D I S C U S S I O N P A P E R S E R I E S IZA DP No. 6252 Analysis o Gender Wage Dierential in China s Urban Labor Market Biwei Su Alas Heshati Deceber 2011 Forschungsinstitut zur Zukunt der Arbeit Institute or the Study o Labor
Analysis o Gender Wage Dierential in China s Urban Labor Market Biwei Su Korea University Alas Heshati Korea University and IZA Discussion Paper No. 6252 Deceber 2011 IZA P.O. Box 7240 53072 Bonn Gerany Phone: +49-228-3894-0 Fax: +49-228-3894-180 E-ail: iza@iza.org Any opinions expressed here are those o the author(s) and not those o IZA. Research published in this series ay include views on policy, but the institute itsel takes no institutional policy positions. The Institute or the Study o Labor (IZA) in Bonn is a local and virtual international research center and a place o counication between science, politics and business. IZA is an independent nonproit organization supported by Deutsche Post Foundation. The center is associated with the University o Bonn and oers a stiulating research environent through its international network, workshops and conerences, data service, project support, research visits and doctoral progra. IZA engages in (i) original and internationally copetitive research in all ields o labor econoics, (ii) developent o policy concepts, and (iii) disseination o research results and concepts to the interested public. IZA Discussion Papers oten represent preliinary work and are circulated to encourage discussion. Citation o such a paper should account or its provisional character. A revised version ay be available directly ro the author.
IZA Discussion Paper No. 6252 Deceber 2011 ABSTRACT Analysis o Gender Wage Dierential in China s Urban Labor Market This paper estiates the gender wage gap and its coposition in China s urban labor arket using the 2009 survey data ro the Chinese Faily Panel Studies. Several estiation and decoposition ethods have been used and copared. First, we exaine the gender wage gap using ordinary least square regression ethod with a gender duy variable. Then, we apply Oaxaca (1973) decoposition ethod with dierent weighting systes to analyze the logarithic wage dierential. To be ore speciic, we prove the existence o saple selection bias caused by the eale s labor orce participation. We eliinate it by using the Heckan s two-step procedure. Epirical results reveal that ale workers generally receive a higher wage than eale workers, and a great deal o this dierence is unexplained. Meanwhile, this unexplained part, which is usually reerred to as discriination turns out to be higher when the adjustent is ade or the selection bias. A urther breakdown o the wage gap shows that aong all the individual characteristics, occupations explain the largest share o the wage gap, ollowed by their working experience. On the other hand, education acts as a contributor or discriination in the labor arket. JEL Classiication: J70, J31, J16, J78 Keywords: discriination, wage gap, decoposition, gender, Chinese labor arket Corresponding author: Alas Heshati Departent o Food and Resource Econoics College o Lie Science and Biotechnology Korea University, East Building Roo#217 Ana-dong, Seongbuk-gu Seoul 136-713 Korea E-ail: heshati@korea.ac.kr
1. Introduction China has one o the largest labor orces in the world with a population o about 1.37 billion. According to the China Statistical Yearbook 2011, ore than hal o the are in eployed in 2011. Though the setting o eployent has changed substantially over the past 3 decades, ull-tie eployent or woen in China has been the nor or the past ive decades (Fang, 2005). The data o eale labor participation ro 1980 to 2009 is reported in Table 1. According to this data, around 70% o the eales in China join the labor orce. This is relatively high copared to any other countries around the globe. However, this apparent equality is quantitative rather than qualitative, with the ajority o woen clustered at the lower end o the job ladder. Along with the reorations in the labor arket woen encounter increasingly blatant discriination in the stages o recruitent, payent, prootion and retireent. Chinese social scientists have ollowed the changes o the woen s eployent and their wages relative to en. Woen s Studies Institute o China indicates that woen have lower average earnings than en. Ever since the econoic transoration o China started, the average ale-eale earning dierentials are widening. According to the data sorted out ro the Chinese Faily Panel Studies (CFPS), the average ale wage is around 2,810 yuan 1 per onth, while the average eale wage is around 2,319 yuan per onth. This is corresponds to only 82.5 percent o the ale s average earnings. Wage dierential has long been noted and debated. Many studies are in place to explain the causes o variations in wages and earnings aong gender, race, and union ebership, etc. Many actors have been identiied as responsible or this state o earning dierential aairs apart ro the obvious dierence in capability between the ales and eales. These include the eployers objectives and their concerns or proit axiization/cost iniization, the absence o an eective enorceent echanis or equal opportunity legislation, gender bias in eployent policies and workplaces, unequal education and on-job training opportunities and prootion possibilities, ale doinated networks and decision aking, as well as traditional social values. The urban labor arket in China has grown assively in the past 3 decades as a result o econoic developent and igration lows leading to an increased wage gap and inequality. Theoretically, the gender wage gap is due to the dierences in individual endowents as well as gender discriination. Most o the studies done on China s urban labor arket pointed out that gender wage dierential and wage discriination have risen ro the 1990 s (Gustasson and Li, 2000; Liu, Xin and Zhang, 2000; Zhang, 2004, aong others). However, in recent decade, the general environent o the labor arket is turning better or the eales. The iproveent is accopanied by the enorceent o the Labor Laws in China, agreeent o International Labor 1 On Deceber 8, 2011, US$1 is valued 6.34 yuan. 2
