IMF conditionality and the theory of special interest politics (1).

Mayer, Wolfgang ; Mourmouras, Alex

INTRODUCTION

The International Monetary Fund (IMF) was created in the aftermath of World War II to promote stability and prosperity in the international monetary and financial system. A key function of the IMF is to make loans to countries facing shortages of foreign exchange. The IMF's financial support gives countries breathing room, while they adjust their economic policies to solve their external payment difficulties. To ensure that borrowing governments follow policies that are consistent with the IMF's purposes, its financial resources are disbursed gradually and on condition that borrowers avoid measures that are destructive to national and international prosperity. The IMF's rules for disbursing assistance have evolved gradually and have been continuously adapted to solve emerging problems--recycling petrodollars in the 1970s, overcoming the debt crisis in the 1980s, and integrating the transition countries and dealing with capital-account crises in the 1990s.

While necessary, conditionality has become somewhat controversial over the years, leading to periodic reviews--within the IMF and by the broader academic community--of the analytical principles underlying it and the practical ways in which it has been applied. The most recent such internal review was conducted during 2000-2002, with extensive external input and participation. The review highlighted the critical importance of country ownership, defined as the ability and willingness of country authorities and other key stakeholders to implement policy programmes agreed with the IMF, which was underscored by the poor record of some programmes that were not based on reasonably broad coalitions in recipient countries. It concluded that while country ownership of policies depends in the first place on the borrowing countries' domestic political economy, it can be influenced by the IMF's methods of interaction with these countries. Reforms in the IMF's processes now under way, including the streamlining and focusing of conditionality and greater tailoring of programmes to the circumstances of member countries, will ensure that IMF-supported reforms will enjoy greater domestic ownership. The importance of promoting country ownership and of limiting conditionality to measures that are critical to the success of programmes is reflected in the revised IMF conditionality guidelines (IMF, 2002).

This paper presents some analytics of IMF conditionality based on the theory of special interest politics (Grossman and Helpman, 2001). This theory appeals to lobbying by special interest groups in order to explain why policymakers in the real world often prefer to use distortionary policies. The next section presents our political-economy model of the IMF, which applies the theory of special interest politics to understand the role of IMF conditionality in shaping economic policies of recipient governments (Mayer and Mourmouras, 2002). The subsequent section discusses how conditionality helps matters in the presence of various domestic and international conflicts of interest. The last section concludes.

THE POLITICAL ECONOMY OF IMF CONDITIONALITY

Overview

To be an effective incentive system (Dixit, 2000), IMF conditionality must guide policymakers in recipient countries to improve their policy choices. This begs the question of why these choices are distorted and how IMF conditionality helps matters. The theory of principals and agents, which focuses on the role of incentives and strategic interactions, is useful in addressing both questions. The IMF is an agent of the international community that created it, established its mandate, and endowed it with financial resources. The conditionality attached to IMF loans provides the international community with certain payoffs. Principal among these are the assurances that these loans will be repaid. Since the IMF does not impose collateral requirements, conditionality is part of the IMF's financial due diligence. In addition, conditionality helps steer borrowing countries toward stable and outward-oriented economic policies and avoid measures that are destructive of national or international prosperity (Article I (v)). Finally, conditionality also guarantees borrowing countries certain rights. Once an IMF loan is committed, borrowers are assured that they will continue receiving IMF financing if they meet the conditions specified in the loan agreement. (2) The IMF cannot arbitrarily 'move the goal-post' after an agreement is signed. The IMF's internal rules also guarantee that borrowing countries are not unduly discriminated against (uniformity of treatment).

Recent developments in the political-economy literature address why economic policy choices are distorted and how IMF conditional loans assist in reducing these distortions. The key innovation has been the move away from the unitary-actor view of borrowing countries. These multi-actor models feature domestic conflicts of interest that prevent recipient-country governments from adopting first-best policies. The lobbying models of the theory of special interest politics is one such approach. Veto-player models are another. (3) In lobbying models, a government's policy choices reflect the joint influence of organised interest groups as well as the pursuit of public welfare. The mathematical foundation of special interest politics is the theory of common agency, which is concerned with strategic interactions of multiple actors (principals) who compete in their efforts to influence a common decision maker (agent). (4)

In this paper, we embed the IMF--modelled as an international lobby that pursues the international public interest--in a standard lobbying model. The combined influence of the IMF and domestic interest groups on government policy choices is analysed using an approach that is standard in the theory of special interest politics. Equilibrium policy choices are the outcome of strategic interactions between the IMF, borrowing governments and domestic lobbies in an interdependent world economy. Conditionality furthers the IMF's broad internationalist objectives and helps resolve conflicts of interest that surface between the IMF and debtor governments. (5) A key feature of our model is the assumption that policy choices have cross-border spillover effects. While national actors--policymakers and lobbies alike--may ignore such externalities, we assume that the IMF takes them into account in setting its conditionality.

The impact of organised special interest groups (SIGs) on economic policy formation has been recognised since at least the work of Olson (1965). Interest groups that overcome the free-rider problems present in collective action can influence policymakers to adopt policies that favour them. In many cases, favours to SIGs come at the expense of general welfare, yet the general public is broadly dispersed and unorganised. The effectiveness of lobbying by vested interests increases with the extent of the imperfection in national economic and political systems. In countries in which the preferences of ordinary citizens are not very well reflected in government decisions, vested interests can assume control of the agenda and dictate policies to the detriment of the general welfare. Imperfections in economic and political systems reflect costly and limited information about the impacts of policy changes and uncertainty about the identity of winners and losers, which makes it difficult for winners to compensate losers. These difficulties are compounded when different countries are involved and place considerable limits on the range of feasible international economic policy choices.

Key to the success of interest groups' efforts to influence policies is policymakers' electoral and selfish concerns. At a general level, all politicians weigh the political benefits and costs of alternative policy choices and choose actions that maximise their overall political support. Political support depends on national well-being (economic growth, poverty reduction, etc.), as well as backing by influential stakeholders. In practice, special interests play a key role in constraining policy reforms in developing and transition countries (the IMF's main borrowers). (6) Despite the spread of democracy and pluralism in recent decades, governments in many developing and transition economies lack sufficient transparency and accountability. Many institutions needed for the effective exercise of political 'voice' and the operation of checks and balances remain weak, and ordinary citizens have limited information about their rights or their government's responsibilities. These imperfections make it easier for vested interests and veto players to exert large political influence. To give but one example, banks, state enterprises and farms in some transition and developing countries were captured by small groups of politically connected people. Some transition-economy enterprises became instruments of vested interests, earning rents for politically powerful groups at the expense of producers or consumers. In the process, growth, equity and aggregate welfare declined more than expected, making transition a much more arduous process than originally anticipated.

