Stopping High Inflation: An Analytical Overview
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

The evidence on stopping high inflation is interpreted in terms of an analytical framework. The evidence suggests that, by using the exchange rate as the nominal anchor, hyperinflations have been stopped almost overnight with relatively minor output costs. In contrast, exchange rate-based stabilizations in chronic-inflation countries have typically resulted in a sluggish adjustment of the inflation rate, sustained real appreciation of the domestic currency, current account deficits, and an initial expansion in economic activity followed by a contraction. These stylized facts are shown to be consistent with the predictions of a staggered-prices, cash-in-advance model. [JEL F41]

Abstract

The evidence on stopping high inflation is interpreted in terms of an analytical framework. The evidence suggests that, by using the exchange rate as the nominal anchor, hyperinflations have been stopped almost overnight with relatively minor output costs. In contrast, exchange rate-based stabilizations in chronic-inflation countries have typically resulted in a sluggish adjustment of the inflation rate, sustained real appreciation of the domestic currency, current account deficits, and an initial expansion in economic activity followed by a contraction. These stylized facts are shown to be consistent with the predictions of a staggered-prices, cash-in-advance model. [JEL F41]

The evidence on stopping high inflation is interpreted in terms of an analytical framework. The evidence suggests that, by using the exchange rate as the nominal anchor, hyperinflations have been stopped almost overnight with relatively minor output costs. In contrast, exchange rate-based stabilizations in chronic-inflation countries have typically resulted in a sluggish adjustment of the inflation rate, sustained real appreciation of the domestic currency, current account deficits, and an initial expansion in economic activity followed by a contraction. These stylized facts are shown to be consistent with the predictions of a staggered-prices, cash-in-advance model. [JEL F41]

There is no subtler, no surer means of overturning the existing basis of society than to debauch the currency. The process engages all the hidden forces of economic law on the side of destruction and does it in a manner which not one man in a million can diagnose.

—John Maynard Keynes (1920, p. 220)

F rom a historical perspective, high inflation is a recent phenomenon. Prior to World War I, episodes of high inflation were mainly confined to the period of the assignats in France, the American War of Independence, and the Confederacy during the American Civil War (see Capie (1986)). The historically rare occurrence of high inflation reflects the prevalence of convertible currencies and commodity monies, which put a natural lid on inflationary forces. In the aftermath of World War I, however, hyperinflation burst onto the stage. Heavy disruptions and reparation payments resulting from the war caused massive fiscal deficits whose monetary financing led to hyperinflation in Austria, Germany, Hungary, and Russia. Following World War II, hyperinflation erupted once again in Hungary, Greece, and China. In 1985, Bolivia experienced the first hyperinflation of the twentieth century not related to a foreign war, civil war, or political revolution.

After World War II, when hyperinflation had already achieved notoriety, a much less dramatic, but equally ominous, phenomenon began to emerge: chronic inflation. Some countries, particularly in Latin America, began to endure high (relative to industrial countries) and persistent rates of inflation, which in many instances have lasted up to the present day. Argentina, for example, has suffered from chronic inflation since the late 1940s and has been unsuccessful in eliminating it, in spite of eight major plans (two per decade) and countless other attempts. A similar story could also be told for Brazil and Uruguay. There are, however, success stories: Chile, Israel, and Mexico managed to reduce three-digit annual inflation rates to about 20 percent during the 1980s. Although such levels of inflation are still quite high by international standards, they are viewed as a significant achievement, considering the resilience of chronic inflation.

High inflation has proved to be fertile ground for researchers. Seminal studies by Cagan (1956) and Sargent (1982) on hyperinflation, and Pazos (1972) on chronic inflation have set the tone for much of the discussion. A reading of the extensive literature suggests that theory and evidence need to be further integrated to enhance our understanding of inflation stabilization in high-inflation countries. With this in mind, this paper interprets the main stylized facts associated with stopping hyperinflation and chronic inflation in terms of a unified analytical framework.

The model used in this paper, based on Calvo and Vegh (1991a), provides a simple and plausible explanation of the dynamics of stabilization policy. The model relies on intertemporal substitution effects as the key channel through which stabilization policies may have real effects, along the lines of Calvo (1986). In addition, the presence of sticky prices introduces dynamic considerations that are key to an understanding of the outcome of inflation stabilization programs. Since programs designed to stop high inflation have usually relied on the exchange rate as the nominal anchor, the analysis will concentrate on exchange rate-based stabilization.1

The discussion is organized around two analytical exercises, which are used to interpret the evidence on stopping hyperinflation and chronic inflation. The first exercise considers a reduction in the rate of devaluation that is fully credible, in the sense that the public views the policy change as permanent. Under these circumstances, inflation falls instantaneously without any output costs. Sticky prices do not prevent an instantaneous adjustment because all price-setting behavior is assumed to be forward looking. It is argued that the analytical exercise of a permanent reduction in the rate of devaluation is, to a first approximation, relevant for interpreting the end of hyperinflation. The reasons are, first, hyperinflationary processes appear to be characterized by the absence of backward-looking behavior; and, second, programs designed to stop hyperinflation usually command high credibility given the unsustainable state of affairs. The evidence reviewed in the paper seems to bear out the prediction that hyperinflation can be stopped abruptly and at relatively small output costs, compared to those that would obtain in low-inflation countries.

A second analytical exercise assumes that the reduction in the rate of devaluation is not credible, in the sense that the public expects the higher rate of devaluation to resume at some point in the future. The fall in the nominal interest rate that results from the lower devaluation rate and the assumption of perfect capital mobility is thus viewed as temporary. Since the cash-in-advance constraint requires that money be used to purchase goods, the opportunity cost of holding money is part of the effective price of consumption. The temporary fall in the nominal interest rate thus reduces the effective price of consumption in the present relative to the future. Hence, demand for both traded and nontraded goods increases and leads to an initial expansion in the nontraded goods sector and a current account deficit. The slow convergence of inflation to the rate of devaluation results in a sustained real appreciation of the domestic currency, which ultimately reduces the demand for nontraded goods. As a consequence, output falls and a recession sets in. The recession may occur either before or when the program ends. The real effects caused by a noncredible stabilization do not depend on whether the program is eventually abandoned, as the public expected, or not.

It is argued that the analytical exercise of a noncredible reduction in the rate of devaluation can be used to interpret inflation stabilization in chronic-inflation countries, where a history of failed stabilizations, together with the ability of the economy to live with high inflation, makes any attempt to stop inflation less than fully credible. The stylized facts are generally consistent with the predictions of the model. In particular, a boom-recession cycle has characterized both failed and successful stabilizations, with the recession often setting in before the program ends (see Kiguel and Liviatan (1992a)). Furthermore, the dynamic adjustment predicted by the model, with U-shaped curves describing the behavior of both inflation and the real exchange rate (defined as the relative price of traded goods in terms of nontraded goods), is consistent with the stylized facts. The inflation “inertia” exhibited by the model, in the sense that the inflation rate remains above the devaluation rate, has been observed in most programs.2

The paper proceeds as follows. Section I discusses the stylized facts associated with stopping high inflation. Section II introduces the model and analyzes the effects of a permanent reduction in the rate of devaluation. Section III discusses the evidence on ending hyperinflation in light of the model. Section IV examines a temporary reduction in the rate of devaluation and matches the analytical results with the evidence on stopping chronic inflation. Section V discusses the quantitative relevance of the lack-of-credibility (or “temporariness”) hypothesis and the role of backward-looking behavior, and offers some concluding remarks.

I. Stylized Facts of High-Inflation Stabilization

This section first discusses the key differences between chronic inflation and hyperinflation, and then reviews the main stylized facts associated with ending both hyperinflation and chronic inflation.

