EARNINGS INEQUALITY AMONG WOMEN: DOES MARRIAGE MATTER? A Thesis by. Sabina Thapa. B.A., Wichita State University, 2008

Transcription

1 EARNINGS INEQUALITY AMONG WOMEN: DOES MARRIAGE MATTER? A Thesis by Sabina Thapa B.A., Wichita State University, 2008 Submitted to the Department of Sociology and the faculty of the Graduate School of Wichita State University in partial fulfilment of the requirements for the degree of Master of Arts May 2012

2 Copyright 2012 by Sabina Thapa All Rights Reserved

3 EARNINGS INEQUALITY AMONG WOMEN: DOES MARRIAGE MATTER? The following faculty members have examined the final copy of this thesis for form and content, and recommend that it be accepted in partial fulfilment of the requirement for the degree of Master of Arts with a major in Sociology. Twyla Hill, Committee Chair Jodie Hertzog, Committee Member Carolyn Shaw, Committee Member iii

4 DEDICATION To my parents Ramesh Kumar Thapa and Shyama Thapa iv

5 ACKNOWLEDGEMENT I owe a deep debt of gratitude to Dr. Twyla Hill for her constant encouragement, guidance and valuable supervision. I would like to express my sincere indebtedness to Dr. Jodie Hertzog and Dr. Charles Koeber for constant mentoring, Dr. David Wright and Dr. Ron Matson for encouragement, and Dr. Carolyn Shaw for comments and suggestions. I am grateful to my parents for continuous inspiration, love, and support they have been providing me and bringing me up to this stage. Therefore I would like to dedicate this thesis work to my parents. I am also grateful to my husband for being so much supportive to my study. Needless to say, I alone am responsible for any deficiencies that may have remained in this study. v

6 ABSTRACT This study aims to investigate the effect of marriage on women s earnings. Using Current Population Survey (CPS) 2010 data, three sets of hypotheses are tested to address the effect of individual level factors, structure level factors and gender/race factors. The results suggest that education, experience, level of occupation, and size of business, among others, are the important factors explaining earnings inequality among women. Marriage has a significant effect on women s earnings and married women have consistently higher income than unmarried women. Some interesting and striking results of this study hold significant sociological and policy importance. vi

7 TABLE OF CONTENTS SECTION Page 1. INTRODUCTION 1 2. LITERATURE REVIEW Individualist Approach Human Capital Theory Selectivity Hypothesis Productivity Hypothesis Structural Approach Dual Economy Theory Segmented Labor Market Theory Gender Model A Composite Model and Research Hypotheses 15 3 RESEARCH METHODOLOGY Data and Sample Variables Dependent Variable Independent Variables Methods of Analysis RESULTS AND DISCUSSION Univariate and Bivariate Results Multivariate Results OLS Regression Results Comparison of Models Partitioning of Variance Discussion Individual Level Structural Level Gender/Race CONCLUSION Implication Limitations 36 REFERENCES 38 APPENDIX 43 vii

8 1. Introduction Over the last several decades, there has been a significant shift in gender-role attitudes about women s role in household work and childrearing, and involvement of women in the labor force. The shift towards more egalitarian gender-role attitudes has helped women to pursue higher education and better their career opportunities (Fan and Marini, 2000). The increase in the age of first marriage of women and the tendency of increasing cohabitation indicate that women these days are foregoing or delaying marriage (Fan and Marini, 2000; Waite, 1995). However, a persistent question arises - does foregoing or delaying marriage provide economic benefits to women or does marriage improve women s earnings? Is there any wage gap between married and unmarried or single women 1? Becker (1985) states that married men will have higher earnings than unmarried or single men because of the comparative advantage of marriage available to married men. But the question still remains as to whether this premium applies to married women. While the effect of marriage on earnings has received considerable attention from sociologists, the focus has been on the marriage premium for men. Empirical studies often find higher wages for married men than unmarried or single men (Chun and Lee, 2001; Korenman and Neumark, 1991). But the theoretical underpinnings and empirical analysis of earnings differences between married and unmarried or single women has not received considerable attention. Some earlier studies focused mainly on gender role specialization, the motherhood penalty and/or racial differences. For example, Hill (1979) reported that married women have weaker labor force attachment and greater absenteeism than single women. Similarly, women with children have had lower wages than childless women (Council of Economic Advisors, 1998). Alon and Haberfeld (2007) reported that minority women earn 1 In this study, unmarried women and/or single women means never married women. 1

