THE UNION MOTHER EFFECT: DOES UNIONIZATION LESSEN THE WAGE GAP FOR MOTHERS?

Transcription

1 THE UNION MOTHER EFFECT: DOES UNIONIZATION LESSEN THE WAGE GAP FOR MOTHERS? A thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy In Public Policy By Alexa Frank, B.A. Washington, DC April 12, 2016

2 Copyright 2016 by Alexa Frank All Rights Reserved ii

3 THE UNION MOTHER EFFECT: DOES UNIONIZATION LESSEN THE WAGE GAP FOR MOTHERS? Alexa Frank, B.A. Thesis Advisor: Thomas Wei, Ph.D. ABSTRACT A persistent gender wage gap and the increasingly larger share of families with female breadwinners make gender wage inequality a problem not only for women, but also for families. The well-researched motherhood wage penalty demonstrates compounded forms of gender inequality, where women with children earn less than their non-mother counterparts even when controlling for a host of other factors. Yet, research also evidences the existence of a union wage advantage, the higher wages for both men and women associated with union coverage compared to those not in unions (Mishel 2012). Little to no research has examined if and how unionization offsets or reduces forms of inequality, such as the motherhood wage penalty. This paper attempts to fill this gap in knowledge, using a cross-sectional difference-in-difference model to examine if the positive wage benefits associated with unions for women lead to reduced negative effects on wages associated with motherhood. Results from the analysis show that, when controlling for socioeconomic and professional factors, the combination of being a mother and in a union does not lead to consistently positive wage effects. Despite discouraging findings when looking at a nationally representative sample of mothers, specific sub-groups, such as younger mothers, Hispanic mothers, and mothers with less than college degree, do experience positive impacts on wages from being in a union. iii

4 This paper is dedicated to my own mother, whose intelligence, dedication, and work ethic taught me early on that women are no less capable than men. Thank you, Alexa Frank iv

5 TABLE OF CONTENTS INTRODUCTION... 1 LITERATURE REVIEW... 3 CONCEPTUAL MODEL... 6 EMPIRICAL STRATEGY... 8 DATA SUMMARY RESULTS Evidence: Wage Premium and Wage Gap Primary Analysis: No Control Variables Primary Analysis: Control Variables and Specifications Primary Analysis: Subgroups Primary Analysis: Industry Subgroups Secondary Analysis: Different Definitions of Motherhood Results Summary LIMITATIONS Theoretical Limitations Empirical Limitations Data Limitations CONCLUSION APPENDIX A BIBLIOGRAPHY v

6 LIST OF ILLUSTRATIONS Figure 1. Conceptual Model... 7 Table 1. Summary Table Figure 2. Mean Wages by Union Status and Mother Type Figure 3. Average Wage Differences for Mothers Compared to Non-Mothers Table 2. Primary Analysis Figure 4.Comparison of Real Wages from Results of No Controls Model Table 3. Sub-groups Figure 5. Mean Motherhood Wage Gaps by Age and Union Status Table 4. Industry Sub-groups Table 5. Definitions of Mother vi

7 INTRODUCTION The gender wage gap has hovered slightly above 20 percent for the last decade, signaling that economic inequality between the sexes persists as a structural problem. Forms of in-group inequality create additional gaps between women, furthering widening the existing wage gap between the sexes. In particular, the motherhood wage penalty, stemming from a double-bind stereotype that assumes mothers to be less valuable or committed workers, shows women with children earning less than their non-mother counterparts even when controlling for a host of other factors. The increasingly larger share of families with female breadwinners make this a problem not only for women, but also for families. Research suggests that this mommy-tax sustains over time and potentially increases for particular types of mothers, such as single, low-income, and non-white mothers (Waldfogel, 1997; Budig and England, 2007). a However, women do benefit from some traditional economic supports - research demonstrates the existence of a union wage advantage for men and women with membership in a union or coverage under a union contract, with evidence suggesting this premium is higher for women than men (Schmitt, 2008; Schmitt and Woo, 2013). b Women covered under a collective bargaining contract benefit more than just with higher earnings, as unionization has been linked to increased access to health insurance, pension plans, and employer-sponsored benefits (Schmitt and Woo, 2013; Anderson, Hegewisch, and Hayes, a For more on this topic, please see Correll, Benard, and Paik, 2007; Harkness and Waldfogel, 1999; Budig and Hodges, 2014; Avellar and Smock, 2003; and Kahn, Garcia-Manglano, and Bianchi, b For more on this topic, please see Bennett and Kaufman, 2007; Keefe, 2015; Mishel, 2012; Anderson, Hegewisch, and Hayes, 2015; and Lovell, Song, and Shaw,

