Santiago Lopez Cantor*, Gilles Hacheme**
Kenza Elass: kenza.elass[at]univ-amu.fr
Camille Hainnaux: camille.hainnaux[at]univ-amu.fr
Daniela Horta Saenz: daniela.horta-saenz[at]univ-amu.fr
Jade Ponsard: jade.ponsard[at]univ-amu.fr
*Public pension schemes act both as inter-temporal income smoothing mechanisms and within-cohort redistribution tools. In this paper we study the role of inequality on the structure of public pension schemes democratically chosen. We use a probabilistic voting model where agents vote both over the contribution and the redistributive degree of the pension scheme and can complement it with private savings. In the absence of credit market frictions, agents only consider the redistributive power of the system and use private savings (or borrowing) to smooth income. Inequality then mechanically increases the redistributive degree of pension. Credit market frictions, through the inability to borrow over future pension, however set an upper bound to contributions that depends negatively on inequality. Then, higher inequality although increasing the redistributive degree of pensions, decrease the chosen level of contribution. These findings are consistent with data: countries with higher inequality are associated with lower public expenditures in their mandatory retirement schemes, higher levels of intragenerational redistribution and the presence of private pensions.
**Despite their high predictive performance, random forest and gradient boosting are often considered as black boxes or uninterpretable models which has raised concerns from practitioners and regulators. As an alternative, we propose in this paper to use partial linear models that are inherently interpretable. Specifically, this article introduces GAM-lasso (GAMLA) and GAM-autometrics (GAMA), denoted as GAM(L)A in short. GAM(L)A combines parametric and non-parametric functions to accurately capture linearities and non-linearities prevailing between dependent and explanatory variables, and a variable selection procedure to control for overfitting issues. Estimation relies on a two-step procedure building upon the double residual method. We illustrate the predictive performance and interpretability of GAM(L)A on a regression and a classification problem. The results show that GAM(L)A outperforms parametric models augmented by quadratic, cubic, and interaction effects. Moreover, the results also suggest that the performance of GAM(L)A is not significantly different from that of random forest and gradient boosting.