Minding the Gaps: Pay Equity in California
Pay equity between women and men has improved in the United States from about 60 percent in 1960 to 82 percent in 2018 (Semega et al. 2021, see Figure 5). California is among the states with the smallest pay gaps and outpaces the national number at 13 percent (Wisniewski 2022). California is unique in that it enacted legislation aimed at eliminating pay gaps by sex and race categories. For example, the California Fair Pay Act of 2015 (Senate Bill 358), effective January 1, 2016, shrinks loopholes employers used to justify unequal wages. It does so by changing the definition of equal work to “substantially similar work” and specifying that people doing similar work at different establishments also must be paid equally. Beyond the gender pay gap, California also amended the Equal Pay Act (1949) with the Fair Pay Act in 2016 (SB 1063 and AB 1676, effective January 1, 2017) and 2017 (AB 168 and AB 46, effective January 1, 2018) to prohibit unequal pay for employees of different race or ethnicities, prohibit employers from asking for a potential employee’s salary history, and expand coverage to include public and private employees, respectively. Given that legislation assumes potential differences in pay for similar work by establishment necessitates controlling for establishment, Firm size is also a potentially important and untested control variable in gender pay gap studies (see Blau and Kahn 2017).
The California Commission on the Status of Women and Girls (CCSWG) led the Pay Equity Task Force to implement SB 358 and create guidance about the law for employees, employers, and unions. I served as CCSWG staff senior research consultant on the task force where I, among other things, organized task force subcommittees by performing a content analysis of web documents advising readers how to address pay equity in the workplace.
The Course
For a social statistics course I teach, we use the U.S. Census American Community Survey (ACS) to analyze the gender pay gap in California. Students deploy the ACS 2014-2018 five-year file to learn about how sex and race categories operate via the institutional pay and occupation structure. The five-year file consists of a weighted sample of 12,581,405 full-time, year-round working people. They test variables indicating three theoretical frameworks, 1) human capital, 2) occupational segregation, and 3) discrimination. While human capital refers to indicators such as education and experience, the latter two are indicated by occupation and demographic characteristics such as race, ethnicity, and gender. Variables sometimes indicate more than one theoretical framework.
At the beginning of term, students learn that less than half of full-time California workers are women (40.7%) who outpace men in each income category except those paying over $100,000 annually, though the two groups appear to break even in the income category just below this, a glass income ceiling of sorts. Women working full-time, year-round in California are overrepresented in the lowest income categories, and men are overrepresented in the highest income categories. The California gender pay gap in this dataset, including outliers, is about 21%.
By midterm, they learn occupation alone is weakly related to gender categories, education category alone is strongly related to income category, and that, once age is controlled for, education remains a strong predictor of income while age is moderately so. Knowing someone’s gender does little to reduce errors in the prediction of occupation category (Lambda = .033), while education category on the other hand is a strong predictor of income category (Gamma = .535). In a multivariate analysis of income as a continuous measure, knowing someone’s age reduces error about 21% (R = .211) of the time and knowing someone’s education level does so about 32% (R = .321) of the time.
Two interesting side points emerge. First, the modal occupation of full-time year-round California workers is a manager once managers and supervisors are recoded into a single group across occupation categories. Second, older workers in these cross-sectional analyses tend to have lower education levels; younger workers are more educated.
In the final analysis, students perform a regression analysis of gender and race categories in addition to continuous age and education (i.e., an ordinal measure treated as interval/ratio) measures to better understand how each variable impacts income.
The scenarios they analyze compare a 25-year-old, white, male with 12 years of education to a 40-year-old, nonwhite, female with 16 years of education. Plugging in the model coefficients yields $18,510 and $34,481, respectively. For the more engaged students, it is eye-opening that being a man indicates someone earns almost $22,000 more a year on average and that being white does so to the tune of about $9,000 a year. Education remains the strongest predictor of income with a standardized coefficient of .337. Significance levels also indicate we can generalize these sample findings to the population. Note: outliers are not removed and income is self-reported. Additional procedures should be done to refine the model.
Conclusion
While the work students do is rigorous using a representative sample of full-time year-round California workers, there remains work to be done. Not only is there a need for firm-level analysis, as pointed out in the introduction to this blog piece, but the shifts in law offer a good starting point for a longitudinal or cross-sectional time-series analysis. The shifts in California law allow a unique opportunity for a natural experiment. Testing the statistical significance of observed associations before and after a California law change will allow us to generalize results should other states follow suit and for researchers interested in how changes in law impact outcomes. That California’s entire working population was exposed to these changes at the same time creates large organic control and treatment groups and further increases the likelihood of being able to generalize results. And, comparing California's pre- and post-changes in law to states in the continental United States will control for larger trends that might otherwise go unmeasured.
Further, local California county women’s commissions are interested in knowing how their counties are doing. County-level analyses may pinpoint what can be done at the local level to eliminate pay gaps favoring one demographic category over others. However, publicly available data yield large margins of error as demographic variables are added to a model, making those data less generalizable when doing a county-level analysis. Restricted-use disclosure avoidance rules also prohibit providing detailed statistical output by county or otherwise— limiting an understanding of the role of local efforts in closing pay gaps unless robust authoritative individual-level data can be accessed and analyzed.
References
-
Semega, Jessica, Melissa Kollar, John Creamer, and Abinash Mohanty. 2021 (updated). Income and Poverty in the United States: 2018. U.S Census Bureau. https://www.census.gov/content/dam/Census/library/publications/2019/demo/p60-266.pdf Accessed on November 18, 2022.
-
Wisniewski, Megan. 2022. In Puerto Rico no gap in median earnings between men and women. U.S. Census Bureau. https://www.census.gov/library/stories/2022/03/what-is-the-gender-wage-gap-in-your-state.html Accessed November 18, 2022.
-
Blau, Francine D. and Lawrence Kahn. 2017. “The Gender Wage Gap: Extent Trends and Explanations.” Journal of Economic Literature v.55. Accessed on August 9, 2023. https://www.nber.org/system/files/working_papers/w21913/w21913.pdf