The Modern Hyperinflation Cycle: Some New Empirical Regularities
Author: Jose Saboin

Contributor Notes

Author’s E-Mail Address: jsaboin@imf.org

Using a database of up to 62 variables for 196 countries over 57 years, a hyperinflation cycle has been characterized to propose a broader setting of stylized facts. Beyond the usual facts, the findings in this paper contribute to the literature of modern hyperinflations in that these cycles occur in contexts where there are (i) depressed economic freedoms, (ii) deteriorated socioeconomic conditions and rule of law, as well as (iii) high levels of domestic conflictivity and government instability. Despite social infraestructure factors improve during stabilization, they keep being substantially lower than the respresentative non-hyperinflation country, suggesting an important role for them in the occurrence of modern hypeinflations. Finally, the role of international financial assistance in stabilization was studied, noting that (i) a clear majority of hyperinflation countries used it, further improving their (ii) economic freedoms, and allowing themselves (iii) greater fiscal flexibility and (iv) more exchange rate stability.

Abstract

Using a database of up to 62 variables for 196 countries over 57 years, a hyperinflation cycle has been characterized to propose a broader setting of stylized facts. Beyond the usual facts, the findings in this paper contribute to the literature of modern hyperinflations in that these cycles occur in contexts where there are (i) depressed economic freedoms, (ii) deteriorated socioeconomic conditions and rule of law, as well as (iii) high levels of domestic conflictivity and government instability. Despite social infraestructure factors improve during stabilization, they keep being substantially lower than the respresentative non-hyperinflation country, suggesting an important role for them in the occurrence of modern hypeinflations. Finally, the role of international financial assistance in stabilization was studied, noting that (i) a clear majority of hyperinflation countries used it, further improving their (ii) economic freedoms, and allowing themselves (iii) greater fiscal flexibility and (iv) more exchange rate stability.

I. Motivation and Introduction

Just when economists thought that hyperinflation was something from the past, it has aroused again: since November 2017, Venezuela has joined the hyperinflation1 club. At the time of writing these lines (September 2018) year-on-year inflation has reached 488,865 percent.2 This exponential increase in prices has been accompanied by a massive contraction of economic activity, although this fall had begun before the hyperinflation process initiated.3 The social implications of this economic collapse have not been different: poverty indexes have increased, eradicated diseases have proliferated4 and, for the years 2017 and 2018, emigration has been the highest recorded worldwide, according to the United Nations. Until today, the Venezuelan society has not envisioned a solution to these problems.

Given these circumstances, other societies might wish to know (or perhaps remember) how they can get into these situations, but also, and more importantly, how to avoid them. In this sense, the present work intends to revive the discussion about the empirical regularities of what in this work is called the hyperinflation cycle. This way, although this research wants to be exhaustive in comparing the behavior of a long list of economic variables (62, to be exact) in and out the cycle, it does not seek to spin finely on the interactions between them, nor the mechanisms of causation among them, nor to discriminate among them. That is, the work is a starting point to observe the generalities of these processes to then advance in depth to their analysis.

The results of the paper indicate that the hyperinflation cycle seems to be a phenomenon that, more than regional, occurs in economies with high presence of natural resource rents (and potentially higher State intervention in the economy) and where economic freedoms have been diminished, especially those related to property rights and the ease of doing business and economic exchange. The cycles also coincide with contexts in which socioeconomic conditions such as employment and real wages deteriorate, where the rule of law and democratic accountability are subdued, the instability of the government increases and there is a greater presence of military personnel over political issues. Although external factors matter, their role may not have been as important as domestic factors, especially those inherent to economic policy: high fiscal deficits, in some cases financed with external debt at the beginning and with seigniorage afterwards; inability to maintain a certain exchange rate regime and restrictions on transactions in the financial account of the balance of payments are the variables whose performance coincides with the advent of a hyperinflation cycle.

The results also suggest that hyperinflation cycles end when there are improvements on three essential fronts: (a) the fiscal and monetary mix: fiscal accounts are closer to equilibrium and base money growth decreases substantially; (b) the interaction with the external sector: barriers to international trade diminish and the exp/imp capacity of the economy increases, the burden of foreign debt on exports regularizes, the resounding level of devaluation of the currency is stopped, hand in hand with less variability of the foreign exchange rate; (c) the structural factors: economic freedoms increase and there is improvement government stability and quality.

The paper continues in the following way: the next sections briefly comment the previous literature and the challenges of studying hyperinflation; section III explains the definition of the hyperinflation cycle and shows the dynamics of each phase; section IV talks about data and methodology; section V discusses the empirical findings for each block of variables. Section VI concludes and sections VII and VIII respectively present references and the appendix.

II. THE CHALLENGES OF STUDYING HYPERINFLATION

Hyperinflation is an uncommon phenomenon. In fact, in what can already be considered a classic in the economic literature, Fischer, Sahay and Végh (2002) —from here on Fischer et. al— examined the main characteristics of hyper and high modern inflations. In their work, authors emphasize that, since the end of the Second World War and until 1996, hyperinflations, despite being a modern phenomenon, had been rare episodes and, rather, very high inflation processes —of about 100 percent— had been much more common.

Regarding the subject that competes this research, hyperinflation, these authors show that in the world there were no hyperinflations during the period 1947-1984. From 1984 until 1996, only 13 countries had experienced this phenomenon. Of which, four were Latin American countries (Argentina, Bolivia, Brazil and Nicaragua), two Africans (Angola and the Democratic Republic of the Congo) and six transition economies (Armenia, Azerbaijan, Georgia, Tajikistan, Turkmenistan, Serbia and Ukraine).5

Although from 1984 to 1996 hyperinflation episodes occurred almost annually, from there the occurrence has been minimal: only 3 countries have registered hyperinflation in the last 20 years. These countries are: Bulgaria (1997) —using the Cagan criterion; Zimbabwe (2007) and currently Venezuela (2017) —using the criterion of Fischer et. al.

