Trade liberalization and unemployment: Theory and evidence from India

Abstract

A widely held view among the public is that trade liberalization increases unemployment. Using state and industry-level unemployment and trade protection data from India, we find no evidence of any unemployment increasing effect of trade reforms. In fact, our state-level analysis reveals that urban unemployment declines with trade liberalization in states with flexible labor markets and larger employment shares in net exporter industries. Moreover, our industry-level analysis indicates that workers in industries experiencing greater reductions in trade protection were less likely to become unemployed, especially in net export industries. Our results can be explained within a theoretical framework incorporating trade and search-generated unemployment and some institutional features of the Indian economy.

Introduction

There now exists a small but growing literature on the relationship between trade and unemployment.¹ Much of this literature is theoretical, with a few exceptions.² A recent contribution based on cross-country analysis and which incorporates developing country experience is Dutt et al. (2009). Using cross-country data on trade policy, unemployment, and various controls, and controlling for endogeneity and measurement-error problems, they find that unemployment and trade openness are negatively related. Using panel data, they find an unemployment-increasing short-run impact of trade liberalization, followed by an unemployment-reducing effect leading to the new steady state. While that paper finds interesting empirical regularities that can be explained using plausible models of trade and search unemployment, the standard criticisms of cross-country regressions apply to that study as well. Countries differ from each other in very important ways that cannot always be controlled for by variables we use in such regressions.

The key empirical studies that analyze the links between trade policy and unemployment for individual developing countries are Attanasio et al., 2004, Menezes-Filho and Muendler, 2007, Porto, 2008. These studies focus on the experiences of Columbia, Brazil, and Argentina, respectively. Using labor force survey data spanning the period before and after Colombia's trade liberalization in the early 1990s, and relying on information on the one digit industry in which unemployed individuals either worked previously or were looking for employment, Attanasio et al. (2004) find the following: while the overall probability of unemployment increased after liberalization, this increase was driven by nontraded sectors such as wholesale and retail trade rather than traded sectors such as manufacturing. On the other hand, based on a unique and very rich linked employer–employee data set which allows formally employed workers to be tracked over time and across production sectors, Menezes-Filho and Muendler (2007) find that Brazil's trade liberalization in the 1990s led to the displacement of formally employed workers from protected industries and that ‘comparative advantage’ industries or exporters did not absorb trade-displaced workers in full. Their complementary analysis of employment survey data reveals that many of these formally employed workers transitioned to informal work or self-employment. For others, trade liberalization was associated with transitions to unemployment.

Porto (2008) examines the links between trade liberalization and unemployment (and wages) in the case of Argentina. However, the focus of his study is a little different from the other two studies. More specifically, Porto investigates how world agricultural trade liberalization can be expected to affect unemployment (and wages) in Argentina using an empirical model of trade, unemployment, and labor supply. His estimates indicate that an increase in the price of Argentine agro-manufactured exports can be expected to lead to both a lower unemployment rate and an increase in labor market participation. Wages also increase given an increase in export prices.

In this paper, we contribute to the empirical trade and unemployment literature using labor force survey data from India, a developing country that has, in the last couple of decades, experienced major trade reforms and where a significant proportion of the population lives below the poverty line. This makes such a study useful for policy analysis.

We carry out our empirical analysis of trade and unemployment at two levels. In addition to an industry level analysis along the lines of Attanasio et al. (2004) (and close in spirit to Menezes-Filho and Muendler, 2007), we also carry out analysis at the state level. Importantly, constitutional arrangements which give India's states considerable regulatory power over economic matters and the large size of these states (larger than the vast majority of countries), with unique ethno-linguistic characteristics, make such an analysis meaningful. In particular, interstate variations in labor regulations (see Besley and Burgess, 2004) and low mobility across Indian states (see Dyson et al., 2004, Topalova, 2010) suggest that treating each state as an independent labor market is a reasonable approximation. Additionally, we adopt broadly the strategy of Topalova, 2007, Hasan et al., 2007b and exploit variations in industrial composition across Indian states (districts in the case of Topalova) around the time of the major trade liberalization of the early 1990s, and the variation in the degree of liberalization across industries over time to construct state-specific measures of protection. This allows us to determine whether states more exposed to reductions in protection experienced increases or decreases in unemployment rates.

We then complement our cross-state analysis of the relationship between unemployment and liberalization with a detailed analysis based purely on changes in industry-level variations in protection over time. More specifically, we examine whether workers in industries experiencing greater reductions in trade protection were more or less likely to become unemployed (relative to the average worker) using the two-stage approach used by Attanasio et al. (2004).

In so far as our state-level analysis is concerned, we find that overall (rural plus urban) unemployment on average does not have any relationship with average protection (weighted average with 1993 sectoral employment as weights) over time and across states. However, there are some conditional relationships between the two variables in certain types of states. In states with more flexible labor markets, there is evidence that on average overall and urban unemployment are positively related to protection. We also find that reductions in protection reduce unemployment in the urban sectors of states with large employment shares in net exporter industries.

Turning to our analysis based purely on industry-level protection, we find hardly any evidence that workers in industries that experienced larger reductions in protection were more likely to be unemployed. In fact, there is some evidence that such workers were less likely to become unemployed, a result that is stronger in states with flexible labor regulations and net exporter industries. There is also some weak evidence that the immediate short-run effect of a tariff reduction may be an increase in unemployment prior to reduction to a lower steady-state unemployment rate.

We show how our empirical results are consistent with the impact of trade liberalization in a two sector model, with labor being the only factor of production and where unemployment arises due to search frictions. We discuss two extreme cases: (a) perfect labor mobility (the Ricardian case), where comparative advantage is exclusively productivity-driven and (b) no intersectoral labor mobility (where labor becomes sector-specific), where comparative advantage, while still dependent on productivity, is also driven by relative sectoral labor force size. Our empirical results fall in between the two extremes, depending on the flexibility of labor markets.³