EFFECTS OF THE BUSINESS CYCLE ON SOCIAL INDICATORS IN LATIN AMERICA AND THE CARIBBEAN: WHEN DREAMS MEET REALITY

After mediocre growth in 2018 of 0.7 percent. Latin America and the Caribbean (LAC) is expected to perform only marginally better in 2019 (growth of 0.9 percent) followed by a much more solid growth of 2.1 percent in 2020. LAC will face both internal and external challenges during 2019. On the domestic front. the recession in Argentina; a slower than expected recovery in Brazil from the 2014-2015 recession, anemic growth in Mexico. and the continued deterioration of Venezuela. present the biggest challenges. On the external front. the sharp drop in net capital inflows to the region since early 2018 and the monetary policy normalization in the United States stand among the greatest perils. Furthermore, the recent increase in poverty in Brazil because of the recession points to the large effects that the business cycle may have on poverty. The core of this report argues that social indicators that are very sensitive to the business cycle may yield a highly misleading picture of permanent social gains in the region.


Introduction
After six years of growth deceleration (including a fall in GDP of almost 1 percent in 2016), the Latin America and the Caribbean (LAC) region had resumed in 2017 what seemed to be a path of modest but increasing growth, led by a rise in GDP of 1.3 percent (Table 1.1 and Figure 1.1). A pickup in oil and copper prices, large net capital inflows into the region, modest recoveries in Argentina and Brazil, and a very gradual pace of monetary policy normalization in the advanced economies, particularly the United States, all contributed to this turnaround in 2017 and, as of last April, growth in 2018 was expected to be 1.8 percent.

. Recent and Forecasted Real GDP Growth in LAC
Notes: Sub-regional values are weighted averages; "f" stands for forecast. For 2019 the World Bank estimates a range of -18 to -30 percent for Venezuela. The 2020 GDP growth forecast for Venezuela is omitted due to the great uncertainty about the country's future economic situation. South America comprises Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Guyana, Paraguay, Peru, Suriname, Uruguay, and Venezuela, RB. Central America comprises Belize, Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua, and Panama. The Caribbean comprises Antigua and Barbuda, The Bahamas, Barbados, Dominica, Dominican Republic, Grenada, Haiti, Jamaica, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, and Trinidad and Tobago. Source: World Bank staff estimates (March 2019).
Unfortunately, the much-anticipated path of increasing growth was not to be, as the region hit several bumps in the road, which reduced last year's growth from the 1.8 percent projection to an estimated 0.7 percent (1.4 percent excluding Venezuela). The sharp contraction in Argentina, the tepid recovery in Brazil after the major recession of 2015-2016, the anemic growth in Mexico in the midst of political uncertainty, and the implosion of Venezuela's economy resulted in growth all but stalling in 2018. 2 2 Note that Brazil, Mexico, and Argentina account for almost 70 percent of the region's GDP. Among the large economies of the region, Colombia was the silver lining in 2018 with a healthy growth of 2.7 percent. Regrettably, the growth prospects this year for LAC (0.9 percent) show no real improvement over 2018, as a result of weak or negative growth in the three largest economies in the region -Brazil, Mexico, and Argentina -and the tragic growth collapse in Venezuela. Argentina starts 2019 immersed in a severe recession, with GDP projected to fall a further 1.3 percent this year following a contraction of 2.5 percent in 2018 (Table 1.1 and Figure 1.2, Panel A). In 2018, the peso depreciated by 66 percent relative to the previous year, inflation is still close to 50 percent, and policy rates had to be raised above 70 percent last October to prevent further depreciation. Despite the unprecedented support from the International Monetary Fund (IMF), reflected in a revised package of 57.1 billion dollars in October 2018, and the central bank's success early this year in stabilizing the peso, the fiscal adjustment needed to comply with the IMF program is taking a heavy toll in terms of economic activity and the peso has come under renewed attack. The government, however, appears firmly committed to complying with the fiscal adjustment agreed with the IMF, but the October's presidential elections will undoubtedly test the government's resolve. economic front, the new administration that took office on January 1 st , 2019 is hitting all the right notes, but the jury is still out on its ability to carry out an ambitious reform program. A major pension reform has already been sent to Congress, which would entail a rise in the retirement age to 65 for men and 62 for women, from the current mid-50's (i.e., after only 30 years of contributions). This reform would cut spending on pensions by more than 1.0 percent of expected GDP over a 10-year period. Whether this reform will be approved by Congress in its current form, however, is far from clear, given that the president's party does not have a majority and would need to rely on coalitionbuilding. In the meantime, the central bank has kept unchanged the current Selic rate at 6.5 percent in a holding pattern waiting for new information on growth and inflation. In 2019, Mexico will show modest, but stable, growth (projected at 1.7 percent after growing 2.0 percent in 2018; see Table 1 however, remain among the highest in the region for large economies (at 8.25 percent), reflecting the central bank's need to defend the peso given the mixed signals from the current administration regarding the future course of economic policies. Even before taking office, the current president spooked markets by announcing the suspension of an already partially built 13 billion dollars new Mexico City airport. Major energy reforms by the previous administration have been put on hold as well, calling into question the future of Mexico's energy policy. On the other hand, the current administration submitted a relatively prudent fiscal budget for 2019, which was approved by Congress in late December 2018. To add to the positive signals, the current administration has also recently announced a slashing of the tax rate for equity IPOs and allowed pension funds to invest in a wider range of instruments. Signals from the new administration have thus been decidedly mixed and only time will tell which orientation will prevail. But, in the meantime, economic policy uncertainty is likely to force the central bank to maintain a tight monetary policy, which will hurt growth. In contrast, in Colombia ( Figure 1.2, Panel C), the policy rate of 4.25 percent should stimulate growth.
But nothing could prepare the region for the escalation of the economic, social, and humanitarian crisis in Venezuela, by far the worst in the region's modern history (Figure 1.3). Economic and social conditions continue to deteriorate rapidly. Declining oil prices -and hence, production and exports of oil -together with highly distortionary policies, from price controls to directed lending, a disorderly fiscal adjustment, monetization of the public sector deficit, and overall economic mis-management have led to hyperinflation, devaluation ( As always, the region's overall growth (0.7 percent in 2018 and a projection of 0.9 percent for 2019) masks a great deal of heterogeneity across different sub-regions (Table 1.1). GDP in South America (SA) remained essentially flat in 2018 (but grew 1.0 percent excluding Venezuela) and is expected to grow by 0.4 percent in 2019 (1.8 percent excluding Venezuela). Central America's (CA) growth was 2.7 percent in 2018 (down from 3.7 percent in 2017), partly due to the political and economic crisis in Nicaragua that led to a fall in GDP of 3.8 percent in 2018, compared to positive growth of 4.9 percent in 2017 (      Figure 1.5 the meltdown in Venezuela and recessions in Nicaragua and Argentina.
Given these differences in growth across countries, what factors may explain this phenomenon? The next section will differentiate between external and domestic factors affecting LAC.