The fact that hyperinflation is uncommon has important repercussions for its analysis. First, the infrequency of hyperinflations implies unusual levels of granularity and heterogeneity in the data, which implies high probability of leading to diversity of criteria in their definition, which in turn has the potential to create biased or impractical analyzes of the phenomenon in general, but also in particular.

An example of the above can be seen in the difference between the definitions of Cagan, Fischer et. al, and the rest.6 While according to the definition of Cagan, hyperinflations of one and two months should be recorded as hyperinflation cases, the definition of Fischer et. al does not include them. This does not mean, however, that one definition is better than the other, but that both conform to the facts of study by their authors. This indeed happens with the definition most used in the literature —and this is what Cagan recognizes with its definition of hyperinflation— since it is based on the empirical regularities found in the European episodes during the interwar period. However, it is remarkable that the Cagan criterion, when applied to other episodes, can identify regularities like those found by this author.

The other repercussion of the rarity of hyperinflations is related to the statistical difficulties of small sample inference, especially those that attempt against the asymptotic properties of the parameters, as well as to unobserved heterogeneity. The combination of two limitations: short duration and low occurrence, presents important obstacles for statistical inference using standard regression methods. It is paradoxical, however, that the empirical regularities of hyperinflation make it one of the phenomena that has contributed the most to the field of macroeconomics.7

In addition to the above, another implication for the analysis of hyperinflations is associated to the lack of availability of high frequency data: while inflation data can be found at a monthly frequency, the data of other variables, like other macroeconomic variables (such as fiscal and external accounts), but especially structural variables and those of public policies (such as access to international financial assistance programs and changes in economic freedoms), whose frequency is usually annual. Although this is not a rigid impediment, since the variables can be aggregated (disaggregated), this process generally makes the estimates less accurate and, therefore, with potential to be biased.8

After the study of Fischer et. al in 2002, technological advances and institutional capacity has allowed researchers to apply new methodologies and incorporate new data, generating new approaches whose results have broaden the analysis of different phenomena. In this sense, the present work seeks to enrich the hyperinflation literature, not only by incorporating new variables, but by using non-parametric methods that, despite being simple, do not stop generating relevant information —as they reduce biases— and that, by being simple, are easy to understand to non-specialized audiences.

III. Stylized Facts: The Dynamics of the Hyperinflation Cycle

This study aims to observe the behavior of different variables during what has been termed in this paper as the hyperinflation cycle and in this way infer (i) under what conditions hyperinflation episodes have begun, (ii) what happens during these episodes, (iii) under what conditions the hyperinflation episode could probably end and (iv) what authorities could do to increase the probability that a disinflation process will be carried out successfully.

The hyperinflation cycle comprises two major stages (rise and fall), which in turn can be divided into two sub-periods each, taking a total of 4 phases. The first stage, the rise, is composed of (i) the phase prior to the hyperinflation period, which has been termed extraordinary acceleration phase or the path towards hyperinflation, defined as the period in which the annual average inflation rate is higher or equal to 50 percent, but less than 500 percent; and (ii) the hyperinflation phase, which is defined as the period in which it is greater than or equal to 500 percent. The second stage, which has been designated as the fall, is composed of (iii) the disinflation phase, defined as the period in which average annual inflation is less than 500 percent, but greater than or equal to 50 percent and (iv) the stabilization phase, which is defined as the period in which it falls below 50 percent and which remains below this threshold for a minimum period of five years. The duration of each hyperinflation cycle is the number of years in which inflation is within these thresholds, allowing the existence of deviations within each threshold (gaps) for periods less than or equal to five years.

The reason why the thresholds of 50 and 500 percent per year have been chosen is inspired in different methodologies applied in previous works, such as those of Dornbusch and Fischer (1993), Bruno and Easterly (1995) and Fischer et. al (2002). However, as mentioned in section I, the definition of a hyperinflation episode always contains some traces of arbitrariness.

In this sense, considering that various adjectives have been used throughout the literature to categorize inflationary episodes and that a single denomination still has not been agreed upon, the extraordinary acceleration phase is thus named due to the fact —found by Fischer et. al— that once annual inflation has exceeded the 50 percent threshold, the probability of it falling or remaining the following year is significantly less than half (33.6 percent), whereas once inflation has exceeded 400 percent a year, the probability that it continues to increase is 67.6 percent. On the other hand, when annual inflation is less than 50 percent, the probability that it is reduced or maintained is significantly greater than half (84.9 percent). See the highlighted area in table 1, below:

Table 1:

Fischer, Sahay and Vegh’s Market Economies Inflation Transition Matrix

(In percent)

article image
Source: Fischer, Sahay y Végh (2002).

Under the criterion used in this paper, 19 hyperinflation cycles were found in 19 countries (Table 2).9 In the Western Hemisphere, there are 6 countries (all Latin Americans): Argentina, Bolivia, Brazil, Chile, Nicaragua and Peru. In the Commonwealth of Independent States, 6 countries: Armenia, Azerbaijan, Belarus, Kazakhstan, Russia and Ukraine.10 In Europe, 3 countries: Bulgaria, Croatia and Poland (although by the time, very influenced by the USSR). In Africa: Angola, Democratic Republic of the Congo (Zaire) and Zimbabwe. Finally, in Asia and the Pacific, 1 country: Indonesia.

The duration of an average hyperinflation cycle is 16-17 years with an annual average inflation of 893 percent (a median duration of 14 years and a median annual inflation of 45 percent).11 The longest cycle was 31 years in the Democratic Republic of the Congo, followed by Argentina (25 years). While the shorter cycles (6-8 years) occurred in economies in transition —although in many of these cases there is not enough data on phase 1, the results here coincide with the aforementioned studies. In Latin America, the country with the shortest cycle was Bolivia, followed by Chile, although the increase in prices in Bolivia was substantially greater than in Chile (Table 2).

