Gender Earnings Gap in Hong Kong

Transcription

1 Gender Earnings Gap in Hong Kong By Yeung Lai Wa Applied Economics Option As Honours Degree Project Submitted to the School of Business in Partial Fulfillment Of the Graduation Requirement for the Degree of Bachelor of Business administration (Honours) Hong Kong Baptist University Hong Kong April

2

3 Acknowledgement I would like to thank Professor Lam Kit Chun for providing me a lot of useful advices and explanations in my Honours Project. Her advices and explanations really stimulated my thought to solve my problems. Doing this Honours Project not only provided me an opportunity to apply the skills and knowledge learned in the courses in University but also enhanced my analytical and organizational skills. I hope I can apply what I have learned in this Honours Project in my future career life. 2

4 Abstract This paper examines the gender earnings gap of males and females in Hong Kong in 2001 and 2006 respectively. It is noted that males had more advancement in earnings power than females in both years. Females earned 25.61% less than males in mean earnings in 2001 and this difference shrank to about 20.21% in By using the standard Oaxaca (1973) and Blinder (1973) decomposition based on male wage structure, it is found that the total gender earnings differential at mean earnings was in 2001 and decreased to in In both years, the gender earnings gap arises from the differential of endowment/personal characteristics is small. Most of the gender earnings gap is unexplained. However, the unexplained potion of the total gender earnings gap decreased a little bit from 2001 towards

5 Table of Contents I. Introduction P.5-P.6 II. Literature Review P.7-P.9 III. Data P.10 IV. Methodology P.11-P.13 V. Empirical Results and Discussions A. Sample Statistics B. Analysis of the Regression Results C. Analysis of gender earnings gap by standard P.14-P.30 P.14-P.18 P.19-P.23 P.24-P.30 Oaxaca(1973) and Blinder (1973) decomposition VI. Limitations and Recommendations P.31 VII. Conclusion P.32 VIII. References P.33-P.34 IX. Appendix P.35-P.36 4

6 I. Introduction It is generally believed that Hong Kong enjoys high gender equality in wage earnings with its spirit of equal and fair treatment to human beings so that people seem to pay not much attention to the issue of gender earnings gap in Hong Kong. However, this may not be true. Previous studies have found that gender earnings gap did exist in Hong Kong with males generally enjoyed higher earnings power compared to females in the past several decades though there was a trend of narrowing gender earnings gap (eg. Lui & Suen, 1993; Chung, 1996; Sung, Zhang & Chan, 2001). Gender inequality not only affects the division of labor, allocation of resources and economic welfare between males and females in the family but also influences the economic incentives faced by female workers beyond the family (Sung et al., 2001). Therefore studying the issue of gender earnings inequality has great economic and social values: It not only raises the public s awareness of this issue but also acts as a useful guide for the government or other related organizations to carry out appropriate policies for better allocation of resources and economic welfare between males and females, thereby brings a more equal and fair labor market. Gender earnings gap in 2001 and 2006 will be estimated and analyzed in this paper. Hong Kong society was still suffering economic downturn caused by the Asian Financial Crisis in However, as time went by, Hong Kong economy was greatly recovered in 5

7 2006. It is worth examining how the gender earnings gap was affected in this 5- year period. Specifically, the main objectives of this paper are the followings: (1). To analyze how large the gender earnings gap is in 2001 and 2006 and how the gap differs from 2001 towards (2). To investigate how large the gender earnings gap is caused by (i). differential of gender characteristics/endowments and (ii). the unexplained factors which may partly arise from market discrimination. 6

8 II. Literature Review The first step for studying the issue of gender earnings inequality is to find out factors which cause gender earnings gap. Blinder(1973) pointed out that part of each wage differential is due to differences in objective characteristics such as education and work experience, while part remains even white-black and male female differences in these traits are controlled for (p.437). This was similar to what Carnoy (1994) had mentioned that the between group earnings differentials could be decomposed into the attribute component and the price component. The attribute component refers to characteristics of a group that affect its earnings which can either be endowed or acquired ( endowed attributes include being female, white and born in poor family, etc. while acquired attributes include educational and occupational levels, etc.) (Chung, 1996). On the other hand, the price component reflects the value of the attribute in the labor market which is based on market preference or discrimination towards an individual possessing that attribute (Chung, 1996). One very important point in studying gender earnings gap is to measure gender earnings differentials. Blinder (1973) and Oaxaca (1973) introduced a decomposition method to measure gender earnings differentials. Based on this method, the gender earnings differentials can be decomposed in two parts: lny m - lny f = β m ( X m X f ) + [ X f (β m β f ) + (α m α f )] (1) where lny is average natural logarithm of earnings, X is a vector of average individual 7