Hyperinflation Versus Chronic Inflation

For practical purposes, the term “hyperinflation” will be defined as in Cagan’s (1956) classic paper.3 At a conceptual level, it is useful to keep in mind the distinction between “hyperinflation” and “chronic inflation,” emphasized by Pazos (1972). Pazos (1972) argues that chronic inflation exhibits two key characteristics. First, it may last for long periods of time; it is not measured in terms of months, as is the case of most hyperinflations, but in terms of years. Second, chronic inflation has an intermediate intensity—higher than that of moderate inflation but much lower than that of hyperinflation—which results from countries’ learning how to live with high and persistent inflation by creating various indexation mechanisms. Specifically, inflation does not have an inherent propensity to accelerate, or if it does, soon reaches a new plateau. In hyperinflations, however, the rate of inflation oscillates freely, before accelerating exponentially in the last six months or so.4

In order to illustrate the differences between chronic inflation and hyperinflation, Table 1 shows annual inflation rates and money growth rates during the last three decades for six Latin American countries (Argentina, Bolivia, Brazil, Chile, Mexico, and Uruguay), Israel, and, for comparison purposes, the United States. Beyond the general notion of high and persistent inflation, there is no generally accepted definition of chronic inflation. Moreover, any definition would probably not withstand the test of time, since the notion of what constitutes high inflation changes over time.5

Table 1.

Inflation and Money Growth in Selected Countries

(In percent per year)

article image
article image
Sources: IMF, International Financial Statistics (various issues); except for 1988–91 M1 figures for Brazil, from Central Bank of Brazil.

Averages are for 1962–70.

In spite of the lack of a clear definition of chronic inflation, some cases are beyond dispute. Countries such as Argentina, Brazil, and Uruguay would be classified as chronic-inflation countries under any sensible definition. In all three countries, average annual inflation has been above 20 percent in all three decades (Table 1).6 Argentina has had triple-digit inflation since 1975 (with the exception of 1986, when, as the result of the Austral Plan, inflation fell to 90.1 percent). Uruguay’s inflation has been above 30 percent since 1972 (with the exception of 1982, when, as a result of the 1978 stabilization program, inflation fell to 19 percent), and above 50 percent since 1984.7 As one would expect—and as the averages presented at the bottom of Table 1 clearly suggest—the proximate cause of inflation was high money growth. Massive money printing, in turn, resulted from the need to finance chronic budget deficits.

Chile, Israel, and Mexico provide cases in which, after reaching three-digit figures, chronic inflation was successfully stopped. During the 1960s and 1970s, inflation in Chile was not much different from that in Argentina, Brazil, and Uruguay. However, as a result of the stabilization plan of 1978, inflation fell to 9.9 percent in 1982, and averaged 20.3 percent during 1981–90. In contrast, Israel and Mexico were low-inflation countries during the 1960s, with Mexico experiencing the same inflation rate as the United States. During the 1970s inflation unraveled and soon reached three-digit figures. Both the Israeli stabilization plan of July 1985 and the Mexican program of December 1987 brought inflation down to around 20 percent where it has remained ever since.

The inflationary processes in the six countries just reviewed (Argentina, Brazil, Chile, Israel, Mexico, and Uruguay) exhibit the main characteristics of chronic inflation suggested by Pazos (1972). As already noted, inflation has been high and persistent. Moreover, inflation does not have an inherent propensity to accelerate, as best exemplified by the case of Uruguay, where inflation has been remarkably stable in the last 30 years. When inflation does accelerate, it soon reaches a new plateau. In Argentina, for instance, inflation accelerated in 1975 to triple-digit figures and remained there for 11 years. In Israel, after 9 years of double-digit figures, inflation jumped to triple-digit figures in 1980 and remained there for 6 years, until the 1985 stabilization plan.

Bolivia provides an excellent example of the explosive nature of hyperinflation. In contrast to the other six countries, Bolivia is certainly not a chronic-in flation country. Annual inflation averaged only 5.5 percent during the 1960s, and was below 10 percent in all but four years during 1961–77. By the end of the 1970s, however, political instability led to high inflation. In 1982 inflation jumped to 123.5 percent, as the debt crisis cut short the inflow of foreign credit and forced the government to resort to monetary financing. High inflation skyrocketed into hyperinflation, reaching 62.7 percent a month in April 1984. Hyperinflation thus started less than two years after annual inflation reached three-digit levels. It lasted 18 months and was stopped by the September 1985 stabilization. From 1987–91 annual inflation averaged 16.9 percent.8

The explosive nature of the inflationary process in hyperinflation is also clear from other historical episodes. In Germany, for instance, annual inflation averaged only 1.8 percent between 1900 and 1913. In July 1914, one month before the outbreak of World War I, note convertibility was suspended to prevent further losses of gold reserves. During the war (1914–18), annual inflation averaged 25.2 percent. In 1919 inflation jumped to 226.4 percent and, after slowing down to 79.3 percent in 1920 and 142.4 percent in 1921, climbed to a monthly rate of 89 percent a month in August 1922. The hyperinflation lasted 17 months and came to an abrupt end in January 1924. Annual inflation averaged 3.8 percent during 1925–29.9, 10

Stopping Hyperinflation: Stylized Facts

Hyperinflations offer a fascinating laboratory for the study of monetary phenomena.11 Astronomical rates of inflation serve to isolate many aspects of reality, which, under normal conditions, are likely to be obscured by other considerations. For further reference, it is useful to list the hyperinflations (and subsequent stabilizations) that will be referred to in this paper.

  • The post-World War I European hyperinflations: Austria (October 1921–August 1922), Germany (August 1922–November 1923), Hungary (March 1923–February 1924), Poland (January 1923–January 1924), and Russia (December 1921–January 1924).12 The common theme in these episodes is how hyperinflation was successfully brought under control by the introduction of drastic fiscal reforms, giving central banks greater independence and restoring (or virtually restoring) convertibility of the domestic currency in terms of the dollar or gold, as emphasized by Sargent (1982) in his paper on the hyperinflations in Austria, Germany, Poland, and Hungary.

  • The post-World War II European hyperinflations: Hungary (August 1945–July 1946) and Greece (November 1923–December 1945).13 The Hungarian stabilization fits the pattern identified by Sargent (1982) of drastic reforms that achieve immediate price stability. In contrast, the Greek stabilization was a more diffuse process, since it took the government over a year and three attempts to finally put the required measures in place.

  • The Taiwanese hyperinflation (1945–49).14 The interesting feature of this episode, which has not achieved the fame of its European counterparts, is that, according to Makinen and Woodward (1989), stabilization apparently took place not only without any fiscal adjustment, but, more important, with no prospects of any future fiscal adjustment. It was only in mid-1952 that massive aid from the United States helped in balancing the budget.

  • The Bolivian hyperinflation (April 1984–September 1985).15 The key components of the stabilization program were a drastic fiscal correction, unification of exchange rates, and a return to full convertibility—capital controls had been in place since late 1982. Unlike other hyperinflations referred to above, which were based on a fixed exchange rate, the exchange rate system in Bolivia during the stabilization plan is best characterized as a dirty float.16 However, when the exchange rate markets were unified, the official exchange rate depreciated by 1.6(H) percent in one day, thus eliminating the parallel market premium practically overnight. The exchange rate stabilized immediately after this initial adjustment, providing a de facto anchor to the system.

For the purposes here, two main stylized facts regarding hyperinflation stabilization need to be emphasized:17 (1) inflation is stopped immediately; and (2) output costs are relatively small.