Conventions related to the protection o woen and equal opportunities in eployent, and increasing concerns on eale workers welare. Thereore in this study, we are interested in the current state o the gender wage gap. Is the pay discriination against the eale workers getting saller? To what extent can this dierential be explained by the individual characteristics? These questions are o priary concern or both researchers as well as decision akers. Our objective is to provide a ore accurate and up-to-date estiation o the wage dierential in urban China. In particular, we apply recent SFPS survey data and use dierent estiation and decoposition ethods to analyze the wage gap. This paper iproves our understanding on the extent o wage discriination in the labor arket in China. It also helps us to identiy the ost inluential actors contributing to the wage dierential. Inoration about wage gap is iportant or the governent to orulate policies or eliinating gender inequality and prooting a ore equalized labor arket in China. The ollowing section provides a background o the labor arket in China. In Section 3 we provide an overview o the literature on wage dierential. Section 4 introduces the data that we used or the epirical analysis. Section 5 describes odels and Section 6 provides analysis o the epirical results. Finally, conclusion and soe policy suggestions are given in Section 7. 2. Labor Market and Gender Wage Dierential in China Central planning and labor arket regulation over the past years, is arked by assigned jobs and equalized wages. Woen in China enjoyed a greater degree o gender equality in the workplace, copared to their counterparts in the copetitive arket econoies. The enhanced equality was although liited to low and iddle rank levels o eployent. While several changes have taken place in the labor arket in the recent years aiing to revive China s econoy, woen in China have begun to loss their previous better equality advantages. These ajor changes are described below. Firstly, in the id-1990s to the early 2000s, the aount o state-owned enterprises (SOE) has been acquisitioned by private copanies. Fro Table 1, we can observe that the share o SOE ell sharply ro 73.6% to 51.1% ro 1995 to 2009. During this reor, illions o workers were laid-o or orced to an early retireent, any o the consequently becae bewildered and unsuccessul job seekers. This radical downsizing o the state-owned corporations and public institutions were accopanied by a series o reors in the huan resource policies and practices in the state sector. These were introduced to cancel out the erosion o job security and welare provision and it tightened the perorance anageent as well as it eased the transoration process. Still, in the ace o these changes, eale laid-o workers ind it hard to survive in the ever ierce copetition against their ale counterparts and other younger job seekers. Most o the have to accept a wage cut in order to get eployed and they 3
crowd up in the low paid service sector. Soe o the stay out o the labor arket and rely on their husband s incoe or the tiny governent support and subsidies. And very ew eales ater losing their eployent, i they are inancially capable try to run sall private businesses. The second ajor change has been the rapid growth o private sectors with the presence o diverse ownership ors, naely ultinational corporations, international joint ventures, doestic private irs, sel-eployed businesses. We can see a stunning expansion o these units ro 1980-2009 that within 3 decades they grew ro zero to nearly a hal share o the total urban eployent units. The labor arket is becoing ore copetitive and lexible. Majority o the workers are categorized by their gender, experience, residence-status 2 and arket capability. The arket econoy gives rise to dierent treatents according to the eployers, workers and consuers discriinatory taste. This places eale workers and those with origin in rural in a unavorable position. Third, continuous urbanization and labor surplus in the agricultural sector has attracted an inlux o igrant rural workers to the rapidly growing urban areas. Due to low skills and lack o education these igrants overcrowd in low-paid segents o the job arket. Hence, they pose huge challenge against urban eale workers, who are captured by an increase in the share o urban population (see Table 1). This is because ost o the igrants are ales, who are stronger and willing to accept lower wages than urban eales. Moreover, they can spare eployers, the consideration o bearing child and early retireent policies i their eployees are eales. Thus an increasing trend in the igrant inlows ro rural to urban places lowers the possibility or eales to get a job in the irst place and lowers the level o payent to eales. Several studies which started ro the 1970s in China prove the level o ipact o the econoic reor and the developent o wage dierential. Maurer-Fazio et al. (1999) suggested that the higher the level o arket-oriented reor in the work unit, the wider will be the gender wage gap in the urban labor arket. Zhang (2004) ound out siilar results and he urther identiied sub-groups o eployees who suer ro the growing wage gaps. These are aong, the lower educated, over-orty-year old, non-sector workers, and blue-collar workers. On the other hand, Gustasson and Li (2000) ound out that young woen and woen with liited education have especially deteriorated during the transoration when copared to en having the sae characteristics. These indings also iply that the enterprise behavior in the arket environent also aects the wage structure. Enterprises ai to axiize proits and have the autonoy or hiring people and deciding the level o copensation; they value, or whatever reason, eale workers less than ale workers, and thus hire ewer woen or pay the less. I the dierences in education and other individual characteristics cannot explain 2 It depends on individual s birthplace containing two types o status: one is rural and the other one is urban. One can tell the residence-status though a certiication o the Chinese citizenship, so-called Hukou. 4
