Impact of family commitment on gender wage discrimination measurement

(1)

Impact of Family Commitment on Gender Wage

Discrimination Measurement

Bachelor Thesis

Qiuhong Piao, 10256342

Thesis supervisor: Jindi Zheng, PhD Date: 1st July 2014

University of Amsterdam

Faculty of Economics and Business

BSc Economics and Business

(2)

Abstract

This paper offers a detail explanation of how gender wage discrimination could be measured. The paper analyzes Altonji, J. G., & Blank, R. M.’s research on gender wage discrimination (1999) and sets their research paper as the main literature review. The aim of this paper is to expand the determination of gender wage discrimination by adding a new socioeconomic characteristic – family commitment, as the additional control variable. The empirical results of this paper show that family commitment does play a role in explaining the gender wage gap and hence affecting the gender wage discrimination. However, the impact of family commitment is not very notable.

Keywords: Gender Wage Discrimination; Family Commitment; Blinder-Oaxaca

decomposition

Acknowledgement

I would like to thank Professor Joseph Altonji for assisting me with the data, Professor Marcel Boumans for offering me valuable suggestions and comments. Last but not least, I would like to thank my thesis supervisor Ms. Jindi Zheng who always gave me prompt replies, guidance, and encouragement.

(3)

discrimination at work. Economists often take Blinder-Oaxaca decomposition to measure discrimination. In the Oaxaca sense, discrimination accounts for the most of the gender wage gap as the controlled variables only account for a small percentage of wage gap (Blinder, 1973; Oaxaca, 1973). In this case, as well as most of the researches on wage discrimination at work, discrimination was measured depend on the set of variables that are observable and relatively direct. For instance, Nielsen (2010) in his paper “Wage discrimination in Zambia: an extension of the Oaxaca-Blinder decomposition” measured wage discrimination by setting year of experience, education, and full-time work as controlled variables. Altonji, J. G., & Blank, R. M. (1999) examined the wage differential between men and women after controlling the differences in socioeconomic characteristics between men and women. Education attainment, age, region of residence, labor force, occupation, and industry are set as the controlled variables/socioeconomic characteristics in their paper. However, it is hardly convincing that the controlled variables used in Altonji & Blank’s research are complete. There might be other relevant variables which will contribute to explain the gender wage gap. This paper aims to conduct a research on one of the unobservable variables and analyze its impact on the traditional discrimination measurement and hence the former conclusion.

Family commitment, more specifically in the terms of commitment to child rising during the formative years, might be a factor that will explain the gender wage gap.

(5)

According to Hotchkiss & Pitts (2005), family commitment can be differ by gender, and have different impact on the labor force outcome for males and females. For instance, females are subject to motherhood penalty while partners of those females receive fatherhood

compensation. Furthermore, mothers seem to dedicate/sacrifice more percentage of their working years to raise their children than the fathers do. This shows that family commitment might be a socioeconomic characteristic that varies between men and women.

This paper adopts Altonji & Blank’s regressions, data sourcing from Bureau of Labor Statistics and tabulating from the Current Population Survey (CPS) in the United States. It re-exam the United States gender wage gap in 2008 by running Blinder-Oaxaca decomposition model with six controlled gender wage predictors as Altonji & Blank did in their paper. Then, this paper will run the model again but with the added control variable – family commitment. This offers a horizontal comparison about the relationship between controlled variables and gender wage discrimination. Finally, this paper uses the Oaxaca command in Stata to run the decomposition, and collects and presents the results in the format as Altonji & Blank did in their papers.

The results of the empirical research show that there is an impact of family commitment on explaining the gender wage gap. Consequently, the percentage of gender wage differential ascribed to discrimination decreases. However, the impact is not so significant. Further research needs to be done by adding more control variables or by using other relevant models to further investigate labor discrimination.

The second section of this paper analyses the existing literatures. The third section offers a description of the data, empirical method, results and followed by a discussion of the empirical results. Section four will draw a conclusion.