9 less than White women. A few studies have provided evidence regarding the effect of marriage on women s earnings. For example, Waite (1995) states that Black women enjoy a marriage premium of about three percent, whereas White women experience a penalty of about four percent. Korenman and Neumark (1992) and Toutkoushian (1998) in their studies did not observe a significant marriage premium for women. Furthermore, Correll, Benard and Paik (2007) state that mothers experience disadvantages in the workplace in addition to those commonly associated with gender and employed mothers suffer a per-child wage penalty of approximately five percent (Budig and England, 2001). However, still there is a lack of a detailed study about the effect of marriage on women s earnings considering other individual level, structural level, and race and ethnic variables. To this end, this thesis uses the 2010 Current Population Survey (CPS) to examine the effect of marriage on women s earnings. This thesis is organized as follows: with this introduction, Section two presents a review of theories on the wage gap which will be used to construct a model and hypotheses that will be tested in this research; Section three explains the data and methodology; Section four presents the univariate, bivariate, and multivariate results and discussions. Finally, Section five concludes the thesis. 2. Literature Review There are various theoretical explanations for earnings differences which can be grouped into three basic approaches individualist approach, structuralist approach and gender approach. The individualist approach explains the earnings gap based upon the individual level characteristics such as education and work experience, whereas structuralist models explain the earnings gap based on occupational roles. Furthermore, the gender models explain the earnings gap based on sex, race and/or ethnicity. Though these models 2

10 generally explain the earnings gap between married and unmarried men, these models have some relevancies while explaining the earnings difference between married and unmarried women. In the following sections I have reviewed these models and attempted to connect those models with potential earnings gaps between married and unmarried or single women. 2.1 Individualist Approach The individualist approach describes sociology as a study of individuals in a social setting where society is regarded as the sum total of the individuals living in the society (Mayhew, 1980). Hence, individual characteristics or human motives are important while studying social behaviors (Mayhew, 1980; Reskin, 2003). In this context, individual models state that earnings difference is the effect of individual level characteristics such as productivity and selectivity (Becker 1985; Nakosteen and Zimmerman, 1997). Within the Individualist approach there are three major theoretical explanations for income (wage) inequality. They are Human Capital Theory, Selectivity Hypothesis and Productivity Hypothesis which are explained below Human Capital Theory Human Capital Theory advanced by Becker (1985) states that individuals are rational beings. Therefore, they make choices to invest in human capital (i.e. education and training) in order to increase their productivity in their jobs thereby future earnings. Individuals with higher productivity are rewarded with higher pay (i.e. those who have invested in human capital will receive higher wages because wage is the reward given for the use of labor productivity) (Becker, 1985). Similarly, it is assumed that as the number of years in work increases, it increases the level of skill through experience. Hence more experienced workers will be more productive and will have higher pay. 3

11 While linking marital status with earnings, the Human Capital Theory states that the attitude of investment in human capital by married and unmarried or single men varies. Married men tend to spend more time and money to gain market specific knowledge and skill because their wives allocate more time and energy to household responsibilities (Becker, 1985; Gray, 1996; Korenman and Neumark, 1991). Waite (1995) further argues that married men have lower rates of alcohol abuse, lower mortality rates and higher sexual satisfaction. It is also argued that employers perceive married men as more committed, motivated and productive (Becker, 1985). As a result married men receive higher wages in comparison to unmarried men. However, this may not be the case for women. Unmarried women tend to invest more in human capital for enhancing their level of education, skills and knowledge for career development in comparison to married women because married women prioritize family commitment over career commitment (Stickney and Konrad, 2007). Married women, in general, prefer part-time work after having a child particularly until the child goes to school (Treas and Widmer, 2000). This makes married women less experienced and less productive at the work place. As level of education and work experience are important factors responsible for wage differences of women (Browne and Askew, 2005), married women tend to earn lower wages than unmarried women Selectivity Hypothesis The second explanation for the marriage premium is put forward by the Selectivity Hypothesis, which states that men with higher earnings, better and more secure jobs, and stronger economic perspective are likely to marry because they are valued more in the labor and marriage market (Becker, 1985; Ginther and Zavodny 2001; White and Rogers, 2000). Nakosteen and Zimmerman (1997) interpret the observance of higher wages for married men as an effect of the mate selection process done by women. Therefore, men with higher 4

12 earnings are more often selected for marriage. For men, their likelihood of marriage depends on their earnings and some personal traits (Chun and Lee, 2001). The attributes that lead to success in the workplace (responsibility, honesty, dedication, etc.) overlap with the attributes that lead to success in finding and keeping a spouse (Becker, 1985; Ginther and Zavodny, 2001). As men with higher income are also less likely to divorce than men with lower income (Waite and Gallagher, 2000), males with higher incomes are most likely to get married and when married, less likely to divorce, therefore are more likely to have higher earnings than unmarried men. However, the Selectivity Hypothesis is silent regarding the link between personal attributes and a marriage premium for women. If high earnings or financial security increases the value of men in marriage market, then women with high earnings or with higher productivity potential are likely to defer to marry or prefer not to marry. Women who are able to support themselves through their own work (or through welfare benefits) will feel less pressure to marry for economic need and may choose not to marry or to form families through cohabitations or non-marital childbearing (Cherlin, 1992 cited in White and Rogers, 2000). They put education advancement and career development in the forefront in order to establish themselves firmly in a career (Lauer and Lauer, 2011) Productivity Hypothesis The productivity hypothesis is based on the role of traditional household specialization or division of labor by sex where men are assumed to join the workforce whereas women are assumed to go into domestic labor. Therefore, men are regarded as more productive in the labor market as they spend more time on their career and labor market goals (Chun and Lee, 2001). Furthermore, married men compared to unmarried or single men have more commitment to their jobs, they are seldom fired and frequently promoted, and in addition 5