Recent applied work, including papers written for the IMF's 2000-2002 conditionality review, underscored the key role of domestic divisions and special interests in limiting the effectiveness of IMF-supported reforms (IMF, 2001a-c). Ivanova et al. (2003) show that the implementation of IMF-supported programmes suffers in countries in which an index of the strength of special interests is strong. The quality of key institutions is important for understanding programme implementation and macroeconomic performance in countries implementing IMF-supported reforms (Nsouli et al., 2004). The importance of overcoming domestic divisions and forging consensus behind reforms is often couched in terms of the need to enhance country ownership for sound policies. Though the evidence--from case studies and econometric work --is indirect, it does point to the key role of ownership for the implementation of IMF-supported reforms (Bird, 2002; Boughton and Mourmouras, 2002; Khan and Sharma, 2003; Willett, 2003). This evidence is similar to findings about World Bank-supported programmes (Dollar and Svensson, 2000).

The empirical studies also conclude that IMF financing and conditionality have had inconclusive impacts on countries' macroeconomic and structural performance. A key theme is the relatively weak record of programme implementation. If programmes are not consistently implemented, their desired effects will not be realised, and the considerable resources countries invest in negotiating programmes will not yield high returns, as Bird (2002) points out. (7) Reconciling the theoretical case for conditionality, as it emerges from the model of special interest politics presented in this paper, with the nuanced empirical evidence is a key challenge in designing conditionality.

A model

The incentive effects of IMF conditionality can be analysed with the help of a simple political-economy model (Mayer and Mourmouras, 2002). Consider a world economy consisting of two groups of countries, those who are creditors of the IMF and those making use of IMF resources (debtors). The IMF is a financial intermediary that represents the interests of both groups of countries. It acquires real resources (capital) T from creditor countries and transfers them (on loan) to the governments of debtor countries. The IMF's objective is to help countries address policy distortions that reduce national and international welfare. We assume that IMF assistance is channeled in socially beneficial ways. We also abstract from time inconsistency and credit risk by assuming that repayment of IMF loans is guaranteed at the contracted (but possibly subsidised) interest rate.

Let W and [W.sup.*], respectively, denote gross national welfare in debtor and creditor countries after IMF resources have been disbursed but before they have been repaid to the IMF and its creditors. Letting b be the gross rate of IMF repayment (unity plus the IMF's rate of charge), then net social welfare in a debtor country after the loan has been disbursed and repaid is equal to gross welfare minus loan repayments, or

(1a) Y = W - bT

Likewise, net social welfare in creditor countries, after IMF loans have been disbursed and repaid, is

(1b) [Y.sup.*] = [W.sup.*] + bT

The IMF's objective is to maximise the weighted sum of net world welfare, or

(2) I = [Y.sup.*] + [gamma]Y

where [gamma] [greater than or equal to] 0 is the weight which the IMF attaches to the utility of debtor countries. This specification of the IMF's objective function is consistent with the Articles of Agreements, which mandate the institution to safeguard the stability and prosperity of the international monetary and financial system. (8)

In addition to the IMF (I), the model features two other actors, the incumbent government of the debtor country (G), and a domestic SIG in that country. As in Grossman and Helpman, the government maximizes a political support function which depends on the general public's welfare and on SIG campaign contributions. We denote by [omega] [greater than or equal to] 0 an index of policy distortions under the control of the government. The country's social welfare would be maximised by following first-best policies (i.e., setting the distortion index to zero). The government's objective function is a weighted sum of SIG contributions, C([omega]), and net social welfare:

(3) G([omega], T) = C([omega]) + aY([omega], T)

The parameter a [greater than or equal to] 0 is the government's concern for its public's welfare relative to interest group contributions. Maximising (3) will result in policy choices that deviate from first-best.

The SIG benefits from distorted economic policies. Its gross utility, U([omega]), increases at a decreasing rate with the degree of distortions ([U.sub.[omega] > 0 and [U.sub.[omega][omega]] < 0). The SIG's net utility is gross utility less political contributions,

(4) V([omega]) = U([omega]) - C([omega])

The SIG designs its contribution schedule C([omega]) to maximise its political influence. The assumption that the interest group's welfare rises with the degree of policy distortions can be derived from microfoundations (see next section), but it can also be justified heuristically as follows. While there are many possible policy distortions, some benefiting and some hurting the SIG, it will pay the interest group to apply pressure only for the adoption of those distortions from which it will benefit. Distortions whose net effect hurts both the SIG and the general public are certainly possible. But the SIG will not push for them and so we can disregard them.

Several additional assumptions are made in what follows. First, IMF assistance is subject to diminishing returns, implying that W rises with T but at a decreasing rate ([W.sub.T] > 0, [W.sub.TT] < 0). Second, social welfare declines at an increasing rate as distortions rise ([W.sub.[omega]] < 0, [W.sub.[omega][omega]] < 0). This assumption is consistent with the fact that the excess burden of distortionary taxes or subsidies is an increasing and convex function of the level of taxes or subsidies. Third, greater distortions reduce the effectiveness of IMF assistance ([W.sub.T[omega]] < 0). This assumption is consistent with the results in the aid-effectiveness literature that foreign assistance is more valuable in good policy environments (Burnside and Dollar, 1998, 2000). Fourth, since creditor countries finance IMF lending, they incur financial costs by providing assistance but reap benefits when debtor countries repay the IMF, so that [W.sup.*.sub.T] < 0 and [W.sup.*.sub.TT] < 0. Fifth, the world economy is interdependent in the sense that the policy distortions of countries making use of IMF resources may affect welfare in creditor countries ([W.sup.*.sub.[omega]] < 0 and [W.sup.*.sub.[omega][omega]], < 0 if a debtor country is 'systemically' important). On the other hand, the IMF provides conditional assistance to small countries (those whose policies do not influence conditions in creditor countries, or [W.sup.*.sub.[omega]] < 0) only to the extent that it has altruistic concerns for them. Finally, we assume [U.sub.[omega]](0) + a-[W.sub.[omega]] (0,0) > 0. This condition states that if the recipient country is left on its own, then it is not in the government's interest to maintain undistorted policies (i.e., if T = 0, then the government will not adopt the first-best policy [omega] = 0).

