A. A New Notion Begins to Take Shape (Late 1970s–Earlyl980s)

In the context of the limited capital mobility of the postwar period, the recipe the IMF applied for dealing with balance of payment difficulties was fairly straightforward. When a country faced current account pressures, the IMF would typically arrange a program featuring a combination of official financing and domestic policy adjustment, the latter including a currency depreciation in cases of “fundamental disequilibrium.” Within this mechanism, quotas—that is each country’s contribution to IMF resources—played the dual role of a constraint on the capacity of each member to borrow and of a ceiling on the resources each member would have to contribute to the common pool.

The ability of the IMF to play its intended role came under stress during the 1970s, as a consequence of worldwide inflation, which reduced the real value of existing quotas, and larger current account imbalances in the aftermath of the oil crises. Faced with the prospect of being unable to meet the members’ needs arising from the first oil shock, the IMF was forced to contemplate at least a partial transformation from a credit union into a financial intermediary. This essentially meant bringing borrowing into the panoply of funding means. The possibility of direct borrowing on financial markets was ruled out at the outset, because it would have fundamentally altered the intergovernmental nature of the institution. As a consequence, the IMF’s financing needs would have to be satisfied by borrowing from member countries. To be sure, this transformation had already begun in 1963, with the creation of the General Arrangements to Borrow (GAB). But the departure from previous practice remained modest until the 1970s. Moreover, in February 1977 IMF Managing Director Witteveen proposed a Supplementary Financing Facility that would link traditional lending based on quotas with funds obtained through borrowing from countries.

The transformation of the IMF into a financial intermediary, however, did not go unchallenged. First, it looked much more controversial in the late 1970s, after international financial markets had taken up the key role in “recycling” the current account surpluses of members of the Organization of Petroleum Exporting Countries. Second, and more important from our perspective, the notion that IMF lending could cause moral hazard and therefore that “taxpayers’ money” should not be used to bail out private investors was increasingly attracting political attention.

It is in connection to this latter problem that the idea that IMF resources, rather than substituting private capital, could be used to sustain financial markets’ confidence and catalyze private capital flows began to take shape. This is most evident in the testimony made in 1977 by Anthony Solomon, Undersecretary of the Treasury for Monetary Affairs, before the Committee on Banking, Finance and Urban Affairs of the U.S. House of Representatives. To defend the proposed Witteveen facility, Solomon made the case that the new IMF window would not be used to bail out the private banks, as the latter typically follow the IMF in their lending decisions:

“The very fact that [countries] are meeting the IMF’s performance criteria and thus continue to be eligible to draw from the IMF tends to represent a kind of Good Housekeeping seal of approval. Good performance under an IMF program tends to result in private capital inflows, private banks being willing to lend more to the country concerned” (U.S. House of Representatives, 1977, p. 72).⁵

The notion that private capital may tend to “follow” IMF lending had also begun to gain currency within the IMF itself. Margaret Garritscn de Vries, for many years the IMF historian, described the mechanism in the following way:

“A close tie between the willingness of commercial banks to lend to a country and the country’s having a stand-by arrangement with the IMF became almost standard after commercial banks encountered repayment problems with Zaïre in 1976 and 1977. … Commercial bank officials were beginning to be more familiar with IMF parlance and policy, and the test which they devised to judge the creditworthiness of a country was whether it could meet the conditions needed for a stand-by arrangement with the IMF in the upper credit tranches. This test induced private commercial banks to have prospective borrowing countries obtain the stamp of approval of the IMF by obtaining a stand-by arrangement in the upper credit tranches.” (De Vries, 1985, pp. 493-94).⁶

De Vries, however, might have overstated the importance attached to the mechanism in the “Fund parlance and policy.” Indeed, references to what we now call catalytic effects in the writings of IMF staff remained for a while episodical.⁷ Moreover, it seems clear from reading such texts that at that time that the seal-of-approval mechanism was not seen as having major implications for the way the IMF operated. Thus, Manuel Guitián (1982), who would later become head of the Monetary and Exchange Affairs Department of the IMF, in a paragraph remarkably entitled “The Fund as Catalyst,” argues’

“a critical side effect of the mix of adjustment and financing that is typically built into the programs supported by Fund resources has been to help members attract flows of capital from sources other than the Fund” (p. 91).

This passage solicits two remarks. First, contrary to Dooley’s (1994) suggestion that the notion of catalytic lending is to be attributed to the Baker Plan of 1985, one can see that not only the notion, but also the label, were known to the IMF staff several years earlier. Second, Guitián does not argue that the effect lies at the core of the IMF role. Rather, he is quite explicit that this is but a “side effect” (although a “critical” one) of IMF programs. The reason probably is that Guitián was still contemplating a world in which balance of payment crises were predominantly being determined by the behavior of the current account.

B. The Debt Crisis (1982–93)

Increased financial liberalization and capital mobility during the 1980s were making the latter assumption increasingly questionable. The capital account was rapidly becoming as important a source of balance of payment pressure as the current account. But capital account pressures are far more complicated to deal with. The problem arose for the first time in August 1982, when Mexico started having trouble in servicing its external debt, in the aftermath of a generalized increase in international interest rates. The problem soon spread to at least eight other countries in Europe, Africa, and Latin America. The IMF was called in to help solve the problem, but it soon became apparent that

“the old approaches would not work, because new financing from the Fund would quickly be siphoned off as the indebted countries would have to repay other creditors” (Boughton, 2000a, p. 171).

In the old approach the IMF could work out the needed amount of “policy adjustment” and “financing” on the assumption that the overall financing gap the country faced was given and easy to measure. Now, the possibility of a drawdown in the foreign sector’s holdings of domestic debt—as well as of an increase in residents’ assets abroad—implied that the size of the overall financing gap depended on the credibility of the promised policy adjustment. Unless one found a way to convince investors not to liquidate their holdings, financing a country’s balance of payment crisis might become like trying to fill up a bottomless pit. Of course, the problem could have been obviated by providing a full bailout of, at least, foreign creditors, but this option was never seriously contemplated in policy circles, on the ground that it would conflict with both the credit union structure and the monetary nature of the IMF (Boughton, 2001), not to mention obvious political difficulties in raising sufficient funds.

With the “Bagehotian” solution ruled out, the official sector’s strategy took at first the form of concerted lending.⁸ The strategy was inaugurated in November 1982, when the then Managing Director of the IMF, Jacques de Larosière, informed a gathering of bankers in New York that the IMF would not approve a request by Mexico for a US$4 billion loan until private banks provided written assurances that they would as a group increase their Mexican exposure by US$5 billion.

