I. Introduction

Commodity exporting countries across the world have benefited largely from the commodity price boom of the last decade. Together with improved macroeconomic management, the associated terms-of-trade boom allowed many emerging market economies to strengthen their economic fundamentals markedly in many dimensions. Latin America—being an important commodity producing region—has witnessed a particularly stark transformation. These developments have fed a growing sense of complacency in the region that this time the macroeconomic response to the terms-of-trade shock has been different, and more prudent, than in past episodes. Whether this is the case, however, remains an open empirical question, that this paper seeks to shed light on. Specifically, we study the current episode—that started around 2002—from a historical perspective, with a focus on the associated terms-of-trade income windfall and the extent to which it has been saved. We focus on Latin America, but draw intra-regional comparisons when relevant.

Some recent studies (IADB, 2008; Izquierdo et al, 2008; Osterholm and Zettelmeyer, 2008; Cespedes and Velasco; 2011; etc.) have looked into the role of external factors—including terms-of-trade shocks—in driving economic activity in Latin America. However, by focusing primarily on the effects of these shocks on output rather than income, they have overlooked an important dimension of the story that is key to assess whether the macroeconomic response to the shock was indeed different this time.

Against this backdrop, we offer two key innovations vis-à-vis these studies. First, we develop a simple, but quite telling, metric of the ‘extraordinary’ income (windfall) arising from the terms-of trade shocks (vis-à-vis a counterfactual) that allows us to grasp the importance of the recent boom in absolute magnitude as well as in historical perspective. With this measure at hand, we then compute marginal savings rates (i.e., the proportion of the windfall saved) to shed light on whether the ‘effort’ to save the windfall has indeed been different this time. Finally, we use these estimates to study the extent to which saving during the boom, and its different types (i.e., domestic versus foreign), affect post-boom income.

We find that the recent terms-of-trade shocks in Latin America have not been larger than those seen during the 1970s. At the same time, the associated income windfall of the recent boom has been considerably larger than in the past, and quite large in absolute terms—averaging close to 15 percent of domestic income on an annual basis and about 100 percent on a cumulative basis. This windfall is only comparable to that received by Middle-East and North Africa (MENA) oil exporting countries. This is our first contribution.

Second, we find that while aggregate (average) savings increased more than in past episodes, the share of the income windfall saved (i.e., the marginal saving rate) has been lower. Together, these findings challenge the growing perception that the macroeconomic response was different this time. Furthermore, it suggests that the marked improvements in fundamentals largely reflect the sheer size of the windfall, rather than a greater effort to save it. At the same time, there is evidence indicating that public sectors have contributed more to saving the income windfall than the private sector this time.

Another important difference with respect to past episodes is that, this time, a larger share of saving has been allocated to domestic investment (as opposed to accumulation of foreign assets). Related to the latter, we find econometric evidence suggesting that, in past episodes, boom-time saving resulted in higher post-boom income but saving tilted towards domestic investment—as seen recently—was been less beneficial to post-boom income than saving in foreign assets. This finding raises questions about the ongoing weakening of external current account balances in Latin America.

The rest of the paper is organized as follows: Section II describes our metric of income windfall and compares Latin America’s current episode with those of other regions, as well as its own past episodes. Section III studies the extent to which windfalls were saved. Section IV presents the econometric exercise to assess the importance of saving the windfall, and its composition, for post-boom income. Section V concludes with a discussion of the key takeaways.

II. The Terms-of-Trade Windfall in Historical Perspective

We study the history of large terms-of-trade shocks, focusing on the period 1970-2012 on a sample of 180 countries. The main data source is the IMF, International Financial Statistics (World Economic Outlook). Our main exercise entails identifying episodes of terms-of-trade booms over the collected sample, as follows:

Terms-of-trade Boom. Define terms of trade as TTⁱ=Pⁱ_X/Pⁱ_M, where Pⁱ_X (Pⁱ_M) stands for the price in U.S. dollars of exports (imports) of goods and services in a given year. Let S (P) be the year in which the boom starts (peaks) identified as local min (max). A terms-of-trade boom episode is an event for which the following conditions hold:

1. The terms of trade increase at least 15 percent from start to peak, (i.e., $\left(\frac{{\mathit{TT}}_P-{\mathit{TT}}_S}{{\mathit{TT}}_S}\right)\ge 0.15)$; and

2. The annual average increase in terms of trade is of at least 3 percent (i.e., $\left(\frac{{\mathit{TT}}_P-{\mathit{TT}}_S}{{\mathit{TT}}_S*\left(P-S\right)}\right)\ge 0.03$.

These thresholds identify 270 episodes, encompassing low-income countries, emerging market economies, and advanced economies (see Appendix Table A1 for the full list of episodes).

A first interesting result from exploring the sample of identified episodes is that the magnitude of the terms-of-trade shocks—measured during the upswing—under the recent boom has not been larger than those seen in the 1970s, as shown in Figure 1 (upper left chart). The recent wave of terms-of-trade shocks, however, has been more widespread (upper right chart) with twice as many episodes as in any of the previous decades, including the 1970s. This likely reflects the widespread increase in commodity prices over the past 10 years. Another key difference with past episodes, is that the recent wave of terms-of-trade shocks has been more persistent. Indeed, the average upswing of the latter is 7 years, compared to 3 years in past episodes (lower left chart and annex Table A3).

Key Features of Terms-of-trade Boom Episodes, 1970-2012^1/

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: Authors’ calculations.^1/ Statistics across all ongoing episodes (upswing) at each point in time.2/ End is reached when at least ⅓ of the increase in terms of trade is reverted (or a new boom takes place). For episodes that did not reach their end, by the end of the sample period, 2012 is taken as the episode end year.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Key Features of Terms-of-trade Boom Episodes, 1970-2012^1/

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: Authors’ calculations.^1/ Statistics across all ongoing episodes (upswing) at each point in time.2/ End is reached when at least ⅓ of the increase in terms of trade is reverted (or a new boom takes place). For episodes that did not reach their end, by the end of the sample period, 2012 is taken as the episode end year.

• Download Figure
• Download figure as PowerPoint slide

Key Features of Terms-of-trade Boom Episodes, 1970-2012^1/

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: Authors’ calculations.^1/ Statistics across all ongoing episodes (upswing) at each point in time.2/ End is reached when at least ⅓ of the increase in terms of trade is reverted (or a new boom takes place). For episodes that did not reach their end, by the end of the sample period, 2012 is taken as the episode end year.

• Download Figure
• Download figure as PowerPoint slide

Looking beyond the upswing, we measure the end of the terms-of-trade boom (lower right chart), as the point in time at which at least a ⅓ of the increase during the upswing is reverted (or a new boom starts). We find that, even if the ongoing boom were to find an abrupt end, the full length from start to end of the episode (i.e., cycle) would still be much longer than those in the past, reflecting in part the relatively rapid reversal of booms in the 1970s and 1980s (on average. past episodes met their end only 1 year after reaching their peaks).