Micro foundations

The reduced-form model just outlined can be derived from rigorous microfoundations. Consider an economy consisting of two sectors, industries [X.sub.1] and [X.sub.2] with production functions [X.sub.j] = h(T)[f.sub.j]([L.sub.j], [K.sub.j]), where [L.sub.j] and [K.sub.j] are labour and capital employed in industry j. Private capital is industry-specific and does not flow between sectors (specific-factor model), while labour is perfectly mobile between sectors ([L.sub.1] + [L.sub.2] = L). Foreign assistance T is used to build up the infrastructure needed to underpin a market economy, broadly construed to include physical infrastructure as well as institutions such as property rights and law enforcement. Assistance results in an unbiased expansion of the economy's production possibilities.

The country in question is small (price taker), markets are perfectly competitive, and the relative world price of good X in terms of good Y is [pi]. A SIG owns private capital specific to industry 1 and makes contributions to the government in exchange for production subsidies paid to industry 1. The subsidy results in misallocation of resources, lowering national income. National income is

(5) W = [pi][X.sub.1], + [X.sub.2] = [pi]h(T)[f.sub.1]([L.sub.1], [K.sub.1]) + h(T)[f.sub.2]([L.sub.2], [K.sub.2])

where h'(T) > 0 and h"(T) < 0. The domestic price of [X.sub.1] in the presence of the subsidy is P = [pi] (1 + [omega]), where [omega] is the production subsidy rate. The return paid to the industry-specific capital employed in industry 1 is

(6) [r.sub.1] = [pi](1 +[omega])h(T)([differential][f.sub.1]/[differential][K.sub.1])

In this setting, the following propositions can he established (proofs are available from the authors). First, a higher production subsidy lowers the country's national income, and this decline is more pronounced the more assistance is received. Second, if assistance increases, then [differential][[differential]W/[differential][omega]/[differential]T < 0. Finally, the welfare of the SIG, U([omega]) = [r.sub.1]([omega])[K.sub.1], rises with subsidies, or [differential]U/ [differential][omega] > 0. These propositions justify the assumptions [W.sub.T[omega]] < 0 and [differential]U/[differential][omega] > 0 made in the previous section.

Political equilibrium

Following Bernheim and Winston and Grossman and Helpman, the outcome of the strategic interaction between the borrowing-country government, the domestic SIG, and the IMF, can be analysed as an equilibrium of a non-cooperative, multi-stage game. The nature of the game (timing and other aspects of the strategic interactions between the three players) depends on the type of assistance offered by the IMF, whether unconditional or conditional.

Unconditional assistance

The political equilibrium with unconditional assistance is ([[omega].sup.0], [C.sup.0]) the outcome of a three-stage non-cooperative game. In the first stage, the IMF decides on the total amount of assistance. In the second stage, the SIG tenders a contribution schedule to the government. This schedule is designed to maximise the SIG's net welfare. In the third stage, the government chooses the index of policy distortions to maximise its political support, taking as given the interest group's contribution schedule and the IMF's assistance.

Without loss of generality, we focus on truthful equilibria. Truthful equilibria are a refinement of Nash equilibria and are unique. Their key property is that SIG contributions accurately reflect the group's gains from increased policy distortions. If the equilibrium is truthful, then in response to an increase in policy distortions the lobby raises its contribution by an amount equal to the marginal utility it derives from the additional distortions: [U.sub.[omega]]([[omega].sup.0]) = [C.sup.0.sub.[omega]],([[omega].sup.0]). As Grossman and Helpman (1994) demonstrate, focusing attention on truthful equilibria does not restrict players" strategy spaces.

The model is solved working backwards. One first focuses on stages two and three to determine the government's choice of [omega] in the presence of the influence-seeking SIG, given the amount of assistance made available by the IMF. In stage one, the IMF's choice of assistance is determined, given the government's policy response to alternative assistance levels. Denote by [[omega].sup.0] the government's choice of the index of policy distortions when the amount of IMF assistance provided is T. The condition [U.sub.[omega]](0) + a[W.sub.[omega]](0,0) > 0 guarantees that [[omega].sup.0] is positive and maximises the government's political support function (3), implying that

(7) [G.sub.[omega]] ([omega], T) = 0 [vector] [C.sub.[omega]]([omega]) + a[Y.sub.[omega]] ([omega], T) = 0

when evaluated at [[omega].sup.0]. Substituting the truthfulness condition [U.sub.[omega]]([[omega].sup.0]) = [C.sup.0.sub.[omega]]([[omega].sup.0]) in (7) yields

(8) [[U.sub.[omega]([[omega].sup.0]) = -a[W.sub.[omega]]([[omega].sup.0], T)

Condition (8) yields the government's positive equilibrium choice of policy distortions when the IMF provides an amount T of unconditional assistance. Even though the general public's welfare in the borrowing country is maximised at [omega] = 0, the first-order condition (8) states that in political equilibrium the government chooses a positive value of [omega]. This choice balances the marginal political benefit to the government from raising distortions (and receiving additional contributions from the lobby) against the marginal political damage which additional distortions impose on public welfare.

We next turn to the question of how the IMF would choose the level of its unconditional assistance. Even if IMF loans were provided without a quid-pro-quo, efficiency would require the IMF to be familiar with the authorities' reaction to different levels of assistance. Varying the level of IMF assistance alters the debtor government's incentives because it affects the marginal damage of distortions on the general public's welfare in that country. If, as shown, higher assistance raises the marginal damage of distortions, the government's reaction function will be downward-sloping. The locus of the government's choices of [[omega].sup.0] traced by varying the amount of IMF assistance is curve RR in Figure 1. The government reaction function is downward-sloping. Its slope depends on the extent of the government's concern for the general public's welfare and on how detrimental distortions are to the effectiveness of assistance,

(9) d[[omega].sup.0]/dT = -a[[W.sub.[omega]T([[omega].sup.0], T) /[[U.sup.[omega] [omega]] ([[omega].sup.0] + a[[W.sub.[omega][omega]]([[omega].sup.0)] < 0

In terms of Figure 1, if the IMF takes into account the government's reaction function, RR, it will choose an amount of assistance [T.sub.0] that puts it on the highest attainable welfare contour, namely [I.sub.0] at point B.