(1) Inflation stops abruptly. In most instances, price stability was virtually achieved overnight following exchange rate stabilization.18 To illustrate this point, Table 2 shows monthly averages for the rates of devaluation and inflation 12 months before and 12 months after the exchange rate was stabilized in eight hyperinflations (Austria, Poland, Germany, Greece, Hungary after each World War, Bolivia, and Taiwan).19 The figures clearly show how anchoring the exchange rate can ensure a swift return to price stability. The most dramatic examples are Hungary 1946 and Germany, where the monthly inflation rate in the 12 months before stabilization averaged 19,800 percent and 455.1 percent, respectively. After stabilization, the monthly average dropped to 1.3 percent and 0.3 percent, respectively. In more moderate hyperinflations, such as Hungary after World War I where monthly inflation averaged 33.3 percent before stabilization, the same phenomenon of overnight price stability took place. Even when price stability was not fully achieved, as in Bolivia and Taiwan, the drop in inflation was substantial. Not surprisingly, in Bolivia and Taiwan, the exchange rate did not stabilize completely.20

Table 2.

Devaluation, Inflation, and Money Growth in Hyperinflations

(In percent a month)

article image
Sources: Austria, Germany, and Hungary 1924: Sargent (1982); Poland: Sargent (1982) and Llach (1990); Greece: Makinen (1984); Hungary 1946: Cagan (1956) and Bomberger and Makinen (1983); Taiwan: Makinen and Woodward (1989); and Bolivia: IMF, International Financial Statistics (various issues), and Morales (1988).Note: The date in parentheses following the country name indicates the month in which the exchange rate stabilized. Money refers to notes in circulation, except in Bolivia and Taiwan where it indicates M1.

In summary, the evidence clearly suggests that in hyperinflationary situations, price stability can be the immediate result of using the exchange rate as the nominal anchor. As discussed in Section III, during hyperinflation virtually all prices are indexed to the dollar or, which amounts to the same thing, quoted in dollars; hence, stabilizing the exchange rate is tantamount to achieving price stability.21

(2) Output costs are relatively small. Although this second stylized fact is certainly more controversial than the first one, a review of the evidence suggests that hyperinflations have been stopped with small output costs compared to those that would have resulted from stopping inflation in low-inflation countries.22 There are two key difficulties in assessing the output effects of stopping hyperinflation. First, available data are usually sparse and unclear. Second, the real effects of stabilization per se are difficult to disentangle from the real dislocations that characterize the transition from hyperinflation to price stability. Moreover, in the European hyperinflations following the two World Wars, the effects of the war itself and the burden of heavy reparations payments have also tended to blur the picture.

Of the nine hyperinflations listed above (Tables 3a3g present data for seven episodes), economic activity appears to have increased following stabilization in three cases—Germany, Greece, and Russia. In Germany the per capita index of industrial production rose substantially in 1924—the year following stabilization—from the depressed level of 1923 that had resulted mainly from the French occupation of the Ruhr and the policy of passive resistance. The rate of unemployment also decreased substantially from its peak of 23.6 percent in the fourth quarter of 1923, to 3.8 percent in the second quarter of 1925. In the summer of 1925, a recession set in. Garber (1982) attributes this recession not to the monetary stabilization per se, but rather to a reallocation of industry, as the capital goods sector had been heavily subsidized during the hyperinflation.

Table 3a.

Real Effects of Stopping Hyperinflations: Germany

(January 1924)

article image
Source: Garber (1982).Note: In Tables 3a3g, the date in parentheses following the country name indicates the month in which the exchange rate stabilized.

Annual data.

Table 3b.

Real Effects of Stopping Hyperinflations: Greece

(February 1946)

article image
Source: Makinen (1984).

Summer 1946.

December 31, 1947.

Table 3c.

Real Effects of Stopping Hyperinflations: Bolivia

(October 1985)

article image
Source: IMF, International Financial Statistics (various issues), and Ministry of Labor.
Table 3d.

Real Effects of Stopping Hyperinflations: Austria

(October 1922)

article image
Source: Wicker (1986).
Table 3e.

Real Effects of Stopping Hyperinflations: Hungary

(August 1946)

article image
Source: Siklos (1989).

Data not available on a regular monthly basis.

Table 3f.

Real Effects of Stopping Hyperinflations: Poland

(February 1924)

article image
Source: Wicker (1986).
Table 3g.

Real Effects of Stopping Hyperinflations: Hungary

(April 1924)

article image
Source: Wicker (1986).

In Greece the sparse available evidence seems to suggest that there were no output costs. In fact, as Table 3b shows, national income increased after the stabilization, and unemployment apparently decreased.23 In Russia, Rostowski (1992) reports that the economy continued to grow strongly after stabilization: industrial output increased by about 30 percent in 1923–24, with higher growth concentrated after stabilization (the reforms took place in February and March 1924), and by about 45 percent in 1924–25.

In two cases—Taiwan and Bolivia—there appear to be no costs associated with stabilization. In Taiwan, Makinen and Woodward (1989) report small, if any, costs after stabilization. The unemployment rate increased only slightly even though Taiwan was forced to absorb nearly a half million mainlanders early in the stabilization. National income kept growing at positive rates, albeit less rapidly than during hyperinflation.24 In Bolivia the stabilization plan in and of itself does not seem to have had much impact on the real sector. As shown in Table 3c, real gross domestic product (GDP) had been falling and unemployment rising since the early 1980s, as a result of internal political chaos and the onset of the debt crisis. There is no apparent reason why this negative trend should have ended in 1985–86. Moreover, the Bolivian economy suffered severe external shocks in late 1985 and 1986: the terms of trade deteriorated substantially as international prices of tin and hydrocarbons collapsed; sales of natural gas to Argentina fell significantly; and large disruptions resulted from a campaign to curtail illegal drug trade. According to Sachs (1986), a conservative estimate would put the export revenue loss at 10–15 percent of gross national product (GNP) in one year. In 1987 real GDP began to increase again, even if at a slow rate.

In two other cases—Austria and Hungary in 1946—there appears to be conflicting evidence. In Austria the number of unemployed increased following the stabilization of September 1922, peaking in February 1923, although it had started to increase even before stabilization. According to estimates provided by Wicker (1986), the unemployment rate peaked in 1926 at 7 percent (Table 3d). Assuming unemployment to be 3 percent in 1921, he concluded that the rise in unemployment of 4 percentage points could be attributed to the stabilization. In contrast, the behavior of real GNP tells a somewhat different story. After increasing by 9 percent in 1922 (the stabilization took place in September), real GNP fell by 1.1 percent in 1923, but rose by 11.7 and 6.8 percent in 1924 and 1925, respectively.25

In Hungary 1946 the unemployment rate began to increase in the first quarter of 1947 (the stabilization took place in August 1946), and reached 11.5 percent in December 1947 (Table 3e).26 In contrast, national income rose by 20 percent in the 12 months following stabilization. Industrial production increased by 9.8 percent in the six months following stabilization. After falling somewhat in the first quarter of 1947, it increased again and, in October 1947, was 35 percent above the prestabilization level. Reinforcing the evidence in favor of very small output costs, studies reported in Siklos (1989) by the Hungarian Institute of Economic Research concluded that in 1946–-47, GDP was only 0.1 percent lower relative to the business cycle peak of 1928–29 and 50 percent higher relative to the business cycle trough year of 1932–33.

In the last two cases—Poland and Hungary in 1924—according to estimates reported by Wicker (1986), the unemployment rate rose substantially, reaching a peak of 12.7 percent in Poland during 1925 (Table 3f) and 14.5 percent in the first quarter of 1925 in Hungary (Table 3g). In Poland, however, even though the number of unemployed increased after the January 1924 stabilization, the level reached in December (159,060) was no worse than before the stabilization (the number of unemployed peaked in January 1922 at 221,444).