13 receive a larger share of the profits distributed according to individual performance (Becker, 1985). Becker (1985) states that men have a competitive advantage in the labor market as women are in-charge of household responsibilities as a result of the division of labor. In this regard, the unmarried or single men have to engage in both the labor market and household work, which causes them to exert more time and energy. Waite (1995) also suggests that married men are likely to benefit from both economic and social benefits such as less alcohol abuse, less stress, longer life expectancy, and higher income and wealth. However, does marriage increase the productivity of women? The existing literature on productivity hypothesis is less responsive to this question. Some empirical studies provide some explanations to this question. Brown and Dickman (2010) in a study of emerging adults work and family commitments reported that college men and women are equally committed to work and family. Similarly Stickney and Konrad (2007) found that women with egalitarian attitudes (career-oriented and independent) have significantly higher earnings than women with traditional attitudes (family oriented and dependent on their men). Despite the increasing labor force attachment of women and increasing egalitarian gender role attitudes, marriage and/or motherhood has significant impacts on productivity and therefore on earnings of women. According to Treas and Widmer (2000), married women are likely to prefer to stay at home or work part-time once they have preschool children and they prefer to go back to full time work only after the children leave home. This discontinuation at work and preference for part-time work may reduce the productivity of married women (Corell, Benard, and Paik, 2007; Peterson and Morgan, 1995). The literature on the motherhood penalty also indicates that mothers looking for employment are less likely to be hired, are offered lower salaries and are perceived as being less committed to a job than fathers or women without children (Correll, Benard, and Paik, 6

14 2007). The motherhood penalty is often linked with decreased productivity of working mothers (Budig and England, 2001; Correll, Benard, and Paik, 2007). Working pregnant women face further disadvantage regarding promotions and discrimination may lead to be fired (Byron, 2010). These circumstances suggest that pregnant women and mothers tend to have lower earnings than non-mothers. In sum, the above explanations of individual models state that individual level characteristics, like level of education and training, work experience and behavioral traits are the factors influencing earnings inequality between married and unmarried women. Unmarried or single women focus on education and career, so have higher human capital whereas married women who may need to give time for family, and working mothers may have to face a motherhood penalty. Therefore, unmarried (non-mothers) women are likely to have higher earnings than married women and working mothers. 2.2 Structural Approach The structural approach assumes that individuals are shaped by the bigger structures of society. Structural theories then, focus on the interrelationships between the larger social structures or institutions of the society, and how these structures and institutions affect individuals in the society (Ritzer and Goodman, 2004). However, in the structural approach, the individual is not the subject matter of analysis in both research and theory construction. Thus the structural approach argues that earnings inequality is a structural phenomenon, and it is determined by organizations and organizational structure. The job positions in organizational structure are based on an organizational hierarchy with owners at the top level, managers at the middle level, and workers at the bottom or floor level. In this structural setting people in higher levels/positions receive higher wages than people in lower levels/positions. Furthermore, Coverdill (1988) states that wages are affected by the structure 7

15 of the market where the company is operating. There are two models of the structural approach that explain the wage difference Dual Economy Theory and Segmented Labor Market Theory, which are explained below Dual Economy Theory The dual economy theory assumes that the economy is not homogenous and, therefore, can be divided into a monopoly sector and a competitive sector (Gordon, Edwards, and Reich, 1982). Sorting of a particular firm in either of the categories depends upon the nature of business, size of the firm, industrial location, and market concentration (Tolbert, Horan, and Beck, 1980). In the monopoly sector or concentrated markets, the company will have high profit, therefore employees in the monopoly sector earn higher wages, have better benefits, more opportunities for mobility, and greater work satisfaction than employees in the competitive sector (Reid and Rubin, 2003). In addition, the monopoly sector requires a stable and trainable workforce, which means education and work experience are the important aspects of gaining entry into the monopoly sector (Coverdill, 1988). In contrast, the competitive market contains small firms with limited markets, low wages, little or no training and skills, minimal job security, and limited career development opportunities (Reid and Rubin, 2003). Hodson (1983) further states that monopoly firms have a higher rate of unionization than competitive firms which may lead to higher wages and greater benefits being provided to workers. Therefore, the gender wage gap is the result of the disproportionate allocation of women to peripheral jobs (Coverdill, 1988). Linking the above theoretical explanations with the marriage premium, as married men are preferred by employers, married men are more likely to be attracted to and employed by the monopoly sector than unmarried men. The attraction of married men in the monopoly sector is also linked with prestige, higher pay and benefits. According to Coverdill (1988), 8