the entire wage dierential between the genders, evidence o discriination is said to be present in the labor arket The type o work unit is another contributor or the wage dierentials.state-owned units behave dierently ro their non-state counterparts which operate in business environent with a closer reseblance to a copetitive labor arket. Meng (1998a) proved that one third o the gender earnings dierentials are attributable in ull to the discriination or those eployees whose jobs were assigned by the local governents, while two-thirds can be attributed to the discriination on the part o workers who obtained jobs through individual job-searching eorts. Another reason or gender wage dierential addressed by Meng (1998b) was gender occupational segregation suggesting that eales usually crowded in lower paid jobs. Wang and Cai (2008) studied the gender wage dierential in the inter-sectors and intra-sectors diensions. By coparison they pointed out that the ain source o lower earnings or eales lied in the unequal pay within the sectors and the earnings gap due to the dierences in the sectoral attainent was relatively sall. It is apparent that China s reors and opening up o its econoy has had signiicant eects on the labor arket. While soe workers are acing a greater level o eployent insecurity and worsening conditions, others are presented with wider occupational choices, job obility and higher rewards or their skills and education. Eployers in general are encountering increasing pressure or quality, adaptability, and perorance enhanceent. This calls or the need o equality in the labor arket and ore coplete legislation to protect the labors rights in China. 3. Theoretical Background Wage dierential, this coon phenoenon in the labor arket around the world has spawned a rich literature on its developent and sources. On one hand, it is iportant to speciy the deterinants o wage which can be used to narrow the gap. On the other hand, it is crucial to ind the explanations to quantiy signiicant wage dierences between the groups which are not justiied by dierential productivity and huan capital investent. This unexplained part is ainly reerred to as discriination. Thus it is iportant through various policies and regulatory easures to reduce the non-huan capital and productivity related portion o the wage gap. At the theoretical level there are three ain econoic genres regarding wage dierential. One is the neoclassical theory steing ro the work o Becker (1957) which suggests that the prejudice is expressed in a discriinatory taste on the part o eployers, workers and consuers. The second one is ocused on the statistical theory o discriination (Aigner and Cain, 1977). The preise o this one is that irs have liited inoration about the skills o the job applicants. This gives the an incentive to use easily observable characteristics such as race or gender, to iner the expected productivity o the applicants.the last one is the segented labor arket approach, 5
which can be traced back to the theory o non-copeting groups in the work o Mill (1885). This approach oves away ro the concept o copetitive labor arket and views the labor arket as being split into sectors that are either doinated by ale or eale workers respectively. Exaples o this approach are job crowding and dual labor arket. Theoretically whether it is better to view wage dierential as an outcoe o essentially copetitive situations or as a product o non-copeting groups at the labor arket is a debatable point. Yet the segented approach has an evident weakness in it. It cannot address the issues on how occupations are segregated and the reasons or discriination still persist or be eroded over tie. The econoetric investigation o discriination started with Becker s seinal study on econoics o discriination in 1957. Since then, the prolieration o the use o icro data enables econoists to analyze the productivity o individuals. In particular, the decoposition technique which was pioneered by Blinder (1973) and Oaxaca (1973) has requently been applied to data acquired ro various countries and at tie periods (For exaples, see Wol and Petrela, 2004; Sith, 2002; Boraas and Rodgers, 2003; Bhandari and Heshati, 2008; Jung and Choi, 2004). This ethod deterines how uch wage dierential between two groups is attributed to the dierences in the characteristic o each group where wage regressions are estiated separately. For general reviews on the easureent o inequality and its decoposition see Heshati (2004a) and (2004b). Weichselbauer (2005) provides a eta-analysis o the international gender wage gap and Jone (1983) provides a critical coent on Blinder s ethod. Several changes have been ade with respect to the original Blinder-Oaxaca ethod. For instance, Reiers (1983) developed this odel by taking account o possible selectivity bias due to the distinction between the oered wages and the observed wages. He claied that discriination, i aected the wage rate largely, it would inluence the individual s decision on working participation. Thereore, the oered wage would be truncated and incidental as it depends on another variable, naely, labor participation as a conditional variable. Cotton (1988) reorulated the Blinder-Oaxaca odel by urther breakdown o the unexplained part, so that both the disadvantage (discriination) iposed on the inority and the advantage (beneit) bestowed on a ajority can be estiated. A ore coplicated transoration o this odel was driven by Neuark (1988), in which he considered the linkage o the Blinder-Oaxaca ethod to a theoretical odel on the eployers discriinatory behavior. 4. The Data The data used in this paper is ro a priary survey designed and conducted by the Peking University or the national progra o Chinese Faily Panel Studies (CFPS) 6
which started in the early 2007. In general, the basic idea o CFPS is to understand social and econoic changes through data at an individual, aily and counity levels. CFPS has covered ost o the questions in 4 panel surveys (PSID, CDS, HRS, and NYLS) 3 in the U.S. with thees covering social, econoic, education, health issues and so orth. We use the data ro the CFPS survey in 2009. The inoration is collected ro the residents in Beijing, Shanghai, and Guangdong on the basis o a structured questionnaire. To ocus on wage deterination in the labor arket, we restrict our saple to civilian wages and salaried eployees. In accordance with the standard practice, we exclude the ollowing ro the analysis: sel-eployed individuals, retirees, students, agricultural workers, the disabled, retired eployees who were rehired and teporary uneployed workers. We also excluded all persons aged 15 or less (China s labor law sets the iniu eployent age at 16 years) as well as respondents who provided incoplete inoration on wages, education, age, or other key variables. Ater the above entioned exclusions, the saple coprised o 1,533 observations including 117 housewives and 1,416 working individuals between the ages o 16 (school-leaving age) and 55 (state retireent age or woen) or 60 (state retireent age or en). For