Dynamic effects of labor supply: A mechanism explaining cross-sectional differences in hours

• We try to reconcile the trends in hours and wages in the US over the last 30 years.
• Suggestion: current hours affect the future probability of moving across wage groups.
• This dynamic mechanism is estimated from the data and empirical validation is found.
• We feed this mechanism to an equilibrium model of heterogeneous agents.
• Model results show that the proposed dynamic story is crucial to replicate the trends.

This paper first establishes the empirical fact that over the last quarter of the 20th century, the average weekly hours worked increased for workers in the highest wage quintile while it decreased for the ones at the lowest. In 1976, a worker in the lowest quintile worked 2.8 hours more per week than a high wage worker (worker in the highest quintile), but by 2006, the low wage worker worked 1 hour less. During this period, there was also a wide increase in wage inequality. The typical mechanism in which hours are only determined by contemporaneous wages cannot simultaneously explain the pattern found in both variables for every quintile.This paper attempts to reconcile these cross-sectional trends in both hours and wages for the US during this time period. As a first step, we show that compositional changes (in education, occupation and age) within quintiles can only explain a fraction of the observed pattern. Next, we propose a mechanism in which individuals' current decisions of how much to work take into account two components: the contemporaneous benefit of the wage received, and also how current hours worked affects the probability of moving across the wage distribution in later periods. The latter dynamic component is estimated from our dataset. We find that changes over time in how hours affect these probabilities provided incentives that differ across the quintiles, and are consistent with the labor supply decisions observed in the data. We incorporate these two components into an equilibrium model of heterogeneous agents with uninsurable income risk. We are able to replicate the decline in hours for the bottom of the distribution as well as the increase at the top. The ratio of hours worked between the two groups delivered by the model also fits the trend found in the data.