Table 2:

Modern Hyperinflation Cycles in the World

article image

Source: Author’s calculations.

The average duration of the first phase (Table 3) is 8-9 years with an annual average inflation of 125 percent (a median duration of 8 years and a median annual inflation of 96 percent). It stands out that only 6 countries did not have this phase, that is, they went from inflation less than 50 percent per year to inflation greater than or equal to 500 percent per year, in one year. These countries are the economies of the Commonwealth of Independent States, whose episodes, according to the data available and the studies mentioned above, reflect more an adjustment to the price level than a continuous and endogenous process of high inflation. The longest acceleration took place in the Democratic Republic of the Congo (15 years), followed by Argentina (12 years). The shortest episode occurred in Angola, Azerbaijan, Bolivia and Chile (1-2 years), followed by Croatia (3 years) and Indonesia (6 years).

Table 3:

Modern Hyperinflation Cycle: Phase 1

article image
Source: Author’s calculations.

The second phase (Table 4), the hyperinflation, has an average duration of 3-4 years with an average annual inflation of 2,912 percent (a median duration of 4 years and a median annual inflation of 1,330 percent), this being the shortest phase. The longest episode took place in Argentina12 and Brazil (7 years),13 followed by Nicaragua (6 years). The shortest episode occurred in Armenia, Bulgaria, Chile, Indonesia, Kazakhstan, Poland and Russia (1-2 years), followed by Azerbaijan, Bolivia and Ukraine (2-3 years).

The third phase (Table 5), disinflation, has an average duration of 3-4 years with an average annual inflation of 195 percent (a median duration of 6 years and an average annual inflation of 171 percent). The longest episode took place in Angola, (7 years), followed by Belarus, the Democratic Republic of Congo and Russia (6 years). The shortest episode occurred in Argentina, Bolivia, Brazil, Croatia, Kazakhstan and Poland (0-1 year), followed by Indonesia, Peru and Ukraine (1-2 years). Only 3 countries did not have this phase, that is, they went from inflation equal to or greater than 500 percent per year to inflation less than 50 percent per year, in less than one year. These are: Nicaragua, Bulgaria, Zimbabwe.

Table 4:

Modern Hyperinflation Cycle: Phase 2

article image
Source: Author’s calculations.
Table 5:

Modern Hyperinflation Cycle: Phase 3

article image
Source: Author’s calculations.

The average duration of the last phase is 5-6 years (Table 6), with an average annual inflation of 15 percent (a median duration of 5 years and an average annual inflation of 13 percent). The longest episode took place in the Democratic Republic of the Congo and Russia (9 years). For the rest, all countries managed to keep inflation below 50 percent per annum 5 years after the start of disinflation, especially Bulgaria, Nicaragua and Zimbabwe.

Table 6:

Modern Hyperinflation Cycle: Phase 4

article image
Source: Author’s calculations.

In summary, the dynamics of the cycle shows that, on average, a hyperinflation cycle has a relatively long incubation period (8-9 years), a relatively short explosive phase (2-3 years), a period of disinflation generally at half of the incubation period (3-4 years) and almost all countries managed to stabilize prices five years after the inflation rate lowered the 50 percent threshold. The countries with shorter cycles were those belonging to the Commonwealth of Independent States, while the countries with the longest cycles were in Africa and Latin America. Coinciding with duration, the highest accumulated rates were recorded in Africa and Latin America and the lowest in the Commonwealth of Independent States. The same occurs with the maximum registered rates, although the highest average rates were found in Latin America.

IV. Data and Methodology

After having observed the hyperinflation dynamics, the next step is to try to determine which are the macro variables14 that underlie these processes. In this sense, to observe what distinguishes these episodes, the median of a series of variables (62 to be exact) that have been considered “key” during the four phases (extraordinary acceleration, hyperinflation, disinflation and stabilization), the two stages (rise and fall) and the complete cycle has been estimated for two samples: the first sample involves the 19 episodes and the other sample comprises the rest of the countries (the control group). Subsequently, a statistical test was performed to determine if the difference between the medians of the two samples is significant at standard confidence levels (Tables 7 to 13)15. Figures 1 to 5 show a stylized representation of these key variables between the two samples at the two stages.

V. Empirical Regularities

A. Hyperinflation and Fiscal and Monetary Factors

The median value of the whole sample for the government’s fiscal balance as a percentage of GDP is -1.9 percent. Controlling for hyperinflation, the median value of this variable was around 3.1 percentage points lower during the rise stage (stage 1) and 1.5 percentage points higher during the fall stage (stage 2) than in the countries without hyperinflation, at the level of significance of 1 percent (see Figure 1, first column, charts 1 and 3, and first row of Table 7 for statistical significance). However, for the whole cycle, it is observed that the fiscal result of the government is 0.5 percentage points higher than in the rest of the sample, which would suggest that a hyperinflation cycle “helps” to improve fiscal accounts, but no statistically significant difference is found.

Table 7:

Modern Hyperinflation Cycle: Fiscal and Monetary Variables. Hodges-Lehman Median Differences 1/

article image

The significance at the levels of 10, 5 and 1 percent is represented as *, **, ***, respectively.

Sources: World Bank, IMF and Author’s calculations.

Figure 1:
Figure 1:

Hyperinflation Cycle: Fiscal and Monetary Variables

Citation: IMF Working Papers 2018, 266; 10.5089/9781484385388.001.A001

Sources: World Bank, IMF and Author’s calculations.

A similar result is observed in tax revenues as a percentage of GDP, which is more than proportionally lower in stage 1 (3.6 percentage points) than in stage 2 (1.5 percentage points). This result may be related to the Keynes-Olivera-Tanzi effect,16 by means of which hyperinflation not only reduces tax collection in real terms, but also erodes the tax base. Additionally, the result for the whole cycle of -2.9 percentage points is interesting: it suggests that a representative country of the hyperinflation cycle collects less taxes than the representative country of the sample.

