managing direcr has traditionally been a European, he acts rarely against U.S. That is not preferences.15 surprising since he is a appointed through process over which U.S. has ve power.16 But 9 Finch (fn. 8), Lars Schoultz, "Politics, Economics, U.S. Participation in Multilateral Development Banks," International Organization 36, no. 3 (1982); Benjamin J. Cohen, "International Debt Linkage Strategies: Some Foreign Policy Implications for United States," in Miles Kahler, ed., The Politics of International Debt (Ithaca, N.Y.: Cornell University Press, 1986). 11 Fred L. Smith, "The Politics of IMF Lending," Ca Journal 4 (Spring/Summer 1984). The U.S. representative is "ordered by law clear his or her decisions with Secretary of Treasury." Swed berg (fn. 6), Kendall W. Stiles, Negotiating Debt: The IMF Process Lending (Boulder, Colo.: Westview, 1991). 13 IMF Annual Report (Washingn, D.C.: IMF, 1995), See Lister (fn. 2). 15 Samuel Lichtensztejn M?nica Baer, Fondo Monetario Internacional y Banco Mundial: Estrate gias y Pol?ticas del Poder Financiero (Mexico City: Ediciones de Cultura Popular, 1987), Miles Kahler notes that U.S. has in past refused support a renewal of managing di recrs tenure when his "accomplishments did not meet American expectations." Kahler, "The United States International Monetary Fund," in Margaret P. Karns Karen A. Mingst, eds., The United States Multilateral Institutions (Bosn: Unwin Hyman, 1990), 94.
5 42 WORLD POLITICS that power rarely needs be wielded The openly. managing direcr typically makes decisions based on a "sense of meeting," derived from comments of various participants ir relative voting power.17 Or powers can be reluctant speak against U.S. for fear that U.S. will later retaliate by exercising its ve power over ir own favored programs.18 Finally, U.S. its like-minded al lies can ger effect international monetary cooperation by forming subsets, or "k-groups," of countries push through certain packages that single parties cannot.19 Given possibility that facrs influence IMF decisions, several scholars have argued that Fund's more powerful members it furr ir own manipulate economic interests.20 The U.S. government, for its part, "has repeatedly ld foreign govern ments that it will not intervene in negotiations between Fund member governments."21 Kahler notes, however, that " U.S. ( or major countries) can still influence programs for friends clients at margin."22 Ors suggest that American politicization of Fund lending is more widespread. A series of case studies conducted for a project directed by Killick Bird reveals that at least one-third of seventeen countries studied secured favorable loan terms on ir IMF programs due intervention of major shareholding countries on ir behalf.23 Stiles concludes that in only one of seven cases exam ined did Fund adopt a ly neutral, technocratic approach lending.24 Such case studies have been useful for providing kind of rich de tailed data that are unavailable by or means, for formulating testable hyposes, for providing initial evidence of role of politics in IMF lending. Despite se advances, we are still unable say much more than that seems politics matter, at least in some cases. This 17 The origins of this procedure date back Fund s early years, when U.S. executive direc r went great lengths muffle strong voice of U.S. power, which neverless was decisive. See Southard (fn. 2), 5-6, Eckaus (fn. 7), 237; Stiles (fn. 12), Ruggie (fn. 2), chap. 1; James A. Caporaso Miles Kahler attribute part of postwar eco nomic cooperation this type of "minilateralism."the creation of Bretn Woods monetary order through U.S. British coordination subsequent adjustments made by G-7 after its breakdown (for example, Plaza Louvre accords) can be profitably undersod in se terms. Caporaso, "International Relations Theory Multilateralism: The Search for Foundations," Kahler, "Multilateralism with Small Large Numbers," in Ruggie (fn. 2). 20 Cheryl Payer, The Debt Trap: The International Monetary Fund Third World (New York: Monthly Review Press, 1974); Swedberg (fn. 6). 21 Kahler (fn. 16), Ibid. 23 Killick(fn.l), Stiles (fn. 12),
6 HIGH POLITICS OF IMF LENDING 43 paper aims two accomplish essential tasks. First, it attempts pro vide first systematic quantitative evidence for wher af politics fects IMF behavior. Second, it proposes a dynamic explanation of how facrs affect interactions of multilateral organizations with ir member states, tests statistical formulation of that argu ment in case of IMF. I first propose a simple macroeconomic model a explanation of IMF lending. I n operationalize se hyposes ger report results of a series of statistical tests conducted on m jointly. I conclude with a discussion of limitations broader implications of this research. A Simple Macroeconomic Model Economists have isolated several important dem- supply-side macroeconomic determinants of IMF lending. Rar than attempt replicate such studies, I take m as a starting point for my analysis. Conway has modeled in IMF participation programs as a func tion of a country's economic environment its past economic per formance.25 He finds statistically significant negative results for lagged variables representing ratio of reserves foreign exchange imports, rate growth of real gross national product, ratio of current account GNP, terms of trade, real rate of interest.26 Vari ables capturing level of development (proxied by share of output from agricultural secr) long short-term debt scks did not attain conventional levels of statistical significance. Lindert tests impact of several variables on interest rate charged on official credir lending fifty-one countries in He obtained statistically significant results for two only variables? log of absolute nominal public publicly guaranteed debt in 1981 log of per capita income?each with a positive coefficient. None of or variables? share of debt held by official credirs, ratio of debt service GNP, ratio of reserves imports, money sck growth, prior default, prior rescheduling, years since first reschedul ing?approached conventional levels of statistical significance. Summarizing his own ors' research, Bird identifies statisti cally significant regression results with negative coefficients for balance 25 Patrick Conway, "IMF Lending Programs: Participation Impact," Journal of Eco ^Development nomics 45, no. 2 (1994). 26 He finds statistically significant positive results for prior participation percentage of avail able funds drawn down. A series of dummy variables for each year had generally significant results. 27 Peter H. Lindert, "Response Debt Crisis: What Is Different about 1980s?" in Barry J. Eichengreen Lindert, eds., The International Debt Crisis in Hisrical Perspective (Cambridge: MIT Press, 1989).