9 characteristics, β is a vector of coefficients, ε and µ are disturbance terms. The first term on the right hand side of the equation [β m (X m X f )] is the earnings differential attributable to endowments (observable characteristics ) while the second term [ X f (β m β f ) + (α m α f )] is the differential due to the coefficients (returns to these characteristics). The second term exists because the market evaluates the identical set of traits possessed by different members of demographic groups (gender groups in this paper) differently and it is a reflection of discrimination (Blinder, 1973). Oaxaca (1973) stated that discrimination against females can be said to exist whenever the relative wage of males exceeds the relative wage that would have prevailed if males and females were paid according to the same criteria. (p.694). Oaxaca (1973) also stated that one difficulty of the wage equation when estimating the gender earnings differentials is that it controls for the major sources of discrimination that many consider to be against women. He explained that by controlling occupations, some effects of occupational barriers which act as sources of discrimination will be eliminated., so another set of equations that do not control for occupations should be estimated. Blinder (1973), Oaxaca (1973), Ng (2007) & Sebaggala (2007) pointed out that another problem this decomposition method involves is the familiar index number problem: in equation (1), the adoption of β m in calculating earnings differential attributable to endowments indicates that the male earnings structure is assumed to prevail without discrimination. However, they stated that female wage structure can also be taken as the 8

10 non-discriminatory benchmark with the following modification of equation (1): lnym - lny f = β f ( X m X f ) + [ X m (β m β f ) + (α m α f )] (2) They pointed out that using both male and female wage structure as the non-discriminatory benchmark can result in estimates with a range of possible values. Nuemark (1988), Cotton (1988), Oaxaca & Ransom (1994) have developed a more general wage decomposition method. This method uses a pooled sample of the two demographic groups as the non-discriminatory wage structure which can be expressed as: lnym - lnyf = ( X m X f ) β * + [(β m β * ) X m + (β * β f )X f ] The second term on the right hand side of the equation [(β m β * ) X m ] measures the extra amount of wage received by males if their sample characteristics were to be rewarded at β *, β * is the non-discriminatory wage structure (Sebaggala, 2007). The third term [(β * β f ) X f ] estimates the female disadvantage that is equal to the difference between the females actual wage received and the wage they should receive if the non discriminatory wage structure was implemented (Sebaggala, 2007). However, Appleton et al. (1999) doubted whether the pooled coefficient can be a good estimator of the non-discriminatory wage structure. It is noted that both the standard Oaxaca (1973) and Blinder (1973) decomposition and the decomposition approach developed by Nuemark (1988) or Cotton (1988) are not perfect and encounter their own problem. 9

11 III. Data The data used in this paper comes from 2001 and 2006 population Census aged between years old. Since the earnings structure of employees and the self-employed may be different, so only employees (work for wage, salary commission, tips or payment in kind) are chosen in this study. Besides, to eliminate the effect low income from the large amount of foreign domestic helpers, foreigners who are not born in Hong Kong or the Mainland of China are excluded. Only those born in Hong Kong and the Mainland of China are selected in this study in order to focus on local employees (Sung et al., 2001). 10

12 IV. Methodology (1). Sample Statistics: (i). Estimate the sample means of the variables (age, monthly earnings, experience, experience squared, schooling, different education level attainment, martial status, birth of place and occupations) (ii). Analyze the sample characteristics with the estimated sample means (2). Model Specification: (i). The model that does not control for occupation (Model 1): Discussion of the empirical results will be mainly based on this model. The standard Oaxaca (1973) and Blinder (1973) decomposition will be applied in this paper. Gosse (2001) pointed out that to get rid of the problem of index problem, the most common way is to do the decomposition based on the male wage structure, so as to standardize the literature. So this paper will also be based on the male wage structure. An Ordinary Least Squares regression for males (m) and females (f) is needed: lnym = α m + β m X m + ε (1) lny f = α f + β f X f + µ (2) where lny is natural logarithm of monthly earnings, X is a vector of individual characteristics, β is a vector of coefficients, ε and µ are disturbance terms. 11

13 The human capital earnings equation by Mincer (1974) will be applied to run the above regressions: log(mearn) = α + β1 (EDUCN) + β2 (EXP) + β3 (EXP)² + β4 (MARIT) + β5 (BORNPL) + ε (3) where log (MEARN) is the dependent variable which refers to logarithm of monthly earnings. The independent variables are the followings: - EDUCN refers to educational attainment (highest level attained) for lower secondary (LOWSEC), upper secondary (UPSEC), post secondary (POSTSEC), university (UNIV) and postgraduate (POSTGRA) with primary and below (PRIMB) as the reference group. - EXP and EXP² refer to years of work experience acquired and its square respectively in which EXP = Age - years of schoolings completed 6. - MARIT refers to marital status which is a dummy variable for married individuals (MARRIED), widowed individuals (WIDOWED), divorced/separated individuals (DIVORCED) with never married individuals (NEVERMARRIED) as the reference group. - BORNPL refers to birth of place which is a dummy variable for individuals born in Hong Kong (HK) with individuals born in China (CHINA) as the reference group. 12