[FIGURE 1 OMITTED]

Conditional assistance

In practice, the IMF's offers of assistance are contingent on the incumbent government's pursuit of distortion-reducing economic policies. The conditional assistance game is a common agency game, with the IMF and the SIG being the two principals. The government's choice of economic policies is a subgame-perfect Nash equilibrium of the following two-stage game. In the first stage of the game, the IMF and the SIG simultaneously present their truthful assistance and contribution schedule, respectively, to the government. In the second stage, the government selects the degree of policy distortions to maximise its political support while taking as given each principal's contribution schedule.

A key proposition in the literature on common agency games is that truthful equilibria do not waste resources and are Pareto optimal for the group of players who actively participate in the game (Dixit et al., 1997, Proposition 3). Our model features three such players: the borrowing government, the SIG, and the IME Since the IMF represents the general public of creditor and debtor nations, it follows that the equilibrium of the conditional assistance game is Pareto optimal for the world as a whole. Efficiency requires maximising the joint welfare of the government and the interest group and the government and the IMF. If this were not true in equilibrium, the SIG (IMF) could offer the government an alternative contribution (assistance) schedule that would be mutually beneficial. In geometric terms, equilibrium requires equality of marginal rates of substitution for the IMF and the SIG-influenced government. The resulting 'international tangency condition' between the indifference curves of the IMF and the government is point C in Figure 1. This tangency point pins down the government's choice of policy distortions and the IMF's assistance level. At this tangency condition, all mutually advantageous political trades between the IMF and the recipient government have been exhausted, taking into account the reactions of the domestic SIG. (9)

The interior conditional assistance equilibrium choice of policy distortions, [[omega].sup.1], can be characterised as follows. Equilibrium requires that the joint welfare of any two active players be maximised. Together with the truthfulness conditions, this yields the following first-order necessary conditions, evaluated at the truthful Nash equilibrium schedules [T.sup.1] and [C.sup.1] and [[omega].sup.1] > 0.

(10) [C.sup.1.sub.[omega]] + a[W.sub.[omega]] + a([W.sub.T] - b)[T.sup.1.sub.[omega]] = 0

(11) [U.sub.[omega]] - [C.sup.1.sub.[omega]] + [C.sup.1.sub.[omega]] + a([W.sub.T] - b)[T.sup.1.sub.[omega]] = 0

(12) [gamma]([W.sub.[omega] + [W.sub.T][T.sup.1.sub.[omega]]) + ([W.sup.*.sub.[omega]] + [W.sup.*.sub.T][T.sup.1.sub.[omega]] + [C.sup.1.sub.[omega]] + a([W.sub.[omega]] + [W.sub.T][T.sup.1.sub.[omega]) - (1 - [gamma] - a)b[t.sup.1.sub.[omega]] = 0

Substitution of (10) into (11) implies, just as in the absence of IMF assistance, that [U.sub.[omega]]([[omega].sup.1]). The contribution schedule equates the interest group's marginal benefit from distortions to the group's marginal cost in terms of additional contributions. Substituting (10) into (12) yields

(13) [T.sup.TE.sub.[omega]] = [gamma][W.sub.[omega]] + [W.sup.*.sub.[omega]]/[gama][W.sub.T] + [W.sup.*.sub.T] + (1 - [gamma])b

This is a key equation that illustrates a number of properties of truthful conditional assistance equilibria. First, as explained in Grossman and Helpman (2001), the truthful assistance schedule, denoted [T.sup.TE] in Equation (13), coincides with the downward-sloping part of the government's indifference curve Go. Second, the equilibrium depends on all the 'deep parameters' of the model, including those describing the preferences of the borrowing government, SIG, and IMF. Third, the conditional assistance schedule is downward-sloping--the IMF rewards reductions in distortions with more assistance. Fourth, the terms of the IMF's assistance reflect the impact of debtor-country distortions on creditor-country economies, as reflected in the value of [W.sup.*.sub.[omega]]. Fifth, the IMF's conditional assistance schedule also takes into account the creditor countries' marginal costs of providing assistance, as reflected in [W.sup.*.sub.T], and the marginal benefits of assistance to recipients, or [W.sub.T]. Sixth, in setting conditionality, the IMF also takes into account the borrowing country's weight in its objective function. Finally, recall that the IMF becomes an active participant--a second principal--in the domestic political game. In other words, the efficient outcome of the conditional assistance games is achieved at a price since the IMF is injected more deeply in the affairs of the recipient country.

Loans versus grants

The effectiveness of IMF assistance also depends on its financial terms, whether loans or grants, as reflected in the value of the repayment parameter b. Raising the value of b amounts to a reduction in the rate of subsidisation of IMF assistance. It can be shown that conditional grants are the most effective form of assistance from the point of view of the IMF (see Mayer and Mourmouras, 2002 for details). Formally,

(14) dI/db = T([W.sub.T] + [W.sup.*.sub.T])/([W.sub.T] - b) < 0 if [gamma] [greater than or equal to] 1

This condition states that if the IMF cares about recipient countries at least as much as it does about its creditors, then reducing the rate of subsidisation raises policy distortions in the assisted countries. In a world economy in which decisions are made only once, the IMF's assistance is most effective if it is provided in the most direct way possible (i.e., through conditional grants). Conditional grants enable the IMF to exercise the most leverage on the policy choices of borrowing governments. Note, however, that while these grants do not have to be repaid, they are not 'free' Grants are disbursed only if borrowing governments reduce their policy distortions, which is costly for them in terms of lost political support.

