International Reserves in Low Income Countries
Have they Served As Buffers?

Contributor Notes

Authors’ E-Mail Address: vcrispolti@imf.org; gtsibouris@imf.org

This paper provides a historical perspective on the role of international reserves in low-income countries as a cushion against large external shocks over the last three decades - including the current global crisis. The results suggest that international reserves have played a role in buffering external shocks, with the resulting macroeconomic costs varying with the nature of the shock, the economy's structural characteristics, and the level of reserves.

Abstract

This paper provides a historical perspective on the role of international reserves in low-income countries as a cushion against large external shocks over the last three decades - including the current global crisis. The results suggest that international reserves have played a role in buffering external shocks, with the resulting macroeconomic costs varying with the nature of the shock, the economy's structural characteristics, and the level of reserves.

I. Introduction

1. This paper is part of the International Monetary Fund’s multi-departmental research project to study reserve adequacy in low income countries (LICs).1 It provides a historical perspective on the role of international reserves as a buffer against large external shocks over the course of the last three decades—including the current global crisis. In particular, the paper seeks to assess whether the macroeconomic costs associated with external shocks were larger in LICs that had lower international reserve holdings prior to a shock event. The following questions are thus addressed: (i) what are the relevant external shock episodes in LICs?; (ii) what are the macroeconomic costs associated with them?; (iii) do such costs vary depending on structural characteristics of the economy, including the level of international reserves?

2. Recent studies on LICs suggest that international reserves may effectively help limit the macroeconomic volatility stemming from exogenous shocks. Though LICs are subject to a wide variety of shocks, it is generally recognized that they are particularly vulnerable to external shocks and natural disasters (Becker et al. 2007). Moreover, the economic costs associated with such shocks are large and seem to vary with the structural characteristics of the economy (Berg et al., 2011). Against this background, international reserves may play an important role in mitigating the impact of shocks and containing macroeconomic volatility (IMF, 2011; Drummond and Dhasmana, 2008).

3. The paper extends previous research on the role of international reserves by examining a wide range of external shocks over the last three decades and by differentiating LICs according to their structural characteristics. For the period 1980–2007, an event-study analysis approach is used to determine the losses in terms of forgone growth of real GDP and consumption per capita associated with different types of shocks (i.e. external-demand, terms-of-trade, climatic, FDI, and aid shocks) and structural characteristics of the economy (i.e. exchange rate regime, export and import concentration, level of indebtedness, and presence of an IMF program). Such losses were then compared across countries with different international reserve holdings in the year prior to the shock episode. For the current crisis period (2008–2010), a four-year event window centered in 2008 was used to assess the impact of the current crisis on several key macroeconomic variables, including real GDP, real per-capita consumption, real per-capita investment, and external current account.

4. The structure of the paper is as follows. Section II describes the methodology used to identify shock episodes during the period 1980–2007, presents the results of the event study analysis for LICs, and checks for robustness. Section III focuses on the current global crisis (2008–2010). Section IV provides concluding remarks.

II. II. Event Study Analysis: 1980–2007

5. This section evaluates the macroeconomic consequences of a range of external shocks faced by LICs in relation to their levels of reserve coverage during the period 1980–2007. For each country, shock episodes are identified and associated macroeconomic costs are derived, while differentiating LICs by their structural characteristics, such as the exchange rate regime, export and import concentration, debt levels, and the presence of a financial arrangement with the International Monetary Fund (IMF).