14 The following decomposition equation will then be applied: lny m - lny f = β m ( X m X f ) + [ X f (β m β f ) + (α m α f )] X m and X f can be found from sample means of related variables while β m and β f can be estimated from the regression equations (1) and (2). Then the explained part [β m ( X m X f )] which is the wage earnings differential attributable to gender characteristics and the unexplained part [ X f (β m β f ) + (α m α f )] which is the differential attributable to returns of these characteristics can be calculated. (ii) The model that controls for occupations (Model 2): The operation of this model is similar to model 1 but the dummy variable OCCUP (occupations) is incorporated into the human capital earnings equation in this model: log (MEARN) = α + β1 (EDUCN) +β2 (EXP) +β3 (EXP )² +β4 (MARIT) + β5 (BORNPL) +β6 (OCCUP) + µ where OCCUP refers to occupations which are dummy variables for Managers, Professionals and Associate Professionals (MAN&PROF), Clerks (CLERK), Service Workers and Shop Sales Workers (SERVICE) and Agricultural and others (AGRI) with Elementary, Plant and Machine Operators and Assemblers (ELEM) as the reference group. 13

15 V. Empirical results and Discussions A. Sample Statistics Table 1: Mean values of variables (2001) Note: standard deviations in parentheses Variables All Male Female Monthly earnings ( ) ( ) ( ) Log of monthly earnings 9.37 (0.72) 9.48 (0.69) 9.24 (0.72) Age (10.23) (10.41) (9.90) Schooling (3.78) (3.72) (3.84) Education Attainment Primary and below 0.16 (0.37) 0.16 (0.37) 0.16 (0.37) Lower secondary 0.20 (0.40) 0.24 (0.43) 0.14 (0.35) Upper secondary 0.38 (0.49) 0.35 (0.47) 0.42 (0.49) Post secondary 0.10 (0.29) 0.09 (0.28) 0.11 (0.31) University 0.13 (0.34) 0.12 (0.33) 0.14 (0.34) Postgraduate 0.03 (0.18) 0.04 (0.19) 0.03 (0.16) Experience (12.20) (12.17) (12.13) Experienced Squared (577.82) (566.46) (543.00) Marital status Married 0.59 (0.49) 0.63 (0.48) 0.55 (0.50) Never married 0.37 (0.48) 0.35 (0.48) 0.40 (0.49) Widowed 0.01 (0.09) (0.05) 0.01 (0.12) Divorced/Separated 0.03 (0.17) 0.02 (0.14) 0.04 (0.20) Birth of place Hong Kong 0.72 (0.45) 0.70 (0.46) 0.73 (0.44) China 0.28 (0.45) 0.30 (0.46) 0.27 (0.44) Occupations Managers 0.07 (0.26) 0.09 (0.29) 0.05 (0.22) Professionals 0.06 (0.24) 0.07 (0.25) 0.06 (0.23) Associate professionals 0.18 (0.38) 0.16 (0.37) 0.19 (0.40) Clerks 0.20 (0.40) 0.10 (0.30) 0.33 (0.47) Service & Shop Sales Workers 0.17 (0.37) 0.16 (0.37) 0.17 (0.38) Elementary, craft, plant and machine operators 0.32 (0.47) 0.42 (0.49) 0.20 (0.40) Agricultural and Others (0.03) (0.04) (0.02) Sample Size

