J Popul Econ (2000) 13: 331±352 999 2000 Sectoral gender wage di erentials and discrimination in the transitional Chinese economy Pak-Wai Liu1, Xin Meng2, Junsen Zhang1 1 Chinese University of Hong Kong, Department of Economics, Shatin, N.T., Hong Kong (Fax: 852-2603-5805; e-mail: jszhang@cuhk.edu.hk) 2 Australian National University, Department of Economics, Research School of Paci c and Asian Studies, Canberra, Australia Received: 5 November 1997/Accepted: 10 January 2000 Abstract. China's economic reform has a ected various ownership sectors to di erent degree. A comparison of gender wage di erentials and discrimination among individuals employed in the three sectors ± state sector, the collective sector, and the private sector ± provides information on the impact of economic reform. Two Chinese data sets from Shanghai and Jinan are used to examine the gender wage gap across the three sectors. It is found that privatization/ marketization of the economy leads to larger wage di erentials as human capital characteristics are more appropriately rewarded. Both data sets show that the relative share of discrimination in the overall gender wage di erential declines substantially across ownership sectors from the state to the private. The increase in gender wage di erential due to marketization is much larger than any increase in di erential that may arise from more gender discrimination. JEL classi cation: J16, J71, P20 Key words: Gender wage di erentials, discrimination, China's economic reform 1. Introduction There has been a recent urry of papers examining cross-country di erences in wage structure (see e.g., Blau and Kahn 1995 and Katz et al. 1995) as well as the changing pattern of wage structure in a single country over time (see Blau All correspondence to Junsen Zhang. We are indebted to Thomas Mroz and two referees for many helpful comments and suggestions, and to the Chinese University for partial nancial support through a direct research grant. Responsible editor: Thomas Mroz.
332 P.-W. Liu et al. and Kahn 1995). One of the major thrusts of the literature is a rekindling of interest in the role of labour market institutions. Card and Freeman (1993) suggest that comparisons of wage structures provide the basis for a `natural experiment', highlighting the potential role of inter-country di erences in labour market institutions and the changing role of a given country's institutions. While major institutional changes are taking place in the former centrally planned economies, the literature is dominated by the analysis of industrialized nations, such as the US, the UK and parts of Western Europe. In particular, there is no study of the impact of institutional changes on the wage structure, especially the gender wage gap, in China.1 The present paper lls the gap in the literature by analyzing how gender wage di erentials and especially gender wage discrimination (i.e. the di erential net of gender di erences in endowments) vary across various ownership sectors. The last two decades have seen a gradual economic transformation from a centrally planned to a market oriented economy in China. Before economic reform, the communist egalitarian ideology led to a rather equal distribution of income across gender groups in the economy (see Meng and Miller 1995). Neo-classical theory suggests that gender wage discrimination may be due to employers' or employees' personal taste (Becker 1957). This personal taste may be traced back as a cultural in uence. Hence, from a theoretical viewpoint, one might expect an increase in wage inequality as central planning gives way to free-market reforms and wages are no longer determined by egalitarian ideology. Such an expectation, however, ignores that economic reforms also increased market competition, which may reduce discrimination. Furthermore, over time employers' personal tastes or the so-called `cultural in uence' may have been changed by a new institutional setting which is di erent from the traditional setting. Speci cally, if after living under the socialist regime for many decades, both employers' and employees' personal tastes have gradually changed toward equal treatment of di erent gender groups, the introduction of free-market system may not bring back the wage inequality that existed before the communist take-over. Thus, whether economic reform brings greater gender wage discrimination in China is largely an empirical question. It is interesting to know whether and how economic reform has a ected the gender wage di erential and discrimination. The ideal way to examine the impact of economic reform on the gender wage di erential and discrimination would be to have two ``comparable'' data sets representing pre- and post-reform situations. However, such comparable data sets are di½cult to construct and are often controversial. Instead, we take a di erent approach by utilizing the institutional and data features of various ownership sectors in a transitional economy undergoing economic reform. As economic reform a ects various ownership sectors to di erent degree, a comparison of gender wage di erentials between di erent ownership sectors can shed some light on the issue. Given that the state sector is the least market oriented, the collective sector is more liberalized, and the private sector has the most free labour market institutions, the comparison of gender wage differentials and discrimination among individuals employed in the three sectors will provide some information on the impact of economic reform. There are several studies of gender wage gaps in former centrally planned economies. Meng and Miller (1995) look at gender wage gaps in China using data from a sample survey of township-, village-, or privately-owned (TVP)
