Women's Employment and Family Income Inequality during China's Economic Transition

Abstract

Economic reforms and trade liberalization have brought profound changes to the Chinese labor market. In this paper, we apply the technique of decomposing the coefficient of variation to examine the impact of changes in married women's employment and earnings on income inequality among Chinese urban households. Using the Chinese Household Income Surveys from 1988, 1995, and 2002, we explore the differences between two phases of economic transition: the gradualist reform period (1988–1995) and the radical reform period (1995–2002). Our analysis shows that the public-sector labor retrenchment of the late 1990s has led to a drastic decline in the employment rates of women, especially those married to low-earning husbands, and the change in women's employment was a major force driving income inequality in post-restructuring urban China.

Keywords:

• Income inequality
• assortative mating
• women's employment
• China
• JEL Codes: D13, J16, P21

Acknowledgments

We are grateful for the comments and suggestions of guest editors, style editors, and three anonymous referees. We also thank the Ford Foundation for its support for the postgraduate economic research and mentoring program for young Chinese women economists.

Notes

¹ While rural–urban disparity is an important source of income inequality in China, our analysis is confined to the urban sector. The data we used provide no information on earnings of wives and husbands in rural households. While farming is the dominant form of economic activity for the majority of married women in rural areas, the contributions of individual members to household production are unobservable in monetary terms.

² For China's labor-market reforms in the late 1990s and its gender outcomes, see Simon Appleton, John Knight, Lina Song, and Qingjie Xia (2002); Xiao-yuan Dong, Jianchun Yang, Fenglian Du, and Sai Ding (2006); John Giles, Albert Park, and Fang Cai (2006); and Margaret Maurer-Fazio, James Hughes, and Dandan Zhang (2007).

³ Due to data limitations, our analysis is confined to the financial aspect of women's labor-force participation, while recognizing the importance of unpaid activities for human well-being.

⁴ Amin and DaVanzo (2004) provide a summary of the empirical findings of studies on this subject.

⁵ Michael M. Kidd and Xin Meng (2001) find that female workers in Chinese state enterprises earned around 86 percent of the average wage received by men in 1981, and the ratio rose slightly to 87 percent in 1987. As a point of reference, the average nonagricultural earnings ratio of female to male is 76 percent in the US, 57 percent in South Korea, and 51 percent in Japan (Joyce P. Jacobsen 1998: 350).

⁶ Xiao-yuan Dong and Louis Putterman (2003) estimate that 20 to 40 percent of workers in state firms were redundant in the early 1990s.

⁷ For the welfare reforms, see Gang Fan, Maria Rosa Lunati, and David O'Connor (1998).

⁸ According to official statistics, the number of childcare centers and kindergartens fell by 14 percent in the urban sector between 1995 and 2003 (China's Ministry of Education 2005).

⁹ Investigating the causes of unemployment in post-restructuring urban China, Du and Dong (2008) find that a person is more likely to be unemployed when one or more other family members (primarily spouse) are also unemployed.

¹⁰ CV is defined as the standard deviation of income divided by the arithmetic mean.

¹¹ While Gini coefficients have most of the desirable properties of CV, they cannot be conveniently decomposed (Maria Cancian and Deborah Reed 1998).

¹² The inclusion of other sources of income should not alter the main results presented below since income from sources other than husbands' and wives' earnings only accounts for a small proportion of total household income in our sample.

¹³ The original decomposition formula is in Reuben Gronau (1982), and the discussion for equation (2) is from Cal R. Winegarden (1987).

¹⁴ Mathematically, the second derivative of CV_HH ² in (1) is greater than zero, indicating that the CF2 is a U-shaped curve.

¹⁵ Sichuan province has a per capita income close to the national average, and hence adding this province to the samples of 1995 and 2002 should not bias the comparison.

¹⁶ Because the CHIS samples are drawn from the NBS's parent sample, they inherit some of the problems of the NBS sample (see Shaohua Chen and Martin Ravallion [1996]). One limitation for our purposes is that the sample over-represents the public sector and excludes migrant workers, creating a downward bias in the estimate of income inequality.

¹⁷ The upper limits of age are set in conformity with China's gender-differentiated retirement policy that we explained earlier. As a sensitivity check, we also performed the decomposition analysis to a sample with husbands and wives both aged between 16 and 60, and the results are substantively similar to those reported in Tables 8 and 9.

¹⁸ The sample for 1988 is larger than those for 1995 and 2002 due to funding cuts in the later years.

¹⁹ A person is considered employed if he/she selects the category of “have a job” in the question regarding employment status in the survey. Earnings of husbands and wives are defined as the sum of wages (or salaries), non-wage compensation, net income of self-employment, and non-monetary benefits minus income tax. Other sources of income include the earnings of other family members, assets income (rent, interests, and dividends), and transfer income. All income measures are discounted by urban consumer price indexes with 2002 as base year. It is noteworthy that we omit outlying observations in earnings that are likely attributable to recalling errors or data-entering mistakes. Because the mistakes are random, the omission of those outlying observations should not bias inequality estimates (Jeffrey M. Wooldridge 2005).

²⁰ The pattern of educational attainment is more indicative of the change in marriage norms than that in spousal earnings since the majority of urban Chinese people entered the labor market and got married only after completing education.

²¹ The stable earnings differentials between working wives and husbands are in line with the trends in sex segregation presented in Table 5. Using data derived from two urban labor-force surveys in 1997 and 2002, Dong et al. (2006) find that public-sector downsizing has led to a dramatic shift of employment from the formal sector to the informal sector. In contrast to the trends of the feminization of informality following structural adjustments in many developing countries, this study shows that the gender effect of informalization in urban China was small, and full-time positions (defined as forty hours or more work per week) were still the predominant mode of employment for both men and women in 2002. We are unable to examine the changing patterns of employment in this paper because the CHIS provides no information on working hours or informality.

²² For instance, the CV of wives' earnings is about twice that of husbands' in Italy (0.80 versus 1.96) (Del Boca and Pasqua 2003) and in the US (0.65 versus 1.27) (Cancian and Reed 1998).

²³ In the face of the rising positive correlation of spouses' earnings in this period, why was the effect of correlation changes so miniscule? This is because the Chinese government increased transfer payments to mitigate the income losses of laid-off workers after public-sector downsizing, which weakened the correlation of wives' earnings with other sources of income, offsetting the impact of the rising positive correlation between spouses' earnings.