Flexible Hours, Workplace Authority, and Compensating Wage Differentials in the US

Abstract

The theory of compensating differentials suggests that workers with flexible schedules will earn less than other workers. Some authors have also contended that the concentration of women in jobs with flexible hours explains a significant part of the gender pay gap. This paper uses data from the US subset of the Comparative Project in Class Analysis to test these hypotheses. These data first indicate that, contrary to popular wisdom, women workers do not have more flexible schedules than men. Second, the really striking differential is by race: black workers have much more rigid schedules than white workers. Third, workers with more authority at the workplace typically have more flexibility than subordinate workers. Finally, the data show that any compensating differentials for flexible hours are small and are offset by returns to workplace authority.

Keywords:

• Flexibility
• work schedules
• compensating differentials
• gender pay gap

ACKNOWLEDGMENTS

I gratefully acknowledge the provision of this dataset and supporting documentation by the Inter-University Consortium for Political and Social Research. Thanks to Eileen Appelbaum and three referees for helpful comments and to Michael Hout for help with the dataset.

Notes

JEL Code: J3

For example, one recent study has found women's jobs to be less hazardous than men's, although they do involve a substantial degree of risk (Joni Hersch 1998).

Author's calculations, including only workers ages 18 – 65 who were not self-employed.

Harold Davis, Patricia Honchar, and Lorina Suarez (1987) found that 53 percent of female job-related deaths in Texas were homicides. These murders were all job-related and not due to domestic violence.

This would reduce omitted variable bias, but the differencing could exacerbate problems of measurement error.

Gariety and Shaffer (2001) controlled for occupation and industry as well as the usual human capital characteristics.

It could explain, for example, why the results on flexibility by gender do not show a white male advantage in the US National Longitudinal Survey of Youth, which only asked if employers “make available” a flexible work schedule (original 1979 youth dataset, 1996 data, author's calculations).

In the interest of brevity, I have suppressed the percentages for “joint decision.” The percentages for “myself,” “someone else,” and “joint decision” always sum to 100 percent. If “joint decision” were shown, the columns in Tables 1 and 2 would sum to 100 percent.

The 1991 CPCA data, however, show higher levels of flexibility than the 1989 Current Population Survey (CPS). For a sub-sample defined the same way as in this paper (same ages and excluding the self-employed), the CPS supplement on work schedules in that year showed 15 percent of white men, 14 percent of white women, and 9 percent each of black men and women had control over their starting and stopping times. There were two main differences in how the question was asked and answered: first, the CPS allowed one respondent to speak for all the workers in the household; second, the CPS question only offered a dichotomous answer – either respondents did or didn't have control over their starting and stopping times.

Authority variables were recoded as follows. The questionnaire first asked a screener, “As an official part of your main job, do you supervise the work of other employees or tell other employees what to do?” People who responded “yes” were asked the series of questions regarding whether they were “directly responsible for” deciding tasks, tools, and pace of work; whether they influenced the pay or promotion of their subordinates; and whether they could discipline a subordinate. All people who responded “no” to the screener were coded as zeroes (“no”) on the subsequent questions. Also, all respondents were asked about the frequency with which someone checked on their work, with six possible responses. For brevity, only the two extreme categories are reported (“never” and “more than once a day”). All respondents were also asked whether they were “personally involved” in policy decisions concerning size of workforce, products/programs/services, methods of production, and budgets. There were two possible ways of participating in these decisions: people who “participated directly” were coded as ones (“yes”) and people who just “gave advice” were coded as zeros (“no”).

Workers on flexible schedules work many hours of overtime (Lonnie Golden 2000). I tried several methods to see if there were compensating differentials for overtime and whether these might be related to the flexibility coefficient. The results did not suggest this, but the sample may have been too small to pick up this effect.

I also attempted the selection model using the composite flexibility variable (day off and arrival/departure time), but the number of respondents for whom F_i  = 1 was too small for reliable estimation.

Variables “chk1” through “chk6” were responses to a question on how often someone in authority checks on your work: “chk1" = never; “chk2” = less than once/week; “chk3” (the omitted category) = about once/week; “chk4” = several times/week; “chk5” = about once/day; “chk6” = more than once/day.

I ran another version (not shown) interacting managerial occupation with these variables, but it did not change the results substantially.

The 95 percent confidence intervals for the mean predicted log wages for the two sectors did not overlap.

I also calculated wage differentials between the flexible and inflexible sectors using OLS rather than the selection model (i.e., without a selection equation and an inverse Mills ratio). Coefficients were not much different between the two models.

Because this selection model was difficult to identify, and because of certain problems with selection models in general (Ross M. Stolzenberg and Daniel A. Relles 1990), I also estimated the OLS parameters for the flexible and inflexible sectors. The results were not substantively different from those presented here.

Additional information

Notes on contributors

Elaine McCrate

JEL Code: J3