China Policy Institute Discussion Paper 2 THE ECONOMICS OF COMMUNIST PARTY MEMBERSHIP ­ THE CURIOUS CASE OF RISING NUMBERS AND WAGE PREMIUM DURING CHINA’S TRANSITION Simon Appleton School of Economics, Nottingham University John Knight Department of Economics, Oxford University Lina Song School of Sociology and Social Policy, Nottingham University Qingjie Xia School of Economics, Peking University © Copyright China Policy Institute November 2005 China House University of Nottingham University Park Nottingham NG7 2RD United Kingdom Tel: +44 (0)115 846 7769 Fax: +44 (0)115 846 7900 Email: [email protected] Website: www.nottingham.ac.uk/china­policy­institute The China Policy Institute was set up to analyse critical policy challenges faced by China in its rapid development. Its goals are to help expand the knowledge and understanding of contemporary China in Britain, to help build a more informed dialogue between China and the UK and to contribute to government and business strategies.
The Economics of Communist Party Membership ­ the Curious Case of the Rising Numbers and Wage Premium during China’s Transition 1 Why is it that, as the Chinese Communist Party has loosened its grip, abandoned its core beliefs, and marketized the economy, its membership has risen markedly along with the economic benefits of joining? We use three national household surveys, spanning eleven years, to answer this question with respect to labour market rewards in urban China. We conceptualize individual demand for Party membership as an investment in “political capital” that brings monetary rewards in terms of higher wages. This wage premium has risen with the growing wage differentials associated with the emergence of a labour market and the continuing value of political status in the semi­marketized transitional economy. However, a demand­side explanation does not explain the fact that the wage premium is higher for the personal characteristics that reduce the probability of membership. We develop an explanation in terms of a rationing of places and a scarcity value for members with those characteristics. JEL Classification: J31, J40, J71, P20, P30 Key words: China; Communist Party; Labour Market; Economic Transition; Wages. 1. Introduction A curious paradox of the reform process in China is that, as the Chinese Communist Party (CCP hereafter) has loosened its grip on the economy and seemingly abandoned its core beliefs, its membership has risen markedly along with the economic benefits from joining. Karl Marx famously envisaged the state “withering away” in the advanced stage of communism. Outsiders have often made a similar assumption about the Communist Party withering away during the process of marketization. In China, the Party exercised considerable control over pay and promotion during the planning period (Groves et al., 1995). However, this has been reduced as China “grows out of the plan” (Naughton, 1995), so the naïve assumption would be that the wage premium to Party membership should fall. This in turn should reduce recruitment among opportunistic would­be members, while the idealists should be deterred by the Party’s abandonment of its core beliefs. Instead, the Party membership has expanded from 3.8% of China’s population in 1978 to 5.2% in 2002. As of June 2002, the CCP had 66.4m members, 5.2% of the total population, making it the largest political organization in the world (Economist Intelligence Unit, 2004). This paper looks in more detail at both the determinants of people joining the Party and the economic rewards to doing so. There has been considerable debate over the role of communist parties in planned economies. The “New Class” theory of Djilas (1957) argued that, although communist parties might come to power with the aim of promoting egalitarian societies, in practice they gave rise to a “redistributive elite”. In the Soviet Union, the Party elite came to be referred to as the nomenklatura, a term identifying those who enjoyed material privileges by virtue of their controlling position in a command economy. As managing a planned economy requires administrative and technical skills, it was subsequently hypothesised – based partly on evidence from communist Hungary – that a professional elite would come to replace the “New Class” of old revolutionaries (Konrad and Szelenyi, 1979). Szelenyi 1 The authors are grateful to the Ford Foundation for funding the data collection and to the CCK Foundation (RG019­U­01) and DfID (under Escor grant R7526) for supporting the research. We are grateful for comments received from seminars and conferences at CERDI­CNRS, Université d'Auvergne, and the universities of Gothenburg, Nottingham and Oxford.