8 2015). c While a body of literature evidences the positive economic benefits of union membership, particularly for women, less studied is how these benefits might counteract or moderate the negative wage impact on wages for women who are mothers (Looze, 2014). d The rise in economic inequality in the U.S., and perpetual wage gaps between and amongst the sexes, necessitates a thorough look at what remedies currently exist to help low- and middle- income workers, particularly mothers that serve as primary income-earners. The potential relationship between union coverage and reduced economic inequality has relevance to policymakers across a wide spectrum of policy goals, most prominently to support pay equity and lift up working families in the form of increased wages and economic supports. A state-level analysis by the Institute for Women s Policy Research of the wage premium for women in unions suggests that, in 32 states, the weekly earnings difference between union women and nonunion women was enough to cover the weekly costs of infant daycare in their respective states (IWPR, 2015). The relationship also has relevance to policymakers with goals of decreasing political focus on tax mechanisms to support low-income families (e.g. EITC), incentivizing unionization as a form of economic mobility and job security for low- and middleincome workers, and combating Right to Work efforts in states. e This paper will examine the effect of unionization on wage gaps between mother and non-mother workers to see if coverage c For more on this topic, please see Schmitt, 2008; MacGillvary and Firestein, 2009; Kerrisey and Schofer, 2013; and Hess, Milli, Hayes, and Hegewisch, d For similar research on this question, but not in regards to mothers, please see Looze, 2015; Budig & England, 2001; Estes & Glass, 1996; Gangl & Ziefle, 2009; and Anderson, Binder, and Krause, e Right to Work states are those with laws that prohibit unions from requiring members or those covered under the unions to pay fees or dues. For more on this topic, please see IWPR 2015; Mishel, 2012; Chetty, Hendren, Kline, and Saez, 2014; Keefe, 2015; Farber, 2005; MacGillvary and Firestein, 2009; Boushey, Farrell, and Schmitt, 2013; Jorgensen and Appelbaum, 2014; and Gould and Shierholz,

9 under a union lessens the wage penalties mothers might suffer. Findings from this hypothesis will contribute to a body of literature that lacks insight on if and how unionization helps working mothers and if political support for unionization could serve as a viable remedy to narrow gender wage gaps within and across industries and demographics. LITERATURE REVIEW Studies on the gender wage gap reiterate the same finding: women receive less pay for equal work, even when controlling for demographic, educational, and labor-market factors. As the wage differential between men and women has changed minimally in the last decade, policy remedies to narrow this gap have resulted in little impact (Bureau of Labor Statistics, 2014). Research highlights a career pay gap between men and women, with one study finding that the overall lifetime earnings for an average full-time working woman are $434,000 less than her male equivalent due to the difference in annual wages and lessened Social Security and retirement payouts (Arons, 2008). Forms of wage gaps between groups of women, such as mothers earning less than non-mothers, highlight the compounded ways in which mechanisms of gender discrimination can harm women s economic security (BLS, 2014; Budig and England, 2001). The motherhood wage penalty explicitly impacts women who choose to have children, a trend evidenced to be potentially more negative for single mothers, middle- and low-income mothers, mothers with more than one child, and non-white mothers (Budig and England, 2001; 3

10 Gangl and Ziefle, 2009). f A seminal study by Jane Waldfogel found that mothers with one child endured a 6 percent wage penalty compared to non-mothers, and women with two or more children suffered a 13 percent wage penalty (Waldfogel, 1997). This problem inflates when one considers the evolving composition of the U.S. labor market: women comprise an increasingly larger share of the workforce, and mothers make up more than half of primary and co-providers for families. In 2013, 40 percent of households with children under 18 featured mothers as the sole or primary source of income (Wang, Parker, and Taylor, 2013). If working mothers represent almost an equal share of primary income earners, the motherhood wage penalty risks sustaining economic inequality between and amongst U.S. families. However, research also evidences the existence of a union wage advantage, the higher wages for both men and women associated with membership in a union or coverage under a union contract compared to those not in unions (Mishel 2012). Some findings suggest that this premium is higher for women in unions than men in unions, and that the gender wage gap between men and women in unions is smaller than the national wage gap (Schmitt, 2008; Schmitt and Woo, 2013; U.S. Bureau of Labor Statistics, 2015c). Using averaged data from 2009 to 2013, Jones, Schmitt, and Woo (2014) found that unionized women workers earned hourly wages approximately 27 percent higher than non-union women. This boost was larger for lowerwage women and in lower-wage jobs, such as hotel and office cleaners, child-care workers, and health aides. Positive findings corroborate what one might expect about the larger collective bargaining and representation power associated with unions. f For more on this topic, please see Budig and Hodges, 2014; Correll, Benard, and Paik, 2007; and Glauber,

11 The tangential benefits of unionization, such as legal representation, support services, social service referrals, and broader policy advocacy for wages and worker protections, justify why underrepresented groups, such as women, specifically reap economic gains from union coverage: union jobs are more stable and predictable, an important factor for low or middle income working families, and provide greater access to quality health insurance, retirement planning, and forms of paid leave, such as sick days, vacation and holiday time, and family and medical leave (Freeman, 2015; Mishel 2012; Schmitt and Woo, 2014; IWPR, 2015; NWLC, 2015; Kerrisey, 2013; IWPR, 2015). g Not only do unions provide economic benefits, they also combat mechanisms that might more negatively impact women, such as Right-to-Work laws that restrict union power by disallowing them from requiring fees; Shierholz and Gould (2011) found that wages in RTW states were approximately 4.4 percent lower for full-time working women and only 1.7 percent lower for full-time working men compared to non-rtw states. If union coverage leads to positive wage effects and broader worker protections and benefits, particularly for women who already fall behind in these areas, the question then becomes: could the positive wage benefits associated with unions possibly mitigate the wage penalties faced by mothers? Little to no research has examined if and how unionization offsets or reduces forms of inequality, such as the motherhood wage penalty. In her assessment of the motherhood wage penalty, Budig (2001) presents findings that unionization is positively associated with mothers g Accordingly, family members also benefit from having a parent in a collective bargaining agreement; evidence suggests that children who grow up in union households have better outcomes than children in nonunion households. A study by Chetty et al. (2014) using geographic data and controlling for parents income found that a 10 percentage point increase in union density was associated with a 4.5 percent increase in children s incomes later in life. 5

