Resumen de Essays in labor income inequality and risk
Tomás Rodríguez Martínez-Zaldivar
In the last 40 years, the world economy went through two striking transformations that shaped the labor markets around the world. First, technological advancement such as computers, the internet, and cheap automatized machines have changed the way production processes combine capital and labor. Second, the increase in the volume and speed that tradable goods are changing national boundaries have altered the global value chain of various industries. Both phenomena have disrupted the labor market of many countries by displacing workers of specific industries and occupations, and by increasing the labor income inequality between workers with different skills. This dissertation aims to understand how the labor market dynamically responded and what are the welfare consequences of these two transformations.
In the first chapter, “Wage Inequality and Occupation Polarization: A Dynamic Perspective”, I argue that job polarization has long lasting effects in the U.S. wage structure. In the past 40 years, the U.S. labor market has experienced a substantial decrease in the labor share of middle wage occupations. The disappearing of these jobs concurrent with the expansion of the ones with high and low wages, so called “job polarization”, has been linked to the expansion of automation technology, as it provided a substitution for routine tasks. In a seminal work, Autor, Levy, and Murmane (2003) suggested that the introduction of computers, a cheap form of automation technology, replaced occupations with higher routine task in the workplace. Hence, routine-intensive occupations, which are typically in the middle of the wage distribution, are displaced as the cost of automation technologies go down.
Concomitant to the phenomena of job polarization, other changes have happened to the U.S. wage and employment structure over this time. A well-known fact is the increase in the education premium simultaneously with the supply of college educated workers (Katz and Murphy (1992) and Katz and Autor (1999)). Less known, but similarly important, are the changes of the experience-wage profiles for different education groups and cohorts (Elsby and Shapiro (2012) and Kambourov and Manovskii (2005)). Low educated workers have experienced a flattening in their wage profiles, while college graduates observed a steepening in theirs. Although the first fact has long been connected to skill-biased technological change (SBTC), most explanations of the second fact have lied on the differences of labor supply of different cohorts (as in Jeong et al. (2015)). In this paper, I argue that changes across cohorts are crucial to understand the inequality dynamics between occupations, education and age groups. Firstly, I show that occupation polarization has different impact across cohorts. Young workers disproportionally moved to low and high-wage occupations in comparison to old workers. In addition, I decompose the polarization for different educational groups. Despite the fact that the phenomenon was widespread over all the distribution of education, I point out that low-educated workers have a higher likelihood to move to a lower wage occupation, whereas college educated workers tend to climb to a higher wage occupation. This antagonistic propensity, induced by the polarization, in the absence of other changes, increased the share of college educated workers in high returns to experience occupations, as well the share of non-college educated workers in the low returns to experience ones. Finally, I document that the returns to experience by occupation is almost entirely increasing at their wage level, meaning that low-wage occupations also tend to have low returns to experience. High-skill occupations have up to 25% higher returns to experience than low skill occupations. Moreover, returns to experience have been increasing in the top of the distribution but decreasing in the bottom, these differences can account to up to 10 p.p.
Thus, providing that there exist complementarities between the type of labor and experience supplied by alternate cohorts, the price of experience will endogenously respond to cross- cohort movements. In fact, since younger cohorts are more educated and are employed in distinct occupations than their older counterparts, the price of experience should increase as observed in the data. Furthermore, changes in returns to experience may potentially affect occupational choice today, and impact the observed cross-sectional occupational polarization.
In light of these facts, I propose a life cycle general equilibrium model with endogenous occupational and educational choice. Workers select an occupation and supply inelastically three type of tasks: abstract, routine and manual. Also, as they get old, they accumulate experience and are able to supply a differentiated type of labor, generating the usual concave life cycle profile of wages. Yet, occupational choice is sticky: only a fraction of workers can switch occupations later in life.
In my model, all the occupations are a composition of the three tasks but in different shares. When a worker chooses an occupation, she is effectively picking a bundle of tasks to supply. This is a departure from the polarization literature that sometimes treats tasks and occupations interchangeable. In the usual polarization models, workers of different ability sort themselves into supply exclusively one task. Biased technological change alters the sorting, pushing more (less) workers to supply abstract and manual (routine) tasks. In my case, routine-biased technical change (RBTC) also affects the price of the three tasks. However, instead of sorting through tasks, RBTC affects sorting through occupations, as it increases the wage of abstract and manual-intensive occupations and decreases the wage of the routine-intensive ones.