16 Table 2: Mean values of variables (2006) Note: standard deviations in parentheses Variables All Male Female Monthly earnings ( ) ( ) ( ) Log of monthly earnings 9.30 (0.74) 9.37 (0.73) 9.21 (0.74) Age (10.61) (10.78) (10.33) Schooling (3.79) (3.76) (3.83) Education Attainment Primary and below 0.12 (0.33) 0.12 (0.33) 0.13 (0.33) Lower secondary 0.19 (0.39) 0.22 (0.42) 0.15 (0.36) Upper secondary 0.38 (0.48) 0.36 (0.48) 0.39 (0.49) Post secondary 0.10 (0.30) 0.10 (0.30) 0.11 (0.31) University 0.16 (0.36) 0.14 (0.35) 0.17 (0.37) Postgraduate 0.05 (0.22) 0.06 (0.23) 0.05 (0.21) Experience (11.79) (11.85) (11.64) Experienced Squared (505.14) (520.12) (482.90) Marital status Married 0.57 (0.49) 0.61 (0.49) 0.53 (0.50) Never married 0.38 (0.49) 0.36 (0.48) 0.40 (0.49) Widowed 0.01 (0.09) (0.06) 0.02 (0.12) Divorced/Separated 0.04 (0.19) 0.02 (0.15) 0.05 (0.22) Birth of place Hong Kong 0.74 (0.44) 0.75 (0.43) 0.74 (0.44) China 0.26 (0.44) 0.25 (0.43) 0.26 (0.44) Occupations Managers and Administrators 0.07 (0.25) 0.08 (0.27) 0.05 (0.23) Professionals 0.07 (0.25) 0.07 (0.26) 0.06 (0.23) Associate professionals 0.18 (0.39) 0.17 (0.38) 0.20 (0.40) Clerks 0.20 (0.40) 0.11 (0.31) 0.31 (0.46) Service & Shop Sales Workers 0.18 (0.38) 0.17 (0.37) 0.19 (0.39) Elementary, craft, plant & machine operators 0.30 (0.46) 0.40 (0.49) 0.19 (0.39) Agricultural and Others (0.04) (0.05) (0.04) Sample Size

17 Table 3: Monthly Earnings of different occupations Note: standard deviations in parentheses Occupations Male Female Male Female Managers and administrators ( ) ( ) ( ) ( ) Professionals ( ) ( ) ( ) ( ) Associate professionals ( ) ( ) ( ) ( ) Clerks ( ) ( ) ( ) ( ) Service & Shop Sales Workers ( ) ( ) ( ) ( ) Elementary, craft, plant and machine operators ( ) ( ) ( ) ( ) Agricultural and others ( ) ( ) ( ) ( ) In 2001, the sample size consists of individuals in total with males (55.71%) and females (44.29%). In 2006, it consists of individuals in total with males (54.25%) and females (45.75%). In general, the monthly earnings of both males and females in 2006 are slightly less compared with In 2001, females earned about 25.61% less than males in mean monthly earnings. This difference shrank to about 20.21% in It is a bit surprised to discover that in both 2001 and 2006, females enjoyed a bit more advancement than males in regard to education attainment. Females had an average of and years of schooling in 2001 and 2006 respectively, a bit higher than that of males 16

18 (10.55 years in 2001 and years in 2006). The proportion of both males and females with higher educational levels (university and postgraduate) in 2006 is slightly higher than that in 2001, with a bit larger increase for females compared with males. Besides, in both years, the proportion of females in the higher educational levels (upper secondary, post secondary and university) is slightly higher than their male counterparts. It can be found that Hong Kong enjoys high gender equality in receiving education with a trend that even favored females both in 2001 and The sample individuals have an average of around 20 years of working experience in both years. Males had about 2 more years of working experience than females. In both years, the married individuals account for about 60% of the total sample. The proportion of married men is a little bit higher than their female counterparts. In 2001, about 28% of the individuals of the sample data are born in China. However, the proportion of the China born individuals decreases a little bit to reach about 25% of the sample data in When focusing on occupations, in both years, less than 20% of the sample individuals are concentrated in the relatively high income occupations (managers, administrators and professionals). The male proportion is higher than that of females in managers and administrators while a similar distribution of males and females is shown in professionals. As shown by table 3, managers and administrators are the highest earnings group, followed by 17

19 professionals. Although males generally earn more than females in these high income groups, the amount that females earn less than males is not very considerable in both years. Associate professionals are the third highest earnings group and account for about 20% of the sample size in both years. Besides, about one third of females in the sample are clerks, nearly triples the number of male clerks in both years. In these two occupations, males and females have similar monthly earnings. About 20% of the sample size is concentrated on the service sector. The proportion of males and females in this sector is quite equal in both years. On the other hand, nearly 40% of males are concentrated in the elementary, craft and machinery sectors and this proportion doubles the proportion of their female counterparts in both years. Besides, females earned much less than males in the service, elementary, craft and machinery sectors. The monthly earnings of females are at least 50% less than their male counterparts in these sectors. It is found in both years, males are largely concentrated in the manual related occupations while a considerable amount of females are concentrated in the clerical sector. It is noted that females earned much less than males in the relatively low income jobs. 18

