Almost six years have passed since Argentina introduced a currency board system called "convertibility." It consists of a one-for-one peg between the peso and the U.S. dollar and full backing of the monetary base with international reserves. In spite of its resounding success in reducing inflation, there is still a considerable degree of skepticism in the world, although not so much in Argentina, about whether convertibility is a suitable monetary arrangement for the long run. Prestigious economists have argued that convertibility was a useful tool to stop hyperinflation in Argentina but that it is not suitable for countries facing less dramatic conditions ( see John Williamson [ 1995] and the speeches by Max Corden, Stanley Fischer, and Williamson in World Bank [1997]). According to them, even Argentina should think of abandoning this system as a way to promote economic growth and higher employment.

International institutions like the International Monetary Fund and the World Bank have been extremely supportive of convertibility in Argentina. But in private they are concerned about what Sebastian Edwards, former World Bank chief economist for Latin America, has called the "exit problem," (i.e., how a country like Argentina can get out of the commitment to a fixed exchange rate without risking price stability).

. Historical Background

In July 1989 the Argentine economy was in shambles. Per capita GDP was 10 percent lower than in 1980, and social indicators had deteriorated dramatically. The fiscal deficit was 7.6 percent of GDP, and the gross investment rate was only 14 percent of GDP. Years of fiscal and monetary indiscipline had resulted in hyperinflation. Street riots and social chaos forced then president Raul Alfonsin to transfer power to newly elected president Carlos Menem five months before the normal time. After 18 months of political hesitation, the Menem government finally decided to tackle the problem at its root and launched the Convertibility Plan. By the end of 1991 annual inflation was below 20 percent, and GDP was growing at 9 percent per year. Inflation continued to fall until it converged to international levels in 1994. A faster convergence was not possible due to the strength of aggregate demand between 1991 and 1993. At the same time, the annual rate of growth in total factor productivity increased from negative figures in the 1982-1990 period to around 6 percent during 1991-1994.

What provoked the miracle? Unquestionably, monetary reform provided the basis for the transformation that followed. Three key elements of the Argentine convertibility system are: (i) it was established by a law; (ii) it abolished price indexation; and (iii) it allows contracts to be denominated in foreign currencies and even allows foreign currencies to be used as alternative means of payment. These three elements were instrumental for the success of convertibility. In particular, dollarization allowed the extension of financial asset maturities, eliminating the short-term debt overhang.' Also, from a political point of view, convertibility was a smashing success. It stopped hyperinflation without producing a recession and without causing regressive income redistribution. This is important since good economic policies need political support to survive in a democracy.

Macroeconomic stabilization was just the starting point. The real core of the program was its insistence on microeconomic reform. By that, we mean a series of structural changes geared at reducing or eliminating distortions in goods and factor markets. The extent and depth of microeconomic reform in Argentina between 1991 and 1996 was unprecedented, maybe greater than in any other country during a similar time span. In addition, these reforms were introduced by a democratic government, with no extraordinary foreign assistance, and during a period in which the external terms of trade were at a historic low. Although some tradable activities were hurt by real appreciation and trade liberalization, convertibility forced entrepreneurs to deal with these problems by introducing more rapid improvements in technology and management. This would not have happened if the possibility of a devaluation had existed. At the same time, the government helped export- and importcompeting industries in various ways. For example, the extensive privatization cum deregulation program in energy, telecommunications, mining, ports, and transportation reduced the relative price of key inputs and increased their supply both quantitatively and qualitatively. The government also helped to mitigate the relative price squeeze on tradable industries by introducing expansionary supply-side shocks (i.e., by reducing or eliminating distortionary taxes). Finally, to avoid unmanageable current-account deficits the economic authorities addressed the problem of increasing national savings early during the reform process by introducing reforms in social security and tax policy which have the implication of taxing consumption rather than saving. In this way, Argentina did not resort to a contraction in investment as a way to reduce the current-account deficit. The factors mentioned above resulted in significant cost reductions and increases in productivity for the corporate sector, which explains why exports increased by about 75 percent between 1992 and 1995 despite the peg of the nominal exchange rate and the appreciation of the real exchange rate between 1991 and 1993.

