Impact of College Degree on Gender Wage Gap

Can college degree shrink the gender wage gap in urban China? An empirical analysis based on CHIP 1995 and 2013

Abstract:

Based on the data set of CHIP 1995 and 2013, this paper estimates the rate of return to higher education for male and female by using extended Mincerian equation, and investigates how college degree affects the gender wage gap by exploiting the Oaxaca decomposition. The result indicates that the rate of return to higher education is not only statistical significant, but also economic significant. Oaxaca decomposition result shows that higher education accounts for 46.1% and -8.5% of the gender wage gap in 1995 and 2013 separately. The unexplained part of the decomposition increases from 46.1% to 71% during that period.

1 Introduction

China has the world’s largest labor force, with an estimated 0.78 billion actively working people as of 2014. The male female wage gap is one of the key issues concerning the vase number of workers in China. Prior economic reforms, China like the former Soviet Union and other eastern European countries, implemented a rigid wage system that compressed wage differentials among workers for the purposes of equity. However, since the economic reforms started, many impediments to gender equality have emerged, resulting in widened gender wage gap. Researchers have found evidence of rising gender earnings differentials in the Chinese urban labor market based on individual data collected from large-scale household survey. Gustafasson and Li (2000) found that female earnings as the percentage earnings had decreased from 84.4 to 82.5% from 1988 to 1995. Ng (2007) showed that the male-female earnings gap had declined in the early reform stage from 1988 to 1991 and then increased substantially in the later rapidly from 1988 to 1999 and then declined from 1999 to 2002.

Regarding the cause of the rising gender earnings gap in China, Gustafsson and Li (2000) found that from 1988 to 1995, the widening gap could not be attributed to men and women’s different productivity characteristics, hence suggesting that discrimination may be the primary cause. Ng (2007) also found that discrimination accounted for the larger proportion of the overall earnings gap. Appleton et al. (2005) also found that gender pay discrimination might have increased in the Chinese urban labor market, especially from 1988 to 1999.

China’s economic reforms, moving from a centrally planned economy to a market based economy, promotes the development of non-state-owned sectors and the restructuring of the wage system. In the market of labor, returns to education are an important factor of wage determination. Some literatures have examined the effect of education on gender wage gap. Zhang (2004) found that from 1986 to 1993, the power of education to explaining wage gap increased from 2.3% to 5%, but there was a fall back to 2% in 1997. Xie and Yao (2005) argued education can explain a small amount of the gender wage gap in 2002. Chen and Hamori (2008) also found education contributed to the discrimination in the labor market in 2005.

Although prior research has examined the gender wage gap and discrimination in China, there is a little literature that analyze how the college degree (or college premium) affect the gender wage differential in China. Based on the above discussion, we have two objectives in this paper. The first objective is to estimate the change return to higher education in urban China at two time points: 1995 and 2013. The second one is to examine how the college degree affect the gender wage differentials and to explore whether the pay discrimination against the female workers getting smaller.

2 Methodology

In the first step, we specify an adapted Mincerian equation for the logarithm of the hourly wage rate where the educational attainment is measured with the dummy variable degree.

where W is the hourly wage rate, D is the dummy variable that classifies graduates and non-graduates, X is the set of employee’s characteristics, e is error term.

In the second step, we estimate how much proportion of gender wag gap can be explained by higher education by exploiting Oaxaca (1973) decomposition method. This procedure splits the total gender wage differential into two components: the part of the differential attributable to gender differences in observable productive characteristics such as education and experience, and the residual gap attributable to differences in the male and female returns to these productive characteristics. This residual component of the wage gap is generally attributed to discrimination or differences in unobserved productive characteristics. More formally, the total gender wage gap is equal to

where and are the estimates of the natural log of male and female wages, from separate wage regressions by gender. and are vectors of the mean values of the male and female productive characteristics, and and are vectors of the estimated coefficient from the male and female wage regressions. This wage differentials can be decomposed as follows,

The first expression on the right-hand side is that portion of the wage differential attributable to differences in the average productive and other characteristics of men and women. The difference in the mean characteristics is multiplied by the estimated coefficients from the men’s regression. These coefficients are interpreted as the men’s wage structure, or the men’s return to these productive characteristics. The second expression on the right is that portion of the wage differential attributable to differences in the male and female regression coefficients; that is, differences in the returns to men and women for the same productive characteristics. The second term is generally attributed to discrimination.

