Volatility and Growth in Latin America
An Episodic Approach
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Author’s E-Mail Address: rsahay@imf.org, rgoyal@imf.org

This paper compares the pattern of macroeconomic volatility in 17 Latin American countries during episodes of high and low growth since 1970, examining in particular the role of policy volatility. Macroeconomic outcomes are distinguished from macroeconomic policies, structural reforms and reversals, shocks, and institutional constraints. Based on previous work, a composite measure of structural reforms is constructed for the 1970-2004 period. We find that outcomes and policies are more volatile in low growth episodes, while shocks (except U.S. interest rates) are similar across episodes. Fiscal policy volatility is associated with lower growth, but fiscal policy procyclicality is not. Low levels of market-oriented reforms and structural reform reversals are also associated with lower growth.

Abstract

This paper compares the pattern of macroeconomic volatility in 17 Latin American countries during episodes of high and low growth since 1970, examining in particular the role of policy volatility. Macroeconomic outcomes are distinguished from macroeconomic policies, structural reforms and reversals, shocks, and institutional constraints. Based on previous work, a composite measure of structural reforms is constructed for the 1970-2004 period. We find that outcomes and policies are more volatile in low growth episodes, while shocks (except U.S. interest rates) are similar across episodes. Fiscal policy volatility is associated with lower growth, but fiscal policy procyclicality is not. Low levels of market-oriented reforms and structural reform reversals are also associated with lower growth.

I. Introduction

Latin America is well known for its high macroeconomic volatility and low growth performance relative to other regions. Less well known are the linkages among different kinds of volatility—for instance, the volatility of macroeconomic policies and macroeconomic outcomes—and their effects on long-term growth. This paper aims to further our understanding of volatility and growth in Latin America by comparing the pattern of volatility across episodes of high and low growth since 1970. Of particular interest is the role of policy volatility in accounting for the region’s growth performance.

To lay out the stylized facts, the paper takes an episodic approach across 17 Latin American countries. Non-overlapping 10-year periods or episodes of the highest and lowest growth are identified for each country during 1970–2004. These periods are not necessarily the standard decadal panels such as the 1970s or the 1990s and may vary across countries. The variables of interest are compared between these two growth episodes for each country, and common lessons are drawn across the countries.

As a data description tool, this approach complements the standard cross-country growth regression exercises and country case studies, and has some useful advantages. First, it avoids the need to define arbitrary time periods for computing averages and volatilities. There is nothing unique about using the standard decadal panels. The approach allows the data to define the periods of best and worst growth performance. Second, it highlights and utilizes the within-country variation in growth performance to identify potentially important explanators for growth for each country. These explanators can then be used to better understand the determinants of growth across countries, including through cross-country regressions.2 But highlighting the within-country variation also helps to address one critique of cross-country regressions commonly heard from policy makers, “My country is different,” by focusing on factors driving growth within each individual country.3 Additionally, the approach identifies outliers in the data. Accounting for growth in these outlier countries could make for useful case studies, and policy recommendations may need to be tailored appropriately.

A second aspect in which the paper differs from previous studies is that it attempts to clearly distinguish among macroeconomic outcomes, policies, and shocks in documenting macroeconomic volatility in Latin America. Surprisingly, and somewhat frustratingly, the term “macroeconomic volatility” has been used rather loosely in the literature. It has been applied to characterize output growth volatility, as well as volatility of other macroeconomic outcomes, policies, and shocks. From the perspective of a policymaker, it is critical to distinguish among factors or variables that are under or outside of his or her control. For example, if Latin America is more volatile than other regions because it experiences more shocks, then the low growth performance can be attributed more to shocks than to bad policy making, although the effects of the shocks would still need to be smoothed by developing a reactive policy framework. On the other hand, if output volatility and low growth are generally caused by variable or inconsistent policies, independent of shocks, a more proactive (and appropriate) policy framework is needed.

In characterizing policy volatility, macroeconomic policy volatility is distinguished from structural policy volatility. The former refers to variability in fiscal and monetary policies, while the latter refers to changes in policies such as product market regulations, trade taxes, regulatory trade barriers, and credit and labor market regulations. Structural reforms and reversals can have independent effects on growth, that is, both the degree of market orientation (the “level” of structural reforms) and the number and intensity of reform reversals can have separate and distinct effects on growth. To analyze the roles of structural reforms and of structural reform volatility, a new index of structural reforms is developed for the Latin American countries from 1970–2004 by combining measures developed by other authors. To our knowledge, such an exercise over an extended time period has not been conducted before—in part, because of lack of measures of structural reforms over long periods of time.

Policymakers in developing and emerging market economies often face financing and other constraints that may limit their policy options. These constraints, which include financial constraints such as lack of access to international capital markets and trade openness and restrictions, could amplify the volatility of macroeconomic outcomes and reduce growth, independent of policies and shocks. This paper examines a broader set of shocks and constraints than has been attempted previously. Within a single framework, the impact of policy volatility (both macroeconomic and structural) on economic growth is analyzed, controlling for shocks, constraints, and political and social characteristics, which directly or indirectly affect macroeconomic outcomes and policies.