the working individuals, there are 844 en (59.6%) and 572 woen (40.4%). We have used a log o the average onthly wages earned ro a ain job, as a dependent variable including three regional duies, ive education level duies and six occupation typed duies. Working experience was not observed in the present data. Hence, we deine the variable o experience as age inus the year o education inus 6 years. Additional variables include a set o duy variables representing the type o household registration, child and arital status. A ull list o variables and their deinitions are presented in Table 2. Table 3 presents the suary statistics o the saple data including the eans and the categorical distribution o the individuals characteristics, or ales, eales and their dierential, which are regarded as the deterination or the wage rate. Here we briely introduce several ain actors based on the current available statistics. Regional location Relatively, Beijing, Shanghai and Guangdong are econoically developed regions when copared to other regions in China. With a higher deand or labor orce, these three regions have ore institutionalized urban labor arkets. Whereas, o the total eployent, their preerential type o econoic activities varies resulting in dierent contribution o ale and eale work orce. Beijing, with its tertiary industry accounting or 73.2% o its gross doestic product shows little dierence in the ale-eale worker proportion. Shanghai, stands as the coercial and inancial 3 PSID (Panel Study o Incoe Dynaics), CDS (Coon Data Set), HRS (Health and Retireent Survey) and NYLS (New York Law School). 7
center o ainland China and it is attracting ore eales to becoe oice workers. It is because gender distinction is less in oices than that in the actories where physical strength attes ore. In contrast, Guangdong has a higher need or ale workers to work in the production and exports sectors. Guangdong owes its growth to the largest share o anuacturing industry and its volue o exports in China. It is interesting to look at the wage rate o these three areas which have relatively siilar econoic level but have dierent preerential in the type o industries. Education level Education is one o the traditional huan capital theory variables, which acts as a proxy or individual copetence. In this paper we deine 5 ladder type sequentially increasing duies or education. According to the conventional studies, education has a positive eect on wages, or it increases the earning capacity through its eect on an individual s productivity. We also expect an increase in the wage rate along with a higher level o educational attainent. However, education has an abiguous eect in ters o explaining the gender wage gap in China. Zhang (2004) ound out that ro 1986 to 1993, the power o education to explain wage gap increased ro 2.3% to 5%, but there was a all back to 2% in 1997. Xie and Yao (2005) argued education can explain a sall aount o the gender wage gap in 2002. Chen and Haori (2008a) also ound education contributed to the discriination in the labor arket in 2005 (see also Li et al, 2005). Fro Table 3, we can see that eales in China enjoy a higher education attainent than their ale counterparts. The dierence becoes evident in the attainent o the college diploa. The largest gap can be observed in the university education, showing that a larger proportion o the eale workers achieved the Bachelor s degree. Though it is eaningul or eales in China to get equal or ore education opportunities than ales, given that eales earn less than the ales, we expect that education levels have negative eects on the explanation o the gender wage gap. Wor k experience Experience is the other traditional variable in the huan capital theory, which is proved to have a non-linear relationship with wage in an earning equation. This is to an extent that work experience is an investent in an on-the-job training. Incentives or this or o huan capital accuulation declines with age. The principal reason is that the present value o any returns-to-training investent declines as the residual work-lie o the worker decreases. Thus, a positive sign on experience and a negative sign on squared-experience are expected. In our saple, in an average, ales possess ore than 4 years o experience when copared to eales. This is attributed to longer schooling period or eales, lower pension age and otherhood leaves. Most o the literatures ound that experience can explain the gender wage dierential to a certain degree (Zhang, 2004; Xie and Yao, 2005; Chen and Haori, 2008a). 8
Type o occupation The skill o an individual worker is assued to play an iportant role or higher earning opportunity (Wang, 2005; Gustasson and Li 2000; aong others). We separate the observations into 6 categories o occupations including two types o workers in service industry, low/high-skilled workers in the anuacturing industry, and junior/senior proessionals. A positive relationship between the onthly wage and the coplexity o the job is expected. Generally, it is ore coon, that eales work in the service industry than ales. Moreover, anuacturing industry has a larger nuber o ale workers than eales. This is attributed to the physical dierence between genders. However, it is surprising to ind ore en working as senior proessionals than eales despite eales receiving ore tertiary education than the ales. We can see that the lower levels job in occupations with higher knowledge and higher social status (e.g. physicians, scientist, university proessors, etc.) are usually illed by eales who are already in the inority group in those occupations. This partially relects the act that eales in China have lower status jobs ater graduation. Even when they ind the right track o proessional career, they ail to be prooted as quickly or as requently as their ale counterparts (Fang, 2005). Household Registration Type China or about 40 years had isolated rural and urban econoies ro each other. One cause or segregation is the strict Household Registration Syste, which required individuals to register with the local authorities to gain residency and thereby, deterined where people lived and worked. (Meng and Zhang, 2001) Even i the restrictions on rural-urban igration have been eased in the late 1980 s; this old syste reains potent. It continues to serve as one o the key institutions perpetuating China s rural-urban disparity. Discriination against rural igrants in ters o occupational