A. Methodology

6. An external shock event is defined as a “significant worsening” in a country’s climatic conditions, terms of trade, external demand, FDI, or aid inflows.2 More precisely, a shock occurs if the annual percentage change of the shock variable (e.g. terms of trade) falls below the bottom 10th percentile of its country-specific distribution (Figure 1).3,4 While alternative definitions of shock events exist in literature, this one presents several advantages including: (i) controlling for heterogeneity among LICs by using country-specific distributions; (ii) capturing rare events by focusing on the bottom 10th percentile of distribution for the shock variable; and (iii) centering the analysis on the reaction to the shock by assuming the same frequency of shocks for each country.5

Figure 1:
Figure 1:

Definition of Shock Episodes

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

7. For each shock event, the analysis identifies its economic impact and measures its cost in terms of forgone growth of real GDP and per-capita consumption.6 In particular, the economic impact of a shock was evaluated by means of a five-year event window capturing the behavior of the relevant macroeconomic variables from one year before the shock episode to three years after. By contrast, the economic cost associated with a shock was determined as the cumulated sum of the estimated growth losses with respect to pre-shock trends over the years following a shock episode.7 The estimated growth loss in terms of GDP or per-capita consumption is computed as the negative difference between their “shock” growth index—based on actual growth––and the “no-shock” growth index––reflecting trend growth of years in which a (Figure 2).8 The number of years in which a country experienced a loss may be interpreted as a measure of duration of the economic cost associated with a shock. Both the impact and the cost of shocks in LICs were examined by differentiating along the structural characteristics of the economy, including the exchange rate regime, export and import concentration, debt levels, and the presence of an IMF program.9

Figure 2:
Figure 2:

Identifying the Cost Associated with a Shock 1

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

1 A negative number represents a loss. The cumulated sum of the annual losses represents the cost in percentage points associated with a shock. In this example, the duration is seven years.

B. Anatomy of Shocks

8. Between 1980 and 2007 LICs were confronted with a large shock event every ten years on average (Table 1).10 The likelihood of a shock to FDI inflows was about sixteen percent, while that of a significant worsening in the climatic conditions or the terms of trade was respectively thirteen and eleven percent. By contrast, shocks to external demand were less frequent. The size of shocks varied depending on their nature. Typically, shocks to external demand and terms of trade had a magnitude of about two standard deviations from the sample mean.11 On the other hand, aid and FDI shocks had a somewhat smaller size, about one standard deviation equivalent to a decline of 5 and 3 percentage points of GDP respectively.

Table 1:

Frequency and Size of Shocks

(average values)

article image

Changes in percentage points of GDP.

Source: April 2010 WEO, IMF.

9. The geographical distribution of shocks showed noteworthy differences (Table 2). Shocks to external demand occurred more frequently in the Middle East, East Asia, and Sub-Saharan Africa. On the other hand, significant changes in the terms of trade and climatic conditions were more likely to happen in the Middle East and East Asia; while, FDI shocks were prominent in South Asia and Sub-Saharan Africa. Finally, drops in aid flows were important in Eastern Europe and Latin America. In terms of size, external demand shocks amounted to almost two standard deviations in Eastern Europe and Middle East; while, shocks to the terms of trade were relatively larger in Middle East, Latin America, and Sub-Saharan Africa. By contrast, the size of FDI and aid shocks significantly exceeded the sample average in East and South Asia, as well as in Latin America.

Table 2:

Frequency and Size of Shocks by Region

(average values)

article image

Changes in percentage points of GDP.

Source: April 2010 WEO, IMF.

C. Macroeconomic Impact and Cost of Shocks

10. Shock episodes were accompanied by a visible deterioration of the macroeconomic situation (Figure 3). The median GDP growth rate decelerated in the aftermath of all types of shocks, particularly after the largest ones (i.e. terms of trade and external demand shocks).12 Similarly, median real per-capita consumption growth remained subdued over the event window, more so in the presence of a significant worsening in external demand. Against this background, the current account deficit generally widened and remained depressed for about two years before returning to pre-shock levels. Reserve coverage in months of imports showed different trends. It fell following shocks to external demand and the terms of trade, possibly reflecting the use of international reserves to finance imports when sources of foreign exchange are limited (Table A.5). By contrast, reserve coverage increased steadily after a shock to FDI and aid flows, suggesting that, in LICs, foreign financing flows have large import content.

Figure 3:
Figure 3:

Macroeconomic Impact of Shocks

(annual percent changes unless otherwise indicated; median values)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: April 2010 WEO, IMF.