16 monopoly markets may have some institutional barriers created through job specification, recruitment process, career development, etc. In contrast in competitive markets, these institutional barriers are less. Therefore women get more job opportunities in the competitive markets than in the monopoly markets. The dual economic theory also suggests that the greater labor force attachment of women in competitive markets is the result of discrimination and gender constraints, not because women have lower human capital (Coverdill, 1988). In addition, with respect to the labor force attachment of married and unmarried women, unmarried women are more likely to get jobs in the monopoly sector than the married women because of discrimination and institutional constraints in the labor markets (Coverdill, 1988). According to O'Connor (O Connor, 1973, cited in Coverdill, 1988), the supply of labor in competitive industries is inflated by workers such as married women, students, and retired workers who want, and will accept lower wages to obtain, irregular work. Because married women are generally involved much more deeply than unmarried women in family activities such as childcare and housework, they may be willing or be forced to forego the wage advantages of monopoly sector workplaces in order to work close to home, have flexible work hours, and so on (Coverdill, 1988) Segmented Labor Market Theory Another structural explanation for earnings inequality is the Segmented Labor Market (SLM) theory. The theory states that there are different job markets and different professionals work in these different job markets. These different job markets are often segmented based on occupation, geography and nature of the industry. The occupational labor markets arise from the division of labor, increasing differentiation and specialization. Since each occupational labor market requires specific skills and knowledge, the workers are 9

17 less likely to switch into another occupational labor market. It also applies in geographic market segments and industry-wise market segments. Therefore, this theory suggests that a wage is directly related to professions and positions in the labor market, not to the workers attributes (Weitzman, 1989). The employees in so called white-collar professions and whitecollar positions are likely to have higher earnings than employees in so called blue-collar professions and blue-collar positions. This theory further segregates labor markets into primary labor markets and secondary labor markets. The jobs in primary labor market are characterized by higher wages, better working conditions, more stable employment, and higher return to human capital (Weitzman, 1989). However, importantly, the SLM theory states that earnings in the secondary labor market, unlike in the primary market, are not related to productivity (Boston, 1990). The empirical literature shows that pay differs significantly for different occupations. The white collar occupations or positions pay comparatively higher wages than the blue collar occupations or positions. The gender wage gap is often attributed to this occupational segregation (Petersen and Morgan, 1995).The occupations that have more female employees are lower paid than the occupations that have more male employees (Baron and Newman, 1990). In other words, women have disproportionately worked in occupations such as teachers, nurses, secretaries, and retail sales clerks that pay relatively low wages and men have disproportionately worked in occupations such as executives, managers, doctors, lawyers, engineers, and scientists that pay comparatively high wages (CONSAD, 2009). In addition, within the white collar professions there are gender wage gaps. The empirical literature on white-collar professions such as lawyers (Noonan, Cocoran, and Courant 2005), physicians (Boulis and Jacobs 2003), scientists (Prokos and Padavic 2005), financial professionals on Wall Street (Roth 2003), and faculty in higher education (Toutkoushian 10

18 1998) indicate that women are earning less than their male counterparts. Furthermore, there is a wage gap between women working in white-collar jobs and blue-collar jobs. The size of this disparity on women s earning is higher than the disparity that exists between men working in white-collar and blue-collar jobs (Petersen and Morgan, 1995). The disparity emerges quickly with a small gender wage gap among college graduates then widens over time as women s professional careers progress (Peterson and Morgan, 1995). The structural explanations of segmented labor theory are silent about the effect of marriage on women s earnings. But as stated by Boston (1990), having never been married is one of the most significant factors in determining the likelihood of upward mobility from blue-collar to white-collar positions and professions because never married women are assumed to be more committed to their work and have less job turnover. Therefore, the unmarried women may have higher earnings than the married women. To sum up, the structural models suggest that occupational level and type, market structure and labor market conditions are responsible for earning differences. Unmarried women are likely to enter into white-collar professions and get jobs in monopoly markets whereas married women are likely to enter into blue-collar profession and get jobs in the competitive market. These disparities in labor markets and occupational segregations result in higher earnings for unmarried women than married women. 2.3 Gender Model In addition to individual and structural approaches, gender theories also explain the earnings inequality between men and women. The gender models are based on gender discrimination at work settings and gender role attitudes in family settings. Regarding gender discrimination within work settings, gender models state that the gender wage gap is caused by discrimination in several processes. Petersen and Morgan 11