A reversal of fortune took place in Argentina in 1995, but it was entirely due to an external shock. The Mexican devaluation crisis of 20 December 1994 had a severe contagion effect on Argentina, where the banking system lost 18 percent of deposits in just three months. Credit contraction resulted in a 4.4-percent reduction in output and a sharp increase in unemployment. Yet, the Convertibility Plan stood unchanged. The crisis was used by the government as an opportunity for speeding up the reform process. Today, the Argentine economy is healthy again, recovering from last year's recession. GDP growth is expected to be 4 percent in 1996 and 6 percent in 1997, and inflation for the 12 months ending in November 1996 is practically zero. Having withstood the crisis, Argentina's monetary system looks much more solid and credible than before. While unemployment is a problem, the Menem administration is determined to fight it the "old-fashioned way" (ise., by introducing greater flexibility in labor markets).

II. Relations with the IMF

The Convertibility Plan received strong support from the international financial community and, especially, from the International Monetary Fund. In July 1991, the IMF approved a stand-by facility for SDR 780 million,2 which was replaced in March 1992 by an extended fund facility (EFF) in the amount of SDR 2,483 million ( 161 percent of quota) . This included augmentation and set-asides to support the debt and debt service reduction operation with commercial bank creditors known as the Brady Plan. In June 1994, the authorities decided to accept the IMF recommendation not to make use of the last two scheduled purchases since there had been a substantial increase in Argentina's voluntary access to international capital markets. Unfortunately, the panic created by the Mexican devaluation crisis of 20 December 1994 drastically reduced this access, forcing Argentina to negotiate a fourth-year extension of the EFF on 6 April 1995. This time the approved amount was SDR 1,537 (100 percent of quota), of which 68 percent was disbursed up front.

When the EFF extension expired in April 1996, the authorities decided that it was convenient for Argentina to remain under the umbrella of an IMF program, and since a fifth year extension of the EFF was not possible, the IMF suggested a new stand-by facility in the amount of SDR 720 million. At the end of 1996, Argentina's outstanding use of IMF credit was close to three times the quota, making it the Fund's third-largest client. IMF support was greatly appreciated by Argentina's government and investors. It was a catalyzer for additional funding from the World Bank, the Inter-American Development Bank, and the Eximbank of Japan, which enabled Argentina to reach an agreement with commercial bank creditors in April 1992. Voluntary private capital flows were reestablished after this agreement, following ten years of interruption due to the 1982 debt crisis. In the aftermath of the tequila effect, the IMF acted swiftly and efficiently for the second time, helping the Argentine government to put together an impressive financial package that stopped a bank run in the second quarter of 1995. This package also included loans from other official institutions, as well as the subscription of a U.S. $2.0 billion "patriotic" bond by the private sector. Even in "normal" times, the endorsement of the Convertibility Plan by the IMF Fund was good for credibility.

Nevertheless, differences of opinion have existed in matters of economic policy between the Argentine authorities and the IMF staff since the beginning of the Convertibility Plan. For example, the IMF was never comfortable with the idea of cutting taxes, no matter how distortionary they were, even if the government was running a fiscal surplus. According to the Fund's credo, accumulating reserves is always better. Thus, the supply-side twist of Argentina's policies sometimes generated some resistance from the IMF. Of course, if Argentina had not been so obstinate, many important reforms such as the privatization of the public pension system would not have taken place. But the main misunderstandings arose from the fact that the IMF never really believed in the currency board system as a longterm arrangement (which seems to be a strange position, coming from the institution created at Bretton Woods). Actually, the difference between convertibility and alternative monetary systems as typically found in LDC's is the same as the difference between rules and discretion. Therefore, by not expressing full faith in convertibility, the IMF took a position against rules and in favor of discretion.