(6)

II. Literature Review

This literature review contains general information about the Blinder-Oaxaca decomposition for linear regression models, the theoretical reasons behind the model which can use to elaborate gender wage discrimination. An analysis of the former researches done about this model will also appear in this section.

II.I Model Framework

Based on the linear model

𝑌𝑖 = 𝑋!_{𝑖𝛽𝑖 + 𝜖𝑖, 𝐸 𝜖𝑖 = 0 𝛾 ∈ 𝓂, 𝒻 (1)}

where Y is the outcome variable and X is a vector containing a set of predictors1 and a constant, 𝛽 represents the slope parameters and the intercept, and 𝜖 is the error term. 𝓂 represents group male while 𝒻 represents group female. Hence, by submitting equation (1), the difference of the mean outcome between the two groups (R) can be represented as

𝑅 = 𝐸 𝑌𝓂 − E Y𝒻 = 𝐸 𝑋𝓂 !_{β𝓂 − E X𝒻} !_{𝛽𝒻 (2)}

where E(𝛽𝛾) = 𝛽𝛾 and 𝐸 𝜖𝛾 = 0 by assumption.

Then rearrange equation (2) according to Jones and Kelley (1984, pp. 325-327), the decomposition equation represents as

𝑅 = 𝐸 𝑋𝓂 − E X𝒻 !_{𝛽𝒻 + 𝐸 X𝒻} ! _{β𝓂 − 𝛽𝒻 + 𝐸 𝑋𝓂 − E X𝒻} ! _{β𝓂 − 𝛽𝒻 (3)}

Equation (3) represents a “threefold” decomposition (Jann, 2008, p. 454) that R could split into three main components,

𝑅 = 𝐸 + 𝐶 + 𝐼 (3!₎ where

𝐸 = 𝐸 𝑋𝓂 − E X𝒻 !_𝛽𝒻;

C= 𝐸 X𝒻 ! _{β𝓂 − 𝛽𝒻 ;}

(7)

𝐼 = 𝐸 𝑋𝓂 − E X𝒻 ! _{β𝓂 − 𝛽𝒻}

Jann (2008, p. 455) also showed an alternative decomposition by adding a nondiscriminatory coefficient vector 𝛽∗_{, and equation (2) can be rearranged as}

𝑅 = 𝐸 𝑋𝓂 − E X𝒻 !_𝛽∗_{+ 𝐸 𝑋𝓂} ! _{β𝓂 − 𝛽}∗ _{+ E X𝒻} ! _𝛽∗_{− 𝛽𝒻 (4)}

Equation (4) is called “twofold” decomposition (Neumark, 1988; Oaxaca, 1988), that 𝑅 = 𝑄 + 𝑈 4!

where

𝑄 = 𝐸 𝑋𝓂 − E X𝒻 !_𝛽∗_;

𝑈 = 𝐸 𝑋𝓂 ! _{β𝓂 − 𝛽}∗ _{+ E X𝒻} ! _𝛽∗_{− 𝛽𝒻}

“Threefold” decomposition and “twofold” decomposition would be further explained in section II.II. This paper will apply “threefold” decomposition and “twofold”

decomposition as two different methods to test the relationship between family commitment and gender wage discrimination.

II.II Theoretical Reasons behind Blinder-Oaxaca Model

Blinder-Oaxaca decomposition (Blinder, 1973; Oaxaca, 1973) is often used by economists to measure discrimination. The outcome variable Y described in section II.I is (log) wage, and R measures the mean outcome difference between male and female, i.e. the gender wage

differential. The three components in the “threefold” decomposition 2 respectively represent

“endowments effect” (𝐸 = 𝐸 𝑋𝓂 − E X𝒻 !_{𝛽𝒻), “differences in the coefficient”}

(C= 𝐸 X𝒻 ! _{β𝓂 − 𝛽𝒻 ), and interaction term, which shows that correlation between}

𝐸 𝑋𝓂 and E X𝒻 ,(𝐼 = 𝐸 𝑋𝓂 − E X𝒻 ! _{β𝓂 − 𝛽𝒻 ).}

Among the three components, “differences in the coefficient” (C) is mostly used to measure differentials due to gender wage discrimination. This component quantifies change in female’s wage when adopting male’s coefficients to female’s characteristics (Jann, 2008, p.