A difficulty with this decomposition is that it values the difference in male in female productive characteristics according to the male returns . Performing the decomposition in this way assumes that the male wage structure is the wage structure that would prevail absent discrimination. The gross wage differential could also be decomposed as follows,

This expression shows that difference in the mean productive characteristics between male and female is valued according to the female return, which indicates that female wage structure would prevail in the absence of discrimination. Although these two decompositions are equivalent, they generally yield different estimates for wage gap decomposition. In our analysis, we assume that in Chinese labor markets the male wage structure is more likely to be non-discriminatory than the female structure, so we use first equation to decompose the gender wage gap.

3 Data

My empirical analysis relies on data from CHIP for the year of 1995 and 2013. The Chinese Household Income Project (CHIP) has conducted five waves of household survey, in 1988, 1995, 2002, 2007 and 2013 to track the dynamics of income distribution in China. These surveys were organized by Chinese and international researchers, with assistance from the China National Bureau of statistics. CHIP 1995 and 2013 data consist of both urban and rural surveys, but in this paper, we will focus on the urban sample because almost all of the university graduates choose to stay in urban areas when they finish their higher education degree. The sample is selected based on individuals aged 16-60 (state retirement age for men) for male, and 16-55 (state retirement age for women) for female. In our research, we restrict our sample to salaried employees, so we exclude self-employed individuals, retirees, students, the disabled, as well as respondents who provided incomplete information on wages, education, age or other variables.

Previous studies have shown that educational levels are negatively correlated with working hours. More formally educated individuals tend to work fewer hours on average, and accordingly the returns to education based on monthly wages may be underestimated. Therefore, we use hourly wage rate as our dependent variable in our estimation. CHIP does not include data about hourly wage rate, but it contains annual wage, monthly wage, and work days per week and work hours per day. With the above information, we can calculate the hourly wage rate.

Explanatory variable also includes a dummy variable, equal to one for communist party membership; a dummy variable equal to one for minorities and zero for Han Chinese; a dummy variable equal to zero of males and one for females; categorical variable of type of contract, equal to zero for short-term contract, one for permanent contract, two for long-term contract; categorical variable of occupational industry. Working experience is not observed in the data set, therefore, we define the variable of experience as age minus the year of education minus 6 years. Education variable is defined as a binary variable, equal to one if an individual receives four-year college education, zero if one receives only high school education. We exclude those individuals with some college education but less than four years. We control for experience and experience squared in estimation. The description of all variables are shown in Table 1.

Summary statistics of the data used in the analysis are reported in Table 2. Simple comparisons of average earnings between individuals with college education and those with only high school education reveal that there exist large wage differentials between education levels in both years. The data also shows that the number of college graduates increased dramatically from 1995 to 2013. In particular, the percentage of individuals receiving college education increased from 24% in 1995 to 56.72% in 2013. Table 2 also shows that men tend to earn more than their women counterparts for nearly 21% in 1995 and 21.15% in 2013.

4 Empirical Analysis

4.1 The returns to higher education in urban China

In this section, we present an estimate of the monetary rate of return of higher education in urban China. As expected, the OLS estimates show that an academic degree has a positive and significant relationship with wages. We first run a regression without occupational characteristics. Results are shown in Table 3. We can see that in 1995, the percent increase on wages for college graduates is 44.9% for females and 30.9% for males. In 2013, the impact of higher education increases to 57.7% for females and 68.6% for males. However, when we control additional variables related to occupation (Table 4), the impact of higher education on wage decreases but still significant. In 1995, the rate of return to higher education is 39.7% for females and 28% for males. In 2013, the rate of return to higher education is 46% for males and 51.6% for females.

Membership of communist party has a significant effect on wage rate in 1995 but not in 2013, which indicates that as free labor market develops and the scales of SOE shrinks in the economy, CCP membership does not play a big role in wage distribution.