2 (1986) later raised doubts about this hypothesis of an evolution towards managerialism, putting forward instead the alternative notion of a “dual elite” whereby the “red” elite of old revolutionaries maintaining administrative control while an “expert” elite of professionals enjoyed social status without political power. In China, this tension between the dual elites appears to have been particularly pronounced: the Cultural Revolution can be seen partly as a struggle between ideological “red” and pragmatic “expert” elites. As a result, the rise of a pragmatic, non­ideological, Party leadership appears to have been delayed in China compared to East European socialist countries and can be dated to the presidency of the Deng Xiaoping in 1979 2 . Deng’s presidency marked the beginning of a reform period that can be characterised as a transition from a command economy to a market economy. What happens to Communist Party members during the transition to the market in formerly planned economies has been the subject of conflicting theories. Nee (1989) argued that Party membership should lose its privileges as marketisation raised the returns to productivity (human capital) and reduced the redistributive power of the Party. However, others have disputed this hypothesis, arguing that during the transition the Party members may be able to convert their political capital into economic wealth (Ronas­Tas, 1994). This process may come about during the de facto privatisation of previously collectively own property and/or through party members securing now lucrative managerial positions. The maintenance or extension of privilege by the old nomenklatura is usually regarded unfavourably because, at the very least, it perpetuates unwarranted inequalities and, at worst, it threatens to subvert a full transition to the market (Frye and Shleifer, 1997). As Bird, Frick and Wagner (1998, p234) argue: “the efficacy of the free market system depends to some degree on its ability to choose different winners and losers than the old system did.” However, Morduch and Sicular (2000) provide a more positive perspective on a similar phenomenon. They observe that, in one rural county in North China between 1991 and 1993, cadres appeared to receive higher benefits from marketisation than others because they were more able to take advantage of new economic opportunities. Morduch and Sicular argue that this may be a favourable development as it implies that pro­market reforms in China may be self­sustaining, since they benefit the agents who have most influence over their future policy. There remains a danger, however, of public disillusionment with the Party and perhaps even pro­market reforms themselves if the reforms are seen to benefit Party members disproportionately – particularly if the benefits come through corruption, nepotism and other illegitimate forms 3 . In this paper, we use national urban household survey data to analyse the determinants of CCP membership and the associated wage premium. The data comes from the China Household Income Project surveys, designed by a team of international scholars, including the authors, and researchers at the Chinese Academy of Social Sciences. Results from the 1988 survey were published in Griffin and Zhao (1993), and those from the 1995 survey were published in Riskin, Zhao and Li (2001). The particular strength of the data is that it provides a very detailed measurement of income and labour supply. However, the data also have the benefits of providing repeated cross­sections spanning over a decade. The surveys used sub­samples of households included in the annual national household income survey of the National Bureau of Statistics (NBS), covering 10 out of 31 provinces in 1988, 11 in 1995 and 6 in 1999. Both the NBS survey generally and the sub­sample 2 Arguably, it is only with the inauguration of Hu Jintao in 2003 that the Chinese leadership could be described as technocratic (for example, for the first time, all Politburo members were university graduates). 3 A survey of 818 migrant laborers in Beijing in 1997­98 showed CCP members were regarded as self­serving. Only 5 percent of the interviewees thought their local Party cadres "work for the interests of the villagers," and 60 percent said their local officials "use their power only for private gains." (Pei, 2002)
