State Policies and Gender Earnings Inequality: A Multilevel Analysis of 50 U.S. States Based on U.S. Census 2000 Data

Abstract

This article examines distinct dimensions of state government intervention in labor markets across states in the United States and investigates the effect of these interventions on gender inequality in earnings. Statistical models that take into account the contextual effects of family policies on gender inequality in earnings are constructed. Results from multilevel models show that progressive state institutional environments supportive of norms of equality help female employees catch-up with their male counterparts with regard to rewards, while states that function as welfare providers and employers exacerbate the gender gap in earnings.

ACKNOWLEDGMENTS

The author would like to thank William P. Bridges, Moshe Semyonov, Rachel Gordon, and Robert Kaestner for their helpful comments on an earlier version of this article. Three anonymous reviewers helped to improve and clarify my arguments. All responsibility for errors remains with the author.

NOTES

Notes

¹ Because much of the comparative literature uses the term “state” to refer to national-level units, there is a high risk of linguistic confusion for readers of this article. In what follows, when I refer to the “state” or the “welfare state,” I am referring to the roles played by state governments in the United States with regard to shaping institutional environments and substantive welfare policies. Although these terms usually denote the state as a national-level political entity in the literature on welfare state development and gender policies, the decentralized nature of the American polity leaves open the possibility that the roles of welfare states as usually seen in other industrialized countries have devolved, in large part, to “states” in the sense of one or another provincial-level entity in the United States.

² In the following discussion, I use the term “liberal” to denote “a willingness to use the state to achieve social purposes” as suggested by O'Connor et al. (1999), in which they distinguish it from other usages of “liberalism” in the political process of decision making. With regard to gender or family-related policies, it emphasizes the roles of the state in providing social services or in alleviating the burden of mothers who have to reconcile domestic and work-related obligations. However, what “liberal” means in the United States differs largely from what this term means in other contexts. In U.S. policy, the former approximates the latter in their meanings, as in the United States, “liberal” means state intervention in the field of social policy and labor markets to achieve the specific goal of increasing labor force participation or lessening income differentiation. Therefore, I use the term “progressive” instead of “liberal” in the following discussion, because the focus of this article is not to characterize U.S. states as “liberal” welfare states or not, but to compare gender policies across U.S. states and evaluate the consequences of these policies for gender outcomes in labor markets. In a similar vein, O'Connor et al. (1999) posited that “we use the term ‘liberalism’ as it is understood in most of the world—as a commitment to a minimal state role and a substantial role for private institutions. This is different from the way it is usually used in discussions of U.S. politics, where ‘liberalism’ connotes what we call ‘social liberalism’, that is, a willingness to use the state to achieve social purposes, while ‘conservatism’ connotes what is elsewhere called ‘neo-’ or ‘traditional liberalism’, that is, minimizing the role of the state” (p. 3).

³ Institutional theory posits that ideology within government bodies channels demands from constituencies to achieve legitimacy within the political systems; I draw on this to argue that institutions channel ideology in a manner that allows one to measure the characteristics of policy in a certain area, for example, the area of family policies. Political institutions such as the state, senate, and/or congress shape social policy processes by either constraining the policy choices available or by channeling the demands of their constituency. In terms of gender inequality, this can help elucidate implicit or explicit preferences regarding gender equity claims. According to the neo-institutional theories of organizations (Meyer and Rowan 1977), “organizations need to cope with uncertainty that exists around their boundaries” (Guthrie and Roth 1999). Guthrie and Roth (1999) also note that institutional environments denote “normative or cultural pressures that arise from legal mandates, which are closely tied to political mandates of Equal Employment Opportunity and Affirmative Action laws” (514–15).

⁴ This paradox can also be explained by the compensating differential theory of wages in economics, which argues that female employees trade lower earnings for the female friendliness that public employment offers (Polachek 1981; Jacobs and Steinberg 1990). This is a supply-side approach to addressing gender gaps in occupational achievement and earnings. However, it is not straightforward to distinguish this line of reasoning from the demand side explanations presented above given the limitations of available data.

⁵ The interaction terms between these two dummies and a female sex are statistically significant, but I did not include them in the models estimated below as they are endogenous to the effects of state-level policy variables on the gender earnings gap (see Mandel and Semyonov 2006).

⁶ The numerator of the variable was retrieved from the “State and Local Government Employment and Payroll Data, March 1999 (Revised June 2001), by State and by Function, Federal, State, and Local Governments” (see http://www.census.gov/govs/www/apesstl.html for details). The denominator, which is the total civilian labor force, was the same as used below (see Footnote 5).

⁷ The interaction terms between these two dummies and females are statistically significant, but I did not include them in the models estimated below as they are endogenous to the effects of state-level policy variables on the gender earnings gap (see Mandel and Semyonov 2006).

⁸ The numerator of the variable was retrieved from the “State and Local Government Employment and Payroll Data, March 1999 (Revised June 2001), by State and by Function, Federal, State, and Local Governments” (see http://www.census.gov/govs/www/apesstl.html for details). The denominator, which is the total civilian labor force, was the same as that used below (see Footnote 5).

⁹ Data were extracted from the Census 2000 Summary File 3 (SF 3) Sample Data, Table: P43 Sex by Employment Status for the Population 16 Years and [15]. The universe for the data is the population group 16 years and older.

¹⁰ Data were extracted from the Regional Economic Information System/Bureau of Economic Analysis/U.S. Department of Commerce, Per Capita Personal Income by States published in 1999.

¹¹ Given the skewed distribution of the FMLA variable, it is possible that regression estimates for the FMLA variable may be biased. To test this possibility, I trichotomized the variable into three categories. The first category included states with FMLA values of 0, the second category included states with FMLA values larger than 0 and no larger than the median, and the third category included states with FMLA values larger than the median. The results remained consistent regardless of the types of FMLA variables included in the models.

¹² Individual covariates include age, education, marital status (1 = married), number of own children, childhood status, and race/ethnicity (black, other, and Hispanic; white is the reference group).

¹³ Female typed occupations are defined as occupations where the proportion of females employed is at least 1.5-fold the nationwide average proportion of female representation (.468) across all census three-digit occupations included in the sample.