Essays on Wage Polarization

This dissertation, in its four essays, considers whether employment polarization entails a similar pattern on wages, inducing higher relative wage growth among low and high-skill workers with respect to middle-skill occupations that perform routine work. In an overall critical view of the previous literature the four chapters below contest the conversion of employment into wage polarization. Instead, labour market institutions are identified as important mediators of the relation between technological change and earnings. Chapter 1 analyses task allocation models that lead to wage polarization due to the automation of middle-skill, routine work. It extends existing models by the introduction of an additional industry responsible for the production of the machines or inputs that substitute routine work. If the production of these machines demands labour and the two sectors have identical technologies it is shown that the substitution of routine work in the final good sector does not lead to employment or wage polarization. If on the other hand the production of “machines” demands highly skilled workers only, the substitution of routine work actually leads to an upgrade of middle-skill workers that tend to occupy tasks previously performed by high-skill workers in the final goods sector. Chapter 2 questions the hypothesis that over the last decades technological progress has increased returns to abstract, cognitive work and reduced those related to physical, usually routine, work activities. It puts forward an alternative interpretation whereby wages are related to employees’ control over work, which is conditioned by the institutions that represent workers’ interests. First, previous nterpretation of O*NET work activities data as measures of task intensity and also the relevance of routine and non-routine work disaggregations for wage outcomes are contested. Second, the estimated multilevel models display a significant degree of heterogeneity between industries in the returns to employee control. Such heterogeneity does not have an intuitive pattern. Although there is an expected negative relation between abstract and physical employee control between industries, several service industries present positive returns to physical control and as many manufacturing and infrastructure industries are characterized by a wage premia related to control over abstract activities. Finally, multinomial logit models at the industry level provide further evidence against technology, proxied by the change in the occupational composition of employment, as the main factor behind wage returns to employee control and identify the role of institutions like professional associations and unions in improving the odds of a positive, or at least avoiding negative random effects estimates for the relation between employee control and wages. The 3rd chapter relies on similar data but considers the relation between work content and the supervision of work. It proposes a novel measure of supervision based on the frequency of supervisory detailed work activities from the O*NET database. We test the hypothesis that the intensity of supervision is increasing on the difficulty to monitor workers, measures by the content of non-routine cognitive work in an industry. At the industry level the analysis suggests supervision increases with the complexity of the activities performed by workers. Hence, in striking contrast to previous studies relying on the ratio of supervisors to employees, industries that employ more workers performing nonroutine cognitive activities also rely on supervision the most. Moreover, at the individual level there is no evidence of a trade-off but only weak evidence of a positive relation between wages and supervision which seems to be a complement rather than a substitute to wage incentives. The 4th and final chapter studies the mediating role of collective bargaining on the effects routine work automation had on wage polarization in Portugal between 1995 and 2012. Relying on a linked employer-employee that covers all firms and workers in the Portuguese economy we find evidence of greater wage premia from the introduction of new collective agreements within job spells for low and high with respect to middle-skill occupations. However, accounting for measures of bargaining power, given by the concentration within each collective agreement of workers in an occupation, the increased returns to low-skill occupations seems to be driven by unions’ strategies to reduce inequality, increasing wages in the bottom of the distribution at the expense of middle-skill employees that constitute by far the largest skill group in Portugal while high-skill workers are usually under-represented in collective agreements with the other two skill groups. There is no clear evidence of wage polarization in the Portuguese economy.