(8)

469), and the existence of such change shows that there is discrimination. The component represents “endowments effect” measures the differential in female-male wage due to differences in predictors in the two groups. In other words, it reflects the mean change in female’s wage if they would have the same characteristics as male. The interaction term, the last component, accounts for the simultaneous effect of differences in endowments and coefficients (Jann, 2008, p. 455).

The two components in the “twofold” decomposition 3 respectively represent

“quantity effect” (𝑄 = 𝐸 𝑋𝓂 − E X𝒻 !_𝛽∗_{), and “unexplained part”}

(𝑈 = 𝐸 𝑋𝓂 ! _{β𝓂 − 𝛽}∗ _{+ E X𝒻} ! _𝛽∗_{− 𝛽𝒻 ). The “unexplained part” is normally}

ascribed to discrimination and the “quantity effect” accounts the differences in endowments4. The “threefold” and “twofold” decomposition methods contribute in revealing the theoretical reasons behind the Blinder-Oaxaca model. That is, decomposing the wage gap into parts that are due to the differences in the characteristics between female and male and a remaining unexplained part which is called discrimination. To put in a different way, if the wage gap could be wholly explained by the differences in the characteristics, then there is no gender wage discrimination.

II.III Prior Research about Gender Wage Discrimination

Altonji, J. G., & Blank, R. M. (1999) investigated wage differentials by race and gender in paper “ Race and Gender in the Labor Market”. They adopted Blinder-Oaxaca model and decomposed the wage differential into “explained” and “unexplained” components by

controlling a set of predictors/ socioeconomic characteristics. The “unexplained” component, which is also the differences in the estimated coefficient, is referred to as the “share due to discrimination” (Altonji, J. G., & Blank, R. M., 1999, p. 3156). The set of predictors that Altonji & Blank controlled in their regressions include education attainment, age, region of residence, labor force, occupation, and industry. The result they obtained is summarized in Table 1.

3_{Refer to equation (4) and (4)’.}

(9)

Table 1

Decomposition of gender labor force participation differentialsa

Females vs Males

1995

Labor force participation difference -0.156

Amount due to

Characteristics -0.008

Coefficients -0.148

a_{Source: Altonji, J. G., & Blank, R. M. (1999).“ Race and Gender in the Labor Market,” in Orley Ashenfelter} and David Card, editors, Handbook of Labor Economics, vol. 3C, 3163 Table 7.

According to results from Table 1, there is a female - male wage gap of 15.6%,

indicating females in 1995 earn 15.6% less than males in the United States. Among the 15.6% wage differential, there is only 0.8% due to characteristics, which means socioeconomic

characteristics only accounts for 5.128%5 of the wage differential. The rest of the wage differential (94.872%) is due to the “unexplained” component which is ascribed to

discrimination. This shows that in 1995, gender wage discrimination is largely account for the gender wage differential in the United States.

However, Altonji & Blank did highlight in their paper that their terminology might be “misleading” (Altonji, J. G., & Blank, R. M., 1999, p. 3156), this is because there must be some important control variables are omitted that will affect gender wage gap and hence affect the results. Besides, Altonji & Blank (1999) also pointed out areas that further research could conduct at which includes expanding current models of gender wage discrimination in terms of time, focusing on other groups other than only on black and white or on males and females, and widening research by collecting data at a sub-national level.

5_{5.128% =} !.!%

(10)

III. Empirical Evidence

This section offers a description of how data was obtained, the empirical method of this paper, and the results of the empirical research. Also, an elaboration and discussion about the results would be included in this section.

III.I Data Description

This paper adopts Altonji & Blank’s regressions, sourcing from Bureau of Labor Statistics and tabulating from the Current Population Survey (CPS) in the United States. The CPS was conducted in March, 2008. After dropping all observations where log wage is less than or equal to 0, the final sample reduced from about 126,127 to 86,932 individuals. Among the 86,932 individuals, there are 44,260 males and 42,672 females. A detailed description of variables used in Blinder-Oaxaca model in this paper can be found in Table 4 in the Appendix.