The Role of External Factors
From the perspective of a small open economy, as those in LAC, external factors play a fundamental role in determining growth (Figure 1.6). Indeed, these have been decisive determinants of the slowdown that the region experienced in the aftermath of the Golden Decade.   The price of commodities, growth in the United States and China, and international liquidity -as captured by the real yield on the 10-year Treasury note -are, by and large, among the most important external factors for the region. Figure 1.6. illustrates their recent behavior. The increasing uncertainty regarding the future path of commodity prices and the slowdown in the Chinese growth rate pose difficult challenges for commodity exporters in the region. In particular, as of mid-March 2019, oil prices have dropped by 17 percent since their October 2018 high, while copper prices have fallen by 8 percent since their January 2018 high. Oil is the main export for Colombia, Ecuador, and Venezuela, and certainly important for Mexico, while copper is the main export for Chile and Peru.
Of course, behind the recent increase in world real interest rates captured by Figure 1 Index in appreciating the dollar (Figure 1.7, Panel B) and, more recently, contributed to a sharp fall in net capital inflows (measured as the 12-month cumulative figure), from a high of 50 billion dollars in January 2018 to virtually zero in January 2019 (Figure 1.7, Panel C). Not coincidentally, this dramatic fall in net capital inflows has been accompanied by a sharp appreciation of the dollar since January 2018 and a corresponding depreciation of emerging markets' currencies ( Figure 1.7, Panel D). The depreciation of domestic currencies in LAC has begun to confront central banks with the monetary policy dilemma analyzed in Végh et al. (2017). Should central banks increase policy rates to defend domestic currencies at the cost of aggravating a possible economic slowdown, or should they lower policy rates to stimulate the economy at the cost of further depreciation and inflation? Having said that, the latest announcement by the Federal Reserve of no more policy rate increases in 2019 and only one in 2020 should provide a breather to the region.
To convey the quantitative importance of external factors in the growth performance in SA, we use an econometric model that estimates the effects of four external variables on the growth rate of SA. 6 The explanatory variables are the growth rate of the G-7 and China, an index of commodity prices, and the real yield on the United States 10-year Treasury note as a proxy for the global cost of capital. Figure 1.8 illustrates the results of the model. The purple line shows the actual growth rate of SA while the orange line shows the growth rate predicted by the model. The resulting predicted series summarizes the average effect of the external factors on the growth rate. Therefore, the difference between both lines can be interpreted as the influence of domestic factors. When actual growth is higher (lower) than predicted, the influence of domestic factors is positive (negative).  The figure makes clear that the deceleration in the region's growth rate since the end of the Golden Decade of high commodity prices was driven by external factors. Additionally, it can be observed that SA's growth rate was notably affected by domestic factors, in particular the Brazilian recession of 2015-2016 (the largest in the country's recent history). Currently, actual and predicted growth coincide, which tells us that SA is generating little, if any, of its own growth and needs to urgently find its own sources of growth, as repeatedly emphasized in this series of reports.

