How Fragile are Male Labor Supply Function Estimates?

Abstract

We estimate male wage and nonwage income effects using linear specifications spanning three techniques (ordinary least squares, fixed effects, and random effects), two wage measures (reported hourly wages and average hourly earnings), and sample stratification by pay scheme (salaried versus hourly paid). Our regressions encompass the one-period static and perfect-foresight life-cycle models. The static model implies exogenous random person-specific effects, a negative nonwage income coefficient, and a positive labor supply substitution effect. The life-cycle model implies endogenous individual-specific effects, a positive wage coefficient, and a zero nonwage income coefficient. Neither the one-period static nor the perfect-foresight life-cycle models are implied by the data for salaried workers while the static model is consistent with the data for hourly paid workers if income taxes are ignored.

We thank Anthony D. Hall, Peter Schmidt, Robin Sickles, Pravin K. Trivedi, Alan Woodland, Junsen Zhang, members of the Department of Economics Workshops at the University of Kentucky and Indiana University Purdue University Indianapolis, and three anonymous referees for valuable comments and Chris Ehlers and Jean Kimmel for help in computing and tabulating results. Dawn Teolis and Lynne Dennison did their usually fine typing of tables and references. Financial support from the University of New Hampshire, Durham and the Department of Statistics — The Faculties of the Australian National University is gratefully acknowledged.