In the case of Emerging Latin America—Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Mexico, Paraguay, Peru, Uruguay, and Venezuela—the data show that within this group, only 9 countries satisfy the above thresholds, delivering 10 episodes during the 2000s. Interestingly, Uruguay and Mexico do not qualify as cases of recent terms-of-trade booms, reflecting the fact that, while being producers of certain commodities, their net commodity exports are relatively small. At the same time, we identify two episodes for Paraguay since 2000.

Similar to the general result highlighted above, Latin America’s current terms-of-trade boom—which started around 2002—is comparable in magnitude and scope (i.e., number of countries experiencing one) to those seen in the 1970s. Identified episodes have a median terms-of-trade cumulative increase of about 80 percent (during the upswing), although there is also significant variation across countries within the group. For example, Venezuela and Bolivia show cumulative terms-of-trade shocks of more than 120 percent while Brazil is on the other extreme, with a shock of only about 35 percent. This compares to 11 episodes in the 1970s, with a median cumulative increase of about 70 percent.

Furthermore, a contemporaneous comparison across regions indicates that Latin America’s recent shock is only comparable to those of the oil exporting countries of the Middle East and North Africa region (Figure 2) and, to a lesser extent, to those of commodity exporting advanced economies (Australia, Canada, and Norway). Indeed, for other emerging market regions, we identify a limited number of episodes (e.g., Emerging Asia) and of much lower magnitude (e.g., Emerging Asia and Emerging Europe).¹

Emerging Latin America and Selected Regions: Terms-of-Trade Booms, 1970-2012¹

(Cumulative percentage change, during upswing)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF, International Financial Statistics; and authors’ estimates.¹Cumulative percentage change in terms of trade (of goods and services) from start to peak of each identified episode (that meet the criteria of at least 15 percent cumulative and 3 percent average increase). Episodes are grouped in 5-year windows according to the date of their first year. Dotted lines indicate group averages. Country groups as defined in Annex Table A2.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Emerging Latin America and Selected Regions: Terms-of-Trade Booms, 1970-2012¹

(Cumulative percentage change, during upswing)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF, International Financial Statistics; and authors’ estimates.¹Cumulative percentage change in terms of trade (of goods and services) from start to peak of each identified episode (that meet the criteria of at least 15 percent cumulative and 3 percent average increase). Episodes are grouped in 5-year windows according to the date of their first year. Dotted lines indicate group averages. Country groups as defined in Annex Table A2.

• Download Figure
• Download figure as PowerPoint slide

Emerging Latin America and Selected Regions: Terms-of-Trade Booms, 1970-2012¹

(Cumulative percentage change, during upswing)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF, International Financial Statistics; and authors’ estimates.¹Cumulative percentage change in terms of trade (of goods and services) from start to peak of each identified episode (that meet the criteria of at least 15 percent cumulative and 3 percent average increase). Episodes are grouped in 5-year windows according to the date of their first year. Dotted lines indicate group averages. Country groups as defined in Annex Table A2.

• Download Figure
• Download figure as PowerPoint slide

This traditional measure of terms-of-trade shock, however, does not provide information on the magnitude of the associated income shock, as it does not take into account the degree of trade openness of these economies.² Next we discuss a metric that takes this dimension into account.

A. A Simple Metric of Terms-of-Trade Income Windfall

Methodology

Our interest lies in quantifying the extent to which real (gross domestic) income in these economies was higher than what it would have been had no terms-of-trade shock occurred (i.e., in the counterfactual).³ To this end, we construct a simple metric that focuses on the difference between actual real (gross domestic income and a similar variable but measured at pre-boom terms of trade. To construct the latter, we exploit the different paths of the GDP deflator and the CPI during terms-of-trade booms.

As in Kohli (2004), define country i’s real (gross domestic) income, at time t, as follows^:4

$\begin{array}{cc}{\mathit{RI}}_{i,t}\equiv \frac{{\mathit{GDI}}_{i,t}}{P_{i,t}^C}=\frac{{\mathit{GDP}}_{i,t}}{P_{i,t}^C}& \left(1\right)\end{array}$

where GDI (GDP) denotes gross domestic income (product) and $P_t^C$ denotes the Consumer Price Index (CPI). Key in this definition is the use of the CPI to deflate nominal income, so as to capture its purchasing power in terms of the average consumption basket.⁵ As the consumer price index differs from the GDP deflator, equation (1) can also be written as ${\mathit{RI}}_{i,t}={\mathit{RGDP}}_{i,t}*\left(\frac{P_{i,t}^Y}{P_{i,t}^C}\right)$, where RGDP_t denotes real GDP (i.e., ${\mathit{GDP}}_{i,t}/P_{i,t}^Y$), and $P_{i,t}^Y$ is the GDP deflator. This simple equation makes it clear that real income tends to diverge from real GDP when the GDP deflator diverges from the consumer price index, which tends to occur when the economy faces large terms-of-trade shocks, as shown next.

Express the GDP deflator in terms of its demand-side components as:

$\begin{array}{cc}{P}_{i,t}^Y=P_{i,t}^C{w}_{i,t-1}^C+P_{i,t}^I{w}_{i,t-1}^I+P_{i,t}^G{w}_{i,t-1}^G+P_{i,t}^X{w}_{i,t-1}^X-P_{i,t}^M{w}_{i,t-1}^M& \left(2\right)\end{array}$

where $P_{i,t}^C$ denotes the consumption deflator and $w_{i,t-1}^C$ its share in GDP in the previous year. Similar terminology applies to investment (I), government consumption (G), exports of goods and services (X) and imports of goods and services (M). Differentiating equations (1) and (2), and combining them, yields:⁶

$\begin{array}{cc}\widehat{{\mathit{RI}}_{\mathit{ı},t}}\approx {R\widehat{\mathit{GDP}}}_{\mathit{ı},t}+\left[P_{\mathit{ı},t}^{\widehat{X,}r}*w_{i,t-1}^X-\widehat{P_{ı,t}^{M,r}}*w_{i,t-1}^M\right]+\left[\widehat{E_{ı,t}}+\widehat{P_t^{*}}-\widehat{P_{ı,t}^C}\right]\left[w_{i,t-1}^X-w_{i,t-1}^M\right]& \left(3\right)\end{array}$

where $\hat{X}$ denotes the annual percentage change of any variable X; $P_{i,t}^{X,r}=P_{i,t}^X/P_t^{*}$ and $P_{i,t}^{M,r}=P_{i,t}^M/P_t^{*}$ are country i’s export and import prices (expressed relative to the U.S. CPI);⁷ $w_{i,t-1}^X$ and $w_{i,t-1}^M$ denote the ratios of exports and imports (of goods and services) to GDP, and E_i,t is country i’s exchange rate vis-à-vis the U.S. dollar (local currency units per U.S. dollar). This equation shows that differences between real GDI and real GDP arise primarily from the income impact of terms-of-trade shocks (as captured by the second term on the right-hand-side) and from the effect of real exchange rate movements (last term).