In summary, of the nine episodes, available evidence suggests that there were no output costs resulting from monetary stabilization per se in five (Germany, Greece, Russia, Taiwan, and Bolivia); there is conflicting evidence for two (Austria and Hungary in 1946), since both unemployment and economic activity increased; and unemployment increased in two (Poland and Hungary in 1924; output data are not avalable in either case).

In sharp contrast, the real costs of disinflation in industrial countries are estimated to be substantial.27 For instance, according to Gordon (1982, p. 39), in the United States (in the period after 1922), there is “abundant evidence that only 10 to 40 percent of nominal demand changes are absorbed by the inflation rate in the first year after such changes.” This implies that a deceleration in nominal spending growth from 10 percent to zero would result in a fall in inflation of only 1–4 percent and a corresponding drop in real output from 9 percent to 6 percent. Gordon (1982) argued that this basic message holds true for other industrialized countries.

An alternative way of looking at the costs of disinflation is to estimate the “sacrifice ratio,” or cumulative percent output loss per percentage point reduction in inflation. Estimates for the United States lie anywhere from 3 to 18 (Sachs (1985)). The evidence, therefore, seems to warrant the conclusion that hyperinflations have been stopped at relatively small cost compared to low-inflation countries.28 As discussed in Section III, the fact that in hyperinflations backward-looking contracts (so prevalent in industrial and chronic-inflation countries) disappear is probably at the heart of the difference in output costs.

Stopping Chronic Inflations: Stylized Facts

The ten episodes of stabilization in chronic inflation countries that serve as background for the theoretical discussion can be divided as follows.29

  • Latin American heterodox programs of the 1960s. This group comprises stabilization plans in Argentina (March 1967), Brazil (March 1964), and Uruguay (June 1968) (see Tables 57 in the Appendix).30 The common elements of these plans were, first, a fixed exchange rate (with periodic devaluations in Brazil) as the main nominal anchor; and second, the use of incomes policies in varying degrees. Whereas price controls were mainly voluntary in Argentina and Brazil, in Uruguay there was a comprehensive freeze similar to that of the heterodox programs of the mid-1980s. All three programs achieved an initial decline in inflation. However, the reduction in inflation was sustained only in Brazil, lasting well into the 1970s, and reaching 12.7 percent in 1973 (Table 1). In Argentina and Uruguay, loss of fiscal discipline led to a resumption in inflation.

  • Southern cone stabilization programs of the late 1970s. This group comprises stabilizations programs in Argentina (December 1978), Chile (February 1978), and Uruguay (October 1978) (see Tables 810 in the Appendix).31, 32 These programs shared two key characteristics in their design. First, all were orthodox programs (that is, there were no price or wage controls). Second, in all three cases the exchange rate policy consisted in announcing a devaluation schedule (the “tablita,” for “little table”) against the dollar, with a decreasing rate of devaluation. Chile eventually fixed the exchange rate in June 1979. In spite of fiscal balance in both Chile and Uruguay, the slow convergence of inflation and the corresponding real appreciation of the domestic currency proved fatal. All three programs ended in dramatic fashion with large exchange rate and financial crises.

  • Heterodox programs of the mid-1980s. This group comprises the Austral Plan in Argentina (June 1985), the Cruzado Plan in Brazil (February 1986), the Israeli plan (July 1985), and the Mexican plan (December 1987).33 A key common element of the four plans was the use of wage and price controls to counter the inertial component of inflation. The Israeli and Mexican plans were successful in bringing down inflation to about 20 percent. In Argentina and Brazil, however, inflation quickly resumed and soon reached higher levels than before the programs.

The key stylized facts shared by most of these programs are (1) inflation converges only slowly to the rate of devaluation; (2) there is a sustained real appreciation of the domestic currency; (3) the current account and the trade balance deteriorate; and (4) real activity increases at the beginning of the program and later contracts.34

(1) Slow convergence of inflation. To illustrate the slow convergence of the inflation rate to the devaluation rate, Figure 1 shows four-quarter changes (that is, changes over same quarter of previous year) of devaluation and inflation for the ten programs just described.35, 36 The evidence clearly suggests that inflation takes a long time to converge, if at all, to the rate of devaluation. This lack of convergence is particularly striking in the Southern cone tablitas. These programs were based on the belief that the inflation rate would quickly converge to the world inflation plus the preset rate of devaluation. However, to the surprise of policymakers, such convergence was remarkably slow. In fact, as Figure 1 shows, inflation convergence was never achieved. To judge from the three heterodox programs of the 1960s, this slow adjustment was nothing new.

Figure 1.
Figure 1.

Inflation and Devaluation

(Percentage change over same quarter of previous year, monthly rates)

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A007

Source: IMF, International Financial Statistics (various issues).

Given the nonconvergence of inflation in the tablitas, it is hardly surprising that some analysts concluded (for instance, Dornbusch (1982)) that wage and price controls should be part of a stabilization package. Fiscal adjustment was viewed as a necessary, but not sufficient, condition for a successful stabilization attempt due to the inertial components of inflation. As a result, all four major programs of the 1980s resorted to price and wage controls.37 As Figure 1 illustrates, convergence continued to be a problem, although not as severe as before. Even in successful programs such as Israel’s and Mexico’s, inflation has yet to converge to the policy-determined devaluation rate.

(2) Sustained real appreciation of the domestic currency. Given the stow convergence of inflation, it is hardly surprising that the real exchange rate, defined as the relative price of traded goods in terms of nontraded goods, fell during most plans (see Figure 2).38 Sustained real appreciation proved to be fatal for most programs, because it led to unsustainable trade and current account deficits. Failed programs are thus characterized by U-shaped curves for the real exchange rate. The largest appreciations occurred in the tablitas. In both Argentina and Uruguay, the real exchange rate (set to 100 in the quarter before the program was implemented) had halved by the time the programs ended (see Tables 8 and 10 in the Appendix).

Figure 2.
Figure 2.

Real Effective Exchange Rates

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A007

Source: Tables 514.Note: A decline in the index denotes an appreciation.

(3) The current account and the trade balance deteriorate. As a result of the real appreciation, the current account deteriorated, as shown in Figure 3.39 In Chile the current account deficit reached 14.5 of GDP in 1981, and in the Uruguayan tablita, it reached 7 percent of GDP in 1980. Current account deficits were financed by large capital inflows. This is particularly evident in the Southern cone tablitas (see Tables 810). In Chile, for instance, capital inflows reached a peak of US $4,800 million (15 percent of GDP) in 1981. The trade balance also deteriorated substantially (see Tables 514). Large increases in imports of durable goods played a key role in the deterioration of the trade balance (see the evidence reported in Drazen (1990)).

(4) Real activity increases at the beginning of the program and later contracts. The peculiar phenomenon of an initial expansion in an inflation stabilization program—traditionally viewed as generating initial output losses—was brought to the forefront by the Southern cone tablitas of the late 1970s. As illustrated in Figure 4, real private consumption (real GDP for Uruguay) increased sharply in the first stages of the tablita programs.40 Later—and in the cases of Chile and Uruguay even before the programs ended—a sharp contraction in consumption (and output) occurred. With the heterodox programs of the mid-1980s, this issue came alive once again. The case of Israel was particularly striking because the same boom-recession cycle was observed in spite of the success of the program.41 The same pattern characterized the heterodox programs of the 1960s, with the exception of Brazil. Kiguel and Liviatan (1992a) analyzed 12 stabilization plans (which include the 10 depicted in Figure 4) and concluded that the boom-recession cycle in real output remains even when considered relative to the trend.

Figure 3.
Figure 3.