Moreover, the model captures the heterogeneity in returns to experience observed in the data. Similarly to Katz and Murphy (1992) and Jeong, Kim, and Manovskii (2015), I embed two type of experience-labor: young and old. Workers supply both types of labor, increasing the amount of old labor supplied and decreasing young labor as they age. The fundamental assumption is that the age supply schedule is task specific. For instance, if the supply schedule is steeper for abstract tasks, then abstract-intensive occupations have steeper wage profiles. This structure not only gives us substantial heterogeneity in the returns to experience by occupation, but also, allows these returns to adjust endogenously over the time, when distinct cohorts change their relative supply of tasks. Therefore, as old and young labor are not perfectly substitutes and occupation movement is costly (so older cohorts do not have the same occupational structure than the young ones), job polarization can potentially drive movements in the returns to experience.
As usual in the literature, the key mechanism is the complementarity between different types of education, tasks and experience. Extending the usual nested CES structure, the model is able to jointly explain the changes in the wage structure under the period studied. In addition to the two demand shifters (SBTC and RBTC), I allow various demographic supply shifters, such as the increase of college share, the increase of abstract and manual-intensive occupations (and decrease of routine-intensive occupations), the increase in occupation mo- bility and the increase in the cohort size (the baby boomers). A major implication of these mechanisms is that cohort effects disappear when demographics adjust, and old and young cohorts have the same education and occupation structure. To capture potential effects from the transition dynamics, the model is calibrated to replicate the 40 years’ transition path of the U.S. labor market. Theoretically, I suggest that assuming that the economy is in the steady state produces an upward bias in the value of going to college.
Quantitatively, I show that the transition matters for inequality. During the transition, because old workers do not go back to college, the college premium is exceptionally higher. Similarly, because occupation switching costs are higher for old cohorts, a demand shift biased against routine occupations, shift disproportionally more young workers. In addition, during a transition, as not every worker can switch out from the dying routine occupations, the excess of supply in these occupations decreases the relative wage even further. To account for this mechanism, I simulate the model to a future steady state, where all cohorts can choose freely their occupation and schooling. In this steady state, college premium is 10% lower than in 2010. Also, the median occupation wage is 22% lower than the 90th occupation and 23% higher than the 10th occupation, in comparison to 34% and 12% in 2010, providing evidence that once the technological transition vanishes, inequality between groups decrease substantially. However, returns to experience responds very little in the new steady state, as in 2010, part of the demographic transition is already completed. The policy recommendation is clear: occupation switching costs drag the transition, preventing workers to move to high paid occupations. Therefore, decreasing these costs is essential to accelerate the transition and increase the wage of vulnerable groups (low educated workers in routine-intensive occupations). The model also allows me to revisit an old question in labor economics: what is the role of demand in shaping inequality between groups in the U.S? To tackle this, I fix the demand shifters and simulate the counterfactual U.S. labor market of the last 40 years. The simulations tell that changes in the demand were the main drivers of inequality. Without these changes, the college premium would have been almost 10 p.p. lower in 2010 than in 1970. This would create a disincentive for workers to go to college, making their share to be 7 p.p. lower. Finally, in the counterfactual world, the difference between the 90th occupation and the median (in terms of average wages) shrinks 7 p.p, while the difference between the median and the 10th occupation stays the same.
In the second chapter, “Trade-Induced Local Labor Market Shocks and Asymmetrical Income Risk”, we investigate empirically the relationship between international trade and inequality in Brazil. One of current pressing economic questions is how economies adjust to changes in inter- national trade. Although early theoretical work points that openness to trade is bound to increase welfare, present evidence has shown that the adjustment can be long and poten- tially costly for certain groups of individuals. In particular, empirical work have shown that international trade can have first order effects on the average income, employment level and labor force participation of affected regions. Nevertheless, while a lot is known about the impact on the income level, few studies focus on the impact on income risk - the distribution of income changes.
This paper aims to shed light on the effects of international trade on the idiosyncratic income risk faced by the workers. It is easy to rationalize how a riskier labor market can have pervasive consequences to the individual welfare - even when average wages are unchanged. First, economic shocks and unexpected income changes often have persistent effects. Jacobson et al. (1993) shows that displaced workers have lower wages even 5 years after displacement. Second, in the presence of borrowing constraints, unexpected income changes are one of the main drivers of welfare and consumption inequality.