Wage di erentials and discrimination in China 333 enterprises.2 According to the channel of obtaining a job, Meng (1998) classi es employees of TVP enterprises into a market group and a non-market group. It is found that gender wage discrimination accounts for a smaller share in the market group than in the non-market group. Newell and Reilly (1996) analyze the gender wage gaps in Russia. However, none of these papers explores the impact of economic reform on gender wage gaps. This present paper is the rst one which examines the relationship between gender wage gaps and di erent rm ownership systems. 2. China's economic reform and institutional features After the launching of economic reform in 1978, the Chinese government adopted a series of policy and institutional changes aimed at increasing the productivity of the economy. While the initial e orts focused on agriculture, the government gradually shifted focus to improving the performance of the non-agricultural sectors. Prior to the reform, these sectors were heavily dominated by state-owned rms operating under central planning. After the reform many decisions were decentralized to the rm level, or at least to the local government level. Production and allocation plans were reduced in quantity and coverage. By 1985 most rms, for example, could change output quantity and variety, production technology, and the timing of production. Beginning in 1986, most newly hired regular workers in state-owned enterprises were given xed-term contracts. Even though the central planning apparatus remained in place, planned inputs and outputs became increasingly small fractions of the total inputs and outputs of rms. Outside-of-plan inputs and outputs could be traded freely among rms. There were also increasing attempts to relax control over the prices for most planned products. In October 1992, a new reform agenda was rati ed by the 14th Party Congress, which proclaimed that China would adopt a ``socialist market economy'' ( Naughton 1995). Much progress was made in 1992±1993 in making the labor system more exible for the 76 million state enterprise workers and the 35 million urban collective workers ( Naughton 1995). Firms were given more autonomy and discretion in wages, bonuses, quits, layo s, and promotion. Some rms even tried to eliminate the ``iron rice bowl'' (the permanent employment system) altogether, shifting workers onto a contract basis. Workers were also given much more freedom in resigning and changing jobs. Before the economic reform, it was extremely di½cult for workers in the state and collective sectors to change jobs (this was one of the unique features of a planned economy). Jobs were often ``assigned'' by government agencies of various levels, and changing of jobs was not simply a matter of individuals' free choice. Hence, economic reform not only gives rms more freedom in wage setting practices, but it also gives workers greater job mobility. As a result, rms of di erent ownership compete for good workers, and workers compete for good jobs in the various sectors. Another key aspect of the economic reform was to make it easier for local governments, collectives, and even private households to set up their own rms outside of the state planning structure. Private and collective enterprises responded swiftly to economic liberalization, and there was rapid growth in the number and output of such rms in response to the new exibility. The former quickly expanded in the retail sector, while the latter became much
334 P.-W. Liu et al. more important in the industrial sector. Not only did these rms produce badly needed goods and services, but they also introduced competitive pressure into the economy, undermining the monopoly power of state-owned rms. For example, the share of national non-agricultural output of the state, collective, and private rms3 changed dramatically from 76, 23.5, and 0.5% in 1980 respectively to 34, 36.6, and 29.4% in 1995 (State Statistical Bureau 1996). In some special economic zones, the proliferation of private rms has led to the domination of such rms in the local economies. As a result of economic reform, one of the main characteristics of the transitional Chinese economy is the coexistence of state, collective, and private rms. State rms used to be tightly controlled by the government planning and administrative apparatus, generally large in size and accounting for large shares of industrial output, assets and so forth (Byrd 1992). Collective rms are either urban collectives nominally owned by urban community entities but actually controlled by local government agencies, or rural collectives owned by township and village governments. Private rms include rms owned by individuals, joint ventures and foreign investors. Despite the economic liberalization since 1979, state rms are still subject to more bureaucratic and administrative intervention/barriers than collective and private rms. 3. The impact of economic reform on gender wage di erentials and discrimination The question of main concern in the present paper is how economic reform a ected the gender wage di erential and discrimination. The impact on the overall gender wage di erential is simple. The communist egalitarian ideology imposed on the Chinese economy before 1979 did not allow much income inequality, and wage dispersion due to human capital characteristics was suppressed. Hence, privatization/marketization of the economy after the reform would allow larger wage di erentials if human capital characteristics are appropriately rewarded as in the market economies. Given that rms under the state ownership system are less privatized/marketized than the collective or the private rms, we expect that the gender wage di erential is the lowest in the state sector, and highest in the private sector. How economic reform a ects discrimination is more involved. On the one hand, according to the economic theory of discrimination by Becker (1957) and Arrow (1972), short-run discrimination is caused by personal taste of individuals, including employers, employees, and consumers. Personal taste can, in turn, be traced to cultural in uence. For example, the Chinese culture discriminates against women more than many other societies mainly due to the patriarchal Confucian tradition. Thus, employers with such a cultural background may pay female employees less than their male counterparts with similar productivity. As a result, it is expected that private/collective rms will exhibit larger wage discrimination since they have more autonomy than state rms in wage setting. On the other hand, economic theory also suggests that under the perfectly competitive market condition, gender wage discrimination will not sustain. This is because those employers, whose personal preference does not discriminate against women, will be able to employ female workers who are as productive as male ones, and hence lower their unit cost of labor. Such labor cost
