Modelagem da permanência no mercado de trabalho formal no estado de São Paulo: uma análise de sobrevivência com dados da RAIS

Abstract

This dissertation investigates the duration of formal employment in the state of São Paulo through survival analysis techniques applied to microdata from the Annual Social Information Report (RAIS). The event of interest was defined as the termination of the employment relationship, while employment ties that remained active at the end of the follow-up period were treated as right-censored observations. The empirical strategy combined descriptive analysis, nonparametric estimation using Kaplan Meier curves with correction for delayed entry (left truncation) and survival modeling through semiparametric (Cox) and parametric approaches, complemented by residual analysis and influence measures to assess the quality of fit. The results reveal substantial heterogeneity in the risk of employment termination according to individual characteristics, establishment attributes, and economic sector. The Cox model clearly described the effect of covariates on the termination risk, while, among the parametric models, the log-normal distribution provided a good overall fit to employment duration and was adopted as the parametric benchmark of the study; its regression extension (an accelerated failure time model), which does not rely on the proportional-hazards assumption, recovered the same ordering of effects as the Cox model, reinforcing the robustness of the findings. It is concluded that survival analysis proved to be an effective and robust framework for studying the dynamics of retention and turnover in the formal labor market, producing a detailed and coherent portrait of employment duration and contributing to more precise diagnoses and to the design of evidence-based public policies. The methodological caveats identified delimit the scope of the results without compromising the consistently observed patterns. As possible extensions of this research, particular emphasis is placed on the explicit treatment of non-proportional hazards, the refinement of parameter specification, and the development of an analysis based on identified RAIS data in a secure research environment, which would allow the follow-up window of the employment relationships to be extended.