19 (1995) outline three major forms of discrimination that cause the gender wage gap. First, women are differently allocated to occupations and establishments through differential access to job markets which is often termed as allocative discrimination. Women face limited access to attractive positions within a work organization at the time of entry and/or at the time of promotion (Hultin and Szulkin, 1999). Women receive less preference for high pay and high ranking positions (Beggs, 2001). Allocative discrimination thus denotes unequal treatment of women in decisions about recruitment and promotion that in turn lead to women generally being employed in occupations, establishments, and jobs with relatively low earnings levels (Hultin and Szulkin, 1999, pg. 455). Among working women, married women and working mothers face more allocative discrimination than unmarried and or nonmothers. Second, the occupations that are primarily held by women are paid lower wages than the similar skilled required occupations that are primarily held by men. This type of discrimination is called valuative discrimination or evaluative discrimination. Although the value of a job should be assessed by the different aspects of the work content that are of importance in the wage-setting process such as demands of qualifications and responsibilities, women are devalued at work by employers or perspective employers assuming women are less productive and less committed than men (Hultin and Szulkin, 1999; South and Spitze, 1994). Therefore, women earn less than men. In addition, the motherhood penalty in terms of pay and promotion is an established example of discrimination and institutional constraints responsible for earnings gap between working mothers and non-mothers. Third, women get lower wages than men within the same occupation and within the same establishment. This form of discrimination is called within-job wage discrimination. Although this kind of discrimination at work is legally prohibited, the increasing trend of 12

20 discrimination charge filings over last decade (EEOC, 2009, cited in Curtis, 2010) indicates the presence of within-job wage discrimination at work. There is lack of transparency on the recruitment and promotion process. The literature suggests that pregnant women and working mothers face even more within-job wage discrimination than single and nonmothers (EEOC, 2009, cited in Curtis, 2010). Besides the above explanations, in a family setting, gender role attitudes play an important role in women s participation in the labor force. Traditionally, women are responsible for household work whereas men are responsible for external affairs like jobs and earnings. These family expectations influence the choices women can make. This separation of women from the labor force systematically decreases women s earnings (South and Spitze, 1994). Women s greater participation in household work than in labor force attachment is an important factor in the gender wage gap (Shelton and Firestone, 1989). However, recent studies show increasing egalitarian gender role attitudes in the family settings and women have an increasing trend of career orientation and labor force attachment (Stickney and Konrad, 2007). Gender theories can also be used to explain the earning differences between married and never married women. Corell, Benard, and Paik (2007) state that mothers experience disadvantages in the workplace in addition to those commonly associated with gender. These women engage more in household work than non-mother women because of childrearing responsibilities. Furthermore, mothers need to give more time to their children, and young children (under 6) demand more time from mothers than older children. Literature suggests that a motherhood wage gap exists and factors such as reduced investment in human capital by mothers, lower work effort by mothers, and discrimination against mothers by employers are responsible for lower earnings of mothers compared with non-mothers (Budig and 13

21 England, 2001; Corell, Benard, and Paik, 2007). In contrast, single non-mother women and/or unmarried women can continue their job and can further invest in human capital for career growth. Never married women are often career oriented. The continuation at work further increases the productivity. These gender based attributes, functions and responsibilities may, therefore, lead to lower pay for married women/working mothers than unmarried or single women without children. In addition to gender discrimination, racial discrimination at work is another important factor for understanding gender earnings inequality (Avery, 2003). The forms of gender discrimination are also applicable in racial discrimination. Traditionally, disadvantaged groups such as Blacks and Hispanics have lesser wages than their White counterparts. Previous studies suggest that Whites have higher earnings than minorities (Blinder 1973; Blau and Kahn, 2002). This differential in earnings is often linked with racial differences in human capital accumulation (Darity, 1982). For example, as described by human capital theories, earnings are tied to human capital acquired by an individual. According to this argument, Blacks tend to acquire or accumulate less human capital therefore they are paid less (Darity, 1982).However, Blacks and Hispanics have less access to accumulate human capital because of socio-economic circumstances. Even if Blacks and Hispanics have competitive human capital than Whites and Asians, Black and Hispanics have less access to better jobs and face discrimination in pay and promotion. Gloria and Hird (1999) found that White students have significantly higher career success than racial and ethnic minority students. Furthermore, Alon and Haberfeld (2007) state that there is a persistent racial and ethnic wage gap among women. These wage gaps are even larger among women with no college education. The effect of marriage on earnings differs for racial groups. Surprisingly, Waite (1995) states that Black women enjoy a marriage premium of about three percent 14