With respect to domestic experiments with orchestrated rescues, all of which are ultimately based on the power of the central bank to hold private banks “hostage” through moral suasion, this international version’s main novelty lay in its formalized character. Under the “financing assurances policy,” as it came to be known, the IMF would suspend access to its own money if creditor banks refused to reschedule the country’s debt (Boughton, 2001, p. 477). Such a high degree of formalization was required to make up for the IMF’s lack of direct powers over commercial banks. It had a serious drawback, though, in that it gave the banks a virtual veto over the approval and financing of adjustment programs. More bluntly, under the new practice it was unclear who was being made hostage to whom.⁹

Yet, the concerted lending strategy was applied with some success to a dozen cases between 1982 and 1986 (Boughton, 2000a, p. 172). By 1985, however, it was no longer working, essentially because of commercial banks’ growing unwillingness to be “concerted.” Both financial and macroeconomic developments concurred to make bankers stiffer. On the financial side, the time gained through the concerted packages had allowed banks to build up considerable provisions against sovereign credit risk. Feeling less exposed to the threat of outright default, banks gradually assumed a tougher negotiating position. On the macroeconomic side, the policy adjustment of the countries benefiting from concerted lending had been insufficient, particularly in the structural area, resulting in protracted stagnation. This made concerted lending increasingly unpopular, as banks feared they would end up throwing good money after bad. Indeed, net bank lending to developing countries declined dramatically in 1985, and turned negative the next year.

All this provided the setting for U.S. Treasury Secretary’s famous October 1985 speech laying out a “Program for Sustained Growth”—later to be known as the Baker Plan. The plan’s underlying idea was that in order to avert massive cuts in lending to developing countries growth had to be revived. To this end, what was needed was a combination of structural reform in recipient countries and the mobilization of financial resources on a large scale. How to achieve the first was fairly straightforward: one only needed to extend the scope of IMF conditionality to structural aspects, something the IMF had already begun doing. The latter task was more complicated. It is here that the notion of “catalysis” came in handy.¹⁰ The Baker plan posited that the provision of sufficient amounts of official financing in the context of structural reform aiming at economic liberalization would catalyze private sector lending to the extent needed to fill whatever financing gap was created in the process of spurring economic growth. It was estimated that the money needed to finance the plan during 1986-88 was US$17-18 billion for official money and US$20 billion for private lending (Boughton, 2001, p. 428).

This first experiment with using COF for crisis management purposes yielded mixed results. Actual official lending was very close to the planned figure, but the sought catalytic effect on private lending was hardly visible.¹¹

Leaving figures aside, the experiment led to widespread frustration. As Paul Volcker puts it, barely a year after the launching of the plan “any sense of enthusiasm was very much gone” (Volcker and Gyohten, 1992, p. 215). Even the most ardent advocates of the plan had to recognize that “by 1988 there was no question that the banks’ willingness to provide new financing was essentially finished” (Boughton, 2001, p. 428). The symbolic coup de grâce to the Baker Plan came when, in 1988, Citicorp provisioned US$3 billion against Citibank’s exposure toward developing countries. While the price of Citicorp stock skyrocketed, the prices of developing countries’ loans on secondary markets plummeted. Not surprisingly, new lending to developing countries nearly dried up over the next two years.

In this environment, proposals to deal with the developing countries’ problems by resorting to some form of debt relief began to gain currency, first in Europe and Japan, then in the United States. The IMF itself began to show dissatisfaction with both concerted lending and catalytic financing by adopting a new operational framework called the “menu approach,” which entailed dealing with individual crises by presenting creditors with a range of options all of which implied replacing old loans with new instruments that more accurately reflected secondary market values.

There was a critical problem, though, before the new strategy could be translated into concrete actions. Under the “financing assurances” policy, the IMF was for all practical purposes incapacitated to force debt relief onto recalcitrant creditors. This policy had to be modified. Thus, the plan concocted by U.S. Treasury Secretary Brady in the late 1980s, envisaging a comprehensive scheme for debt reduction, also encouraged the IMF to reconsider its financing assurances policy to reduce the risk that some creditors might hold out for a protracted period, thereby undermining the success of the debt reduction strategy in individual cases. The IMF Board endorsed the Brady Plan in May 1989. Under the new operational framework, an accumulation of arrears to banks could be tolerated even pending agreement between the country and its creditors, provided “negotiations continue and the country’s financing situation does not allow (arrears) to be avoided.” The new policy meant adapting COF to the needs of crisis resolution, as the IMF, through its lending-into-arrears, would now pressure creditors to reach agreement with the country’s authorities on the appropriate amount of debt reduction.

The Brady Plan can be counted among the successes of international cooperation. By 1993, eight Latin American countries, through a variety of operations, achieved a reduction of their external obligations by US$42 billion, out of an initial stock of obligations of about US$104 billion (Boughton, 2001, p. 552). Over the same period, rescheduling agreements with bilateral official creditors achieved through the Paris Club totaled about US$35 billion The LDC debt crisis was finally over

C. The Emerging Market Crisis (1994–2002)

After the launch of the Brady Plan, discussions on international crisis lending stopped for several years—inter alia, because of the lack of major new balance of payment crises. However, the idea that IMF lending could catalyze private financial inflows had not lost its appeal, particularly within the staff and management of the IMF (see, for example, IMF, 1992, and Guitián, 1992, p. 20).

As a matter of fact, the catalytic effect lay at the heart of “precautionary” programs aimed specifically at promoting financial markets’ confidence through the IMF’s seal of approval, rather than providing member countries with official finance. “Precautionary” programs-programs in which the country authorities state their intention not to draw—became quite popular in the early 1990s, especially in those countries with a recent history of government intervention in the economy, and therefore with a structural credibility deficit (Dhonte, 1997). By the outbreak of the Asian crises, such programs accounted for one-third of all Stand-By and Extended Fund Facility (EFF) arrangements. Moreover, the ratio between actual and potential borrowing in all outstanding IMF arrangements—excluding those off-track—declined significantly between the early 1980s and the middle of the 1990s (Cottarelli and Giannini, 1997).

Interestingly, however, when Mexico was hit by a devastating capital account crisis, in late 1994, the catalytic effect was not invoked. Indeed, as Kenen (2001, p. 22) remarks, the package that was mounted to cope with the crisis (totaling about US$50 billion, mostly from the IMF) had all the features of a Bagehot-style LOLR operation:

“The official financing for Mexico was unique in more than its unprecedented amount. Although other crisis-stricken countries obtained large-scale official financing in subsequent years, the amounts involved were not intended—nor were they sufficient—to pay off those countries’ short-term foreign debts. The size of the Mexican package, by contrast, was meant to extinguish the whole stock of tesobonos, to help Mexico cope with large dollars withdrawals from its banking system, and to help it rebuild its reserves.”

Of course, no one either at the IMF or the U. S. Treasury really thought that the package should be used to pay back the Tesobonos. Only, the package was big enough to ensure holders of Tesobonos that the Mexican authorities would have sufficient money to honor their short-term debt, even if that implied redeeming it in full.

But why was COF not tried out for Mexico?¹² Probably, an important chunk of the answer is political in nature (one could maliciously ask why a Bagehot-style operation of such a size proved so easy to arrange, after all?). But this is hardly the whole story. One should add to it that until then no one had yet contemplated the idea of using COF as a crisis management tool while financial markets stayed open. As we have seen, the Baker Plan had been concocted in quite different circumstances, since all the crisis countries had already suspended external debt service long time before.