In practice, the repeated nature of the interactions between the IMF and debtor countries vastly complicates the answer to the loans versus grants question. While more subsidised assistance purchases greater reductions in policy distortions, improvements in policies may not be durable if the underlying source of the distortions--the influence-seeking activity of SIGs--is not addressed. In a multi-period version of the model in which the IMF provides assistance for only a limited time, conditionality will help for only as long as the IMF stays financially engaged. To the extent that the interest group remains active, policy distortions will increase following the withdrawal of IMF assistance. Mayer and Mourmouras (2004) analyse policies to avoid recidivism in such an environment.

IMPLICATIONS OF SPECIAL INTEREST POLITICS FOR THE IMF

The analysis of IMF conditionality presented in this paper is based on a stylised political-economy model, suitably modified to incorporate the IMF's role in the world economy. Based on this general approach, this section draws some additional implications of the theory of special interest politics for the design and application of IMF conditionality.

At a general level, the theory points to the critical need to consider the political feasibility of reforms when designing IMF-supported programmes. While IMF conditionality counteracts government incentives to pursue distortionary policies, the recipients' domestic political economy places clear limits on the IMF's leverage. The imperfections in economic and political systems of countries making use of IMF resources impose clear limits on the ability and willingness of governments to pursue first-best economic policies. Such policies are preferred by IMF staff both on technical grounds and because the IMF is a public-interest institution. However, as illustrated by the model, insisting on first-best programmes in the presence of organised vested interests could lead to problems in implementation and be counterproductive. Taking into account the limits on feasible reforms at the programme-design stage could result in more realistic, better-designed, and better-implemented adjustment programmes.

The theory makes a strong welfare case for using IMF conditionality to counter policy distortions. We have portrayed a 'representative' IMF that uses conditionality to push for the reduction of policy distortions and achieves a Pareto optimal allocation for the world economy. The results of conditional lending are superior relative to either the case of no IMF assistance or to the case where assistance is unconditional. This basic efficiency result will survive so long as the IMF's objective function includes all the citizens of the world and the strategic interactions between the IMF, the authorities and their domestic constituents do not waste resources.

On the other hand, the distribution of the gains from the financial relationship between the IMF and borrowing governments depends on the nature of the strategic game being played. In general, the outcome of the conditional assistance game depends on the number and nature of interest groups, who has the ability to make offers, and the bargaining process. In the model of this paper, conditional assistance raises world economic welfare, including of the general public in the recipient country which benefits both from improved policies and from access to IMF funds. On the other hand, the government's political support in the conditional assistance equilibrium is unchanged relative to the no assistance case. This is a special result, reflecting the perfect-information nature of the model and the convention--that is standard in the common agency literature--that principals have the power to make take-it-or-leave it offers. Even though the IMF engages in extensive information gathering and in frequent negotiations with officials during the programme design and implementation stages, in reality its information does not match that of country authorities. The authorities' informational advantage presumably leads to a more equitable distribution of the gains from the conditional assistance game than portrayed in the formal model.

International conflicts of interest

IMF conditionality is most commonly justified by appealing to international conflicts of interest. Such conflicts arise naturally when sovereign governments' policy objectives clash. Beggar-thy-neighbor policies, such as competitive exchange rate devaluations and resort to tariffs and other trade protection in the 1930s, were key factors behind the decision to create the IMF in 1944. International policy spillovers continue to be an important justification for the IMF's existence. This role for the IMF is reflected in our modeling approach, in which it takes the international externalities of recipient countries' policies into account when setting the terms of its conditionality. In the model, international externalities are captured in a stylised fashion, by inserting the debtor countries' aggregate distortion index into the creditors' social welfare function. This may be justified by appealing to the deadweight losses that countries suffer when their trading partners adopt inefficient trade policies, such as tariffs, quotas, or undervalued exchange rates. (10) The political trades that creditor countries are willing to engage in (through the IMF) reflect the extent to which debtor countries' policies affect their welfare: an increase in the absolute value of [W.sub.[omega].sup.*] makes the IMF's truthful assistance schedule steeper. In other words, the IMF will willingly provide more assistance for a given reduction in distortions to those debtor countries whose policies have greater impact on creditor-country welfare (are systemically more important). This finding is also consistent with Stone's (2002) conclusion that the length of punishment for non-compliance with conditionality is shorter in systemically important countries.

While conditional assistance is effective in addressing policy distortions originating in debtor countries, it is not clear how much leverage the IMF has in addressing policy failures in creditor countries. To address the international externalities caused by industrial country policies, including agricultural subsidies, trade barriers, or inappropriate exchange rate policies, requires endowing international institutions such as the IMF with instruments that go beyond that of conditional lending. Thinking about these issues is clearly very important but is beyond the scope of the present paper.

Domestic conflicts of interest

Conditionality will not be implemented unless key groups and stakeholders in recipient countries support or acquiesce to the policy programmes being supported by the IMF. In an ideal setting in which losers could be identified and compensated by those gaining from the enactment of welfare-enhancing reforms, all such reforms would be unanimously approved. However, as already discussed, domestic political and economic systems are imperfect and elected governments have considerable latitude to pursue personal and electoral objectives. These imperfections also give organised interest groups the opportunity to exert disproportionate influence on government decision-making. Interest groups are able to gain access to decision-makers and provide them with specialised information about the impact of different policy choices and the strength of support for a particular policy. As illustrated by the model, political contributions may also buy interest groups more favourable policies, such as trade protection, subsidies, or tax advantages.

The presence of domestic groups opposing reforms provides a rationale for IMF conditionality even in countries that are so small that their choices result in negligible cross-border externalities. Yet, because the IMF's mission is commonly defined as the amelioration of the international spillovers discussed in the previous section, the insertion of IMF conditionality in domestic conflicts of interest is controversial. Many observers would like to return to the simpler world in which the IMF only lends to countries whose policies pose a threat to the regional or global economy. But, as stated in Article I of its charter, the IMF's purpose is to give members the 'opportunity to correct maladjustments in their balance of payments without resorting to measures destructive of national or international prosperity.' Countries can use IMF assistance to achieve a more orderly adjustment of internal disequilibria, and reduce structural rigidities and ineffective institutions--such as those associated with central planning for example.