3 used here were designed to be representative of urban China. However, they are subject to the limitation that they include only those households with urban registration (hukou). Consequently the analysis in this paper is confined only to such households and excludes the most rural­urban migrant households, since these are denied urban hukou. The outline of the rest of this paper is as follows. Section 2 provides a theoretical framework for modelling CCP membership, considering both the demand for membership by individuals and the screening of potential members by the Party. Section 3 uses household survey data from 1988, 1995 and 1999 to model party membership empirically. The effects of membership on wages are considered using the same data in Section 4. Section 5 examines the critical period of mass retrenchment in the state sector and looks at how the fortunes of CCP members varied according to whether they suffered unemployment or not. Section 6 concludes. 2. Theoretical framework: the economics of Communist Party membership To be a member of Chinese Communist Party is utterly different from being a member of any political party in the Western sense. In Western political systems, parties compete against each other for power and so are relatively open in admitting new members, since these provide funds and personnel for this political competition. For the individual, joining a party in the West may be seen as primarily marking an ideological affiliation and membership is unlikely to have a direct effect on one’s career or economic circumstances. By contrast, in China, the CCP has a monopoly on political power and historically has maintained its grip on power by virtue of a loyal membership, carefully vetted on admission and subject to scrutiny thereafter 4 . Moreover, the dominant position of the CCP and its occupation of important administrative positions implies that membership of the Party may bring real economic benefits to the individual. In this section, we expand on these distinctive features of membership of the CCP: first, the Party’s active role in screening prospective members and second, the economic cost­benefit analysis that may underlie the individual’s demand for CCP membership. The Party’s screening of prospective members Joining the CCP requires two decisions: that the individual decides to join and that the Party decides to admit the individual. We refer to the former decision as the individual’s demand for membership and the latter decision as reflecting the Party’s screening of members 5 . The process of joining the Party has three broad stages. First, the individual must make a formal application; although Guo (2002) states that the initial approach nonetheless usually comes from the Party. Second, there is a period of observation and training. The individual must participate in study sessions and community service while subject to daily monitoring. One or two CCP members act as liaisons (lian xi ren) while background checks on the applicant are performed and opinions sought from other CCP members. Third, after admitting a new member, they are subject to a one­year probationary period. Even after their probation, members are subject to a degree of scrutiny by the Party and are automatically expelled if they fail to attend meetings for six months without good reason. Individuals’ decisions to join the CCP are essentially irreversible 6 . Loss of Party membership is seen as very serious and in some cases could lead to individuals losing their jobs and having their career prospects blighted. 4 There are eight “democratic” parties in China, but these are relatively small (with total membership of around half a million) and, as virtual satellite organizations of the CCP, do not provide genuine political competition. Historically, these parties have tended to recruit people from non­laboring classes who are nonetheless supportive of the CCP. 5 The terminology here is potentially confusing. Solotaroff (2003) refers to party recruitment policy as being the “demand side” and individual choosing to join CCP being the “supply side”. 6 For example, from their sample of 2096 people in Shanghai and Tianjin in 1993, 17.8% of whom had joined the Party, Bian, Shu and Logan (2001) found no one who had withdrawn from membership of the Party.
4 During the process of admission, the Party is screening prospective members to see if they are desirable recruits. It looks for recruits who are committed to its values and will participate actively in politics. Beyond this, it is possible to discern several important overall trends in the CCP’s criteria when screening new members (Bian, Shu and Logan, 2001). In the revolutionary period prior to 1949, the CCP was engaged in first an armed struggle against the nationalist government, initially acting underground and ultimately engaging in conventional warfare. Revolutionary ideals and loyalty to the Party were the qualities sought in new members but class background was often used as loyalty filter. Similar criteria existed in the planning period, between 1950 and 1978. Parental membership of the CCP was added to class background as a way of screening for loyalty. However, political commitment and participation was also required from all members. Owing to the key role of the Party in administering the economy and the machinery of government, the Party membership became increasingly urbanized and professional competence rose in importance. Exemplary workers were often targeted for recruitment. As mentioned previously, the Cultural Revolution represented something of a backlash against the increasing managerialism of the Party and during this period of