7 44 WORLD POLITICS of payments, per capita income, current account, reserves.28 Col lectively, se studies find statistically significant positive regression coefficients for inflation, access private bank credit,29 domestic credit growth, government spending. With exception of Conway, low predictive power of se models suggests that are y underspec ified. Rar than attempt test validity of distinct competing macroeconomic models, this paper draws upon large body of exist ing research identify putative economic determinants of Fund activities. First, balance of payments position of a country is ini tial baseline upon which its in IMF participation programs is evaluated. An improvement in balance of payments is stated primary goal of most IMF lending programs,30 without a payments deficit, a country should neir need nor be eligible for Fund lending.31 When faced with a payments deficit, a country can eir run down its re serves or borrow internationally.32 In context of debt crisis, run ning down reserves was not a viable long-term solution, most reliable source of international was borrowing IMF. Specifically, de terioration in balance of payments is expected increase chances of a receiving loan from IMF. Second, a country's debt position should affect its dem for supply of an IMF loan for distinct reasons. On dem side, a heavier debt burden increases developing countries' need for external fi nance service that debt. In terms of some supply, have argued that more heavily indebted countries have more over bargaining leverage IMF because of ir importance global financial stability.33 In ad dition, some view IMF loans as a payoff foreign credirs.34 Assuming that those lenders wield influence within executive board, IMF loans more likely will go countries where credirs are more ex heavily posed. Lindert found that more heavily indebted countries did re ceive official credir loans, but at higher interest rates than smaller debrs, due reluctance of Norrn taxpayers finance more 28 Bird (fn. 1,1995). 29 This result supports catalytic impact of IMF as a lending providing "seal of approval" that en courages private banks resume lending a country that has an negotiated agreement with Fund. Bird (fn. 1,1995), 122. A negative result would suggest a substitution effect between IMF private lending. 30 Spraos (fn. 8); Finch (fn. 8). 31 Bird (fn. 1,1995), Ibid., See Jahangir Amuzegar, "The IMF under Fire," Foreign Policy 64 (Fall 1986). 34 See Waiden Bello David Kinley, "The IMF: An Analysis of International Monetary Role in Third World Debt Crisis, Its Relation Big Banks, Forces Influencing sions," Multinational Monir 4 (1983). Fund Its Deci s
8 HIGH POLITICS OF IMF LENDING 45 concessionary terms.35 For purposes of this study, fact that larger debrs were more likely receive loans is of primary interest. Third, level of per capita income of a country also may influence its secure ability Fund assistance. Killick notes that IMF's hisr narrow ically focus on balance-of-payments considerations has given way in recent years a broader view that acknowledges relation ship between balance of payments growth.36 Lindert reports that official favoritism for poor countries resulted in lower interest rates on official loans, Bird argues that poorer countries are less likely borrow on private capital markets refore have a higher rela tive dem for IMF loans.37 Countries with lower per capita incomes should be more likely request receive a loan from IMF. Fourth, if Doctrine of Economic Neutrality is followed, a poor credit hisry should decrease chances of a receiving loan. After massive defaults of 1930s, debr nations were cut effectively off from credit for several years. Many have argued that we should expect similar outcomes now.38 Specifically, past failures IMF uphold loan re quirements should make it more difficult receive additional loans. Finally, both neomarxist modern economy interpreta tions would two suggest additional macroeconomic indicars that should affect Fund decisions due