Experience has positive sign on wage rates and squared experience has negative sign in the both two years, which indicates experience has an inverted-U shape relationship with wage. The relation is characterized by an increase in wages at a worker’s early age and then the wage is peaked at somewhere and followed by a decrease at the later stage. It can be explained that if a worker is in the labor market for a long time, the return of his experience is likely to fall as his skill might be out-of-date due to the rapid development of technology.

Types of contract is significant in 2013. Male workers who have long-term contract and have permanent contract generally earn 27.8% and 30.8% more than male workers who only have short-term contract respectively. OLS results imply that the rate of return to college degree is larger for females in both 1995 and 2013 compared with males, although the rate of return to college degree is large and positive for both men and women. In particular, the OLS results indicate a difference of about 11.7 percentage points in 1995 and that of 5 percentage points in 2013 of college premium between men and women.

4.2 Decomposition of the wage gap

Table 5 reports the result of decomposition of the estimated wage gap among genders. The dependent variable of the analysis is the log of wage rate. The total wage differential has been decomposed into two parts; the first part is due to the difference of genders’ endowments, and the second part is due to the differences in the parameters of the wage function which can be attributed to the labor-market discrimination and to other omitted variables.

The mean log hourly wage difference in our sample is 0.217 in 1995 and 0.2 in 2013. Then how much proportion of the difference is caused by discrimination? Table 5 shows that discrimination accounts for about 0.1 out of 0.217 in 1993. In other words, discrimination accounts for 46.1% of the estimated gender wage differential.

Since we are analyzing the male-female differential, a positive sign indicates that factor is working to increase the differential. On the other hand, a negative sign indicates a decrease the differential. For the explained part of decomposition, higher education degree accounts for 23.13% of the estimated male female wage gap.

For the estimates of 2013, discrimination accounts for 71% (=0.142/0.2) of the gender wage differential. College degree accounts for 0.017(marginal significant at 5% level) of the wage gap with a negative sign, which indicates that higher education narrows the gender difference on salaries by 8.5% (0.017/0.2).

Among all the characteristics, experience has the strongest power in the explained part, it explains about 59% and 54.5% of the gender wage gap respectively for 1995 and 2013.

5 Conclusion

This paper intends to analyze the how the college premium affect the gender wage differential in China’s urban labor market. To attain this, we identify the wage level and gender wage differential determinants using Mincerian earning function with gender dummy variable. Second, we exploits Oaxaca methods to analyze the composition of wage differential. In addition, we investigate how much of the total gender wage differential comes from sources such as discrimination and those due to the individual characteristics. The decomposition results show that discrimination apparently is an overwhelming reason for the low wages of females. Higher education degree plays an important role in increasing the gender wag gap in 1995, and is a contributor to gender discrimination in China’s urban labor market.

We should note that the unexplained differential is not an exact measure of discrimination, since there is an absence of detailed controls for all possible relevant factors of job characteristics and person-specific skills. Therefore, the magnitude of discrimination is likely to be overestimated. In the future research, we will include more specific worker characteristics to get more accurate results.

References

[1] Appleton, et al. Has China crossed the river? “The evolution of wage structure in urban China during reform and retrenchment”. Journal of comparative economics, 2005, pp390-396.

[2] Björn A. Gustafssonï¼ŒLi Shi. “Inequality and Public Policy in China”. Cambridge University Press, 2000.

[3] Chen, G.F., Hamori, S. “An empirical analysis of gender wage differentials in urban China”. Kobe University Economic Review, 2008, pp25-34.

[4] Emily N. Johnson, Gregory C. Chow. “Rates of Return to Schooling in China”. Pacific Economic Review, 1997, pp101-113.