22 whereas White women experience a penalty of about four percent. In sum, assessing gender and racial discriminations is important to understanding income inequality. 2.4 A Composite Model and Research Hypotheses The individual models state that individual characteristics such as education and skill, work experience, and behavioral traits (honesty, commitment, etc.) are important factors explaining wage differences whereas structural models state that structural variables like job position (rank), occupation, and market structure are important factors to explain this wage gap. The gender models further state that the bias towards particular gender, race or ethnicity, and marital status affect wage differences. Since these models individually are not sufficient to fully explain the wage gap between married and unmarried women, this thesis attempts to synthesize individual models, structural models and gender models, and proposes an alternative model for earnings inequality. The composite model presented in Figure 1 below shows that individual level characteristics like level of education, and work experience determine the level of earnings of individuals. This relationship is further affected by gender issues like race, ethnicity, marital status, etc. as explained by gender theories. The structural variables like occupation level, industry, etc. determine the level of earnings as explained by structural theories. The structural variables may also influence the relationship between individual level characteristics and level of earnings. Furthermore, as explained by gender theories, gender issues also influence the structural variables thereby the level of earning. As explained by individual theories, individuals having higher education qualification, and more work experience will have higher earnings than individuals with lower education qualification and less work experience. Similarly, as explained by structural theories, individuals working in a higher position in the job hierarchy and working in white-collar 15

23 professions will have higher earnings. Furthermore, as stated by gender theories, unmarried women will attain higher job positions in organizational hierarchy. When mapping these relationships, it is hypothesized that unmarried women will be found to invest more in human capital, be sorted into higher positions and white-collar professions, and have less household responsibilities. Therefore, unmarried or single women will have higher earnings than married women. Figure 1: A Composite Model Independent Variables Dependent Variable Individual level variables Education Work experience Structural variables Job position Job occupation Wage gap (women) Gender/Race variables Gender Race Marital status The three sets of hypotheses are formulated as follows and will be tested in this thesis. Hypothesis 1a: Net of other factors, with an increase in level of education, there will be an increase in earnings. Hypothesis 1b: Net of other factors, with an increase in number of years at work (measured by age), there will be an increase in earnings. Hypothesis 2a: Net of other factors, women in white-collar positions (in organizational structure) will have higher earnings than that of blue-collar positions. 16

24 Hypothesis 2b: Net of other factors, women in high-skill occupations will have higher earnings than that of low-skill occupations. Hypothesis 2c: Net of other factors women working in large size organizations will earn more than women working in small size organizations. Hypothesis 2d: Net of other factors women working in government sector will earn more than women working in private sector. Hypothesis 2e: Net of other factors, women in the Midwest region will earn less than women in other regions. Hypothesis 2f: Net of other factors, women working full time full year will earn higher than women working part time or part year. Hypothesis 3a: Net of other factors, unmarried women will have higher earnings than married women. Hypothesis 3b: Net of other factors, women without a child under 6 will have higher earnings than women with a child under 6. Hypothesis 3c: Net of other factors, minority women will earn less than White and Asian women. 3 Research Methodology 3.1 Data and Sample This study uses the Current Population Survey (CPS) 2010 data set (Bureau of Census, 2010). The CPS is a random, representative sample survey of the American population conducted by US Bureau of Census for the Bureau of Labor Statistics. This survey mainly focuses on factors such as individuals, families and households, employment, salary/wages, occupations, working hours, education level, etc. The CPS is comprised of a nationally 17

25 representative sample. There are a total of 158,879 observations in the CPS data set. However, as per the objective of the study some sample restrictions are being imposed. The sample restrictions imposed are as follows: only individual level data is selected; only the female respondents (age range 18 through 64) are selected and responses from noncivilians and military spouses are excluded; only the female respondents who are currently married and never married are selected, and, only those respondents with annual earnings greater than $258 and less than $200,000 are selected. To make the sample more representative of the true population, a relative weight has been applied to all cases. The relative weight index is created by dividing standard weight by mean of standard weight. Therefore the final sample consists of 32,862 cases comprising percent married women and about percent unmarried women. Of total cases, cases (35.7 percent) are from the South region, 7716 cases (23.5 percent) from the Midwest region, 7243 cases (22.1 percent) from the West region, and 6169 cases (18.8 percent) from the Northeast region. The regional sample distribution is similar to the total population distribution. 3.2 Variables Based on the literature surveyed, data available in the CPS data set, and the scope of the study the following variables are used in this study Dependent Variable The dependent variable of this study is annual income or earnings which is measured by total annual wages (as available in CPS data set) in raw U.S. dollars. This is an interval level variable with values ranging from $258 to $200,000, exclusive. 18