Table 2 presents the sample statistics of the estimated coefficient for log wages of males and females. The first two Males and Females columns on the left show the results of re-test of coefficients of Altonji & Blank’s six socioeconomic characteristics. EDUCATION ATTAINMENT includes the record of number of years of education received by each individual/respondent of the survey. AGE includes the real age of the respondents and the sample individuals selected are individuals aged from 18 to 64 years old. LABORFORCE includes dummy variables that indicating if the respondents are currently in the labor force or not. REGION includes region dummy variables of 9 areas; the 9 location dummies are used to test if there is any regional labor market difference. Similarly, 10 OCCUPATION and 13 INDUSTRY dummies are also included to test if there is any occupation and industry labor market difference. The two columns on the right indicate the results of estimated coefficients for log wages of males and females after adding FAMILY COMMITMENT as the seventh predictor. This paper sets 5 years old as children’s formative year. Reason being parents are most needed during their children’s formative year, number of children under 5 years old would be a good indicator of family commitment. Hence, FAMILY COMMITMENT includes the number of children under 5 years old.

(11)

Table 2 Coefficient Estimates for log wagesa

MALESb _FEMALESc _MALESb _FEMALSc

2008 FAMILY COMMITMENT X X 0,110*** 0,074*** EDUCATION ATTANIMENT 0,118***d _0,130*** _0,117*** _0,129*** AGE 0,014*** 0,009*** 0,015*** 0,010*** LABORFORCE 0,190*** 0,176*** 0,179*** 0,182*** REGION NEW ENGLAND 0,097*** 0,148*** 0,099*** 0,149*** MIDDLE ATLANTIC 0,113*** 0,149*** 0,115*** 0,149*** EAST-NORTH CENTRAL 0,064*** 0,070*** 0,064*** 0,069*** WEST-NORTH CENTRAL -0,015 0,027** -0,014 0,026** SOUTH ATLANTIC 0,050*** 0,101*** 0,055*** 0,102*** EAST-SOUTH CENTRAL -0,014 -0,028* -0,013 -0,028*

WEST-SOUTH CENTRAL 0(omitted) 0(omitted) 0(omitted) 0(omitted)

MOUNTAIN 0,076*** 0,082*** 0,074*** 0,081***

PACIFIC 0,120*** 0,175*** 0,122*** 0,177***

OCCUPATION

(1) MANAGERIAL, PROFESSIONAL, AND RELATED

OCCUPATIONS 0,417*** 0,271*** 0,408*** 0,272***

(2) PROFESSIONAL AND RELATED OCCUPATIONS 0,337*** 0,186*** 0,329*** 0,186***

(3) SERVICE OCCUPATIONS -0,104*** -0,206*** -0,104*** -0,204***

(4) SALES 0,121*** -0,110* 0,117*** -0,106*

(5) ADMINISTRATIVE SUPPORT -0,089** -0,083 -0,082** -0,079

(6) FARMING, FISHING, AND FORESTRY OCCUPATIONS 0(omitted) 0(omitted) 0(omitted) 0(omitted)