Fiscal Adjustment in LAC: A Progress Report
Unfortunately, and as illustrated in Figure 1.9, the region's fiscal situation continues to be rather weak, despite some marginal improvements. In 2019, 27 out the 32 countries in the region will have an overall fiscal deficit, a slightly better performance relative to 2018 (when 29 out of 32 countries had an overall fiscal deficit). Further, the median fiscal deficit for the region in 2019 will be in fact a tad lower than in 2018 (2.1 percent of GDP in 2019 compared to 2.4 percent in 2018). In the same vein, the median fiscal deficit for South America in 2019 is expected to fall by 1.0 percentage points (i.e., from 3.8 percent of GDP in 2018 to 2.8 in 2019). Another welcome development is the large number of primary budget surpluses, especially in MCC (Mexico, Central America, and the Caribbean), which suggests that fiscal consolidation efforts are underway. Leaving aside the case of Venezuela, the most worrisome case is that of Brazil, which is expected to have an overall deficit of 6.9 percent of GDP in 2019 and primary deficit of 1.2 percent of GDP.
To assess the fiscal trends more accurately, Figure 1.10 shows the overall and primary deficits for 15 LAC countries. Panel A illustrates the case of South America and Mexico. Although, on average, the  overall deficit has improved by 2.3 percentage points and the primary deficit by 2.2 percentage points, the figure clearly shows that fiscal consolidation efforts vary considerably across countries. Specifically, we can see consistent fiscal improvements in Argentina, Ecuador, and Peru even if, except for Ecuador, overall deficits remain high. 7 Brazil, again, stands out for its enormous overall deficit.
The picture looks less encouraging in the case of Central America and Dominican Republic ( Figure  1.10, Panel B). In fact, during the four-year period 2016-2019, the average overall fiscal deficit and primary deficit have not changed much. Further, of the seven countries in this panel, there is none that shows consistent reductions in the overall deficit, although some countries, like El Salvador, show repeated improvements in the primary deficit.
Fiscal deficits, of course, continue to add to the region's public debt burden, which now averages 59.4 percent of GDP ( Figure 1.11), with seven countries (Jamaica, Barbados, Venezuela, Argentina, Belize, Antigua and Barbuda, and Dominica) having a debt ratio above 80 percent. 8 It is also worth noting that Brazil has a debt ratio of 77 percent of GDP, reflecting the country's precarious fiscal situation. Not surprisingly, the weak fiscal situation and correspondingly higher debt-to-GDP ratios have an impact on the countries' credit ratings, making access to international capital markets more difficult and costlier. To illustrate this, Figure 1.12 shows a scatter plot that links debt-to-GDP ratios (on the vertical axis) against Fitch long-term credit ratings (on the horizontal axis). As expected, the regression 7 For Ecuador, data correspond to the non-financial public sector (NFPS) as opposed to the central government. This is an important difference to keep in mind in the case of Ecuador because the fall in the central government's overall deficit has been much smaller (from 5. line (significant at the 1 percent level) indicates a negative relationship implying that the higher the debt levels, the lower the credit ratings. Further, while the average risk premium for investment grade countries is 186 basis points, the one for non-investment grade countries is more than twice as high (384 basis points).  In the case of Uruguay, Fitch specifically referred to persistent fiscal deficits and high and rising debt burden while, in the case of Mexico, the rating agency cited uncertainty over the overall economic policy course and the continuity of energy reforms from the previous administration, as well as growing risks of contingent liabilities materializing from stateowned oil company Pemex. Among the non-investment grade countries in the region, four saw their credit ratings downgraded since January 1, 2018. In particular, Brazil's credit rating was downgraded from BB to BB-in February 2018 and Nicaragua's was downgraded twice (in June and November 2018).
As a complement to Figure 1.12, Figure 1.13 shows the EMBI spreads for 16 countries in the region. Several observations are worth making. First -and not surprisingly -the Fitch credit ratings are highly and negatively correlated with the spreads (i.e., the higher the credit rating, the lower the spread), with a correlation of -0.78 (significant at the one percent level). Second, with the exception of Mexico, the green bars denoting investment grade countries are the ones with the lowest spreads. Thirdly, the average spread for LAC countries (310 basis points) is about twice as high as that of Asian countries.  Finally, it should come as no surprise that the two highest spreads are for Argentina and Ecuador (both currently under IMF programs).

Poverty in LAC: Trends and Cycles
Since the main focus of this report in the following chapters will be the effects of the business cycle on various social indicators -particularly poverty -we conclude this first chapter by providing a brief and very broad overview of poverty in the region.
As is well-known, monetary poverty reflects the share of the population below some income threshold. Naturally, different income thresholds may be used to evaluate monetary poverty. One commonly-used threshold is 1.9 dollars per person a day (2011 PPP), typically referred to as extreme monetary poverty. 9 As detailed in World Bank (2018b), extreme poverty stood at 10 percent of the world's population in 2015, down from 36 percent in 1990. While this is a remarkable feat, 10 percent equates to 736 million people in the world still living in extreme poverty. In LAC, only 4 percent of the population lives in extreme poverty. Further, as illustrated in Figure 1.14, Panel A, the reduction in extreme poverty has been quite remarkable, falling from 13 percent in 1995 to 4 percent in 2017. 9 PPP refers to purchasing power parity; see Appendix E.  Given the low incidence of extreme poverty in LAC, as in many other upper-middle-income countries around the world, a more informative threshold commonly used for upper-middle-income countries Of course, these dramatic gains in terms of the reduction of both extreme and monetary poverty vary considerably across countries, as illustrated in Figure 1.15. 10 While many LAC countries have essentially eliminated extreme poverty or reduced it way below 10 percent, it continues to be very high in countries such as Honduras, and, particularly, Haiti. In contrast, monetary poverty is still widespread in the region with almost two-thirds of countries (11 out of 18 in Figure 1.15) having a poverty rate above 20 percent.

FIGURE 1.15. Latest Poverty Rates for LAC Countries
Notes: Poverty rates for the year 2017, except for Dominican Republic and Mexico (2016), Guatemala and Nicaragua (2014), and Haiti (2012) This heterogeneity in poverty rates across countries in the region is obviously lost when regional aggregates are considered, such as in Figure 1.14. In fact, poverty has increased sharply in some countries in LAC since the end of the Golden Decade. In particular, Brazil, which represents one third of the region's population, has seen an increase in monetary poverty of about 3 percentage points between 2014 and 2017. The recent increase in poverty in Brazil as a result of the recession highlights the important fact that the business cycle may have significant repercussions on poverty. In effect, we would expect periods of slowdown/recession to reverse part of the gains in the reduction of poverty that are achieved in good times (such as the Golden Decade). While obvious, this fact seems to have been often overlooked by the poverty literature, which tends to measure the effect of growth on poverty without distinguishing between the trend and the cycle in GDP. 11 Hence, during good times, we would want to control for the cyclical effects on poverty before celebrating those gains as permanent. The next chapters, the core of this report, will analyze in great detail the impact of the business cycle on the behavior of social indicators, particularly poverty.  Figure 1.14. Source: SEDLAC (CEDLAS and World Bank). 11 See, among many others, Bourguignon (2003) and Ferreira et al. (2013).