Wage di erentials and discrimination in China 335 advantage will allow them to compete successfully against those employers who discriminate against women. In the long-run, therefore, market competition will eliminate gender wage discrimination in the market place (see, for example, Arrow 1973). Thus, while the relaxation of strict government egalitarian control and more freedom in wage setting and employment may increase discrimination, more market competition between and within ownership sectors can reduce discrimination. Therefore, whether economic reform brings greater gender wage discrimination in China is largely an empirical issue. The earlier discussion about economic liberalization of Chinese state rms also suggests that the impact of economic reform on gender wage gaps that we estimate is not the full impact. If state rms had not been reformed at all, then a comparison of state and private rms would have revealed the full impact of privatization. However, given that state rms have also been reformed to a certain extent toward greater marketization even though they remain less autonomous than private/collective rms, a comparison between state and private/collective rms will only reveal the direction and lower bound of the impact of marketization and privatization on gender wage gaps in China. This re ects the transitional nature of the Chinese economy. As discussed earlier, economic reform a ects not only rms' wage setting practices but also workers' job mobility. The latter enables more self-selection of workers among the ownerships sectors. Thus, it should be noted that gender wage discrimination across the sectors can also be a ected by sectoral di erences in self-selection. However, for self-selection to a ect gender wage discrimination, self-selection must be gender-speci c, i.e. males and females who self-select into working in di erent sectors have di erent unobservable characteristics which a ect wages.4 In the absence of evidence or theory on gender-speci c self-selection, we believe that di erent wage setting practices across the sectors are the main driving force. 4. Data description This paper uses two data sets which suit our purpose. Both data sets include a rich set of variables that are normally viewed as earnings determinants, and most importantly, both have information on the ownership structure of the rm (state, collective or private) that an individual employee works for. The rst data set is from a sample survey in Shanghai conducted by the Institute of Population Studies, Shanghai Academy of Social Sciences in early 1996, and the second data set is from a rural-urban migration sample survey in Jinan conducted by the Institute of Population Studies of the Chinese Academy of Social Sciences in collaboration with the Shangdong Statistical Bureau in July 1995. Located in the east coast, a well-known nancial and industrial center, Shanghai is one of the largest cities in China. In 1995 it has a population of about 13 millions and covers 6342 square kilometres (Shanghai Statistical Bureau 1996). As the capital city of Shandong province which is also in the east coast, Jinan has 5.3 million population with an area of 8227 square kilometres in 1995 (Shandong Provincial Government 1996). The real annual income per capita for urban households was 7196 and 4265 yuans in Shanghai and Shandong respectively in that year. In terms of the real income per capita
336 P.-W. Liu et al. in 1995, Shanghai and Shandong ranked the rst and twelfth respectively among China's thirty municipalities and provinces (State Statistical Bureau 1996). The Shanghai data set is from a survey of urban residents and oating population in early 1996, with a sample of 3000 individuals aged 15 to 64. This survey comprises individuals who work in the state, collective, and private sectors. The survey was designed to cover 2500 local residents and 500 migrants. Strati ed, multi-stage sampling was used in the survey. Districts/ counties (hereafter both referred as ``district'' for simplicity) were used as the rst-stage sampling units. The street communities (SCs) and resident committees ( RCs) were used, respectively, as the second- and the third-stage sampling units. In the RCs selected, 250 local residents and 50 migrants were randomly selected for interview. This sampling proportion of 5 local residents to 1 migrant was based on the overall composition of the Shanghai population. Hence, the sample was representative of the Shanghai population. The Jinan rural-urban migration sample survey covered 1504 individuals over 15 years old who have migrated from rural areas to Jinan and got a job in the urban labour market. Multi-strata non-equal proportional sampling method was adopted. The sampling procedure was arranged in three stages. First, four districts were chosen to represent the di erent patterns of development within Jinan proper. Second, adopting the sampling method with probability proportional to sizes, 3 resident committees were selected in each district from a total of 68 resident committees in the four districts, yielding a total of 12 resident committees, and the sample size in each committee was also calculated. Third, by using the occupational structure of rural migrants in Jinan as reference, the sample size in each occupation in each district was determined and within each occupation in each district, using random sampling, the sample members were selected in the resident committees. Responses were double-checked in the eld to eliminate errors and omissions. The quality of data entry was strictly controlled by logical checks and double data entry. These methods of data processing resulted in a high quality data set which can be a good representative of rural-urban migrants in Jinan proper. Note that the Jinan data set consists of only migrants. This may seem to be a limitation but we believe that it is an advantage in the following sense. It has been suggested that there is di erential availability of jobs in the various sectors to migrants and non-migrants in the Shanghai data set, and this makes it di½cult to assess whether any sectoral gender wage pattern is due to sectoral di erences in wage policies or to the di erential availability of jobs.5 Hence, the Jinan data set naturally eliminates the problem of di erential job availability across sectors for migrants and non-migrants. As we will show later, the same pattern of gender discrimination across the sectors is observed in both Jinan and Shanghai data sets, thus indicating that the pattern of gender wage discrimination is more likely to be caused by sectoral di erence in wage policies, rather than by di erential availability of jobs to migrants and nonmigrants across the various sectors. 