There might be several reasons why international trade is linked to idiosyncratic income changes. Increase in the exposure to trade induces reallocation of labor within and across industries, sometimes associated to long unemployment spells and loss of human capital (Dix-Carneiro (2014)). Further, firms endogenously respond to changes in the international markets by opening and closing plants (Cosar et al. (2014)). As far as ex-ante similar workers follow different labor market trajectories in response to those events, changes in the trade flows can affect the distribution of idiosyncratic economic shocks. On top of the effects from the reallocation of workers, which tend to be transitory, higher integration with international markets can increase permanently income risk. To the extent integrated markets have a higher degree of pass-through from the shocks of the global economy to the local demand of labor, an increase in trade can leave labor income more exposed to changes in the international environment. In addition, even after the economy have adjusted to the trade shocks, the permanent shift in sector and occupation composition might leave the labor market more or less risky.
Our analysis uses Brazilian data to understand the impact of trade on risk. Brazil has been widely regarded as an ideal country to study trade shocks. First, it experienced a variety of changes in its trade dynamics, from the trade liberalization of the early 90’s to the more recent commodity-manufactures trade boom with China in the 2000’s. Second, its sheer size, combined with various natural resources and divergent human capital accumulation, provides a large number of local labor markets with different comparative advantages that may be subject to heterogeneous trade shocks. In the spirit of Autor et al. (2013) and Costa et al. (2016), we exploit the increase in the Brazil-China trade volume between 2000 and 2012 at the national level, together with local industry composition, to construct variation across time and space of the impact of the “China shock” in the Brazilian labor market. Our identification approach relies on within local labor market changes in trade exposure, effectively comparing changes of labor income risk of regions affected by trade with regions that have been somewhat untouched by it. Yet, as in much of the literature, the shift-share estimates would be biased if there are time and region-varying unobserved factors correlated with both the Brazilian labor market and the increase in trade with China, such as sector specific productivity growth or changes in demands for certain goods due to rise in income. Therefore, we use variation in the trade flows of China with the rest of the world (excluding Brazil) to create instruments for the Brazil-China imports and exports. To the extent that the Chinese trade flows with the rest of the world is unrelated to the Brazilian labor market, these are valid instruments.
To construct measures of labor income risk at fine defined local labor markets, we use longitudinal administrative labor data that covers the universe of formal workers in Brazil. In particular, we construct the distributions of n−years income changes using the residual income of workers that have not moved out of their regions. Given the recent focus of non- normal income growth highlighted by Guvenen et al. (2019), our examination is not limited to the variance but also focus on the asymmetry and tails of the distribution. Guided by a region-specific stochastic income process, we interpret the moments of income changes over long and short time horizon as persistent and transitory income shocks.
The empirical results can be summarized as follows. The increase in imports from China is associated with an increase in the dispersion in both the long and short time horizon distribution of income growth. This effect is robust and economically strong. Suppose we rank all regions by their income growth dispersion in 2005. In our preferable specification, an increase in imports of US$ 1000 per worker is sufficient to bring a region in the 25th centile to the 50th centile in the rank dispersion risk. There is no evidence that an increase in the exports to China leads to changes in the dispersion of the labor income risk. In the case of the higher moments, Chinese imports are associated to an increase in the negative income changes in the 5-year distribution, as well as the increase in the kurtosis in the 1-year income growth distribution. Again, the impact of Brazilian exports to China in both the asymmetry and tails is minimal. Finally, we estimate that an increase in imports of US$ 1000 per worker can decrease the average yearly income up to 25%.
Afterwards, we quantify the welfare cost given by the increase in labor income risk from the “China shock”. To tackle this question, we estimate two stochastic income processes: one targeting the empirical moments before the increase in trade flows, and the other targeting the counterfactual moments implied by our causal estimates. We input our estimates into an off-the-shelf incomplete markets model and compute the utility losses of being in the riskier labor market. We found that an unborn individual would be willing to forgo up to 12.43% of his consumption to not be part of this labor market. When decomposing the sources of the increase in risk, results show that the utility losses come mainly from the increase in dispersion. This is because the impact of trade on the higher moments is relatively small relative to the dispersion, and its effects are transitory, making them easy to insure and lowering their welfare influence.