attainent and wages has been ound in several literatures (e.g. Wang and Zuo, 1999; Wang and Cai, 2006; Zhao, 2000; Xin, 1995; Chi et al., 2007). In our saple the proportion o urban residents or both genders are nearly the sae, reaching to around 62% o the total nuber o observations. As the total workorce coprises o a large share o igrants, it is iportant to capture the eect o workers residency on their onthly wages. Given an earning dierential between these two groups, a positive sign on this duy variable is expected when the urban residence is assigned as 1. Marriage and Child status Finally we consider the status o arriage and child or the worker assuing that getting arried and having a child ay have dierent eects on one s earnings. We propose the hypothesis that couples usually have a larger expenditure than that o a single one s. Thus the orer ones ay have incentive to earn higher wages. Previous studies have shown that, arriage had positive eects on a person s wage rate in China (Chen and Haori, 2008; Xie and Yao, 2005). The relationship between the children and the workers wage rate has seldo been 9
addressed in the previously conducted wage dierential studies. Previous studies, which ainly ocused on eales, ound out that others earn lower hourly wages than do woen without children due to less working experience and teporary part-tie eployent (Hill, 1979; Waldogel, 1997). Here, we also consider a negative eect o children on ales or eales onthly wages as tie and eorts are needed or parents to take care o their children. 5. Speciication o the Model In this section, we analyze how the various characteristics o individuals contribute to the gender wage dierential through application and coparison o several alternative approaches. The irst step involves the identiication o whether the gender wage dierential exists statistically and signiicantly or not. In doing, so we estiate the ollowing Mincerianearning equation (see Heckan et al., 2003) with an additional gender duy variable: (1) ln Wi = X iβ + Mγ + ε i where, W i is the onthly wage o the worker i, X i is a vector o individual characteristics (see Table 2 or the list o the variables), Mis a duy variable or ale, β and γ are vectors o unknown coeicients to be estiated, and ε i is a disturbance ter which is assued to satisy the usual properties. The coeicient γ is the estiated coeicient o the gender duy variable in this earning equation. It supposedly easures the wage penalty or the wage dierential or the eales ater controlling several characteristics such as education, experience, occupation, arriage, child, and Hukou in the earnings deterination. As entioned previously the ain objective o this paper is to analyze the coposition o the gender wage gap and investigate how individual characteristics can explain the gap. Thereore to begin with, we deine the gross wage dierential as: W (2) D = 1 W where, W /W is the observed ale-eale wage ratio. Then, we introduce the standard Mincerian earning equation or ales and eales separately written as: g g (3) ln Wi = X i + ε i where, the subscript g denotes worker s gender, g=(ale, eale). The variables included in these equations are the sae as those in the equation (1) but with the absence o the gender duy, M. Estiated wage equations or dierent sexes can be orulated based on the estiated 10
coeicients as ollows: (4) ln W = ˆ β X and lnw ˆ = β X where, W is the geoetric ean o the onthly wages, X is a vector o the ean values o the regressors and βˆ is the vector o the corresponding estiated coeicients. By taking the log or o equation (2) the onthly wage dierential between ales and eales can be expressed as: (5) ln D = lnw lnw = X ˆ β X ˆ β Following Oaxaca (1973), the above equation can be urther transored or decoposition purpose as: (6) ln D = ( X X )[ Ω ˆ β + ( I Ω ) ˆ β ] + [ X ( I Ω ) + X Ω ]( ˆ β ˆ β ) where, I is an identity atrix and Ω is a diagonal atrix o weights. This states that the ean dierence in log onthly wage is decoposed in to two parts. One part is o the dierential which is due to their average endowents (the irst ter on the right-hand side), and the eects o discriination and the other oitted actors, as revealed by the second ter. In general, we will get dierent easures to analyze the wage dierential, depending on the choice o weights in the atrix Ω. I Ω=1, it suggests that discriination penalizes the eale by preventing the ro earning according to the ale s wage-oer unction. I Ω=0, it iplies that the discriination gives the ale advantages. In another words, ales are paid ore than what they would get in a non-discriinatory world. These two equations represent two extree cases. The decoposition o dierences in earning is the estiate o what a ale worker receives i he aced the eale wage structure or vice versa. It is ost likely that the real values would all in between the two extree cases. Thereore, we add Reiers s (1983) ethod as a coparison. In this odel Ω=0.5I, shows the eployers preerence or ales and their distaste or eales, distort both groups wages. Thereore, neither group s observed wage-oer unction would be likely to exist in a non-discriinatory world. Following the Reiers s ethod the wage dierential in (6) is reduced to: (7) ln D = 0.5( X + X )[ ˆ β ˆ β ] + 0.5[ X X ]( ˆ β + ˆ β ) Moreover, Reiers pointed out that i participation in the labor arket is not rando, 11
then the estiation is subject to saple selection bias. In our case, this proble occurs when we take in to consideration the eale worker s population. Woen with higher expected wage rate or those who preer to be a housewie are reluctant to join the labor orce. As a result, their wages ay not be observed and this absence would aect the to-be-estiated wage structure. To encounter this kind o selective proble Reiers (1983) applies Heckan (1979) two-step selection correction estiation procedure (see also Wooldridge, 2006). The irst step is to speciy a labor orce participation equation or woen in the or o a probit unction in order to get an inverse Mill s ratio to correct the possible selection bias. Speciically, the dependent variable is a binoial variable assigned as 0 i the eale is a housewie and as 1 i the eale works. Explanatory variables include duies or the region, education level, Hukou, Marriage and Child status. Experience and occupations are not considered or housewives who have no related inoration. The second step involves estiation o the wage equation by OLS with the inverse o the Mill s ratio (MR), which is predicted ro the probit unction as an additional explanatory variable. Here the saple is liited to only eales working. With the adjustent ter, the equation (1) and (4) changes into the ollowing ors: (8) ln W i = X i + δ MR + ε i (9) ln W = ˆ β X + ˆ δ MR and the wage gap decoposition in equation (7) changes as: (10) ln D = 0.5( X + X )[ ˆ β ˆ β ] + 0.5[ X X ˆ ]( β ˆ + β ˆ ) δ MR here, δˆ is the estiation o the covariance between the errors in the probit and wage equation and M R is the estiated inverse Mill s ratio to correct or the selection bias resulting ro the dierences in eales probability o working. 