Finally, the first term on the right-hand-side can be decomposed into trend growth and deviations from trend $\left({\widehat{\mathit{RGDP}}}_{i,t}={\widehat{\mathit{RGDP}}}_{i,t}^{\mathit{trend}}+\mathrm{\Delta }{\widehat{\mathit{RGDP}}}_{\mathit{ı},t}\right)$ to yield:

$\begin{array}{cc}{\widehat{\mathit{RI}}}_{i,t}\approx {\widehat{\mathit{RGDP}}}_{i,t}^{\mathit{trend}}+\mathrm{\Delta }{\widehat{\mathit{RGDP}}}_{\mathit{ı},t}+\left[\widehat{P_{\mathit{ı},t}^{X,r}}*w_{i,t-1}^X-\widehat{P_{\mathit{ı},t}^{M,r}}*w_{i,t-1}^M\right]& \\ +\left[\widehat{E_{\mathit{ı},t}}+\widehat{P_t^{*}}-\widehat{P_{\mathit{ı},t}^C}\right]\left[w_{i,t-1}^X-w_{i,t-1}^M\right]& \left(4\right)\end{array}$

Stripping out the effect of changes in the real exchange rate (last term), equation (4) forms the basis for our estimates of the income windfall.⁸ It provides us with the key components to build a counterfactual level of real income. Specifically, one can construct two variants of such measure, depending on the assumed counterfactual:

• (i) A purely exogenous metric based on the difference between actual income (stripped out from exchange rate effects) and income at pre-boom trade prices. This measure captured by the third term on the right-hand-side of equation (4) assumes that, in absence of the terms-of-trade shock, the same level of output would have been achieved but with relative prices as those prevailing before the boom.

• (ii) A broader measure that tries to capture also the increase in real output arising from the shock, by imputing any deviation of output from its trend also to the terms-of-trade income windfall. This means computing the difference between actual income (stripped out from exchange rate effects) and projected income at trend growth, where the latter would grow at the rate of trend output. This measure encompasses the second and third terms on the right-hand-side of equation (4).

Real Gross Domestic Income

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Real Gross Domestic Income

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

• Download Figure
• Download figure as PowerPoint slide

Real Gross Domestic Income

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

• Download Figure
• Download figure as PowerPoint slide

Our interest lies primarily on the first measure, even though it is likely to provide a lower bound estimate of the income windfall (as output also tends to respond positively to favorable terms-of-trade shocks). The focus on this lower bound estimate poses stringent requirements on the data, strengthening the results. Furthermore, the second measure might be contaminated by the effect of domestic factors, which could raise endogeneity issues, while the first is an exogenous measure of the terms-of-trade income shock.⁹

Finally, using equation (4), we compute the annual income windfall as:

$\begin{array}{cc}{\mathit{WI}}_{i,t}\equiv \frac{\lceil {\mathit{RI}}_{i,t}-{\mathit{RI}}_{i,t}^{*}\rceil }{{\mathit{RI}}_{i,t}^{*}}& \left(5\right)\end{array}$

where RI_i,t is an index of actual real income, and ${\mathit{RI}}_{i,t}^{*}$ is the corresponding counterfactual, constructed as:¹⁰

$\begin{array}{cc}{\mathit{RI}}_{i,t}=\{\begin{array}{cc}1\hfill & \mathit{if}t=T-1\\ {\mathit{RI}}_{i,t-1}*[1+\mathrm{\Delta }{R\widehat{\mathit{GDP}}}_{\mathit{ı},t}+[\widehat{P_{ı,t}^{X,r}}*w_{i,t-1}^X-\widehat{P_{ı,t}^{M,r}}*w_{i,t-1}^M]]\hfill & \mathit{if}t>T-1\end{array}& \left(6\right)\end{array}$

and

$\begin{array}{cc}{\mathit{RI}}_{i,t}^{*}=\{\begin{array}{cc}1\hfill & \mathit{if}t=T-1\\ {\mathit{RI}}_{i,t-1}^{*}*[1+\mathrm{\Delta }{R\widehat{\mathit{GDP}}}_{\mathit{ı},t}]\hfill & \mathit{if}t>T-1\end{array}& \left(7\right)\end{array}$

Thus, the annual income windfall measures the vertical distance between real income and real income at pre-boom terms of trade; and the cumulative windfall the area between the two. Both are expressed as shares of real income at pre-boom terms of trade.

Finally, in order to construct cross-section statistics, and given that the duration of each event varies over time and across countries, we normalize the length of each episode and report statistics that aggregate information at different fractions of the lifetime of each episode. In particular, each episode’s (start to peak) length is discretized into ¼, ½, and ¾ fractions.

Results

The magnitude of the positive income shocks associated with recent terms-of-trade boom is evident from Figure 3, which presents the key measures of real income for subsamples of the identified booms. As shown, such effect has been especially pronounced in the recent episode and even more so for Emerging Latin American and the MENA oil exporting countries.¹¹ Furthermore, this income shock is particularly relevant for countries whose output did not grow faster than previously anticipated during the recent terms-of-trade boom, but whose income levels increased markedly on account of the improved trade prices. Within Latin America, Bolivia and Chile are striking cases of this (see Annex Figure A1).

Real Gross Domestic Income during Terms-of-trade Booms, 1970-2012 ^1/

(medians, normalized lengths of episodes, Index T-1=100)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF, International Financial Statistics; and authors’ estimations.^1/ Episode lengths are normalized in order to allow aggregation, and series are reported at fractions of the lifetime of each event. Real Domestic Income as defined in Section II.A (i.e., excluding real exchange rate effect). Country groupings as defined in Annex Table A2.2/ Episodes taken place since 2000.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Real Gross Domestic Income during Terms-of-trade Booms, 1970-2012 ^1/

(medians, normalized lengths of episodes, Index T-1=100)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF, International Financial Statistics; and authors’ estimations.^1/ Episode lengths are normalized in order to allow aggregation, and series are reported at fractions of the lifetime of each event. Real Domestic Income as defined in Section II.A (i.e., excluding real exchange rate effect). Country groupings as defined in Annex Table A2.2/ Episodes taken place since 2000.

• Download Figure
• Download figure as PowerPoint slide

Real Gross Domestic Income during Terms-of-trade Booms, 1970-2012 ^1/

(medians, normalized lengths of episodes, Index T-1=100)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF, International Financial Statistics; and authors’ estimations.^1/ Episode lengths are normalized in order to allow aggregation, and series are reported at fractions of the lifetime of each event. Real Domestic Income as defined in Section II.A (i.e., excluding real exchange rate effect). Country groupings as defined in Annex Table A2.2/ Episodes taken place since 2000.

