WPS5791 Policy Research Working Paper 5791 Fewer Jobs or Smaller Paychecks? Aggregate Crisis Impacts in Selected Middle-Income Countries Gaurav Khanna David Newhouse Pierella Paci The World Bank Poverty Reduction and Economic Management Network Poverty Reduction and Equity Unit & Human Development Network Social Protection and Labor Unit September 2011 Policy Research Working Paper 5791 Abstract This paper reviews evidence from 44 middle-income considerably across countries. For a given drop in gross countries on how the recent financial crisis affected jobs domestic product, earnings declined more in countries and workers’ incomes. In addition to providing a rare with larger manufacturing sectors, smaller export assessment of the magnitude of the impact across several sectors, and more stringent labor market regulations. In middle-income countries, the paper describes how labor addition, overall employment became more sensitive to markets adjusted and how the adjustments varied for growth in gross domestic product. These findings have different types of countries. The main finding is that implications that go beyond the recent financial crisis the crisis affected the quality of employment more than as they highlight (i) the limitations of focusing policy the number of jobs. Overall, the slow-down in earning responses on maintaining jobs and providing alterative growth was considerably higher than that in employment, employment or replacement income for the unemployed, and the decline in gross domestic product was associated and (ii) the critical role of fast-track data systems that are with a sharp decline in output per worker, particularly capable of monitoring ongoing labor market adjustment in the industrial sector. In several counties, hours per during economic downturns, in supporting the design of worker declined and hourly wages changed little. But effective policy responses. both the magnitude and nature of the adjustments varied This paper is a joint product of the Poverty Reduction and Equity Unit, Poverty Reduction and Economic Management Network; and the Social Protection and Labor Unit, Human Development Network. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at dnewhouse@worldbank.org or ppaci@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Fewer Jobs or Smaller Paychecks? Aggregate Crisis Impacts in Selected Middle-Income Countries Gaurav Khanna†, David Newhouseǂ, and Pierella Paciǂ † Department of Economics, University of Michigan ǂ World Bank, Washington DC Introduction While policy responses to economic downturns tend to focus on protecting employment, evidence from the developing world shows that employment reduction is not the only, or even the most significant, effect. Although economic crises are difficult to predict, their recurrence is a salient feature of emerging market economies. Nevertheless, many developing countries continue to lack an effective policy infrastructure that can mitigate the impacts of economic downturns on employment opportunities. Moreover when this infrastructure exists it mostly focuses on either maintaining existing jobs, providing alterative employment, or replacing income for workers that lose formal sector jobs via unemployment benefits. However, job losses are only one of the possible ways in which labor markets adjust to economic downturns and the prevailing adjustment mechanism depends on the nature of the shock and on the characteristics of both the economy and the labor market. During past crises, workers in developing countries typically faced large declines in hourly wages or to employment reallocation across sectors rather than major reductions in employment. 2 The experiences of the global downturn of 2008-2009 highlights a common disconnect between the labor market impact of crises and policy responses. Identifying the labor market channels through which the economic downturn is transmitted is a precondition for effective targeting of policy interventions.3 If first- round labor market adjustments are concentrated in specific jobs, sectors or geographic areas, targeted employment interventions to protect those most immediately affected may effectively mitigate crisis impacts.If, by contrast, most of the adjustment occurs through generalized wage declines, standard employment-focus policies are not the most effective instruments to mitigate the impact of a crisis. But lack of fast-track data to monitor changes in main labor market indicators in a timely manner often prevents effective targeting of policy responses. This is painfully highlighted by the disconnect— described is some detail in Robalino, et al (2011)—between the prevailing labor market adjustment experienced during the global downturn of 2008-2009 and the majority of policy responses implemented. Much of this disconnect was due to the lack of adequate information on the prevailing adjustment mechanism in the developing world. In general, most analysis of the magnitude and nature of labor market adjustments to the 2008 economic shock has been limited to OECD countries, and quantitative cross-country evidence on the impact of the great recession on labor market outcomes in developing countries has been scarce.4 This paper summarizes evidence from 45 middle income countries on how the 2008 crisis affected labor markets. There are four main findings. First, as in previous crises, adjustments occurred more through the quality of jobs than their quantity. For the average country in our sample, earnings growth slowed by 3 times as much as employment growth, and unemployment changed little. Second, for several countries, hourly wage growth did not collapse and the adjustment came mainly via large reductions in hours worked. Third, while labor market adjustments generally tracked GDP declines, there was considerable 2 This was particularly true of the East Asian crisis in the late 1990’s (Betcherman and Islam, 2001, Fallon and Lucas, 2002) but it is also true for Mexico’s ―tequila‖ crisis which resulted in a drop in real wages of 21% between 1994 and 1996 while aggregate employment increased slightly (McKenzie 2002). 