[5] Ronald Oaxaca. “Male-Female Wage differentials in Urban Labor Market”. International Economic Review, 1973, pp694-695

Appendix

Table 1 Description of all the variables in the earnings equation

Variables

Description

lrwage

log of hourly wae rate

gender

0 for male, 1 for female

marr

0 for married, 2 for single

ethnc

0 for minor ethnicity, 1 for Han

memb

0 for others, 1 for communist party

leduc

0 for high school, 1 for college

expe

experience

expesq

squared experience

contract (reference: short-term)

type of contract

permanent

long-term

sector: (reference sector: government organization)

Reference group: government organizations, social organizations

Agriculture

Manufacturing

Mining

Construction

Transportation and communication

Commerce and trade

Real estate Public utility

Health and social welfare

Education culture

Scientific research information technology

Finance and insurance

Reference group: government organizations, social organizations

Agriculture

Manufacturing

Mining

Construction

Transportation and communication

Commerce and trade

Real estate Public utility

Health and social welfare

Education culture

Scientific research information technology

Finance and insurance

Agriculture

Manufacturing

Mining

Construction

Transportation & communication

Commerce & trade

Real estate & Public utility

Health & social welfare

Education & culture

Scientific research &

information technology

Finance & insurance

Table2Â Descriptive statistics of major variables

1995			2013
All	Male	Female	All	Male	Female
wage	2.7044	2.9411	2.4277	20.8758	22.586	18.6433
marr	.8681	.8569	.8828	.8775	.8854	.8672
leduc	.2400	.3137	.1535	.5672	.5520	.5870
ethnicity	.0335	.0319	.0344	.0443	.0433	.0458
memb	.2488	.3429	.1452	.3338	.3947	.2542
expe	19.1979	20.42	17.5127	19.1324	20.7409	17.0328
Observation	3422	1852	1570	3020	1710	1310

Table3Â Log wage estimates males and females

1995		2013
male	female	male	female
lrwage	lrwage	lrwage	lrwage
1.marr	0.0946	-0.0234	0.162*	0.0628
(1.73)	(-0.32)	(2.25)	(0.88)
1.ethnc	0.00329	-0.0135	-0.0236	-0.0875
(0.05)	(-0.16)	(-0.28)	(-0.93)
1.memb	0.0824**	0.183***	0.0686	0.0442
(2.78)	(3.91)	(1.80)	(0.92)
1.leduc	0.309***	0.449***	0.577***	0.686***
(10.56)	(9.79)	(13.67)	(14.00)
expe	0.0313***	0.0602***	0.0243**	0.0517***
(4.53)	(5.87)	(3.13)	(5.76)
expesq	-0.000317*	-0.000987***	-0.000416*	-0.00122***
(-2.16)	(-3.84)	(-2.50)	(-5.35)
_cons	0.220***	-0.0804	2.093***	1.780***
(4.36)	(-1.25)	(30.73)	(23.09)
N	1852	1570	1710	1310
=”* p<0.05Â p<0.01 Â * p<0.001″