(7) CONSTRUCTION AND EXTRACTION OCCUPATIONS 0,083** -0,011 0,080** -0,001

(8) INSTALLATION, MAINTENANCE, AND REPAIR

OCCUPATIONS 0,102** 0,150** 0,100** 0,154**

(9) PRODUCTION OCCUPATIONS -0,047 -0,213*** -0,046 -0,210***

(10) TRANSPORTATION AND MATERIAL MOVING -0,125*** -0,168*** -0,124*** -0,166***

INDUSTRY

(11) AGRICULTURE, FORESTRY, FISHING, AND HUNTING -0,376*** -0,328*** -0,387*** -0,326***

(12) MINING 0,233*** 0,185** 0,221*** 0,183**

(13) CONSTRUCTION -0,019 0,072 -0,029 0,068

(14) MANUFACTURING 0,068 0,134** 0,057 0,132**

(15) WHOLESALE AND RETAIL TRADE -0,097* -0,091 -0,102** -0,091

(16) TRANSPORTATION AND UTILITIES 0,092* 0,160** 0,081 0,160**

(17) INFORMATION 0,001 0,071 0,006 0,072

(18) FINANCIAL ACTIVITIES 0,114 0,109* 0,102* 0,108*

(19) PROFESSIONAL AND BUSINESS SERVICES 0,032 0,073 0,46 0,071

(20) EDUCATION AND HEALTH SERVICES -0,140*** -0,028 -0,146*** -0,03

(21) LEISURE AND HOSPITALITY -0,254*** -0,125* -0,252*** -0,124*

(22) OTHER SERVICES -0,212*** -0,095 -0,218*** -0,096

(23) PUBLIC ADMINISTRATION 0,149*** 0,138** 0,137*** 0,137**

CONSTANT 0,933*** 0,867*** 0,903*** 0,808***

a _{Source: Current Population Survey, March 2008.}

b _&c_{Sample size of Males and Females are 44260 and 42672, respectively.} d_{*, **, and*** denote significance at 10, 5, and 1%, respectively.}

(12)

III.II Methodology

In Altonji & Blank’s research (1999) on gender wage discrimination, they measured gender wage gap by controlling six main socioeconomic characteristics in their regressions, and referred the “unexplained” component as gender discrimination in labor market. However, as what they also evaluated in their paper, the validity of such measurement depends largely on if all the predictors which females and males differ are controlled (Borjas, 2010). There must be other socioeconomic characteristics other than education attainment, age, region of

residence, labor force, occupation, and industry that would also play a role in affecting gender wage gap.

This paper adds family commitment as the seventh predictor. Hotchkiss & Pitts (2005) believes that family commitment can be differ by gender, and have different impact on the labor force outcome for males and females. For instance, females are subject to motherhood penalty while partners of those females receive fatherhood compensation. Furthermore, mothers seem to dedicate/sacrifice more percentage of their working years to raise their children than the fathers do. Thus, this paper tests if family commitment would be a factor that has an impact on the gender wage gap, and hence further affect the results of gender wage discrimination. It will first re-exam the United States gender wage gap in 2008 by running a similar model as Altonji & Blank did in their paper. Then, this paper will run the model again but with the added control variable – family commitment. In this way, there is a horizontal comparison about the relationship between controlled variables and gender wage discrimination. To finalize the research, this paper uses the Oaxaca command in Stata to run the decomposition. Both “threefold” and “twofold” decompositions that mentioned in section II.I and II.II. will be conducted in this paper as these two decomposition methods are

provided in the Stata Journal while introducing “oaxaca” command. Besides, by adopting two different methods, this paper would prevent obtaining biased/wrong results as the results of the two methods should not vary too much.

III.III Results

Table 3 presents the results of Blinder-Oaxaca decomposition for linear regression with both “threefold” and “twofold” decomposition methods. The first column, Males vs Females 1, uses data with six socioeconomic characteristics; the column Males vs Females 2 uses data with seven socioeconomic characteristics including the newly added FAMLIY

(13)

In general, males earn 24.2% more than females in the sample data that this paper selected6_{. Start by analyzing vertically, in Males vs Females 1 case, “threefold”}

decomposition method concludes that the amount due to differences in the coefficients weights 89.6%7of the gender (log) wage differential. This shows that discrimination significantly contributes to the gender wage gap. In the “twofold” decomposition method, discrimination weights even more, 98.9%.

In Males vs Females 2 case, after adding FAMILY COMMITMENT as a predictor, Endowments in the “threefold” decomposition method increases from 0.0098567 to

0.0127258. This indicates that FAMILY COMMITMENT dose play a part in explaining the gender wage gap. Correspondingly, weights of discrimination to the total gender wage differential decreases from 89.6% to 89.3%. In the “twofold” decomposition method, percentage of gender wage gap that can be “explained” increases from 1.07% to 1.75%. Meanwhile, the discrimination weight decreases from 98.9% to 98.2%.