20 B. Analysis of Regression Results Table 4: Regression Results Male Female Male Female CONSTANT *** (356.99) *** (291.21) *** (321.11) *** (298.29) LOWSEC *** (5.68) *** (7.70) *** (8.71) *** (3.78) UPSEC *** (24.67) *** (34.21) *** (25.15) *** (24.77) POSTSEC *** (37.31) *** (43.14) *** (33.80) *** (33.29) UNIV *** (56.68) *** (58.02) *** (51.44) *** (49.97) POSTGRA *** (55.10) *** (47.95) *** (54.22) *** (49.46) EXP *** (37.73) *** (34.50) *** (38.63) *** (38.39) EXP² *** (-32.85) *** (-27.75) *** (-32.61) *** (-32.44) MARRIED *** (15.94) (0.57) *** (15.54) *** (2.76) WIDOWED (1.03) (0.37) ** (2.15) * (1.78) DIVORCED (1.15) * (1.77) (1.41) (0.25) HK *** (16.23) *** (16.65) *** (16.13) *** (21.64) R-square Adjusted R- square N Note: t statistics in parentheses. The asterisks *, **, *** represents statistically significant at 10%, 5% and 1% significance level respectively. 19

21 Table 4 shows the Ordinary Least Squares regression results of male and female earnings equations for model 1 in 2001 and 2006 respectively. The adjusted R square shows that about 40% of the variations of the dependent variable (log monthly earnings) of the male and female earnings equations can be explained by their corresponding independent variables in both years. This indicates that the variations of the dependent variable can moderately be explained by their corresponding regression equations in both years. In both years, all the coefficients of both male and female earnings equations are statistically significant at 1% of significance level with only a few exceptions. All the coefficients have the signs as we expected in both years. Both males and females have positive return to education. The higher the educational level, the higher the rate of return. Females have a higher rate of return compared to males at each educational level with the exception of lower secondary level in To illustrate, compared with the female individuals attained primary level or below, females who attained university earned % more in 2001 and % more in However, their male counterparts only earned % more in 2001 and % more in 2006 compared with male individuals attained primary level or below. Besides, the rate of return of males in lower and upper secondary level increases slightly from 2001 to 2006 while decreases a little bit in post secondary, university and postgraduate level in this period. However, the rate of return of females in each educational level decreases significantly from 2001 to

22 The coefficient of Experience and Experience Square is positive and negative respectively which is within our anticipation. The coefficient of Experience Square is negative because the age profile is parabolic (Kaufman & Hotchkiss, 2006). Mincer & Polachek (1974) stated that as job related investment on training controls a return at work, the older we are, the shorter the duration of work experience and this will weaken our incentive to invest in our job skills. Besides, they pointed out that as we ages, our human capital depreciates. To illustrate, we first take a look at the earnings function: lny = lny 0 + α 1 School + α 2 EXP + α 3 EXP² The increase in earnings from an additional year of work experience is (lny)/ (EXP) = α 2 + 2α 3 EXP (Kaufman & Hotchkiss, 2006) By applying this equation, in 2001, the earnings of males increased by about (0.0011)(1) =5.95% for the first year of experience. However, their earnings only increased by about 1.84% in the twentieth year of experience. It is found that in both years, the rate of return of experience at mean for females (1.99% in 2001 and 1.91% in 2006) is higher than that of males (1.59% in 2001 and 1.74% in 2006) and the differential is less in The earnings of married, widowed and divorced/separated individuals are higher than those of never married individuals. When focusing on married individuals, married males earned about 20% more than never-married males in both years. This may be due to less 21

23 housework responsibility of married men which allows them to be more productive at work (Becker, 1985). It may also be possible for what Nakosteen and Zimmer (1987) had found that men possessing higher earnings are more likely to get married. Besides, it is found that in 2006, married women earned about 4% more than never married women. This positive earnings premium for females may be because nowadays some households hire domestic helpers to help married women to do the housework (Jeronimo & Eduardo), especially in the middle and upper class in Hong Kong. Furthermore, it is found that the earnings premium arise from marriage is much larger for males than females. This may be because married women are usually responsible for the major part of the housework (Becker, 1985). Although the division of housework between men and women nowadays seems to be more equal than before in Hong Kong, many women still take over the main housework. Hong Kong Young Women s Christian Association had interviewed a total of 2089 individuals (916 males and 1173 females) about their division of labor in housework in The result shows that over 80% of the female interviewees are responsible for doing the major tasks of housework and taking care of their children. Males born in Hong Kong earned about 17% and 19% more than those born in China in 2001 and 2006 respectively. On the other hand, Hong Kong born females earned about 21% and 28% more than those born in China in 2001 and 2006 respectively. A larger earnings differential is shown between Hong Kong born and China born females compared with their 22

24 male counterparts. This may be partly due to employer discrimination, and it may also be possible that compared with Hong Kong born females, China born females are more influenced by traditional attitudes and they usually have heavier housework, so they tend to work less (Sung, Zhang, Ng, and Hempel, 2002). Table 5 (see appendix) shows the Ordinary Least Square regression results of male and female earnings equations for model 2 in 2001 and 2006 respectively. This model has added occupational dummies into the earnings equations in hopes of investigating the impact of the distribution of different occupations of the males of females on the gender earnings differential, but this approach has assumed that occupations are given exogenously (Sung et al., 2001). However, occupations are endogenous. In case occupational determination or segregation is associated with labor market discrimination, this approach would not be appropriate (Gunderson, 1989). Therefore this model won t be discussed in detail in this paper. It is just for reference. 23