3 See for example Paci, et al (2011). 4 The analysis of the present crisis in developing countries often focuses on changes in GDP (Calderon and Didier, 2009), Blanchard, Faruqee, and Das (2010). Other analyses that focus on labor markets are sometimes based on simulations using pre-crisis data (Global Econometric Trends, 2009 & 2010, World Economic Outlook 2010, Habib et al, 2010) 2 variation in outcomes across countries. For a given decline in GDP growth, countries with smaller manufacturing sectors, larger export sectors, and less costly redundancy regulations were better able to protect earnings. Finally, the responsiveness of employment growth to changes in GDP growth increased during the crisis, leading simple projections to understate the extent of employment loss during the crisis. Data and Methodology The paper uses data from published quarterly labor market statistics of 45 middle income countries. 37 of these countries have information on both GDP growth and employment growth, while 31 have information on GDP growth and earnings growth. Most of the countries lie in Latin America or in the Europe and Central Asia region (ECA). We divide the remaining countries into two regions: East Asia and Sri Lanka, and the Middle East and Africa.5 The main source of primary data sources on labor market and GDP are country statistics published by the relevant country statistical offices and provided, on a subscription basis, to the World Bank by three private data providers—Haver Analytics, CEIC, and EMED—that collect them daily. These data were supplemented by ILO’s laborsta database. 6 Data on other country indicators--such as per capita GDP, and the size of the manufacturing and export sector-- come from the World Development Indicators database. Data coverage and sources vary across countries and indicators. GDP, employment, and unemployment statistics are available quarterly for almost all countries in the sample. 7 Many countries, however, do not report earnings and even fewer provide hours. Moreover, hours and earnings are sometimes reported as changes in indices rather than levels. Since data are often missing for particular quarters, the analysis is limited to year-on-year changes for comparable quarters. Tables A2 and A3 in the appendix list the types of surveys used to generate employment and earnings indicators. In three quarters of the countries analyzed, employment statistics were derived from labor force surveys, while in the remaining one fourth, employment levels were obtained from establishment surveys or administrative data. As a result, in a quarter of countries analyzed employment trends may refer only to formal employees. By contrast earnings data comes from establishment surveys in 70 percent of cases and the countries, and in these countries under-represents informal or self-employed workers. In a few countries, (Mexico, Colombia, and Turkey), earnings data from establishment surveys only cover the industrial or manufacturing sectors, and most countries in Eastern Europe only collect earnings data for non-agricultural workers. Finally employment and earnings data often do not cover the full country. At least 16% of countries – typically countries in Latin America, as well as China – only collect employment data in urban areas. Results should be interpreted with appropriate caution. Much of the analysis focuses on the change in the wage bill as a convenient summary measure of the total magnitude of the labor market adjustment borne by workers. The wage bill is the product of employment and average monthly earnings. Given the nature of the data, changes in the wage bills often reflect changes in establishment earnings and total employment. Many of these countries have large informal sectors and the cyclicality of earnings in formal and informal establishments may differ, adding additional uncertainty to our estimates of the precise role 5 The five countries from the Middle East and Africa are Egypt, The West Bank, South Africa, Morocco and Mauritius. 6 A detailed list of sources and comments on the data are available from the authors upon request. 7 Usually, quarterly data is available. However, Peru and Indonesia only have half-yearly results. 3 of earnings in the adjustment process may not be precise. It is areassuring, however, that data on earnings from seven middle income countries also tend to show moderate to large earnings adjustments (Cho and Newhouse, 2011). The analysis compares growth rates during the crisis year with averages over the prior two years. Focusing on the change in the growth rate, rather than post-crisis declines, is important to accurately measure the effect of the crisis relative to pre-crisis trends. 8 Extending back two years allows a comparison with changes immediately preceding the crisis, while mitigating the influence of the rapid price increases in food and fuel in 2008. GDP growth in crisis-affected countries fell sharply in the fourth quarter of 2008. Therefore, the post-crisis period includes the four quarters from Q4 2008 to Q3 2009, while the pre-crisis period includes the eight quarters from Q1 2006 to Q3 2008.9 The Extent and Nature of the Labor Market Adjustment The cost of the crisis to workers was considerable but delayed and gradual. As shown in Figure 1 the average wage bill, which had been growing by nearly 10 percent over the pre-crisis years of 2007 and 2008, came to a near standstill in 2009, with the growth dropping to less that 1 percent. 10 The drop was only slightly less than the fall in GDP growth over this period (9.4 percentage points) but the decline in wage bill growth began later than that of GDP and continued longer. This is in line with findings from previous crisis that suggest long term impacts on workers may be deep and protracted. 