Current Account

(In millions of U.S. dollars)

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A007

Source: IMF, International Financial Statistics (various issues).Note: Data are quarterly, except for Argentina 1967, Brazil 1964, Uruguay 1968, Chilean tablita, and Uruguayan tablita, for which annual data were used. For Israel, the figures correspond to the trade balance. A minus number indicates a deficit.
Figure 4.
Figure 4.

Real Private Consumption

(Percentage change over same quarter of previous year)

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A007

Source: Tables 514.Note: Annual percentage changes for Argentina 1967, Brazil 1964, and Uruguay 1968. For the Uruguayan tablita, figures correspond to real GDP.

II. The Model

This section presents an analytical framework, based on Calvo and Végh (1991a), which will prove useful in interpreting the evidence presented in the previous section. Since the model is described in detail elsewhere, most formal derivations will be bypassed and emphasis will be placed on the results and their intuition.

Consider a small open economy with predetermined exchange rates. There are two (nonstorable) goods: a tradable good, c*, and a nontradable (or home) good, c. The representative consumer maximizes

0[log(ct)+log(ct*)]exp(βt)dt,(1)

where β denotes the constant and positive subjective discount rate.

The consumer is required to use money to carry out purchases. The cash-in-advance constraint is thus

α(ct/et+ct*)=mt,(2)

where α is a positive constant, and e, the real exchange rate, is defined as the relative price of traded goods in terms of home goods; that is, e ≡ EP*/P, where E is the nominal exchange rate (in units of domestic currency per unit of foreign currency), P* is the (constant) price of the traded good in foreign currency, and P is the domestic price of the home good.42 Real money balances in terms of traded goods are denoted by m; that is, m ≡ M/EP*, where M stands for nominal domestic money balances.

The consumer holds an internationally traded bond, b, which bears a constant real interest rate (in terms of traded goods) equal to r. Real financial wealth, a, is thus at = mt + bt. The lifetime constraint faced by the consumer is

a0+0(yt/et+yt*+τt)exp(rt)dt=  0(ct/et+ct*+itmt)exp(rt)dt,(3)

where a0 denotes initial real financial wealth; y and y* stand for output of home and traded goods, respectively; τ denotes real transfers from the government; and i stands for the domestic nominal interest rate. Equation (3) equates the consumer’s lifetime expenditure to his or her lifetime resources. The consumer’s expenditure includes the “rental” cost of real money balances, im.

The consumer’s optimization problem consists in choosing optimal paths of ct, ct* and mt to maximize his or her lifetime utility (equation (1)) subject to the cash-in-advance constraint (equation (2)), and the intertemporal budget constraint (equation (3)), given his or her initial real financial wealth, and the paths of yt, yt*, τt, it, and et. The first-order conditions are (in addition to equations (2) and (3))

1/ct*=λ(1+αit),(4)

and

ct=etct*,(5)

where λ is the (time-invariant) Lagrange multiplier associated with constraint (3), which can be interpreted as the marginal utility of wealth.43 Equation (4) is the familiar condition whereby the consumer equates the marginal utility of consumption of traded goods to the marginal utility of wealth times the “price” of traded goods. In the present context, the relevant price of the traded good—which will be referred to as the effective price—consists of the market price (unity) plus the opportunity cost of holding the a units of real money balances that are necessary to purchase the good, αi. Equation (5) indicates that the consumer equates the marginal rate of substitution between traded and home goods to the relative price of traded goods in terms of home goods (that is, the real exchange rate).

Perfect capital mobility implies that

it=r+ϵt,  (6)

so that the nominal interest falls, one-to-one, with the rate of devaluation.

Consider now the supply side of the economy. It will be assumed that the economy is endowed with a constant flow endowment of the tradable good, y*. The supply of the home good is demand determined and follows the staggered-prices model of Calvo (1983).44 Calvo (1983) showed that the rate of change of the inflation rate is negatively related to excess demand. Formally

π.t=θDt,(7)

where π(≡ Ṗ/P) is the rate of inflation of home goods; and D, a measure of excess demand in the home goods market, is defined as Dt = ytyt where y can be interpreted as the full-employment level of output. Equation (7) can be derived in a framework in which firms set prices in a nonsynchronous manner, taking into account the expected future path of excess demand and of the average price prevailing in the economy. Since only a small number of firms may change their individual prices at any point in time, the price level is a predetermined variable; but inflation is free to jump, because it reflects changes in individual prices charged by firms. When higher demand emerges, some firms increase their individual prices, and thus, inflation rises. Over time, the proportion of firms that have yet to respond to excess demand declines, so that inflation falls over time. Hence, the change in inflation is a negative function of aggregate demand, as equation (7) indicates.

Imposing equilibrium in the home goods market (that is, ct = yt) and using equation (5) and the definition of excess aggregate demand, equation (7) can be rewritten as

π.t=θ(yetct*).(8)

Differentiating e = EP*/P with respect to time yields (recalling that P* is assumed constant)

e.t=(εtπt)et.(9)

For given paths of c* and the policy variable ε, equations (8) and (9) form a dynamic system for π and e.45

To close the model, aggregate resource constraints need to be considered. Assuming that the government transfers back to the public interest income on net foreign assets and revenues from money creation, it can be shown (see Calvo and Végh (1991a)) that the economy’s resource constraint is

k0+y*/r=0ct*exp(rt)dt(10)

where k0 denotes the economy’s initial stock of foreign bonds. Equation (10) states that the present value of tradable resources must equal the present value of consumption of traded goods. Under the assumption that domestic credit just compensates the consumer for the depreciation of nominal money balances, the current account is given by

kt=y*+rktct*,(11)

which indicates that the current account balance is the difference between traded goods income and consumption of traded goods.

Permanent Reduction in Devaluation Rate

Consider now a permanent reduction in the rate of devaluation. Specifically, suppose that at time 0 (the present), policymakers announce that the rate of devaluation will be reduced immediately from εh to εl. Moreover, the announcement is fully credible; that is, the public expects the rate of devaluation to remain at the lower level, εl, forever.

At time 0, the nominal interest falls by the same amount as the rate of devaluation, as indicated by equation (6). Since the policy is fully credible, the nominal interest rate is expected to remain at the lower level, r + εl forever. This implies that the consumption of traded goods does not change. The reason is that a constant nominal interest rate, no matter what the level is, implies, by first-order condition (4), that consumption of traded goods is constant over time. Even if the effective price of consumption is reduced, the fact that it remains constant over time implies that there are no incentives to engage in intertemporal consumption substitution. Since tradable resources do not change, consumption of traded goods must remain at the same level.

From the system (8) and (9), it follows that, given that c* is not affected by permanent changes in the rate of devaluation, a fall in π that exactly matches the fall in ε immediately moves the system to a new steady state. Naturally, the (average) inflation rate of the economy, which is a weighted average of the inflation rate of home goods, π, and that of traded goods, ε, also falls instantaneously to its new level, εl. Therefore, an exchange rate-based stabilization program that is fully credible reduces the inflation rate instantaneously at no real costs.

It is worth emphasizing that, even though the price level is sticky and individual prices are set in a staggered manner, no real effects result from a reduction in the devaluation rate that is perceived as being permanent. This shows, as emphasized by Calvo and Végh (1991a), that price level rigidity does not, by itself, imply stickiness in the inflation rate. The reason is that in this framework, firms act in a forward-looking manner.46

III. Stopping Hyperinflation

Despite its simplicity, the exercise just undertaken analyzing the effects of a permanent reduction in the rate of devaluation provides a useful conceptual framework for discussing the end of hyperinflations. The reason, as this section argues, is that the model captures two distinguishing characteristics of hyperinflations: the absence of backward-looking behavior, and the presence of a high degree of credibility. Hence, it seems reasonable to identify, if only as a first approximation, the above analytical exercise with experiences of stopping hyperinflation.