The paper finds, as expected, that volatility of outcomes and policies is higher in episodes of low growth compared to episodes of high growth. Surprisingly though, both the level and volatility of shocks are similar across episodes, with the strong exception of U.S. interest rates. Extreme events—currency crises, banking crises, and debt defaults—are more frequently associated with low growth episodes, but are nonetheless quite common during other episodes as well. Output generally falls during such events, especially simultaneous crises involving a currency crisis. The volatility of discretionary fiscal policy is associated with lower growth, and although there is strong evidence of fiscal policy procyclicality across the region, there is no significant difference between procyclicality in high-growth episodes compared with low-growth episodes. Low levels of market-oriented reforms are also associated with lower growth, as are structural reform reversals. Indeed, there have been several instances of reform reversals in virtually all countries in the region, notwithstanding the progress made since 1970.

The rest of the paper is organized as follows: Section II outlines the methodology of the episodic approach. Sections IIIVII document, respectively, the volatility of macroeconomic outcomes, shocks, macroeconomic policy volatility, structural reforms and their reversals, and constraints faced by policymakers during the high and low growth periods. The different factors associated with the growth performance are brought together in Sections VIII and IX. The final section offers some concluding remarks. An appendix lists the data sources.

II. The Building Blocks

The robust negative cross-sectional relationship between volatility and growth was established in a seminal paper by Ramey and Ramey (1995) (Figure 1). Prior to their paper, growth and business-cycle volatility were presumed to be unrelated (see, for example, Lucas 1987). Ramey and Ramey’s result is of special importance for Latin America if indeed the region were volatile.

Figure 1.
Figure 1.

Real GDP Growth and Volatility, 1970-2004

Citation: IMF Working Papers 2006, 287; 10.5089/9781451865479.001.A001

Source: World Bank, World Development Indicators

In a comprehensive study conducted at the Inter-American Development Bank, Hausmann and Gavin (1996) documented the extent of macroeconomic volatility in Latin America, and found that Latin America has been more volatile than any other region in the world, with the exception of Africa and the Middle East. This finding remains true even when more recent years are included. During 1970–2004, the volatility of growth (measured by the standard deviation of real GDP per capita growth) was higher in Latin America than in industrial countries and emerging Asian countries but lower than in the Middle East and Africa (Figure 2).

Figure 2.
Figure 2.

Regional GDP Growth and Volatility, 1970-2004

(In percent)

Citation: IMF Working Papers 2006, 287; 10.5089/9781451865479.001.A001

Source: World Bank, World Development Indicators

Within Latin America, the cross-sectional evidence confirms a significant and negative relationship between volatility and growth, even in a small sample of 17 countries (Figure 3). Strikingly, and as is well documented, average per capita growth in each of the countries in the region over the past 35 years has been low—ranging from -2 percent per annum in Nicaragua and -1 percent per annum in Venezuela to an increase of about 2½ percent per annum in Chile. Volatility of growth has been the highest in Nicaragua and lowest in Colombia.

Figure 3.
Figure 3.

Latin America: Real GDP Growth and Volatility, 1970–2004

(In percent)

Citation: IMF Working Papers 2006, 287; 10.5089/9781451865479.001.A001

Source: World Bank, World Development Indicators.

Several cross-sectional studies on growth have included some measures of macroeconomic volatility to study their impact on growth. However, there is little consensus yet on the channels through which different types of macroeconomic volatility affect growth. There is a need, in particular, to understand better whether this volatility is induced by policies, shocks, or underlying structures of the economies, and how it affects growth.

Figure 4 provides a schematic representation of the channels through which growth may be affected by the volatility of macroeconomic outcomes, shocks, policies, constraints, and socio-political institutions. Broken arrows indicate the direction of causation. Two-way arrows imply that causation could go both ways. Thus, for example, exogenous shocks affect policies and outcomes, but not vice versa. On the other hand, financing constraints and trading restrictions can affect the choices that policymakers may face, and at the same time policies could influence these constraints and restrictions. While it would be highly ambitious and beyond the scope of this paper to explore all possible relationships, an attempt is made to investigate the relationships of particular interest from a policy perspective.

Figure 4.
Figure 4.

Channels through which Policy Volatility Affects Growth: A Schematic Representation

Citation: IMF Working Papers 2006, 287; 10.5089/9781451865479.001.A001

A. Methodology: An Episodic Approach

There are several ways of identifying the high and low growth periods in each country. One approach is to examine performance during each decade (for example, the 1970s, 1980s, 1990s, and so on). This is overwhelmingly the approach followed in panel long-term growth studies. The disadvantage is that it cuts the data at arbitrary points in time—at the start and end of each decade. It may well be the case, for example, that the high or low growth periods start in the middle of a decade, which would not be captured by using arbitrary cut-off points in time.

A second approach is to define the ranges for “high” and “low” growth. For example, a high growth period can be defined as one where growth per annum is more than x percent for at least y number of years, and similarly for low growth periods. Several papers try to identify and explain either growth accelerations or decelerations. Ben-David and Papell (1998), Rodrik (1999) and Hausmann et al. (2004) focus on growth accelerations, Jones and Olken (2005) look at growth decelerations, and Berg et al. (2006) study the duration of growth spells following an acceleration.

Accelerations and decelerations in previous work are identified using either structural break analysis or arbitrary growth thresholds, or a combination of the two. The problem of the first analysis is that high volatility in developing countries makes it difficult to identify structural breaks. The problem with the second is the converse: one cannot be sure whether growth increases (even persistent ones) that are not identified as structural breaks are random or not. In addition, all of these analyses, by definition, only focus on explaining either the onset or the length of high growth episodes, and hence ignore other aspects of the growth phenomenon that might be relevant to a country’s long-run growth performance. For example, if the levels of per capita income and economic development differ significantly across countries, then imposing the same cut off point for all countries may not make sense. Studies have shown that countries with lower per capita incomes tend to grow faster than those with higher per capita incomes (absolute convergence theory). Finally, more relevant for this study, if the high or low growth episodes last for very short periods of time, then it is computationally hard to measure volatility during those periods (at least a few years of data are needed)—a key variable of interest in this paper.