15 10 Percent 5 0 -5 -10 GDP growth Employment growth Real earnings growth Wagebill growth Figure 1: The employment adjustment was delayed and gradual 8 . A simple example makes this clear. China was far more affected by the financial crisis in 2009 than Morocco. But a simple comparison of post-crisis rates of growth would have given the false impression that Morocco suffered more. This is because Morocco’s GDP grew only half as fast China’s, even after the crisis, and employment growth in the two countries was nearly the same. However, in the pre-crisis period the Chinese economy had been growing more rapidly, so GDP growth slowed by 6 percentage points more in China than it did in Morocco. Employment growth also slowed by 3.5 percentage points more. 9 Using eight quarters of pre-crisis data rather than four may mitigate the impact of the food and fuel crisis of 2007. 10 The sample used to construct Figure 1 consists of the 24 countries that reported GDP growth, employment growth and real earnings growth in each quarter 4 Reduced growth of earnings was the main culprit for the slowdown in the wage bill. 11 Real earnings growth began to decline in the first quarter of 2008, as increases in food and fuel prices led to accelerating inflation. The sharp decline in GDP growth in the fourth quarter of 2008 was mirrored by a relatively large decline in nominal earnings growth: nearly three quarters of the 8.7 percentage point slowdown in the wage bill, among countries for which data is available, came from the reduction in earnings growth, which fell 6.5 percentage points. This is largely consistent with the evidence from past crises, as pointed out in the introduction (For more details see footnote 1). The simultaneous drop in inflation of 2.8 percentage points, due to easing demand for commodities, prevented an even steeper drop in the wage bill and led to a small increase in the rate of real wage growth of 0.7 percentage points (Figure 2). 20 15 10 5 Percentage 0 Hours growth CPI growth -5 Real earnings growth Nominal earnings growth -10 Figure 2: For 8 countries - Hours pulled down earnings, while prices mitigated the impact Hours fell considerably in several countries. The slowdown in hours tracks closely the reduction in earnings growth (See Figure 2).12 On average, in the 14 countries for which data are available, growth in hours worked fell by a striking 7 percentage points but the drop was highly concentrated in only half of these countries (Figure 3). The decline in hours worked not only accounted for the entire decline in earnings but also allowed for a slight increase in real hourly wages (0.7 percentage points) (Figure 4). However, the extent to which the reduction in hours was an active policy response rather than simply an adjustment mechanism is unclear. For example, an ILO survey of 54 countries (10 LICs, 10 lower middle income, 17 upper middle income and 17 HICs) shows that reductions in hours worked was a common response to the downturn in countries as different as China, Vietnam, the Philippines, Indonesia, Mexico, Argentina, Colombia and Jordan (ILO, 2009). 11 Table A1 in the appendix shows that these conclusions hold when examining pre and post-crisis average growth in a slightly larger sample of 28 countries that report quarterly, semi-annual, or annual data. 12 Figure 2, however, underestimates the extent of the fall in hours, because it only reflects the 8 countries that report hours quarterly 5 5 Percentage points 0 -5 -10 -15 -20 Figure 3: Change in the growth of hours worked Decomposition of decline in Decomposition of decline Decomposition of wage bill growth in earnings growth increase in real wage % due to real growth earnings 74.7% % due to hours % due to prices % due to 25.3% employment % due to nominal % due to real wages wages -20.0% 20.0% 60.0% 100.0% -500.0%-300.0%-100.0% 100.0% 300.0% 500.0% Figure 4 : Decomposing the change in the growth of wage bill, earnings and wage rates Because of the relatively minor impact of the downturn on employment, output per worker grew more slowly than in previous years. Among the 33 countries for which the relevant data is available GDP growth fell by 7.9 percentage points. Of that drop, just over 6 percentage points were due to a decline in growth in per adult GDP while slower employment growth rate contributed only 1.7 percentage points and activity rates added very little (only 0.1 percentage points) (Figure 5). 13 This is consistent with the patterns for employment and hours discussed above; while employment growth slowed little, the growth rate of the number of hours worked dropped steeply, meaning that each worker spent less time producing output. 13 In line with our findings, some high income countries, like the US, had sharp rises in productivity – something that the US had not seen in past crises. 6 Slowdown in Growth in GDP per adult (Pct points) 7 6 5 4 3 2 1 0 due to productivity due to employment due to activity Figure 5: Decomposing the slowdown in the growth of GDP per adult But employment shifts out of more productive sectors also played an important role. Small aggregate employment changes masked significant shifts out of industry and into the agricultural and service sectors.14 The fall in industrial employment is not surprising, since manufacturing exports suffered during the crisis in many countries and entry barriers are lower for family businesses in retail trade and agriculture, facilitating employment shifts into these sectors. 15 Even within the service and industrial sectors, however, the high productivity industrial and service sectors suffered most. Figure 6 below shows changes in employment growth across different sectors for a subsample of nine countries with available level of disaggregation. 