• Download Figure
• Download figure as PowerPoint slide

A historical comparison of our estimates of the cumulative income windfall points to a much larger effect during the recent wave than in past ones, including those of the 1970s (Figure 4 and Tables 1–2). This reflects a higher degree of trade openness in most countries in the sample, and the longer duration of the recent event.

Income Windfall of Terms-of-trade Booms, 1970-2012 ^1/

(share of GDP)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: Authors’ calculations.^1/ Statistics across all ongoing episodes (upswing) at each point in time.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Income Windfall of Terms-of-trade Booms, 1970-2012 ^1/

(share of GDP)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: Authors’ calculations.^1/ Statistics across all ongoing episodes (upswing) at each point in time.

• Download Figure
• Download figure as PowerPoint slide

Income Windfall of Terms-of-trade Booms, 1970-2012 ^1/

(share of GDP)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: Authors’ calculations.^1/ Statistics across all ongoing episodes (upswing) at each point in time.

• Download Figure
• Download figure as PowerPoint slide

Table 1.

Key Statistics of Income Windfall (medians) ^1/

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Of goods and services, for upswing period (start to peak).

^5/In percent of income under the counterfactual (i.e., real domestic income at pre-boom terms-of-trade).

Table 1.

Key Statistics of Income Windfall (medians) ^1/

			Episode		Terms-of-Trade Shock 4/		Income Windfall (price effect) 5/
			Length (years)
Country Group		Number of Episodes	Upswing 2/	Cycle 3/	Cum. (%)	Annual Avg. (%)	Cum.	Avg
Latin America 11		35	3	4	43.1	14.1	15.3	5.2
	Current	10	9	10	80.0	8.6	92.5	15.1
	Past	25	2	3	33.2	15.7	11.8	4.5
EM Asia		16	2	4	27.7	9.8	18.1	6.8
	Current	3	3	3	32.4	7.9	51.2	10.3
	Past	13	2	4	26.8	11.1	14.9	5.6
EM Europe		15	4	5	30.4	8.5	16.7	7.4
	Current	6	5	6	29.9	5.5	32.4	4.2
	Past	9	3	5	30.4	14.7	16.7	8.0
MENA Oil Exporters		39	3	5	83.4	24.8	80.8	23.7
	Current	7	8	8	87.3	10.5	178.3	35.2
	Past	32	2	5	81.2	36.0	51.7	20.3
Advanced Ec. (excl. MENA)		21	4	5	28.8	7.2	19.0	5.0
	Current	3	9	9	62.7	7.0	64.1	7.1
	Past	18	4	5	28.5	7.4	16.2	4.9

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Of goods and services, for upswing period (start to peak).

^5/In percent of income under the counterfactual (i.e., real domestic income at pre-boom terms-of-trade).

View Table

Table 1.

Key Statistics of Income Windfall (medians) ^1/

			Episode		Terms-of-Trade Shock 4/		Income Windfall (price effect) 5/
			Length (years)
Country Group		Number of Episodes	Upswing 2/	Cycle 3/	Cum. (%)	Annual Avg. (%)	Cum.	Avg
Latin America 11		35	3	4	43.1	14.1	15.3	5.2
	Current	10	9	10	80.0	8.6	92.5	15.1
	Past	25	2	3	33.2	15.7	11.8	4.5
EM Asia		16	2	4	27.7	9.8	18.1	6.8
	Current	3	3	3	32.4	7.9	51.2	10.3
	Past	13	2	4	26.8	11.1	14.9	5.6
EM Europe		15	4	5	30.4	8.5	16.7	7.4
	Current	6	5	6	29.9	5.5	32.4	4.2
	Past	9	3	5	30.4	14.7	16.7	8.0
MENA Oil Exporters		39	3	5	83.4	24.8	80.8	23.7
	Current	7	8	8	87.3	10.5	178.3	35.2
	Past	32	2	5	81.2	36.0	51.7	20.3
Advanced Ec. (excl. MENA)		21	4	5	28.8	7.2	19.0	5.0
	Current	3	9	9	62.7	7.0	64.1	7.1
	Past	18	4	5	28.5	7.4	16.2	4.9

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Of goods and services, for upswing period (start to peak).

^5/In percent of income under the counterfactual (i.e., real domestic income at pre-boom terms-of-trade).

Table 2.

Emerging Latin America: Key Statistics of Income Windfall ^1/

Source: authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak).

^2/From start to peak.

^3/Start to end, with end identified as year when at least ⅓ of the shock is reverted (or 2012 if the latter is not met).

^4/Of goods and services, for upswing period (start to peak).

^5/In percent of income under the counterfactual (i.e., real income at pre-boom terms-of-trade).

^6/Results for Argentina are based officially reported data. The IMF has, however, issued a declaration of censure and called on Argentina to adopt remedial measures to address the quality of the official GDP and CPI-GBA data.

Table 2.

Emerging Latin America: Key Statistics of Income Windfall ^1/

	Episode			Terms-of-Trade Shock 4/		Income Windfall (price effect) 5/
		Length (years)
Countries	First Year	Upswing 2/	Cycle 3/	Cum. (%)	Annual Avg. (%)	Cum.	Avg
Argentina 6/	2003	9	10	85.1	9.5	89.7	10.0
Bolivia	2003	10	10	131.1	13.1	190.1	19.0
Brazil	2006	6	7	37.2	6.2	17.1	2.9
Chile	2003	9	10	90.2	10.0	184.7	20.5
Colombia	2004	9	9	68.7	7.6	47.7	5.3
Ecuador	2002	10	11	74.9	7.5	116.8	11.7
Paraguay	2002	3	3	89.2	29.7	55.6	18.5
Paraguay	2008	5	5	38.8	7.8	95.3	19.1
Peru	2003	10	10	61.9	6.2	84.6	8.5
Venezuela, Republica Bolivariana de	2003	10	10	233.3	23.3	303.9	30.4

Source: authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak).

^2/From start to peak.

^3/Start to end, with end identified as year when at least ⅓ of the shock is reverted (or 2012 if the latter is not met).

^4/Of goods and services, for upswing period (start to peak).

^5/In percent of income under the counterfactual (i.e., real income at pre-boom terms-of-trade).

^6/Results for Argentina are based officially reported data. The IMF has, however, issued a declaration of censure and called on Argentina to adopt remedial measures to address the quality of the official GDP and CPI-GBA data.

View Table

Table 2.

Emerging Latin America: Key Statistics of Income Windfall ^1/

	Episode			Terms-of-Trade Shock 4/		Income Windfall (price effect) 5/
		Length (years)
Countries	First Year	Upswing 2/	Cycle 3/	Cum. (%)	Annual Avg. (%)	Cum.	Avg
Argentina 6/	2003	9	10	85.1	9.5	89.7	10.0
Bolivia	2003	10	10	131.1	13.1	190.1	19.0
Brazil	2006	6	7	37.2	6.2	17.1	2.9
Chile	2003	9	10	90.2	10.0	184.7	20.5
Colombia	2004	9	9	68.7	7.6	47.7	5.3
Ecuador	2002	10	11	74.9	7.5	116.8	11.7
Paraguay	2002	3	3	89.2	29.7	55.6	18.5
Paraguay	2008	5	5	38.8	7.8	95.3	19.1
Peru	2003	10	10	61.9	6.2	84.6	8.5
Venezuela, Republica Bolivariana de	2003	10	10	233.3	23.3	303.9	30.4

Source: authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak).