The change in results in both decomposition methods after adding FAMILY COMMITMENT proves that it plays a role in affecting the measurement of gender wage discrimination as it contributes in explaining the gender wage gap. Nevertheless, it is important to take note that the change (±0.5%) is not very significant. In other words, it is important to understand that gender wage discrimination still largely account for the gender wage gap in 2008 in the United States.

6_{Labor force participation difference simply measures the gender (log) wage differential, selected predictors} will not affect the overall gender wage differential of the sample.

7_89.6%=!.!"#"$$%

(14)

Table 3

Decomposition of gender labor force participation differentials

Males vs Females 1 Males vs Females 2

2008 "threefold"

Labor force participation

difference 0.2423614 0.2423614 Amount due to Endowments 0.0098567 0.0127258 Coefficients 0.2171009 0.2165231 Interaction 0.0154041 0.0131125 Weights of discrimination 89.6% 89.3% 2008 "twofold"

Labor force participation

difference 0.2423614 0.2423614 Amount due to Explained 0.0025929 0.0042381 Unexplained 0.2397686 0.2381233 Weights of discrimination 98.9% 98.2% III.IIII Discussion

Though this paper has drawn a clear conclusion, there are still areas of limitations and improvements regarding this paper that are worth discussing.

Firstly, in section II.III., Altonji & Blank’s regression results shows that “unexplained” component weights 94.872% of the gender wage gap in 1995 in the United States. However, the empirical research conducted in this paper has come out with the results of 98.9%

discrimination weights by applying the “twofold” decomposition method. This indicates that level of gender wage discrimination changes over time. This might because of the predictors also changes over time and hence affect the discrimination level. Thus, further research could investigate on the gender wage discrimination over time.

This paper only investigates on one of the potential predictor – family commitment. The results show that though there is an impact of family commitment on the gender wage discrimination measurement, the impact is not so large that would alter the previous findings.

(15)

Further research should continue dedicating on testing other socioeconomic characteristics and review the relationship between those newly set predictors and the gender wage discrimination. Actually, there are prominent scholars who have come out with more significant results. For instance, Nyhus & Pons (2012) in their paper “Personality and the gender wage gap” concluded that differences in the personality trait scores could explain up to 11.5% of the observed gender wage gap. Such researches prove that additional research on the gender wage discrimination would be greatly valuable.

Lastly, this paper only tests the gender wage discrimination in the United States. However, as shown in Figure 1., gender discrimination does not only matter in the United Sates, but also in other countries throughout the world. In addition, though there are papers investigate gender wage discrimination in other countries. Those selected countries are normally developed western countries where data is relatively easier to obtain. However, gender discrimination in Asian countries could be more severe. To name a few, South Korea, a country with bastion of male privilege, it is not hard to imagine gender wage discrimination might be worse in South Korea. Hence, a cross-country research on the gender wage

discrimination would be very fruitful as well. Figure 18

8_{Source: Olivetti, C., & Petrogolo, B. (2008).“ Unequal Pay or Unequal Employment? A Cross-Country} Analysis of Gender Gaps,” Journal of Labor Economics 26, 623, Fig.1.

(16)

IV. Conclusion

To conclude, this paper adopts Altonji, J. G., & Blank, R. M.’s research paper as the main literature review. It builds on the research that Altonji & Blank did in their paper and apply two different decomposition methods and finally comes out with a meaningful result. That is, family commitment is a socioeconomic characteristic that would affect gender wage

discrimination measurement, and hence affect the weights of discrimination on the total gender wage gap. However, this paper does point out the areas that future research should conduct at.

(17)

Reference List

Altonji, J. G., & Blank, R. M. (1999).“ Race and Gender in the Labor Market,” in Orley Ashenfelter and David Card, editors, Handbook of Labor Economics, vol. 3C, 3144-3213.

Borjas, G. J. (2010). Labor Economics (6th_ed.)

Blinder, A. S. (1973).“ Wage discrimination: Reduced form and structural estimates,” Journal of Human Resources 8, 436-455.