6. Epirical Results 6.1 Deterinants o Monthly Wage The results o regressions analyses have been shown in Table 4.1 and 4.2. These tables contain the results o the estiated earning equation or the total saple o workers with a gender duy variable, and the results o the estiated earning equations or eales and ale workers separately. For eales, there are two earning equations, one which adjusts or the saple selection bias, while the other does not. Fro the irst colun o Table 4.1, we can identiy the contribution o the worker s characteristics and the eect o discriination on the wage gap between the ale and eale workers. Overall, the odel its the data relatively well, and ost o the 12
coeicients have expected signs and are signiicant at less than 5% to the level o signiicance. The coeicient o gender duy shows that ale workers on an average earn about 5.9% ore than their eale counterparts, keeping the other variables a constant. This result conirs the existence o wage discriination in China. Other characteristics have dierent eects on wages. Generally, workers ro Beijing have the sae wage rate as workers in Guangdong, as the coeicient is not signiicant at the 10% level o signiicance. However workers in Shanghai enjoy on an average a 4.5% higher wage rate than workers ro other two regions. This is consistent with the report given by the Scientiic Socialis Research Institution in China (2011) where it indicated that Beijing and Guangzhou have a lower average wage and consuption levels than that in Shanghai. It also relects that wage varies in the doinated sectors i the saples are collected randoly. For exaple, in the case o Shanghai, larger share o workers are eployed in inance and coerce sector which leads to a higher wage rate than the other two locations, whereas Guangzhou has ore workers in anuacturing while Beijing has ore workers in education and governent sectors. A positive association is obtained consistently between the successive higher levels o education. According to our results, i we set the junior/below junior graduates as the reerence category, every upgrade in education attainent is accopanied with 5% increase in wage rate and with a 10% rise i one attains urther studies ater his undergraduate course. This result can be supported by Bargain et al. (2007) who proved the phenoenon that the average wage o the work orce increased in return ro the education level. This is due to the rapid iproveent in educational attainent which has taken place in China. Experience has an inverted-u shape relationship with wage. Such relations are illustrated in any previous literatures (Liu, Meng and Zhang, 2000; Zhang, 2004). The relation is characterized by an increase in wages at a worker s early age then the wage is peaked at soewhere and ollowed by a decrease at the later stage. It can be explained that i a worker is in the labor arket or a long tie, the return o his experience is likely to all as his skill ight be out-o-date due to the rapid developent o technology. On the other hand, his incentive or learning new skill is weaker than that or the younger ones. Hence, his expected returns are saller due to the approaching o retireent. Types o occupation are the greatest contributors to the dierence o wage rates. Various types o occupation would entail a wide range o copensation due to the level o skill sets required. White-Collar workers earn 11.2% higher than people who take up low-paid service jobs. Nevertheless, coeicient or low-skilled workers in the industrial sector is sall and is statistically insigniicant showing a siilar wage rate with low-paid service workers. Those with special skills such as trainings in achinery, driving etc. receive 14.1% ore than the reerence group. Larger gaps can be observed i a person works as a proessional (physicians, proessor, lawyer, etc.), with 25.2% higher wage or the juniors and a substantial aount o 55.3% higher wage or the 13
seniors. As expected, urban residents get higher wage rates than rural residents. Several reasons can be addressed to explain this phenoenon. Firstly, the productivity dierence between the urban and rural residents is because the orer usually has an easier access to tertiary education and ore job-training opportunities in order to get a higher-paid job. Second, the eployer s discriination is another contributing actor or the earning dierence between the urban and rural iigrants (Luo, 2008). Third, the two-tier labor arket, or labor arket segregation in China yields to the inequality (Knight et al., 1999; West and Zhao 2000; Goh et al., 2008). Moreover, urban residents enjoy other advantages like wider personal networks, accoodation, and inoration or jobs etc. Marriage possesses an expected positive eect on the wage rate. A arried worker has 4.2% higher wage than others who are single or divorced. This dierence ay be because the couples expenditure is usually higher than that or singles. This, in turn serves as a stiulator or couples to seek or prootion or higher paid job in order to aintain and urther iprove their living standard. Also, arried workers tend to be older than singles thus they get higher wages with longer working experience. On the other hand, wage rate is shown to be 6.4% lower i a worker has a child, considering the aount o tie and energy that will be distracted ro working or child rearing. The results o the earning unction or only ales are presented in the third colun o Table 4.1, with the R-squared value o 0.5789. The coeicients o workers characteristics are siilar to those attained ro pooled earning unction. In Table 4.2, we can see the results o two estiated alternative equations or eale workers. The irst colun shows the results ro the equation with no consideration o saple selection bias. While the third colun presents the