Trade and Labor Market Adjustment: Recent Evidence from Brazil

Trade and Labor Market Adjustment: Recent Evidence from Brazil Rafael Dix-Carneiro Duke University, NBER and BREAD January 25, 2018 This chapter reviews recent evidence on how the Brazilian labor market adjusted to the trade liberalization episode of the early 1990s. The results I discuss are important for at least three reasons in the context of this session on Trade and Labor Market Adjustment. First, the Brazilian trade liberalization is a valuable episode from which to learn about trade and labor market adjustment. This episode consisted of large unilateral import tariff reductions between 1990 and 1995. Average tariffs fell from 31% to 13% and there was ample variation across sectors. Importantly, trade liberalization in Brazil can be approximately viewed as a once-and-for-all event. Import tariffs were gradually reduced between 1990 and 1995, but import tariffs remained relatively constant thereafter. This allows us to trace the evolution of liberalization s effects over time. Second, Brazil has excellent data sources on the labor market, covering a period including the trade liberalization episode. Among these data sources are the Relação Anual de Informações Sociais (RAIS), an administrative dataset starting in 1986 which provides high quality information on all formal-sector workers and firms in the country. In particular, it is possible to track workers over time across firms, sectors and regions. The research discussed here also employs multiple rounds of the Decennial Demographic Census covering the period ranging from 1970 to 2010 to obtain information about the Brazilian informal sector or about the labor market as a whole. 1 Together, these data make possible a comprehensive I thank Brian Kovak, Rodrigo Soares and Gabriel Ulyssea for collaboration on the topics discussed here. Special thanks to Brian Kovak for comments on this chapter. 1 The informal sector accounts for approximately half of overall employment in Brazil. A worker is considered informal if she is informally employed by a firm (off the books and invisible to the government) or if she is self-employed. In each case, the worker does not receive the benefits or regulatory protections 1

and detailed analysis on how the Brazilian labor markets adjusted to trade. Third, it complements the evidence discussed by my co-panelists Gordon Hanson and Lori Kletzer, which focuses on a developed country such as the United States. Given Brazil s status as an important middle-income country and its very different labor market structure compared to the United States, it is interesting to contrast these two countries experiences adjusting to globalization. In addition, although the rise of China was probably the single most important development in the global economy in the past 30 years, it is usually agreed by economists that the emergence of China is a done deal and that it is unlikely that we will witness another comparable episode in our lifetimes. In contrast, even though Brazil went through a major trade liberalization episode in the 1990s, it still remains a relatively protected economy, with import tariffs across sectors averaging 10.4%. 2 Therefore, understanding how Brazilian labor markets adjusted to trade liberalization is useful to inform policy makers planning another wave of trade liberalization. Details about trade liberalization in Brazil can be found in Dix-Carneiro and Kovak (2017b), but Figure 1 illustrates that: (1) trade liberalization led to large import tariff reductions; and (2) there was ample variation on how sectors were affected. For instance, sectors such as Agriculture and Mining were virtually unaffected by changes in trade policy, whereas sectors such as Apparel, Rubber, Pharmaceuticals, and Autos faced large declines in protection. A recent but prominent literature has exploited the fact that sector-specific shocks, such as those illustrated in Figure 1, can have substantially different effects on labor market outcomes across regions within a country. 3 To understand why, suppose that workers face large barriers of mobility across regions within a country. In that case, we would expect labor demand in an Apparel town to fall relative to labor demand in an Agriculture town, as tariff cuts were much steeper in Apparel compared to Agriculture. This intuitive idea was formalized and rationalized by Kovak (2013) who showed how sector-specific tariff changes can be translated into region-specific labor demand shocks, depending on differences in industry composition across locations. In short, Kovak (2013) shows that the regional labor demand shock induced by trade liberalization is given by a weighted average of sectorspecific tariff declines, where weights are given by sector-specific employment shares in that present in the formal labor market. 2 Author s calculations using 2010 UNCTAD TRAINS data and similar level of aggregation as Figure 1. The 25 th percentile of the distribution of 2010 import tariffs is 5.3% and the 75 th percentile is 13.9%, with sectors being protected with over 30% tariffs (clothing and footwear). 3 For example, see Topalova (2010), Autor et al. (2013), Kovak (2013) and Hakobyan and McLaren (2016). 2