16 Employment growth in agriculture and low-productivity service sector fell slightly (less than 1 percentage point) while growth in high-productivity services fell 2.5 percentage points. Evidence from a larger sample of countries on more aggregate sectors suggests similar patterns. Both activity and unemployment rates, however, were largely unaffected. Contrary to evidence from previous crises, on average labor force participation rates remained almost unchanged (from 54.0 percent to 54.3 percent) suggesting that added and discouraged workers effects may have balanced out. Unemployment rates also hardly changed (from 9.9 percent to 10.3 percent), In other words, while very mild added worker effects may have increased participation, a smaller share of those in the labor force were employed.17 14 In the sample of 15 that report sectoral employment, agricultural employment growth increased by 1.5 % points, service sector growth by 0.2 % points and industrial employment growth fell by 3.3% points 15 Past crises estimates suggest a similar story: during the 1997 Asian crisis, some 30-40% of displaced urban workers are estimated to have moved to agricultural jobs (Manning 2000). Evidence indicates that worker reallocation and the quality of new jobs created are procyclical, rather than countercyclical (Bowlus 1993). The returns to labor in these small-scale sectors are low, but displaced workers often lack better options during downturns (Mead and Liedholm 1998). 16 High productivity services include: Transport and communications, real estate and financial services, and personal services. The remaining sectors, including public administration, real estate, education, hotel, retail trade, social work, and private household, are classified as low-productivity services. 17 Reduced earnings may lead to a rise in labor market participation. Some crises appear to have driven more women and children into the labor market, especially in rural areas (Manning 2000, McKenzie 2004 and Ezemenari et al 2003). However, in others participation rates appeared unresponsive. And in some cases underemployment rates rose significantly due to decreased labor demand (e.g. Mexico during the Peso crisis (McKenzie 2004), Thailand and 7 High productivity Services Manufacturing Construction Low productivity Services Agriculture -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 Percentage points Figure 6: Change in Employment Growth In which countries were adjustments unexpectedly large? Examining average labor market changes across such a varied sample of countries does not tell the whole story. While no country escaped the crisis unharmed, there was a great deal of heterogeneity across our sample of countries in the way labor markets adjusted. As suggested in the introduction, much of this heterogeneity is associated with fundamental differences in the severity of the shock, the structure of the economy, and the nature of labor market institutions. Figures 7 and 8 show how labor market outcomes varied according to the size of the shock. Countries more severely affected by the shock tended to adjust primarily through earnings, rather than employment. (Figure 7).18 Nevertheless unemployment increased faster in countries with larger GDP declines (Figure 8). However, both graphs clearly show substantial heterogeneity even for countries with identical GDP declines. This raises the question of which country characteristics helped cushion the labor market from the negative impact of economic shocks. South Korea in 1997/8 (Sarkar and Kumar, 2002), suggesting that workers had little ability to increase their labor supply. 18 The left side of figure 8 shows an increasing divergence between earnings and employment effects as the size of the shock increases. 8 10 8 Change in change (pct points) 6 0 4 Change 2 -10 0 -20 -2 -25 -20 -15 -10 -5 0 -25 -20 -15 -10 -5 0 Change in GDP growth Change in GDP growth (pct points) Earnings growth change Employment growth change Unemployment Participation Fitted earnings change Fitted employment change Fitted unemployment Fitted participation Figure 7 Figure 8 To assess the role of these other factors, this paper classifies countries along seven dimensions. The first two are region and level of per capita GDP, which are intended to capture structural differences in the economy and in labor market institutions. Countries are also classified according to the relative size of the manufacturing and export sector, two areas of the economy where crisis impacts were most severe. Exports were also affected by countries’ exchange rate regimes, as those with currencies pegged to the dollar suffering larger falls in exports, since the dollar appreciated roughly 15 percent against the Euro in the latter half of 2008. 19 Finally, countries are grouped according to their level of labor regulation (high/low) to investigate the expectation that stringent labor market regulations, if enforced, discouraged employers from shedding workers during a downturn. Two measures of firing costs from the 2009 Doing Business Indicators are used for this purpose: the difficulty of firing index and the firing cost.20 Only a few of these characteristics are closely correlated with each other. 21 19 We use the IMF’s de facto exchange rate-regime to classify countries by their currency regime. http://www.imf.org/external/np/mfd/er/2006/eng/0706.htm 20 Countries are classified as high manufacturing, high export, difficult to fire, and high firing cost countries, based on whether the indicator exceeds the median across countries in the sample. The difficulty of firing index measures whether redundancy is allowed as a basis for terminating workers, and the existence of rendundancy regulations and requirements. The redundancy cost indicator measures the cost of advance notice requirements, severance payments and penalties due when terminating a redundant worker, expressed in weeks of salary Additional information on the redundancy indices is available at http://www.doingbusiness.org/Methodology/Surveys/EmployingWorkers.aspx 21 In only 4 cases do the pair-wise correlations between characteristics exceed 0.25.Latin American countries are most likely to be upper middle income and least likely to be lower middle income. Europe and Central Asian countries and middle income MICS tend to have larger manufacturing sectors, but Latin American countries are less likely to have large manufacturing sectors. Finally, countries with high redundancy costs are far more likely to be in East Asia or Latin America than Europe, and are less likely to be middle income countries. Somewhat surprisingly, the correlation between a high difficulty of redundancy and high redundancy cost is weak and negative. 