On the other hand, the median value of the growth rate of base money (a proxy variable of seigniorage) is 91.6 percentage points higher in stage 1 than in any other part of the sample, at 1 percent confidence; while it is also 24.3 points higher in stage 2, also at 1 percent, and throughout the cycle it is 44.3 percentage points higher than in the rest of the sample, also at 1 percent, being the median value of the entire sample of 13.2 percent. These results suggest that an extraordinary growth of base money is an invariable fact during all phases of the hyperinflation cycle (see Figure 1, second column, charts 1 and 3, and second row of Table 7 for statistical significance).

The results of this block of variables are in line with the previous findings of the literature, which suggest a relationship between fiscal deficits, seigniorage and inflation.

B. Hyperinflation and External Factors

External factors matter. Especially when we want to try to answer the famous question: was it bad policy or bad luck? In several occasions external factors have been attributed part of the responsibility for the advent of a hyperinflation cycle (the bad luck hypothesis) while most of the occasions to domestic factors such as agents behavior (the bad policy hypothesis). Regarding hyperinflation, this study finds that the scale is tilted toward the bad policy side.

External factors can manifest themselves in several ways, among which we can find: (i) shocks in the terms of trade (affecting the real exchange rate and, therefore, prices; also, less income in foreign currency can induce greater seigniorage and, consequently, higher inflation), (ii) volatility in international capital markets (volatility in the markets can increase the cost of external debt and/or reduce its supply, possibly inducing seigniorage) (iii) restrictions on external financing (high interest rates, poor macroeconomic management, low institutional quality, diplomatic conflicts, etc., can reduce access to capital markets, inducing seigniorage) (iv) restrictions on foreign trade (trade wars, controls on current account flows, can act as a shock to external accounts, reducing hard currency supply, increasing relative prices, therefore inducing seigniorage).

To understand the role of external factors in the hyperinflation cycle, an export price index has been used as a proxy variable of changes in the terms of trade. Regarding the role of market volatility, the Libor interest rate was used in real terms (using the U.S. GDP deflator) and the behavior of the VIX index. Given that several countries with hyperinflation have high levels of income from natural resources, a variable that measures the proportion of this income over GDP is also included, to see the role of price booms and collapses in countries dependent on them. To observe the role of trade restrictions and external financing, variables of the balance of payments have been considered, such as tariffs applied to tradable products and the balance of the current and financial accounts and their components. Table 8 below presents the results of the median difference tests for the external sector variables:

Table 8:

Modern Hyperinflation Cycle: External Variables. Hodges-Lehman Median Differences 1/

article image

The significance at the levels of 10, 5 and 1 percent is represented as *, **, ***, respectively.

Sources: World Bank, IMF, Saint Louis Fed and Author’s calculations.

Although the difference in the median value of the price of exports in hyperinflation episodes is statistically significant at 1 percent, this is only slightly lower (-0.2 percentage points), with an exactly equal performance in the rise and fall stages, coupled with the fact that in the rise stage the difference is not statistically significant. This result would be reducing the emphasis of the hypothesis of the important role this variable might have at the beginning of a hyperinflation cycle. In sum, although the role of external prices is significant, its impact seems marginal.

In terms of market volatility, the median value of the VIX index is not different during the cycle, although it does not have statistical significance either.17 However, during the first stage of the cycle, the volatility of the markets is 1.6 percent lower than on any other side of the sample. This result could be suggesting, among other things, that a hyperinflation cycle may develop despite greater stability in the capital markets. However, the real interest rate in the U.S. is up to 2 percentage points higher in the rise stage, at 1 percent significance, while in the rest of the stages (including the complete cycle), the difference is not statistically significant. Thus, higher borrowing costs could be an important factor.

Regarding restrictions to international trade, only a difference in medians is statistically significant. This variable, which is a quotient that measures the size of the tariff applied to tradable goods in the economy of origin, shows that, during the second stage of the cycle, this rate is up to three percentage points higher than in the rest of the sample, at the five percent level.

As another proxy variable to the performance of current foreign transactions, the balance of the current account as a percentage of GDP was used. For this variable, it stands out that —at least at the traditional levels of confidence— no difference in medians is statistically significant. This implies that the signs and magnitudes may differ across countries. However, the sign of the differences in medians during the two stages of the cycle contains relevant information: in the first stage, the balance of the current account as a percentage of GDP is 0.4 percentage points higher than the rest of the sample. In the second stage, it is 1.5 percentage points lower than in the rest of the sample.

This result can have different implications. Among them, we could have, given that external prices are somewhat lower during the cycle: (i) lower imports (both by reductions in economic activity and/ or restrictions —foreign and domestic— to the capital account that, in response to different policy options, may result in import cuts), (ii) lower volume of exports (due to falls in economic activity or trade restrictions), (iii) lower (or greater) payments of external debt and less repatriation of dividends (both due to the occurrence of defaults or to refinancing/restructuring of public external debt, as well as due to falls in economic activity and controls) and/or (iv) falls in current transfers.18

To test hypothesis (i), the behavior of prices and volumes of imports has been incorporated. In the case of prices, there are no significant differences, but in the case of volumes, it is observed that, in the first phase, the growth rate of imports is 5.7 percentage points lower, while during the second stage, it is higher by 4.5 percentage points, at the 5 percent level. In the case of export volumes (hypothesis ii), a positive difference of 3 percentage points is observed in the second stage of the cycle, at the 5 percent level. These results mainly suggest a deterioration and subsequent recovery of the absorption and export capacity of the economy during the cycle.