But the Mexican full-rescue package raised a number of problems that would later generate the attempt to use catalytic lending as a tool for crisis management also for major emerging markets. First, the sheer size of the package was hard to replicate, should more than one country be hit by a crisis. Second, it generated moral hazard concerns. Third, and perhaps more importantly, the idea that taxpayers’ money should be used to bail out private banks was hard to sell. Thus, it is not surprising that immediately after the announcement of the Mexican program references to COF as a market-friendly crisis management tool became more frequent.

For example, the Interim Committee of the IMF already in April 1995 refers explicitly to the “catalytic role” that the IMF could play in assisting members in coping with sudden market disturbances. Later that year Michel Camdessus, makes it plain that the Mexican package is to be regarded as an exception:

“We have agreed on procedures through which the Fund can respond rapidly—as it did in Mexico—to help put appropriate adjustment measures in place and avoid spill-over effects. We have made clear, however, that the use of such procedures will be limited to truly exceptional circumstances. The extent of our support will depend, as always, on the strength of the country’s own adjustment effort, and we will take appropriate steps to ensure that its support remains catalytic in nature” (Camdessus, 1995).

More extreme measures, reminiscent of the way the LDC debt crisis had been handled, were also contemplated. Thus, at the Halifax summit, in June 1995, the G7 recommended an in-depth reflection on crisis management tools. The task was taken up by the G10, which set up a working group whose report (the so-called “Rey Report,” from the name of his chairman) was published in May 1996. The report did accept the importance of catalytic lending.¹³ But it also warned that, given the difficulty of arranging a concerted package when a large share of “lending” takes place through bonds held by a myriad of investors, a temporary suspension of debt payments (a standstill) might sometimes prove desirable, as a way to gain time and better assess the sources of market turbulence. The Rey Report, however, met with fierce opposition from the private sector. A counter-report published by the Institute of International Finance in September 1996 called its approach “misguided,” on the twofold ground that it would face enormous implementation problems and, if implemented, it would fuel moral hazard on the debtor side. As a result, its recommendations never won enough consensus to be adopted.¹⁴

With the concerted approach increasingly regarded as difficult to implement, and the standstill approach facing strong opposition, by default the catalytic approach saw its appeal as a crisis management tool unquestionably rise.¹⁵ The Asian crises provided a first testing ground for the new technique.

Several excellent descriptions of the Asian crises are already available (see, for example, IMF, 2001, and Kenen, 2001). Here, it will suffice to note that COF—with IMF resources provided in large amounts yet falling short of Bagehot-style LOLR financing—featured in all the main Asian crises, notably in Thailand, Indonesia, and Korea. Indeed, as noted by IMF staff “most of the capital account crisis programs [of the late 1990s] were predicated on just such a catalytic effect of official financing.” (Ghosh and others, 2002, p. 8). As it happens, the catalytic effect failed to materialize, leading to sharper domestic adjustment, further injection of IMF resources and, in some cases (such as Korea) the recourse to stronger forms of concerted lending.

Russia’s crisis, in the summer of 1998, although different in both causes and ending, confirmed that the effectiveness of COF in the new environment could not be taken for granted. There, too, the existence of an IMF-supported program failed to provide sufficient confidence to markets. Yet, for want of better alternatives the official community continued to bet on the idea. Thus, the report on international financial crises published by the so-called Group of Twenty-Two (G22, 1998), reiterates that:

“In most cases when a country experiences payment difficulties, the combination of adjustment and financing of a typical IMF programme can be expected to restore market confidence and catalyze private capital flows.”

While the new report was being published, a new crisis broke out, this time in Brazil. The financial package the official community mounted this time totaled about US$42 billion, almost half of which from the IMF. The package was heavily front-loaded, as about 90 percent of the sum was to be made available within 13 months. Both the size of the package and its time profile were clearly unprecedented. Yet, the package remained, according to our definition, catalytic in nature, since Brazil’s short-term debt exceeded at the time US$100 billion. After the announcement of the program, markets stabilized for a while, but then, in late December 1998, new pressures emerged in the foreign exchange market. This time, financial conditions improved only after the floating of the real, and the announcement of a roll-over agreement between the Brazilian government and international banks. Such deals, contrary to Korea’s, had not been actively promoted through moral suasion at the international level, but it proved successful. Brazil’s external problems subsided. Eventually, only a fraction of the US$42 billion package was actually drawn. The whole episode was widely read as a success story for COF, although some commentators invited caution in drawing lessons from it.¹⁶

After Brazil, official thinking on crisis management proceeded along two parallel paths. A first path led to giving center-stage to the notion of catalytic lending, which was in fact enshrined in repeated public statements. The other led to the notion of “private sector involvement” (or PSI, a general term encompassing a range of options, from concerted lending to mandatory solutions), which clearly underscored the belief that catalytic lending may not be sufficient in a number of cases. Indeed, some debt restructuring operations were actively promoted, if somewhat implicitly, in a number of small countries with limited financial ramifications.

This two-pronged approach culminated in the “Prague Framework” issued at the 2000 IMF Annual Meeting, which hinged around three propositions. The first simply reiterated the belief that, in some cases, catalytic lending would be sufficient to overcome a crisis. The second pointed to the usefulness of voluntary PSI in many circumstances. The third announced that “in exceptional cases, countries might impose capital or exchange controls as part of payments suspensions or standstills, in conjunction with IMF support for their policies and programmes, to provide time for an orderly debt restructuring.” In such a context “IMF lending into arrears might be appropriate.”

In practice, however, experimentation continued only along the first two propositions of the Prague Framework. Thus, COF was again invoked, this time coupled with PSI, in two prominent cases, Turkey and Argentina. Both countries had requested an IMF program in December of 1999, for precautionary purposes in the case of Argentina, to strengthen the credibility of disinflationary policies, in the case of Turkey. Both programs initially worked, but then for a number of reasons, including policy slippages, financial conditions deteriorated. In line with the Prague Framework, in the course of 2000 the programs were increased in size (in IMF technical jargon, they were “augmented”) and expanded to include a PSI component. But this provided only little respite. Financial conditions stabilized in Turkey only after a further “augmentation,” at the end of 2001, brought financial support up to Bagehotian levels, namely about 1.500 of Turkey’s quota in the IMF. In Argentina, a third augmentation, in August 2001, proved unable to restore calmer financial conditions. The following December, the authorities then declared a moratorium on foreign debt to private creditors.

In light of these experiences, it is not surprising that in recent months the emphasis on COF as a crisis management tool has been declining. As Eichengreen (2001) puts it:

“the main lesson of the Turkish and Argentine experiences is that the markets are likely to disappoint official hopes. In the climate of uncertainty that invariably surrounds a crisis, waiting has option value. Investors have an incentive to wait and see whether the commitment to reform is sustained instead of being first to provide new money.”