12 wages. However, her work does not examine whether union representation might work to reduce the penalty itself. This paper attempts to fill this gap in knowledge, shedding early light on whether unionization s positive association with wages could lead to reduced motherhood wage gaps and, if so, how this manifests across different types of mothers, such as single mothers, nonwhite mothers, mothers of different ages, and mothers with young children. Positive findings could reinvigorate the current discussion on how to promote the economic security of women and families, and encourage policymakers to invest political resources in unions as a remedy toward economic equality. CONCEPTUAL MODEL Unions work to compress wage distributions as a way of increasing pay equality across workers. Conceptually, this narrowed distribution would lead to smaller wage gaps in the union sector compared to the non-union sector. A model for this analysis would measure the union wage premium and the motherhood wage penalty experienced by a union mother, and determine if the positive union advantage reduces the negative motherhood penalty. This would manifest in the form of a smaller wage penalty for mothers in unions compared to mothers not in unions. 6

13 Figure 1. Conceptual Model Figure 1 shows a hypothetical model of how the union wage premium, and the more compact distribution of wages within unions, and the motherhood wage penalty might jointly affect wages. More specifically, it illustrates the hypothesis that the interaction between the positive wage benefits associated with union coverage and the negative wage penalties associated with motherhood leads to a reduction in the wage penalty. This hypothesis assumes that: unions can successfully compress the wage distribution, women in unions do receive a wage premium, women who are mothers suffer a wage penalty, the union wage premium holds for females with children, and the motherhood wage penalty exists for women in unions. 7

14 The relationships illustrated in Figure 1 feature two key variables of interest that affect wage levels: union coverage, which leads to a union wage premium, and motherhood, which results in a motherhood wage penalty. Additional variables that affect wage levels, and might affect motherhood or unionization, include occupational, residential, and personal characteristics. Given this model, if unionization and motherhood jointly impact wages, and lead to a motherhood wage penalty and a union wage premium, then the union premium for women might be large enough to reduce or offset the penalty suffered by mothers in unions. Accordingly, this paper will examine if the positive wage benefits associated with union coverage for women lead to reduced negative effects on wages associated with motherhood. EMPIRICAL STRATEGY Operationalizing the conceptual model requires calculating the union wage premium for a woman and testing if that wage premium lessens a calculated wage penalty for the same woman who is a mother and also in a union. However, these individual constructs and comparisons are not easily calculated from the data. To operationalize this hypothesis, I will employ a crosssectional difference-in-difference model to examine if the wage gap between comparable mothers and non-mothers in unions is smaller than the wage gap between comparable mothers and non-mothers not in unions. An interaction term of motherhood and unionization will show if union mothers receive a positive wage boost, leading to a wage penalty for mothers in unions that is less negative than for mothers not in unions. 8

15 The model will aim to control for differences between mothers and non-mothers in all respects except motherhood, and between women in unions and not in unions in all respects except union status. To compare wage outcomes across women in similar professional and familial situations, the sample will be restricted to working-age females (ages 16-64) who are employed and earn above $1 per hour and less than $200 per hour. The primary coefficients for this hypothesis include: a dependent log hourly wage variable that includes wages and overtime, an independent indicator for union coverage (individuals who are union members or covered under a union contract), an independent measure of motherhood (any women that report having one or more children), and an independent interaction term (the difference-in-difference estimator) for women who are mothers and also in unions. A log hourly wage variable will be used to standardize comparison across wages, and better illustrate the magnitude of possible effects. Other independent variables available in, or to be constructed through, the data include: an age variable; a squared age variable to account for the nonlinear relationship between age and wages; indicators for married, never married, and single; race variables for blacks, Hispanics, and other; indicators for educational attainment categories, including less than high school, high school, some college, college, and advanced degree; a variable for usual hours worked per week; a squared usual hours worked per week variable; an indicator for whether one is a manager; an indicator for whether one is an hourly worker; an indicator for whether one works full-time (35+ hours per week); an indicator variable for whether the individual is a resident of a Right-to-Work state; an indicator for whether the individual lives in a rural area; an indicator for whether one 9

16 works in the public sector; and indicators for 10 occupational categories and 13 industry categories (see Appendix A). h The primary estimating equation to test the positive wage benefits associated with unions, the negative wage penalties associated with motherhood, and the interaction effect for being a mother in a union, becomes: yi = β0 + β1unioni + β2motheri + β3unionmotheri +β4xi + ui where subscript i represents an individual woman in the data, yi represents the dependent outcome (log hourly wages), xi includes all of the aforementioned control variables, and ui is a random error term. To best understand the nuanced effects of possible wage discrimination, mother will be defined three separate ways: any woman with one or more children, any woman with a single child aged 3 to 5, and, more broadly, the total number of children. The mother variable to be employed in the primary analyses will include any women who report having children, regardless of the child s age or number of children. Other definitions will be employed in a secondary analysis for comparison. The second group (any women with a single child aged 3 to 5) recognizes that wage penalties or premiums may vary depending on the time since the birth of a child. This group of women had a similar amount of time to rejoin the labor force and experience post-child wage effects. The third group, number of children, will demonstrate the linear wage impact for each additional child. h The raw CPS data includes an industry variable with 14 categories (using the 2012 Census Industry Classification) and an occupation variable with 11 categories (using the 2012 Basic CPS record layout). This analysis employs these codes, excluding the armed forces code in each category. Full industry and occupation codes can be found in Appendix A. 10