Similarly, to identify the role of external debt payments (hypothesis iii), we have observed the behavior of the variable that measures interest payments on foreign debt as a percentage of total exports of goods, services and net factor payments. This variable shows an interesting dynamic: it is 11 percentage points higher (and significant at 1 percent level) during the rise stage, while in the fall stage, especially during the beginning of disinflation (phase 3), it is 0.6 percentage points lower than the rest of the sample (although this difference is not statistically significant). That is, external debt interest payments increase significantly when domestic prices rise, then fall when the disinflation process begins and then regularize during the stabilization phase.

This behavior of the external debt service to exports during hyperinflation —especially the statistical relevance during the rise period— is interesting in the sense that it has potential to explain the contractionary role of fiscal policy in such episodes. It suggests, among other things, that increases in public spending based on increases in external debt have the capacity to alter the future allocation of resources (via fewer imports and increases in seigniorage, for example), affecting not only the government’s capacity to generate goods and services but also that of the private sector (via the crowding out effect, increased borrowing costs, etc.), which in turn exacerbate the deterioration in economic activity and trigger other problems (such as shortages19), that reinforce the process of exponential increase on domestic prices.

The above described mechanism could be amplified in economies with a high level of natural resource dependency where the State also exercises a strong presence in economic activity, since committing exports proceeds to pay debt could have a deeper impact than in the rest of countries. This is the case of Angola, the Democratic Republic of the Congo, some Commonwealth countries, Russia, Bolivia, and Chile.

The median value of the financial account balance as a percentage of GDP is 1.8 percentage points lower during the hyperinflation cycle, although it is not statistically significant. It is noteworthy, however, that while this difference is not substantial in the first stage (between 0 and -0.2 percentage points), it is substantial in the second (between -1 and -1.8 percentage points). Inquiring through the components of the financial account, we observe that the flows of foreign direct investment (FDI, from now on) are 0.5 points of GDP lower (-0.4 for incoming flows and -0.1 for outgoing flows) during the cycle, at 1 percent level. These lower FDI flows are greater in the first stage than in the second, which suggests a greater inflow once inflation begins to fall and stabilize. The median value of portfolio investment is notsignificantly different from the rest of the sample in each of the four phases of the cycle, which suggests that the rest of the contraction of the financial account during the cycle comes from the other investment account and variations in international reserves. In fact, the median value of international reserve flows as a proportion of GDP is lower by 3.6 points, at the 1 percent level.

The results of the financial account could help conjugate the hypothesis that the lower current and financial accounts balances during the cycle are mainly due to (i) a greater burden of external debt service as a percentage of exports, (ii) lower FDI inflows and (iii) lower cash flows, lower commercial and financial loans and (iv) net losses of international reserves.

In sum, from the analysis of external factors, it can be inferred that exogenous external factors (such as shocks in the terms of trade and volatility in international financial markets) are not very different during the hyperinflation cycle, although the role of external interest rates seems to be important. Rather, factors associated with economic policy variables, such as high fiscal deficits, intervention on current and capital account transactions, the exchange rate regime, basic economic freedoms, are the ones that differ significantly. See Figure 2 below:

Figure 2:
Figure 2:

Hyperinflation Cycle: External Variables

Citation: IMF Working Papers 2018, 266; 10.5089/9781484385388.001.A001

Sources: World Bank, IMF, St. Louis FED, and Author’s calculations.

C. Hyperinflation and the Real and Financial Sectors

This section goes on to explore the impact of the hyperinflation cycle in the two most important exchange sectors of the economy: real and financial. This section will evaluate the unanimous view of the economic literature that high inflation, in this case hyperinflation, is detrimental to economic growth. As well as the case that, in hyperinflations, the financial system contracts significantly, influencing the effect this sector has in the economy.20

In this sense, for the block of real variables, the growth rate of all components of aggregate demand (GDP) and the real exchange rate have been considered; while for the financial variable block, credit to the private sector and bank deposits (as proxy variables of the performance of the banking sector) were also considered.

Table 9:

Modern Hyperinflation Cycle: Real and Financial Variables. Hodges-Lehman Median Differences 1/

article image

The significance at the levels of 10, 5 and 1 percent is represented as *, **, ***, respectively.

Sources: World Bank, IMF and Author’s calculations.

The median value of 3 different variables that measure the performance of the economy: real GDP growth, real GDP growth per capita, and GDP per capita in constant dollars, is lower during the entire hyperinflation cycle, at least at the 5 percent level. Similarly, although there is a negative difference of medians throughout the cycle for consumption and investment, these are not statistically significant. However, public consumption contracts by 1.5 percentage points, at the 10 percent level. In the case of the real exchange rate, while the median value is lower by 13.2 percentage points (which suggests a real depreciation), this difference is not statistically significant. Finally, investment and household savings as a proportion of real GDP are between 4 and 5 percentage points lower than in the rest of the sample, at the 1 percent level.

In the case of the financial sector, during the whole cycle, the average value of credit to the private sector as a proportion of GDP is 13.3 percentage points lower while the ratio of bank deposits/GDP is also lower by 14.8. percentage points, at 1 percent confidence in both cases. This result goes hand in hand with the previous findings of the literature on hyperinflations.21 See Figure 3 below as a summary of these findings.

Figure 3:
Figure 3:

Hyperinflation Cycle: Real and Financial Variables

Citation: IMF Working Papers 2018, 266; 10.5089/9781484385388.001.A001

Sources: World Bank and Author’s calculations.

D. Hyperinflation, Other Economic Policy Variables and Structural Factors

In the previous sections it was described how, in this sample, and beyond a series of exogenous factors, an unsustainable macroeconomic policy mix highlights as the characteristic factor during all the hyperinflation episodes. However, these are not the only variables that the authorities of a country have to influence their economy. There are other economic policies, such as exchange rate policy (via the design and management of the exchange rate regime), as well as development policies (via changes or reforms that affect the structure and functioning of the economy), that also deserve to be considered.