Coral, R., Carlos, G., & Olga, C. (2011).“ The measurement of gender wage discrimination: the distributional approach revisited,” Journal of Economics Inequality, vol. 9(1), 57-86.

ChangHwan, K. (2010).“ Decomposing the Change in the Wage Gap Bewteen White and Black Men Over Time, 1989-2005: An Extension of the Blinder-Oaxaca Decomposition Method,” Sociological Methods & Research, vol. 38, no. 4, 619-651.

Dennis, J. A., & Glen, G. C. (1977).“ Statistical Theories of Discrimination in Labor Markets,” Industrial and Labor Relations Review, vol. 30, no. 2, 175-187.

Ellen, K. N., & Empar, P. (2012).“ Personality and the gender wage gap,” Applied Economics, vol. 44 (1), 105-118.

Hotchkiss, J. L., & Pitts, M. M. (2005).“ Female labor force intermittency and current earnings: switching regression model with unknown sample selection,” Applied Economics, 37, 545-560.

June E. O’N., & Dave M. O’N. (2005).“ What do wage differentials tell us about labor market discrimination?” National Bureau of Economic Research.

Jones, F. L., & Kelley, J. (1984).“Decomposing differences between groups: A cautionary note on measuring discrimination,” Sociological Methods and Research 12, 323-343.

Jann, B. (2008).“ The Blinder-Oaxaca decomposition for linear regression models,” The Stata Journal 8, no. 2, 453-479.

Marjorie, L. B., & Chung C. (2014).“ Re-examining the models used to estimate disability-related wage discrimination,” Applied Economics, vol. 46 (12), 1393-1408.

Newman, S., & Oaxaca, R. (2004).“ Wage decompositions with selectivity corrected wage equations: A methodological note,” Journal of Economic Inequality 2, 3-10.

(18)

Neumark, D. (1988).“ Employers’ discriminatory behavior and the estimation of wage discrimination,” Journal of Human Resources 23, 279-295.

Nielsen, H. S. (2010).“ Wage Discrimination in Zambia: an extension of the Oaxaca-Blinder decomposition,” Applied Economics Letters, vol. 7 (6), 405-408.

Oaxaca, R. (1973). “Male-female wage differentials in urban labor markets,” International Economic Review 14, 693-709.

Olivetti, C., & Petrogolo, B. (2008).“ Unequal Pay or Unequal Employment? A Cross-Country Analysis of Gender Gaps,” Journal of Labor Economics 26, 621-654.

Oaxaca, R. L. (1988).“ Searching for the effect of unionism on the wages of union and nonunion workers,” Journal of Labor Research 9, 139-148.

(19)

Appendix

Table 4

Description of variables used in Blinder-Oaxaca model

Variable Description Mean Std. Dev

FAMILY COMMITMENT Number of children under 5 years old. 0.21 0.514

EDUCATION ATTANIMENT Record of number of years of education

receieved by each individual/repspondent. 7.70 1.349

AGE Real age of the respondents who aged

from 18 to 64. 40.17 122.000

LABORFORCE Variable indicates 1 if the respondent is in

the labor force; 0 if not. 1.95 0.225

NEW ENGLAND Varialbe indicates 1 if the respondent is in

New England; 0 if not. 0.11 0.307

MIDDLE ATLANTIC Varialbe indicates 1 if the respondent is in

Middle Atlantic; 0 if not. 0.09 0.292

EAST-NORTH CENTRAL Varialbe indicates 1 if the respondent is in

East-North Central; 0 if not. 0.12 0.320

WEST-NORTH CENTRAL Varialbe indicates 1 if the respondent is in

West-South Central; 0 if not. 0.12 0.320

SOUTH ATLANTIC Varialbe indicates 1 if the respondent is in

South Atlantic; 0 if not. 0.18 0.385

EAST-SOUTH CENTRAL Varialbe indicates 1 if the respondent is in

East-South Central; 0 if not. 0.04 0.205

WEST-SOUTH CENTRAL Varialbe indicates 1 if the respondent is in

West-South Central; 0 if not. 0.08 0.278

MOUNTAIN Varialbe indicates 1 if the respondent is in

Mountain; 0 if not. 0.11 0.309

PACIFIC Varialbe indicates 1 if the respondent is in

Pacific; 0 if not. 0.15 0.359

(1) MANAGERIAL, PROFESSIONAL, AND RELATED OCCUPATIONS

Varialbe indicates 1 if the respondent is in Managerial, Professional, and related occupations; 0 if not.