Introduction
When examining the evolution of social indicators over recent decades, we should always keep in mind that any change in the underlying indicator can be decomposed into a transitory component, typically driven by cyclical factors, and a more persistent or "permanent" component that responds to structural factors. Taking this distinction into account is critical for policymakers since policies and programs implemented to address the cyclical behavior of social indicators will be necessarily different from those designed to improve structural factors. Moreover, measuring the success in the fight against poverty using social indicators with large cyclical components could be misleading since the analysis would be highly sensitive to the time span under study. In other words, a policymaker would draw very different conclusions if the response of poverty were evaluated during a boom or a complete (boom-bust) business cycle. In fact, the importance of the cyclical component in social indicators is magnified for the case of emerging markets subject to large external shocks, such as changes in the terms of trade, global liquidity, and world economic activity. All these shocks are cyclical in nature and thus will tend to amplify emerging markets' business cycles and, in turn, the transitory components of social indicators.
This chapter is devoted to understanding the role of transitory versus structural components in the evolution of relevant social indicators such as unemployment, monetary poverty, or unsatisfied basic needs (UBN). 12 Given that income is one of the most important drivers of economic and social welfare, this chapter uses the business cycle (i.e., the transitory component of national income) and long-term income changes to proxy for the transitory and permanent components of our set of social indicators, respectively.

How Cyclical are Social Indicators?
The first key message that follows from a simple trend-cycle decomposition is that the relative importance of transitory versus permanent changes differs greatly across social indicators. To highlight the size and importance of these differences, Figure 2.1 normalizes to 100 to all four measures for the year 2003 and follows the improvement of social conditions until 2014 (the period in-between the two vertical bars). This period is typically referred to as the Golden Decade due to the long-lasting boom in commodity prices. Depending on which social indicator we focus on, a very different picture emerges. Both unemployment and monetary poverty had a strong response to the 13 For sure, there are, among others, two social indicators that are typically associated with structural factors: the Multidimensional Poverty Index (MPI) and the Human Capital Index (HCI), the latter recently developed as part of the 2019 World Bank Human Capital Project. Unfortunately, the MPI series are not comparable over time, which is obviously crucial for our analysis and a sufficiently large dataset is not yet available for the HCI. These measures, however, will be a highly relevant resource for future research on permanent social gains.  For a casual observer standing in the year 2014, taking the large cyclical gains in unemployment and monetary poverty at face value would lead to an over-optimistic (and, in fact, misleading) evaluation of the permanent improvements in social conditions in the region. This biased view of reality becomes evident once the economic cycle begins to take a turn for the worse in 2013 and a large part of these social gains quickly start to dissipate. Had she been more careful, our casual observer could have prevented such over-optimism (or conveying a misleading picture) by either controlling for the cyclical component of unemployment and monetary poverty or simply basing her analysis on measures uncorrelated with the business cycle such as the UBN indicator.
The variance decomposition presented in Figure 2.2 for a sample of 15 LAC economies and Figure  2.3 for a worldwide sample formalizes the above intuition. The height of the bars in both figures denotes the share of the cyclical component of real GDP per capita in the total variance of each indicator. 14 Specifically, the share of the total variance explained by the business cycle is much higher for unemployment and monetary poverty than for structural measures of social welfare such as the 14 As shown in Appendix D, the shares of the cyclical and trend components add up to 100 percent.  In particular, negative real or monetary shocks would lead to short-term rises in unemployment. Our results for the U.S. economy confirm, as expected, that the share of the overall unemployment variance explained by the cyclical component of output is around 90 percent. In sharp contrast, since UBN comprises factors that are structural in nature and thus much less responsive to the business cycle, we would indeed expect the trend component to play a much more important role.

FIGURE 2.3. Contribution of Cyclical Component of Real GDP per Capita to Total Variance of Social Indicators (World Sample)
As follows from the above discussion, an interesting quantitative difference arises between LAC and the world when it comes to the share of unemployment explained by the cyclical component of output (74 percent in LAC versus 48 percent for the world sample). Why would this be the case? Without taking a stand into possible structural differences in labor markets between LAC and other emerging markets and how they may respond to temporary shocks, it is worth pointing out that we can account for most of the gap based on the higher output volatility experienced by LAC economies, typically exposed to large external shocks. For a given and similar structural reaction of unemployment to transitory shocks, the share of the variance explained by such shocks grows mechanically with their volatility. 16 In fact, a simple example suggests that, all else equal, the above difference (between, roughly, 70 and 50 percent), can be explained by a difference in output cycle variances of around 60 percent (compared to an actual difference of around 50 percent). 17 Unlike unemployment, the UBN indicator for LAC and the HDI for the worldwide sample are mostly driven by changes in structural factors, such as improvements in housing, education, and health, which are typically carried out over long periods of inclusive economic growth. Finally, changes in monetary poverty will, by construction, depend on the evolution of income per capita and changes in its distribution. 18 How much the business cycle affects economic welfare will ultimately depend on the existence of automatic stabilizers such as unemployment benefits and/or other policy buffers. Since both the underlying macroeconomic volatility and the effectiveness of different social policies may vary substantially across countries, we would expect that the relative importance of the business cycle in explaining changes in monetary poverty would also vary significantly across countries. This is precisely what we find in the data, leaving us with a very important corollary to our first insight: not only is the share of the cyclical component of output different across social indicators but, in the case of monetary poverty, it is also heterogenous across countries.    monetary poverty indicator are highly heterogenous across economies. As a result, when analyzing long term gains in social welfare across the region, our casual observer may be misled not only by focusing on cyclical indicators like unemployment but also by concentrating on countries like Argentina or Uruguay where monetary poverty is idiosyncratically cyclical.