^2/From start to peak.

^3/Start to end, with end identified as year when at least ⅓ of the shock is reverted (or 2012 if the latter is not met).

^4/Of goods and services, for upswing period (start to peak).

^5/In percent of income under the counterfactual (i.e., real income at pre-boom terms-of-trade).

^6/Results for Argentina are based officially reported data. The IMF has, however, issued a declaration of censure and called on Argentina to adopt remedial measures to address the quality of the official GDP and CPI-GBA data.

This pattern is also visible in the case of Emerging Latin America (Figure 5). Estimates of the income effect of the recent boom for the latter group are quite large in absolute terms, with a median of about 87 percent of a year’s GDP cumulative or about 15 percent per year. This means, that only taking into account the price effect of the terms-of-trade shock, real income has been, on average, at least 15 percent per year higher than what it would have been otherwise. Within Latin America, Venezuela, Bolivia, and Chile stand out as having benefited the most, with cumulative (annual average) windfalls of as much as 300 (30) percent of income in the first case, and close to 200 (20) percent in the case of both Bolivia and Chile. Not surprisingly, Brazil is at the other end of the distribution, with significantly lower windfall estimates, reflecting a lower reliance on commodities and its lower degree of trade openness.

Emerging Latin America and Selected Regions: Income Windfall, 1970-2012 ^1/

(share of annual GDP)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF International Financial Statistics, and authors’ estimates.^1/ Cumulative and annual average income windfall, as share of GDP, from start to peak of each identified episode (that meet the criteria of at least 15 percent cumulative and 3 percent average increase). Episodes are grouped in 5-year windows according to the date of their first year. Bars indicate cumulative values, dots indicate annual averages, and dotted lines indicate group averages for cumulative values. Country groups as defined in Annex Table A2.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Emerging Latin America and Selected Regions: Income Windfall, 1970-2012 ^1/

(share of annual GDP)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF International Financial Statistics, and authors’ estimates.^1/ Cumulative and annual average income windfall, as share of GDP, from start to peak of each identified episode (that meet the criteria of at least 15 percent cumulative and 3 percent average increase). Episodes are grouped in 5-year windows according to the date of their first year. Bars indicate cumulative values, dots indicate annual averages, and dotted lines indicate group averages for cumulative values. Country groups as defined in Annex Table A2.

• Download Figure
• Download figure as PowerPoint slide

Emerging Latin America and Selected Regions: Income Windfall, 1970-2012 ^1/

(share of annual GDP)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF International Financial Statistics, and authors’ estimates.^1/ Cumulative and annual average income windfall, as share of GDP, from start to peak of each identified episode (that meet the criteria of at least 15 percent cumulative and 3 percent average increase). Episodes are grouped in 5-year windows according to the date of their first year. Bars indicate cumulative values, dots indicate annual averages, and dotted lines indicate group averages for cumulative values. Country groups as defined in Annex Table A2.

• Download Figure
• Download figure as PowerPoint slide

These effects are quite large also in relative terms to those of other regions. For example, the median cumulative windfall was just over 50 percent in Emerging Asia, 30 percent in Emerging Europe, and close to 65 percent in commodity-exporting advanced countries. Only oil-exporting countries in the MENA region show income windfalls of larger magnitude than those of Latin America, reaching a median of about 170 percent of GDP. This highlights the order of magnitude of the terms-of-trade windfall received by Latin America, which has been not only larger that its own past experiences, but greater than most other regions. The share of countries that benefited from terms-of-trade booms has also been much higher in Latin America than in other regions.

III. Saving Patterns During the Terms-of-Trade Booms

In this section we study aggregate saving patterns during these episodes of terms-of-trade booms by looking at both average as well as marginal rates. The goal is to document differences across time and regions in the macroeconomic responses to these shocks, in terms of the extent to which income windfalls were saved. Such responses have an important bearing on economic performance after the boom, as discussed in section IV, especially in cases where booms were followed by busts (i.e., sharp drops in terms of trade).

It is important to highlight however that, in this paper, we do not attempt to assess the optimality of the macroeconomic response to the terms-of-trade shock (i.e., the determinants of saving and investment decisions). The latter entails, among other things, linking such response to the perceived persistence of the shock as well as other contemporaneous external perturbations, which we address in a separate paper.

Marginal saving rates are of particular interest, as they give an indication of the country’s aggregate effort to save the additional (marginal) income arising from the terms-of-trade shock. Since the allocation of savings may also play a role, as shown later, we also decompose aggregate savings into domestic (i.e., investment) and foreign saving, relying on the current account identity (S = I + CA).

A. Average Saving Rates

A comparison of Latin America’s average saving rates across episodes points to a visible difference between the recent episode and the previous ones (Figure 5 and Tables 3–4). In particular, we observe that during the recent event, the median saving rate has increased by about 4 percentage points of GDP relative to pre-boom levels, as opposed to only 2 percentage points in past episodes. This has been accompanied by a remarkable increase in the investment rate (of about 5 percentage points of GDP), in clear contrast with the past. As a result, and despite higher saving rates, the recent episodes do not display higher current account balances than those of the past. In addition, a look into the dynamics within the episode shows that current account improved during the first stages of the current episode, but have deteriorated more recently, reflecting the pickup in investment along with the decline of saving rates.¹²

Table 3.

Average and Marginal Saving Rates during Terms-of-trade Booms (medians) ^1/

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Average of 3 years previous to the terms-of-trade boom (in percent of GDP)

^5/Aggregate average rates (percent of GDP).

^6/Aggregate marginal rates, in percent of income windfall (computed on the basis of average saving and investment rates of 3 years prior to the terms-of-trade boom).

Table 3.