Figure 1: Tariff Changes Change in ln(1+tariff), 1990-95 0.00-0.05-0.10-0.15-0.20-0.25 Agriculture Metals Apparel Food Processing Wood, Furniture, Peat Textiles Nonmetallic Mineral Manuf Paper, Publishing, Printing Mineral Mining Footwear, Leather Chemicals Auto, Transport, Vehicles Electric, Electronic Equip. Machinery, Equipment Plastics Other Manuf. Pharma., Perfumes, Detergents Petroleum Refining Rubber Petroleum, Gas, Coal Industries sorted based on 1991 national employment (largest on the left, and smallest on the right). region. These shocks are called Regional Tariff Reductions. Figure 2 shows how labor demand was affected across local labor markets in Brazil in response to trade liberalization by plotting the distribution of these regional tariff reductions. Darker regions are mostly specialized in sectors facing small tariff declines, such as Agriculture. In contrast, lighter regions are mostly specialized in sectors facing larger tariff declines, such as Apparel and Autos. It is important to notice that there is ample variation on how regions were affected by trade liberalization. Dix-Carneiro and Kovak (2017b) investigate how region-specific labor market outcomes have responded to these local shocks induced by the tariff cuts, tracking the evolution of these effects over time. Figure 3 shows that regions facing larger (negative) shocks induced by liberalization experienced declines in formal-sector employment relative to regions facing smaller shocks. 4 These effects gradually increased following the beginning of liberalization 4 The effects we review in this chapter can only reveal relative effects of Brazil s trade liberalization on local labor markets. This is a well-known limitation of reduced-form estimates in the presence of important general equilibrium effects, which is a common feature of all the trade and local labor markets literature. 3

Figure 2: Local Labor Demand Shocks Induced by Liberalization Regional Tariff Reductions Manaus Belém Fortaleza Recife Salvador 8% to 15% Brasília 4% to 8% 3% to 4% 1% to 3% -1% to 1% Curitiba São Paulo Belo Horizonte Porto Alegre Lighter regions faced larger tariff reductions, whereas darker regions faced smaller tariff reductions. Labor demand in lighter regions was negatively affected relative to darker regions. and only stabilized 15 years later. In particular, these results show that formal-sector employment adjustment in response to trade liberalization was large but very slow. Before I discuss how regional wages responded to these local labor demand shocks induced by liberalization, it is instructive to comment on how wages were expected to evolve. A model that is often employed by economists assumes that workers are not very mobile across regions in response to local economic shocks in the short run, but are increasingly mobile as time unfolds. This view is consistent with the dynamic effects illustrated in Figure 3, which shows formal-sector employment gradually responding to the local shocks induced by liberalization. Therefore, following a decline in labor demand in regions facing larger tariff cuts (relative to 4

Figure 3: Effects of Liberalization on Formal-Sector Regional Employment 1.5 Pre- liberaliza6on (chg. from 1986) Liberaliza6on Post- liberaliza6on (chg. from 1991) 0.5 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010-0.5-1.5-2.5-3.5-4.5-5.5-6.5 The graph shows the effect of trade-induced local labor demand shocks (from 1990 to 1995) on the change in regional (log) formal employment from 1991 to the year listed on the x-axis. Negative estimates imply larger employment declines in regions facing larger tariff reductions (relative to less affected regions). Dashed lines represent 95 percent confidence intervals. Pre-liberalization trends are measured relative to 1986. See Dix-Carneiro and Kovak (2017b) for details. the national average), we would expect wages to fall at impact as workers are stuck in these regions in the short run. However, as workers gradually move away from these harder-hit locations toward other locations, equilibrium wages would move up the local labor demand curve up to the point where wages are re-equalized across regions. This logic suggests that wages in harder-hit locations would decline in the short run relative to the national average, but would then gradually recover, so that long run effects on wages across regions should be negligible. Surprisingly, Figure 4 shows a very different adjustment pattern. Dix-Carneiro and Kovak (2017b) document that wages in harder-hit locations steadily decline for years and never recover. The long-run effect is three times as large as the short run effect. The effects of liberalization on wages are also very slow and persistent. These effects are in sharp contrast to the economic logic described in the previous paragraph. What explains them? 5