11. Structural vulnerabilities exacerbated the impact of shocks (Table A.5 and A.6). External demand shocks had a larger macroeconomic impact in highly-indebted and commodity-exporting economies, mainly reflecting their greater dependence on exports as source of financing.13 After a terms-of-trade shock, countries with a fixed exchange rate regime showed a protracted slowdown in economic activity and a decline in per-capita consumption. On the other hand, the macroeconomic consequences of climatic shocks were more visible in islands and commodity-exporting countries, perhaps reflecting their fairly concentrated production base. Finally, a significant worsening in FDI and aid flows had a fairly prolonged impact in countries without an IMF program around the time of the shock episode.

12. The macroeconomic costs associated with external shocks were large and persistent (Figure 4). In presence of an external-demand shock, cumulative losses expressed as forgone GDP growth were as high as 2 percentage points of GDP over four years. The macroeconomic costs related to a significant worsening in FDI inflows and climatic conditions were typically modest, approximately 0.2 percentage points of GDP growth in a year. By contrast, shocks to the terms of trade were not accompanied by any output loss. Costs in terms of foregone growth of real per-capita consumption were larger. After a terms-of-trade shock, consumption growth remained below trend for thirteen years resulting in a cumulated loss of 18 percentage points. While this result seems to overstate the cost of termsof-trade shocks in LICs, previous research suggests that such shocks have a prolonged impact on households’ purchasing power and hence on consumption (IMF, 2008). In the case of a shock to external demand, costs amounted to 7 percentage points over six years; whereas cumulative losses were rather modest and short-lived under FDI and climatic shocks.

Figure 4:
Figure 4:

Costs by Type Shocks

(losses with respect to pre-shock trend in percentage points; median values)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: April 2010 WEO, IMF.

13. The magnitude and persistence of costs varied significantly depending on the structural characteristics of the economy (Figure A.1 and A.2). Generally, costs in terms of real GDP growth were larger and more persistent in islands, as well as commodity-exporting and -importing economies. Countries with a flexible exchange rate regime prior to a shock event showed hardly any loss, highlighting the role of exchange rate as shock absorber. Interestingly, an early program engagement with the IMF in the proximity of a shock event was generally associated with very limited losses. Finally, costs related to a worsening in climatic conditions were fairly large in island economies with cumulative losses reaching approximately 14 percentage points of GDP after a decade. The analysis in terms of real per-capita consumption growth gave similar results. Namely, countries with a fixed exchange rate regime normally suffered larger consumption losses than economies with a flexible regime—confirming previous findings on the role of the exchange rate. Also, consumption sharply dropped with respect to the pre-shock trend in commodity-exporting and -importing countries following shocks to external demand and the terms of trade. In islands and highly-indebted economies, costs associated with a significant worsening in the terms of trade exceeded 20 percentage points of consumption in ten years. Notably, the presence of an IMF-supported program did not produce tangible benefits in shielding consumption, perhaps reflecting the role played by fiscal consolidation in such arrangements.

D. The Role of International Reserves

14. International reserves holdings helped contain the economic costs associated with an external shock. In particular, the role of international reserves was assessed by grouping LICs according to whether their reserve coverage was strictly above or below three months of imports in the year preceding the shock event and by comparing losses between the two groups. 14 Under most shocks, annual losses declined as reserve holdings approached three months of imports, suggesting that in LICs the standard rule of thumb of three months of imports is at most a lower bound (Figure 5).