results generated ro the Heckan-two step procedure adding to the negative and signiicant inverse Mill s ratio as an explanatory variable in the second wage equation to eliinate the possible selection bias. On coparison the results we note that, the coeicients, other than experience and occupation, have changed signiicantly suggesting that housewives do induce changes to the eects o deterinants or eales wage structure. The returns o education becoes saller, so as experience and type o occupations. Children and arriage are not signiicantly related to the wage rate. This indicates that i we correct or the bias, the eales wage structure will change less rapidly than in a biased estiation. According to Reiers, we explain this situation by saying that woen who have high-wage opportunities, given their observed characteristics, have even better opportunities outside the wages and salary sector. Hence, they are less likely to be included in the labor arket. In the next step we present the result o the wage gap decoposition using the Oaxaca decoposition technique with dierent weights. The dierences are then attributed to the relative endowents o huan capital between the ales and eales and the labor arket discriination coponents. 14
6.2 Decoposition o the Wage Gap Table 5 reports the result o decoposition o the estiated wage gap aong genders. The total wage dierential has been decoposed into two parts; the irst part is due to the dierence o genders endowents, and the second part is due to the dierences in the paraeters o the wage unction which can be attributed to the labor-arket discriination and to other oitted variables. We show this easure o discriination using three dierent sets o weights (Ω is equal to 0, 0.5 and 1.0) under two kinds o estiations (with and without correction or saple selection bias). The ean log onthly wage dierence in our saple is 0.071. Since selection bias ter is negative and signiicant, OLS estiates would be downward biased too, as would a wage-dierential decoposition based on the OLS results. The correction or selectivity in our saple widens the gender dierential to 0.116. Thus, the degree o underestiation o bias due to dierences in the probability o woen working in the labor arket is relatively large. It is interesting to ask how serious is this proble o discriination in producing these dierences? The results ro the adjusted earning unction are ore stable, indicating that discriination accounts or around 86% percentage o the estiated wage dierential. While the biased unctions give the results that the discriination ranges ro 79% to 87% averaged at 83%. These results are generally siilar with those o the previous studies o gender wage dierentials in China: or instance Chen and Haori (2008a) who proved that discriination counted to about 70% using Oaxaca ethod, and 75% using the Reiers ethod o the wage dierential in 2005; Liu et al. (2000) also showed that discriination in Shanghai labor arket was about 88% using the en-weighted value, 93% using the eale-weighted value, and 90% using the Cotton ethod in 2000. A sall variation ay arise ro using dierent datasets, assigned to weights and ethods. However, it is sae to draw a conclusion that in the past decade, discriination has been a ain contributor or wage dierential in China s urban labor arket. It is obvious that the three estiates o discriination are quite siilar or the adjusted unction. However, luctuations in percentage o discriination can be ound i they are derived ro unctions with saple selection bias. These changes could be attributed to large coeicients o education and other variables in the biased earning unctions or eales. While discriination reains large in China, we try to explain the causes o this phenoenon in light o analyzing the eects o the personal characteristics. Table 6 displays the eects o the individuals and the labor arket characteristics on the explained and unexplained part o the wage dierential. Though the weights are dierent the scale o the coeicients eects are nearly the sae. It is surprising to ind that education only accounts or 6% o the wage gap with a negative sign, indicating that education enlarges the wage gap between eales and ales. While in the unexplained part, education shows a positive eect and explains 23% o the discriination. This is because eales have higher education attainents than 15
ales while they are especially less paid on jobs. It is ore obvious or high school graduates and Master or Ph.D. holders. Those two types o eployees generally work in physical-work and senior proessional areas respectively, whereas ales receive higher wages than eales. Sae results have been ound in Chen and Haori (2008a). Another reason can be that, education acts ore like a signal or an eployer to identiy a person s ability in China s labor arket (Li et al., 2005). I a eale wants to prove her ability, being aware o the existence o gender discriination, she ight resort to a higher education degree and accept a lower wage given or her educational attainent. Aong all the characteristics, occupation has the strongest power in the explained part. This explains 27% the wage gap. It is coon to ind that the gender pay gaps are sall within the narrowly-deined occupations (Gunderson, 1989). The eects in relation with two types o workers are especially strong, naely the low-skilled workers and senior proessionals. It is worth to note that, the shares o ales are higher than the shares o eales in these two types o occupations whereas ales college and tertiary education attainents are less than the eales. This happens when the discriination o eployent exists. It is relected by preventing eales ro entering the high-wage sector in the labor arket and eventually leads to the increased wage gap. Chen and Haori(2008b) in the discriination part, say that the eect is about 24%. Low-skilled workers and Junior proessionals suer ro discriination ost. This can be ostly explained by ales physical advantages and eployers preerences. Instead o explaining the gender wage gap; experience is ore likely to contribute to the discriination. For paraeters o regions, Beijing and Shanghai generally have the sae aount o eects, indicating discriinations against eale workers. Marriage explains the wage gap to a certain degree. It ay suggest that Chinese woen, like any elsewhere, bear priary responsibility or household chores and childcare (Loscocco and Wang, 1992). Despite the signiicant dierentials between the arried eales and ales, (our results in Table 4.2) show that arried woen earn ore than single ones. This result is siilar to Hughes and Maurer-Fazio s (2002) inding where they explained this arriage preiu by entioning that eployers preer arried workers perhaps they view arried workers as ore stable and less obile-or that the wage syste contains equity provisions and views arried workers as ore needy. But i the second reason stands, Child should also explain the wage gap rather contribute to ore discriination since eales with children have higher needs than those without. Thus we consider that eployers ay preer arried eales, but once i they had children, they ight be considered not as copetitive as their counterparts and thus receive a lower wage. 7. Conclusion and Policy Suggestions This paper intends to analyze the gender wage dierential in China s urban labor 16