Absence of Backward-Looking Behavior

The disappearance of long-term nominal contracts in hyperinflation episodes is a recurrent theme in the literature (see, for instance, Cagan (1989)). Furthermore, a key characteristic of hyperinflations is that at some point all prices become indexed to the nominal exchange rate. Wage contracts, for instance, are renegotiated more frequently as inflation accelerates. At first, wage readjustments are based on a cost of living index. As the interval between readjustments becomes shorter, however, the cost of living index must be replaced by another index that is available on a weekly or even daily basis. The quotation of a foreign currency—usually the U.S. dollar—provides such an index; the dollar quotation is available on a continuous basis and is widely circulated.

The de facto indexation of all prices in the economy to the foreign exchange implies that nominal contracts virtually cease to exist. Thus, all backward-looking behavior is eradicated from the economy. Since one of the key features of the model presented in Section II is the absence of backward-looking behavior, the model can be taken to apply to hyperinflation episodes.

High Credibility

It has been argued (see, for instance, Kiguel and Liviatan (1988)) that two characteristics of hyperinflationary processes make a stabilization attempt more credible than attempts to stop chronic inflation.

First, the need for seigniorage (that is, revenues from money creation) as the cause of high inflation comes across more clearly in hyperinflations than in chronic inflations. Since the fiscal nature of the inflationary process is more obvious, the public may become more easily convinced that closing the budget deficit is enough to ensure price stability. In contrast, in chronic-inflation countries, the relationship between inflation and revenues from money creation is much less clear, which may raise doubts in peoples’ minds as to whether a fiscal reform—even if successfully implemented—may be enough to halt inflation. Thus, given that in hyperinflations the source of the inflationary process is easily identified and widely agreed upon, the announcement of a stabilization program that includes a fiscal reform should command a high degree of credibility.

The second factor that may increase the credibility of a stabilization program is that hyperinflation creates such a chaotic social and economic environment that the public becomes convinced that the situation is untenable. This sense of urgency in tackling the problem is likely to lend more credibility to a stabilization program. In contrast, chronic-inflation countries learn how to live with high inflation by adopting various indexation mechanisms. An example of this ability to adapt to chronic inflation can be observed in the behavior of real revenues from taxation. In hyperinflations, the Olivera-Tanzi effect drastically reduces real revenues from taxation; this is not the case in chronic inflations (see Kiguel and Liviatan (1988)).

There is also more direct evidence on the presence of high credibility in hyperinflation stabilizations. As emphasized by Sargent (1982), the real monetization of the economies in which hyperinflation was successfully stopped is dramatic.47 In the aftermath of stabilization, monthly growth rates of notes in circulation above 10 percent alongside a basically stable price level have been observed in many episodes (Germany, Hungary 1946, Greece, and Poland), as shown in Table 3. In Hungary 1946, for instance, notes in circulation increased by a factor of 4.9 in the 12 months following stabilization.48 Such enormous increases in real money balances can only be attributed to a drastic reduction in inflationary expectations, as a result of the high credibility of the program.

Matching Theory with Evidence

It has been argued that backward-looking behavior all but disappears during hyperinflations and that attempts to stop hyperinflations enjoy-high credibility. Therefore, the analytical experiment undertaken in Section II, which assumes full credibility in the context of a model in which there is no backward-looking behavior, may be considered a reasonable first approximation to understanding hyperinflation stabilization.

The next step is to match theory with evidence. If full credibility is assumed, the model predicts that stabilizing the nominal exchange rate should lead to an immediate halt in inflation at no real cost. This prediction is in accordance with the evidence provided in Section I: price stability has been achieved overnight with relatively small output costs. The costs that did emerge are more attributable to real dislocations—sometimes policy induced, such as Germany’s policy of subsidizing the capital goods industry with the inflation tax—than to the monetary stabilization itself.

Admittedly, the model is a highly stylized version of the real world and ignores many important aspects, in particular supply-side and structural considerations that have played an important role in past hyperinflations. But its usefulness lies precisely in its simplicity, because it isolates what is regarded as key aspects in actual episodes. Furthermore, by enriching the formulation of the supply side, the same model could be used to tackle real sector considerations.

IV. Stopping Chronic Inflation

This section will interpret the evidence presented in Section I on stopping chronic inflation in terms of the analytical model developed in Section II.49 The key assumption behind the analytical exercise developed in this section is that there is lack of credibility. Unlike programs designed to stop hyperinflation, stabilization attempts in chronic-inflation countries are likely to suffer from lack of credibility. Chronic-inflation countries have learned to live with high inflation, and as a result, the incentives to eradicate inflation are much lower than in hyperinflations. Furthermore, given past failures, policymakers will have a hard time convincing a skeptical public that the current stabilization plan will be sustained over time.

The spreads between domestic and devaluation-adjusted foreign interest rates provide at least a crude measure of credibility. Large positive spreads suggest lack of credibility, since they indicate that the public expected a devaluation (under a fixed exchange rate) or a larger devaluation than the one preannounced (in a tablita program). Tables 814 in the Appendix show the presence of large spreads throughout the programs, except, as one would expect, in the quarters preceding a non-scheduled devaluation.50 In the Argentine tablita (Table 8), for instance, the spread reaches its maximum of 45.7 percent (in annual terms) in the third quarter of 1980. In the following quarter the spread becomes negative, reflecting the devaluation that finally ended the program. In Mexico (Table 14), the spread has been positive since the beginning of the program. It declined, however, during 1989 as the authorities switched from a fixed exchange rate to a constant (in absolute terms) daily devaluation.

Lack of credibility will be modeled as temporary policy. Suppose that at time 0 policymakers announce a permanent reduction in the devaluation rate, but the public believes that the stabilization will be abandoned at some time T in the future. Under this scenario, the public acts as though the reduction in the devaluation rate were temporary. Hence, the effects of temporary policy may be reinterpreted as arising from noncredible policy. Formally, then, consider a temporary reduction in the rate of devaluation. Suppose that initially (that is, prior to time 0), the rate of devaluation is ϵh. At time 0, the expected path becomes

εt=εl,for0t<T(12a)
εt=εh,for  tT,(12b)

where T > 0 and εh > εl. The rate of devaluation falls at time 0 but is expected to increase back to its original level at time T.

The lower rate of devaluation during the period [0,T)—hereafter referred to as the “transition”—implies, by the assumption of perfect capital mobility (equation (6)), that the path of the nominal interest rate is given by it = r + εl, for 0 ≤ t < T, and it = r + ϵh, for tT. Since the nominal interest rate is time invariant (that is, its time derivative is zero), the first-order condition (equation (4)) indicates that consumption of traded goods is time invariant, even though its level may change. Because the nominal interest rate is lower during the transition than it is after T, equation (4) shows that consumption of traded goods is higher during the transition than afterwards (recall that λ does not change at T). The reason is that the effective price of consumption is lower during the transition. Since the resource constraint (equation (10)) must be satisfied for any equilibrium path, consumption of traded goods during the transition will be above initial permanent income of traded goods, while consumption of traded goods after T will be below initial permanent income of traded goods (see Figure 5, panel A).51 Otherwise, the resource constraint would be violated.

Figure 5.
Figure 5.

Temporary Reduction in Devaluation Rate

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A007

The current account path, which follows from the consumption path just described and equation (11), is illustrated in Figure 5, panel B. At time 0, the current account jumps into deficit because of the sudden increase in consumption of traded goods. During the transition (that is, for 0 ≤ t < T), the trade deficit remains constant but the current account deteriorates because interest income on net foreign assets declines. At time T, the current account jumps into balance. As discussed in Section I, the current account has deteriorated substantially throughout the different programs (see Figure 3). The current account deficit usually reflects sizable trade deficits (see Tables 514 in the Appendix).

