Insuring Unemployment Out of Your Own Pension
Causal Evidence from Mexico’s Retiro Parcial por Desempleo
I study the labor market effects of Mexico’s Retiro Parcial por Desempleo (RPD), which allows unemployed workers to withdraw pension savings. Using rich administrative data and a fuzzy regression discontinuity design exploiting a sharp eligibility threshold, I first show that eligibility increases program take-up by 3.6 percentage points twelve months after displacement. Leveraging these take-up data, I estimate the local average treatment effect on compliers and find that RPD use prolongs time out of formal employment by 36 weeks over three years, without gains in reemployment wages, job stability, or cumulative formal earnings. Testing differences between subgroups directly, the cost concentrates on workers with below-median prior earnings and on men; the analogous gradients by age and by pandemic exposure are suggestive but not statistically distinguishable. No subgroup shows a detectable gain in three-year cumulative earnings, and the only reemployment-wage gain significant at the 5% level accrues to workers already above the median. Because the additional weeks are weeks without pension contributions, the cost compounds until retirement: the loss in accumulated balance is several times the sum withdrawn, and roughly two fifths of it stems from the induced unemployment rather than from the withdrawal itself. The results suggest that while RPD provides short-term liquidity, it does not improve long-term outcomes and may exacerbate inequalities.
unemployment insurance, income substitution, Mexico, labor economics, RPD
Reuse
Citation
@misc{degetau2026,
author = {Degetau, Esteban},
title = {Insuring {Unemployment} {Out} of {Your} {Own} {Pension}},
date = {2026-08-13},
url = {https://www.estebandegetau.com/RPD_LM/},
langid = {en},
abstract = {I study the labor market effects of Mexico’s Retiro
Parcial por Desempleo (RPD), which allows unemployed workers to
withdraw pension savings. Using rich administrative data and a fuzzy
regression discontinuity design exploiting a sharp eligibility
threshold, I first show that eligibility increases program take-up
by 3.6 percentage points twelve months after displacement.
Leveraging these take-up data, I estimate the local average
treatment effect on compliers and find that RPD use prolongs time
out of formal employment by 36 weeks over three years, without gains
in reemployment wages, job stability, or cumulative formal earnings.
Testing differences between subgroups directly, the cost
concentrates on workers with below-median prior earnings and on men;
the analogous gradients by age and by pandemic exposure are
suggestive but not statistically distinguishable. No subgroup shows
a detectable gain in three-year cumulative earnings, and the only
reemployment-wage gain significant at the 5\% level accrues to
workers already above the median. Because the additional weeks are
weeks without pension contributions, the cost compounds until
retirement: the loss in accumulated balance is several times the sum
withdrawn, and roughly two fifths of it stems from the induced
unemployment rather than from the withdrawal itself. The results
suggest that while RPD provides short-term liquidity, it does not
improve long-term outcomes and may exacerbate inequalities.}
}