26 3.2.2 Independent Variables To capture the theoretical explanations posited by three major theoretical models (individual models, structural models, and gender models), three sets of independent variables are used in this study which are explained below. Individual level variables: There are three individual level variables level of education attainment, age, and agecentered which are considered in this study. The CPS survey originally has 17 categories for the level of education attainment variable. For convenience, this variable is grouped into 5 categories ( less than high school diploma, high school diploma or equivalent, some college degree, college/bachelor degree, and advance degree ) in ordinal level. Another individual level variable is age which is a proxy measure for work experience. This variable is a ratio level variable value ranging from 18 to 64 inclusive. The sample restriction on this variable is guided from general understating about the age for entering into the formal labor force and retiring from the labor force. Furthermore, as suggested by the literature, there is a non-linear relationship between age and earnings. Therefore a new Agecentered variable is created to capture this non-linear effect by standardizing and squaring the variable age. Structural level variables: There are five major structural variables used in this study namely size of the business, level of occupations (4 levels occupations), working status (full time/part time), employer s type (government/private), and geographic region. The size of business is represented as the number of workers in the company. For methodological purpose, the size of business is divided into small business (employees 1-99), medium business (employees ), and large business (employees above 500). To capture the effect of size of business on earnings, two dummy variables are created as medium business and large business considering small business as a reference category, that is, 19

27 medium business takes the value 1 if size of business is medium otherwise takes a value of 0. Similarly, in the large business dummy variable, large business takes the value 1 if the size of business is large otherwise takes a value of 0. To capture the effect of 4 levels of occupation, three dummy variables are created considering Blue collar low skill job as a reference category. The first dummy variable White collar high skill takes value 1 if the level of occupation is white-collar high skill, otherwise takes value 0. Similarly another dummy variable White collar low skill takes value 1 if the level of occupation is white-collar low skill, otherwise takes value 0. The next dummy variable Blue collar high skill takes value 1 if the level of occupation is blue-collar high skill, otherwise takes value 0. Similarly, for the working status of women, a dummy variable is created that takes value 1 if she is working full-time full-year and takes value 0 otherwise. For employers type, a dummy variable is created that takes value 1 if respondent is working in government organization (federal, state, and local) and takes value 0 if working in private sector. Finally, to capture the effect of geographic region on earnings, a dummy variable is created which takes value 1 if respondent is from the Midwest region and takes value 0 otherwise. Gender/Race variables: As explained by gender theories, gender/race variables such as gender, marital status, having children, and race and ethnicity are important factors explaining earnings inequality. There are three gender/race variables considered in this study. They are marital status of women, having a child age under 6, and race/ethnicity. A dummy variable that takes value 0 if the woman is never married and value 1 if she is married. Another dummy variable is created that takes value 0 if a woman doesn t have a child age under 6, and value 1 if a woman has children age under 6. For race/ethnicity, the available race/ethnicity variable is transformed into minority and nonminority (binary) based dummy 20

28 variable. The Whites and Asians are considered as non-minority and Blacks, Hispanic and Others are considered as minority. This dummy variable takes value 0 if respondent is White and/or Asian and takes value 1 otherwise. The motive behind putting White and Asian in the same reference category is both of these racial categories have similar earnings. 3.3 Methods of Analysis In this study the analysis of data is undertaken at three stages. First univariate analyses are performed where frequency, mean, median and standard deviation (where applicable) are computed and analyzed. At the next stage, bivariate analyses are performed between/among dependent and independent variables and the bivariate relationships are tested by t-tests, and ANOVA tests. At the final stage the multivariate analyses are performed developing three multiple regression models: total model, unmarried women only model and married women only model'. Ordinary Least Square (OLS) method is used to estimate the parameters. 4. Results and Discussion 4.1 Univariate and Bivariate Results Table 1 and Table 2 present univariate and bivariate results. The average annual income of women in the sample is about $ with standard deviation of $ The average annual income of married women is $ and unmarried women is $ The t-test for the mean difference of annual income of married and unmarried women has t- value of (p<0.001, df=26103). This suggests that the null hypothesis of equality of mean of annual income of married and unmarried women is rejected at level of significance. Therefore, married women have significantly higher earnings than unmarried women. 21

29 Within individual level factors, about two-thirds women in the sample have less than a bachelor s degree and about 23 percent have bachelor s degree and only about 11 percent have advance degrees. This pattern is similar within married and unmarried sub-sample groups. The average annual income of women with less than a high school diploma is $15381, with a high school diploma is $22569, with some college is $25180, with a bachelor s degree is $37354, and with an advanced degree is $ The One-way ANOVA test results for the equality of average annual income with different levels of education attainment suggest that level of annual earnings differs significantly across the different levels of education attainment (F =2476.8, p<0.001, df= (4; 32856)). Furthermore, posthoc analysis (Scheffe test) indicates that there are no homogeneous subsets. The results suggest that level of income increases as level of education attainment increases. The average age of women, in sample, is about 38.1 years with standard deviation of 12.7 years. The average age of unmarried women is 29.4 years whereas the average age of married women 42.3 years. Therefore, unmarried women are younger than married women, and therefore, married women could potentially have greater work experience than unmarried women (t= , p<0.001, df=24208). Based on four levels of occupations, about 43.9 percent of women are working in white-collar high-skill occupations and about 33.3 percent are working in white-collar low-skill occupations. Therefore, about 77.2 (= ) percent of women are working in white-collar level occupations, that is, the majority of working women are working in white-collar level occupations. There are very few, only about 3.2 percent of women working in blue-collar high-skill level and about 19.6 percent women are working in bluecollar low-skill level occupations. The pattern of distribution among married women is similar to the full sample. However, in the case of unmarried women, about 35 percent work 22