In fact, increasing attention is now being given to the third leg of the Prague Framework. The speech given in November 2001 by the First Deputy Managing Director of the FMF, Anne Krueger, in which she proposed a formalized framework for restructuring a country’s external debt in case this was deemed unsustainable, set the ball rolling in the new direction. In subsequent discussions, the original proposal was somewhat refined and qualified. But its gist found its way into the official position at the IMF Spring Meetings, in April 2002. There, the word “catalytic” disappeared from both the IMFC and the G7 communiqués. The latter contained in its stead an Action Plan “to increase predictability and reduce uncertainty about official policy actions in the emerging markets.” The plan outlined “a market-oriented approach to the sovereign debt restructuring process in which new contingency clauses would be incorporated into debt contracts.” It also supported “the IMF on proposed approaches to sovereign debt restructuring that may require new international treaties, changes in national legislation, or amendments of the Articles of Agreement of the IMF.” At the subsequent Annual Meetings, in September 2002, the IMFC invited the IMP to make “a concrete proposal” along these lines by the Spring of 2003, Once the technical work is completed, pressure for change will have reached a critical stage.

A. Case Studies

Few papers have focused on case studies. An early example of the case-study approach is Killick (1995), who discusses the experience with IMF-supported programs in 17 countries, concluding that there is no evidence of catalytic effects on private capital inflows.¹⁷

A comprehensive discussion of country experiences is also presented in Bird, Mori, and Rowlands (2000).¹⁸ This paper looks at the experience of 17 countries during the 1980s and 1990s and draws four conclusions: (i) involvement by the multilateral will not guarantee an inflow of capital from other sources; (ii) what really matters is the perceived commitment by a government to a policy agenda that is seen as sound and internally consistent: (iii) catalysis with the multilateral is likely to be stronger and more positive in the case of bilateral aid flows, and (iv) the nature and extent of catalysis may differ between the IMF and the World Bank, with the IMF having a stronger but negative effect.

A significant contribution to the case-study approach has recently been given by the IMF staff in their study of IMF-supported programs in capital account crises (Ghosh and others, 2002). This paper looks at the experience of eight programs for large emerging markets during the 1990s, and points mostly at the absence of strong catalytic effects:

“It is striking that, in every [emphasis in the original] instance in our sample, the outcome [for private capital inflows] was worse than projected. Even in cases where the magnitude of the error was small, such as Argentina (1995), the program does not appear to have had a strong catalytic effect, at least in the very short run.” (p. 11).

The resulting financing gap had to be filled with larger current account adjustment than originally envisaged. The amounts were not negligible, as Figure 1 testifies. The discrepancy reached a peak in the cases of Korea and Thailand, where actual current account adjustment was 13-15 points larger than planned, but remained sizable also afterwards, in Turkey and Argentina The only exception to this pattern is Brazil in 1999, where the outturn was about 1 percentage point below the planned figure.¹⁹

Current Account Adjustment in Selected Crisis Countries

(As a percentage of GDP)

Citation: IMF Working Papers 2002, 193; 10.5089/9781451859829.001.A001

Sources: IMF, World Economic Outlook database; Ghosh and others (2002).* Implied adjustment equal to planned current account minus actual current account in the year before the program.** Implied adjustment equal to actual current account in the year of the program minus actual current account in the year before.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Current Account Adjustment in Selected Crisis Countries

(As a percentage of GDP)

Citation: IMF Working Papers 2002, 193; 10.5089/9781451859829.001.A001

Sources: IMF, World Economic Outlook database; Ghosh and others (2002).* Implied adjustment equal to planned current account minus actual current account in the year before the program.** Implied adjustment equal to actual current account in the year of the program minus actual current account in the year before.

• Download Figure
• Download figure as PowerPoint slide

Current Account Adjustment in Selected Crisis Countries

(As a percentage of GDP)

Citation: IMF Working Papers 2002, 193; 10.5089/9781451859829.001.A001

Sources: IMF, World Economic Outlook database; Ghosh and others (2002).* Implied adjustment equal to planned current account minus actual current account in the year before the program.** Implied adjustment equal to actual current account in the year of the program minus actual current account in the year before.

• Download Figure
• Download figure as PowerPoint slide

B. Effects on Capital Inflows

Several papers have focused on testing statistically how the existence of an IMF-supported program, or the size of IMF financing, have affected private capital inflows. Various methodologies have been used, with the predominant result being that catalytic effects, if any, are small. Three approaches have been followed (Table 1).

Table 1.

Private Capital Inflows and IMF Lending

Table 1.

Private Capital Inflows and IMF Lending

	Methodology	Question	Sample	Main Findings
I. Papers estimating the demand for IMF credit
Bird and Orme(1981)	Regression analysis.	What determines the drawings by program countries?	31 developing countries that made drawings in 1976 and (an unspecified number of countries with drawings in) 1977 (annual data).	In 1976 significant positive correlation between drawings from the IMF and borrowing from euromarkets.
Cornelius (1987)	Regression analysis.	What determines the IMF drawings?	Eleven sub-Saharan countries, during 1975–77 and 1981–83.	Negative and nonsignificant correlation between drawings and borrowing from foreign and international capital markets.
Joyce (1992)	Regression analysis.	What determines the decision of a country to enter an IMF- supported program?	45 countries, annual data for 1980–94.	The commitment of new funds from private creditors relative to imports has a negative impact on requests for IMF assistance, but it is not significant.
Bird (1994); results also published in Bird (1995)	Regression analysis.	What determines the drawings in program countries?	1980–85 for 27 countries.	Positive, but weak, correlation between drawings and net disbursement of debt creating private finance.
II. Papers based on before-after tests
U.S. Treasury	Inspection of changes in capital inflows.	Does bank lending to program countries increase or decrease with the inception of a program?	36 program countries during 1970–75.	In most cases bank lending increased.
Reichmann and Stillson (1978)	Mann-Whitney test for difference in frequency distribution.	Is the increase in net foreign assets observed statistically significant?	85 program during 1963–72 (quarterly data).	Improvement observed in two-thirds of cases, but statistically significant only in less than one-fourth of cases.
McCauley(1986)	Before-after analysis.	Docs a program affect bank lending to a country in the year following the inception of a program?	99 programs covering 56 countries during 1976–81.	Generally supportive of the catalytic approach, with only one- third of countries showing a decline in private lending. However, the frequency of these cases increases over time (to half by 1981),
Killick. Malik, and Manuel (1991), same results reported also in Killick (1995)	Before-after analysis of differences from base (pre-program value); time horizon: four-year period.	Is the size of private long-term capital inflows different from the pre-program period over a four year horizon?	16 developing countries with IMF programs commenced in 1979–85.	Small effects and negative on net capital accounts, with net repayment of foreign loans and no increase in disbursements.
Schadler and others (1995)	Inspection of capital inflows.	What is the effect of a program on capital inflows?	45 IMF loans from mid-1988 to mid–1991.	Only in one-third of cases there were large increases in capital inflows.
Conway (1998)	Regression analysis.	Is the probability of exiting a crisis (defined as a deviation of the reserve to import ratio from a certain threshold) affected by the existence of an IMF program?	84 countries, quarterly data for 1973–92.	Countries participating in IMF programs are less likely to enter crisis, but this benefit is lost for countries for extended periods of IMF programs.
III. Papers estimating capital inflows models
Hajivassiliou (1987)	Regression analysis.	What determines the supply of new loans?	79 countries during 1970–82, annual data.	There is a negative (although barely significant) correlation between the supply of new loans and the lagged presence of IMF support or request for a rescheduling.
Rodrik(1996)	Regression analysis.	What determines net private capital inflows?	87 developing countries, averages of six-years periods (1970–93).	Effect of IMF lending not significantly different from zero, negative in some cases.
Adji et al. (1997)	Regression analysis.	What determines foreign direct investment?	Annual data for 23 developing countries during 1970–81.	No significant effect of the amount of IMF-lending on foreign direct investment.
Bird and Rowlands (1997)	Regression analysis.	What determines new lending commitments (net of repayments)?	90 developing countries, annual data for 1974–89.	The catalytic effect on private flows is either negative (and significant) or nonsignificant; some positive and significant effect on official inflows, not private.
Bird and Rowlands (2000)	Regression analysis; and questionnaires.	What affect private (portfolio FDI, and oilier debt) and bilateral capital inflows?	96 developing countries during 1980–95, annual data; and 15 interviews, with financial market managers.	No evidence of a positive catalytic effect and some evidence of a negative one for the private inflows: some (weak) effect on bilateral inflows. No evidence that the catalytic effect has strengthened over time.
Edwards (2000)	Regression analysis.	Do private inflows (commitments) and FDI reach positively to IMF programs?	106 countries, annual data for 1979–95.	No catalytic effect from the existence of an IMF program, even in countries with a good record of compliance with past programs; negative effects for countries with IMF programs that experienced past compliance problems.
Marchesi (2001)	Regression analysis.	Does an IMF program facilitate debt rescheduling by commercial creditors?	93 developing countries, annual data during 1983–95.	The presence of an IMF program increases the probability that a country benefits from debt restructuring.