25 C. Analysis of male-female earnings gap by standard Oaxaca (1973) and Blinder (1973) decomposition (Model 1) Table 6: Decomposition Results Education LOWSEC UPSEC POSTSEC UNIV POSTGRA Subtotal for Education Gap contributed by variable (-0.33%) (-65.00%) (-21.04%) (-24.24%) (4.49%) ( %) Gap due to Gap due to Gap Gap due to Gap due to endowment coefficient contributed endowment coefficient differential differential by variable differential differential (4.00%) (-4.33%) (13.11%) (6.56%) (6.55%) (-12.01%) (-52.97%) (-20.50%) (-9.95%) (-10.55%) (-8.10%) (-12.94%) (-10.09%) (-6.26%) (-3.83%) (-7.64%) (-16.60%) (-23.47%) (-15.62%) (-7.85%) (7.27%) (-2.78%) (4.82%) (7.49%) (-2.67%) (-16.48%) (-89.62%) (-36.13%) (-17.79%) (-18.35%) EXP (65.84%) (55.77%) (10.08%) (42.76%) (77.07%) (-34.31%) EXP² (-44.28%) (-39.89%) (-4.39%) (-10.80%) (-58.54%) (47.75%) Marital Status MARRIED (52.45%) (6.84%) (47.41%) (62.77%) (9.51%) (53.26%) WIDOWED (0.00%) (-0.46%) (0.45%) (-0.38%) (-1.34%) (0.97%) DIVORCED (-0.55%) (-0.35%) (-0.20%) (0.44%) (-0.77%) (1.20%) Subtotal for Marital status (53.69%) (6.03%) (47.67%) (62.83%) (7.40%) (55.43%) 24

26 Table 6: Decomposition Results (Continue) Birth of place Hong Kong (-14.10%) (-2.00%) (-12.09%) (-35.61%) (1.18%) (-36.79%) Subtotal for all variables (-44.95%) (3.42%) (-48.36%) (23.04%) (9.32%) (13.72%) estimated constant differential (144.94%) (76.96%) Explained potion (endowment differential) (3.42%) (9.32%) Unexplained potion (coefficient differential estimated constant differential) (96.58%) (90.68%) Total differential (100%) (100%) Note: (i). Percentage of the gender earnings differential contributed by variables in terms of total gender earnings gap in parentheses; (ii) The log mean earnings differential between males and females is in 2001 and in Table 6 shows the explained (due to endowment differential) and unexplained amount (due to coefficient differential and estimated constant differential) that contribute to the gender earnings gap by different variables and the explained and unexplained potion of the total gender earnings gap by using the standard Oaxaca (1973) and Blinder (1973) decomposition for Model 1 in both 2001 and Since males are taken as the comparator group, a positive number in the explained / unexplained potion indicates a higher earnings power for males (increasing the gender 25

27 earnings differential) while a negative number indicates a higher earnings power for females which decreases the gender earnings differential (Gosse, 2001). It is found that the total gender earnings differential at mean earnings is in 2001 and decreases to in It can further be split into two components, the explained and unexplained potion. The Explained Gender Earnings Gap The explained amount of the gender earnings differential is in 2001 and is increased to in The explained potion only accounts for a very small proportion of the total gender earnings gap (3.42%) in However, it climbs a little to reach 9.32% in This shows that the gender earnings differential due to the observable characteristics or endowments is small in both years, but this differential enlarges a little bit in The unexplained Gender Earnings Gap The amount of unexplained gender earnings differential is in 2001 and is decreased to in It is found that the unexplained potion, which may partly arise from discrimination, accounts for most of the total gender earnings gap, say, 96.58% in 2001 and it decreases to 90.68% in Sung et al. (2002) had found that the unexplained potion accounts for 102.8% of the total gender earnings gap in 1996 using the standard Oaxaca 26

28 (1973) and Blinder (1973) decomposition. It can be found that the unexplained potion of the gender earnings gap has decreased from 1996 towards Gender earnings gap contributed by different variables 1. Education Education accounts for % and % of the total gender earnings gap in 2001 and 2006 respectively which helps to decrease the gender earnings gap. It indicates that females exceed males (by having higher rate of return) in regard to education attainment. This may be because women may accept the earnings which may undervalue their characteristics due to market prejudice or discrimination towards them and the higher educated women can better tackle these market disadvantages so as to compete with men (Dougherty, 2003). The results also indicate that the education advancement for females has declined from 2001 towards Among the % of the gap contributed by education in 2001, only a small proportion (-16.48%) is due to the endowment differential between males and female, the major gap comes from the coefficient differential resulting from higher rate of return of education to females. However, in 2006, the proportion of the gap arises from endowment differential and the rates of return to education are similar. When taking account of different levels of education attainment, almost all the education levels help to decrease the gender earnings gap, with the exception of postgraduates. In both 27