discord, intellectuals were regarded with suspicion. The reform period from 1979 onwards saw a marked change in the criteria for admission to the party. With the prioritization of economic growth, the CCP sought to recruit economically productive members. Education replaced class background as an explicit criterion for membership. Reaction to the student’s movement of 1989 entailed a more negative attitude towards recruiting educated young people and a ban on private entrepreneurs entering the Party. This did not endure and in 2001 President Jiang Zhemin openly invited private business people to join. Often already having entered by “the backdoor”, entrepreneurs were now openly acknowledged as part of the “advanced productive forces” that the CCP was now supposed to stand for according to Jiang’s “Three Represents” theory 7 . In summary, the CCP’s screening criteria have an important effect on who joins, although the particular criteria used have changed markedly over time. The individual’s demand for Party membership In some contexts, joining a political party can be thought of as an act of what Sen (2002) terms “commitment”, a choice not directly related to considerations of personal material well­being. This might be thought to apply in particular to the CCP, whose writings stressed the importance of members selflessly putting their personal interests behind those of the Party. However, while ideological reasons may have led some to join the CCP in the revolutionary period, the level of commitment is likely to have diminished as the realities of life under communist rule became apparent. Political campaigns such as the mass mobilizations during the Cultural Revolution can be seen in part as an attempt to recreate the ideological fervor that drove Party members before 1949. But in the reform period, it seems increasingly unlikely that people joined the CCP for ideological reasons as the party gradually abandoned most of the communist ideology that was its raison d’etre. Traditional communist precepts were overturned by contrary slogans such as that of Deng Xiaoping that “to get rich is glorious”. After over twenty years of reform, this culminated in the CCP lacking a clear ideological basis 8 . A survey of university students in Shanghai in 1988 suggested widespread cynicism about the reasons people became Communist Party members (Guo, 2002). A total of 2,083 students in 18 universities were asked about their friends’ motivations for joining the CCP. 49% of respondents picked the response agreed that “In reality they want a “Party card” 7 Under Jiang Zhemin’s “Three Represents Theory”, the CCP is supposed to represent "advanced productive forces, advanced Chinese culture and the fundamental interests of the majority." 8 This is pithily expressed in two quotes from CCP members cited by Rosenthal (2002):"What does the Communist Party stand for now? Nothing. Stability maybe. But really no ideals at all.” and "There is almost no one now who believes in the party for its ideals. If I walked into my former office and started praising the party, people would look at me like some weirdo."
5 which they can use as capital to receive future benefits”. Only 4% chose the option “They believe in Communism and want to make a contribution”. Indeed, far from people joining the CCP for ideological reasons, the perception that party membership provides material benefits and is a vehicle for opportunistic careerism may have deterred more idealistic people from joining the Party during the reform period. Instead, joining the Party can be conceived as being based on a cost­benefit analysis of private material advantage. Just as education is regarded as an investment in human capital, CCP membership can be regarded as an investment in what is sometimes termed political capital. The costs of membership are not primarily financial – membership dues are low. The costs may be largely up front – subjecting oneself to the long and intensive screening procedure described above. Subsequent costs take the form of time devoted to CCP activities and submission to scrutiny and discipline. However, party membership may bring various forms of benefit: additional income, additional perks, higher status and greater power or influence. The non­economic benefits can be ends in themselves as well as providing means of obtaining economic benefits. During the planning period, economic benefits might be expected from Party membership because of the CCP’s pervasive and formalized role in the allocation of labor. As Groves et al. (1995, p.876) state: “the Communist Party … functioned more or less as the personnel department of this enormous corporation, maintaining dossiers and tracking managerial careers.” The Party approved promotions and certain posts of responsibilities were reserved only for party members. Favorable treatment of Party members might be regarded as an unintended by­product (or corruption) of the planning system. However, while discrimination is often regarded as dysfunctional