Abstract: This paper resolves the empirical puzzle in the public-private wage literature: why studies using similar data reach contradictory conclusions about wage premiums and penalties. Utilizing rich French administrative panel data (2012-2019), this study has two main contributions: first, it presents a set of new, intuitive yet previously undocumented stylized facts about wage dynamics, sectoral mobility, and gender differences across sectors. Two, the results reveal that the modest hourly wage gaps conceal substantial disparities in lifetime earnings and employment stability. Women, in particular, gain a significant lifetime earnings advantage in the public sector, driven by higher retention, better-compensated part-time work, and more equitable annual hours compared to the private sector, where gender gaps remain larger, especially for those with higher education. In contrast, highly educated men experience a lifetime penalty in public employment due to rigid wage structures. By flexibly modeling sectoral transitions, transitions into and out of employment, and earnings heterogeneity using an Expectation-Maximization algorithm, this study shows that both premiums and penalties depend systematically on gender, education, and labor market experience. The analysis reveals that significant unobserved heterogeneity remains in wage dynamics. These findings unify prevailing narratives by providing a comprehensive, descriptive account of sectoral differences in transitions, part-time work and wages by gender.

Abstract: This paper proposes a unified framework for studying labor market sorting that bridges two dominant empirical approaches: the Abowd-Kramarz-Margolis (AKM) variance decomposition and match surplus/value function methods. I leverage SHapley Additive exPlanations (SHAP), rooted in cooperative game theory, to decompose wages into contributions from worker types, firm types, and their interactions. This framework nests both AKM (when interactions are negligible) and the Bonhomme-Lamadon-Manresa (BLM) approach (when complementarities are present). The SHAP-based decomposition provides a principled method to (i) test for additive separability, (ii) quantify worker-firm complementarities, (iii) recover match surplus functions, and (iv) infer bargaining power. I develop a two-step estimation procedure combining BLM-style classification with machine learning models and TreeSHAP algorithms. The framework offers model-agnostic interpretation, allows nonlinear interactions, and provides individual-level wage attributions that aggregate to standard variance decompositions. This unification enables researchers to assess whether the restrictive additive structure of AKM is appropriate or whether the richer BLM framework is necessary.

Abstract: The dual objective of this research is to first quantify the extent to which job transitions relate with pecuniary motive using administrative data for France coupled with a finite mixture approach. The novelty here is that we define a job not only as a match of worker i with firm j, but also in occupation k with unrestricted movements of workers to different firms, firm types, and occupations. We employ k-means clustering to categorize firms and Expectation-Maximization (EM) algorithms to classify workers, incorporating occupational data to refine worker classification. Building on the AKM (Abowd, Kramarz, and Margolis) literature, we re-evaluate the extent to which previously identified worker effects can be attributed to occupational choices instead. Then, we decompose the variance of a linear model of log wages with person and firm fixed effects with occupations. We find that occupations explain significant shares of what the literature estimates to be worker and firm effects. We also introduce the practical use of Partial Information Decomposition (PID) to identify unique, redundant and synergistic contributions of firm and occupation effects to worker effects.

Abstract: This study examines the impact of privatization on female full-time employment, with implications for UN SDG 5 on gender equality. Using forecasted treatment effects and bootstrapping, data from 1997-2007 are analyzed to estimate the Average Treatment Effect on female full-time employment. Deterministic trends forecast counterfactual outcomes, while bootstrapping estimates the ATE's confidence intervals. Findings reveal a significant decline in female full-time employment post-privatization. This case study highlights the importance of considering gender costs in privatization policies and underscores the need for robust methods in policy analysis. Further research should explore long-term implications and mechanisms affecting gender equality.