influence of "low politics": trade investment exposure of firms based in IMF's major princi pal shareholder, U.S. Authors grouped loosely within a neomarxist (or dependency) perspective argue that capitalists in core states, es pecially U.S., dictate IMF at policy expense of nations of periphery.39 A economist more concerned with impact of domestic on politics economic foreign policy also might posit, without necessarily adopting concomitant exploitation argument, that well organized export enterprises multinational corporations (MNCs) pressure U.S. government protect ir interests on executive board. Neomarxism suggests that higher levels of U.S. exposure lead a greater likelihood of receiving an IMF loan because attendant 35 Lindert (fn. 27), Killick(fn.l). 37 Lindert (fn. 27), 243; Bird (fn. 1,1995), Barry Eichengreen has questioned impact of "default on penalty" future credit access. Eichengreen, "The U.S. Capital Market Foreign Lending, ," in Jeffrey D. Sachs, ed., Developing Country Debt World Economy (Chicago: University of Chicago Press, 1989), 247. Cf. Jonathan Ean Mark Gersovitz, "Debt with Potential Repudiation: Theoretical Empiri cal Analysis," Review of Economic Studies 48 (April 1981). 39 See E. A. Brett, "The World's View of IMF," in Latin America Bureau, ed., The Poverty Bro kers: The IMF Latin America (London: Latin America Bureau, 1983); Manuel Pasr, "The Effects of IMF Programs in Third World: Debate Evidence from Latin America," World Development 15 (Fall 1987); Swedberg (fn. 6). pol
9 46 WORLD POLITICS icy conditionality promotes expansion of global capitalism. A less more ex explored, domestic-politics interpretation yields ambiguous pectations. The inflow of foreign exchange resration of in ternational creditworthiness would be expected benefit U.S. exporters foreign invesrs, while dem-reduction compo nents of IMF typical program would suggest a negative impact for se variables.40 High Politics imf The international aspects of IMF lending have received far less rigorous analysis. Two rudimentary strs of thought comprise this genre, but neir has been fully developed nor adequately tested. I label first, more common str " proximity" hypo sis. Simply put, friends of U.S. are more likely receive loans than are its enemies. In addition case studies described above, Bello Kinley argue that U.S. disregarded Fund's eco nomic criteria pressured Fund approve loans ly friendly South Africa, El Salvador, Haiti.41 The IMF has also de nied loans economically worthy enemies of U.S., such as Vietnam.42 In sum, more a closely country aligns with U.S., higher probability it will receive a loan from IMF. These arguments have not yet been nor fully developed conceptually thoroughly tested empirically. To illustrate an intuitive analytical foun dation for this argument facilitate its testing, I construct a con tinuous voting space, scaled from 0 1, where 1 represents tal agreement 0 complete discord with United States on a single broad dimension of foreign policy affinity (such affinity could be easily measured by votes in a majority-rule international voting arena, such as United Nations General Assembly). Figure 1 is a graphic represen tation of this space a schematic portrayal of proximity hyposis. Countries at point A, at far left-h side of voting space, have little chance of a receiving loan, while those at point C, at far are right, much more likely receive Fund support. Alignment near middle, at point B, has little or no effect. 40 The net effect of DFI exposure may depend on secral location of investment. If it serves primarily domestic market, a negative result might be expected. If it serves mostly export markets, a positive result would be more likely. The impact of export exposure may on depend wher product exported is a final or an consumption good (negative) input in export secr 41 (positive). Bello Kinley (fn. 34), Susumu Awanohara, "Fiscal Interdiction: U.S., Japan Block IMF Effort Support Vietnam," Far Eastern Economic Review, September 28,1989.