9 Higher firing costs are associated with Countries with larger export sectors, smaller employment loss and higher and less stringent regulation better earnings loss maintained previous trends in earnings without incurring large East Asia (omitted) employment costs. Figure 9 shows how *** earnings and employment are Region Europe Latin America *** associated with each set of country characteristics after accounting for the Middle East ** declines in countries’ GDP growth. 22 Low middle income (omitted) The most striking results are those Income Middle middle income ** associated with exports and firing Upper middle income regulations. High export countries were much better able to maintain Structur Small manufacturing sector … earnings, at negligible cost to e Large manufacturing sector employment. Countries with higher Exports Less dependent (omitted) redundancy costs were better able to More dependent ** protect employment, on the order of 1 ** Fixed or Peg (omitted) to 2 percentage points. However, this Exchange may have partially come at the rate Managed float expense of greater earnings reductions. Full float Countries with more costly and Low difficulty of redundancy cumbersome firing regulations had Regulation High difficulty of redundancy two to three percentage point greater Low redundancy cost declines in earnings. The relationship between firing regulations and High redundancy cost * earnings declines is only suggestive source Firm survey and not statistically significant, Data Household survey (omitted) however, due to the imprecision of the earnings measures. Change in employment growth -4 -2 0 2 4 6 Change in earnings growth Note: Bars indicate coefficients from fourteen separate cross-sectional regressions of changes in employment and earnings growth on change in GDP growth, data source, and each country characteristic. Stars indicate statistical significance at the 10 , 5, and 1 percent level. There are 35 or 36 observations in the employment regressions and 30 or 31 in the earnings regressions. Figure 9 22 Figure 9 reports coefficient estimates from 7 regressions. Each bar represents the coefficient from a regression of employment or earnings growth on the change in GDP growth, its square, cube, and quartic, and one set of dummy variables.Stars indicate significance level. For earnings and employment equations, the regression also controlled for whether the data was derived from a firm or a household survey. 10 Increases in unemployment, after Unemployment rose less in European and factoring in the size of the shock, lower redundancy cost countries were larger outside of Europe and in countries with higher redundancy East Asia (omitted) costs. After controlling for GDP Eastern Europe and Central Asia growth, unemployment increases Region ** were particularly low in Eastern Latin America Europe and Central Asia (Figure 10). Middle East This suggests that concerns that European institutions exacerbated Low middle income (omitted) unemployment increases during the Income ** crisis were exaggerated, at least for Middle middle income the middle income European Upper middle income *** countries in our sample. Activity rates and unemployment rates Structure Small manufacturing sector… increased in higher-income countries, as laid-off workers in these countries Large manufacturing sector were more able to afford to search for Less dependent (omitted) work, in part perhaps because Exports spouses or younger members of the More dependent household rejoined the labor force. Finally, countries with higher Exchange rate Fixed or Peg (omitted) redundancy costs experienced Managed float significantly larger increases in unemployment. During the crisis, Full float unemployment rates increased most Low difficulty of redundancy for youth, who may have faced particular difficulty obtaining Regulation High difficulty of redundancy employment in countries with higher redundancy costs. Low redundancy cost High redundancy cost *** -1.5 -1.0 -0.5 0.5 0.0 1.0 1.5 2.0 Change in activity rate trend Change in unemployment rate trend Note: See notes from figure 10. All data are from household surveys. Unemployment regressions contain 36 or 37 observations, while activity regressions contain 29 observations. Figure 10 Figure 10 11 How accurate were projections of employment changes? Employment became more sensitive to changes in GDP during the crisis. We estimated the relationship between employment growth and GDP growth before and after the crisis, on the 37 countries that reported data for each of these variables. Our preferred estimates control for time-invariant characteristics of the country. They indicate that the employment elasticity rose during the crisis, from 0.16 before the crisis to 0.25 after the crisis (Table A4).23 This result is of importance since forecasts of employment changes are often based on methods that combine estimates of employment elasticities with GDP projections. The ILO, for example, predicts present and future employment based on regressions of employment rates on GDP growth, controlling for time-invariant country characteristics (ILO, 2010, p.79). If the relationship between GDP growth and employment growth changes during the crisis, extrapolating from the past may lead to inaccurate estimates of employment loss. Thus predictions based on pre-crisis elasticities significantly underestimated employment loss during the crisis. Projections of employment growth derived from pre-crisis elasticities predicted employment growth of 0.47 percent. However, in the countries analyzed, employment actually declined on average, although slightly, by -0.08 percent. Therefore, on average the employment estimates overstated employment growth by 0.55 