17 To support the hypothesis, one would expect a positive sign on the union coverage coefficient, a negative sign on the motherhood coefficient, and a positive sign on the interaction term between motherhood and unionization. A positive sign on the interaction term implies that union mothers receive a wage boost and that the wage penalty for mothers in a union is less negative than for mothers not in a union. Extended versions of this analysis use subgroup regressions to compare the wage outcomes of specific groups. These stratified samples include black women, Hispanic women, women that live in a Right-to-Work state, women that are single (divorced, widowed, or separated) or never married, women ages 21-42, students, women ages 43-64, women with less than a Bachelor s degree, and women that earn under $15 per hour. i These subgroup analyses will inform early insight on if and how unionization moderates possible wage penalties across various categories of mothers. In these instances, we restrict regression c. to include only the relevant group of mothers (e.g. black mothers) and observe changes in the unionmother coefficient. While an individual fixed effects model might better isolate the effect of motherhood and unionization on wages over time across individuals, this analysis uses repeated cross-sectional data in its stead. The key assumptions for a consistent and unbiased difference-in-difference model include confidence that there are no omitted variables that differentially affect the wage gap between union and nonunion sectors and between mothers and non-mothers. The feasibility of this assumption is discussed later in the Limitations section of this paper. i Women ages and were chosen to reflect distinct age groups. Younger women might be more likely to be actively having children and motherhood wage gaps would conceivably have the strongest presence. Women earning under $15 per hour were also chosen to examine women lower on the wage distribution. 11

18 DATA SUMMARY This analysis will use the Current Population Survey (CPS) Outgoing Rotation Group (ORG) Extracts for from the Center of Economic and Policy Research (CEPR). j The monthly survey design of the CPS yields a repeated cross-sectional and nationally representative database with measures for employment, wages, demographics, education, and family - that provides insight on the changing socioeconomic landscape of American families and the labor market (U.S. Bureau of Labor Statistics, 2002). The CPS is known for its high response rate of approximately 90 percent, one of the highest amongst government household surveys, increasing confidence in its data despite its changing sample. While each month is considered an individual cross-section, the survey employs a scheme one-quarter of the sample enters the survey each new month, stays in the sample for four consecutive months, leaves for eight months, and then returns for a final four months before permanently leaving the sample. In each month, onequarter of the households are in their fourth consecutive month, also known as their outgoing interviews. These outgoing interviews from each month, which include additional questions on wages and earnings, are gathered together into a single annual file, called the Outgoing Rotation Group (ORG) Uniform Extracts, or the Quarter Sample. This analysis uses the 2012, 2013, and 2014 ORG Uniform Extracts as prepared by CEPR. By choosing three recent and consecutive years, model estimates will utilize a larger sample and more strongly reflect the situation of the j The CPS is a computerized questionnaire administered monthly by the Census Bureau through in-person and telephone interviews to those 15 years and older. Each month, the questionnaire reaches a probability-selected sample of approximately 60,000 occupied households from all 50 states and the District of Columbia. Surveys are administered monthly on the week that includes the 19th, with questions referring to activities from the prior week. 12

19 current labor market. Due to the repeated cross-sectional design of the survey, individuals only appear once in any year, but may reappear in following years of the ORG Extracts. k After restricting the sample to females of working age (16-64) who are self-reportedly employed at the time of the survey, the data includes 233,765 observations, with 12 percent of women in unions (n=28,193), 35.4 percent of women having one or more children (n=82,713), and 4.5 percent of these women with children working in a union (n=10,534). Table 1 shows mean characteristics of the four groups of interest - union mothers, non-union mothers, unionnon-mothers, and non-union non-mothers across a range of demographic factors included in the model. The average union mother in the sample is approximately 40 years old, has 15 years of education, earns $25.41 per hour, and works slightly more than 38 hours per week. Compared to the average union mother, a non-union mother is younger, slightly less educated, earns almost $5 less per hour, and works more than 2 hours less per week. These trends also hold for women without children union non-mothers are older, more educated, earn more, and work more compared to non-union non-mothers. For those who report having children, the average number for both union and non-union women is Notably, more non-mothers report being married, regardless of union status, and the proportion of women in unions is drastically less in Right-to- Work states, regardless of motherhood status. l k While this presents the issue of duplicates in a multi-year sample, this analysis assumes that each woman in each year could conceivably represent a unique observation. More, specifying which duplicates to remove becomes difficult due to household ID numbers as the most unique label in the data. This complicates the process of differentiating individuals within a single household across years. Accordingly, this analysis will not remove duplicates, and count each woman in each year as a unique observation. l Data for this analysis includes 25 Right-to-Work States. As of February 2016, West Virginia became the 26 th Right-to-Work state in the U.S., however this analysis preceded that legislation and does not include workers from West Virginia in its Right-to-Work category. 13