The IMF can assist countries in overcoming the pernicious influence of domestic vested interests by helping them to amplify the voice of governmental or non-governmental civil society organisations. As more interest groups organise in borrowing countries, including those representing the broad public interest, the proportion of the population represented in the political trades rises. In the limit, if all the citizens are represented in some domestic interest group, the political equilibrium is Pareto optimal for the country. Since the identity of the under-represented--including the poor, disenfranchised, or marginalised groups--is likely to vary from country to country, and within a given country at different times, considerable field work may be required to ensure that IMF processes help give such groups voice, including in poverty and social impact analysis. The participatory features of the Poverty Reduction Strategy Process are a step in the direction of giving groups voice and could be emulated in other IMF-supported programmes.

Inefficient policy instruments

The tension between the IMF and member governments extends to the choice of policy instruments. As the special interest politics approach shows, governments operating under the influence of interest groups prefer to use tariffs, export taxes, subsidies and other distortionary policy instruments because these tools better facilitate political trades. On the other hand, the IMF and other IFIs promoting broad national and international goals are keen to see distortionary policy instruments eliminated.

Key to understanding the revealed preference of interest groups for inefficient means of taxation and redistribution is a classical proposition from public finance. This proposition, which is due to Diamond and Mirrlees (1971), asserts that the government can achieve socially optimal redistribution by taxing final goods only, even in second-best situations in which lumpsum taxes and transfers are not available. Taxes or subsidies on intermediate inputs distort production decisions and are inefficient. When governments are influenced by special interests, however, the Diamond-Mirrlees result does not hold. Groups that overcome the free-rider problems and fixed costs involved in organising will expend resources to lobby for inefficient policy instruments. The more inefficient the tools of redistributive politics, the better is the bargaining position of organised interest groups vis-a-vis policymakers (Grossman and Helpman, 2001, p. 279).

Conditionality as a commitment technology

One way in which IMF conditionality helps governments eschew inefficient policy instruments is by providing an external 'commitment technology.' As is well known, governments that are unable to commit to sound policies in the future will be tempted to adopt inferior, time-inconsistent policies. One-time, surprise taxes on income from capital already in place, including unanticipated inflation, involve a zero excess burden ex post and are efficient from an optimal tax point of view. Anticipating such behaviour, the private sector will reduce its productive capital investment and hold less domestic money (dollarisation) than in the 'commitment' equilibrium. While fully representative governments face incentives to resort to capital levies, the incentive to engage in such tactics increases as the degree to which a government cares about its citizens declines.

By making the IMF's assistance contingent on the adoption and maintenance of non-distorting fiscal and monetary policies, the IMF helps member countries come closer to first-best policies. Government commitments to the IMF work synergistically with those to other international organisations (such as the World Bank and the World Trade Organization). The aggregate political influence of these external actors assists governments in their efforts to steer clear of inefficient taxation and redistribution. Decisions to reduce policy distortions will be consistent with political equilibrium only if the government's valuation of IFI membership to outweigh short-run political costs. In normal times, this will depend on the relative political influence of IFIs and domestic SIGS, which depends on the SIGs' gains (and the IFIs' losses) from inefficient policies, the government's time discount factor and remaining tenure in office, and the other factors illustrated by the model. In crisis times, the political cost-benefit calculus may change quite dramatically. When the country is in the midst of a dire external crisis, the IFIs' bargaining power increases and they can confront governments with take-it-or-leave-it proposals that proscribe inefficient policy instruments. On the other hand, if domestic conditions are not dire, then more accommodative strategies may be needed.

Veto players

In practice, government policymaking involves the interaction of multiple actors in the executive, legislative and judiciary branches of government. Some of these actors have the power to block reform measures agreed between the IMF and the executive. The theory of 'veto players' (Tsebelis, 2001) analyses policy formation as equilibrium interactions among veto players. The number, identities, and political influence of these veto players depends on each country's constitutional and political organisation. In presidential systems, the legislature (one or two chambers) and the chief executive are veto players. The congress has agenda-setting powers since it can make take-it-or-leave-it offers to the president. The president can accept these offers or veto legislative bills, in which case a qualified majority in congress can over-rule the veto. In parliamentary systems, the chief executive is selected by parliament. Veto players are parliamentary parties and their coalitions since their agreement is needed for the implementation of policy changes. In other systems, the military may be a de facto veto player. And, in any political system, the public can become an effective veto player through mass street protests.

Our application of the theory of special interest politics to IMF conditionality abstracts from within-government heterogeneity. For the sake of tractability, the model of the political economy of IMF conditionality presents a unitary view of government in borrowing countries. In reality, conflicts of interest between the executive and legislative branches are a key obstacle to the implementation of reform measures agreed between the IMF and member governments and backed financially by the IMF. Recent models (Drazen, 2002) analyse the consequences for IMF conditionality of the presence of veto players. To the extent that their policy preferences differ from those of the executive, veto players use their leverage to ensure that adopted policies are closer to their ideal positions. Like lobbying models, veto player models of conditionality involve interactions of multiple players. In Drazen's model, an agreement signed by the IMF and the executive must be approved or implemented with the cooperation of the legislature. Conditionality helps improve policies relative to unconditional assistance, much as in models based on the theory of special interest politics and common agency.

Conditionality and cooperation

Conditionality can be useful in promoting cooperative arrangements ('trust') in group politico-economic interactions in member countries subject to domestic divisions. According to game theory's folk theorem, cooperation can evolve spontaneously as the outcome of non-cooperative interaction among selfish agents in a repeated interaction. (11) In some countries, domestic divisions are so severe that such equilibria do not arise in the absence of external influence. This might be the case when the government is under the control of a single ethnic, religious or regional group that looks after its own narrow interests at the expense of the general welfare. In countries with divided polities, each group's incentives to cooperate in the one-shot game are therefore limited. The group in power is a 'bandit' in Olson's sense, pilfering society's resources while it can. If two such groups alternate in power in an infinitely repeated interaction, incentives for mutual cooperation arise that do not exist in the one-shot game (Dixit, 2001). Such self-enforcing inter-group cooperation is more likely the greater is each group's patience (i.e., the lower is each group's subjective rate of time preference) and the greater is each group's risk aversion. Cooperation is also more likely the greater is the likelihood of switch in power by either party. While some persistence in power is consistent with the emergence of spontaneous cooperation, such an outcome is made impossible if one or both groups have a highly persistent lock on power.

