BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//AMSE//Event Calendar//FR
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:event-9048@www.amse-aixmarseille.fr
DTSTAMP:20260430T002403Z
CREATED:20260430T002403Z
LAST-MODIFIED:20260430T002403Z
STATUS:CONFIRMED
SEQUENCE:0
SUMMARY:phd seminar - Santiago Lopez Cantor*\, Gilles Hacheme**
DTSTART:20220315T100000Z
DTEND:20220315T113000Z
DESCRIPTION:*Public pension schemes act both as inter-temporal income smoot
 hing mechanisms and within-cohort redistribution tools. In this paper we st
 udy the role of inequality on the structure of public pension schemes democ
 ratically chosen. We use a probabilistic voting model where agents vote bot
 h over the contribution and the redistributive degree of the pension scheme
  and can complement it with private savings. In the absence of credit marke
 t frictions\, agents only consider the redistributive power of the system a
 nd use private savings (or borrowing) to smooth income. Inequality then mec
 hanically increases the redistributive degree of pension. Credit market fri
 ctions\, 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 pensi
 ons\, decrease the chosen level of contribution. These findings are consist
 ent with data: countries with higher inequality are associated with lower p
 ublic expenditures in their mandatory retirement schemes\, higher levels of
  intragenerational redistribution and the presence of private pensions.**De
 spite their high predictive performance\, random forest and gradient boosti
 ng are often considered as black boxes or uninterpretable models which hasÂ
  raised concerns from practitioners and regulators. As an alternative\, weÂ
  propose in this paper to use partial linear models that are inherently int
 erpretable. Specifically\, this article introduces GAM-lasso (GAMLA) and G
 AM-autometrics (GAMA)\, denoted as GAM(L)A in short. GAM(L)A combines para
 metric and non-parametric functions to accurately capture linearities and 
 non-linearities prevailing between dependent and explanatory variables\, a
 nd a variable selection procedure to control for overfitting issues. Estim
 ation relies on a two-step procedure building upon the double residual meth
 od. We illustrate the predictive performance and interpretability of GAM(L
 )A on a regression and a classification problem. The results show that GA
 M(L)A outperforms parametric models augmented by quadratic\, cubic\, and i
 nteraction effects. Moreover\, the results also suggest that the performan
 ce of GAM(L)A is not significantly different from that of random forest a
 nd gradient boosting.\\n\\nContact: Kenza Elass: kenza.elass[at]univ-amu.fr
 Camille Hainnaux: camille.hainnaux[at]univ-amu.frDaniela Horta Saenz: danie
 la.horta-saenz[at]univ-amu.frJade Ponsard: jade.ponsard[at]univ-amu.fr\n\nP
 lus d'informations: https://www.amse-aixmarseille.fr/en/events/santiago-lop
 ez-cantor-gilles-hacheme-0
LOCATION:MEGA
URL;VALUE=URI:https://www.amse-aixmarseille.fr/en/events/santiago-lopez-cantor-gilles-hacheme-0
CONTACT:Kenza Elass: kenza.elass[at]univ-amu.frCamille Hainnaux: camille.ha
 innaux[at]univ-amu.frDaniela Horta Saenz: daniela.horta-saenz[at]univ-amu.f
 rJade Ponsard: jade.ponsard[at]univ-amu.fr
TRANSP:OPAQUE
END:VEVENT
END:VCALENDAR