30 in white-collar high-skill level occupations and about 26 percent women work in bluecollar low-skill level occupations. The average earnings of women working in White-collar high skill occupations is $39354, White-collar low skill occupations is $25046, Blue-collar high skill occupations is $27402, and Blue-collar low skill is $ The F statistic (ANOVA test) value is (p<0.001, df = (3, 2849)) which indicates that the average earnings of women are not equal for different levels of occupation. The posthoc analysis results indicate that the earnings are significantly different across the different levels of occupations. Women working in White-collar high skill occupations earn the highest income whereas women working in Blue-collar low skill occupations earn the lowest income. Interestingly, women working in Blue-collar high skill occupations earn higher income than women working in White-collar low skill occupations. The majority of women in the sample are working in large sized businesses in the sample. About 51.7 percent are working in large sized organizations. About 34.4 percent women are working in small sized organizations, with only about 13.9 percent are working in medium sized business. The pattern is similar within sample sub-sets. The average earnings of women for small business, medium business and large business categories are $23683, $31403 and $33049 respectively. The F-test for the equality of average annual earnings of women across the size of business they employed has F-statistic (p<0.001, df= (2, 32858)). The significant F-value suggests that the average annual earnings of women are not equal across size of businesses. The posthoc analysis suggests that the women s earnings vary by the size of business they work for. Specifically, the annual earnings of women increase as the size of business increases. 23

31 On an average women work about 38 hours per week with a standard deviation of 10.4 hours. However some women work only about one hour per week and some women work about 99 hours per week. About 60.4 percent women are working in a full time job throughout the year whereas about 15.4 percent women are working in a part time job throughout the year. About 12.6 percent of women work full time job partly throughout the year and about 11.5 percent of women work a part time job partly throughout the year. The first two categories suggest the continuous labor force attachment of women whereas the latter two categories may suggest discontinuity of women in the labor force. Therefore, about 75.8 (= ) percent of women are attached to the labor force throughout the year. The Chi-square test (χ 2 = , p<0.001, df = 3) suggests that the labor force attachment of women is related to marital status of women. About 54 percent of unmarried women work full time jobs throughout the year whereas about 60 percent married women work full time jobs throughout the year. Similarly about 17 percent of unmarried women work part time jobs throughout the year whereas about 15 percent of married women work part time jobs throughout the year. Within structural model factors, as shown in Table 2, there is earnings inequality between women living in Midwest region and other regions (Northeast, South, and West). The t-test of equality of average income of women between Midwest region and other regions suggests that women in the Midwest region have lower earnings than women in other regions. The t-value is 5.72 (p<0.001, df=32860). However, the Cohen s d suggests that the t-test result of significant difference is not meaningful. In the sample, about 80 percent of women are working in private sector jobs whereas about 20 (= ) percent of working women are working in government sectors. The t-test result (t = , p<0.001, df=32280) for the equality of women s earnings based 24

32 on the employers type (private sector and government sector) suggests that women working in government sector jobs earn more than women working in the private sector. The annual wages for women working in government sector is $ whereas for women working in private sector is $ For the gender/race model, about 80 percent of working women in the sample do not have children under 6. The average annual income of women with a child under 6 is about $28087 and women who do not have a child under 6 is about $29970 (t = 6.44, p<0.001, df=32860). The t-test results indicate that women without a child have a higher annual income than women with a child. However, in both cases Cohen s d values suggest that effect size is small. Therefore, this statistically significant difference is not meaningful. Finally, the racial distribution of women in the sample is categorized into five categories. About 68.3 percent of women in the sample are White non-hispanic, about 12.7 percent are Hispanic, about 12.2 percent are Black non-hispanic and only about 4.8 percent are Asian. There are also about 2 percent who are identified as Other non- Hispanic women in the sample. When combined, there are about 73 percent White, and Asian women and about 27 percent minority women in the sample. The t-test shows that the annual income between minority women and non-minority (White and Asian women) is not equal. The t-value is 26.7 (p<0.001, df = 32860). That means, minority women have significantly lower annual earnings than non-minority (White and Asian) women. The average annual income of minority women is $ whereas the average annual income of non-minority women is $