The episodic approach used in this paper attempts to address some of the shortcomings of the above approaches by taking a sufficiently long period of time so that volatility can be measured sensibly and by allowing the data to define the start and end of the high and low growth periods. Specifically, 10-year moving averages of GDP per capita growth are computed, and the 10 consecutive best and worst years identified for each country during the 35-year period (1970–2004). In other words,

High growth period in country i = decade corresponding to max (git), where t is a 10-year interval corresponding to 1970–79, 1971–80, …, 1995–2004, and g is the average annual per capita GDP growth rate in that 10-year interval, and

Low growth period in country i = decade corresponding to min (git), where g and t are defined as above.

Non-overlapping 10-year periods of low and high growth are found for 14 of the 17 Latin American countries. The exceptions are Costa Rica, Uruguay, and Venezuela. For these countries, the lowest and highest decades are picked such that the distance between the lowest (highest) and next lowest (next highest) were the smallest in terms of average per capita growth rate during the decade and such that the periods were non-overlapping. These periods of high and low growth are unique for each country, and form the basis of the subsequent analysis.

III. Level and Volatility of Macroeconomic Outcomes

A. Output Growth

Several interesting facts emerge from defining the high and low growth periods according to this approach (Table 1 documents the level and volatility of output growth). The high and low growth periods do not generally coincide across the countries. Rather, the starting and ending periods for the episodes differ, particularly with regard to the highest growth decade. The 1990s were not the best years, unlike as is sometimes believed. The 1970s remain the best years for about half the countries, while for the others there is no systematic pattern: for six countries, the 1990s were the best; for two, they started in the mid 1980s; and for one, the starting point was the mid 1990s. The best three performers in their respective best decades were Chile (1988–97: 6.1 percent per annum); Paraguay (1972–1981: 5.9 percent per annum); and Brazil (1971–80: 5.8 percent per annum). But even during their respective best decades, growth in the Latin American countries remained below the blistering pace set by the emerging Asian economies.

Table 1.

Latin America: Growth and Volatility, High and Low Years, 1970-2004

article image
Sources: World Bank, World Development Indicators; and IMF staff calculations.

Ratio of volatility in low decade to high decade.

Simple average.

During their worst decade, all countries recorded near zero or negative growth rates. The 1980s were, as anticipated, the worst or “lost decade” for most but not all countries, the exceptions being Chile, Colombia, Paraguay, Uruguay, and Venezuela. The worst performers during their respective worst decade were Nicaragua (1979–88:−6.7 percent per annum), El Salvador (1979–88) and Peru (1982–91).

In most countries, the volatility of growth was substantially higher during the low growth period. This is most evident in Nicaragua and Chile, but in Argentina and Venezuela, on the other hand, volatility was fairly high and similar in both the high and low growth periods. As a region, volatility was more than twice as high during the worst decade, as compared to the best decade. The negative relationship between volatility and growth in nearly every Latin American country bears out in the time series the cross country evidence found in previous studies.

B. Inflation, Devaluation, and Fiscal Balance

Latin American countries have also experienced substantial volatility in other macroeconomic outcomes, such as inflation, exchange rates and fiscal balances. The growth literature since Fischer (1993) has treated these variables as “explanatory” variables or proxies for policies, but conventional wisdom would argue that they are endogenous and should fall into the category of “outcomes”.

Table 2 documents the level and volatility of inflation, devaluation of exchange rates, and the fiscal balance during the best and the worst decades. Average inflation, devaluation, and fiscal imbalance (as a percent of GDP) were two to four times higher during the low growth period compared with the high growth period. While devaluations may move the economy to equilibrium to ensure competitiveness and hence be associated with expansions (Edwards, 1986), large devaluations may reflect disorderly adjustments to macroeconomic imbalances, namely crises that result in output drops (see below). Table 2 indicates that relatively small devaluations were associated with higher output growth, but larger devaluations were associated with output drops.

Table 2.

Latin America: Inflation, Devaluation, and Fiscal Balance During High and Low Growth Decades, 1970-2004

article image
Sources: World Bank, World Development Indicators; IMF, World Economic Outlook; and IMF staff calculations.

Simple average.

The relationship between the fiscal balance and growth is not clear cut. If fiscal balances worsen to stimulate output to prevent a recession, output can be expected to increase. But if the worsening reflects unsustainable policies—a plausible interpretation given the 10-year periods under consideration (presumably longer than business cycle responses)—it can negatively affect output. The latter view has often been dubbed the “credibility factor” or the “expansionary effect of fiscal contraction” (see Giavazzi and Pagano, 1990). Table 2 suggests the dominance of the credibility factor since high growth periods are associated with low fiscal imbalances. In every country except Ecuador, fiscal imbalances were higher during the low growth period.4

The volatilities of inflation, devaluation, and fiscal balance were, on average, higher during the low growth period, but by a larger magnitude (three times) for inflation and devaluation. The exceptions are few, but even in those cases the volatilities across high and low growth periods do not differ much. There are significant differences among countries, however, with respect to inflation and devaluation volatility during the low growth period. Colombia, Panama, and Paraguay have relatively low volatility of monetary outcomes, while Argentina, Bolivia, Brazil, Nicaragua, and Peru experienced the highest volatility.