Figure 5:
Figure 5:

Annual Losses and Level of Reserves 1

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

1 A negative number represents a loss.Source: April 2010 WEO, IMF.

15. Countries with higher levels of international reserves in the year preceeding a shock event were able to better cushion economic activity (Figure A.3 and A.5).15 In terms of real GDP growth, countries with reserve coverage above three months of imports did not suffer any losses except in the case of shocks to FDI inflows (Figure 6). On the contrary, economies with less than three months of imports generally encountered significant costs—particularly in the face of an external demand shock with cumulative losses reaching approximately 13 percentage points over eight years. Higher reserves were also associated with a smoother adjustment of real per-capita consumption growth (Figure 6). Except for the case of FDI shocks, costs in terms of forgone consumption growth were generally about 2 percentage points in countries with more than three months of imports. By contrast, countries with lower reserve coverage presented an average loss of about 1 percentage point for nearly 6 years. In the case of a terms-of-trade shock, cumulative losses were more than twenty times bigger than those of economies with more than three months of imports, suggesting that benefits of entering a shock episode with higher reserve holdings may be substantial.

Figure 6:
Figure 6:

Costs of External Shocks by Variable and Type of Shock

(losses computed with respect to pre-shock trend in percentage points; median values)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: April 2010 WEO, IMF.

16. The structural characteristics of the economy had a bearing on the role of international reserves. Reserve coverage above three months of imports was not a safe haven when it came to GDP losses—particularly for island economies and countries with an exchange rate peg in the face of a terms-of-trade shock (Figure A.3 and A.4). Similarly, cumulative losses were significant in oil-exporting and commodity-importing countries after a significant worsening of external demand. By contrast, reserves above three months of imports appeared to be an adequate cushion against all shocks in highly-indebted economies. In terms of per-capita consumption growth losses, international reserves seem to have allowed for consumption smoothing (Figure A.5 and A.6). Under a fixed exchange regime, costs were somewhat smaller in countries with more than three months of imports prior to the shock episode. Among countries with reserve coverage below three months of imports, losses were larger in commodity-importing countries, islands, and highly-indebted economies. Interestingly, countries without an IMF program generally experienced a more prolonged adjustment than countries with an IMF program in the proximity of a shock episode.

E. Robustness Analysis

17. The robustness of results was determined by using alternative definitions of shocks and related losses. In particular, shocks were redefined in terms of: (i) a different threshold (i.e. 25th percentile) to capture a broader range of events, while controlling for heterogeneity among LICs; and (ii) a different distribution (i.e. whole-sample) to abstract from a country’s specific circumstances, while maintaining the focus on rare events. Furthermore, costs associated with shocks were recomputed differentiating between “closed” and “open” events in each country in order to assess whether permanent factors might bias the analysis. A “closed” event is defined as a situation in which a country’s “shock” growth index catches up with its “no-shock” index within the sample period (Figure 2). By contrast, an “open” event (corresponding to some notion of permanence) is identified by a case in which the “shock” index is consistently above or below the “no-shock” index.

18. The results based on the 25th percentile threshold confirm previous findings.16 Compared to the analysis based on the 10th percentile threshold, shock episodes were more frequent and had a smaller magnitude, but their relative importance did not change significantly (Table 3). Also, the role of international reserves was broadly confirmed though the magnitude and duration of costs differed somewhat from previous analysis (Figure 7). Namely, countries with reserve coverage above three months of imports prior to the shock event did not experience any loss in terms of forgone GDP growth after an external demand shock. Similarly, countries with higher reserve buffers suffered lower consumption losses under all shocks, particularly terms of trade shocks.

Table 3:

Frequency and Size of Shocks (25th Percentile)

(average values)

article image

In percentage points of GDP.

Source: April 2010 WEO, IMF.
Figure 7:
Figure 7:

Costs of External Shocks by Variable and Type of Shock (25th percentile)

(losses computed with respect to pre-shock trend in percentage points; median values)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: April 2010 WEO, IMF.

19. Looking at shock episodes from the entire sample distribution did not alter the results.17 In this case, shocks were less frequent but of larger magnitude, reflecting the fact that the bottom 10th percentile of the whole-sample distribution comprises events that are generally larger than those falling below the bottom 10th percentile of country-specific distributions (Table 4). Results based on this approach were broadly in line with the previous analysis, including on the potential role of reserves as a buffer against external shocks. Costs in terms of GDP growth were nil for countries with reserve coverage above three months of imports prior to the shock event (Figure 8). Likewise, economies with more than three months of imports were able to smooth consumption in the aftermath of shocks to the external demand and the terms of trade.