What Drives the Cyclical Behavior of Monetary Poverty in LAC?
The variance decomposition analysis used so far in this chapter to evaluate the relative importance of the business cycle in explaining the evolution of social indicators rests on two key factors. The first is the elasticity of social indicators with respect to changes in the business cycle (what some may call "poverty multipliers"); that is, how much social indicators react to transitory changes in income levels. Differences in these poverty multipliers are typically associated with structural or policy differences across indicators and countries. For example, in the case of the monetary poverty indicator, differences across countries may be explained by structural factors such as the distribution of income or the level of income per capita (in particular, how far away is the mean income from the poverty line) as well as social policies, such as conditional cash transfers or unemployment benefits, to insulate the most vulnerable against income shocks. The second factor is the output cycle volatility relative to its trend. All else equal, the higher the volatility of transitory shocks relative to long term growth, the higher will be the share of the social indicator explained by the business cycle (Figures 2.5 and 2.6).

Volatility of the Cyclical Component of the Real GDP per Capita
Since Chapters 3 and 4 will be partly devoted to better understanding the poverty multipliers, this section focuses on analyzing the effects of output volatility on the cyclicality of social indicators.
Output volatility becomes crucial to explain another relevant stylized fact uncovered by Figures 2.2 and 2.3: LAC displays a substantially larger cyclicality of unemployment and monetary poverty than the rest of the world. 19 Figure 2.5 helps further visualize these differences comparing the cyclicality of monetary poverty in LAC to a sample of emerging East Asian economies. As the figure shows, the cyclicality of monetary poverty is more than three times higher in LAC than in East Asia.
While these sizeable differences in the average cyclical shares are difficult to account for based on structural or policy differences across regions, we know from a long-standing empirical literature that LAC is one of the most volatile regions in the world (see, for example, Végh et al., 2018). Countries in LAC are highly exposed to volatile external factors such as commodity prices, international liquidity, and movements of goods and capital as well as frequent political and institutional instability. The large volatility in the region's business cycle could explain why LAC's social indicators are more driven by transitory movements in output than in other emerging markets.

The Perils of Random Sampling
This chapter has shown that not all social indicators are created equal, especially when it comes to their degree of cyclicality. These stylized facts lead to a powerful policy warning: given the prevalence of temporary gains and losses in the measures of social welfare, a policymaker focusing on an indicator with a relatively short time span of available data could be highly misled in her efforts to evaluate permanent improvements in social conditions.
To illustrate this warning, consider a simple example with two countries and two different indicators: monetary poverty, which will be quite cyclical, and UBN, which will be rather unresponsive to changes in the business cycle. Further, to factor in the volatility dimension, we purposefully pick two large economies in our region with very different shares of cyclicality in monetary poverty, Argentina and Chile (Figure 2.7, Panels A and B, respectively).  1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Figure 2.7 display the evolution of real GDP per capita, monetary poverty, and UBN for Argentina and Chile, respectively. We can already see large differences in both the business cycle and monetary poverty across the two economies. Argentina endures large swings in real GDP that are matched by relatively large responses in the monetary poverty indicator. In sharp contrast, the amplitude of Chile's business cycle is very small and both real GDP and monetary poverty seem to be clearly driven by a trend. Interestingly, in both cases, the UBN indicator is driven by a trend regardless of the underlying business cycle.
Suppose now that our policymaker is trying to evaluate the long-run improvement of social conditions but is restricted to one decade of data. Finally, Figure 2.7, Panels E and F translate the previous results into a density function for the 10-year moving differences in monetary poverty and UBN, respectively. The differences in the variance of the distributions for monetary poverty in Figure 2.7, Panel E tell us that a policymaker from Argentina is more exposed to relatively large errors than one from Chile. The fat tails in the distribution for Argentina also indicate that those errors could be extremely large. Using instead UBN as an indicator not only lowers the chances of sample-dependent errors but also eliminates the differences across countries, as shown in Figure 2.7, Panel F.

Are Structural Indicators such as HDI Structural Enough?
This chapter has emphasized the potential bias implicit in cyclical social indicators when trying to evaluate permanent improvements in social conditions. A simple solution would be to extract the trend component of each indicator and use that trend in the evaluation of long-run gains in poverty reduction. The problem with this approach is the risk of sample-dependent bias explained above due to the short sample of available data. An alternative solution comes from the type of indicator used. Throughout the chapter, we have seen that social indicators based on structural factors such as UBN or HDI are much less influenced by transitory shocks. The large trend component in these indicators implies that we can better proxy long-run improvements in social conditions with relatively short periods of available data.
Having said that, indicators such as HDI are also subject to some degree of cyclicality because of the composition of the index. Figure 2. and shows their share of cyclicality for LAC. While structural components such as education and life expectancy are quite unrelated to the business cycle, the measure of income included in the index is, as expected, much more influenced by the business cycle. Again, including such cyclical components in the construction of structural measures like HDI would affect some countries more than others depending on the underlying volatility of the business cycle, as shown in Figure 2.8, Panel B where the cyclicality of the income component is twice as large in Argentina than in Chile.

What Have We Learned?
This chapter has uncovered a set of important stylized facts regarding the cyclical nature of social indicators. First, not all social indicators are created equal. Indicators such as unemployment or monetary poverty tend to comove with the business cycle. Second, structural indicators such as UBN or HDI mainly co-move with long-run economic growth. Thirdly, contrary to our findings in unemployment, UBN, and HDI, the degree of cyclicality in monetary poverty is highly heterogenous across countries.
These differences may be due to variations in the amplitude of business cycles across countries or the effectiveness of anti-poverty measures designed to insulate the most vulnerable individuals from transitory shocks in the economy. Finally, both unemployment and monetary poverty tend to be more cyclical in LAC than in other emerging markets. Part of this difference may be due to the higher volatility of output endured by our region compared to other emerging markets.
These stylized facts are particularly relevant for policymakers trying to evaluate the long-term changes in social conditions using relatively short time series. Our results alert about the possibility of sampledependent conclusions when using highly cyclical social indicators. This risk is higher for countries exposed to large income cycles. To avoid these biases, policymakers should rely on structural indicators dominated by trend components to evaluate long-term improvements in social welfare.