20 Table 1. Summary Table Individual Characteristics Mothers 14 Non-Mothers Union Non-union Union Non-union Hourly Wage ($1-$200) $25.41* $20.64 $25.39* $21.32 Age (16-64) years* years years* years White, race 62.4%* 59.3% 70.1%* 71.1% Black, race 14.5% 14.1% 12.1%* 10.3% Hispanic, race 15.8%* 18.9% 10.4%* 11.3% Other, race 7.3% 7.7% 7.3%* 7.2% Lives in a rural area 11.7%* 14.2% 12.1%* 15.8% Lives in a 'Right-to-Work' state 26.7%* 51.2% 27.3%* 50.7% Family characteristics Married 75.5%* 67.7% 83.2%* 84.4% Single (widowed, separated, divorced) 13.4%* 15.9% 11.5%* 9.2% Never married 11.1%* 16.4% 5.3%* 6.4% Number of children (0-9) Mother, one child aged 3 to % 25.7% - - Mother, one child 23.2% 25.0% - - Mother, two children 20.1% 20.2% - - Mother, three children 6.5% 6.8% - - Educational characteristics Years of Education (1-24) 15.4 years* years years* years Current student.2%*.9%.6%* 1.2% Less than high school 2.7%* 6.5% 3.0%* 4.7% High-school degree 15.0%* 25.4% 22.9%* 29.7% Some college 24.5%* 32.2% 25.1%* 31.1% College degree 28.2%* 23.7% 24.8%* 23.1% Advanced degree 29.4%* 12.0% 24.1%* 11.4% Professional characteristics Usual Hours Worked Per Week 38.5 hours* 36.2 hours 39.2 hours* 37.6 hours Full-Time (35+ hours per week) 85.1%* 74.5% 87.8%* 79.8% Part-Time (<35 hours per week) 14.8%* 25.5% 12.2%* 20.2% Manager 3.4%* 9.5% 4.1%* 10.5% Public Sector 62.6%* 13.3% 63.6%* 15.1% Total Sample 10,534 72,179 10,229 63,760 * Indicates t-test difference in means was significant at p<0.1. Sample includes weighted means of women ages that report being employed. Data from Center for Economic Policy and Research CPS ORG Extracts

21 RESULTS Evidence: Wage Premium and Wage Gap Before looking at the interaction of union status and motherhood, it is important to confirm that the union wage premium and the motherhood wage penalty exist in the data. Figure 2 compares the mean wages of union women and non-union women, supporting past evidence that unionization is positively associated with wages and that this holds for mothers; across all groups of women and mothers, those in unions earn higher average wages. However, contrary to expectations of the motherhood wage penalty, union mothers earn higher average wages than union non-mothers ($25.41 and respectively). This might derive from that fact that, while almost all other groups of mothers do earn less than non-mothers, women with two children earn the highest mean wages ($26.21 in a union, $21.63 not in a union). Figure 2. Mean Wages by Union Status and Mother Type 15

22 While not shown here, mothers with two children had the highest rates of marriage and highest average years of education compared to women with more or less children, and might conceivably be at the peak point of their careers. The data also showed that - excluding women with two children - rates of marriage, education, and hours worked per week all decreased as women reported having more children, evidencing why mothers might earn less than nonmothers. In addition to demonstrating the union wage premium, Figure 2 also evidences the motherhood wage penalty; despite the notable exception of women with two children, all other groups of mothers earn lower mean wages than non-mothers. Figure 3 begins to illustrate the potential relationship between unionization and motherhood, showing that the motherhood wage penalty is consistently smaller in union sectors (dark green) than non-union sectors (light green). Consistent with Figure 2, mothers with two children earn a wage benefit compared to non-mothers, and this benefit is larger for union mothers. Collectively, the data confirms the existence of the union wage premium and the motherhood wage penalty, while Figure 3 illustrates that, without controlling for any other factors, the wage penalty endured by mothers is smaller in unions. 16

23 Figure 3. Average Wage Differences for Mothers Compared to Non-Mothers Primary Analysis: No Control Variables Table 2 shows results from the primary difference-in-difference (DD) model of union coverage, motherhood, and an interaction of the two variables (the DD estimator) all regressed on mean hourly wage. As stated previously, Mother includes any women that reported having children, regardless of age of child or total number of children. In the No Controls model, which excludes any control variables and only compares wages across union and motherhood status, the DD estimator, Union Mother, is positive and statistically significant. Unionization seemingly reduces the motherhood wage penalty in this instance, as the positive coefficient implies that the gap is less negative than it would otherwise be specifically, union mothers 17