10 HIGH POLITICS OF IMF LENDING 47 AB C Voting scale loan probability Impact (-) on (neutral) (+) Figure 1 The Political Proximity Hyposis Despite existence of case plentiful studies, previous research has not effectively evaluated this argument. Furrmore, several studies have documented numerous cases where U.S. "enemies" are rewarded or "friends" punished. In many instances Fund has made loans leftist governments, such as Manley's in Jamaica in 1979 East European Soviet bloc countries of Hungary, Yugoslavia, Romania, each of which ranked among p fifteen IMF loan recipients from This evidence seems paint a picture of IMF lending as an a technocratic process, economic at neutrality its best. But se loans may not have been on economic justifiable purely grounds, eir. Asset compares results of a regression equation designed predict size of IMF loans based on solely economic criteria with size of actual loans received by three East coun European tries conclude that actual lending exceeded a predicted lending by significant margin.44 A less static interpretation of se anomalies introduces "politi cal movement" hyposis, less-developed second str of po litical argument. Movement ward or away from U.S. on international issues may be at least as as important absolute alignment of a particular country. at Hinting this idea, Horowitz asks wher IMF should use loans entice countries like Romania Hungary away from Soviet bloc.45 This notion is con sistent with cases of Hungary, Yugoslavia, Romania, all of whom distanced mselves ly from Soviet Union (that is, moved closer U.S.). In contrast, neir Czechoslovakia nor 43 Amuzegar (fn. 33); Valerie J. Asset, The Soviet Bloc in IMF IBRD (Boulder, Colo.: Westview, 1988). 44 Ibid, Irving Louis Horowitz, "The 'Rashomon Effect': Ideological Proclivities of IMF," in Robert J. Myers, ed. The Political Morality of International Brunswick, N.J.: Transaction Books, 1987), 96. Political Dilemmas Monetary Fund (New
11 48 WORLD POLITICS A B Movement on voting scale Impact on loan probability (-) (neutral) (+) Figure 2 The Political Movement Hyposis Pol, more consistently faithful Soviet allies, received any IMF fund ing during this period.46 Frey applies Hirschman's neutrality model formulate a model of bilateral aid-giving process in a bipolar world where aid recipients can play two superpower donors off one anor.47 On a more gen eral level, McKeown models aid as a relationship formally sequen tial bargaining game between lending principal borrower.48 The lender exchanges aid for a realignment by developing country ward position of lender. The borrower moves from its "ideal" a new point equilibrium point where marginal utility of addi tional aid received equals domestic loss incurred by anor move away from its ideal position. I adapt extend se central in sights hyposize that movement ward U.S. increases a country's probability of a receiving loan from IMF. I portray lending process as a dynamic game between each bor rower a single lender. I do not model this interaction in formal game-oretic terms, nor do I model relationship between U.S. IMF. Rar, I assume U.S. plays role of principal within IMF, generate testable hyposes about relationship be tween Fund borrowing countries, evaluate m empiri cally. If data confirm se hyposes, it would strongly suggest, but not directly confirm, a predominant U.S. presence in an increas ingly important multilateral organization. Such a conclusion would have important implications for study of international institutions as regimes, well as for multilateral management of interna tional economy. 46 Asset (fn. 43), Bruno S. Frey, International Political Economies (London: Basil Blackwell, 1984), chap. 5; Albert O. Hirschman, "The Stability of Neutralism: A Geometrical Note," American Economic Review 54 (March 1964). 48 Timothy J. McKeown, "Resolving 'Conditionality Paradox' in U.S. Bilateral Foreign Aid" (Manuscript, University of North Carolina, Chapel Hill, n.d.).
12 HIGH POLITICS OF IMF LENDING 49 A B C D Voting scale 0 -> <- Impact on (-) (+) loan probability Figure 3 The Impact of Political Movement over Time Figure 2 captures basic argument of movement hy posis. For simplicity's sake, I present this as a linear relationship, though future research might loosen this assumption. It is based on same 0-1 voting space. Rar than measure a country's absolute polit ical alignment, however, it charts change (or in realignment) that position from one time period anor. The maximum distance a country can move vis-?-vis lender is?1.0. Countries that make large movements ward U.S., such as at point B, have a greater chance of IMF receiving credit than those that make movements away from U.S., such as at point A. Figure 3 brings spatial tem poral sides of sry ger illustrate effect of re alignment from one voting cycle next. A country shifting from point A point B has a better chance of a receiving loan than one moving from point D point C, even though point C is still closer lender's position than is point B. Operationalization of Hyposes Because ory tells us that both economic facrs affect IMF a lending, model that excludes eir category is, by definition, A misspecified. combined economy approach addresses one of Bird's main concerns: [S]ome countries may be able muster more support amongst membership of Executive Board than ors. The problem is that such  facrs are difficult model formally include in econometric estimation, but ir exclusion may dem functions which eco explain why rely exclusively upon nomic characteristics will leave much of sry unld.49 It may be difficult model test facrs econometrically, but it is certainly possible. 49 Bird (fn. 1,1995),