in a competitive economy (Becker, 1957), the reverse may be true in a communist system. As Walder (1995, p.323) states “career rewards for the politically loyal has been one of the foundations of communist party rule”. There are prima facie reasons to expect the economic benefits from Party membership to be eroded as China moves towards the market. Urban reforms in the 1980s largely removed the formal role of the CCP in determining promotions within state­owned enterprises (SOEs hereafter). The Party still approved the appointment of enterprise managers, but these managers then had autonomy over promotions within the enterprise. Greater managerial autonomy within the SOE sector, coupled with an expansion of private sector, might be expected to lead wages to be set according to worker productivity rather than political loyalty 9 . However, the evidence of an increasing wage premium on CCP membership immediately contradicts this. A more sophisticated hypothesis is that the initial stage of marketisation involves two changes which could increase the private value of CCP membership. One is the fact that productive, personal or power relations can now attract monetary rewards, a symptom of which is the general rise in income inequality and the widening of wage structures. The other is the creation of potential rents in an interventionist, semi­marketized economy: CCP membership may assist rent­seeking, whether in the labour market or elsewhere. Econometric specification The selection process can be formalized as follows. Let P * represent the unobserved net utility placed on CCP membership by an individual and V * represent the unobserved net utility to the party of the individual’s membership. We postulate the index functions P * = a´ X + u (1) V * = b´Z + v (2) where X is a vector of personal characteristics influencing individual preferences, Z is a vector
9 At the time of writing, Party membership is still required for certain top posts, for example, CEOs of large SOEs, Heads of Government Divisions and even school Head Teachers. 6 of characteristics that the party values in its members, and u and v are error terms. If P* is positive, the person wants to join the party, and if V * is positive, the party wants him to join. Where M is a dummy variable indicating party membership, the decision rule is M = 1 iff P* > 0 and V* > 0 M = 0 iff P* < 0 or V* < 0. If u and v have the standard properties, the probability of a person wanting to join is Φ(a´X ) and the conditional probability of being chosen is Φ (b´Z), where Φ(.) is the standard normal cumulative distribution function. On the assumption of independence between these choices, the probability of observing that a person belongs to the CCP is Φ(a´ X)Φ( b´Z ). In the absence of rationing, only Φ(a´ X) is relevant: all the parameters in b´Z except the constant term are zero. Identification of the separate equations is problematic if both decisions are influenced by the same set of variables (Z= X). In the empirical analysis of Section 3, we use a simple binary probit to model whether urban workers are Party members: M * = c´ X + ε M = 1 iff M* > 0 (3) The significant coefficients c in the estimated equation must be examined for consistency with screening (CCP preferences) and demand (personal preferences) interpretations. Some insight into these alternative interpretations of the determinants of CCP membership can be provided by looking at how the wage premium for membership varies with individual characteristics. Wage functions can be estimated separately for party members and non­ members: lnY = dp´ X + sp if M=1 (4) = dn´ X + sn if M=0 where lnY is the log wage and s is the error term. The wage gains from party membership are thus: E(lnY |M=1) ­ E(lnY |M=0) = (dp­ dn)´X + E(sp|M=1) – E(sn|M=0) (5) The differences in vectors of coefficients dp and dn can then be compared with the vector c in the party membership equation. If membership is primarily demand­determined, one might expect a correspondence of c and (dp ­ dn): characteristics that increase the economic benefit from membership also increase the probability of membership. However, such a pattern will not necessarily be observed in a rent­seeking equilibrium. A higher benefit (rent) for members of one type may induce more people of that type to join the Party, driving the average benefit for them down to equal the cost of membership. However, a positive correspondence between c and (dp ­ dn) may arise if the costs of membership to individuals are heterogeneous. Although the monetary costs (Party dues) may not differ, some individuals are likely to be more averse to the non­ pecuniary costs (being subject to scrutiny, having to attend dull meetings, etc.). Higher wage benefits will be required to induce such people to join. Hence one will tend to see higher wage premia for groups with higher rates of membership. Conversely, if membership is determined by the Party’s screening ­ a rationing process, rather than demand ­ one might expect a negative relation between c and (dp ­ dn). Suppose membership provides rents that only members of a certain type can compete for. If the Party restricts the numbers of members of one type, there will be less competition for the rents available for this type and thus higher average rents. For example, suppose Party membership is required for lucrative posts of responsibility within two groups of workers (say, employees at factories A and B). If the Party acts to restrict the supply of workers of one type (say, at factory A) who become members, then the lucky few who are members will have a greater chance of being given a lucrative post. In this instance, lower probabilities of