0.14 0.344

(2) PROFESSIONAL AND

RELATED OCCUPATIONS Varialbe indicates 1 if the respondent is in _{Professional and related occupations; 0 if}

not.

0.21 0.404

(3) SERVICE OCCUPATIONS Varialbe indicates 1 if the respondent is in

Service occupations; 0 if not. 0.15 0.360

(4) SALES Varialbe indicates 1 if the respondent is in

Sales sector; 0 if not. 0.10 0.296

(5) ADMINISTRATIVE

SUPPORT Varialbe indicates 1 if the respondent is in _{Administrative support; 0 if not.} 0.14 0.342 (6) FARMING, FISHING,

AND FORESTRY OCCUPATIONS

Varialbe indicates 1 if the respondent is in Farming, fishing, and forestry

occupations; 0 if not.

0.01 0.087

(7) CONSTRUCTION AND EXTRACTION

OCCUPATIONS

Varialbe indicates 1 if the respondent is in Construction and extraction occupations; 0 if not.

0.06 0.233

(8) INSTALLATION, MAINTENANCE, AND

REPAIR OCCUPATIONS

Varialbe indicates 1 if the respondent is in Installation, maintenance, and repair occupations; 0 if not.

(20)

(9) PRODUCTION

OCCUPATIONS Varialbe indicates 1 if the respondent is in _{Production occupations; 0 if not.} 0.07 0.247 (10) TRANSPORTATION

AND MATERIAL MOVING Varialbe indicates 1 if the respondent is in _{Transportation and material moving}

sector; 0 if not.

0.06 0.232

(11) AGRICULTURE, FORESTRY, FISHING, AND HUNTING

Varialbe indicates 1 if the respondent is in Agriculture, forestry, fishing, and hunting industry; 0 if not.

0.01 0.105

(12) MINING _{Varialbe indicates 1 if the respondent is in}

Mining industry; 0 if not. 0.01 0.082

(13) CONSTRUCTION Varialbe indicates 1 if the respondent is in

Construction industry; 0 if not. 0.07 0.253

(14) MANUFACTURING _{Varialbe indicates 1 if the respondent is in}

Manufacturing industry; 0 if not. 0.11 0.316

(15) WHOLESALE AND

RETAIL TRADE Varialbe indicates 1 if the respondent is in _{Wholesale and retail trade; 0 if not.} 0.13 0.338 (16) TRANSPORTATION

AND UTILITIES Varialbe indicates 1 if the respondent is in _{Transportation and utilities; 0 if not.} 0.05 0.222

(17) INFORMATION _{Varialbe indicates 1 if the respondent is in}

Informanation industry; 0 if not. 0.02 0.148

(18) FINANCIAL

ACTIVITIES Varialbe indicates 1 if the respondent is in _{Financial industry; 0 if not.} 0.07 0.247 (19) PROFESSIONAL AND

BUSINESS SERVICES Varialbe indicates 1 if the respondent is in _{Professional and business services}

industry; 0 if not.

0.09 0.293

(20) EDUCATION AND

HEALTH SERVICES Varialbe indicates 1 if the respondent is in _{Education and health services industry; 0}

if not.

0.22 0.412

(21) LEISURE AND

HOSPITALITY Varialbe indicates 1 if the respondent is in _{Leisure and hospitality; 0 if not.} 0.08 0.269

(22) OTHER SERVICES _{Varialbe indicates 1 if the respondent is in}

Other services industry; 0 if not. 0.04 0.195

(23) PUBLIC

Impact of family commitment on gender wage discrimination measurement