13 50 WORLD POLITICS The analysis focuses not on size of loans, interest rate, or or conditions but simply on decision lend. The structure of Fund leaves more room for facrs enter in pro cess of loan approval than in formation of terms of loans mselves. The ly appointed executive board generally votes yes or no on a complete package that has been assembled by staff, based largely upon market conditions its own modeling fore casting. Furrmore, Fund rules on confidentiality make data collec tion on most of terms conditions of loan packages impossible.50 This section presents hyposes introduced above in form of a pooled logit model of IMF lending eighty-seven developing countries from The dichomous nature of depen dent variable requires use of logit estimation, which treats rela tionship between a categorical dependent variable ( probability of receiving a loan) continuous independent variables as a nonlin ear one that approaches both 0 1 asymptically.52 The basic logit can be expressed symbolically as: ln[(p(l = = = l)iv)/(l-(p(l 1)?/)] b- + ^(BalPay)?2(Debt) -?3(PerCapY) -?4(Default) +?5(USX) +?6(USDFI) + +?7(PolProx)?g(PolMove) L is a dichomous variable coded 1 if country i received an IMF St-by Arrangement (SBA) or Extended Fund Facility (eff) loan during calendar year t, 0 if it did not.53 P(L = 1). is estimated 50 Regressions were run on amount of loan divided by GNP, general results were sim ilar those reported here, particularly for variables. 51 This figure represents all of developing countries, as defined by IMF, for which data were available. See IMF, AnnualReport (Washingn, D.C: IMF, 1986), 162. Data for indicar of polit ical alignment used here are not available before For some countries, data are available only for certain years. See Appendix B for a list of countries used in data 52 analysis. See John H. Aldrich Forrest D. Nelson, Linear Probability, Logit, Probit Models (Beverly Hills, Calif: Sage, 1984). 53 Two or IMF lending programs, Structural Adjustment Facility (SAF) Enhanced Structural Adjustment Facility (ESAF), are not included in this analysis for a number of reasons. First, only low-income developing countries qualify for SAF ESAF loans. A large number of countries in sample would refore not qualify for se programs, while all are eligible for SBA EFF pack ages. Second, SAF ESAF are structural adjustment rar than economic stabilization programs. To include m in analysis would a require different underlying macroeconomic model than that specified for SABs EFFs. Third, 1987 was first full year of operation for SAF 1988 for ESAF. SBA Only EFF programs were operational throughout entire time period examined here. See Polak (fn. 6,1991); Susan Schadler, Adam Bennett, Maria Carkovic, Louis Dicks-Mireaux, Maruo Mecagni, James H. J. Morisink, Miguel A Savastano, "IMF Conditionality: Experience under St-By Extended Arrangements. Part I: Key Issues Findings," IMF Occasional Paper, no. 128 (1995). Compared number of SBAs EEFs, re have been few SAF ESAF loans
14 HIGH POLITICS OF IMF LENDING 51 that a probability country will receive a loan in year t54 is inter bq cept term. "BalPay" is balance of payments position; "Debt" is debt burden; "PerCapY" is per capita income; "Default" is a measure of credit hisry; "USX" is amount of U.S. exports country; "USDFI" is amount of U.S. direct foreign investment in coun try; "PolProx" is proximity U.S. in international vot ing space described above; "PolMove" is movement ward U.S. within same space. The economic variables are lagged by one year variables by one two years establish direction of causality. To pose a more test challenging for model, I isolate a number of critical macroeconomic facrs expected affect Fund be havior select several statistical proxies for m. The balance of payments is operationalized in several different variables. is BOPf.M overall balance of payments of a country in year /-?.55 ABOPiVis change in overall balance of payments from t-1 /. PCBOP^ APCBOP.^ are per capita balance of payments change in per capita balance of payments, respectively.56 CACCI\ _2 is current account, ACACCTV is change in current account. CACCT/GNPitA ACACCT/GNPit capture ratio of current account GNP change in that ratio.57 Since higher payments deficits are thought increase chances of receiving a loan, all coefficients should be negative. A country's debt burden is measured by following variables: DEBT^j is level of absolute public publicly guaranteed debt in year /-i, ADEBTy is change in that level of debt from t-1 t. PCDEBTV APCDEBTV compute per capita debt figures. A se made. run Regressions on a variable including all of se programs ger yielded results generally consistent with those reported in following section. 54 Logit transforms this variable, which has a nonlinear relationship independent variables, in log-odds of a receiving loan, which has a linear relationship independent variables. The new dependent variable, or logit, is n on regressed independent variables using maximum like lihood estimation (MLE). Data for this variable were gared from IMF, Annual Report (Washingn, D.C: IMF, various issues). 55 All economic variables except ratios are expressed in millions of 1990 U.S. dollars, using 1990 U.S. GDP deflar reported in IMF, International Financial Statistics Yearbook. (Washingn, D.C: IMF, various issues). 56 These variables make figures for large small countries more comparable. I also tested ratio of balance of payments GNP change in that ratio with same substantive results. Data are from IMF, International Financial Statistics Yearbook (Washingn, D.C: IMF, various issues). 57 The World Banks debt ratios (DEBT/GNP, INT/GNP, RES/DEBT) appear have been multiplied by 100. To make across comparisons units consistent, I multiplied CACCT/GNP ratios calculated from (but not listed in) World Bank data by 100. World Bank, World Debt Tables (Wash ingn, D.C: World Bank, various issues); idem, Global Development Finance (Washingn, D.C: World Bank, various issues).