percentage points. This is a significant overestimate, given that employment growth in total declined about 2 percentage points. Table 1: Pre-crisis employment elasticities understated employment losses. Pre-crisis Post-crisis Elasticity 0.16 0.25 Average projected employment 0.47 growth (based on post-crisis GDP) Average actual employment growth 1.88 - 0.08 Average overestimate of 0.55 employment growth Average magnitude of discrepancy 1.9 Forecasting employment changes on the basis of GDP alone typically produces inaccurate estimates for individual countries. The extent of the discrepancy between actual and predicted employment ranged from -10.5 to 6.5 percentage points. On average across the countries, the prediction was off by a magnitude of 1.9 percentage points.24 From evidence to policy responses Evidence on labor market adjustments is an important input in the design of policy responses. While economic growth has bottomed out and appears to be recovering in the advanced economies, labor 23 These estimates come from regressions, based on quarterly data, of annual percent change in employment growth on annual percent change in employment, controlling for country fixed effects. Separate regressions were run for the 8 quarters before the crisis and for the four during the crisis. Results are shown in appendix table A4. The increase in the employment elasticity from 0.16 to 0.25 is not statistically significant, but leads to economically meaningful discrepancies between predicted and actual employment. 24 In other words, the average of the absolute value of the discrepancy across countries and quarters is 1.9 12 markets recover more slowly and can often take years to bounce back from sharp downturns (Rogoff and Reinhardt, 2009). In the meantime, the nature of the labor market adjustment determines winners and losers. For example, employment declines tend to concentrate losses among an unlucky few, while earnings declines spread losses more evenly amongst the workforce. The appropriate mix of policies to respond to the crisis depends partly on the nature of the labor market adjustment. Income replacement programs such as unemployment insurance or public works are most effective in mitigating the impact of job loss. In contrast, when the labor market adjustment occurs through earnings, then income maintenance program—such as cash transfers or income tax credits— become particularly important. Targeting the working poor through income support programs is therefore a priority in these circumstances. 25 Most policies adopted in response to the crisis concentrated on employment generation.26 During this crisis, support to SMEs was implemented by a larger share of developing countries than any other type of policy.27 The follows a pattern seen in past crises, as employment generation programs were implemented in response to past crises in Argentina, Mexico, Korea, and and Thailand. Meanwhile, wage subsidies were implemented in response to past crises in Costa Rica, Argentina, and Malaysia. Public works programs have also been implemented in response to crises, as in Indonesia in 1998. None of these programs, however, directly benefited the informal sector workers, and relatively few provide income support to workers who maintained their jobs.28 The emerging evidence presented in this paper suggests that effective policy packages should support earnings and household income. Responses taken in European OECD countries -- such as partial unemployment insurance, expanding cash transfers to poor workers, and temporary wage subsidies–may be priority interventions in those countries where hours and earnings adjustments dominated. 25 Surprisingly little is known about how well passive and active labor market program work during severe downturns, despite the large literature on the impact of active and passive labor market programs on labor market outcomes (Betcherman et al 2004; Fretwell et al 1999; Auer et al 2005), 26 For more detail see Paci et al 2011 27 This is according to a survey of policy responses to the crisis in 52 developing countries conducted by the World Bank and the International Labor Organization. See Robalino, Newhouse, and Rother (forthcoming) for details. 28 Argentina and Mexico concentrate on employment generation while Brazil had a large unemployment insurance scheme (Marquez 2000). Costa Rica and Argentina also had wage subsidy programs. Argentina’s Jefes program promoted employment in Argentina’s private sector, to recover lost jobs in companies with sufficient capacity (Marshall 2004). In Mexico, though there were not many legal reforms of labor market policies, changes took place largely through collective bargaining. Social security reform was implemented in 1995, but did not involve changes in employer level contributions (Marshall 2004). In Korea, job creation and a Wage Claim Guarantee was the thrust of the policy response to the East Asian Crisis (Phang and Kim, 2001). They also had a job training program to encourage reemployment (Hur 2001). An Employment Insurance Scheme that was active from 1995 was extended to cover all enterprises in 2008 (Phang and Kim, 2001). Even in Thailand the emphasis of labor market policies in response to the crisis was direct job creation rather than self-employment. Informal workers were not protected, but likely to benefit from ―income support activities‖ such as tax reductions on VAT, price supports for rice, etc. (Betcherman and Islam 2001). In Indonesia, public works were mainly to provide emergency income and create social capital, with ultimate goal to create a total of 226 million person-days of employment. While Malaysia already had a public works program in place, they initiated wage subsidies in August 1998. In response to the 1997 crisis, The Philippines initiated an Emergency Loans program, while providing wage subsidies and training (Betcherman and Islam 2001). 