In this sense, for the first block (exchange rate policy), three variables have been included: (i) the type of exchange rate regime (based on of Itzenski, Reinhart and Rogoff22 database), understanding that changes in the exchange regime have repercussions on other policy variables as well as on the economy; (ii) an index of restrictions and controls to the financial account (using the Chin and Ito23 Index), understanding that the level of regulation of international financial transactions impacts the flow of capital and, therefore, may increase seigniorage, and (iii) the depreciation rate of the nominal exchange rate, understanding — based on the extensive literature on the role of the nominal exchange rate as an economic policy variable— that country authorities intervene in the foreign exchange market so as to smooth out fluctuations as to promote the exports competitiveness24 and that, in turn, large nominal exchange rate depreciations have an impact on the aggregate level of prices, the fiscal, monetary and external accounts, among others. (Dornbusch, 1993; Morales, 1989).

For the second block (structural changes), three variables have been included: (i) Heritage Foundation’s economic freedom index,25 understanding that changes in economic freedoms have an impact on the performance of the economy, (ii) access to multilateral financial assistance programs (particularly programs with the International Monetary Fund), understanding that these programs aim to improve the macroeconomic performance of countries in times of stress and (iii) country risk indicators from the International Country Risk Guide (ICRG),26 which measure socioeconomic, institutional and political conditions, understanding that the hyperinflation cycle can affect (and be affected by) the behavior of these variables.

Table 10:

Modern Hyperinflation Cycle: Exchange Rate Policy and Structural Variables. Hodges-Lehman Median Differences 1/

article image

The significance at the levels of 10, 5 and 1 percent is represented as *, **, ***, respectively.

Sources: World Bank, IMF, Heritage Foundation, ICRG and Author’s calculations.

Regarding the role of the exchange regime, it is necessary to explain first that the Itzenski et. al Index moves on a scale of 1 to 15, with the first levels being “more pegged” regimes while the highest levels reflect regimes that float freely, depreciate in free fall or in which a dual exchange rate system with parallel markets exists.27 That is, as the index increases, there is more variability in the currency.

In this sense, it is observed that, during the complete cycle, the median value of the index is 6 positions higher than the rest of the sample, at the 1 percent level. In stage 1, the median value of the index increases 10 positions, while in stage 2 it increases only 3, also at the 1 percent level. During the first 3 phases, the median value of the exchange regime becomes between 9 and 10 positions “more flexible” than the median value of the rest of the sample, while, in the last phase (the stabilization), the index is only one position higher than the rest of the sample’s median, but this difference is not statistically significant. The median value of the exchange regime for the whole sample is 4 (i.e. de facto peg).

In other words, in the first stage, a fixed exchange rate regime changes to a free fall and/or dual market one. While in the stabilization stage, countries return to a regime of less fluctuation/more stability of/in the foreign exchange rate. This result suggests capacity, willingness and credibility losses in the management of exchange rate policy by the authorities during the cycle, perhaps as a consequence of the application of controls to foreign currency transactions, financing restrictions in capital markets, among others.

The previous results go hand in hand with the dynamics of the nominal exchange rate. For the whole cycle, the median depreciation rate of the official nominal exchange rate is almost 40 percentage points higher than the rest of the sample, this difference being significant at 1 percent level. In the first stage, the depreciation rate is 94.5 percentage points higher than in the rest of the sample, while in the second stage the difference in medians is higher by 11.7 percentage points, both at 1 percent confidence (the median value of the depreciation rate of the nominal exchange rate in the entire sample is 0).

Finally, the median value of the Chinn and Ito capital control index (ranging from -2 to 2, where a negative value reflects greater restrictions) decreases by 0.7 during the hyperinflation cycle, at 1 percent level. In the rise stage, this difference increases to -0.9 (also at the 1 percent), while in the fall stage it only decreases 0.5 points (although the difference is not statistically significant), highlighting that, in the stabilization phase, there are practically no differences with the median value of the rest of the sample (the median value of the index for the whole sample is -0.1).

The fact that the median value of the Chinn and Ito index for the entire sample is zero illustrates how, during the rise stage, countries have opted for a stricter stance towards capital flows, while, in the fall stage, things “return to normal”. That is, although the controls on the financial account do not reflect a statistically significant change on capital flows, the negative sign of the financial account for the entire cycle suggests two things: (i) greater control of these transactions does not achieve a change in the financial account at the first stage, that is to say, there is still a net negative balance (i.e. the net outflows of capital do not decrease, in any case they are contained), while (ii) their relaxation in the second stage deteriorates the net negative balance of said account (i.e. net capital outflows increase).

These findings show that, during a hyperinflation cycle, tightening the control over the financial account has the usual implications found in the literature on capital flows28 and is even counterproductive because it exacerbates the external constraint in a context of deterioration of the local currency, possibly generating pressures on the nominal exchange rate and seigniorage. Relaxing this control, however, seems to go hand in hand with a lower pressure on the nominal exchange rate, less seigniorage and greater (absolute) capital inflows.

The results of the exchange rate policy set are consistent with the results for other policy variables mentioned up to this point, since they show that the exchange rate regime is rather accommodating price increases; and capital controls rather exacerbate the effects of other economic policy variables. See Figure 4 below:

Figure 4:
Figure 4:

Hyperinflation Cycle: Exchange Rate Policy Variables

Citation: IMF Working Papers 2018, 266; 10.5089/9781484385388.001.A001

Sources: World Bank, Chinn and Ito (2016), Itzenski, et. al (2016) and Author’s calculations.

With respect to the set of structural variables, in the case of economic freedoms (measured by the index of economic freedom, which ranges from 0 to 100 and where the higher the index, the greater the freedom) it is observed that, throughout the cycle, the median value of the aggregate index is 16 points lower than the rest of the sample, at 1 percent confidence. In the first stage, the index is lower by 21.1 points, at 1 percent confidence. In the second stage, the index is lower by 14.8 points, also at 1 percent confidence. For comparison, see Figure 5, column 1, charts 1 and 3, and so on.