15 52 WORLD POLITICS ries of ratios captures debt service burden for each economy. DEBT/GNP./_t is ratio of long-term tal debt scks (public private) GNP, tracks ADEBT/GNP., change in that ratio. INT/GNP,, AINT/GNP., are variables for ratio of inter est payments GNP its change, RES/DEBT v ARES/DEBT v measure ratio of reserves GNP change in that ratio.58 Because a heavy debt burden increases debrs' dem for loans because Fund is hyposized give greater supply con sideration larger debrs,59 all coefficients should be ex positive cept those for RES/DEBT Vi ARES/DEBTV, which are expected be negative. PCAPY?M represents per capita GNP, computed from data reported in World Bank's World Debt Tables Global Development Finance in IMF's International Financial Statistics, Lower-income coun tries should be more likely receive loans, so a negative coefficient is expected. DEFAULT v is a dummy variable coded 1 if a country has had a prior IMF program canceled any time since 1975 ( first full year of EFF program) through year /-_?, 0 orwise. Since a bad credit his ry should adversely affect likelihood of future loans, its coefficient should be negative.60 USX. j is level of U.S. exports a country. According neomarxist hyposis, it should be positively signed. The domestic level economy perspective has ambiguous expectations for direction of this effect.61 USDFI. tt-i r is value of sck of U.S. direct investment for all industries in a country. are Expectations similar those for USX V_r62 The voting space depicted in Figures 1 3 is measured by an index of KVOTE/V_v agreement between country / U.S. in year t-2. Calculated as a decimal between 0 1, this variable measures degree of coincidence between votes of sample 58 These figures are from World Bank, World Debt Tables (Washingn, D.C.: World Bank, var ious issues); idem, Global Development Finance (Washingn, D.C.: World Bank, various issues); with population data taken from IMF, International Financial Statistics Yearbook (Washingn, D.C.: IMF, various issues). 59 Adequate data on exposure of U.S. banks in particular countries are unavailable. In any event, largest credir banks are likely be based in U.S. IMF's or principal shareholder countries. 60 A variable measuring tal number of cancellations that a country experienced from 1975 through t-1 did not yield statistically significant results. Data were gared from IMF, Annual Report (Washingn, D.C.: IMF, various issues). 61 Data are from IMF, Direction of Trade Statistics Yearbook (Washingn, D.C.: IMF, various issues). 62 Data have been taken from U.S. Department of Commerce, Survey of Current Business, vari ous issues.
16 HIGH POLITICS OF IMF LENDING 53 country U.S. in United Nations General Assembly (UNGA) on ten approximately fifteen issues in each session that U.S. Department of State has deemed votes. key Under man Congressional date, State Department has compiled voting records of all UN member nations on se selected issues since 1983 General As sembly in its annual on publication, Practices Report Congress Voting in United Nations.63 Appendix A lists key votes identified in this report for years examined here. In accordance with a proximity hyposis, positive coefficient is anticipated.64 Using UN voting patterns measure international align ment is one solution problem of testing variables lamented by Bird. For its part, U.S. government has proclaimed that examining UN votes makes it "possible make judgments about whose values views are harmonious with our own, whose policies are consistently opposed ours, whose practices fall in between."65 But not all UN votes are equally important. In reference key votes, same states report that votes "only that can legitimately be read as a measure of support for United States are those which we identified as important us, on which we lobbied or nations"66 The validity of UNGA voting records has been debated extensively Using se annual reports, I coded votes in agreement with U.S. 1.0, votes in disagreement with U.S. 0.0, abstentions or absences by sample country n added divided se numbers by tal number of key votes each year come up with annual decimal measure for each country. This method differs slightly from technique of discarding absences absten tions from tal count of UNGA votes used in Charles W. Kegley Jr. Steven W. Hook, "U.S. Foreign Aid U.N. Voting: Did Reagan's Linkage Strategy Buy Deference or Defiance?" Interna tional Studies Quarterly 35 (September 1991). Rar than not count those nonvotes on "key" issues, I interpret m as neutral. 64 The transmission of United States foreign policy preferences from State Department is not necessarily direct in case of multilateral development banks IMF, where Treasury plays a critical role. See Schoultz (fn. 10). The (American) deputy managing direcr has typically been "a Treasury man,' reinforcing close ties between that agency of U.S. government IMF." Kahler (fn. 16), 94. Furrmore, Kahler argues that Treasury maintains tight control over U.S.-Fund relations that "or agencies that might attempt politicize IMF for broader foreign policy goals tended be excluded from direct access it." Kahler (fn. 16) 94,97. On or h, Joanne Gowa notes that Treasury has an adopted ordering of priorities that "subordinates dems of international monetary order imperatives of domestic economic policy foreign security pol some icy," suggesting coordination?or at least compatibility?between different agencies within executive branch. Gowa, Closing Gold Window: Domestic Politics End of Bretn Woods (Ithaca, N.Y.: Cornell University Press, 1983). The present analysis of policy output (as opposed in teragency input) is an indirect test of se two competing hyposes. Future work should address interagency more dynamics directly. 65 U.S. Department of State, on Report Congress Voting Practices in United Nations (1985), Ibid., See Soo Yeon Kim Bruce Russett, "The New Politics of Voting Alignments in United Nations General Assembly," International Organization 50, no. 4 (1996); Steven K. Holloway Rodney Tomlinson, "The New World Order General Assembly: Bloc at Realignment UN in Post-Cold War World," Canadian Journal of Political Science 28, no. 2 (1995); Leona Pallansch