Table 4:

Frequency and Size of Shocks (Whole Sample Distribution)

(average values)

article image

In percentage points of GDP.

April 2010 WEO, IMF.
Figure 8:
Figure 8:

Costs of External Shocks by Variable and Type of Shock (Whole Sample Distribution)

(losses computed with respect to pre-shock trend in percentage points; median values)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: April 2010 WEO, IMF.

20. The analysis did not seem to be biased by permanent factors. Specifically, cumulative losses associated with “closed” events were broadly consistent with those presented in section D. However, the magnitude of losses was generally smaller and their duration shorter, particularly under external-demand shocks for GDP losses and under terms-of-trade shocks for consumption losses (Figure 9). This suggests that the role of international reserves as a buffer against external shocks is not significantly affected by the inclusion of permanent factors (i.e. “open” events) in the analysis.

Figure 9:
Figure 9:

Macroeconomic Performance by Level of Reserves and Structural Characteristics (“Closed” Events)

(losses computed with respect to pre-shock trend in percentage points; median values)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: April 2010 WEO, IMF.

III. Global Financial Crisis: 2008–2010

21. This section focuses on the shock episode related to the current global crisis. Taking 2008 as the shock year, it evaluates the macroeconomic impact of the crisis and the potential role of foreign exchange reserves as a buffer against external shocks while differentiating by LICs’ structural characteristics.

22. The global crisis had a significant impact on LICs (Figure 10). In 2008, these economies faced an unprecedented surge in oil and food prices, which increased countries’ policy challenges to maintain macroeconomic stability (IMF, 2008). In 2009 and early 2010, the spillovers from the financial crisis in the advanced countries put LICs under additional strain as external demand and sources of foreign financing declined markedly (IMF, 2009). As a result, in LICs the median GDP growth and real per-capita consumption dropped by respectively 3 and more than 4 percentage points between 2007 and 2009. Also, external imbalances widened markedly. In 2008, the median current account deficit deteriorated by about 3 percentage points of GDP, mainly reflecting the sharp worsening in the terms of trade. Though international prices softened thereafter, external accounts remained under pressures due to the adverse spillovers from the global financial crisis. Against this background, median reserve coverage experienced large swings. In 2008, it fell by almost one month of imports following a lower accumulation of international reserves and higher import prices (Table A.8).18 In 2009, median coverage increased by one and a half months of imports as countries adjusted to the shock and the SDR allocation from the Fund became available to all IMF members.19

Figure 10:
Figure 10:

Macroeconomic Impact of the Crisis

(median values)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: September 2011 WEO, IMF.

23. International reserves helped LICs withstand the impact of the global crisis (Figure 11). Countries that entered the crisis with reserve coverage above three months of imports were able to better buffer economic activity and to smooth consumption than economies with a lower level of international reserves in 2007. In these countries, the median accumulation of international reserves came almost to halt in 2008 falling to 3 percent (30 percent in 2007) and remaining consistently below that of economies with fewer reserves thereafter (Table A.8). Interestingly, economies with a slender level of reserves showed a sharper correction in median real per-capita investment, possibly indicating that absent a comfortable level of reserves LICs curtailed investment in an attempt to protect consumption.

Figure 11:
Figure 11:

Macroeconomic Impact of the Crisis by Level of Reserves

(median percent changes)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: September 2011 WEO, IMF.

24. As in the event study analysis, the potential benefits of holding international reserves differed according to the structural characteristics of the economy (Figure 12). Median GDP growth declined in island economies with reserve coverage below three months of imports prior to the shock event, while remaining largely positive in islands with higher reserves (Table A.7). Similarly, a rather slow pace of economic growth was experienced in countries with high debt, fairly concentrated exports, and no financial arrangement with the IMF. In terms of real per-capita consumption growth, highly-indebted countries with reserve below three months of imports confirmed their vulnerability to external shocks. By contrast, higher reserve coverage could not prevent a significant adjustment of consumption patterns in commodity-importing and island economies.