Inquiring through the components of the index, it is observed that freedoms to do business, labor and exchange, decrease significantly, at the 1 percent level. The significant deterioration in fiscal (at 5 percent) and monetary (at 1 percent) freedom rates validates the findings in part A of this section. Likewise, freedom from corruption also deteriorates significantly during the cycle, at 1 percent confidence.

While these results are revealing and suggest that during the hyperinflation cycle restrictions to economic activities increase, two characteristics that underlie the sample should be highlighted and weighted: while only 3 of the 19 countries that have had a hyperinflation cycle show data coinciding with the index (since it starts in 1995), it is relevant that the median value of all countries (i.e. even when they were not at hyperinflation) that have had hyperinflation is up to 16 points lower than the representative value the rest of the sample shows, suggesting that hyperinflations are a phenomenon that occurs in contexts in which economic freedoms are substantially reduced29.

Likewise, indicators of political and social risk show that, during the hyperinflation cycle, there is a statistically significant deterioration in socioeconomic conditions (this being an index that considers variables of unemployment, consumer confidence and poverty). This result may reflect the closure of companies and the consequent loss of jobs, as well as the loss of purchasing power during hyperinflations. For a more detailed discussion of these issues see Bailey (1956), Braumann (2001).

Similarly, there is a deterioration in democratic accountability (at 1 percent confidence), the quality of institutions (at 5 percent) and the rule of law (at 10 percent); as well as greater instability in the government (10 percent) and greater internal conflicts (5 percent). It also highlights that, throughout the hyperinflation cycle, there is a greater presence of the military in the government (5 percent). Among other things, these results coincide with findings like those of Alesina and Drazen (1991), who show how institutional conflicts and political economy problems delay stabilizations.

Figure 5:
Figure 5:

Hyperinflation Cycle: Structural Variables

Citation: IMF Working Papers 2018, 266; 10.5089/9781484385388.001.A001

Sources: Economic Freedom Report, International Country Risk Guide and Author’s calculations.

The other economic policy variable that has been considered key during a hyperinflation cycle is multilateral assistance. Sargent (1982), in his seminal work on how the great European hyperinflations of the interwar period came to a halt, highlighted as an essential factor the role of international assistance not only in stabilizing prices but also in reactivating economic growth.

In this sense, it is observed that in 18 of the 19 hyperinflation cycles a program of financial assistance with the IMF has been implemented, Angola being the only country that did not request assistance. In the first stage, the presence of a program is observed in 45 percent of the cases. While, in the second, it is observed in 62 percent of the cases. If we observe the behavior of this variable in each of the four phases, three interesting facts stand out: (i) the absence of programs in phase 2 (hyperinflation), 66 percent of the cases; (ii) the greater presence of programs in phase 4 (stabilization), 67 percent of cases and (iii) that on the road to hyperinflation, the first phase, the presence of programs was 51 percent, suggesting that in half of the cases, a hyperinflation cycle developed despite having multilateral assistance.

The countries that used this assistance the longest were Democratic Republic of the Congo (12 years, 7 of them in phase 1, the rest in phase 4) and Argentina (12 years, starting one year before starting phase 2, until the end of the cycle). The countries that used this assistance the least were Zimbabwe (2 years, only at the beginning of phase 1) and Chile and Brazil (3 years, only in phases 2 and 3).

Table 11:

Financial Assistance Programs During Modern Hyperinflation Cycles.

article image
Source: Author’s calculations.

To evaluate the impact of financial assistance in the hyperinflation cycle, the same median difference test was carried out, but comparing the years in which the countries had assistance against the years in which they did not have it. In terms of the complete cycle, it is highlighted that in the countries that had an assistance program, both, base money growth and nominal exchange rate depreciation were lower (-29.2 and -34.3 percentage points, respectively).

In the first stage, assistance recorded a higher balance of payments and higher GDP growth, despite registering a higher cost of imports. In the second stage, countries that deflated and stabilized using IMF assistance registered a fiscal result 1.8 points of GDP lower (i.e. a greater deficit) than the countries that stabilized without using it. They also showed a larger contraction on base money growth (17.3 points lower). There were also lower volumes of imports, a greater proportion of external resources destined to serve the external debt and a lower contraction in the final consumption expenditure of the public sector.

Table 12:

Modern Hyperinflation Cycle: Selected Variables in Case of Financial Assistance.Hodges-Lehman Median Differences 1/

article image

The significance at the levels of 10, 5 and 1 percent is represented as *, **, ***, respectively.

Sources: World Bank, IMF, Heritage Foundation, ICRG and Author’s calculations.

Finally, it highlights that, at both stages of the cycle, the economic freedom index was higher for the representative country that went to the IMF.

These results show that, throughout the cycle, access to multilateral financial assistance programs entails, in terms of statistical relevance, more fiscal flexibility, lower growth of base money, less variation in the nominal exchange rate, greater burden of external debt on exports and greater economic freedoms. In the fall stage, apart from those mentioned above, a lower growth of imports, lower growth of public consumption and greater economic freedoms stand out. There are no differences in the exchange regime, nor in the flows of foreign direct investment.

E. Within Analysis

Given that the analysis presented in the four previous subsections only shows the comparison between countries, this section would be doing the within countries one. Table 13 below shows the difference in medians between the rise and fall stages, although only showing the variables that turned out to be statistically significant.

When prices rise, the fiscal and monetary variables show the expected signs. Regarding the external sector, while lower market volatility is observed, higher interest rates in the markets are shown. There is also a contraction on imports, a greater burden of external debt interest on exports, and lower capital flows and international reserves. Although there are no significant differences regarding the financial system (at least at the 10 percent level), GDP growth, public consumption and investment are lower.30 As for exchange rate policy, there is evidence of a massive depreciation of the exchange rate, coupled with a much less stable exchange rate regime which, by symmetry, translates into lower depreciation and a much less volatile exchange rate regime in the fall stage.31 Finally, it highlights how, at the rise stage, economic freedoms are lower and there is greater government instability.