5. Methodology and model speci cation A general Mincer (1974) type human capital earnings equation is speci ed as ln w ˆX b u, where w is monthly earnings,6 X a vector of individual
Wage di erentials and discrimination in China 337 characteristics and u the random error term. Blinder (1973) and Oaxaca (1973) developed similar decomposition approaches to partition the gender wage di erential into components caused by two factors: a di erence in productivity, and an ``unexplained'' component that is often referred to as discrimination. The Blinder-Oaxaca approach requires estimation of human capital equations for males and females separately. Their decomposition of the wage di erential can then be written as: ln W m ln W f ˆ X m X f ^b m X f ^b m ^b f 1 where ^b are the OLS estimate of the parameters b from the human capital equation, and a bar over a variable denotes the mean value. On the right hand side of (1), the rst term is attributable to di erent endowments (productivity), and the second term is the unexplained wage di erential due to di erences in coe½cients, which is often attributed to discrimination. The second term includes di erences in both slopes and intercepts. (A further decomposition between the slopes and intercepts is not appropriate (Jones 1983).) A practical consideration associated with the adoption of the Blinder- Oaxaca approach is the index number problem. This refers to the fact that the decomposition of the gender wage gap is not unique. In equation (1) the weights used for the rst and second terms are ^b m and X f respectively. This is sometimes referred to as a male-weighted decomposition, in the sense that the current male wage structure would be adopted in the absence of discrimination. These weights can be replaced by ^b f and X m to yield a femaleweighted decomposition. Cotton (1988) argues that derived weights using averages of the male and female coe½cients will be more accurate than the other estimations.7 Under this approach the non-discriminatory wage structure, b, is de ned as ^b ˆ f m ^b m f f ^b f, where f m and f f are the proportions of men and women among total employees. We will report maleweighted, female-weighted and Cotton decomposition so as to show the range of the discrimination e ect. In terms of reality and policy, if the male wage structure is used for adjusting female wages, then the male-weighted decomposition would be relevant. If female wages are adjusted upward, but male wages are adjusted ``downward'' through eliminating certain privilege or lower wage/bene t growth than otherwise, then Cotton's weighted average would be more relevant. While we try to keep the inclusion of variables in X as consistent as possible across the two data sets, the choice of variables in X is partly constrained by the availability of variables in each data set. For the Shanghai data set, the vector of X includes years of schooling, rm tenure and its square, other job experience and its square, duration of training, and three dummy variables for communist party membership, for being a cadre, and for males.8 For the Jinan migrant data set, the vector of X includes years of schooling, city work experience and its square, other job experience and its square, duration of training, and four dummy variables for the place of origin, for marital status, for being a former farmer, and for males. The de nition of variables is given in the data appendix. The summary statistics of the variables for each data set are presented in Table 1 and 5. In Shanghai, the mean log earnings di erential between males and females are 0.29, 0.38 and 0.39 in state, collective and private sectors respectively. Thus, the gender wage di erential was similar in collective and
338 P.-W. Liu et al. private sectors, but it was about 10 percentage points higher than the di erential in the state sector. In Jinan, the gender di erentials are 0.20, 0.30 and 0.42 respectively for the three sectors, indicating the di erential increases by about 10 percentage points from the state to the collective sector, and from the collective to the private sector. Therefore, economic reform, more speci cally marketization and privatization of economic activities, seems to widen the overall gender wage di erentials.9 These wage di erentials, however, do not necessarily re ect gender wage discrimination as the latter can also be caused by the gender personal endowment di erentials. Table 1 shows that the average education is slightly higher in the state sector than in collective/private sectors in the Shanghai data. Men and women seem to have almost identical educational attainment on average in the state sector, while men in the collective/private sectors have one more year of schooling than women. The rm tenure pattern is expected, longest (shortest) in the state (private) sector, and longer tenure for men than for women. The pattern in other job experience is expectedly di erent; individuals in the private sector have more experience in other jobs than those in the state sector. It is also obvious that a larger proportion of male employees are communist party members and/or cadres in comparison to female employees in any sector. Most individuals were married. Some similar patterns in endowments can also be noted in the Jinan data set (Table 5), but the gender or sector di erential is not as obvious as in the Shanghai data set. This is because workers in the Jinan data