Chapter 3 Decomposition of Fall in Monetary Poverty during the Golden Decade in LAC Introduction
The previous chapter showed that while some social indicators (such as the unemployment rate) are tightly linked to the business cycle and will thus fluctuate considerably in the short run, others (such as UBN and HDI) follow essentially the long-run trend in GDP per capita, and hence bear almost no relation to the business cycle. Interestingly, in the case of the LAC region, monetary poverty lies somewhere in between the two extremes just mentioned, with large heterogeneity across countries.
In particular, whether monetary poverty behaves more like the unemployment rate (i.e., affected mainly by the business cycle) or UBN and HDI (i.e., determined essentially by the long-run trend in GDP per capita) crucially depends on the volatility of the business cycle. In high volatility countries (such as Argentina), changes in monetary poverty are much more closely related to the business cycle than in low volatility countries (such as Chile).
Given this insight, a natural question arises: how much of the fall in monetary poverty during the Golden Decade was permanent and how much was transitory? As argued in Chapter 1, the LAC region made tremendous progress in terms of reducing monetary poverty during this period, with the headcount ratio for the 5.50-dollar poverty line falling by almost 20 percentage points (from 42.2 percent in 2003 to 23.4 in 2014). Needless to say, the relative importance of permanent versus temporary reductions in poverty has critical implications for (i) assessing the "true" magnitude of the fall in poverty and how lasting it may be, and (ii) the public policies that may be put in place to address/influence permanent versus temporary falls in poverty (which in turn are likely to differ across countries depending on output volatility). To decompose the fall in monetary poverty between permanent and transitory components, we will follow a very helpful methodology proposed in a seminal paper by Datt and Ravallion (1992).

The Datt and Ravallion Decomposition
Using microdata from household surveys, Datt and Ravallion (1992) decompose the change in poverty between two points in time into (i) a "growth component" and (ii) a "redistribution component." The growth component is identified as the change in monetary poverty between two years that would have occurred if the income of each household member had changed in the same proportion as the national mean income, keeping constant the shape of the income distribution (measured by the Lorenz curve).
The redistribution component reflects the change in monetary poverty that would have occurred if the income distribution had changed as it did, but with no changes in mean income. 20 21 Applying Datt and Ravallion's (1992) Lustig, 2010 and. As will be further discussed in Chapter 4, the resources allocated to these programs increased as a percentage of GDP in the 2000s, as did the number of countries implementing them and the share of the population covered (Stampini andTornarolli, 2012 andCecchini andMadariaga, 2011). These programs have a higher redistributive impact than social spending in general because they target specifically those in greatest need. Further, the fact that transfers are in cash implies that the programs have a direct impact on income inequality, the most common measure of inequality (Gasparini, Cruces, and Tornarolli, 2016).
To further split the growth component into a "cyclical income component" and "trend income component," we need to know the growth rate of the mean income trend during the period under analysis. 24 If one had a long time series of mean income, one would calculate its trend component by using any trend-cycle filter. Unfortunately, for most countries in LAC, data are not available for very long periods (say, for more than 25 years). For this reason, and as is common practice in this literature (Ahluwalia et al., 1979;Bourguignon and Morrison, 2002;Sala-i-Martin, 2006;and Bhalla, 2002), we proxy the average trend growth in mean income by the average trend growth in real GDP per capita. 25 That is, almost half of the fall in monetary poverty was due to temporary income factors and about 20 percent to permanent income gains. In terms of population, this would imply that around 50 million people got out of poverty in LAC during the Golden Decade due to cyclical income gains. Therefore, one could argue that this group might be at risk of falling into poverty again, as the temporary gains achieved during the expansionary phase of the business cycle continue to dissipate.

Decomposition for Commodity Exporters and Importers
Panel A in Figure 3.3 presents the same decomposition as • First, it is worth noting the significant cross-country heterogeneity observed in the relative importance of the redistribution component (Figure 3.3, Panel A). A myriad of policy and non-policy factors could lie behind such heterogeneity. While a detailed analysis would fall outside the scope of this report, some obvious factors come to mind. First, political ideology: even though redistributive policies have increased markedly over the last 25 years regardless of ideology, left-of-center governments are typically more likely to engage in additional redistribution than right-of-center ones. Second, bureaucratic quality or governments' efficiency in designing and implementing social programs. Thirdly, how progressive government transfers are. Fourth, reductions in labor income inequality, perhaps due to a reduction in the skill premium. Finally, factors that affect inequality, such as returns to capital, private transfers, and remittances, will also have an impact on the redistribution component (since for given mean income, the share of the population living in poverty will be affected). 28 • Second, and as follows from Figure 3.3, Panel B, the trend income component is more important for commodity importers than exporters. Moreover, in some cases like Honduras and the Dominican Republic, the cyclical income component is actually negative. Why? While one tends to link higher commodity prices to higher incomes, it should be kept in mind that this is only true for commodity exporters. In contrast, for commodity importers, higher commodity prices are typically associated with lower incomes. This explains why Central America and the Caribbean had lower growth rates than South America during the Golden Decade. In other words, the Golden Decade was essentially a South American story. Not surprisingly, then, several commodity importers had a negative cyclical income component as actual income was below trend income. 29 • Thirdly, within commodity exporters, there is a large heterogeneity across countries in terms of the relative importance of the trend component (  Lustig et al. (2013), and Messina and Silva (2018). 29 While the negative contributions of the cycle in Honduras and Dominican Republic may seem strikingly big at first (-50 and -70, respectively, as shown in Figure 3.3, Panel A), it should be noted that these figures are expressed as a percentage of the total improvement in poverty which, in both countries, was quite small during the Golden Decade. Thus, these big numbers are the result of two relatively mild forces pushing in the same direction; that is, a relatively poor macroeconomic performance on top of relatively small gains in the fight against poverty. For instance, the 50 percent negative impact of the cycle found in Honduras was due to an increase of 4 percentage points in monetary poverty due to the negative business cycle over an overall decrease in monetary poverty of just 8 percentage points (one of the smallest in the region).   Cyclical Income Trend Income the discussion in Chapter 2, in countries like Chile and Mexico (where income growth is dominated by the trend), the role of the trend income component is more important than in countries like Argentina and Brazil (where income growth is dominated by the cycle). Put differently, even after "controlling" for the redistribution component, we obtain the same results regarding cycle and trend as in Chapter 2.