In sum, there is a high co-movement in the volatility of all macroeconomic outcomes, and this volatility is much higher during the low growth period compared to the high growth period in virtually all countries. However, some countries did experience low volatility in both high and low growth periods.

C. Extreme Events—Currency Crises, Banking Crises, and Sovereign Debt Defaults

Large devaluations coincident with low growth in Table 2 could reflect extreme events such as currency crises. Gupta, Mishra, and Sahay (2003) found that about 60 percent of the currency crises in a sample of nearly 200 crisis events around the world from 1970 to 2000 were contractionary. To understand better the contractions during low growth episodes in Latin America, Table 3 documents the extent of macroeconomic crises—sovereign debt defaults, currency crises or banking crises.5 Concurrent crises would in particular be expected to affect output considerably.

Table 3.

Latin America: Crises During High and Low Growth Decades, 1970-2004 1/2/

article image
Sources: Gupta, Mishra, and Sahay (2003); Chuhan and Sturzenegger (2005); and IMF desk economists.

Data are from 1975-2000 from Chuhan and Struzenegger (2003).

A dummy variable takes a value of 1 if there is a crisis in a given year for a particular country, and 0 otherwise. In the case of sovereign debt defaults, it records not only the year in which the government defaulted but also the subsequent years in which they remained in default.

Sum.

The most remarkable, though not entirely unexpected, observation is the sheer number of crises in Latin America. There were a total of 312 crisis observations (combining all three types of crises) during the 1970–2004 period, of which 223 occurred during high and low growth episodes. In other words, nearly 9 crisis observations were recorded on average each year in Latin America. Or, there were about 6 crisis observations per country per decade since 1970.

The number of currency crises and years of sovereign debt default are at least twice as high during low growth episodes, but there is little difference between the numbers of banking crises in the two sub-periods. There were 6 currency crises, 15 banking crises, and 50 debt default observations in the high growth years as compared to 34 currency crises, 14 banking crises, and 104 debt default observations in the low growth years. It may seem surprising that there were several crisis observations during the high growth period. However, the negative effect on output may not have been too high or long lasting if the severity of the crisis was small, if there were no concurrent crises, or if the policy response was quick and appropriate. One can reasonably expect the negative impact to be highest in the first year of a crisis compared with the subsequent years. Hence, it is not altogether surprising that crisis observations were recorded during the high growth period.

Every country in Latin America experienced a crisis of one kind or another. But Colombia did not experience a sovereign debt default and Panama did not experience a currency crisis (which is not surprising given dollarization). Argentina has experienced the maximum number of crises—almost one each year (there were 29 crises over the 35 year period). At the other extreme, Colombia experienced the least number of crises during the entire sample period (6). Most countries had substantially more crises during the low growth years, except Colombia, Uruguay, and Venezuela.

One reason for the association between crises and low long-term growth could be the concurrence of crises. Table 4 documents concurrent crises. The most common occurrence of simultaneous crises during 1970–2004 was currency crises and debt defaults covering 13 countries at various points in time (a total of 26 crisis-years). At the other extreme, concurrent banking and currency crises occurred in only two countries—Argentina and Uruguay.6 Banking crises and debt default together occurred in 10 countries. All three crises occurred simultaneously in only four countries.

Table 4.

Latin America: Concurrent Crises and Growth, 1970-2004 1/

article image
Sources: Gupta, Mishra, and Sahay (2003); Chuhan and Sturzenegger (2005); and IMF desk economists.

Cumulative real GDP per capita growth is reported in brackets for the year in which the crises occurred, the previous year, and the subsequent year.

Simple average.

The impact of concurrent crises on output may be measured by the cumulative effect on output one year before, during, and after the year of the crisis. Not surprisingly, the largest declines occurred when all three crises occurred simultaneously (average of nearly 9 percent fall in output). Output also fell significantly when a banking crisis or a debt default occurred at the same time as a currency crisis. Surprisingly though, a debt default combined with a banking crisis may have led to a growth deceleration but not to a fall in output.

In sum, extreme events appear to have been common in Latin America, occurring in both high and low growth periods, but twice as often in the latter. There is a wide variation in the experience across the countries, with Argentina recording the highest and Colombia the lowest number of crises. Debt defaults are the most common type of crisis, followed by currency crises and banking crises. One type of crisis by itself may not lead to a fall in output, but usually does so when combined with other types of crises.

IV. Magnitude and Volatility of Shocks

Latin America is known to be subject to frequent shocks, and the frequency and intensity of exogenous shocks—whether positive or negative—could be one reason for the large volatility of macroeconomic outcomes. Shocks could affect output directly or complicate policy making which in turn could affect output. Tables 5 and 6 document global and country-specific shocks that hit the countries in their respective high and low growth periods.

Table 5.

Latin America: Global Shocks–G7 Growth and World Real Interest Rates, 1970-2004

article image
Sources: IMF, World Economic Outlook; and Datastream.

Simple average.

Table 6.

Latin America: Country-Specific Shocks–Terms of Trade, Natural Disasters, and Official Aid, 1970-2004

article image
Sources: IMF, World Economic Outlook; and EM-DAT.

Simple average.

Two types of global shocks are considered—the growth rate of G-7 industrial countries and U.S. real interest rates. The former affects exports (a substantial part of the exports over this period were to industrial countries), while the latter affects the cost and ability to borrow abroad. Oil prices also constitute a global shock, but are captured in the country-specific terms of trade shocks.