were migrants with more homogenous characteristics. A higher proportion of men were married than women, re ecting that married men were more mobile and willing to take risk in migrating to more pro table urban areas than married women. Furthermore, female migrants were about ve years younger than male ones whose average age is 26. Not surprisingly, these migrant workers were much younger than workers in the Shanghai data. It is interesting to note that workers in the state sector have higher earnings than those in the other two sectors in Shanghai, whereas the opposite is true for men in Jinan. The main reason is di erent types of jobs or occupations held by migrants and non-migrants. In the Shanghai data, the majority were urban residents and many occupied important positions in the state sector. The fact that the state sector had higher average earnings indicates that the state sector was, on average, more attractive to workers. On the other hand, the Jinan data set consists solely of migrants. Their jobs or occupations, especially in the state sector, were at a lower level than those held by urban residents. Hence the average earnings in Jinan are of a di erent sectoral pattern from those of Shanghai. 6. Empirical ndings and discussion 6.1. Evidence from Shanghai Table 2 provides the results from wage equation estimation using the pooled sample of males and females for the three sectors. White's (1980) heteroskedasticity-consistent covariance matrix was used in calculating standard errors for the estimated coe½cients. The general picture is that the rate of return to labor productivity related variables (except total days of training) is
Wage di erentials and discrimination in China 339 Table 1. Summary statistics for Shanghai sample State Collective Private Males Females Males Females Males Females Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Age 42.98 8.64 39.74 8.07 42.06 8.13 41.71 8.61 35.72 9.54 33.69 10.14 Years of schooling 10.84 3.04 10.70 2.76 9.24 2.23 8.69 2.78 9.82 3.19 8.72 3.54 Firm tenure 19.18 10.74 15.67 8.71 14.32 11.10 13.83 9.59 7.23 8.59 6.56 8.67 Firm tenure 2 483.02 452.40 321.16 293.61 327.64 413.02 282.91 316.26 125.65 257.04 117.63 244.17 Other job experience 3.75 7.56 3.52 7.35 6.19 9.47 5.42 9.46 7.23 9.63 5.76 8.18 Other job experience 2 71.19 200.48 66.36 181.30 127.39 312.88 118.47 305.04 144.56 276.56 99.71 200.05 Total days of training 61.77 158.40 49.61 139.47 44.73 109.52 24.39 64.43 36.23 128.97 37.99 136.98 Dummy for party member 0.24 0.42 0.12 0.32 0.24 0.43 0.09 0.28 0.15 0.36 0.06 0.24 Dummy for cadre 0.09 0.29 0.05 0.21 0.16 0.36 0.04 0.20 0.01 0.10 0.03 0.16 Dummy for married 0.99 0.08 0.99 0.07 1.00 0.00 0.99 0.06 0.88 0.32 0.87 0.34 Log(monthly earnings) 6.86 0.45 6.57 0.46 6.69 0.42 6.31 0.46 6.67 0.63 6.28 0.61 Monthly earnings 1052.19 497.45 791.38 366.72 880.86 429.60 604.42 266.23 957.80 646.95 636.28 399.38 No. of observations 865 599 160 278 181 150
340 P.-W. Liu et al. Table 2. Human capital equations by sector in Shanghai State Collective Private Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Years of schooling 0.0320 8.03 0.0418 5.15 0.0455 4.15 Firm tenure 0.0145 3.34 0.0354 5.27 0.0504 4.18 Firm tenure 2 0.0003 3.29 0.0009 5.29 0.0011 2.77 Other job experience 0.0241 5.22 0.0166 2.80 0.0273 2.81 Other job experience 2 0.0006 3.74 0.0004 2.61 0.0008 2.37 Total days of training 0.0002 2.70 0.0005 2.00 0.0003 1.42 Dummy for party 0.0643 2.04 0.1006 1.62 0.2528 2.10 Dummy for cadre 0.1022 2.20 0.1020 1.34 0.1801 0.75 Dummy for sex 0.2661 11.26 0.3373 8.07 0.3400 5.27 Constant 6.0442 94.80 5.6541 64.54 5.6010 51.54 No. of observations 1464 438 331 Adjusted R 2 0.17 0.30 0.24 higher in the private sector than in the other two sectors. The return to education seems somewhat low in all sectors, only around 3 to 4.5%.10 It is interesting to note that the variables which appear to be non-related to labor productivity have either no impact or opposite impact for the private sector wage determination as compared to the state sector. For example, the dummy variable for being a communist party member has a positive impact in the state/collective sectors, but has a negative impact in the private sector. In the state/collective sectors, political factors are often accounted for in pay schemes. In contrast, party membership does not count in the private sector, since those people may be good at politics but they may not be used to the competitive market. Further, the coe½cient on the sex dummy is much higher for the private and collective sectors than for the state sector, suggesting a smaller gender wage di erential in the state sector.11 Finally, note that the adjusted R 2 is much higher in the non-state sectors. More restrictive government control in the state sector reduces the importance of human capital variables in explaining earnings di erentials. Poor ts of the Mincerian functions and relatively modest ( pecuniary) returns to education are consistent with the evidence from Russia in 1992 (Newell and Reilly 1996) and a recent study of the return to schooling in China (Johnson and Chow 1997). The above wage equation is then estimated for the two gender groups separately within each sector and the results are reported in Table 3. The differences in the wage structure for the total sample among the three sectors basically hold for the sample of male employees. For females, however, the rate of return to education is the highest in the state sector and lowest in the private sector. The rate of return to rm tenure and to other job experience, on the other hand, is much higher for the private sector than for the state and collective sectors. The results of the wage regression for male and female samples are then used to carry out the Blinder-Oaxaca decomposition. Table 4 reports the decomposition of the overall wage di erential into the explained and unexplained components in terms of log di erentials, monetary values, and proportions. In terms of monetary values, the unexplained amount is smaller in the state/collective sector than that in the private. For example, under the Cotton decomposition, the unexplained amount is about 249, 248, and 275 yuans respectively in the state, collective and private sectors.