7 being a member will be associated with higher returns to membership. Another issue arising in the paper concerns the interpretation of the apparent increase in the wage premium associated with CCP membership. Preliminary findings indicate that a premium exists and that it has been rising over time. This raises a standard problem: does M cause Y, ceteris paribus, or does Y cause M, or does some third factor, J, cause both Y and M? It is possible that the Party, at least in recent years, has sought out economically successful people for membership, and that it has done so increasingly over time. It is also possible that unobserved ‘ability’ is a determinant both of CCP membership and of wages, and that both of these relationships have become more important over time, i.e. the Party has become more meritocratic and merit has been more rewarded in the labour market. This, too, could explain the apparent premium and its rise over the years. These problems can be viewed as arising from correlations between the unobserved factors, ε, which determine party membership and those unobserved factors, s, that determine wages. Consistent estimates can be dealt with using the sample selection correction suggested by Heckman (1979). This requires that we have instruments for membership: variables that are closely correlated with M but are not direct determinants of wages. In the data that we have, information on parental CCP membership would appear a priori to be a good instrument. A second way of correcting for this problem is to use panel data. If the unobserved determinants that cause biases are time­invariant, then they will be removed by using a fixed effects estimator. We are able to construct a short panel for the late 1990s using recall data on wages. Unfortunately, we do not know when the respondents joined the CCP, so we are not able to obtain a fixed effects estimate of the overall wage premium for Party membership 10 . However, we are able to use the panel to see if there is a change in the wage premium over time after controlling for the unobserved time­invariant characteristics of individuals, such as their “ability”. 3. Empirical determinants of CCP membership What characteristics increase the probability of workers being CCP members? In this section we use binomial probit models to estimate the empirical determinants of an urban worker being a Party member in a given survey. We confine the analysis to the sample of urban workers earning wages, since our primary focus is on the effect of party membership on wages 11 . Two sets of explanatory variables are included. One is of personal characteristics: sex, age, ethnicity, education and city of residence. The other set is of job­related characteristics: dummy variables for variety of categories of occupation, ownership and industrial sectors. We estimate one model for each of the survey years, noting the rise in party membership in our data 23.5% in 1988 to 24.5% in 1995 and 26.9% in 1999 (descriptive statistics are given in Table 1). Table 2 reports the results of the probit models, while Table 3 considers the predictions of the model evaluating at the means of the explanatory variables (Table 3 refers). Pairwise Wald tests were conducted for whether the coefficients in the models changed significantly over time. Of the personal characteristics, being male, educated and experienced all increased the probability of being CCP membership in each of the three surveys. However, the “pure gender gap” in the probability of being in the Party narrowed over time. Evaluating at the means of other explanatory variables, the probability of a woman being in the Party doubled from 8% in 1988 to 16% in 1999; for men, the corresponding increase was from 22% to 24% 12 . There 10 That is to say, we do not have data on income before and after joining the Party. However, if it is possible that such information may not be informative: people do not leave the CCP (so reducing ‘events’), and the economic benefits may flow only some time after the event. 11 In what follows we do not repeat the caveat that our statements apply only to urban workers and not to rural­urban migrants, although this should be borne in mind. 12 Wald tests imply significant changes in the coefficients on being female and on years of education between each survey interval.