Figure 5, panel C, illustrates the time path of inflation of home goods.52 On impact, there are two effects on the inflation rate of home goods that go in opposite direction. On the one hand, the lower rate of devaluation exerts a dampening effect on inflation of home goods. On the other hand, aggregate demand increases (see below), which tends to push inflation up. In the absence of the aggregate demand effect, inflation of home goods would fall, one-to-one, with the rate of devaluation (which is what happens under full credibility). However, because of the aggregate demand effect, the rate of inflation of home goods falls by less than the rate of devaluation, and could even increase. Inflation falls over time at the beginning of the program but later begins to rise in anticipation of the expected resumption of the higher devaluation rate.53 When time T is reached, the authorities face the choice of abandoning the program (thus validating the public’s expectations) or sticking to the lower rate of devaluation. If the authorities abandon the program, inflation continues to increase toward its initial level, as in Figure 5, panel C. If the authorities stick to the stabilization plan, then inflation jumps downwards at time T and converges from below to the lower devaluation rate.

The (average) inflation rate of the economy is a weighted average of the inflation rate of traded goods, ε, and the rate of inflation of home goods, π with the weights depending on the weight of traded and home goods in the utility function. Since the devaluation rate is constant, the inflation rate exhibits the same qualitative behavior during the transition as inflation of home goods. The model thus predicts that, due to lack of credibility, the inflation rate exhibits inertia. The U-shaped time profile of inflation predicted by the model is typical of that observed in failed programs (which include all the programs depicted in Figure 1, except Israel and Mexico).54 The behavior of inflation in successful plans, such as Israel and Mexico, is consistent with the assumption that policymakers stuck to the stabilization plan.

Panel D in Figure 5 illustrates the behavior of the real exchange rate.55 The real exchange rate falls during the transition (that is, there is a real appreciation of the domestic currency), because the inflation rate of home goods is always above the rate of devaluation. At time T, whether or not the plan is abandoned, the real exchange rate begins to increase (that is, there is a real depreciation).56 If the plan is not abandoned, inflation falls below the devaluation rate to generate the real depreciation. This U-shaped time profile of the real exchange rate predicted by the model is the one that has generally been observed in practice (see Figure 2).

Consider now the time path of consumption of home goods illustrated in Figure 5, panel E. On impact, as already noted, consumption of home goods increases as follows from equation (5) and the increase in c*. Intuitively, since the relative price of home goods in terms of traded goods has not changed, there are no incentives to change consumption of home goods relative to consumption of traded goods. During the transition, consumption of home goods falls as the real exchange rate decreases (that is, the relative price of home goods increases). If T is large enough, the economy enters into recession (that is, output falls below its full-employment level) before the plan is expected to be discontinued. For a small T, output will remain above its full-employment level during all of the transition. At time T, consumption of home goods jumps downwards, accompanying the fall in consumption of traded goods. After T, the real exchange rate increases, and therefore consumption of home goods rises.

The evidence depicted in Figure 4 for private consumption is consistent with the predictions of the model, in that there is an early consumption boom followed by a contraction. Moreover, in two of the longest programs (Uruguay, 17 quarters; and Chile, 18 quarters), consumption fell before the end of the programs. In Uruguay, as Table 10 indicates, the annual rate of change of consumption was negative in 1980 and the four-quarter change in real GDP turned negative in the last quarter of 1981, one year before the end of the program. In Chile, the four-quarter change in both private consumption and real GDP turned negative in the fourth quarter of 1981. two quarters before the end of the program. If one identifies the actual duration of the program with T, this fact is consistent with the prediction that the higher is T, the more likely it is that the contraction will take place before the program ends. Even if the program is not abandoned at T, the behavior of real variables remains the same, as argued above. Hence, the model is also capable of rationalizing consumption patterns such as Israel’s, where the recession took place in spite of the success of the program.

Consider now the time path of the domestic real interest rate (that is. rd = i – π), which is illustrated in Figure 5, panel F, The inflation rate of home goods falls by less on impact than the nominal interest rate does; therefore, the domestic real interest rate falls on impact. During the transition, the domestic real interest rate stays below its initial level. At time T, the domestic real interest rate jumps upwards, owing to the sudden increase in the nominal interest rate. The domestic real interest rate falls afterwards toward its unchanged steady-state value, given that inflation increases.

The predictions of the model with respect to the behavior of the real interest rate seem to hold for the Southern cone programs but not for the rest of the programs.57 The Uruguayan tablita (Table 10) is the best example: the (ex post) real lending rate fell substantially when the program was implemented and was negative during the first three quarters of 1979. In 1980 it became positive and reached a peak of 48.2 percent (in annual terms) in the fourth quarter of 1981, undoubtedly contributing to the contraction in economic activity. In the Argentine tablita (Table 8), the real lending rate was also negative in the first two quarters of the program (even if it increased with respect to the fourth quarter of 1978). In the Chilean tablita (Table 9), the real lending rate fell from very high levels, reaching 5.8 percent, six quarters into the program.

In the rest of the programs for which data are available (the four heterodox programs of the mid-1980s), real interest rates increased (see Tables 1114 in the Appendix). A possible explanation for this phenomenon lies in the use of additional nominal anchors. The Israeli 1985 plan, for instance, had an explicit target for bank credit, which was to be achieved by a combination of higher reserve requirements and a higher discount rate (see Barkai (1990)). The idea was to offset the expansionary effects of a large increase in credit to the private sector. In fact, credit to the private sector declined in real terms at the beginning of the program, which may explain the initial (although brief) downturn in the last two quarters of 1985. In contrast, in the Southern cone tablitas, real credit grew strongly. Analytically, high interest rates can be rationalized by modeling the use of additional monetary anchors (see Calvo and Végh (1991b)).

In summary, a noncredible, exchange rate-based stabilization yields predictions that are generally consistent with the stylized facts. In particular, the model generates a boom-recession cycle and U-shaped curves for both inflation and the real exchange rate, as has been observed in most programs. The key missing ingredient in explaining high real interest rates at the beginning of the program seems to be the presence of additional nominal anchors.

V. Final Remarks

This last section discusses the quantitative importance of lack of credibility (or the “temporariness” hypothesis) in explaining the observed consumption booms and the role of backward indexation, and then offers some concluding remarks.58

Quantitative Relevance of the Temporariness Hypothesis

A crucial, and relatively unexplored, question is the quantitative relevance of the temporariness hypothesis. Given that intertemporal elasticities of substitution are generally low, it is not clear whether models that rely on intertemporal substitution effects can explain the magnitude of observed rises in consumption. Since what triggers all effects in the model is the temporary (as perceived by the public) fall in the nominal interest rate relevant for consumption decisions, a first question is whether nominal deposit interest rates (that is, the interest rate relevant for consumption decisions) have fallen in different programs. Figure 6 shows that in the four heterodox programs of the mid-1980s (Austral, Cruzado, Israeli, and Mexican), deposit rates did indeed fall dramatically.59 In the Southern cone tablitas, however, the fall was considerably less, particularly in Uruguay where the deposit rate soon reached its initial level.

Figure 6.
Figure 6.

Nominal Deposit Interest Rates

(In percent per year)

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A007

Source: Tables 814.