Wage di erentials and discrimination in China 341 Table 3. Human capital equations by sector and gender in Shanghai State Collective Private Males Females Males Females Males Females Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Years of schooling 0.0258 5.18 0.0433 6.44 0.0525 3.59 0.0411 4.12 0.0527 3.12 0.0350 2.44 Firm tenure 0.0077 1.43 0.0308 3.64 0.0346 3.71 0.0331 3.47 0.0380 2.23 0.0622 3.55 Firm tenure 2 0.0002 1.61 0.0008 3.27 0.0009 3.94 0.0009 3.26 0.0008 1.56 0.0013 2.17 Other job experience 0.0263 4.69 0.0195 2.34 0.0234 2.80 0.0102 1.24 0.0302 2.24 0.0203 1.34 Other job experience 2 0.0007 3.67 0.0004 1.23 0.0005 2.23 0.0003 1.34 0.0010 2.32 0.0004 0.65 Total days of training 0.0002 2.23 0.0002 1.23 0.0003 1.10 0.0008 1.85 0.0002 0.51 0.0006 1.76 Dummy for party member 0.0814 2.20 0.0360 0.61 0.1604 2.01 0.0024 0.03 0.3165 1.85 0.0634 0.29 Dummy for cadre 0.1601 2.96 0.0462 0.52 0.1102 1.20 0.0338 0.26 0.2424 0.52 0.0950 0.35 Constant 6.4286 81.15 5.8189 52.82 5.8601 38.88 5.6945 51.78 5.9543 34.64 5.6195 41.26 No. of observations 865 599 160 278 181 150 Adjusted R 2 0.11 0.08 0.27 0.14 0.09 0.25
342 P.-W. Liu et al. Table 4. Blinder-Oaxaca decomposition for Shanghai Male-weighted Female-weighted Cotton Value % Value % Value % State Total di 0.285 100 0.285 100 0.285 100 [260.81] [260.81] [260.81] Explained 0.022 8 0.002 1 0.012 4.3 [20.87] [ 2.608] [11.21] Unexplained 0.263 92 0.288 101 0.273 95.7 [239.94] [263.42] [249.60] Collective Total di 0.382 100 0.382 100 0.382 100 [276.44] [276.44] [276.44] Explained 0.062 16 0.025 7 0.039 10.1 [44.23] [19.35] [27.92] Unexplained 0.320 84 0.357 93 0.343 89.9 [232.21] [257.09] [248.52] Private Total di 0.393 100 0.393 100 0.393 100 [321.52] [321.52] [321.52] Explained 0.042 11 0.074 19 0.056 14.2 [35.37] [61.09] [45.66] Unexplained 0.351 89 0.319 81 0.337 85.8 [286.15] [260.43] [275.86] Note: The main entries under the column ``value'' represent the log differential decomposition, while the numbers in the square bracket are the corresponding monetary values (yuans). See Table 1 for average earnings by sector/gender. It also indicates that most of the gender wage di erential in the monetary value is unexplained, suggesting a substantial level of wage discrimination in each sector. In the following discussion of comparison of results across the various sectors, we put our primary emphasis on the log wage di erential and especially on the proportion that is explained and unexplained.12 Table 4 shows three important di erences between state and collective/ private rms. First, the explained log di erential due to endowments increases from state to collective/private rms, as discussed before. Second, the unexplained log di erential also increases, suggesting that economic liberalization increases wage discrimination. Finally, Table 4 indicates that although the non-state sectors have a larger log gender wage gap than the state sector, a greater proportion of such gap can be explained by di erences in employees' endowments. For example, under the Cotton weighting, only 4.3% of the total log wage gap in the state sector can be explained by di erences in personal characteristics, whereas the explained portion accounts for about 10 and 14% of the total log wage gaps for the collective and private sectors, respectively. The decomposition results indicate that in Shanghai, while more privatization increases wage discrimination in absolute terms, it reduces the proportion of discrimination in the overall gender wage gap. As mentioned earlier, Meng (1998) classi es employees of TVP enterprises into a market group and a nonmarket group and nds that gender wage discrimination accounts for a smaller share in the market group than in the non-market group. Thus, our
Wage di erentials and discrimination in China 343 Table 5. Summary statistics for Jinan sample State Collective Private Males Females Males Females Males Females Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Age 25.73 7.43 20.52 2.61 26.91 7.81 21.43 5.05 27.85 8.82 20.02 3.77 Years of schooling 9.18 1.77 8.78 1.69 8.63 1.82 8.74 1.78 8.80 1.73 8.45 1.84 City work experience 2.81 2.25 2.01 1.62 2.34 2.51 1.77 1.59 2.43 1.85 1.74 1.29 City work experience 2 12.95 23.83 6.64 12.20 11.76 40.91 5.62 12.35 9.32 13.77 4.68 8.35 Other job experience 6.79 7.53 2.82 2.34 9.01 8.00 3.98 5.54 9.70 8.45 2.92 3.42 Other job experience 2 102.73 305.32 13.43 22.87 145.15 262.51 46.21 205.79 164.99 269.70 20.11 87.37 Total training days 27.26 72.08 33.59 92.68 25.68 95.55 12.17 38.19 30.07 71.50 9.64 30.28 Dummy for origin 0.97 0.18 0.97 0.16 0.97 0.18 0.97 0.17 0.93 0.25 0.97 0.18 Dummy for married 0.49 0.50 0.07 0.25 0.56 0.50 0.12 0.32 0.54 0.50 0.04 0.20 Dummy for former farmer 0.62 0.49 0.40 0.49 0.63 0.48 0.53 0.50 0.70 0.46 0.57 0.50 Log(monthly earnings) 6.07 0.27 5.87 0.29 6.16 0.26 5.86 0.25 6.16 0.26 5.74 0.32 Monthly earnings 448.39 128.53 366.80 102.02 488.00 126.07 360.93 87.95 491.23 145.46 326.84 116.31 No. of observations 336 147 443 104 150 125
344 P.-W. Liu et al. Table 6. Human capital equations by sector in Jinan State Collective Private Coef. T-Ratio Coef. T-Ratio Coef. T-Ratio Years of schooling 0.0037 0.36 0.0247 3.43 0.0374 2.89 City work experience 0.0056 0.32 0.0132 1.11 0.0434 1.46 City work experience 2 0.0002 0.09 0.0004 0.78 0.0019 0.42 Other job experience 0.0220 3.91 0.0102 2.06 0.0278 3.93 Other job experience 2 0.0006 5.15 0.0001 0.56 0.0005 2.87 Total training days 0.0004 2.39 0.0002 1.00 0.0012 4.00 Dummy for origin 0.1223 1.11 0.0401 0.67 0.1422 2.43 Dummy for married 0.0325 0.83 0.1537 3.86 0.0001 0.00 Dummy for former farmer 0.0124 0.44 0.1122 5.77 0.0461 1.38 Dummy for males 0.1603 5.17 0.1944 6.98 0.2493 7.25 Constant 5.9004 41.34 5.5848 61.99 5.4334 37.65 Number of observations 483 547 275 Adjusted R 2 0.18 0.38 0.55 result is consistent with Meng (1998) in that gender wage discrimination becomes relatively less important in accounting for the overall wage gap as rms and the market become more liberalized. 