The Role of Shock Absorbers
Given the importance of the cyclical income component in the fall in poverty during the Golden Decade and the fact that this period included the Global Financial Crisis of 2008-2009 (where real GDP per capita in LAC fell on average by 1.6 percent) a natural question is how structural social policies dealing with poverty have been able to address recessionary environments (i.e., as a shock absorber).      (Figure 3.5, panels A and B, respectively).
• Second, the role of redistributive policies also changed in a very important dimension; that is, in dealing with a recessionary environment (i.e., acting as a shock absorber). Before the Golden Decade -and in line with its limited structural role as well -the redistribution component played no role in helping the economy cope with changes in poverty caused by the business cycle. During the Tequila Crisis (1995)(1996), for example, poverty in LAC increased by 2.2 percentage points, mainly driven by the cyclical components in Argentina and Mexico ( where we can see that cyclical considerations would have called for an increase in poverty but this was exactly offset by redistributive policies, so that poverty did not change at all. As the next chapter will show, the nature of these shock absorbers was not the same as in developed countries (where they operate mainly through unemployment insurance), but rather by an unintended consequence of how the conditional cash transfers work over the business cycle. The number of beneficiaries tends to increase in bad times but not fall in good times, which helps poor people in bad times but imparts a downward rigidity in these types of programs that make them unsuitable to handle cyclical shocks and, in turn, could tend to perpetuate some inherent problems the LAC region faces in terms of reducing informality.

Chapter 4 Social Transfers in LAC: In Search of Shock Absorbers Introduction
As discussed in Chapter 3 -and in the poverty literature in general -redistributive policies have become more important in reducing poverty in LAC since the onset of the Golden Decade (although, on average, the contribution of redistributive policies was smaller than the cyclical income boost). In practice, moreover, redistributive policies have also acted as shock absorbers and prevented, for example, a reversal in social gains during short recessionary periods such as the Global Financial Crisis (2008)(2009). What were the main factors behind the successful expansion of redistributive policies, which even proved resilient in response to some harsh economic conditions? As suggested below, the reasons are two-fold: (i) the incremental implementation of social programs throughout the region in a much greater scale than before, and (ii) higher social spending during recessions (i.e., shock absorbers). Unfortunately, the role of redistributive policies as shock absorbers has been, more often than not, a welcome but unintended consequence of programs that had been originally designed to address long-run (structural) poverty conditions rather than to smooth out the business cycle. An important policy lesson is thus that the region should develop, in addition to existing structural programs, social safety net tools that can support the poor and most vulnerable over the business cycle.

CCT Programs in LAC: An Overview
One of the major drivers of the increase in social assistance spending throughout the region was undoubtedly the implementation of programs based on conditional cash transfers (CCTs). As indicated in Figure 4.1, in 1998 only 3 out of the 15 LAC countries analyzed in the previous chapters had CCTs: Brazil (PETI), Honduras (PRAF/IDB II), and Mexico (Progresa). The Golden Decade saw a rapid expansion of these programs and, by 2018, all the countries had sophisticated CCTs, with an average expenditure equivalent to 0.4 percent of GDP. As commonly defined in the literature, conditional cash transfers are, by design, "structural" programs that target long-term poverty (Ravallion, 2016). By providing monetary incentives in exchange for investments in health and human capital accumulation, CCTs' aim is to break the intergenerational transmission of poverty.

Shock Absorbers: A Pending Agenda
If CCT programs are among the most popular structural tools for long-term poverty reduction, then policymakers must also have cyclical instruments at their disposal to smooth household income over the business cycle. Without proper cyclical buffers (i.e., shock absorbers), social gains could be at risk as people could fall below the monetary poverty line in response to negative income shocks (Baéz et al., 2017). To this effect, perhaps the best-suited shock absorber is unemployment insurance, as it guarantees that, even in the extreme event of job loss, a minimum income will be available to maintain a healthy living standard during a predetermined period of time.
As shown in Figure 4.1, unemployment insurance is, by and large, work in progress in the region's social agenda. As of 2018, only 6 out of 15 countries had unemployment insurance as part of the social security system (with Ecuador being the last one to adopt it in 2016).

Country Program Description
Argentina Enactment year: 1991 Eligibility: Private salaried workers dismissed without just cause who have 6-8 monthly contributions to social security. Duration of coverage: 2-12 months. If the unemployed is 45+ years old, coverage is extended for 6 months.
Value of benefit: The benefit consists of a basic allowance plus the payment of family allowances, medical coverage, and recognition of seniority for retirement. The basic allowance is variable, calculated as a percentage of a reference value of half the best monthly net salary in the 6 months that preceded unemployment, with decreasing monthly installments.