1.0 Figure 4: Effects of Liberalization on Formal-Sector Regional Earnings Pre- liberaliza6on (chg. from 1986) Liberaliza6on Post- liberaliza6on (chg. from 1991) 0.5 0.0 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010-0.5-1.0-1.5-2.0 The graph shows the effect of trade-induced local labor demand shocks (from 1990 to 1995) on the change in regional (log) formal earnings from 1991 to the year listed on the x-axis. Negative estimates imply larger employment declines in regions facing larger tariff reductions (relative to less affected regions). Dashed lines represent 95 percent confidence intervals. Pre-liberalization trends are measured relative to 1986. See Dix-Carneiro and Kovak (2017b) for details. Two ingredients are essential to explain the effects documented in Figures 3 and 4: slow and incomplete inter-regional labor mobility, and dynamics in labor demand. In particular, these dynamics in local labor demand are driven by: (a) Slow adjustment of capital across regions (because of slow depreciation and new investment being gradually directed toward less affected regions); and (b) slow changes in local productivity driven by agglomeration economies. Using Demographic Census data, we show that migration does not systematically respond to the trade-induced shocks, corroborating the hypothesis of imperfect inter-regional labor mobility in response to shocks. On the other hand, Figure 5 illustrates that firm exit increased very gradually following liberalization, and that investment (measured as firm entry) responded almost immediately and permanently. As suggested by Dix-Carneiro (2014), 6

the slow reallocation of capital led to a steady amplification of the initial local labor demand shock, making workers in harder-hit regions even less productive over time compared to those in more favorably affected regions. Dix-Carneiro and Kovak (2017b) also suggest that agglomeration economies amplify the labor market effects of trade liberalization: as firms in harder-hit regions leave the market, the productivity of remaining local firms gradually declines, further reducing local wage and employment growth. Using a simple model of local labor markets, they show that capital reallocation and agglomeration economies together can explain the quantitative scale of the wage and employment effects they document. Figure 5: Effects of Liberalization on Establishment Entry and Exit 5 Pre- liberaliza5on (chg. from 1986) Liberaliza5on Post- liberaliza5on (chg. from 1991) 4 Exit 3 2 1 Entry Pretrend 0 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010-1 - 2 Exit Pretrend Entry - 3 The graph shows the effect of trade-induced local labor demand shocks (from 1990 to 1995) on (log) cumulative regional formal establishment entry or exit from 1991 to the year listed on the x-axis. Positive exit estimates and negative entry estimates imply larger rates of exit and smaller rates of entry in regions facing larger tariff reductions (relative to less affected regions). Dashed lines represent 95 percent confidence intervals. Pre-liberalization trends are measured relative to 1986. See Dix-Carneiro and Kovak (2017b) for details. The evidence reviewed so far focused on how aggregate formal-sector regional-level labor market outcomes responded to liberalization. To complement this evidence, Dix-Carneiro and Kovak (2017a) analyzed how individual labor market trajectories of workers responded to 7