View Table

Table 1.

Private Capital Inflows and IMF Lending

	Methodology	Question	Sample	Main Findings
I. Papers estimating the demand for IMF credit
Bird and Orme(1981)	Regression analysis.	What determines the drawings by program countries?	31 developing countries that made drawings in 1976 and (an unspecified number of countries with drawings in) 1977 (annual data).	In 1976 significant positive correlation between drawings from the IMF and borrowing from euromarkets.
Cornelius (1987)	Regression analysis.	What determines the IMF drawings?	Eleven sub-Saharan countries, during 1975–77 and 1981–83.	Negative and nonsignificant correlation between drawings and borrowing from foreign and international capital markets.
Joyce (1992)	Regression analysis.	What determines the decision of a country to enter an IMF- supported program?	45 countries, annual data for 1980–94.	The commitment of new funds from private creditors relative to imports has a negative impact on requests for IMF assistance, but it is not significant.
Bird (1994); results also published in Bird (1995)	Regression analysis.	What determines the drawings in program countries?	1980–85 for 27 countries.	Positive, but weak, correlation between drawings and net disbursement of debt creating private finance.
II. Papers based on before-after tests
U.S. Treasury	Inspection of changes in capital inflows.	Does bank lending to program countries increase or decrease with the inception of a program?	36 program countries during 1970–75.	In most cases bank lending increased.
Reichmann and Stillson (1978)	Mann-Whitney test for difference in frequency distribution.	Is the increase in net foreign assets observed statistically significant?	85 program during 1963–72 (quarterly data).	Improvement observed in two-thirds of cases, but statistically significant only in less than one-fourth of cases.
McCauley(1986)	Before-after analysis.	Docs a program affect bank lending to a country in the year following the inception of a program?	99 programs covering 56 countries during 1976–81.	Generally supportive of the catalytic approach, with only one- third of countries showing a decline in private lending. However, the frequency of these cases increases over time (to half by 1981),
Killick. Malik, and Manuel (1991), same results reported also in Killick (1995)	Before-after analysis of differences from base (pre-program value); time horizon: four-year period.	Is the size of private long-term capital inflows different from the pre-program period over a four year horizon?	16 developing countries with IMF programs commenced in 1979–85.	Small effects and negative on net capital accounts, with net repayment of foreign loans and no increase in disbursements.
Schadler and others (1995)	Inspection of capital inflows.	What is the effect of a program on capital inflows?	45 IMF loans from mid-1988 to mid–1991.	Only in one-third of cases there were large increases in capital inflows.
Conway (1998)	Regression analysis.	Is the probability of exiting a crisis (defined as a deviation of the reserve to import ratio from a certain threshold) affected by the existence of an IMF program?	84 countries, quarterly data for 1973–92.	Countries participating in IMF programs are less likely to enter crisis, but this benefit is lost for countries for extended periods of IMF programs.
III. Papers estimating capital inflows models
Hajivassiliou (1987)	Regression analysis.	What determines the supply of new loans?	79 countries during 1970–82, annual data.	There is a negative (although barely significant) correlation between the supply of new loans and the lagged presence of IMF support or request for a rescheduling.
Rodrik(1996)	Regression analysis.	What determines net private capital inflows?	87 developing countries, averages of six-years periods (1970–93).	Effect of IMF lending not significantly different from zero, negative in some cases.
Adji et al. (1997)	Regression analysis.	What determines foreign direct investment?	Annual data for 23 developing countries during 1970–81.	No significant effect of the amount of IMF-lending on foreign direct investment.
Bird and Rowlands (1997)	Regression analysis.	What determines new lending commitments (net of repayments)?	90 developing countries, annual data for 1974–89.	The catalytic effect on private flows is either negative (and significant) or nonsignificant; some positive and significant effect on official inflows, not private.
Bird and Rowlands (2000)	Regression analysis; and questionnaires.	What affect private (portfolio FDI, and oilier debt) and bilateral capital inflows?	96 developing countries during 1980–95, annual data; and 15 interviews, with financial market managers.	No evidence of a positive catalytic effect and some evidence of a negative one for the private inflows: some (weak) effect on bilateral inflows. No evidence that the catalytic effect has strengthened over time.
Edwards (2000)	Regression analysis.	Do private inflows (commitments) and FDI reach positively to IMF programs?	106 countries, annual data for 1979–95.	No catalytic effect from the existence of an IMF program, even in countries with a good record of compliance with past programs; negative effects for countries with IMF programs that experienced past compliance problems.
Marchesi (2001)	Regression analysis.	Does an IMF program facilitate debt rescheduling by commercial creditors?	93 developing countries, annual data during 1983–95.	The presence of an IMF program increases the probability that a country benefits from debt restructuring.

