Abstract :
[en] This dissertation examines the impact of institutions on the distribution of jobs and wages, with a special focus on European countries. We are more specifically interested in labor market polarization, a phenomenon which has notably been observed in the US between the 1980s and the 2010s. While wage polarization describes an increase in wages at both ends of the distribution relative to the middle, job polarization refers to an increase in the employment share of both low- and high-skill jobs relative to middle-skill jobs. A now standard explanation of this phenomenon is routine-biased technical change (RBTC). According to this approach, technical progress favors the substitution of machines --- and, indirectly, of high-skill workers --- for middle-skill labor, which leads in fine to the polarization of the distribution of earnings and jobs.
While labor market polarization has also been observed in Europe, the intensity of this phenomenon depends on the country considered. Since developed economies have a similar access to technology, other determinants of the distribution of jobs and wages have to be considered to explain these cross-country differences. This dissertation considers institutions and demonstrates that a highly institutionalized labor market mitigates the twofold phenomenon of polarization. Institutions --- which are typically country-specific --- can thus partially explain cross-country differences in labor market polarization.
This dissertation consists of three essays. In the first essay, we implement decomposition methods to show the impact of institutions on the wage structure. Our strategy makes use of the difference between the public and the private sector in terms of institutionalization of the wage-setting process. Decomposing the change in wage quantiles for both sectors and operating a between-sector comparison of the results for a set of European countries, we reach the conclusion that institutions are able to mitigate the RBTC-induced polarization of the wage structure.
In the second essay, we develop a theoretical model based on the Acemoglu and Autor (2011) task-based framework. We contribute to this framework by including, in a Ricardian model of the labor market à la Acemoglu and Autor, an institutional device which mitigates wage polarization, based on the results of the first essay of this dissertation. While this device can be thought as unions operating in a centralized and coordinated bargaining regime, it is not restricted to this interpretation. Our model predicts that the institutions we consider, by mitigating wage polarization, have an anti-polarizing impact on the change in employment: while job polarization still follows skill-biased technical change, it is less pronounced in a highly institutionalized labor market.
In the third and last essay, we test the predictions of the model presented in the second essay by empirically assessing the impact of institutions on job polarization. For each country studied in the first essay and for each year of the period 1992-2017, we build a measure of job polarization, based on the employment levels observed in selected, ranked and aggregated occupational categories. We then build a composite index of institutionalization and use panel cointegration techniques to estimate the long-run relationship between this index and each component of job polarization. Finally, we take into account both the reverse-causality problem implied by such a study and the potential delayed response of the variables by resorting on panel vector autoregressive models and structural impulse response analysis. Our results indicate that de-institutionalization fosters both components of job polarization, confirming the structural interpretation of the model introduced in the second essay.
By combining different approaches and methods, this dissertation thus shows that the impact of technical change on the wage and occupational structure is mediated by country-specific institutional settings. Institutions do have an impact on labor market polarization, and can clearly be used to foster or inhibit this phenomenon.