Table 4 Log wage estimates males and females with occupation variables

1995		2013
male	female	male	female
lrwage	lrwage	lrwage	lrwage
1.marr	0.0869	-0.0408	0.133	0.0444
(1.58)	(-0.55)	(1.89)	(0.65)
1.ethnc	0.00443	-0.0323	-0.00162	-0.0786
(0.06)	(-0.38)	(-0.02)	(-0.87)
1.memb	0.0776**	0.154**	0.0439	0.0459
(2.58)	(3.23)	(1.12)	(0.95)
1.leduc	0.280***	0.397***	0.460***	0.516***
(8.66)	(7.79)	(9.79)	(9.26)
expe	0.0325***	0.0638***	0.0249**	0.0462***
(4.68)	(6.18)	(3.27)	(5.27)
expesq	-0.000341*	-0.00109***	-0.000439**	-0.00110***
(-2.30)	(-4.18)	(-2.69)	(-4.95)
1.contract	-0.0157	0.0210	0.278***	0.325***
(-0.16)	(0.21)	(5.22)	(5.44)
2.contract	-0.00246	-0.0455	0.308***	0.382***
(-0.02)	(-0.45)	(6.48)	(7.48)
1.sector	-0.0492	-0.267	-0.204	-0.209
(-0.56)	(-1.86)	(-1.09)	(-0.95)
2.sector	-0.0270	-0.137*	-0.00683	0.217*
(-0.63)	(-2.34)	(-0.11)	(2.54)
3.sector	0.0452	-0.142	0.195	0.00623
(0.43)	(-0.92)	(1.89)	(0.04)
4.sector	-0.113	-0.265*	0.0833	0.128
(-1.35)	(-2.07)	(0.95)	(0.98)
5.sector	0.0860	0.0505	0.109	0.196
(1.39)	(0.57)	(1.51)	(1.64)
6.sector	-0.0936	-0.122	-0.105	-0.0443
(-1.86)	(-1.92)	(-1.55)	(-0.62)
7.sector	0.116	-0.160	0.118	0.232*
(1.46)	(-1.57)	(1.56)	(2.24)
8.sector	0.0294	-0.0108	0.000305	0.137
(0.40)	(-0.12)	(0.00)	(1.51)
9.sector	0.0472	-0.0380	-0.0180	0.0207
(0.84)	(-0.51)	(-0.28)	(0.31)
10.sector	0.193**	0.0269	0.337***	0.365***
(2.64)	(0.24)	(4.17)	(3.57)
11.sector	0.236*	0.175	0.314***	0.391***
(2.37)	(1.66)	(3.48)	(4.06)
_cons	0.236*	0.00511	1.929***	1.620***
(2.25)	(0.04)	(22.66)	(16.27)
N	1852	1570	1710	1310
=”* p<0.05Â p<0.01Â Â * p<0.001″

Table 5Â Results of the decomposition

1995	2013
lrwage	lrwage
overall
group_1	0.901***	2.855***
(62.99)	(154.36)
group_2	0.684***	2.654***
(40.03)	(122.31)
difference	0.217***	0.200***
(9.74)	(7.03)
explained	0.117***	0.0582***
(9.91)	(3.48)
unexplained	0.0997***	0.142***
(4.68)	(5.37)
explained
marr	-0.000782	0.00182
(-0.59)	(1.19)
expe	0.128***	0.109***
(5.85)	(4.45)
expesq	-0.0871***	-0.0953***
(-4.33)	(-3.95)
leduc	0.0502***	-0.0170
(8.44)	(-1.91)
ethnc	0.0000354	0.0000996
(0.24)	(0.29)
memb	0.0199***	0.00605
(4.08)	(1.47)
contract1	-0.0000303	0.0145***
(-0.06)	(3.66)
contract2	0.000603	0.00627**
(0.35)	(2.79)
contract3	0.000650	-0.000476
(0.43)	(-0.22)
sector1	0.0000652	-0.00519*
(0.06)	(-2.26)
sector2	-0.00136	-0.000166
(-1.54)	(-0.17)
sector3	0.00145	-0.000873
(1.18)	(-0.39)
sector4	-0.000146	0.00125
(-0.36)	(0.85)
sector5	-0.00176	0.000511
(-1.53)	(0.33)
sector6	0.00130	0.00337
(1.38)	(1.21)
sector7	0.00614**	0.0235***
(2.69)	(4.18)
sector8	0.00000204	0.00198
(0.03)	(1.25)
sector9	-0.000107	0.000690
(-0.40)	(0.74)
sector10	-0.0000184	0.00870*
(-0.09)	(2.47)
sector11	0.00215*	0.00343
(2.02)	(1.53)
sector12	-0.00219	-0.00389
(-1.69)	(-1.74)
unexplained
marr	0.111	0.0771
(1.25)	(0.86)
expe	-0.582*	-0.379
(-2.35)	(-1.63)
expesq	0.313*	0.276*
(2.43)	(2.04)
leduc	-0.0231*	-0.0315
(-2.05)	(-0.74)
ethnc	0.00122	0.00343
(0.40)	(0.60)
memb	-0.0156	-0.000392
(-1.45)	(-0.02)
contract1	-0.0000788	0.0111
(-0.03)	(0.82)
contract2	-0.0303	-0.00315
(-0.56)	(-0.17)
contract3	0.00828	-0.0103
(0.57)	(-0.86)
sector1	-0.0126	0.00877
(-1.94)	(0.76)
sector2	0.00180	0.000446

Order Now