29 years, we discover that male postgraduates exceed their female counterparts and increase the gender earnings gap. Their higher earnings power comes from their endowment advancement over female postgraduates. 2. Experience Experience is the main contributor of the gender earnings gap that favors males in both years. It accounts for 65.84% and 42.76% of the gender earnings gap in 2001 and 2006 respectively. We discover that the endowment differential in which males exceed females accounts for most of the gap arise from experience. The coefficient differential only accounts for a small proportion. The fact that experience accounts for such a considerable amount of the gap can be explained by what Blinder (1973) had found in his estimates: women shows a flatter age-earnings profile than men, in which women have less advancement in exhibiting a rise of their earnings over their life cycle compared with men. Mincer & Polachek (1974) pointed out that there may be a discontinuous labor force participation of married women, especially for mothers since they may shift their time in doing housework and raising their children. This discontinuity may affect some young women to have less job training in their premature employment compared with their male counterparts with comparable education. They also stated that married women s non-participation in the labor force during childbearing may 28

30 depreciate their skills learned at school and obtained at work. So women exhibit a less steep earnings profile in their life cycles. 3. Marital Status Marital status is also a major contributor of the gender earnings gap in both years. It contributes 53.69% of the gender earnings gap in 2001 and increases a little bit to 62.83% in Most of the gap comes from the coefficient differential which is the differential between how male earnings equation would value the characteristic of marital status of the female earnings equation and how female earnings equation actually values them (Blinder, 1973). This indicates that marital status may be a significant source of discrimination. The endowment differential is not significant in this case. 4. Birth of place The Hong Kong born group that favors females decreases the gender earnings gap. They decrease 14.10% of the gap in 2001 and even more, say, 35.61% in A considerable amount of this gap arises coefficient differential, the differential between how male earnings equation would value the characteristic of birth of place of the female earnings equation and how female earnings equation actually values them (Blinder, 1973). This implies that birth of place may also be a source for gender discrimination which can t be overlooked. 29

31 5. Estimated constant The differential between the estimated constants (a part of the unexplained potion) of male and female earnings equation is the gender earnings gap when no other variables are controlled for. This differential accounts for most of the gender earnings gap in both years. It contributes even more than 100%, say, % of the gender earnings gap in

32 VI. Limitations and Recommendations Sung et al.(2002) pointed out that the unexplained part of the gender earnings gap may be overestimated since part of the unexplained earnings gap might come from variables such as working hours and intensity of effort rather than discrimination. Since these variables may determine productivity, so the explained gender earnings gap might be underestimated if these variables are omitted. If the Census data contains more productivity-related variables, the potion of the explained and unexplained gender earnings gap will be more accurate..besides, the importance of discrimination may be underestimated when the sample individuals are segregated into different occupations since occupational segregation may be associated with labor market discrimination (Oaxaca, 1973). Estimation of discrimination will be more accurate if the effect of occupational segregation is incorporated into the model. In addition, Gosse (2001) stated that the decomposition method only measures the post hiring earnings differentials. He pointed out that if the hiring process is subjected to market discrimination, then decompositions will underestimate the impact of discriminations on earnings. In such case, pre-market discrimination affecting the gender earnings gap that is related to productivity is omitted. The decomposition result will be more accurate if more information about pre-hiring situation such as the information about the hiring process is provided in the Census data. 31

33 VII. Conclusion This paper aims to examine the amount of gender earnings gap in 2001 and 2006 and how the gap differs in this five-year period. Besides, it also examines how large the gender earnings gap is caused by observable gender characteristics (endowments) and unexplained factors (which may partly come from market discrimination) respectively. Data from 2001 and 2006 Hong Kong Population Census is used to examine the gender earnings gap using the standard Oaxaca (1973) and Blinder (1973) decomposition based on the male wage structure. It is found that although the earnings power of males are generally higher than that of females in both years, the gap shrinks in this 5 year period. The total gender earnings differential was in 2001 and decreased to in It is also found that the endowment differential between males and females is not the major cause of the gap. Most of the gender earnings gap is unexplained which may partly arise from discrimination. However, the unexplained potion of the total gender earnings gap decreased a little bit from 2001 towards