Analysis of the demand for IMF credit. Several papers have explored the issue of why countries need IMF credit. These papers do not typically focus on the existence of catalytic effects, but are often quoted in this context because the right-hand side of the econometric equations describing the demand for IMF-credit usually include the amount of private capital inflows. The sign of the coefficient on this variable is seen as having a bearing on the catalytic issue: a positive sign would indicate complementarity between IMF credit and private credit (as one would expect when catalytic effects are in place); a negative sign would indicate substitution. The first of these papers (Bird and Orme, 1981) does find a positive and significant correlation for a cross section of 1976 programs, but no correlation for the 1977 cross section. A positive correlation, albeit a weak one, is found also in Bird (1994) (published also in Bird, 1995). Other papers (Cornelius, 1987; Joyce, 1992) find negative (and typically nonsignificant) coefficients.²⁰

Before-after tests. This approach focuses on whether the inception of an IMF-supported program involves a statistically significant increase of capital inflows (or of variables related to them, such as foreign exchange reserves) with respect to the pre-program period. The first of these studies is the short paper (see Section III) prepared by the U.S. Treasury in 1977 as background for the discussion on the Witteveen Facility (U.S. Treasury, 1977). This analysis of 36 Stand-By Arragements (SBAs) during 1970-75 finds that in most cases SBAs were followed by an increase in bank lending. Richmann and Stillson (1978), although not directly aiming at measuring the existence of catalytic effects, also find that net foreign assets typically increase as a result of IMF-supported programs, but note that in only one-fourth of their sample this increase is statistically significant. Some evidence of catalytic effects in also found by McCauley (1986): in only one-third of the programs considered in his sample did bank lending decline in the year following the inception of a program. The relative frequency of outflows, however, increases to 50 percent by the end of his sample (1981). A more negative assessment of the existence of the catalytic effects can be found in Killick, Malik, and Manuel (1991).²¹ They find that, on a two-year time horizon, the effect on private capital inflow is negative or small: the improvement in the overall balance of payment position, which they also observe, is thus generally due to the improvement of the current account.

They also note that not only net inflows are not much affected by programs, but that gross flows are not affected either (there is no increase in gross disbursements). Another paper finding weak catalytic effects was prepared by IMF staff (Schadler and others, 1995). This paper, covering 45 IMF loans from mid-1988 to mid-1991, finds that only in one-third of the cases there were large increases in capital inflows. A more sophisticated variant of this approach is found in Conway (1998), who explores whether the probability of a foreign exchange crisis is reduced by the existence of an IMF-supported program. However, as the “crisis” is defined as foreign exchange reserves falling below a certain threshold, and as the behavior of foreign exchange reserves is modeled through a simple time series model, this approach essentially focuses again on whether on average the reserve position increases after the inception of a program. The results show that reserves do increase when a program is in place, but that this is no longer true for countries under a program for an extended period.

Capital inflows models. The before-after approach does not address the “counterfactual” problem. A solution to this problem is to model directly the behavior in capital inflows, and evaluate whether this behavior was changed by the existence of an IMF-supported program (or by the size of their financing). Early results in this area were obtained by Hajivassiliou (1987), who, while not focusing directly on the existence of catalytic effects, does find a nonsignificant (and indeed negative) effect on the supply on new private loans of the lagged presence of an IMF program. In this paper, however, the dummy measuring the existence of a program also measures the existence of a previous rescheduling (being built to signal the existence of previous payment difficulties). A more direct attempt to assess the existence of catalytic effects on capital inflows is Rodrik (1996),²² who finds that the effect of IMF lending on net private capital inflows was not significantly different from zero, and negative in many cases. No effect of IMF-lending on foreign direct investment is found by Adji and others (1997). Similar conclusions were reached by Bird and Rowlands (1997), and Bird and Rowlands (2000), with more sophisticated capital inflows model, although they do find some evidence of catalytic effects on bilateral official inflows. One should note that the dynamic specification of these papers is absent or very basic, and one wonders whether the use of longer lags, and perhaps higher frequency data, would have helped. However, the fact that annual data do not reveal catalytic effects suggests that, if those effects are present, they do not show up very rapidly, and are thus not very valuable for crisis management.²³ Edwards (2000) also fails to identify positive catalytic effects, after controlling for the record of past implementation of IMF programs: indeed, he finds that, while a program does not have catalytic effects even following a record of good program implementation, it does have negative catalytic effects if coupled with a recent history of nonperformance. The only paper that does find catalytic effects on private capital flows is Marchesi (2001), who, however, focuses on a very specific form of such flows, namely the decision to reschedule existing external obligations: a program would raise the probability of a rescheduling. In principle, the decision to reschedule a loan can be seen as equivalent to granting a new loan and, so, as a new capital inflow. However, the paper does not make allowance for rescheduling decisions that were part of concerted strategies from those that were due to genuine catalytic effects. And there is no doubt that the IMF did play a major role in several cases of concerted rescheduling (Milivojevié, 1986). Moreover, the paper does not control for the effect of a change in policies on the decision to reschedule; the latter, rather than the IMF’s seal of approval, may be responsible for the increased probability of a rescheduling.²⁴

C. Effects on Spreads

One obstacle in estimating capital inflows equations is that capital movements are typically very noisy. In addition, as noted, a decline in net capital inflows may merely reflect an improvement in the external current account. Thus, some paper have focused on how the interest rates charge on a country’s debt is affected by a program (Table 2).

Table 2.

Spreads and IMF Lending

Table 2.

Spreads and IMF Lending

	Methodology	Question	Sample	Main Findings
Ozler(1993)	Regression analysis.	Does the existence of an IMF-supported program affect the spread over Libor on commercial bank loans?	1,525 commercial bank loan contracts for 26 countries over 1968–81.	The effect is positive (i.e., the existence of a program raises the spread). The effect is statistically significant, although less so in specifications with more variables.
Haldane(1999)	Inspection of spreads.	Does a program affect spreads on secondary bond markets?	Seven large emerging markets during 1996–99.	A large spread persists after the program initiation. Indeed the spread “typically widens further in the immediate aftermath of a program.”
Eichengreen and Mody (2000)	Regression analysis.	Does a program affect the spread at launch on international bonds?	All bond issues by emerging markets from 1991–1 to 1999–IV.	Negative effect of the existence of a program on the spread. The effect is statistically significant in some regressions, in particular for EFF programs and, to a lesser extent, SBAs, but only in equations including the duration of programs with a positive sign (i.e., the effect is negative, but declines over time). The size of the effect is, however, small.
Mody and Saravia (2002)	Regression analysis.	Does a program affect the spread at launch on international bonds?	All bond issues by emerging markets from 1991–1 to 199–IV.	Broadly in line with the previous paper, but, in addition: the size of a program matters, with stronger effects for larger programs; precautionary programs are those where the effect is stronger (and indeed, the catalytic effect does not matter for nonprecautionary programs).

View Table

Table 2.