17 54 WORLD POLITICS adopt self-identified measure of alignment of principal U.S. foreign policy decision-making body: UNGA key votes. There is also evidence that State Department tracked such data in a similar manner previous Congressional mate, that it considered UN votes a reliable indicar of that U.S. allocated aid alignment, on basis ofthat alignment. In a 1964 memo direcr of Food for Peace Dick Program, Reuter, Lansdale noted that "at critical moments in world's recent hisry, U.S. votes 'bought' subtly indirectly support its st in General Assembly. The 'buying' is in terms of U.S. assistance voting country."68 Furrmore, Lansdale's a measure analysis employed of alignment similar cur rent State use Department of key votes, votes on charting only cold war issues. It also appears that at least some recipient countries take U.S. vigi lance of UN voting seriously. Argentina, for a example, previously leader in Non-Aligned Movement, modified its voting stance in UNGA reflect better its improved relations with United States in early 1990s.69 In a 1997 interview, Carlos Escud?, a former adviser Argentina's minister of foreign relations, revealed that "with respect some important United Nations resolutions, re was direct contact between Argentina United States, Argentina voted in a manner favorable United States."70 More generally, between Argentina altered its UN votes move from fourth, most anti-u.s. stance in UN a position similar that of Turkey.71 MKVOTE. _j measures movement in alignment be tween sample country U.S. within voting space from Frank Zinni Jr., "Demise of Voting Blocs in General Assembly of UN? A Multidimensional Scal ing Analysis" (Paper at presented annual meeting of Sourn Political Science Association, At lanta, 1996); Brian W. Tomlin, "Measurement Validation: Lessons from Use Misuse of UN General Assembly Roll-Call Votes," International Organization 39, no. 1 (1985); Kenneth J. Menkhaus Charles W. Kegley Jr., "The Compliant Foreign Policy of Dependent State Revis ited: Empirical Linkages Lessons from Case of Somalia," Comparative Political Studies 21, no. 3 (1988). 68 Ed Lansdale, "Memo Re: Long Range Impact FPF-II," April 24,1964, National Archives, Record Group 59, Lot file 67D554, Under Secretary for Political Affairs, Records of Special Assistant , Box 2.1 thank Tim McKeown for me providing with a transcription of this document. 69 For sent example, Argentina troops 1991 Persian Gulf conflict. Carlos Escud?, "Entrevista a Escud? realizada por Lorena Kniaz" (1997), cited May 19, 1999, CapilHill/Congress/4359/reporta.html. 70 Ibid. 71 Ibid.; Carlos Escud?, from author, February 16,1999.
18 HIGH POLITICS OF IMF LENDING 55 year t-2 year t-1, measured in UNGA key votes.72 From movement I a hyposis, expect positive coefficient. Findings Interpretations Table 1 presents results of three different versions of model. The combination of pooled data a categorical dependent variable presents unique diagnostic challenges.73 Column 1 presents results for basic logit model, with no correction for aucorrelation. It ap pears provide a good overall fit: -2 times log-likelihood ratio (-2 x LLR) for model is , with p < We can reject null hyposis that none of independent variables individually or a amount collectively explain significant of variation of dependent variable.74 The model correctly predicted percent of out comes. In terms of individual coefficients, PCAPY. ' tt-l, tt-2 KVOTE. 0 are significant at 0.90 level of confidence; BOPy 1 PCDEBTV t at 0.95; DEBT/GNP., v INT/GNP.,,, RES/ AI?T/GNP.,, DEBTVJ, MKVOTE,Jat 0.99; all with anticipated signs. To test correct for aucorrelation, I employed binary time series-cross-section estimation technique formulated by Beck, Katz, Tucker.751 constructed a series of nine (T-1) dummy variables coded 1 if it had been... (1, 2, 3, T-1) years since a country last re ceived an IMF loan, 0 orwise. If se nine variables are collectively in a significant log-likelihood ratio test, it is an indication of aucorre lation. The correction for aucorrelation is simply inclusion of temporal dummy variables in estimation. Once corrected, new coefficients for original variables of interest should better satisfy 72 Because of UNGA s voting calendar, voting variables have a longer lag structure than economic variables. The fact that UNGA votes are taken in last four months of calendar year means that re is a 67 percent chance that a given loan decision will be made before UNGA meets in a given year. The chances that such a decision will be made before session is complete final votes are tallied approaches 100 percent. movement at Conversely, t-1 occurs immediately before next calendar year s loan cycle begins. 73 James A. Stimson, "Regression in Space Time: A Statistical Essay,"American Journalof^Polit icalscience 29, no. 4 (1985); Nathaniel Beck jonathan N. Katz, Richard Tucker, "Taking Time Se riously: Time-Series-Cross-Section Analysis with a Binary Dependent Variable," American Journal Political Science 42, no. 4 (1998). 74 This assumes a Chi Square distribution for -2 x LLR figure. While this assumption may not be entirely valid for individual level data, strong results are still encouraging. 75 Beck, Katz, Tucker (fn. 73). This approach is designed for longitudinally dominant data with typically twenty or more time periods. The authors have not yet tested this explorary method on shorter time periods like one used here (T = 10). Richard Tucker, conversation with author, August We may refore have somewhat less confidence in a negative diagnostic for aucorre lation than in one positive obtained here.