Brazil
Enactment year: 1986 Eligibility: Salaried workers dismissed without just cause who have 12 monthly contributions to Social Security. Duration of coverage: 3-5 months.

Value of benefit:
The benefit consists of a basic allowance, which takes, as a reference, the average monthly salary in the 3 months that preceded unemployment. Chile

Enactment year: 2001
Eligibility: Private salaried workers. Workers may access the Individual Severance Account (ISA) and the Solidarity Severance Fund (SSF). To access the ISA, workers must have at least 6-12 monthly contributions, depending on the contract type, and at least 12 contributions in the case of the SSF. Duration of coverage: 1-13 months in the ISA; 3-5 months in the SSF, depending on the contract type.

Value of benefit:
The benefit consists of a basic allowance plus the payment of family allowances, medical care coverage and recognition of seniority for retirement. The basic allowance is variable, with monthly payments defined as percentages of an average monthly salary (of the last 6-12 months, depending on the contract type).

Colombia
Enactment year: 2013 Eligibility: Workers that contributed to the Family Allowance Fund for 1-2 years during the 3 last years. Duration of coverage: 6 months. Value of benefit: Contributions to the health and pension systems (taking as base a legal minimum monthly wage), a food bonus (of 1.5 times the legal minimum monthly wage, distributed equally for 6 months), access to the prevailing family subsidy, and assistance in the search of new employment.

Ecuador
Enactment year: 2016 Eligibility: Salaried workers with (i) at least 24 contributions to the social security system (with the last 6 contributions in consecutive months), (ii) 60 days of unemployment, and (iii) involuntarily unemployment. Duration of coverage: 5 months.

Value of benefit:
The benefit consists of a basic allowance that varies per month, and is computed as a percentage of the average monthly salary of the 12 months that preceded unemployment with decreasing payments.

Uruguay
Enactment year: 1981 Eligibility: Unemployed or suspended private salaried workers. Duration of coverage: 6 months in the case of dismissed workers, and 4 months in the case of suspended workers.

Value of benefit:
The benefit consists of a basic allowance that varies per month and is computed as a percentage of the average salary of the 6 months that preceded unemployment with decreasing payments.
On average, unemployment insurance in the region offers benefits for up to 6 months, with payments decreasing gradually over time to increase incentives for job search. In these countries, the average gross replacement rate for the first year of unemployment is 20 percent, which is less than half the OECD average. 32 To make matters worse, given the eligibility requirements, only an average of 17 percent of the unemployed in these LAC countries receive unemployment benefits (with the average falling to 6 percent if Chile and Uruguay are excluded). As discussed by the International Labour Organization (2017) and Izquierdo et al. (2019), the high degree of informality is one of the main factors limiting the collective funding of unemployment insurance schemes and, thus, the possibility of increasing their coverage.

Behavior of CCT Programs throughout the Business Cycle
As argued above, the progressive expansion of social spending (particularly in the form of CCTs) since the beginning of the Golden Decade is surely a key factor behind the large contribution of redistributive policies to the reduction of poverty discussed in Chapter 3. It would thus prove useful to study the behavior of real CCT spending in individual LAC countries to assess its sensitivity to the business cycle. This is an important question because, in a region with limited shock absorbers (like unemployment insurance), the fact that redistributive policies have contributed to poverty reduction, even in recessions, may indicate that other social policies are acting, unintendedly, as shock absorbers.  Figure 4.2, Panel C). As discussed in Izquierdo et al. (2019), the underlying force behind this phenomenon is that, as both unemployment and poverty rise during recessions, a larger share of the population becomes eligible for CCT benefits and takes advantage of these programs to compensate for the fall in household income. Therefore, CCTs, which are de jure structural policy tools, become de facto shock absorbers. • Second, real CCT spending is rigid, especially downward. Figure 4.2 shows that once real CCT spending expands, it is very unlikely to decrease. This is particularly the case for the steep spending increases experienced during economic slowdowns (i.e., even after many years, real CCT spending does not return to pre-recession levels). This has contributed to the significant  Real Expenditure in CCTs Real GDP Cycle (right axis) growth in the total expenditure of these programs, which has risen by a factor of 2, 5, and 15 times that of the first year in the case of Argentina, Brazil, and Chile, respectively. From the perspective of the policymaker, both the de facto shock absorber function and rigidity of CCT spending should be corrected to prevent the buildup of fiscal imbalances that may later compromise the funding of the programs. If CCTs are intended to alleviate structural poverty, then instances in which real CCT spending is suddenly cut down -such as fiscal adjustment efforts in Argentina (2017, Panel A) and Brazil (2015-2017, Panel B) -should be avoided to protect the most vulnerable. This downward rigidity is particularly evident since it has occurred during a period of tremendous income growth.
In sum, the steady increase in social transfers has been an important driver of the prolonged contribution of redistributive policy to poverty reduction in LAC. However, there are crucial caveats that must be corrected so that each individual policy instrument can fulfill its intended role. In the case of CCTs, efforts must be made to ensure that the programs are truly structural and target the most vulnerable. In the case of shock absorbers, such as unemployment insurance, coverage must be extended (which may require other policy efforts such as reducing informality). Only in this scenario will each social instrument work as intended and the response of poverty to the business cycle be greatly reduced.  Real GDP Growth (as percentage) models. 33 All possible ensemble combinations with two independent variables in the full model result in two estimated models. By alternating each independent variable between included versus excluded, the final shares are derived as a weighted average marginal/incremental contribution to the overall fit made by an independent variable across all models in which the independent variable is included. These values are equivalent to Shapley values (Shapley, 1953). Since the shares of the variance of the social indicators explained by the output cycle and output trend are reported in relation to the R 2 of regression (2), they will add up to 1 (i.e., 100 percent of the R 2 ).