The level and volatility of G-7 growth do not differ much between the high and low growth episodes, but US real interest rates differ significantly. The difference in G-7 growth rates between the two episodes is small (0.4 percent), and the volatility of growth, while small in both episodes, was actually higher during the high growth episode. But the average US real interest rate was significantly lower (about 500 basis points) and the volatility twice as low during the high growth years. In only three countries—Chile, Uruguay, and Venezuela—was the cost of external borrowing higher in the high growth period, which suggests that higher external borrowing costs need not be a drag on growth.

Country-specific shocks including terms of trade shocks, natural disasters, and changes in official development assistance (ODA) could also impact volatility and growth. Natural disasters are not commonly looked at, but disasters could create tremendous havoc to societies, especially smaller ones. For example, in 1998, Hurricane Mitch in Honduras and Nicaragua killed more than 10,000 people and caused as much as US$8.5 billion in damage. The levels and volatility of official development assistance, while somewhat influenced by domestic policies, are largely outside the control of country authorities (see Easterly, 2003, and Barro and Lee, 2002).

The average positive terms of trade shock during the highest growth period is surprisingly negligible (less than ½ percent). The average negative shock during the lowest growth period is also surprisingly small (just over -1 percent). However, these averages mask wide differences in individual country experiences. Nearly 6 of the 17 countries experienced worse (and negative) terms of trade shocks during their best growth decade. Of the remaining countries where the relationship between growth and terms of trade changes go in the expected direction, the difference between the positive shock during the best years and the negative shock during the worst years is significant only in Chile, Ecuador, and Mexico—all primary commodity producers. Volatility, on the other hand, is high and similar during both periods. Countries exhibiting the highest volatility are Ecuador and Paraguay (during the best growth decade) and Chile (during the worst growth decade). Thus, on average, there is no significant difference in terms of trade shocks between the two growth periods.

Measured by the number of disasters per million people affected, the smaller countries are the most affected by disasters: Costa Rica, Ecuador, Honduras, Nicaragua, and Panama. Interestingly, the number of disasters were much higher during the high growth period. This is perhaps not entirely surprising since sharp declines in output in the year of the disaster are usually followed by a rise in construction activity the following year, financed in large part by grants, and often an improvement in infrastructure that were neglected for a long time.

The relationship between foreign aid and growth has been the subject of considerable recent debate. Some studies have argued for a positive relationship between aid and growth (Hansen and Tarp, 2000, Burnside and Dollar, 2000, Dalgaard, Hansen and Tarp, 2004, and Clemens, Radelet and Bhavnani, 2004), but more recent studies have not found a strong association (Easterly, 2003, Easterly, Levine and Roodman, 2004, and Rajan and Subramanian, 2005a and 2005b). The evidence from the episodic approach confirms this latter finding, although ODA was marginally higher (less than ½ percent of GDP) during the high growth period.7 Surprisingly, the volatility of ODA was quite small, a finding that contradicts the experience of Africa and cross regional studies (Bulíř and Hamann, 2003).

To summarize, of all the exogenous shocks—global and country specific—only the U.S. real interest rate stands out as being significantly different in the highest growth episode compared to the low growth episode.

V. Macroeconomic Policy Volatility

The stance and volatility of policy are perhaps the hardest measures to conceptualize and compute. Growth regressions commonly use the fiscal balance (as a share of GDP) and inflation (or inflation volatility) as proxies for fiscal and monetary policies. However, both measures are highly endogenous, and are, at best, weak proxies.8 Even if policies could be measured, should the focus be on the discretionary component or also on the “rules”? The “rules” are important in their own right, but may differ widely across countries, making cross-country comparisons difficult. In some studies, the “rules” used in industrial countries are replicated for developing countries, which is unlikely to be appropriate (see Ter-Minassian, 2005, and Clements et al., 2006, for a criticism of such approaches).

The literature is more advanced in measuring fiscal rather than monetary policy. Even with respect to fiscal policy, however, Fatás and Mihov (2003) note that there is no consensus on the appropriate methodology for constructing a cyclically-adjusted measure because of simultaneity problems in determining output and the budget.

In this paper, measures of fiscal policy are replicated from the literature while an attempt is made to define new measures of monetary and structural policy volatility. Monetary policy volatility has been measured previously by inflation volatility. To get at the policy aspect more directly, this paper examines instead exchange rate regime changes based on the definition of exchange rate regimes in Reinhart and Rogoff (2004). Structural reform policy volatility (or reform reversals) is based on a new index that is constructed on the basis of existing measures of structural reforms from various sources and is discussed in the subsequent section.

A. Fiscal Policy Volatility

Two concepts of fiscal policy volatility are studied: the first set of measures is based on standard deviations of fiscal expenditures, while the second are measures of fiscal policy procyclicality. The extent of expenditure volatility and procyclicality during high and low growth episodes are documented in, respectively, Tables 7 and 8.

Table 7.

Latin America: Fiscal Policy Volatility During High and Low Growth Decades, 1970-2004

article image
Sources: IMF, International Financial Statistics; IMF, World Economic Outlook; and Fatas-Mihov (2003).

Using a broader measure of real fiscal expenditures (total central government expenditures), Fatas and Mihov (2003) is re-estimated over 1970-2004. The residual from the regression, which is the measure of discretionary fiscal policy, is reported.

Simple average.

Table 8.

Latin America: Fiscal Policy Procyclicality, 1970-2004 1/

article image
Sources: Kaminsky, Carmen and Vegh (2004); and authors’ estimates.

Significance levels are denoted as follows: 1 percent (***), 5 percent (**), and 10 percent (*).