34 VIII. References Appleton, S., J. Hoddinott & P Krishnan.(1999).The gender wage gap in three African Countries. Economic Development and Cultural Change, 47(2): Becker, G.. S.(1985). Human Capital, Effort, and the Sexual Division of Labor. Journal of Economics 3, S33-58 Blinder, A. S. (1973) Wage Discrimination: Reduced Form and Structural Estimations. Journal of Human Resources, 8, Carnoy, M. (1994). Faded Dreams. Melbourne: Cambridge University Press. Chung, Y. P. (1996). Gender Earnings differentials in Hong Kong: The Effect of the State, Education, and Employment. Economics of Education Review, 15, Cotton, J. (1988). On the decomposition of wage differentials. Review of Economics and Statistics, 70, Dougherty, C. (2003, August). Why is the rate of Return to Schooling Higher For Women Than For Men? UK: Centre for Economic Performance, School of Economics and Political Science. Gosse, M. (2001). The Gender Pay Gap in the New Zealand Public Service. Working Paper No.15, New Zealand: State Services Commission. Gunderson, M. (1989). Male-Female Wage Differentials and Policy Reponses. Journal of Economic literature, 27, Jeronimo O. M., & Eduardo L.G. Rios-Neto. Marriage Premium among men and women in Brazil, downloaded at 26/04/2006, Available: Kaufman, BE, & Hotchkiss, JL. (2006). The Economics of Labor Markets. 7th Edition, Dryden Press Lui, H.-K. & W. Suen (1993). The Narrowing Gender Gap in Hong Kong: Asian Economic Journal, 7(2),

35 Mincer, J. (1974). Schooling, Experience and Earnings, New York: National Bureau of Economic Research. Mincer, J. & S. Polachek. (1974). Family investment in Human Capital: Earnings of Women. Journal of Political Economy, Vol. 82 (2, part II): S76-S108. Nakosteen R. & Zimmer M. (1987). Marital Status and Earnings of Young Men. Journal of Human Resources, 22(2): Neumark, D. (1988). Employer s discrimatory behaviour and the estimation of wage discrimination. Journal of Human Resources, 23, Ng, Y. C., (2007, March). Gender Earnings Differentials and Regional Economic Development in Urban China Review of Income and Wealth, 53(1): Oaxaca,R.(1973). Male-Female Wage Differentials in Urban Labor Markets. International Economic Review, 14, Oaxaca, R.L. & M.R. Ransom.(1994). On discrimination and the decomposition of wage differentials. Journal of Econometrics, 61:5-21. Sebaggala, R. (2007). Wage Determination and Gender Discrimination in Uganda. Research Series No.50, Economic Policy Research Centre (EPRC), Uganda: Makerere University. Sung, Y. W., Zhang J.-S. & Chan C.-S. (2001). Gender wage gap differentials and Occupational Segregation in Hong Kong, Pacific Economic Review, 6:3, Sung, Y. W., Zhang J. S., Ng S.-H. & Hempel P. (2002). Feasibility on Equal Pay For Equal Work of Equal Value. Research Report 5, Hong Kong: Hong Kong: Equal Opportunity Commission. 和諧家庭研究系列 : 香港性別意識與家庭分工狀況調查 香港基督教女青年會 [ 網上 ] 下載於 26/04/2006 網址 : 34

36 IX. Appendix Table 5: Regression Results (Model 2) CONSTANT LOWSEC UPSEC POSTSEC UNIV POSTGRA EXP EXP² MARRIED WIDOWED DIVORCED HK MAN&PROF CLERK Male Female Male Female *** *** *** *** (374.49) (293.95) (343.61) (304.90) *** *** *** (4.32) (4.63) (7.40) (-0.10) *** *** *** *** (13.96) (20.79) (14.70) (11.26) *** *** *** *** (19.77) (24.87) (18.62) (16.54) *** *** *** *** (31.82) (33.01) (29.16) (27.14) *** *** *** *** (35.97) (31.50) (34.28) (29.74) *** *** *** *** (35.66) (31.56) (38.73) (35.89) *** *** *** *** (-30.76) (-24.37) (-32.24) (-28.52) *** *** ** (15.93) (0.76) (15.08) (2.28) ** * (0.29) (0.73) (2.28) (1.93) (1.16) (1.23) (1.35) (0.29) *** *** *** *** (13.24) (13.25) (11.60) (14.10) *** *** *** *** (40.21) (36.05) (50.58) (47.41) *** *** *** *** (5.31) (15.93) (11.35) (26.34) 35

37 Table 5: Regression Results (Model 2) (continue) Male Female Male Female SERVICE *** (15.49) *** (10.52) *** (18.81) *** (18.42) AGRI * (-1.68) (0.57) ** (-2.04) *** (5.22) R- Square Adjusted R-Square N Note: t statistics in parentheses. The asterisks *, **, *** represents statistically significant at 10%, 5% and 1% significance level respectively 36