the trade-induced labor demand shocks and through what margins they adjusted. Following individual workers over time using RAIS, Figure 6 shows that workers initially employed in tradable sectors in harder-hit locations tend to spend less and less time employed in the formal sector relative to workers initially employed elsewhere. Perhaps surprisingly, this effect grows over time. Although Figure 3 documents that formal employment gradually declines in regions facing steeper tariff cuts relative to the national average, we would expect that individual labor market outcomes would eventually recover as workers are able to migrate away from harder-hit regions. This suggests that individual workers did not systematically migrate away from negatively affected locations a fact that is corroborated using longitudinal data. Tradable sector workers initially employed in harder-hit locations are more likely to switch to non-tradable sectors in response to liberalization. However, this response does not offset the large losses in employment in tradable sectors. Finally, Dix-Carneiro and Kovak (2017a) document that the formal employment trajectories of workers initially employed in non-tradable sectors are almost as affected as those of tradable sector workers. This shows important spillovers from tradable to non-tradable sectors locally. These spillovers across sectors raise concerns about policies providing targeted compensation for workers in industries experiencing increased import competition, such as Trade Adjustment Assistance in the United States. When regional labor markets are reasonably integrated across sectors, even workers whose industry did not directly face a trade shock experience the labour market effects of that shock, and policies with industry targeting will fail to address declining earnings and employment rates for these indirectly affected workers. Figure 6 shows that workers initially employed in harder-hit regions are less and less likely to be found working in the formal sector, but it does not tell us how exactly these workers adjust. Unfortunately, the Brazilian administrative data does not have information on workers once they leave the formal sector, so it is unclear if workers who leave the formal sector are unemployed, out of the labor force, self-employed or informally employed. To have a picture of what happens to trade-displaced workers once they leave the formal sector, Dix-Carneiro and Kovak (2017b) exploit data from the Demographic Census. The advantage of using such data is that we can identify workers who are formally employed, informally employed or not employed. The disadvantage of such data is that, in contrast with the administrative data, we cannot follow workers over time. Therefore, Dix-Carneiro and Kovak (2017a) investigate how the structure of local labor markets responded to the trade-induced local shocks. They document that, in the medium run (1991 to 2000), non-employment and informal 8

Figure 6: Effects of Liberalization on Workers average months of being formally employed per year 1.0 Liberaliza6on Post- liberaliza6on 0.0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010-1.0-2.0-3.0-4.0-5.0-6.0-7.0 The graph shows the effect of trade-induced local labor demand shocks (from 1990 to 1995) on individual cumulative average number of months formally employed per year from 1991 to the year listed on the x-axis. Negative estimates imply that workers initially in regions facing larger tariff reductions spend a smaller average share of the relevant years formally employed than workers in other regions. See Dix-Carneiro and Kovak (2017b) for details. employment increase in harder-hit locations relative to the national average. However, in the long run (1991 to 2010) non-employment does not respond to local trade shocks, but informal employment strongly increases in harder-hit regions. Together with the individual worker results, it seems that trade-displaced workers spend time unemployed or out of the labor force, but eventually find re-employment in the informal sector. Therefore, the informal sector seems to partly smooth the labor market outcomes of trade displaced workers. Without this fall-back sector, trade-displaced workers would likely have experienced even longer non-employment spells. Dix-Carneiro and Kovak (2017b) and Dix-Carneiro and Kovak (2017a) document that the Brazilian economy adjusted very slowly to trade liberalization. It is therefore important to understand what are the implications of such slow adjustment for the gains from trade. This 9