24 receive a 5.2 percent wage boost that reduces or offsets the motherhood wage gap compared to the wage gap between mothers and non-mothers who are not in a union. The results of this first model also validate early findings from the data that women in unions earn higher wages (the positive coefficient on covered by a union ) and mothers earn lower wages (the negative coefficient on the respective mother variable). Table 2. Primary Analysis No Controls Full Controls Real Wages, No Overtime Weekly Pay Non- Hourly Workers Self- Reported Responses No Occupation/ Industry Controls No Occupation/ Industry/Age Controls Covered by a Union 0.207*** 0.088*** 0.086*** 0.089*** 0.060*** 0.095*** 0.090*** 0.077*** (0.007) (0.006) (0.006) (0.006) (0.009) (0.009) (0.007) (0.006) Mother *** 0.026*** 0.026*** 0.023*** 0.026*** 0.022*** 0.028*** ** (0.004) (0.004) (0.004) (0.004) (0.006) (0.005) (0.004) (0.003) Union Mother 0.052*** ** ** ** * ** *** ** (0.009) (0.008) (0.008) (0.008) (0.011) (0.011) (0.009) (0.009) Constant 2.88*** 1.892*** 1.889*** 3.353*** 1.475*** 1.928*** 1.563*** 2.137*** (0.003) (0.026) (0.026) (0.027) (0.059) (0.035) (0.0251) (0.016) Observations 156, , , ,763 61,374 71, , ,772 R-squared Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1. Data from Center for Economic Policy and Research CPS ORG Extracts uses weighted values. Control variables in all models include, unless otherwise specified: married, single, age, age (squared), rural, right-to-work state, less than high school, high school degree, some college, college degree, advanced degree, usual hours worked per week, hours worked per week (squared), full-time, manager, hourly worker, white, black, Hispanic, other race, public sector, industry indicators, and occupational indicators. To visualize these findings, Figure 4 illustrates the results of the No Controls model using actual wages rather than log wage. Women in unions earn approximately $4.06 more than non-union women, while mothers earn 68 cents less than non-mothers. The combination of being a mother and being in a union results in a positive wage boost of 71 cents. These results suggest that the union wage boost in this model not only reduces the wage gap, but leads to hourly wages 18

25 for union mothers that are 3 cents higher than union non-mothers ($25.41 per hour compared to $25.38 per hour). While this positive interaction effect is encouraging, particularly due to its size and statistical significance, the No Controls model does not account for potential confounds that might affect one s union or motherhood status. Figure 4. Comparison of Real Wages from Results of No Controls Model 19

26 Primary Analysis: Control Variables and Specifications The Full Controls model in Table 2 shows results of the same DD model, but with the inclusion of occupational, resident, and personal characteristics. m Once control variables are added to the model, the positive union effect remains the same, however mothers now appear to earn more than non-mothers, a notable contradiction to the literature. The interaction coefficient also becomes negative, suggesting that the motherhood wage benefit, in this particular instance, is 2 percentage points less for mothers in unions. While these reduced wage benefits for mothers in unions are small in magnitude, results indicate that, when comparing across like groups of women, unionization fails to serve as a consistent wage boost for working mothers. Table 2 also shows several sensitivity analyses, to determine if the results of the Full Controls model change with different model specifications. These include use of real wages without overtime instead of log hourly wage, use of weekly pay instead of log hourly wage, restriction of the model to just non-hourly workers, restriction of the model to only self-reported (non-proxy) survey responses, exclusion of occupational and industry control variables, and exclusion of occupational, industry, and age controls. Looking at weekly pay counteracts a more narrow approach of using hourly wages, while only looking at non-hourly workers considers the possibility that wage effects of motherhood and unionization might be more strongly demonstrated in salaried positions, where earnings are typically higher and might result in larger m Control variables in all models, unless otherwise specified, include: married, single, age, age (squared), rural, right-to-work state, less than high school, high school degree, some college, college degree, advanced degree, usual hours worked per week, hours worked per week (squared), full-time, manager, hourly worker, white, black, Hispanic, other race, public sector, industry indicators, and occupational indicators. 20

27 differences that the analysis can capture. One could also speculate that self-reported responses, as compared to proxy responses where someone else reported survey responses for an individual, would include more accurate data proxy responses for variables such as wages and hours worked might be upwardly biased in a way that underestimates the penalties associated with motherhood. Lastly, excluding specific control variables that might show wide variance in wage gaps, such as occupation, industry, or age, also represent specifications that have potential to influence results. Despite justifications for these sensitivity checks, results remain the same a positive wage benefit for mothers and a smaller wage benefit for union mothers with all negative interaction coefficients statistically significant and similar in magnitude to the Full Controls model. Primary Analysis: Subgroups The inconsistency inherent in some of the variables relative to wages calls for subgroup analyses (of industry, occupation, age, and others) as a possible way of replicating the positive findings in Table 2. Table 3 applies the fully-controlled DD model to multiple restricted samples, with results primarily confirming that in the Full Controls model some groups still receive a wage benefit from motherhood and mothers in unions receive reduced wage benefits. However, this trend reverses for women who are Hispanic, single or never married, aged 21-42, have less than a college degree, or earn under $15 per hour these women benefit from being both a mother and in a union. Hispanic women, women ages 21-42, and women with less than a Bachelor s degree show the largest positive interaction effect for being a mother in a union (6.7 21