13 Sound diagnostics are also important to understand the causes and implications of different adjustment patterns. Labor markets can adjust to downturns in different ways, with different implications for workers. The nature of these adjustments is at least partially determined by countries’ labor market policies and institutions, and has important implications for the welfare of the poor. Future downturns are bound to occur, and additional research based on sound labor market diagnostics can help unravel the link between policies and institutions, labor market adjustments, and outcomes for the poor. 14 APPENDIX: Table A1: Average Changes in Labor Market Indicators, Before and After the Crisis Indicator Number of Pre-crisis Average Difference countries percent change, post-crisis GDP growth 28 7.3 -2.4 -9.7 Wage bill growth 28 9.4 1.1 -8.3 Employment growth 28 1.8 -0.2 -2.0 Earnings growth 28 7.3 1.3 -6.0 Hours worked growth 14 1.7 -5.2 -6.9 Real wage growth 14 6.5 7.2 0.7 Nominal wage growth 14 12.9 12.2 -0.7 CPI growth 14 7.0 6.0 -1.0 Unemployment rate 28 9.0 9.7 0.7 Participation rate 24 53.0 53.3 0.3 Notes: Pre-crisis period is an average of year on year changes over eight quarters from Q3 2006 to Q32008. Post-crisis period is an average between Q4 2008 and Q3 2009. 15 A.2 Table A2: Country summary Country Data Source Incom Severity Change in growth rate e Group (percentage points) group Employment Earnings GDP Wage Bill Employment Earnings Hours East Asia -5.64 -5.38 -0.14 -5.13 0.22 Urban China Firm Firm LMIC Mild -5.00 0.88 -1.52 2.55 Indonesia Household Household LMIC Mild -1.81 -3.98 -1.53 -2.39 0.22 Malaysia Household N/A HMIC Medium -8.92 0.21 Philippines Household Legislated LMIC Mild -4.68 -1.59 0.62 -2.26 Sri Lanka Household Legislated LMIC Mild -3.76 -19.52 1.38 -20.93 Thailand Household Household MIC Medium -9.68 -2.67 0.03 -2.65 Europe/Central -11.84 -11.30 -2.70 -8.32 -5.66 Asia Albania Household N/A LMIC Mild -0.16 0.89 Armenia Household Firm LMIC Severe -24.31 -13.77 -0.84 -9.39 Belarus Household Firm MIC Medium -7.48 -5.07 -2.12 -3.15 Bosnia Firm Firm MIC N/A -17.26 /Herzegovina Bulgaria Firm Firm MIC Medium -9.07 -6.29 -5.66 -0.04 -0.01 Georgia Firm N/A HMIC Severe -13.31 -1.19 Kazakhstan Household Firm MIC Medium -7.87 -1.09 -2.16 1.04 Kyrgyz Republic Firm Firm LMIC Medium -7.82 -0.44 Latvia Household Firm HMIC Severe -21.92 -28.68 -6.20 -22.47 -0.23 Lithuania Household Firm HMIC Severe -20.08 -22.30 -6.19 -16.35 Macedonia Household Household MIC Medium -6.35 12.11 0.10 11.57 Moldova Household Firm LMIC Medium -7.78 -4.25 -3.28 0.08 Montenegro Household N/A HMIC N/A 1.31 Poland Household Firm HMIC Mild -4.75 -8.39 -2.83 -5.31 -0.31 16 Romania Household Firm MIC Severe -12.20 -8.69 -0.95 -9.50 -0.28 Russia Household Firm MIC Severe -14.96 -20.11 -3.86 -15.87 Serbia Household Firm LMIC Medium -7.62 -18.67 -3.42 -15.71 -18.84 Tajikistan Household Household LMIC N/A 1.03 Turkey Household Household HMIC Severe -12.42 -9.60 2.69 -12.49 -14.29 Ukraine Firm Firm LMIC Severe -23.10 -23.45 -3.68 -18.90 Latin America -6.82 -5.35 -1.89 -2.47 -5.08 Urban Argentina Household Firm HMIC Medium -5.72 1.42 -0.91 2.39 -11.59 Urban Brazil Household Firm MIC Medium -7.14 -1.60 -1.30 -0.25 -6.22 Chile Firm Firm HMIC Medium -6.46 -0.76 -2.14 1.45 Urban Colombia Household Firm MIC Medium -7.14 -5.17 -1.55 -3.46 -1.85 Urban Ecuador Household Household LMIC Mild -4.05 -8.34 -0.72 -1.32 Jamaica Household N/A MIC Mild -3.17 -3.75 Mexico Household Firm HMIC Medium -9.49 -18.13 -2.73 -15.50 -9.72 Peru Firm Firm MIC Medium -8.17 -2.17 -2.14 0.12 -3.39 Trinidad and Household N/A HMIC N/A 0.03 Tobago Uruguay Household Household HMIC Medium -5.27 2.36 2.29 Venezuela Household Firm HMIC Medium -7.72 -8.04 -1.27 -6.61 Dominican Firm Industry MIC Medium -9.10 -4.38 Republic Paraguay N/A Firm LMIC Medium -8.40 -3.85 Middle -2.80 -5.71 -2.25 0.25 1.89 East/Africa Egypt Household N/A LMIC Mild -2.58 -3.61 Jordan Household N/A HMIC Medium -5.76 Mauritius Household N/A MIC Mild -3.31 -3.69 South Africa Household Household MIC Medium -5.88 -5.71 -4.10 -1.53 West Bank/Gaza Household Household HMIC Mild -0.40 2.04 1.89 Strip 17 A.3 Table A3: Data discrepencies. Indicator Coverage/Source Number of countries Whole country 27 Employment Only urban 6 Unclear/Largely urban 4 Only formal 3 All persons 28 Employment Unknown 7 Total 37 LFS/HHS 29 Employment Establishment survey 8 Total 37 Industry/Non agriculture 9 Earnings All sectors 19 Unknown 3 Total 31 LFS/HHS 7 Establishment survey 22 Earnings Legislated 2 Total 31 Note: Excludes Bosnia and Herzogovina, Montenegro, and Tajikistan due to lack of GDP data. Table A4: Regression results from elasticity regressions FE POLS RE Pre Crisis Crisis Pre Crisis Crisis Pre Crisis Crisis VARIABLE Employmen Employmen Employmen Employmen Employmen Employmen S t Growth t Growth t Growth t Growth t Growth t Growth GDP growth 0.160* 0.252*** 0.0230 0.199*** 0.0897 0.225*** Constant 0.726 0.282*** 1.719*** 0.205 1.190** 0.200 Observations 244 144 244 144 244 144 R-squared 0.035 0.276 0.001 0.247 Number of 37 37 37 37 countries Robust standard errors *** p<0.01, ** p<0.05, * p<0.1 REFERENCES Auer, P., U. Efendioglu, J. Leschke. 2005. Active labour market policies around the world. International Labour Organization. Agenor, P. 2002. ―Why Crises are Bad for the Poor.‖ Development Outreach. World Bank Institute. Betcherman, G., K. Olivas, and A. Dar. 2004. ―Impacts of Active Labor Market Programs: New Evidence from Evaluations with Particular Attention to Developing and Transition Countries.‖ World Bank Social Protection Discussion Paper 0402. Betcherman, Gordon and Islam, Rizwanul (2001) ―Summary of East Asian Labor Markets and the Economic Crisis: Impacts, Responses, and Lessons‖ World Bank. Betcherman, Gordan, Karina Olivas, and Amit Dar, 2004, ―Impacts of Active Labor Market Programs: New Evidence from Evaluations with Particular Attention to Transition Countries‖, Social Protection paper 0402 Blanchard, Faruqee and Das (2010), The Impact of the Crisis on Emerging Market Countries, IMF working paper Bowlus, A. 1993. ―Job match quality over the business cycle.