Correlation of the output gap and the deviation of real central government fiscal expenditures from trend, both measured using the Hodrick-Prescott filter.

Regression of the log deviation of output from trend (output gap) and log deviation of terms of trade from trend on log deviation of real central government fiscal expenditures.

Simple average for correlation, and entire pooled sample for the Alesina-Tabellini regressions.

Standard deviation of fiscal expenditures

Two measures of standard deviations are used:

  1. the standard deviation of central government fiscal expenditures as a share of GDP (which captures both the “rules” and the “discretionary” components), and

  2. a Fatás and Mihov (2005) based measure of discretionary fiscal policy. Discretionary fiscal policy is defined as the residual from a regression of real government spending growth on current macroeconomic conditions: output growth (which is then instrumented to correct for simultaneity), lagged government spending growth and standard controls:

ΔGit=αi+βiΔYit+γiΔGit1+δiXit+εit

where Git is the log of real government consumption in country i in year t, Yit is the log of real GDP, and Xit is a set of controls (a time trend, inflation, and inflation squared). Discretion, therefore, is defined as changes in expenditure that are unrelated to macroeconomic conditions or the cycle. Volatility is defined as the standard deviation of the residual, εit.

The average level of fiscal expenditures was about 3–4 percentage points of GDP lower during the high growth decade compared with the low growth decade, while the average volatility was significantly lower (Table 7). This pattern holds in nearly all countries, although the magnitudes differ. Therefore, reducing fiscal expenditures may not hurt growth on a longer-term basis. Chile is an extreme example where fiscal expenditures were more than 10 percentage points of GDP lower in the high growth period.9

The Fatás-Mihov measure of discretionary fiscal policy volatility reveals somewhat different results—the volatility of fiscal policy is only marginally higher in the low growth period. Discretionary government expenditure can be expected to rise during the low growth period, reflected in the higher average level in the low growth decade. But there is wide variation across countries, with Nicaragua a clear outlier. Excluding Nicaragua, the average reduces in the low-growth period compared with the high growth period, suggesting the existence of financing constraints for many countries.

Expenditure procyclicality

Procyclical fiscal policies can affect long-term growth indirectly by amplifying business cycles and increasing output growth volatility. To the extent that growth volatility has a negative impact on growth, procyclical fiscal policies will adversely affect long-term growth.

Here, two measures of fiscal policy procyclicality are used: one is based on Kaminsky, Reinhart and Vegh (2004, henceforth, KRV); and the other is based on Alesina and Tabellini (2005, henceforth, AT). Specifically:

  1. KRV procyclicality measure correlates the cyclical components of real government spending and real GDP. If the correlation is positive, policy is procyclical; and

  2. the AT measure of procyclicality is obtained from country-by-country regressions of changes in government spending as a share of GDP (Sit) on the cyclical components of real GDP (or the output gap) and terms of trade.

ΔSit=ϕi+λiGDPgapit+φiToTgapit+θiSit1+μit

If the coefficient on the output gap (λit) is positive, that is, if an increase in the cyclical component of GDP is positively related to changes in spending as a share of GDP, then policies are defined to be procyclical.

Table 8 indicates that there is sufficient evidence for procyclical fiscal policies in Latin America. According to the KRV measure over the whole sample period, all countries (except Ecuador) pursued procyclical policies, of which nearly half were strongly significant. The pattern generally holds for the high and low growth subperiods. The Alesina-Tabellini measure also confirms the relationship, with higher and more significant coefficients.

B. Monetary/Exchange Rate Policy Volatility

The inability to find good measures of monetary policy is notorious in the literature. It is particularly difficult in the case of non-industrial countries where data are hard to come by or where financial markets are not well developed. Kaminsky, Reinhart and Vegh (2004) developed a measure based on the Taylor rule, but they themselves did not use it as they encountered several problems. Additionally, the Taylor rule may not be useful in a cross-country context of heterogeneous countries.

This paper develops a measure (albeit an extreme one) to capture exchange rate regime changes. Exchange rate regimes are defined according to Reinhart and Rogoff (2004). Regimes are scored on a 15-point scale on an annual basis for each country, ranging from one extreme (currency board) to the other (free floating). Based on this scoring, two variables are defined: the number of exchange rate regime changes and the intensity of regime changes for each subperiod. The intensity of change is the absolute difference in the score obtained on the 15-point scale.

A priori, the effect of exchange rate regime changes on growth is not obvious. There is some evidence that countries with fixed exchange rate regimes grow slower than those with more flexible regimes (Ghosh et al, 1997), which if true suggests that a move to greater flexibility would help countries grow faster and vice versa. On the other hand, if exchange rates are fixed and there are clear signs of overvaluation, then a move to a more depreciated rate would help growth even if the type of exchange regime itself does not change. In a similar vein, the number and intensity of changes could signify the extent of uncertainty (and hence hurt growth) or could indicate that the authorities are responding flexibly to the external environment (and hence spur growth).

Table 9 shows the number and intensity of changes in the exchange rate regime for each subperiod. The overall picture is not striking. The total number of changes are similar in the two subperiods and the number of changes per country in a 10-year period is a little over one. The intensity of changes varies widely by country, but the pattern across the two subperiods on average is not very different.

Table 9.

Latin America: Monetary Policy Volatility, 1970-2004

article image
Source: Reinhart and Rogoff (2004).

Data are from Reinhart and Rogoff (2004), where exchange rate regimes are scored on a 15-point scale. A change in their rating of an exchange rate regime is counted as a change in the exchange rate regime. The number of changes computes the number of times there were changes to the rating in the given sample period. The intensity of a change is computed as the absolute difference in the rating.