28 percent, 3.5 percent, and 2.2 percent higher wages respectively), with these positive wage boosts all showing statistical significance. Table 3. Sub-groups Black Hispanic Lives in RTW state *Single or Never Married Ages Ages Less than Bachelor's Degree Earns under $15/hour Covered by a Union 0.116*** 0.067*** 0.058*** 0.099*** 0.053*** 0.089*** 0.109*** 0.037*** (0.020) (0.023) (0.011) (0.017) (0.014) (0.007) (0.008) (0.007) Mother *** *** *** (0.011) (0.009) (0.005) (0.008) (0.005) (0.005) (0.004) (0.003) Union Mother * 0.067** ** ** 0.022* (0.025) (0.026) (0.014) (0.020) (0.015) (0.012) (0.012) (0.010) Constant 1.999*** 2.051*** 1.801*** 1.979*** 2.475*** 2.654*** 2.766*** 1.916*** (0.076) (0.060) (0.0372) (0.045) (0.027) (0.027) (0.020) (0.020) Observations 12,933 15,369 58,016 28,507 59,678 69,142 77,685 52,275 R-squared Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1. Data from Center for Economic Policy and Research CPS ORG Extracts uses weighted values. All models are restricted to exclude relevant observations or variables (e.g. All race variables were excluded from the Black and Hispanic models). Control variables in all models, unless otherwise specified, include: married, single, age, age (squared), rural, right-to-work state, less than high school, high school degree, some college, college degree, advanced degree, usual hours worked per week, hours worked per week (squared), full-time, manager, hourly worker, white, black, Hispanic, other race, public sector, industry indicators, and occupational indicators. In contrast to black mothers, Hispanic mothers earn a 6.7 percent wage benefit from unionization - Hispanic mothers might work in low-wage jobs or industries where union representation has a larger presence. Racial stereotypes of Hispanics that differs from that of blacks, likely unaccounted for in the model, may also play a role in these differing results across non-white groups of mothers. Mothers with less than a Bachelor s degree earn a 2.2 percent wage boost for being in a union this suggests that women with lowers levels of education might benefit more from unionization than women with post-secondary degrees. The positive 22

29 wage effects for younger mothers also starkly contrasts that of older mothers mothers ages receive a 3.5 percent boost in wages for being in a union, compared to mothers aged whom earn less from being in a union. Younger mothers are conceivably in prime years of both fertility and professional development this life-cycle stage, compared to that of older mothers, suggests that that these women might soon have children, have recently given birth, or have younger children, all of which might restrict their opportunities or outcomes in the labor market. Accordingly, unions might provide a stronger economic boost for these women. Figure 5. Mean Motherhood Wage Gaps by Age and Union Status Figure 5 illustrates the inconsistent relationship between age and the motherhood wage gap, particularly when broken down by union status. While the wage gap in unions (light green) 23

30 appears less negative from ages 25 to 35, mothers start to earn more than non-mothers around the age of 40, and this wage premium becomes smaller for mothers in unions after age 45. These findings deviate from historical evidence on the motherhood wage gap and help explain results that show the null overall impact of unionization on mothers wages. The variation in wage gaps across age clarifies why union coverage may not have a broad positive impact, if at all, for mothers to overcome the impact of other socioeconomic or individual characteristics. This warrants the question of whether unionization is truly a causal factor in reducing wage gaps, or just a contributive or external factor. However, sub-group analyses suggest that union coverage might be a stronger tool for more specific demographics of mothers, such as younger mothers who are in the prime years of fertility and professional development. While positive wage effects for other groups of union mothers in the sub-group analysis, including unmarried women and women earning less than $15 per hour, fail to demonstrate statistical significance (see Table 3), all of the subgroups with positive wage impacts highlight demographics where unionization might cause a positive boost on mothers wages. Primary Analysis: Industry Subgroups The inconsistent wage impacts seen in the subgroup analysis, such as across age or different races, indicate the value of further exploration into other key variables, like industry. Theoretically, workers in some industries might benefit more from union coverage. Table 4 shows the primary Full Controls DD model restricted to eight different industries while no results are statistically significant, five industries show positive wage benefits that range from.8 percent (Public Administration) to 19.7 percent (Construction). This suggests that mothers in 24

31 these industries might receive an increased wage boost from being in a union, but this cannot be confirmed. Overall, the mixed findings when looking at specific industries fail to contribute any evidence to unionization as a consistent counter to possible motherhood wage penalties. Table 4. Industry Sub-groups Agriculture Construction Manufacturing Wholesale/ Retail Trade Transportation/ Utilities Information Public Admin. Education/ Health Services Covered by a Union *** *** 0.178*** 0.080*** 0.085*** (0.141) (0.096) (0.027) (0.027) (0.028) (0.057) (0.019) (0.008) Mother 0.092* *** *** *** *** (0.050) (0.033) (0.012) (0.009) (0.022) (0.027) (0.017) (0.006) Mother, in a union (0.213) (0.121) (0.038) (0.037) (0.037) (0.083) (0.026) (0.011) Constant 2.121*** 1.976*** 1.949*** 2.098*** 1.752*** 1.725*** 1.336*** 1.913*** (0.326) (0.293) (0.117) (0.059) (0.175) (0.219) (0.140) (0.043) Observations 597 1,491 9,092 14,730 3,417 2,179 7,057 53,029 R-squared Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1. Data from Center for Economic Policy and Research CPS ORG Extracts uses weighted values. These models exclude occupational and industry variables. Control variables in all models, unless otherwise specified, include: married, single, age, age (squared), rural, right-to-work state, less than high school, high school degree, some college, college degree, advanced degree, usual hours worked per week, hours worked per week (squared), full-time, manager, hourly worker, white, black, Hispanic, other race, public sector, industry indicators, and occupational indicators. Secondary Analysis: Different Definitions of Motherhood Table 5 applies the Full Controls model from Table 2 across the three different definitions of motherhood: any women with children, women with one child aged 3-5, and number of children. Much like the sensitivity analyses, these model specifications consider the limitations of the primary analysis, which uses any women with children as the sample for 25