In countries with divided polities, IMF conditional assistance can be designed to strengthen the domestic warring groups' incentives to cooperate. As the IMF is involved in repeated interaction with member countries" governments, the conditions for meeting the folk theorem could be applicable to this interaction. These conditions are well known. The IMF's and member governments' time discount factors must be sufficiently high, so that the 'shadow of the future' is large enough to make cooperative strategies worthwhile. It is an open question what types of strategies are appropriate for the IMF to follow to induce cooperation. Since the IMF and borrowing countries engage in repeated interactions, it may be optimal for the IMF to use a combination of 'stick and carrot' incentives, as opposed to relying on trigger strategies alone in which aid is cutoff permanently following noncompliance. In countries with poor track records of policy implementation and weak ownership, it may be appropriate for IMF loans to be made available on the basis of result-based conditionality. Disbursing loans after a country has met certain verifiable conditions and not on the basis of promises to deliver policy changes in the future may discipline both the IMF and borrowing countries.

The IMF and special interests in practice

As emphasised by the public choice literature, IMF conditionality may be influenced by political pressures from shareholders that use the IMF to obtain advantages in their bilateral economic and financial relationships with debtor countries. Some critics of the IMF (Stiglitz, 2002), have put forward an extreme version of this 'capture theory" of the IMF. They claim that the institution does not represent the interests of developing countries, having been captured by the financial lobbies of creditor countries. While Stiglitz's theory is the polar opposite of the public-interest view of the IMF presented in this paper, it can be evaluated using the same tools of the theory of special interest politics. Altering the IMF's objective function in the way suggested by Stiglitz will reduce the effectiveness of conditionality, since the IMF ceases to be representative of the poor and disenfranchised in developing countries.

But while the IMF may not be fully representative, Stiglitz's theory flies in the face of the IMF's constitution, which focuses on the global public interest. In addition, as already indicated, changes in the IMF's operations in low-income countries in the late 1990 s have increased the voice and participation of poor in decision-making, including in the design and implementation of national poverty reduction strategies being supported by the IFIs. The IMF counters the influence of special interests regardless of their country of origin. In a number of occasions, the IMF has advocated the need for industrial countries to improve market access for developing countries. For example, in an open letter published in May 2002, the heads of the World Trade Organization, World Bank and IMF called on the industrial countries to follow through on their commitment from the Doha Round of Trade negotiations to help developing countries better engage in the global trading system. The IMF has also refocused its surveillance to provide more detailed and critical analysis of trade policies in industrial and large developing countries. The IMF has repeatedly called for wealthy countries to fulfil their commitment to provide Overseas Development Assistance equivalent to the United Nations target of 0.7 percent of GNP. Finally, the IMF also speaks up against governance and corruption problems in countries receiving IFI resources.

A comprehensive analysis of the IMF's objectives and incentives is, however, beyond the scope of the paper. The extent to which the IMF can resist political pressures depends on its institutional independence and the degree to which it can apply its rules-based approach to lending. Suffice it to say that while the IMF's choices cannot deviate too much from those of its member governments (Martin, 2000), the institution does enjoy a reasonable degree of discretion that helps insulate it from outside pressures. The IMF's operational independence derives from the powers delegated to it by its membership and also by informational advantages vis-a-vis individual members (the IMF typically has more troops on the ground in debtor countries than individual member countries).

CONCLUDING REMARKS

Left on their own, some countries' economic and political systems would generate highly inefficient and distorted economic policy choices. IMF loans and their associated conditionality assist these countries in their efforts to reduce these distortions. This paper illustrated the incentive effects of IMF conditionality with the help of a perfect-information model of the world economy. In this model, in which the IMF is a benevolent lobby that maximises world welfare, conditional loans help governments counter the political influence of interest groups that lobby for policies that are harmful to national and international welfare. Conditionality improves world resource allocation relative to either unconditional assistance or no assistance while economising on the amount of assistance--conditional assistance achieves a greater 'bang' relative to unconditional assistance. However, while IMF conditionality can help reconcile the narrowly focused objectives of national authorities with the IMF's broader internationalist objectives, its application must be tailored to the domestic political realities in borrowing countries. The need for conditionality was also affirmed in the IMF's conditionality review. The review validated the need for country-tailored programmes and a critical mass of ownership in order to ensure reasonably good programme implementation.

As already discussed, while the theoretical case for conditionality is strong, the empirical evidence regarding its application is more varied. Policy programmes agreed between the IMF and governments of member countries and backed by IMF financial assistance are often disrupted by economic and political shocks and by policy disagreements. In practice, such 'on-off' application of conditionality compromises its effectiveness. Understanding the reasons behind the relatively poor record of implementation of IMF-supported programmes and finding ways to make programs more robust to economic and political shocks is a priority of future research. The key to reconciling the empirical evidence with the rosy analytical predictions lies in relaxing some of the strong assumptions made in this paper. Chief among these are the incentive problems which altruistic lenders, including the IMF, face. Abstracting from issues of time inconsistency (the Samaritan's dilemma) helps ensure tractability but may paint an excessively favourable picture of the conditions in which the IMF is operating--perfect information, no credibility problems, etc. Working out the implications of relaxing these assumptions will also need to be a priority in future work.

(1) The views expressed in this paper are those of the authors and should not be attributed to the International Monetary Fund.

(2) While technically IMF financing is a currency exchange and not a formal loan, this distinction is not important in practice. The IMF is treated as a senior creditor and almost always gets repaid. Instances of protracted arrears to the IMF are rare and involve a small number of countries with intractable economic, political, and social problems. On the other hand, the problem of 'evergreening of IMF loans,' or defensive lending, is occasionally an important issue in decisions to renew IMF financing. The present paper focuses on the incentive issues facing debtor countries a and abstracts from incentive problems that the IMF may face.

(3) Mayer and Mourmouras (2002) develop a lobbying model of the IMF. Drazen (2002) presents a veto player model.

(4) Common agency was developed by Bernheim and Winston (1986) and applied to the positive analysis of trade policy and public finance by Grossman and Helpman (1994, 2001) and Dixit et al. (1997).

(5) Such conflicts would persist even if debtor governments were not under the influence of special interests but rather pursued legitimate national goals (such as redistribution) that created international policy externalities.