The sum is computed for the number of exchange rate regime changes while a simple average is computed for the intensity of regime changes.

VI. Structural Policy Volatility and Reform Reversals

There are several indices of structural reforms for Latin America, but it is difficult to find a sufficiently long, consistent time series for 1970–2004. Sources include the Heritage Foundation (Index of Economic Freedom measure: 1995–2005, annual data), Fraser Institute (Economic Freedom of the World measure: 1970–2003, every five years until 2000 and annual thereafter), Morley et al. (structural reform index measure—for Latin America: 1970-95, annual data), Lora (structural reform index measure—for Latin America: 1985–99, annual data), and the World Bank (Country Policy and Institutional Assessment, CPIA: 1977–2004, annual data). Given the need for annual data from 1970-2004, the measures from the Heritage Foundation and Morley et al. (1999; henceforth, Morley) were used. The CPIA was dropped because the confidentiality requirements of the World Bank prohibit the reporting of individual country data. Fraser was dropped because it does not contain annual data until 2000, while Lora’s index commences from 1985.

Combining appropriately-transformed measures of Morley for 1970–95 with the Heritage Foundation for 1995–2004 provides a consistent new series for 1970–2004 that we call the composite Heritage-Morley index.10 Nicaragua and Panama were dropped from the sample since they are not reported by Morley. The components of the two measures measure similar types of reforms: both include measures of trade policy reforms, government intervention, tax reforms, foreign investment policy, property rights, domestic financial reforms, privatization, labor market regulations, and regulatory burdens on business.11

On the whole, the region has experienced quite significant progress in structural reforms during 1970–2004, but there has been considerable variation across countries. Figure 5 illustrates the progress in structural reforms for the region and for selected countries. Argentina was the most advanced reformer in 1970, but fell behind most other countries in the region by 2004. At the other extreme, Chile was one of the laggards in 1970 but advanced to the top by 2004. Costa Rica maintained a relatively high level of reform orientation during the period, while Ecuador, which started out at the low end, progressed very slowly. Venezuela is the only country whose level of market orientation by 2004 had regressed below the level in 1970.

Figure 5.
Figure 5.

Latin America: Progress in Structural Reforms, 1970–2004

Citation: IMF Working Papers 2006, 287; 10.5089/9781451865479.001.A001

Source: Authors, Heritage-Morley Composite Index.

Figure 6 illustrates the periods of progress and regress in implementing market-oriented structural reforms for the region as a whole. The shaded area indicates the period of reversals (which is a decline in the level of the index). The area under the curve could be considered the “cost” of reform reversals—it shows the intensity of reform reversals and the time taken to revert to the level of reforms achieved before the reversals occurred. For the region, there was quite significant progress in implementing reforms particularly in the latter half of 1980s and during the 1990s. However, there were general periods of reform reversals in the early 1980s and the early 2000s. The reversals during the early 1980s coincided with a period of extreme macroeconomic instability and debt crises, against the backdrop of an unfavorable external environment. The reversals in the more recent period coincide with crises in some countries but not in others, and occurred when the external macroeconomic environment has been conducive to implementing reforms. A closer look suggests that the reversals commenced before the crises began, but were exacerbated by the crises.

Figure 6.
Figure 6.

Latin America: Structural Reforms and Reversals, 1970–2004

Citation: IMF Working Papers 2006, 287; 10.5089/9781451865479.001.A001

Source: Authors, Heritage-Morley Composite Index.

The variation across countries in the initial extent of reforms in 1970 and the pace of reforms over 1970–2004 is documented in Table 10. Countries are grouped according to high (above-median) and low (below-median) levels of initial reforms, and by fast (above-median) and slow (below-median) pace of reform during the sample period. The first number in parentheses refers to the absolute level of market orientation in 1970, while the second reports the change in the structural reform measure. Brazil is at the median in terms of both the level and the pace of reforms. The most market-oriented countries in 1970 that made the fastest progress are Costa Rica and El Salvador, while those that made the slowest progress are Argentina, Mexico, Venezuela, and Honduras. The least market-oriented countries in 1970 that made the fastest progress are Bolivia, Chile, Guatemala, Uruguay,12 and Peru, while those that made the least progress are Ecuador and Colombia.

Table 10.

Latin America: Fast and Slow Reformers, 1970-2004 1/

article image
Source: Authors.

The composite Heritage-Morley index of structural reforms is used to construct this table. Data on Nicaragua and Panama are unavailable.

Pace of reform is measured by the change in index value from 1970 to 2004.

Initial level of market orientation is measured by the value of the index in 1970.

Brazil was the median country in both the initial level of market orientation and in the pace of reform during 1970-2004. The brackets next to each country contains the initial level of the index and the change in index value from 1970 to 2004.

Not only did the initial level and the pace of reforms differ across countries in the region but the volatility of reforms also differed markedly across countries. Figures 78 plot the progress in market-oriented reforms and reform reversals for each of the countries. Argentina, Bolivia, Ecuador, Paraguay, and Venezuela stand out for their marked reversals of reforms in recent years.

Figure 7.
Figure 7.

Latin America: Structural Reforms and Reversals (Fast Reformers), 1970–2004 1/

Citation: IMF Working Papers 2006, 287; 10.5089/9781451865479.001.A001

Source: Authors, Heritage-Morley Composite Index.1/ Brazil is excluded from the fast reformers.
Figure 8.
Figure 8.