Average and Marginal Saving Rates during Terms-of-trade Booms (medians) ^1/

					Saving rates
			Episode Length (years)		Pre-boom 4/			Boom (Upswing)
					Average 5/			Average 5/			Marginal (percent of IW) 6/
Country Group		Number of Episodes	Upswing 2/	Cycle 3/	Total	Domestic	Foreign	Total	Domestic	Foreign	Total	Domestic	Foreign
Latin America 11		35	3	4	17.2	18.9	−2.5	20.7	21.8	−0.9	66.7	46.6	20.1
	Current	10	9	10	17.2	17.0	−0.1	21.4	22.0	0.7	45.2	41.2	4.0
	Past	25	2	3	17.2	20.2	−3.4	19.1	21.3	−2.2	80.6	46.6	34.0
EM Asia		16	2	4	22.0	26.8	−4.4	23.1	25.9	−0.9	50.5	11.7	38.7
	Current	3	3	3	27.1	23.2	3.9	29.5	25.8	0.7	52.9	21.0	31.9
	Past	13	2	4	21.4	27.1	−5.2	21.2	26.0	−1.0	41.1	−8.3	49.4
EM Europe		15	4	5	19.4	20.9	−2.6	22.7	24.5	−0.9	30.5	41.1	−10.6
	Current	6	5	6	17.6	20.6	−3.6	19.3	25.2	−7.5	28.6	51.3	−22.7
	Past	9	3	5	20.5	25.3	−2.3	23.0	23.1	−0.5	32.5	34.9	−2.4
MENA Oil Exporters		39	3	5	24.4	23.0	−1.3	33.2	21.3	6.8	79.0	7.3	71.7
	Current	7	8	8	34.3	19.6	8.2	45.1	28.4	14.2	77.2	29.7	47.6
	Past	32	2	5	22.6	23.8	−2.4	31.2	21.1	4.2	80.7	1.7	79.1
Advanced Ec. (exc. MENA)		21	4	5	22.4	25.0	−3.2	24.8	24.0	−0.2	46.5	10.6	35.8
	Current	3	9	9	22.3	20.2	2.1	24.0	23.0	1.3	52.5	66.9	−14.5
	Past	18	4	5	23.0	26.4	−3.9	26.0	24.1	−0.6	43.6	3.1	40.5

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Average of 3 years previous to the terms-of-trade boom (in percent of GDP)

^5/Aggregate average rates (percent of GDP).

^6/Aggregate marginal rates, in percent of income windfall (computed on the basis of average saving and investment rates of 3 years prior to the terms-of-trade boom).

View Table

Table 3.

Average and Marginal Saving Rates during Terms-of-trade Booms (medians) ^1/

					Saving rates
			Episode Length (years)		Pre-boom 4/			Boom (Upswing)
					Average 5/			Average 5/			Marginal (percent of IW) 6/
Country Group		Number of Episodes	Upswing 2/	Cycle 3/	Total	Domestic	Foreign	Total	Domestic	Foreign	Total	Domestic	Foreign
Latin America 11		35	3	4	17.2	18.9	−2.5	20.7	21.8	−0.9	66.7	46.6	20.1
	Current	10	9	10	17.2	17.0	−0.1	21.4	22.0	0.7	45.2	41.2	4.0
	Past	25	2	3	17.2	20.2	−3.4	19.1	21.3	−2.2	80.6	46.6	34.0
EM Asia		16	2	4	22.0	26.8	−4.4	23.1	25.9	−0.9	50.5	11.7	38.7
	Current	3	3	3	27.1	23.2	3.9	29.5	25.8	0.7	52.9	21.0	31.9
	Past	13	2	4	21.4	27.1	−5.2	21.2	26.0	−1.0	41.1	−8.3	49.4
EM Europe		15	4	5	19.4	20.9	−2.6	22.7	24.5	−0.9	30.5	41.1	−10.6
	Current	6	5	6	17.6	20.6	−3.6	19.3	25.2	−7.5	28.6	51.3	−22.7
	Past	9	3	5	20.5	25.3	−2.3	23.0	23.1	−0.5	32.5	34.9	−2.4
MENA Oil Exporters		39	3	5	24.4	23.0	−1.3	33.2	21.3	6.8	79.0	7.3	71.7
	Current	7	8	8	34.3	19.6	8.2	45.1	28.4	14.2	77.2	29.7	47.6
	Past	32	2	5	22.6	23.8	−2.4	31.2	21.1	4.2	80.7	1.7	79.1
Advanced Ec. (exc. MENA)		21	4	5	22.4	25.0	−3.2	24.8	24.0	−0.2	46.5	10.6	35.8
	Current	3	9	9	22.3	20.2	2.1	24.0	23.0	1.3	52.5	66.9	−14.5
	Past	18	4	5	23.0	26.4	−3.9	26.0	24.1	−0.6	43.6	3.1	40.5

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Average of 3 years previous to the terms-of-trade boom (in percent of GDP)

^5/Aggregate average rates (percent of GDP).

^6/Aggregate marginal rates, in percent of income windfall (computed on the basis of average saving and investment rates of 3 years prior to the terms-of-trade boom).

Table 4.

Emerging Latin America: Average and Marginal Saving Rates during Terms-of-trade Booms ^1/

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Average of 3 years previous to the terms-of-trade boom (in percent of GDP)

^5/Aggregate average rates (percent of GDP).

^6/Aggregate marginal rates, in percent of income windfall (computed on the basis of average saving and investment rates of 3 years prior to the terms-of-trade boom).

^7/Results for Argentina are based officially reported data. The IMF has, however, issued a declaration of censure and called on Argentina to adopt remedial measures to address the quality of the official GDP and CPI-GBA data.

Table 4.

Emerging Latin America: Average and Marginal Saving Rates during Terms-of-trade Booms ^1/

				Saving rates
		Episode Length (years)		Pre-boom 4/			Boom (Upswing)
				Average 5/			Average 5/			Marginal (percent of WI) 6/
Country Group	First Year	Upswing 2/	Cycle 3/	Total	Domestic	Foreign	Total	Domestic	Foreign	Total	Domestic	Foreign
Argentina 7/	2003	9	10	16.1	14.6	1.5	24.3	21.9	2.4	105.9	96.4	9.5
Bolivia	2003	10	10	12.0	16.2	−4.3	21.7	15.9	5.8	74.6	15.1	59.5
Brazil	2006	6	7	17.7	16.4	1.4	18.0	19.1	−1.0	29.8	114.3	−84.5
Chile	2003	9	10	22.7	23.9	−1.2	23.4	22.2	1.2	27.1	14.0	13.2
Colombia	2004	9	9	16.2	17.3	−1.2	19.9	22.3	−2.3	91.5	116.3	−24.8
Ecuador	2002	10	11	20.1	18.0	2.1	22.4	22.2	0.2	42.8	58.6	−15.8
Paraguay	2002	3	3	14.2	16.7	−2.5	17.9	16.2	1.7	36.9	13.8	23.1
Paraguay	2008	5	5	17.6	16.7	0.9	13.6	14.9	−1.3	−10.2	5.8	−16.0
Peru	2003	10	10	16.8	19.1	−2.4	21.2	22.0	−0.8	74.6	57.8	16.8
Venezuela, Republica Bolivariana de	2003	10	10	29.2	24.3	4.9	33.6	24.1	9.4	47.5	24.6	23.0

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Average of 3 years previous to the terms-of-trade boom (in percent of GDP)

^5/Aggregate average rates (percent of GDP).

^6/Aggregate marginal rates, in percent of income windfall (computed on the basis of average saving and investment rates of 3 years prior to the terms-of-trade boom).

^7/Results for Argentina are based officially reported data. The IMF has, however, issued a declaration of censure and called on Argentina to adopt remedial measures to address the quality of the official GDP and CPI-GBA data.