III. Why Do Governments Devalue?

Actually, the question should be why governments devalue without prior announcement at times when there is no monetary imbalance. Although corrective devaluations are very common, this paper does not focus on them. Instead, we are interested in a monetary policy that is both discretionary, in the sense that it cannot be fully anticipated by economic agents, and consistent, in the sense that it does not produce a systematic loss in reserves. The condition for consistency is easy to establish. If there is no increase in real money demand, an increase in domestic credit or net domestic assets of the central bank must be followed by a proportionately higher or equal devaluation. On the other hand, an increase in real money demand reduces the size of a devaluation needed to avoid monetary disequilibrium (i.e., a persistent loss in reserves), given an increase in domestic credit. Alternatively, if the government devalues, the rate of expansion in net domestic assets cannot exceed the rate of devaluation unless real money demand is also growing.

Having defined a monetary policy that is both discretionary and consistent, the next question is why governments and the IMF like it so much. We have identified at least six reasons: (i) to collect the inflation tax; (ii) to increase the level of reserves; (iii) to improve fiscal conditions; (iv) to bail out domestic debtors and banks; (v) to increase real output and reduce unemployment; and (vi) to correct external imbalances. A seventh reason, to control interest rates, has little relevance in financially open economies, since under high capital mobility real interest rates are mainly determined by foreign interest rates and country risk, which are mostly exogenous and therefore independent of monetary policy. Of the six reasons listed above, (i), (iv), and (v) are particularly appealing to governments while (ii), (iii), and (vi) are particularly appealing to the IMF.

The most seductive element of discretionary monetary policy for sovereign governments is that it allows them to collect the inflation tax. Although this tax can be collected even if inflation is perfectly anticipated, the proceeds are much higher if it is not; hence, the appeal of monetary discretion. As noted above, to avoid losing reserves the government must accompany an increase in domestic credit with a "consistent" devaluation. The increase in money supply can then be spent by the government in various ways or can be used by the central bank to raise credit to the private sector. As is true of any tax whose rate changes at the discretion of the government, the inflation tax is time-inconsistent (Finn Kydland and Edward Prescott, 1977; Guillermo Calvo, 1978). This means that inflationary targets are not credible. When this happens, the increase in the domestic credit and the nominal exchange rate needed to collect the same amount of money in real terms will rise, which is how Argentina got hyperinflation in the first place.

Alternatively, the government can engineer a devaluation that is not accompanied by an increase in domestic credit (or that is accompanied by a proportionately smaller increase in domestic credit). In this case, central-bank reserves accumulate as people respond to the higher demand for nominal money balances by selling foreign assets to the central bank. Here too the government collects the inflation tax, but in this case the "proceeds" are used to accumulate international reserves. A devaluation can improve fiscal accounts by reducing public wages, pensions, and interest payments on the non-indexed peso-denominated public debt in real terms, which is just another way of collecting the inflation tax. Finally, when a government devalues, part of the proceeds of the inflation tax does not accrue to the government itself but to those who have borrowed pesos. If loan insolvency is a major concern, a devaluation can help improve bank balance sheets by increasing the likelihood of repayment, thus avoiding centralbank intervention.

We now turn to the last two motives for a devaluation: employment and currentaccount stabilization. As long as the real wage is above the equilibrium level and the nominal wage is "sticky," a devaluation can reduce unemployment by reducing real wages. This would also increase real output, as firms are induced to hire more labor. This effect, usually referred to as increase in competitiveness, has been emphasized by the traditional literature on fixed versus flexible exchange rates, leading to the conclusion that a flexible exchange-rate system reduces vulnerability to external shocks. In other words, the expansionary effect of a devaluation can offset the contractionary effect of a deterioration in the terms of trade or an increase in the foreign interest rate. As argued below, we believe that this emphasis has been both excessive and misleading.