Latin America: Structural Reforms and Reversals (Slow Reformers), 1970–2004

Citation: IMF Working Papers 2006, 287; 10.5089/9781451865479.001.A001

Source: Authors, Heritage-Morley Composite Index.

All countries experienced reform reversals during 1970–2004. There were 167 reversals in total, excluding Nicaragua and Panama. The highest numbers of reversals were in Argentina, Brazil, Colombia, and Venezuela. The most intense reversals were in Argentina, Chile (during the early part of the sub-period), Paraguay, and Venezuela. The least intense were in Costa Rica, El Salvador, Guatemala, and Uruguay.

Structural reform performance was better during the high-growth episode compared with the low-growth episode. Table 11 documents the average level and trend (or progress) in structural reforms as well as the number and intensity of reform reversals during the high and low growth episodes in each country.13 During the high-growth episode, the degree of market orientation of structural policies was higher, countries made more progress in structural reforms, the number of reform reversals was fewer, and the intensity of reform reversals was lower.

Table 11.

Latin America: Structural Reform Levels, Trends, and Reversals During High and Low Growth Decades, 1970-2004 1/

article image
Source: Authors, the Heritage-Morley Composite Index.

The index ranges from 0 to 100, with a higher number denoting a more reformed economy. Data from 1995-2004 are from the Heritage Foundation, normalized to range from 0 to 100. Data from 1970-1994 are from Morley (1999), normalized to range from 0 to 100, and normalized such that the numbers in 1995 correspond to the Heritage Foundation 1995 numbers.

The low-medium-high characterization is based on whether the intensity of reversals is in the 1st, the 2nd and 3rd quartiles, or the 4th quartile, respectively.

Simple average, except for number of reversals where the total number of reversals is provided.

VII. Constraints Faced by Policymakers

Policy choices in implementing reforms or responding to shocks may be constrained if country authorities face financing, trading or institutional constraints. If two countries faced the same shock, the volatility of policies could be higher in one country if the policymakers in that country faced greater constraints. This section examines three types of constraints: financing constraints, restrictions in external trade, and socio-political features.

A. Financing Constraints

Financing constraints could be measured by access to external capital markets, financial liberalization, capital flows, depth of domestic financial sectors, and levels of public debt. Table 12 documents these measures of financial integration and constraints. The first column indicates a country’s access to external private capital markets. This measure, taken from Gelos, Sahay, and Sandleris (2004), is an annual binary variable that takes the value of 1 if the country had access to capital markets, zero otherwise. For example, during its best growth period, Argentina had access to external capital market for 70 percent of the time, while during its worst growth period, it had access 40 percent of the time. On average, Latin American countries had greater access to capital market during their high growth periods. Chile and Uruguay are the only countries that had lower access to external capital during their high growth period.

Table 12.

Latin America: Financial Integration and Constraints During High and Low Growth Decades, 1970-2004

article image

Dummy variable covering 1980-2000, with 1 indicating access to international capital markets. Source: Gelos, Sahay, and Sandleris (2004).

Dummy variable from 1970-2000, with 1 indicating a liberalized financial sector. Source: Kose, Prasad, and Terrones (2006).

The ratio of capital flows to GDP. Data covers 1970-2000. Source: Kose, Prasad, and Terrones (2006).

Data covers 1970-2004. Sources are WDI and WEO.

Data covers 1970-2003. Source is WDI.

Simple average.

The second column indicates the degree of financial liberalization, which is also a one-zero annual dummy variable. More financially liberalized countries experienced higher growth. Only Brazil, Colombia, Mexico and Venezuela were financially less liberalized in their high growth period, but three of these four countries experienced their highest growth period during the 1970s when the liberalization was lower in the world, in general.

The third column indicates capital flows as a share of GDP. On average, capital flowed into the countries during their high growth periods and flowed out during the low growth periods. Colombia, Nicaragua, Paraguay, and Venezuela are exceptions, however. The penultimate column provides a measure of financial depth, namely, the M2-to-GDP ratio. On average, financial depth was the same during the high and low growth periods. On the other hand, debt was higher in the low growth period in virtually all countries, though not significantly so (exceptions are Colombia, Nicaragua, Uruguay, and Venezuela).

B. Trade Integration

Trade openness is expected to be related to higher growth (Dollar and Kraay, 2002, and Chang et al., 2005). Two measures of trade integration are considered: a binary trade liberalization dummy and trade (exports+imports) as a share of GDP. Trade is indeed more liberalized and the economy more open on average during periods of high growth than periods of low growth, but only marginally so (Table 13). Country experiences vary significantly. Five countries—Colombia, Costa Rica, Ecuador, Mexico, and Paraguay—grew faster when their trade regimes were much less liberal, though the best years for each of these countries were in the 1970s when trade regimes in the world were generally less free. Measures of trade openness have been criticized, however, for not capturing the channels through which trade liberalization facilitates growth (see, for instance, Srinivasan and Bhagwati, 2001). Hence, such measures need to be viewed with some caution.

Table 13.

Latin America: Trade Liberalization and Integration During High and Low Growth Decades, 1970-2004

article image
Sources: Kose, Prasad and Terrones (2006); and IMF, World Economic Outlook.

Trade liberalization is a 0-1 dummy from Kose, Prasad and Terrones (2006), where 1 denotes liberalized regime. The data cover 1970-2000.

Trade openness is exports plus imports as a share of GDP, from Kose, Prasad, and Terrones (2006) from 1970-2000 and from WEO from 2001-04.

Simple average.