View Table

Table 4.

Emerging Latin America: Average and Marginal Saving Rates during Terms-of-trade Booms ^1/

				Saving rates
		Episode Length (years)		Pre-boom 4/			Boom (Upswing)
				Average 5/			Average 5/			Marginal (percent of WI) 6/
Country Group	First Year	Upswing 2/	Cycle 3/	Total	Domestic	Foreign	Total	Domestic	Foreign	Total	Domestic	Foreign
Argentina 7/	2003	9	10	16.1	14.6	1.5	24.3	21.9	2.4	105.9	96.4	9.5
Bolivia	2003	10	10	12.0	16.2	−4.3	21.7	15.9	5.8	74.6	15.1	59.5
Brazil	2006	6	7	17.7	16.4	1.4	18.0	19.1	−1.0	29.8	114.3	−84.5
Chile	2003	9	10	22.7	23.9	−1.2	23.4	22.2	1.2	27.1	14.0	13.2
Colombia	2004	9	9	16.2	17.3	−1.2	19.9	22.3	−2.3	91.5	116.3	−24.8
Ecuador	2002	10	11	20.1	18.0	2.1	22.4	22.2	0.2	42.8	58.6	−15.8
Paraguay	2002	3	3	14.2	16.7	−2.5	17.9	16.2	1.7	36.9	13.8	23.1
Paraguay	2008	5	5	17.6	16.7	0.9	13.6	14.9	−1.3	−10.2	5.8	−16.0
Peru	2003	10	10	16.8	19.1	−2.4	21.2	22.0	−0.8	74.6	57.8	16.8
Venezuela, Republica Bolivariana de	2003	10	10	29.2	24.3	4.9	33.6	24.1	9.4	47.5	24.6	23.0

Source: Authors’ estimations.

^1/For episodes with at least 15 percent cumulative and 3 percent annual average terms-of-trade shock (from start to peak). Subgroups as defined in Annex Table A2.

^2/From start to peak.

^3/Cycle is defined as start to end. End is identified when at least ⅓ of the shock is reverted.

^4/Average of 3 years previous to the terms-of-trade boom (in percent of GDP)

^5/Aggregate average rates (percent of GDP).

^6/Aggregate marginal rates, in percent of income windfall (computed on the basis of average saving and investment rates of 3 years prior to the terms-of-trade boom).

^7/Results for Argentina are based officially reported data. The IMF has, however, issued a declaration of censure and called on Argentina to adopt remedial measures to address the quality of the official GDP and CPI-GBA data.

Developments in Latin America, however, are dwarfed by those seen in the MENA oil exporting countries. The latter experienced, during the recent episode, an increase in saving rates of about 11 percentage points of GDP, and starting from levels twice as high as those seen in Latin America. Investment also picked up markedly in these oil exporters, leading to a sharp weakening of their current account balances towards the ½ life of the episode, although external balances still hovered around 10 percentage points of GDP (as opposed to around zero in Latin America).

B. Marginal Saving Rates

Average saving and investment rates, however, do not inform us about the extent to which the income windfall is saved, as they are computed over the overall level of income, as opposed to just the extraordinary component. Marginal rates, on the other hand, give a metric of the latter, and thus an indication of the ‘effort’ made to save the additional income. It is, therefore, likely to be more relevant for explaining the dynamics of the economy during the boom as well as their post-boom impact.

We compute marginal saving rates in a way that captures the increase in savings as a proportion of the estimated income windfall.¹³ To do this, the economy’s average saving rate during the boom (s) is decomposed into a ‘norm’ saving rate ($\overline{s}$)—that we proxy by the average saving rate in the three years preceding the boom—and the marginal saving rate (s^W) or share of the income windfall saved, as follows:

$\begin{array}{cc}s=\overline{s}*\frac{{\mathit{RI}}^{*}}{\mathit{RI}}+s^W*\frac{\left(\mathit{RI}-{\mathit{RI}}^{*}\right)}{\mathit{RI}}& \left(8\right)\end{array}$

The average saving rate can be seen as a weighted average of the norm saving rate and the saving rate on the extra income (windfall). Equation (8) can be re-arranged to derive the marginal saving rate as:

$\begin{array}{cc}{s}^W\equiv \frac{\left(s*\mathit{RI}-\overline{s}*{\mathit{RI}}^{*}\right)}{\mathit{RI}-{\mathit{RI}}^{*}}& \left(9\right)\end{array}$

And this expression shows that the marginal rate can be computed as the additional saving during the boom with respect to what saving would have been had no terms-of-trade boom occurred, measured as share of the income windfall.

We also compute the marginal rates for domestic and foreign savings in the same way that the average saving rates (S) can be decomposed into domestic investment (I) and foreign asset accumulation (CA) using the current account identity (CA=S-I). In this case, the measures are interpreted as the marginal saving either channeled to increase the stock of domestic capital (investment) or to improve the country’s net foreign asset position (through a strengthening of the external current account). Specifically, we compute the marginal investment rate (s^WD) as follows:

$\begin{array}{cc}{s}^{\mathrm{WD}}\equiv \frac{\left(\mathit{ir}*\mathit{RI}-\overline{\mathit{ır}}*{\mathit{RI}}^{*}\right)}{\mathit{RI}-{\mathit{RI}}^{*}}& \left(10\right)\end{array}$

where ir stands for the investment rate and $\overline{\mathit{ır}}$ is the corresponding norm (again, measured as the pre-boom 3-year average). As before, this estimate has an intuitive interpretation: how much of the extra income (relative to the counterfactual) is allocated to gross capital formation, relative to the corresponding allocation in the counterfactual. Finally, using the current account identity we derive the foreign saving component, as

$\begin{array}{cc}{s}^{\mathit{WF}}=s^W-s^{\mathit{WD}}& \left(11\right)\end{array}$

For the latter, the same interpretation applies, although referring to the share of income windfall allocated to the accumulation of net foreign assets.

These metrics reveal that Latin America’s marginal saving rates in the recent episode were actually lower than those from past episodes (Table 3). Indeed, the last episode has a median saving rate of 45 percent of the income windfall, compared a median of 80 percent during past events, although marginal rates for previous episodes depict a relatively high dispersion. These figures suggest, therefore, that Latin America’s macroeconomic response to the recent terms-of-trade boom, in terms of its effort to save the windfall has not been greater than during past episodes. There is, however, significant variance across countries within Latin America (Table 4). Bolivia, Colombia, and Peru appear to have saved a larger proportion of their income shocks than the rest, while Paraguay, Chile, and Brazil stand on the other extreme. In the case of Chile, however, a relatively low marginal saving rate reflects a relatively high starting average rate.¹⁴

Interestingly, MENA oil-exporting countries—the other group with similar or larger income windfall—appear to have saved a much larger share of the windfall than Latin America, with a median marginal saving rate of about 80 percent. Furthermore, this is about the same as during this region’s past episodes, suggesting that common global shocks (other than terms of trade) cannot fully explain the drop in marginal savings rates in Latin America. Another important difference between these two regions is in the split between domestic and foreign saving, with MENA countries displaying a balanced allocation in the recent boom, as opposed to Latin America, whose allocation has been heavily biased towards domestic investment. The breakdown between domestic and foreign saving suggest that, other than Paraguay in 2002 and Bolivia in 2003, all other episodes show a larger share of their marginal savings being allocated to domestic investment, as opposed to improving their net foreign asset position (through the current account). In some cases, there is even a (marginal) deterioration in the country’s net international asset position.

