How is the worker’s weekly hour related to wage over the life cycle? The U.S. evidence

ABSTRACT

The current study estimates the relationship between weekly hours and weekly wage over the life cycle of a representative sample of workers. Recognizing the endogeneity of these two variables, the study estimates both equations in a simultaneous equations framework and demonstrates that the relationship between weekly hours and weekly wage is not uniform over the worker’s life cycle. These two variables are negatively related when the workers are young and have a positive relationship when they are matured adults. This conclusion remains valid for both men and women. Our robustness check further confirms that workers respond to wage increases differently at different stages of their working career. This has interesting policy implications. Any policy to influence the worker’s hours decision through wage incentive must consider the stage of his/her working career.

KEYWORDS:

• Weekly hours
• weekly wage
• life-cycle relation
• two-stage estimation

JEL CLASSIFICATION:

• J22
• J29
• J30

Acknowledgements

The authors wish to thank Ramon Castillo, Miles Finney, Sunil Sapra and the participants of the Economics and Statistics Seminar at California State University – Los Angeles for helpful comments. The usual caveats apply.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The distinction between the prime-age adults and matured adults is clearly arbitrary. These two terms are used simply to distinguish between the two age groups – one older than the other – used in this study.

² The dummy variable full-time or part-time used as an explanatory variable in a typical wage regression is traditionally generated from the worker’s weekly hours of work. This variable usually assumes the value 1 when a worker works for at least 35 h a week, and is zero, otherwise.

³ Unlike several earlier studies, the current study estimates weekly wage equations and not hourly wage equations to avoid the division bias associated with the latter (Borjas 1980).

⁴ The weekly wage was obtained by dividing the annual wage income by the number of weeks worked during that year.

⁵ Due to extremely high non-response rate, the school enrolment variable is not available in the 2010 survey, and consequently this variable has not been used in this period’s hour equation.

⁶ With a t-value of 1.44, the importance of this variable in the hour equation in the year 2000 cannot be ignored completely.

⁷ The finding that the hour wage relationship varies during the life cycle of a worker is revealed when the endogeneity of wage in the hour equation is recognized and the hour equation is estimated by 2SLS. This finding, however, remains disguised when this endogeneity is ignored and the hour equation is estimated by OLS. The unreported OLS results can be obtained from the authors on request.

⁸ With a t-value of 1.47, the importance of wage in the 1990 hour equation cannot be ignored completely.

⁹ The detailed results of this table can be obtained from the authors on request