The empirical question is therefore: can the observed fall in nominal interest rates account for a sizable fraction of the rise in private consumption in a temporariness model? A first attempt to answer this question can be found in Reinhart and Végh (1992). Using estimates of the intertemporal elasticities of their own and other available estimates, they concluded, based on a very simple simulation exercise, that the temporariness hypothesis accounted for an important fraction (over 60 percent) of the actual increase in consumption in the four heterodox programs of the 1980s.60 In contrast, it only accounts for less than 15 percent in the tablitas. The key difference, as discussed above, is that in the four heterodox programs, nominal interest rates dropped sharply, whereas in the Southern cone tablitas, the fall was considerably less. An important caveat is that durable goods, which are likely to play a crucial role in any empirical application, are not modeled. Elasticities of substitution are bound to be higher if a distinction is made between durable and nondurable goods, as found by Fauvel and Sampson (1991) for Canada.61 Therefore, although further work is warranted, Reinhart and Végh’s (1992) results suggest that some other factors (for instance, negative or falling real interest rates) may have played an important role in the tablitas.

Backward Indexation

In the model presented in Section II, inflation “stickiness” results from lack of credibility. Alternatively, inflation inertia could result from backward-looking behavior.62 In this spirit, Rodriguez’s (1982) assumed adaptive expectations and showed how, due to an initial fall in real interest rates, there would be a consumption boom. In a setting of rational expectations, Calvo and Vegh (1992) incorporated backward-looking contracts in a model similar to that of Section II and, in contrast to Rodriguez (1992), concluded that a fully credible, exchange rate-based stabilization will lead to a recession if the intertemporal elasticity of substitution is smaller than the elasticity of substitution between traded and home goods (which seems to be the relevant case in practice; see Ostry and Reinhart (1992; this issue)). Intuitively, the real appreciation of the domestic currency has both expansionary effects—since it reduces the domestic real interest rate—and recessionary effects—since it increases the relative price of home goods. Hence, backward-looking contracts, per se, may not be capable of explaining the observed expansion. If, in addition, there is lack of credibility, then the net effect on consumption will still depend on the configuration parameter, but an increase in consumption becomes more likely since the temporariness effect is always expansionary. Hence, the key ingredient in explaining a consumption boom appears to be lack of credibility and not backward indexation.

Conclusions

A unified theoretical framework has been used to interpret the main stylized facts associated with stopping both hyperinflation and chronic inflation. The model predicts that, in the absence of backward indexation, a credible stabilization program stops inflation in its tracks with no real effects. This experiment was taken as a reasonable first approximation to a hyperinflationary situation. Although the evidence regarding the output costs of stopping hyperinflation is spotty and subject to debate, a good case can be made that hyperinflation has, in fact, been stopped suddenly with no major costs. The contractionary forces that have come into play seem to result from real distortions brought about by hyperinflation rather than by monetary stabilization itself.

A noncredible reduction in the rate of devaluation has been used to interpret stabilization in chronic-inflation countries. The model predicts a slow convergence of the inflation rate, a boom-recession cycle in the home goods sector, a current account deficit, and real appreciation of the domestic currency. This is consistent with the key stylized facts observed in these episodes. The model also predicts a fall in domestic real interest rates on implementation of the program. This prediction is consistent with the experience of the Southern cone tablitas, but not with that of the major heterodox plans of the mid-1980s.

The model used in this paper, based on Calvo and Végh (1991a). appears to offer a reasonably good description of reality and should prove useful as a benchmark for interpreting the outcome of stabilization programs in high-inflation countries. Naturally, many issues remain to be further analyzed, but three seem to stand out.

First, the behavior of real interest rates in the aftermath of stabilization programs deserves further attention, from an analytical point of view. Contrary to common perceptions, lack of credibility does not necessarily generate high real interest rates. As mentioned above, additional nominal anchors are likely to be a key ingredient. In this respect, the issue of tight credit policy during stabilization policy, an issue raised by Barkai (1990) for the case of Israel, seems crucial.

Second, the role of credibility needs to be endogenized. The model analyzed in the paper takes the existence of credibility, or the lack thereof, as exogenous to the model. As long as credibility has an exogenous component—given, for instance, by past experiences—the essence of the model should not change. Common sense suggests, however, that credibility is gained or lost as a program unfolds, since the evolution of the different variables (for instance, the real exchange rate) provides valuable information regarding the sustainability of the program.63 Casual evidence suggests that credibility often follows an inverted U-shaped pattern, rising at first and decreasing later. This credibility pattern has been found in the Cruzado Plan (Agénor and Taylor (1992; this issue)). Guidotti and Végh (1992) developed a model of endogenous credibility in which this pattern arises.

Finally, it would be important to model supply-side considerations that may come into play in periods of high inflation, and study the interaction between real factors and monetary stabilization. This is bound to be particularly useful in understanding stabilization programs in Eastern Europe and the former Soviet Union where high inflation and structural changes are likely to coexist in the foreseeable future.

APPENDIX Statistical Appendix

This Appendix contains tables with selected macroeconomic indicators for eleven stabilization programs in high-inflation countries; Bolivia 1985 (Table 4); Argentina 1967 (Table 5); Brazil 1964 (Table 6); Uruguay 1968 (Table 7); Argentina 1978 (Table 8); Chile 1978 (Table 9); Uruguay 1978 (Table 10); Argentina 1985, Austral Plan, (Table 11); Brazil 1986, Cruzado Plan (Table 12); Israel 1985 (Table 13); and Mexico 1987 (Table 14).

Subject to data availability, the same variables are reported for each episode to facilitate comparisons across different episodes. Four-quarter changes indicate percentage change over same quarter of the previous year.

Table 4.

Bolivian 1985 Stabilization

article image
article image
Sources: IMF, International Financial Statistics (various issues); except columns (2) and (13), from Central Bank of Bolivia, and column (17), from Central Bank of Bolivia, SAFCO, and Fund estimates.Note: All rates of change and interest rates are expressed in percent per year. Horizontal line indicates the beginning of the program. The spread between the parallel and the official rate has been negligible since 1987. Columns (11), (12), (13), (15), (16), and (17) report annual figures; column (7) is computed using the one-quarter-ahead inflation rate; column (8) is defined as the ratio of (1 plus) the deposit rate to (1 plus) the three-month LIBOR rate times (1 plus) the one-quarter-ahead devaluation rate; column (16) includes errors and omissions (a positive sign indicates a capital inflow); and column (17) refers to the balance of the nonfinancial public sector (a minus sign denotes a deficit).
Table 5.

Argentine 1967 Stabilization

article image
article image
Sources: IMF, International Financial Statistics (various issues); except columns (6) and (8), from Fundación Mediterránea; column (7), from Di Telia and Dornbusch (1989); column (9), from Kiguel and Liviatan (1989); and column (12), from Di Telia (1983).Note: All rates of change and interest rates are expressed in percent per year. Horizontal lines indicate the beginning and end of the program. All columns except (1), (2), and (3) report annual figures; column (11) includes errors and omissions (a positive sign indicates a capital inflow); and column (12) refers to the nonfinancial public sector (a minus sign denotes a deficit).
Table 6.

Brazilian 1964 Stabilization

article image
article image
Sources: IMF, International Financial Statistics (various issues); except columns (8) and (9), from Kiguel and Liviatan (1989); and column (12), from Lemgruber (1977).Note: All rates of change are expressed in percent per year. Horizontal lines indicate the beginning and end of the program. Columns (6) through (12) report annual figures; column (11) includes errors and omissions (a positive sign indicates a capital inflow); and column (12) refers to the operational balance of the nonfinancial public sector (a minus sign denotes a deficit).
Table 7.

Uruguayan 1968 Stabilization

article image
article image
Sources: IMF, International Financial Statistics (various issues); except columns (6), (7), and (12), from Viana (1990); and columns (8) and (9), from Kiguel and Liviatan (1989).Note: All rates of change are expressed in percent per year. Horizontal lines indicate the beginning and end of the program. Columns (6) through (9), (11), and (12) report annual figures; column (11) includes errors and omissions (a positive sign indicates a capital inflow); and column (12) reports figures for the nonfinancial public sector (a minus sign denotes a deficit).