Spreads and IMF Lending

	Methodology	Question	Sample	Main Findings
Ozler(1993)	Regression analysis.	Does the existence of an IMF-supported program affect the spread over Libor on commercial bank loans?	1,525 commercial bank loan contracts for 26 countries over 1968–81.	The effect is positive (i.e., the existence of a program raises the spread). The effect is statistically significant, although less so in specifications with more variables.
Haldane(1999)	Inspection of spreads.	Does a program affect spreads on secondary bond markets?	Seven large emerging markets during 1996–99.	A large spread persists after the program initiation. Indeed the spread “typically widens further in the immediate aftermath of a program.”
Eichengreen and Mody (2000)	Regression analysis.	Does a program affect the spread at launch on international bonds?	All bond issues by emerging markets from 1991–1 to 1999–IV.	Negative effect of the existence of a program on the spread. The effect is statistically significant in some regressions, in particular for EFF programs and, to a lesser extent, SBAs, but only in equations including the duration of programs with a positive sign (i.e., the effect is negative, but declines over time). The size of the effect is, however, small.
Mody and Saravia (2002)	Regression analysis.	Does a program affect the spread at launch on international bonds?	All bond issues by emerging markets from 1991–1 to 199–IV.	Broadly in line with the previous paper, but, in addition: the size of a program matters, with stronger effects for larger programs; precautionary programs are those where the effect is stronger (and indeed, the catalytic effect does not matter for nonprecautionary programs).

The first paper to include a dummy for the existence of an IMF-supported program is özler (1993). He finds that programs are associated with an increase in spreads, a results that is interpreted as indicating that IMF-supported programs signal (or are associated with) payment difficulties. The effect is statistically significant, albeit less so in specifications including more variables. More recently a similar conclusion was reached by Haldane (1999) who focuses on the behavior of secondary market spreads on the bonds issued by large emerging economies. He finds that a large spreads persists after the initiation of a program and that, indeed, the spread “typically widens further in the immediate aftermath of a program.”

Haldane’s paper, however, looks only at seven countries in a relatively short time frame, and does not test statistically his conclusions, reaching them from visual inspection of the spreads. A more systematic approach is followed in Eichengreen and Mody (2000). This paper has two advantages with respect to others: it is based on a large sample and on high frequency data (thus reducing the risk of averaging between before- and after-the-program data). The paper looks at the spread at launch on bond issues during the 1990s. The strongest catalytic effect are found for EFF loans (the IMF’s longer-term loans) and, to a lesser extent SBA loans (the traditional shorter term loans) in equations when the number of quarters under a program is also included with a positive coefficient. This means that a program lowers the spread but the effect declines over time. The authors conclude that their results “will hearten official observers concerned to document the “catalytic effect” of IMF programs, although they also suggest cautions about the type of programs that exercise this effect.” ²⁵Two qualifications should be made in interpreting these results. First, the regressions do not control for the change in economic policies that typically characterizes an IMF-supported program. This means that it is impossible to distinguish between the decline in the spread that is due to sounder policies, from that arising from the IMF’s seal of approval: as noted, catalytic effects should refer only to the latter.²⁶ Second, the effect of the programs remains, albeit significant, small: the spread is lowered by some 9 percent, or 24 basis points in the sample average. This is not much compared with the average volatility of spreads in international markets. Similar conclusions are reached in Mody and Saravia (2002), with the presence of a program lowering spreads on average by 8 percent. In addition, this paper shows that the size of a program does matter: an additional program size of 10 percent of a country’s long-term debt lowers spreads by about 20 percent. While this effect remains fairly low, it underscores that, more than the seal of approval, what matters is the amount of financial support that the IMF makes available. A second interesting finding relates to precautionary programs (those for which either the authorities declared that there was no intention of actually borrowing, or those for which no borrowing took place): the paper shows that the whole negative effect on spreads comes from these programs. This of course may be tautological: if a program succeeds, this should be reflected in both a decline in spreads and the fact that the country does not need to borrow from the IMF. In any case, this results suggests that, if catalytic effects are present, they do not arise under crises situations (when programs are not precautionary).

A. Five Channels for the Catalytic Effect

B. Channel-Specific Problems

We now turn to the discussion of why some of the requirements summarized in Table 3 are unlikely to be realized in a number of practical cases, and especially under conditions prevailing in capital account crises. Four requirements may be particularly hard to meet in some circumstances.

Requirement 1 (“sufficient information”). This requirement takes two forms. The information channel requires that the IMF has better information than the market. All other channels require that there are no major information asymmetries between the IMF and the program country. As to the first aspect, the process of financial consolidation and technical improvements in information technology have reduced the private sector’s disadvantage in gathering and processing macroeconomic information. Consequently, the resources invested in these tasks by the private sector have increased and the information edge of the IMF may have declined, with respect to, say, the late 1970s. Note also that some information that becomes more critical in capital account crises (that related to financial transactions and intermediaries) may be more easily accessible to financial analysts than to the IMF. As, to the assumption of no major information asymmetries between the IMF and the program country, there are also some problems.

Requirement 2 (“no constraint on policy design”). The policy design channel (Channel 1) assumes that the IMF can identify and promote the best program for each country. Leaving aside the issue of possible errors in policy design, a more intrinsic problem arises from the very nature of the IMF as a multi-country agency. For lack of a better term, we will refer to this problem as “policy contagion.” Suppose the best program that could be designed for a country requires a measure that markets fear may be introduced also in other countries. If the IMF had to deal only with one country, its unconstrained advice would be to introduce that measure. However, the IMF may not go ahead because of the implications that this may have for other countries. To make the case more concrete, the measure in question may be the introduction of some form of capital control or of debt restructuring. The fact that the IMF is supporting this measure in one country (where it could be implemented rapidly, thus minimizing disruptions) might be perceived by markets as indicating that the same approach would be followed for other program countries. The latter perception could trigger speculative attacks. Fearing this, the IMF would stop short of including in a program a measure that would be optimal for that country. This is an extreme case, involving a strong (and very contagion-prone) measure. But the point is more general (it affects, for example advice regarding the choice of the exchange rate regime). IMF advice to a country does not take place in a vacuum, and this may constrain its policy advice, particularly in crisis management packages where more radical (and contagion-prone) measures may be needed.

Requirement 5 (“lending”). Lending, of course, takes place (at least potentially) in all IMF programs. The real question, however, is whether the “limited” lending that characterized catalytic programs is enough to raise the propensity of the private sector to invest in the program country, particularly, in a capital account crisis. This issue is particularly problematic for the insurance channel (Channel V), as highlighted by Zettelmeyer (2000), Eichengreen and Mody (2000), and Ghosh and others (2002). In particular, the analytical model in Zettelmeyer (2000) demonstrates that limited financing does not eliminate the risk for creditors to be locked-in in pre-default situations; indeed, limited financing may prompt a speculative attack as investors take the opportunity of closing their positions.

Requirement 7 (“reasonable and predictable penalty for policy slippages”). The penalty for noncompliance in the execution of IMF-supported programs could be clearly defined in the “old days.” Lack of implementation of critical measures would cause the interruption of a program and of IMF financial support. The latter could be resumed after the implementation of corrective measures. This is still the way things work in a number of countries, those less involved in international financial markets, mostly pure cases of “external current account crises.” But in capital account crises, the penalty for noncompliance cannot be set a priori: it depends on the markets’ reaction. The latter may be disproportionately large due to band-wagon effects. Consequently, the magnitude of the penalty is also difficult to predict.³⁵ The unpredictability of the penalty can have severe effects on the IMF’s willingness to promptly react to lack of compliance, and hence on its credibility and the strength of catalytic effects. Fearing an excessive market response, the IMF may not respond in a timely way to inappropriate policies, thus undermining its role both as provider of information (Channel II), provider of credibility (Channel III), and provider of a screening device (Channel IV).