Figure 12:
Figure 12:

Macroeconomic Performance by Level of Reserves and Structural Characteristics

(average annual percent change between 2008 and 2010, unless otherwise indicated)

Citation: IMF Working Papers 2012, 007; 10.5089/9781463930561.001.A001

Source: September 2011 WEO, IMF

IV. Conclusions

25. This paper investigates the issue of reserve adequacy in LICs. Specifically, it examines the role of international reserves as a buffer against different types of shocks (i.e. external-demand, terms-of-trade, climatic, FDI, and aid shocks) over the last three decades. The paper expands on previous research by examining a wide variety of external shocks and explicitly taking into account LICs’ different structural characteristics (i.e. exchange rate regime, export and import concentration, level of indebtedness, and presence of an IMF program).

26. The event study analysis for 1980–2007 indicates that LICs faced large external shocks, whose macroeconomic costs varied based on the nature of the shock, the structural characteristics of the economy, and the level of international reserve holdings. During the three years following any shock event, macroeconomic performance weakened significantly with median real GDP growth and external accounts deteriorating—particularly in countries with a high level of debt, fairly concentrated export base, and limited exchange rate flexibility. Also, costs with respect to pre-shock trends were considerable both in terms of forgone real GDP and per-capita consumption growth. After a terms-of-trade shock, costs were rather large and persistent in commodity-exporting economies (including oil-exporters), islands, and countries without an IMF program in the proximity of the shock event. Against this backdrop, countries with international reserve holdings above three months of imports in the year before the shock event were generally able to better cushion economic activity. However, higher reserve coverage was not panacea—especially in countries with an exchange rate peg—suggesting that the standard rule of thumb of three months of imports is at most a lower bound and that the appropriate level of reserve holdings may vary with the structural characteristics of the economy.

27. International reserves helped LICs withstand the impact of the current global crisis (2008–2010). As the world faced an unprecedented surge in commodity prices and the challenges faced by financial markets in advanced countries, LICs with reserve coverage above three months of imports in 2007 were able to better cushion GDP growth and consumption from the spillovers of the crisis. Nonetheless, the potential benefits of holding international reserves varied markedly depending on the structural characteristics of LICs.

VI. Appendix

This study is based on a panel of 71 LICs comprising 17 island economies, 17 commodity exporting countries, 6 oil exporters, 20 commodity importers, and 64 highly-indebted economies (Table A.1). The sample period spans from 1980 to 2007 for the event study analysis and from 2007 to 2010 for the study on the impact of the current crisis. The data used for the event study analysis (1980–2007) mainly come from the April 2010 WEO database with a few exceptions that are documented in Table A.1. To reduce the number of missing observations, historical data on per-capita consumption were reconstructed using growth rates from the Penn World Tables (version 6.3). For the current crisis period, the April 2010 WEO data were updated with the September 2011 WEO database. All computations were performed with Stata® 11/SE and all code is available on request.

Table A.1:

List of Variables

article image
Table A.2:

List of Countries and Structural Characteristics of the Economy

article image

As defined by the April 2010 World Economic Outlook.

Countries that on average had a share of commodity imports to total merchandise imports exceeding 40 percent over the sample period.

Countries that on average had a debt-to-GDP ratio exceeding 50 percent over the sample period.

Table A.3:

Shock Variables (1980–2007)-Summary Statistics

(average percent changes)

article image
Source: April 2010 WEO, IMF.
Table A.4:

Macroeconomic Variables (1980–2007)-Summary Statistics

(percent changes unless otherwise indicated; average values)

article image
Source: April 2010 WEO, IMF.
Table A.5:

Macroeconomic Impact of Shocks

(annual percent changes unless otherwise indicated; median values)

article image
Source: April 2010 WEO, IMF.