‖ Panel Data and Labour Market Dynamics: 21-41. Bresciani, F., D. Gilligan, G. Feder, T. Onchan, and H. Jacoby. 2002. ―Weathering the Strom: The Impact of the East Asian Crisis on Farm Households in Indonesia and Thailand.‖ World Bank Research Observer 17:1-20. Calderon, Cesar and Didier, Tatiana (2009) ―SEVERITY OF THE CRISIS AND ITS TRANSMISSION CHANNELS‖ World Bank Casacuberta, C. and N. Gandelman. 2009. ―Productivity, Exit and Crisis in Uruguayan Manufacturing and Services Sectors.‖ Mimeo. Cho, Yoonyoung and David Newhouse, 2011, ―How did the Great Recession Affect Different Types of Workers? Evidence from 17 Countries‖, Policy Research Working Paper no. 5636. Di Giovanni, J., and A. Levchenko, 2009. ―Trade Openness and Volatility.‖ The Review of Economics and Statistics 91(3), 558-585. Diwan, Ishac. 2001. ―Debt as Sweat: Labor, Financial Crisis, and the Globalization of Capital.‖ World Bank. Washington, DC. Available on line at http://info.worldbank.org/etools/docs/voddocs/150/332/ diwan.pdf. Processed. ————. 2002. ―The Labor Share during Financial Crisis: New Results.‖World Bank.Washington, DC. Processed. Dollar, David and Mary Hallward-Driemeier. 2000. ―Crisis, Adjustment, and Reform in Thailand's Industrial Firms.‖ World Bank Research Observer 15(1): 1-22. Ezemenari, K., Chaudhury, N. and Owens, J. (2003) Gender and risk in the design of social protection interventions, Social Protection Discussion Paper Series No 0231, Washington, DC: World Bank (2004). Fallon, P. and R. Lucas. 2002. ―The Impact of Financial Crises on Labour Markets, Household Incomes, and Poverty: A Review of Evidence.‖ The World Bank Research Observer 17(1):21-43. Fretwell, David, Jacob Benus, and Christopher O’Leary. 1999. ―Evaluating the Impact of Active Labor Programs: Results of Cross Country Studies in Europe and Central Asia.‖ Social Protection Discussion Paper 9915. Gutierrez, C. Orecchia, C, Paci, P. and Serneels, P (2007). ―Does Employment Generation Really Matter for Poverty Reduction‖ World Bank Policy Research Working Paper 4432. Habib, Bilal, Ambar Narayan, Sergio Olivieri, and Carolina Sanchez-Paramo, 2010, ―Assessing Ex-ante the Poverty and Distributional Impact of the Global Crisis in a Developing Country: A Micro-Simulation Approach with Application to Bangladesh.‖ World Bank Policy Research Working Paper 5238 Hamermesh, Daniel (1986). ―The Demand for Labor in the Long Run,‖ Handbook of Labor Economics, ed. O. Ashenfelter and R. Layard, I: 429–71. Amsterdam, Netherlands: Elsevier Science Publisher. Hamermesh, Daniel (1993). Labor Demand. Princeton, NJ: Princeton University Press. Hill, Hal and Takashi Shiraishi. 2007. ―Indonesia After the Asian Crisis.‖ Asian Economic Policy Review 2(1):123-141. Japan Center for Economic Research. Hiren Sarkar and Ray Kumar, 2002, ―Social Impact of the Asian Economic Crisis‖ in Protecting Marginalized Groups During Economic Downturns: Lessons from the Asian Experience, United Nations. Hur, Jai Joon (2001) ―Economic Crisis, Income Support, and Employment Generating Programs: Korea’s Experience‖ UN ESCAP seminar: “Evaluation of Income/Employment Generating Programs to Alleviate Socio-Economic Impacts of the Economic Crisis� International Labor Organization. 2009. ―Protecting people, promoting jobs.‖ G20 Leaders’ Summit ILO Report. International Labor Organization, 2010. Global Employment Trends, January 2010. IMF World Economic Outlook (Washington, DC, IMF, October 2009). Manning, C. 2000. ―Labour market adjustment to Indonesia’s Economic Crisis: Context, Trends and Implications.‖ Bulletin of Indonesian Economic Studies 36(1):105-136. Marquez, Gustavo (2000) ―Labor Markets and Income Support: What Did we Learn from the Crises?‖ IADB Working Paper #425 Marshall, Adriana (2004) ―Labour market policies and regulations in Argentina, Brazil and Mexico: Programmes and impacts‖ ILO Project on Employment challenges in Argentina, Brazil and Mexico; Consejo Nacional de Investigaciones Cientificas y Tecnicas, Buenos Aires. McKenzie, David J. 2004. ―Aggregate Shocks and Urban Labor Market Responses: Evidence from Argentina's Financial Crisis.‖ Economic Development and Cultural Change 52(4):719-58. McKenzie, David. 2002. ―How do Households Cope with Aggregate Shocks? Evidence from the Mexican Peso Crisis.‖ World Development 31(7):1179-1199. Mead, Donald and Liedholm, Carl. (1998) "The Dynamics of Micro and Small Enterprises in Developing Countries" World Development 26(1): 61-74. Narjoko, Dionisius and Hal Hill. 2007. ―Winners and Losers during a Deep Economic Crisis: Firmlevel Evidence from Indonesian Manufacturing .‖ Asian Economic Journal 21(4):343-368. Phang, Hanam and Kim, Dong-Heon (2001) ―Policy Options for Income Support and Active Labor Market Programs: A Synthesis of the Korean Experience‖ World Bank Manila Conference Paper, March 1-2, 2001. WB Conference: “Labor Market Policies: Its Implications to East and Southeast Asia� Reinhart, C. and K. Rogoff. 2009. ―The Aftermath of Financial Crises.‖ American Economic Review 99(2):466-72. Robalino, David, David Newhouse, and Friederike Rother, forthcoming, ―Labor and Social Protection Policies During the Crisis and Recovery‖ Sarkar, Hiren and Raj Kumar, 2002, ―Protecting Marginalized Groups During Economic Downturns: Lessons from the Asian Experience‖, New York: United Nations Sasin, Marcin (2008). Moroccan migration: the profile and the impact on households Background paper for the Regional Study on Migration in the MENA region - World Bank. Sinha, Anushree. (2009). ―Impact of Global Slowdown on Employment and Income in India: A CGE Model Analysis.‖ Mimeo James P. Smith, Duncan Thomas, Elizabeth Frankenberg, and Kathleen Beegle (2002) ―Wages, Employment, and Economic Shocks: Evidence from Indonesia‖, Journal of Population Economics, Vol. 15, pp 161-193 World Bank. 2008. ―Double-Jeopardy: Responding to High Food and Fuel Prices.‖ G8 Hokkaido- Toyako Summit Paper World Bank, 2009, ―Averting a human crisis During a Global Downturn: Policy Options from the World Bank’s Human Development Network‖