6.2. Evidence from Jinan The results of the wage equation estimation for Jinan are reported in Table 6. There are a number of similarities and di erences in its structure of wage determination compared to Shanghai. First, as in Shanghai, the rate of return to education is quite low. One more year of education brings about an increase of 2 to 4% in wages in the private and collective sectors, and the return is almost zero in the state sector. The much lower return to education in Jinan's state sector than in Shanghai's state sector may be due to the fact that most of the migrant workers in Jinan were in temporary positions (with or without contracts). Thus, the migrant workers in Jinan were subject to not only government restrictions but also local rms' discrimination. Second, in terms of statistically signi cant variables and R 2, labor productivity related variables are more important in the private and collective sectors than in the state sector. The gap in R 2 between the state sector and the collective/private sector is much larger here than in Shanghai. Finally, as in Shanghai, controlling for the human capital variables and other individual characteristics, the coe½cients on the sex dummy suggest that women in Jinan earn signi cantly less than men, and the wage di erential is the least in the state sector and the largest in the private sector.13 The above wage equation is then estimated for males and females separately within each sector and the results are reported in Table 7. Not surprising, schooling is (statistically) signi cant in the collective/private sectors but not in the state sector. Except for other job experience and training days, there appears to be no clear systematic pattern in the coe½cients for males and females across the three sectors. The reward to other job experience and training days is higher for females than for males, suggesting that these two variables do not contribute to discrimination.
Wage di erentials and discrimination in China 345 Table 7. Human capital equations by sector and gender in Jinan State Collective Private Males Females Males Females Males Females Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Coe. T-Ratio Years of schooling 0.0075 0.61 0.0009 0.05 0.0231 3.03 0.0501 2.71 0.0443 3.64 0.0383 1.77 City work experience 0.0025 0.12 0.0075 0.21 0.0264 2.01 0.0781 2.22 0.0643 2.08 0.0008 0.01 City work experience 2 0.0006 0.26 0.0007 0.17 0.0010 1.57 0.0031 0.87 0.0042 0.94 0.0010 0.11 Other job experience 0.0182 3.09 0.0458 1.65 0.0102 1.95 0.0146 1.18 0.0142 2.05 0.0486 2.54 Other job experience 2 0.0005 4.04 0.0015 0.59 0.0001 0.74 0.0001 0.47 0.0002 0.97 0.0009 1.74 Total training days 0.0003 1.61 0.0005 1.62 0.0001 0.88 0.0009 1.92 0.0009 2.60 0.0030 2.15 Dummy for origin 0.1997 1.39 0.0157 0.15 0.0285 0.42 0.0737 0.67 0.0835 1.85 0.2291 1.85 Dummy for married 0.0675 1.74 0.1979 1.72 0.1512 3.46 0.2119 2.09 0.0567 0.92 0.0484 0.26 Dummy for former farmer 0.0065 0.20 0.0650 1.04 0.0990 4.57 0.1549 3.17 0.0464 1.08 0.0669 1.26 Constant 6.0871 32.12 5.7456 27.58 5.7779 59.41 5.4532 26.74 5.5951 43.08 5.5108 22.77 No. of observations 336 147 443 104 150 125 Adjusted R 2 0.12 0.08 0.29 0.30 0.48 0.26
346 P.-W. Liu et al. Table 8. Blinder-Oaxaca decomposition for Jinan Male-weighted Female-weighted Cotton Value % Value % Value % State Total di 0.203 100 0.203 100 0.203 100 [81.59] [81.58] [81.58] Explained 0.055 27 0.050 24 0.023 11.3 [22.03] [ 19.58] [9.219] Unexplained 0.148 73 0.253 124 0.180 88.7 [59.56] [101.16] [72.36] Collective Total di 0.299 100 0.299 100 0.299 100 [127.07] [127.07] [127.07] Explained 0.106 35 0.143 48 0.113 37.9 [44.47] [60.99] [48.16] Unexplained 0.193 65 0.156 52 0.186 62.1 [82.60] [66.08] [78.91] Private Total di 0.426 100 0.426 100 0.426 100 [164.39] [164.39] [164.39] Explained 0.157 37 0.302 71 0.223 52.4 [60.82] [116.72] [86.14] Unexplained 0.269 63 0.124 29 0.203 47.6 [103.57] [47.67] [78.25] Note: The main entries under the column ``value'' represent the log di erential decomposition, while the numbers in the square bracket are the corresponding monetary values (yuans). See Table 5 for average earnings by sector/gender. As in the case of Shanghai, Table 8 shows that the explained log differential due to endowments increase from state to collective/private rms. The unexplained log di erential, however, remains more or less the same across the three sectors. It is important to see again that although the nonstate sectors have a larger log gender wage gap compared to the state sector, in these sectors a greater proportion of such gap can be explained by the differences in employees' endowments. For example, under the Cotton weighting, only about 11% of the total wage gap in the state sector can be explained by personal characteristics, whereas the explained portion accounts for about 38 and 52% of total wage gaps in the collective and private sectors, respectively. The decomposition results indicate that, like in the Shanghai data set, more privatization induces less gender wage discrimination in percentage terms. The same conclusion on the diminishing importance of discrimination from the state to collective/private sectors appears quite robust, especially given that the Shanghai and Jinan data sets are so di erent in many aspects. 7. Discussion and interpretation As discussed earlier, there are many factors which a ect gender wage discrimination. These include employers' personal preferences, labor market institutions, and the degree of marketization. Personal preferences are shaped