Table 13:

Modern Hyperinflation Cycle: Selected Variables at Stages 1 and 2. Hodges-Lehman Median Differences 1/

article image

The significance at the levels of 10, 5 and 1 percent is represented as *, **, ***, respectively.

Sources: World Bank, IMF, Heritage Foundation, ICRG and Author’s calculations.

VI. CONCLUSIONS

At the beginning of this paper, it was proposed to infer (i) under what conditions the hyperinflation cycles have begun, (ii) what happens during these episodes, (iii) under what conditions the hyperinflation episode could probably be closed and (iv) what could be done by authorities to increase the probability that a disinflation process will be carried out successfully and at the lowest possible cost.

Now the initial questions will be responded:

(i) The hyperinflation cycle seems to be a phenomenon that, more than regional, occurs in economies with high presence of natural resource rents (and potentially higher State intervention in the economy) and where economic freedoms have been diminished, especially those related to the ease of doing business and trade, as well as those related to rigidities and inefficiencies in the labor market. Although external factors matter, their role may not have been as significant as domestic factors, especially those inherent to economic policy: high fiscal deficits, in some cases financed with external debt at the beginning and with seigniorage afterwards; inability to maintain a certain exchange rate regime and restrictions on transactions in the financial account of the balance of payments are the variables whose performance coincides with the advent of a hyperinflation cycle. The cycles also coincide with contexts in which socioeconomic conditions such as employment and real wages deteriorate, where the rule of law and democratic accountability are subdued, the instability of the government increases and there is a greater presence of military personnel over political issues.

(ii) In hyperinflation (the second phase of the four that comprise the cycle) base money grows exorbitantly, there is a greater loss of international reserves, the deterioration in the aggregate demand of the economy is massive, the financial sector contracts, the currency loses all (or almost all) its value, social and political conditions, as well as economic freedoms, deteriorate to the lowest levels recorded in the sample. All of this, even though the external factors seem to be better with respect to the previous phase.

(iii) Although it is not intended to be exhaustive, since each country requires “a tailored suit”, the results of this sample suggest that hyperinflation cycles end when there are improvements on three essential fronts: (a) the fiscal and monetary mix: fiscal accounts are closer to equilibrium and base money growth decreases substantially; (b) the interaction with the external sector: barriers to international trade diminish and the exp/imp capacity of the economy increases significantly, the burden of foreign debt on exports regularizes, the resounding level of devaluation of the currency is stopped hand in hand with less variability of the foreign exchange rate; (c) the structural factors: economic freedoms increase and there is greater government stability.

(iv) In this paper, it was proposed to evaluate the role of multilateral assistance during the hyperinflation cycle. The reason is simple: countries that face these cycles find themselves in a position where, on their own, recovery would look uphill. It was observed that the representative country that has used it registers lower base money growth, smaller variation of the exchange rate, a greater burden of foreign debt payment to exports (perhaps suggesting a lower haircut on external debt outstanding) and greater economic freedoms. No significant differences were observed on economic growth,32 the exchange rate regime, and foreign investment flows. These results, rather than suggesting that countries that had assistance managed to have greater fiscal flexibility and lower monetary growth (an interesting result), invite analyzing the particularities of each case from an optimistic perspective, since it is precisely in the design of these programs, where the “tailored suit” applies the most.

Looking ahead, the empirical regularities found here propose multiple avenues to advance in the analysis of hyperinflation cycles. Questions like: Why is there no substantive evidence of hyperinflation between 1947 and 1984? What coincidences are in the institutional arrangements and/or political systems of the countries that have suffered this phenomenon? What is the true impact of external factors on hyperinflations, is it as marginal as it looks in this study? How is the financial system affected beyond the balance sheet? What is the behavior of the labor market, can it be documented in detail? What role do parallel markets play? Why do countries that used assistance programs with the IMF appear to have more fiscal flexibility than countries that do not? are those that occur to the one who write these lines and are open to be answered by anyone who has a great interest in continuing to document the hyperinflation cycle.

Appendix I. Data

The tables below show from where are obtained and a brief description of each of the variables studied.

Table A1:

Fiscal and Monetary

article image
Table A2:

External

article image
article image
Table A3:

Real and Financial

article image
article image
Table A4:

Other Economic Policy and Structural

article image
article image
article image
article image

APPENDIX II. ADDITIONAL CHARTS

A01ufig1

Hyperinflation Cycle: Episodes by Country

Citation: IMF Working Papers 2018, 266; 10.5089/9781484385388.001.A001

Sources: World Bank and Author’s calculations.

References

  • Alesina, Alberto, and Alan Drazen, 1991, “Why Are Stabilizations Delayed?American Economic Review, Vol. 81, pp. 1170-88.

  • Braumann, Benedikt, 2001, “High Inflation and Real Wages,” IMF Working Paper 01/50 (Washington: International Monetary Fund).

  • Bruno, Michael, and William Easterly, 1995, “Inflation Crises and Long-Run Growth,” NBER Working Paper No. 5209 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Bruno, Michael, and Stanley Fischer, 1986, “The Inflationary Process: Shocks and Accommodation,” in The Israeli Economy: Maturing Through Crises, ed. by Yoram Ben Porath (Cambridge, Massachusetts: Harvard University Press), pp. 347-71.

    • Search Google Scholar
    • Export Citation
  • Cagan, Phillip, 1956, “The Monetary Dynamics of Hyperinflation,” in Studies in the Quantity Theory of Money, ed. by Milton Friedman (Chicago: University of Chicago Press), pp. 25-117.

    • Search Google Scholar
    • Export Citation
  • Calvo, Guillermo A., and Carlos A. Végh, 1993, “