arket. To attain this, irst we identiy the wage level and gender wage dierential deterinants using Mincerian earning unction with gender duy variable. Second we use Oaxaca ethods with three dierent weighting systes to analyze the coposition o wage dierential and eploy the Heckan-two-step procedure to eliinate the possible saple selection bias attributed to the dierences in the probability o working. In the last section we have seen how uch o the total gender wage dierential coes ro sources such as discriination and those due to the individual characteristics. The latter category is controllable and can be ipacted by policy, such as education levels, geographic sector and occupation; while others coes ro characteristics that are beyond direct inluence, such as experience, arriage and child status. Discriination apparently is an overwheling reason or the low wages o eales. Ater adjustent or selection bias, the estiated discriination is around 85% o the gender wage gap. Occupation plays an iportant role to explain the wage gap while education, instead o explaining the wage gap, is a contributor to the gender discriination in China s urban labor arket. Married woen earn less than their arried ale counterparts but they earn ore than the unarried eales, while having children has a sall eect on discriination. It is iportant to note that the unexplained dierential is not an exact easure o discriination, as there is an absence o detailed controls or all possible relevant actors o job characteristics and person-speciic skills. In Macpherson and Hirsch (1995), this dierential is likely to overestiate the agnitude o discriination. However, when the unexplained dierential constitutes to a large percentage o the total dierential, the possibility o gender discriination cannot be copletely ruled out (Blau, Ferber, and Winkles, 1998). What do our indings ean or those including state, associations and civil society organizations who are concerned about iproving the econoic situation o eales in China? Firstly, anti-discriination eorts are in all its ors needed regardless o their wage gap eects. Governent is encouraged to reinorce the ipleentation o laws to orbid gender discriination in ters o wages, eployent, opportunities and proote equality; proote gender equality through ass edia and education syste; and increase the nuber o eale eployees in the public sector. Second, in light o our analysis, we ind discriination to be severe oreost in the high-wage sector. With this respect, governent could encourage the enterprises to eploy ore eale workers by providing various incentives. Third, while aintaining equal opportunity or educational attainent or both the genders, governent should strive to iprove the quality o education as well. For education, is a tie and oney consuing investent it should act as a role to iprove one s productivity rather than siple signaling. Only through iproved quality cobined with necessary incentives and regulations can education narrow the gender wage gap rather than increasing it. Thus state intervention and incentive provision and regulations are aong the easures to be 17
cobined with education and general public awareness to tackle wage discriination eectively. Reerences Aigner, D.J. and Cain, G.G. (1977).Statistical Theories o Discriination in Labor Markets.Industrial and Labor Relations Review 30(2), 175-187 Bargain, O., Bhauik, S.K., Chakrabarty, M. and Zhong, Z. (2007).Returns to Education and Earnings Dierences between Chinese and Indian Wage Earners, IZA Discussion Paper. Becker, G.S. (1957). The Econoics o Discriination. University o Chicago Press, Chicago. Berndt, E.R. (1994) The Practice o Econoetrics Classic and Conteporary.Addison-Wesley Publishing Copany, U.S. Bhandari A. and A. Heshati (2008). Wage Inequality and Job Insecurity aong Peranent and Contract Workers in India: Evidence ro Organized Manuacturing Industries.ICFAI Journal o Applied Econoics 7(1), 80-111. Blau, F., Ferber, M. and Winkler, A. (1998).The Econoics o Woen, Men, and Work. Upper Saddle River, New Jersey: Prentice Hall. Blinder, A.S. (1973). Wage Discriination: Reduced For and Structural Estiates, The Journal o Huan Resources 8(4), 436-455. Boraas, S. and Rodgers, W.M. (2003). How does gender play a role in the earnings gap? An update, Monthly Labor Review 126(10), 9-15. Cai, F. and Wang, M.Y. (2008). Gender Earnings Dierential in Urban China.Review o Developent Econoic 12(2), 442-454. Chen,G.F. and Haori, S. (2008a). An Epirical Analysis o Gender Wage Dierentials in Urban China.Kobe University Econoic Review 54, 25-34. Chen, G.F. and Haori, S. (2008b). Do Chinese Eployers Discriinate Against Feales When Hiring Eployee? Econoic Bulletin 10(14), 1-17. Chi, Wei, Li, Bo, and Yu, Q.M. (2007). Decoposition o Changes in Earnings Inequality in China: A Distributional Approach. MPRA Paper No. 3807. Cotton, J. (1988). On the Decoposition o Wage Dierentials.The Review o Econoics and Statistics 70(2), 236-249. Fang, L.C. (2005). HRM, Work and Eployent in China.Routledge, New York, U.S. Goh, C.C., Luo, X.B., and Zhu, N. (2008). Incoe Growth, Inequality and Poverty Reduction: A Case Study o Eight Provinces in China.China Econoic Review 20, 485-496. Gunderson, M. (1989).Male-Fael Wage Dierentials and Policy Responses.Journal o Econoic Literature 27(1), 46-72. Gustasson, B. and Shi, L. (2000).Econoic transoration and the gender earnings gap in urban China.Journal o Population Econoics 13(2), 305-29. Heckan, J.J. (1979). Saple selection bias as a speciication error.econoetrica, 47: 153-161. 18