question has been addressed by Dix-Carneiro (2014) who showed that slow reallocation of workers and capital toward export oriented industries leads to substantially lower gains from trade compared to traditional models (which assume that the new equilibrium is reached instantaneously). In that paper, I estimate 11% to 26% lower gains from trade compared to a situation where reallocation occurs immediately. In Dix-Carneiro (2014), I also estimate that adjustment costs are very heterogeneous across the population. For example, older, less educated and female workers face substantially higher barriers to mobility across sectors. These workers in import-competing sectors experience substantial losses following liberalization, so governments willing to compensate the losers from trade should pay particular attention to workers with these characteristics. A recent literature has been emphasizing that globalization can have important effects beyond the labor market. 5 For instance, in the Brazilian context, Dix-Carneiro et al. (forthcoming) show that Brazilian regions exposed to larger tariff reductions experienced a temporary increase in crime following liberalization. These results highlight an additional dimension of adjustment costs to trade shocks. Given that crime generates substantial externalities, these results show that the adjustment costs triggered by trade shocks can go well beyond the individuals directly affected by them. In conclusion, recent work on the Brazilian experience following the 1990 s trade liberalization episode showed that the adjustment of the labor market was slow and that not every region benefitted in the same way. As discussed above, the welfare implications of this slow adjustment can be substantial. In addition, the pattern of adjustment was more complex than conventional theories would predict, highlighting several features absent in existing models of trade and labor markets. First, to understand labor market adjustment in response to globalization, capital and labor markets must be studied simultaneously. Second, the large response of informal employment to trade liberalization (especially in the long run) highlights an additional margin of adjustment that has been under-studied in the literature. Finally, globalization can have important effects beyond the labor market, leading to important externalities within regions. 5 For example, recent studies have estimated the effects of trade shocks on crime (Iyer and Topalova, 2014), the provision of public goods (Feler and Senses, 2016), health and mortality (McManus and Schaur, 2016; Pierce and Schott, 2016), household structure (Autor et al., 2015) and political outcomes (Dippel et al., 2017; Autor et al., 2016; Che et al., 2016). 10

References Autor, D., D. Dorn, and G. Hanson, The Labor Market and the Marriage Market: How Adverse Employment Shocks Affect Marriage, Fertility, and Children s Living Circumstances, 2015. Mimeo.,,, and K. Majlesi, Importing Political Polarization? The Electoral Consequences of Rising Trade Exposure, 2016. NBER Working Paper, No. 22637. Autor, David H., David Dorn, and Gordon H. Hanson, The China Syndrome: Local Labor Market Effects of Import Competition in the United States, American Economic Review, October 2013, 103 (6), 2121 68. Che, Y., Y. Lu, J. Pierce, P. Schott, and Z. Tao, Does Trade Liberalization with China Influence U.S. Elections?, 2016. Mimeo. Dix-Carneiro, Rafael, Trade Liberalization and Labor Market Dynamics, Econometrica, 2014, 82 (3), 825 885. and Brian K. Kovak, Margins of Labor Market Adjustment to Trade, 2017. NBER Working Paper, No. 23595. and, Trade Liberalization and Regional Dynamics, American Economic Review, October 2017, 107 (10), 2908 46., Rodrigo R. Soares, and Gabriel Ulyssea, Economic Shocks and Crime: Evidence from The Brazilian Trade Liberalization, American Economic Journal: Applied, forthcoming. Feler, L. and M. Senses, Trade Shocks and the Provision of Local Public Goods, 2016. Mimeo. Gold, R. Dippel C.and, S. Heblich, and R. Pinto, Instrumental Variables and Causal Mechanisms: Unpacking the Effect of Trade on Workers and Voters, 2017. NBER Working Paper No. 23209. Hakobyan, Shushanik and John McLaren, Looking for Local Labor Market Effects of NAFTA, The Review of Economics and Statistics, 2016, 98 (4), 728 741. 11

Iyer, Lakshmi and Petia B Topalova, Poverty and crime: evidence from rainfall and trade shocks in India, 2014. Harvard Business School BGIE Unit Working Paper 14-067. Kovak, Brian K., Regional effects of trade reform: What is the correct measure of liberalization?, American Economic Review, 2013, 103 (5), 1960 1976. McManus, C. and G. Schaur, The Effects of Import Competition on Worker Health, Journal of International Economics, 2016, 102, 160 172. Pierce, J. and P. Schott, Trade Liberalization and Mortality: Evidence from U.S. Counties, 2016. Mimeo. Topalova, Petia, Factor immobility and regional impacts of trade liberalization: Evidence on poverty from India, American Economic Journal: Applied Economics, 2010, 2 (4), 1 41. 12