A glance at the cross-time dynamics of Latin America’s marginal saving rates during the recent episode points to a gradual decline after a short-lived initial increase (Figure 6). These dynamics, as discussed above, could reflect changing perceptions of the persistence of the terms-of-trade shock, if being increasingly perceived as more protracted, as the boom evolved.¹⁵ At the same time, different dynamics in the MENA region do not lend support to the latter hypothesis. Furthermore, the breakdown of marginal savings rate into domestic and foreign savings shows, as average rates do, a growing share of the windfall being allocated to domestic capital formation rather than to improving countries’ net foreign asset position. This pattern is also consistent with the notion that the perceived persistence of the shock may have increased over time, leading to further domestic investment (including in commodity related sectors). While such behavior can be reconciled with a need to accumulate physical capital, especially because these are developing economies, it remains an empirical issue whether domestic investment or foreign asset accumulation is preferable, in terms of increasing post-boom income, during periods of terms-of-trade booms. This is studied in Section IV.

Emerging Latin America and MENA Oil Exporters: Aggregate Savings, Investment, and Current Account during Terms-of-Trade Booms¹

(Percent of GDP; medians, and 25th and 75 percentiles)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: IMF International Financial Statistics; and authors’ calculations.¹Episode length is normalized in order to allow aggregation, and series are reported at fractions of the lifetime of each event.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Emerging Latin America and MENA Oil Exporters: Aggregate Savings, Investment, and Current Account during Terms-of-Trade Booms¹

(Percent of GDP; medians, and 25th and 75 percentiles)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: IMF International Financial Statistics; and authors’ calculations.¹Episode length is normalized in order to allow aggregation, and series are reported at fractions of the lifetime of each event.

• Download Figure
• Download figure as PowerPoint slide

Emerging Latin America and MENA Oil Exporters: Aggregate Savings, Investment, and Current Account during Terms-of-Trade Booms¹

(Percent of GDP; medians, and 25th and 75 percentiles)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: IMF International Financial Statistics; and authors’ calculations.¹Episode length is normalized in order to allow aggregation, and series are reported at fractions of the lifetime of each event.

• Download Figure
• Download figure as PowerPoint slide

Public and Private Sectors’ Saving Patterns

So far, we have discussed patterns of savings and investment at the aggregate level (i.e., for the economy as a whole). An interesting and important issue, however, is whether there are differences in behavior between the public and the private sectors. To study this issue, we compute the corresponding average saving and investment rates for each of them.¹⁶

We find that the public sector appears to have responded more prudently than the private sector to the recent terms-of-trade shock (Figure 7). This is evidenced by the sharp increase in average savings rates of the former, while private sector saving rates have remained broadly stable. At the same time, the dynamics suggests a more recent ‘relaxation’ in the public sector efforts to save the windfall. The latter, together with a sharp increase in investment rates has largely contributed to weakening current account balances in the region. Interestingly, for past episodes we find broadly similar results regarding public and private sector savings. A main difference, however, is the sharp response of investment rates this time, as opposed to the past. It is important, however, to highlight that the comparison with past episodes is subject to significant data limitations, and thus results should be interpreted with caution. Furthermore, this analysis is based on averages rather than marginal rates, and therefore not the best indication of the effort made by these two sectors of the economy. Computing their marginal savings rate is left for future research as it entails estimating private disposable income, thus requiring further data gathering.

Emerging Latin America and MENA Oil Exporters: Windfall Saving Rates¹

(Cumulative, share of income windfall ; medians, and 25th and 75th percentiles)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: IMF, International Financial Statistics; and authors’ estimations.¹ Episode length is normalized in order to allow aggregation, and series are reported at fractions of the lifetime

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Emerging Latin America and MENA Oil Exporters: Windfall Saving Rates¹

(Cumulative, share of income windfall ; medians, and 25th and 75th percentiles)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: IMF, International Financial Statistics; and authors’ estimations.¹ Episode length is normalized in order to allow aggregation, and series are reported at fractions of the lifetime

• Download Figure
• Download figure as PowerPoint slide

Emerging Latin America and MENA Oil Exporters: Windfall Saving Rates¹

(Cumulative, share of income windfall ; medians, and 25th and 75th percentiles)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Source: IMF, International Financial Statistics; and authors’ estimations.¹ Episode length is normalized in order to allow aggregation, and series are reported at fractions of the lifetime

• Download Figure
• Download figure as PowerPoint slide

Emerging Latin America. Saving, Investment and Balance ^1/

(in percent of GDP, change from T-1, group simple averages)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF International Financial Statistics, and authors’ estimations.^1/ Breakdown public/private for past episodes is based on a subsample of episodes, due to limited data availiability.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Emerging Latin America. Saving, Investment and Balance ^1/

(in percent of GDP, change from T-1, group simple averages)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF International Financial Statistics, and authors’ estimations.^1/ Breakdown public/private for past episodes is based on a subsample of episodes, due to limited data availiability.

• Download Figure
• Download figure as PowerPoint slide

Emerging Latin America. Saving, Investment and Balance ^1/

(in percent of GDP, change from T-1, group simple averages)

Citation: IMF Working Papers 2013, 103; 10.5089/9781484356241.001.A001

Sources: IMF International Financial Statistics, and authors’ estimations.^1/ Breakdown public/private for past episodes is based on a subsample of episodes, due to limited data availiability.

• Download Figure
• Download figure as PowerPoint slide

IV. Saving during the Boom and Post-Boom Income

In this section we study how saving patterns during the boom (and their different allocations) affect post-boom income, by way of a simple cross-section econometric exercise. The goal is to gauge the extent to which saving during the boom delivers higher income afterwards, as one would expect, and more interestingly, whether it matters if the saving is channeled to domestic purposes (investment) or to improving the external position of the country (through the current account). Thus, the following specification is estimated using OLS:

where ${\mathit{RGNDY}}_i^{\mathit{Post}}$ denotes post-boom real gross national disposable income for episode i, which is measured as the level at 5 years after the boom; and ws_i stands for the corresponding windfall saving. We use gross national disposable income—rather than gross domestic income as in the first part of the paper—to take into account the net financial income from abroad (the income balance of the external current account), which incorporates the return on marginal changes in the country’s net foreign asset position. This let us study whether returns on domestic or foreign savings have been higher.