19 Table 1 Logit Coefficient Estimates of imf Lending, Independent With Temporal Refined with Variables Basic Logit (1) Dummies (2) Dummies (3) PCBOP/V_, x 10-4** x 10"4** x 10"4** BOP.M (0.911 x lo'4) (0.926 x 10"4) (0.583 x 10"4) ABOP x x 10"6 ( x 10-6) ( x 10"6) x 10" x 10"4 ( x 10-4) ( x 10~4) x 10" x 10"4 APCBOPt., ( x 10"4) ( x 10'4) CACCT._t x 10" x 10'4 (.956 x 10-4) (0.974 x 10"4) ACACCT, x 10" x 10" x 10"4** (0.976 x 10-4) (1.031 x 10-4) (0.632 x 10"4) CACCT/GNPiVi * (0.026) (0.027) (2.044) ACACCT/GNP (0.024) (0.025) DEBT.M x 10~* x 10"6 ( x 10"*) ( x 10"6) ADEBTV x 10" x 10"5 (6.63 x 10~5) (6.730 x 10'5) PCDEBT. _, x 10"4** x 10"4* x 10"4* (4.58 x 10"4) (4.655 x 10"4) (4.340 x 10"4) APCDEBT x 10" x 10"4 DEBT/GNPiV_; ADEBT/GNP., INT/GNP.^ AINT/GNR, RES/DEBT._, ARES/DEBT, (12.84 x 10"4) ( x 10"4) *** *** *** (0.003) (0.003) (0.003) (0.003) (0.004) 0.284*** 0.274*** 0.267*** (0.065) (0.069) (0.065) 0.503*** 0.519*** 0.516*** (0.100) (0.102) (0.099) -O.026*** *** ** (0.010) (0.010) (0.010) x 10" x 10"4 (0.021) (0.022) x 10"4* x 10" x 10"4* PCAPY._, ( x 10-4) (2.122 x 10"4) (2.052 x 10"4) DEFAULT iw 0.394* 0.412* 0.467** (0.231) (0.238) (0.226) x 10" x 10"6 USX._t ( x 10-*) ( x 10"6)
20 HIGH POLITICS OF IMF LENDING 57 Table 1 (cont.) Independent With Temporal Refined with Variables Basic Logit (1) Dummies (2) Dummies (3) USDFI jv_, x 10" x 10~5 (8.800 x 10"5) (8.880 x 10"5) KVOTEiV * (0.716) (0.728) (0.677) MKVOTE.^ 2.756*** 2.858*** 2.711*** (0.795) (0.856) (0.827) Intercept *** *** *** (0.476) (0.522) (0.490) Correctly predicted (%) Model x p-value p< < p < p d.f. 33 d.f. 21 d.f. N = 746. Stard errors are in parenses below estimates. Coefficients for temporal dummies not reported. * Significant ** Significant *** Significant at p < 0.10 at p < 0.05 at p < 0.01 level. level. assumption of error independence. This test revealed a high likelihood of aucorrelation (log-likelihood ratio = with 9 d.f, p < 0.025). Columns 2 3 in Table 1 report results of full model a more refined model correcting for aucorrelation. Tests were also conducted for multicollinearity heteroskedastic ity. Neir revealed any indications of problems. Despite presence of a large number of potentially overlapping economic variables, none of variables in model exhibited high degrees of collinearity with or variables (eir collectively or individually). To test for het eroskedasticity, I a incorporated variable for GNP test impact of country size a series of eight dummy variables capture effect of geographic region. The results of se tests did not con approach ventional levels of statistical so significance, I retained assumption of homoskedasticity. The parameter estimates of corrected full model (Column 2) yield several interesting, albeit tentative, findings. The only significant balance of payments variable is overall balance, BOPjV r As ex pected, its coefficient is negatively signed, suggesting that a country