(6) On rent seeking and vested interests see Krueger (1974), Havrylyshyn and Odling-Smee (2000), Helhnan and Kaufmann (2001), and IMF (2001c).

(7) The Working Paper version of this paper reviews in greater detail the recent empirical literature on the implementation and effectiveness of IMF-supported programmes. See also Ul Haque and Khan (1998).

(8) The theory of special interest politics can be used to analyse the impact of conditional IMF financial assistance on economic policy choices in debtor countries under alternative assumptions about the IMF's objective function. See the section on Implication of special interest politics for the IMF.

(9) By contrast, the unconditional assistance equilibrium (point B in Figure 1) is not Pareto optimal. Starting from the point of tangency between the IMF's indifference curve and the government's reaction function, there exist unexploited mutual welfare gains for the IMF and the recipient government.

(10) Domestic distortions, such as the production subsidy considered in the political economy of IMF conditionality, may have international impacts as well.

(11) See Rasmusen (2001, pp. 112-114) for a statement, Axelrod (1984) provides a thorough discussion of the evolution of cooperative arrangements in environments in which self-interested players must interact without the presence of a central authority.

REFERENCES

Axelrod, R. 1984: The evolution of cooperation. Basic Books: New York.

Bernheim, BD and Winston, MD. 1986: Menu auctions, resource allocation, and economic influence. Quarterly Journal of Economics 101(1): 1-32 (February).

Bird, G. 2002: Conditionality and ownership: how to improve the implementation of IMF programmes Working paper, University of Surrey: UK.

Boughton, JM and Mourmouras, A. 2002: Is policy ownership an operational concept? IMF Working Paper 02/72.

Burnside, C and Dollar, D. 1998: Aid, the incentive regime, and poverty reduction. Working paper, Macroeconomics and Growth Group, World Bank.

Burnside, C and Dollar, D. 2000: Aid, policies, and growth. American Economic Review 90(4): 847-868.

Diamond, P and Mirrlees, J. 1971: Optimal taxation and public production: I. production efficiency, and II. tax rules. American Economic Review 61:8-27 and 261-278.

Dixit, A. 2001: Some lessons from transaction-cost politics for less-developed countries. Olson Memorial Lecture delivered at the University of Maryland (October). Available at: http://www/bsos.umd.edu/umccc/olson_memorial_lecture_series_schedule.htm.

Dixit, A. 2000: IMF programs as incentive mechanisms. Mimeo, International Monetary Fund (June).

Dixit, A, Grossman, GM and Helpman, E. 1997: Common agency and coordination: general theory and application to government policy making. Journal of Political Economy 105(4): 752-769 (August).

Dollar, D and Svensson, J. 2000: What explains the success or failure of structural adjustment programmes? Economic Journal 110 (October): 894-917.

Drazen, A. 2002: Conditionality and ownership in IMF lending: a political economy approach. IMF Staff Papers 49(Special Issue): 36-67.

Grossman, GM and Helpman, E. 1994: Protection for sale. American Economic Review 84(September): 833-850.

Grossman, GM and Helpman, E. 2001: Special interest politics. MIT Press: Cambridge, MA.

Havrylyshyn, O and Odling-Smee, J. 2000: Political economy of stalled reforms. Finance and Development 37(3). Available at http://www.imf.org/external/pubs/ft/fandd/2000/09/havrylys.htm.

Hellman, J. and Kaufmann, D. 2001: Confronting the challenge of state capture in transition economies. Finance and Development 38(3): (September). Available at bttp://www.imf.org/external/pubs/ft/fandd/2001/09/hellman.htm.

International Monetary Fund. 2001a: Conditionality in IMF-supported programs--policy issues (February 16). Available at http://www.imf.org/external/np/pdr/eond/2001/eng/policy/index.htm.

International Monetary Fund. 2001b: Structural conditionality in IMF supported programs (February 16). Available at http://www.imf.org/external/np/pdr/cond/2001/eng/struct/index.htm.

International Monetary Fund. 2001c: Strengthening country ownership of fund-supported programs (December 5). Available at http://www.imf.org/external/np/pdr/cond/2001/eng/strength/120501.htm.

International Monetary Fund. 2002: Guidelines on conditionality (September 25). Available at http://www.imf.org/External/np/pdr/cond/2002/eng/guid/092302.pdf.

Ivanova, A, Mourmouras, A, Mayer, W and Anayiotos, G. 2003: What determines the implementation of IMF-supported programs?. IMF Working paper 03/08 (January).

Khan, MS. and Sbarma, S. 2003: IMF conditionality and country ownership of adjustment programs. World Bank Research Observer 18:249-273 (Fall).

Krueger, A. 1974: The political economy of a rent-seeking society. American Economic Review 64(3): 291-303.

Martin, LL. 2000: Agency and delegation in IMF conditionality. Manuscript, Department of Government, Harvard University (October).

Mayer, W. and Mourmouras, A. 2002: Vested interests in a positive theory of IFI conditionality. IMF Working paper 02/73.

Mayer, W. and Mourmouras, A. 2004: The political economy of foreign assistance: grants versus loan rollovers. IMF Working paper 04/4.

Nsouli,, S, Atoian, R and Mourmouras, A. 2004: Institutions, program implementation, and macroeconomic performance. IMF Working paper (forthcoming).

Olson, M. 1965: The logic of collective action. Harvard University Press: Cambridge, MA.

Rasmusen, E. 2001: Games and information: an introduction to game theory. Blackwell Oxford.

Stiglitz, J. 2002: Globalization and its Discontents. Norton: New York.

Stone, RW. 2002: Lending credibility: the international monetary fund and the post communist transition. Princeton University Press: Princeton.

Tsebelis, G. 2001: Veto players: how political institutions work. Princeton University Press and Russell Sage Foundation: Princeton, NJ.

Ul Haque, N and Khan, MS. 1998: Do IMF-supported programs work, a survey of the cross-country empirical evidence. IMF Working paper 98/16 (December).

Willett, T. 2003: IMF conditionality and the new political economy of ownership. Working Paper, Claremont Graduate School.

WOLFGANG MAYER (1) & ALEX MOURMOURAS (2)

(1) University of Cincinnati, Cincinnati, OH 45221, USA. E-mail: amourmouras@imf.org;

(2) International Monetary Fund, Washington, DC 20431, USA.