Finally, to the extent that the currentaccount balance is the difference between income and absorption, a devaluation can reduce the current-account deficit either by increasing income or by reducing absorption. We noted above that a devaluation can be expansionary on output. Is it also contractionary on absorption? This is certainly possible since a nominal devaluation can reduce private wealth, wages, and pensions in real terms (this is nothing but the inflation tax as it impinges on people's assets and income). Consequently, the trade balance should generally improve. Moreover, if part of the external debt is denominated in local currency, as with Mexico's Cetes, a devaluation could also translate into a larger improvement in the current account.

IV. The Costs of Discretionalty

Discretionary devaluations increase the exchange-rate risk considerably, thereby raising domestic real interest rates. Also, they reduce the marginal productivity of capital, since exchange-rate instability typically discourages improvements in management and the adoption of better technologies which increase capital productivity. In addition, exchange-rate risk has the effect of concentrating public and private debt in the short run, which raises the exposure of these sectors to changes in interest rates. These effects unambiguously lower investment and, hence, growth. In other words, a high real interest rate is the price that a government has to pay for having a discretionary monetary policy. Are there any offsetting effects on welfare? It appears that such effects do not transpire. In the first place, even if a devaluation were capable of increasing real output, it could not increase growth. On the other hand, a devaluation may be contractionary even in the short run if it leads to a large immediate increase in country risk. In this case, a flexible exchangerate system would exacerbate external shocks, generating larger output losses. This was clearly the case for Argentina, where the abandonment of convertibility at the time of the tequila shock would have had disastrous financial and real consequences.

Although an exchange-rate devaluation may be an expeditious way of reducing unemployment if monetary wages are sticky, it is not if real wages are sticky.3 In this case, the only way to reduce unemployment is through policies that shift the labor demand curve upward, either by raising productivity or by reducing nonwage labor costs. Furthermore, even if a devaluation were capable of reducing real wages, easy resort to this instrument would postpone increases in productivity and labor-market deregulation.

Certainly, increasing inflation-tax revenue does not provide a good argument for discretionary devaluations since few taxes are so distortionary and regressive as the inflation tax. Even if one looks at this question from the narrow point of view of the fisc, what the government has to gain from the inflation tax it loses by paying higher interest on public debt and by concentrating it in the short term. Similar considerations apply to the stabilizing effects of a devaluation on government expenditures and the current account. To assume that these effects are lasting or permanent is to assume too much. In fact, devaluations postpone more permanent fiscal adjustments that could be accomplished by reducing public employment, privatizing public enterprises, and streamlining excessively generous pension benefits. With respect to the currentaccount deficit, the only way in which a reduction can be sustainable is if national saving increases in the long run, something that a devaluation cannot guarantee. By the same token, avoiding a banking crisis by levying a capital tax on deposits to bail out borrowers and banks is a questionable practice that does not improve social welfare.

V. The Case for Convertibility

The alternative to monetary discretion is a monetary rule. One possible rule is not to increase net domestic assets at all to avoid losing reserves, while the exchange rate is fixed. In this case, the central bank cannot finance the treasury or extend credit to commercial banks. This policy is called convertibility in Argentina and a currency board system elsewhere. Its main benefit is that it reduces exchange-rate risk (hence, real interest rates).

Although convertibility is a monetary rule, it is not the only possible one. Another rule would be to fix money supply or to increase it at a preannounced rate (Milton Friedman's x-percent money rule). In an open economy, this requires a clean float, namely, a fully flexible exchange-rate system. If there is no shift in real money demand, annual inflation converges to the rate of money growth. However, if real money demand increases or decreases, the nominal exchange rate changes in the opposite direction. The problem of using a monetarygrowth target is that it cannot accommodate real money demand shocks, therefore resulting in exchange-rate volatility. In conclusion, if a country decides that monetary rules are better than discretion, as Argentina did in 1991, then convertibility may be the best choice.