Wage di erentials and discrimination in China 347 largely by culture, and in the case of a socialist planned economy, also in uenced by ideology and dominated by the preference of the central planner. Historically, the Chinese culture may shape personal preferences of male employers to be discriminatory against females in wage setting. However, under the centrally planned economy before the economic reform in 1979, the egalitarian preference of the central planner took precedence over personal preferences to mandate a labor market institution which set wages to reduce inequality across individuals as well as gender, rather than to reward productivity di erences. This compressed the gender wage gap and suppressed gender wage discrimination in China prior to the reform. Economic reform brings about two institutional changes which impact on gender wage discrimination. First, economic reform leads to decentralisation which removes or weakens central planning control on wage setting in enterprises. Employers are given greater discretion in setting wages, and they can set them according to their personal preferences, if they choose. Their personal preferences against females shaped by traditional culture are no longer suppressed and they are given expression in the form of gender wage discrimination in much the same way as in other market economies. Second, economic reform leads to marketization which brings into play market competition. Market competition unleashes two forces that operate in opposite directions. On the one hand, to remain competitive, employers must remunerate their employees according to their productivity. Since employees are di erent in human capital endowments, gender wage gap increases. On the other hand, under perfect competition, employers cannot a ord to remunerate employees on criteria other than productivity. Employers who discriminate pay for their bias through higher production costs, but under perfect competition they will be driven out of business. Therefore, gender discrimination cannot persist in the long run under perfect competition. To the extent that there is increasing competition but not perfect competition in the markets of a transitional economy, gender wage discrimination will diminish but will not be eliminated. The relation between economic reform and gender wage discrimination in a transitional economy is therefore intricate as a number of forces unleashed by the reform are in play. The relative importance of their impact on the gender wage gap or gender wage discrimination is an issue that can only be resolved by appealing to empirical data. The results obtained from the Shanghai and Jinan data sets can be interpreted within this framework of countervailing forces. First, the impact of marketization on widening gender wage gap in the two municipalities in China is dominant as we observe substantial widening of wage di erentials across ownership sectors from state to private in both data sets. Increased marketization rewards productivityrelated human capital characteristics and this e ect overwhelms any discrimination reduction e ect of market competition that may exist to yield a widening of gender wage gap. Second, results from the Shanghai data set show increasing gender wage discrimination, in absolute terms, across ownership sectors from the state to the private. This is consistent with the interpretation that economic liberalisation and decentralisation allow more autonomous decision making on wage and employment in the collective and private sectors, leading to rising discrimination which is not curbed by market competition as the degree of competition is still relatively low. However, the Jinan data set seems to indicate a
348 P.-W. Liu et al. closer balance between rising discrimination due to decentralised wage setting and declining discrimination because of market competition as gender wage discrimination remains more or less stable across ownership sectors. Last but most important in our view, despite increasing amount of gender wage discrimination (in log di erential) observed in Shanghai from the state to collective/private sectors, both data sets show that the relative share of discrimination in the overall gender wage di erential declines substantially across the ownership sectors. This result is consistent with the earlier interpretation that the impact of marketization on widening gender wage gap is dominant. The increase in gender wage di erential due to marketization is much larger than any increase in di erential that may arise from more gender discrimination. 8. Conclusion The experience from the developed countries indicates that gender wage differentials tend to be smaller in countries with centralised collective bargaining that emphasize egalitarian wage policies in general (e.g. Sweden, Norway and Australia). It tends to be larger in countries that have decentralized, marketoriented wage determination with enterprise-level bargaining (e.g. United States and Canada) (Gunderson 1994, Blau and Kahn 1995). Further, former centrally planned economies tend to have the smallest gender wage gaps (Meng 1996). This cross-country pattern between the degree of marketization and gender wage gap and discrimination is also observed within a former centrally planned economy now in transition, namely China. After two decades of economic reform, there is a robust co-existence of state, collective and private enterprises in China. Across the ownership sectors from state to collective and private, there is an increasing degree of decentralization and marketization, and diminishing central planning control, re ecting the transitional nature of the Chinese economy. This provides the natural setting for testing the intricate relation between economic reform, the degree of marketization and gender wage discrimination. Our ndings, based on a representative sample and a selected sample of migrants in two municipalities in China, are in general consistent with the cross-country pattern. Besides being broadly consistent with ndings of other country studies, our results add new insights to the understanding of the dynamic e ect of economic reform and marketization on gender wage discrimination. Even though economic reform increases the gender wage gap and possibly gender wage discrimination in absolute terms, our analysis further shows that marketization arising from economic reform actually reduces gender wage discrimination in relative terms. In other words, while gender wage discrimination increases in the sense of a further widening of the wage di erential between gender as the economy liberalizes, its importance as a share of the wage differential diminishes. In the process of marketization, increased market competition drives the employers to reward productivity-related characteristics more than before. The widening di erential returns to individual human capital endowments accounts for an increasing portion of the gender wage gap. Ipso facto, gender wage discrimination becomes relatively less important. This