32 mothers. Limiting the analysis to mothers with one child aged 3-5 considers historical literature that suggests mothers with young children are likely to suffer negative wage effects in the labor market. This definition of motherhood restricts the sample the most, but focuses on groups of mothers who might suffer more negative wage penalties and thus might benefit most from the positive impact of unionization on wages. Similarly, number of children as a substitute for motherhood considers the possible linear impact that each additional child could have on wages. Despite comparison across different definitions of motherhood, results from Table 5 still fail to provide evidence for the primary hypothesis. In each of the three models, mothers earn more than non-mothers and union mothers receive a smaller wage benefit. Table 5 also shows the positive and negative correlations for included control variables. These relationships confirm past research being black or Hispanic, and living in a rural area or a right-to-work state are all associated with lower mean wages, while marriage, full-time working status, and higher levels of education are associated with higher mean wages. While being single is interestingly associated with a positive wage impact, these effects are statistically insignificant and too small in magnitude to warrant much merit. 26

33 Table 5. Definitions of Mother Union Mother (all mothers) Mother, one child aged 3-5 Number of children Covered by a Union 0.088*** 0.078*** 0.087*** (0.006) (0.007) (0.006) Mother 0.026*** 0.021** 0.010*** (0.004) (0.009) (0.002) Union Mother ** ** (0.008) (0.026) (0.004) Married *** 0.026*** 0.036*** (0.005) (0.008) (0.005) Single (0.006) (0.010) (0.006) Age 0.028*** 0.028*** 0.029*** (0.001) (0.001) (0.001) Lives in a rural area *** *** *** (0.004) (0.005) (0.004) Lives in Right-to-Work State *** *** *** (0.003) (0.004) (0.003) High School, GED 0.142*** 0.147*** 0.143*** (0.006) (0.009) (0.006) Bachelor's Degree 0.426*** 0.424*** 0.428*** (0.007) (0.011) (0.007) Advanced Degree 0.563*** 0.537*** 0.564*** (0.008) (0.012) (0.008) Full-Time (35+ hours) 0.126*** 0.146*** 0.126*** (0.007) (0.010) (0.007) Race, Black *** *** *** (0.005) (0.007 (0.005) Race, Hispanic *** *** *** (0.004) (0.007) (0.004) Constant 1.892*** 1.851*** 1.891*** (0.026) (0.036) (0.026) Observations 129,772 65, ,772 R-squared Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1. Data from Center for Economic Policy and Research CPS ORG Extracts uses weighted values. Control variables in all models include: married, single, age, age (squared), rural, right-to-work state, less than high school, high school degree, some college, college degree, advanced degree, usual hours worked per week, hours worked per week (squared), full-time, manager, hourly worker, white, black, Hispanic, other race, public sector, industry indicators, and occupational indicators. 27

34 Results Summary Overall, the analyses fail to provide compelling evidence that unionization reduces the motherhood wage gap. While across many instances, union mothers earn higher wages than nonunion mothers, union coverage fails to have a consistent or significant positive boost on mothers wages. Results of subgroup analyses suggest that unionization might cause positive wage effects for specific groups of mothers, such as Hispanic mothers, younger mothers, and mothers with less than a college degree. However, the totality of the analysis indicates that this trend does not hold across a nationally representative sample. Most interesting is the finding that, broadly, mothers earn more than non-mothers. This may result from the large sample, which includes employed women up to 64 years old who might have overcome negative wage effects associated with motherhood and distorted the averages. Certain characteristics controlled for in the model (such as occupation, industry, and age) also might elucidate why a more granular comparison of union and non-union mothers contradicts the original hypothesis. Subgroup analyses of industry and age show inconsistent wage impacts when looking at the variation within these variables. This within-variable variation across wages helps clarify why the motherhood wage gap could, on average, appear larger in unions, as union mothers in some industries or at certain ages are not benefiting as much as others. 28

35 LIMITATIONS Theoretical Limitations Theoretical limitations primarily relate to the economic context of unions relative to the modern economy. First, as the share of union members in the workforce fell from 26.7 percent in 1973 to 13.1 percent in 2011, the broader decline in unionization in the last four decades suggests that labor unions no longer have the numbers to make an impact on a larger scale (Mishel, 2012). n Due to globalization and rapid technological advancements, the twenty-firstcentury labor market has also come to prioritize highly educated and skilled workers that offer advanced competency and innovation, and bring value that modern technology does not. This evolution of the labor market links to the weakening impact of unions by creating less demand for low- and middle-skill workforces that are likely to join unions and benefit from their representation (Western and Rosenfeld, 2011). Globalization has also lead to the breakdown of borders a more synthesized, interdependent economy suggests that the influence of unions in compressing wage distributions may extend to beyond just those workers covered by a union contract. In industries, occupations, or regions with a high density of unionized workplaces, nonunion employers frequently adopt union standards, or make an effort to improve compensation and workplace practices to the same standard as unions, in order to appear on par with competing employers (Mishel, 2012). Researchers call this labor market response the union threat effect: the degree to which n In the same time period that unionization declined almost 50 percent, the hourly wages of the average worker rose by less than 11 percent, representing what some academics call a lost decade of wage growth. 29