C. Channel Inconsistencies

It would be somewhat reassuring if the seven requirements for effectiveness we have identified could be shown to have no adverse interaction. However, this is not the case: some requirements, while making some channel stronger, at the same time tend to weaken the others. Thus, attempts to enhance one channel, for example through changes in the way the IMF operates, may overall yield ambiguous results. There appear to be three main “channel inconsistencies,” respectively associated with “conflicts of interest in blowing the whistle,” “transparency and the conditionality process,” and “conditionality and insurance services.”³⁶

(A) “Conflicts of interest in blowing the whistle.” The information channel (Channel II), the commitment channel (Channel (III) and the screening channel (Channel IV) all require that the IMF is an unbiased observer and that “blowing the whistle” is not costly for the IMF. However, when the IMF is involved in policy design (Channel I) and in insurance (Channel V) blowing the whistle is costly.

Channels II, III and IV require that the IMF plays its role fairly: it should provide the relevant information to the market (Channel II) and enforce the implementation of the program policies, interrupting the program in case of noncompliance (Channels III and IV). However, the larger is the cost that the IMF faces when it blows the whistle, the larger will be the disincentive in playing its role effectively.

Unfortunately the IMF does face large costs in blowing the whistle. According to some IMF-watchers, because of the IMF’s political nature—the IMF is administered by a Board of Directors who are political appointees—these costs are primarily political. If this were the whole story, the solution would be to give the IMF more independence, as indeed proposed in De Gregorio and others (1999). But there are costs that arise from the very function the IMF is required to play:

• No matter how much stress is put on program ownership, IMF-supported programs are perceived as IMF-programs. In a way, this is desirable, as the fact that the IMF-expertise is used to build the program should support the catalytic effect (Channel I). But this implies that the IMF suffers a cost in terms of reputation if the program does not work. Note that the cost is there even if the program fails because of lack of implementation: the IMF could in this case be accused of having supported a government that did not deserve credit, possibly because of a “bias towards lending.”³⁷

• As already remarked, stopping a program can trigger a crisis with major costs for the program country. A crisis involves also costs for the IMF. Not only because the program country is an IMF member, but also because of possible contagion effects to other IMF members. Thus, from the perspective of an international agency, it may be preferable to wait to “blow the whistle” as this allows to “buy time,” giving financial markets time to differentiate and reducing, in this way, contagion effects. The cost of this for the IMF is a weakening of its credibility, and thus a weakening of catalytic effects. Needless to say, this cost is larger the more a countries is exposed to the risk of a capital account crisis.³⁸

Note also that these costs exist not only when a program is already in place, but also when the IMF has to decide whether to start a new program.

(B) “Transparency and the conditionality process.” Channel II requires transparency, but transparency enhances problems in the conditionality/negoliation process required by channels III and IV.

As stressed by Mussa and Savastano (1999), the conditionality mechanism is not based on a “contract,” but is rather a “process:”

“A typical IMF-supported program is not set in stone at its inception, either proceeding subsequently in exact accord with the initial plan, or terminated because of some minor deviation. [D]isbursements proceed automatically if all the performance criteria are met as initially specified. This rarely happens all the way through an arrangement. Instead, if various conditions are not met, deviations may be accommodated with ‘waivers,’ projections may be revised, and numerical targets changed. Those who participate in the process of IMF-supported programs, from both sides, do so with full awareness of their fundamental iterative, open-loop character.”

The “constructive imperfection” of this process is plain in these words. Economic policy-making is not an exact science. The process of steering an economy back on track is a continuous one and may require, for example, waiving some conditions that had been initially regarded as critical, or, vice versa, halting the program if a sufficiently number of conditions initially regarded as not critical was not met, or, again, changing conditions. In virtually all of these cases, judgment and negotiation will play a critical role.

Transparency, in particular the publication of Letters of Intent—which is yet critical to enhance the information channel—makes this process more difficult to manage. For example, failure to implement a certain measure may trigger a crisis, even when the IMF would have been willing to grant a waiver. Conversely, the granting of a waiver may be seen as a watering down of the program, which could be construed as based on political reasons.

It may be argued that these problems could be solved through increased transparency. For example, the IMF could be more explicit or more prompt in clarifying what is critical and what is not, and under which circumstances it would be willing to support a change in policies. This is a naive view. Not only is the range of events that a program can face too large to be subject to pre-specification. More fundamentally, pre-specification would be inconsistent with the negotiation aspects of an IMF-supported program. Suppose financial markets reacted negatively to the lack of implementation of a measure included in a published Letter of Intent. If the IMF immediately stated publicly its willingness to waive the condition, so as to stem a speculative attack at its inception, it would lose any leverage in negotiating remedial actions. Thus, in these circumstances, what the IMF might end up doing would be to provide some half-hearted support which would not be fully convincing for the markets (it might indeed be even counterproductive).

The critical point here is that the negotiation/conditionality process requires time, and transparency, coupled with high capital mobility, dramatically cuts the time available for the process to proceed effectively and to lead to an optimal redesign of programs as they proceed. In this environment, conditionality may become perverse: not only does it not provide credibility, but it may end up being used as a coordinating device for speculative attacks, with program benchmarks/performance criteria assuming inappropriately the role of triggering devices.³⁹

(C) “Conditionality and insurance services. “The commitment and screening channels presuppose conditionality, and this creates uncertainty about the availability of IMF resources; but certainty that sufficient resources will be made available is key to the working of the insurance channel.

Channel IV requires that liquidity is available “with no strings attached.” Ever since Bagehot (1873) it has been recognized that the attempt to tighten the type of “security” required for granting liquidity support during a panic would have disastrous results. But this is inconsistent with conditionality.⁴⁰ It could be argued that this inconsistency could be addressed by eliminating the conditionality in pure “liquidity” crises. This is the logic underlying the recommendations of the Meltzer Commission: liquidity support should be restricted to countries that have pre-qualified by meeting a (limited) number of performance criteria ahead of a crisis. There is a difficulty, though, in this logic, since countries, unlike the representative individual of most economic models, may be subject to sudden regime changes. No matter how good the historical record has been (and typically for most IMF borrowers “historical” means a decade or so), the domestic political process may engender a coalition that is simply “unwilling” to accept the burden of past policies. It is not accidental, in our view, that in most cases capital account crises burst out in the vicinity of a major election, in the run-up to which one or more of the candidates has won large popular support precisely by promising a “regime change.” A measure of conditionality, and of discretion by the crisis manager in assessing the credibility of the ruling coalition, is therefore inevitable. But the insurance service provided by an insurer that feels the need to insure itself against the adverse event is unlikely to meet with success in the eyes of private investors.