8 was also been a significant rise in the importance of education in determining Party membership. For example, compare two workers with otherwise average characteristics: one who had college education (15 years) and the other who had senior middle school education (nine years). In 1988, such a college graduate had a 22% predicted probability of being a Party member, rising to 38% in 1999. By contrast, the probabilities for a comparable worker with senior middle school education were virtually unchanged at 12% in 1988 and 13% in 1999. The chances of less educated workers being Party members fell, ceteris paribus, in the same interval. Experience also appeared to have become increasingly important in raising the probability of Party membership, although the changes were less marked than with gender and education 13 . Only in the most recent survey did people of non­Han Chinese ethnicity have a significantly lower probability of being Party members (a 15% chance compared to 21% for Han Chinese, evaluating at the means of other variables). Turning to the work­related variables, factors that make it more likely individuals were that party members include being a white collar worker, being employed in an SOE and working in government administration. That white collar workers were more likely to be Party members might seem surprising given the Party’s ideological claim to represent the proletariat. However, the finding reflects the CCP’s historical role in administering many aspects of life in China. In our probit models, the single most statistically significant variable is the dummy variable for being a blue collar worker, which negatively reduces the probability of being a Party member (the default being white collar) 14 . At the mean of other explanatory variables, white collar workers had a 33% probability of being in the party in 1999 whereas blue collar workers had only an 11% probability. This differential is lower in relative terms than it was in 1988, with a significant fall between 1988 and 1995 in the absolute size of the coefficient on the dummy variable for being a blue collar worker. It is unsurprising that working in a SOE raises the probability of being a CCP member, while working in the private sector lowers it. The Party has less influence in the private sector, so recruitment there may be more costly for the Party and offer lower benefits for workers. However, there is tentative evidence that the Party has made particular progress in recruiting more members from the private sector. Our models predict that, for individuals with otherwise average characteristics, the probability of being in the CCP in 1988 was 16% for those in the state owned sector compared to 1% for those in the private sector. By 1999, these predicted probabilities had changed to 23% for SOE workers and 7% for private sector workers. However, this evidence of change is rather tentative: Wald tests revealed that the change was only statistically significant at the 11% level. Turning to the influence of industrial structure on the probability of Party membership, we can see a range of effects. Working in government administration is associated with the most positive impact on the probability of being a Party member, while working in education has the most negative effect 15 . Party membership is expected of workers employed in government administration, being taken as a sign of loyalty to the government and required for performing confidential or important tasks of the state. The finding about education is more puzzling and may reflect a historic suspicion of “intellectuals” within the CCP. Outcomes for older teachers in the surveys may have been influenced by the experiences in the Cultural Revolution, when teachers were often the first target of the Red Guard movement. This experience may have fostered lingering mutual distrust between the Party and the teaching profession. 13 Potential experience enters in a quadratic form, with an inverse­U effect. However, this reflects a diminishing effect of years of experience on the probability of being a Party member rather than a non­ monotonicity. The turning point of the quadratic – around 47 years of experience in the first two surveys rising to 93 years in the second survey – is more than double the mean years of experience in the samples. 14 Blue collar workers are defined here as industrial, service and commercial workers; white collar workers are taken here to be managers and administrators, clerks and those in professional or technical occupations. 15 The fact that we find people working in the education sector over­represented in the Party in Table 1 therefore reflects the fact that such workers have other characteristics favorable to Party membership, rather than that working in the education sector per se makes Party membership more likely.
9 4. The wage premium for CCP membership Communist Party members earn higher wages in urban China. In the 1988 survey, the mean daily wage for urban workers who were members was 29% higher than the mean for those workers who were not members. In the 1995 and 1999 surveys, the differential had risen to 33%. However, not all of these differentials can be attributed solely to membership of the Party because, as we have seen, membership is systematically related to observable determinants of wages, such as experience and education. To try to isolate the influence of membership on wages, we estimate wage functions which control for observable determinants of wages (Table 4 refers). To allow Party membership to have a differential effect of workers with different characteristics, we estimate separate wage functions for members and non­members. Whether estimating separate wage functions generates bias was explored by augmenting the wage functions with sample selectivity correction terms derived from the probits for CCP membership. For the 1999 survey, dummy variables for parental membership of the Communist Party are possible instruments for the selectivity corrections as they may meet two conditions. Firstly, parental Party membership is likely to be correlated with own membership. This may work via either demand factors (e.g. one’s parents act as role models) or supply factors (e.g. one’s parents vouch for one’s character). Secondly, parental party membership may not have strong direct effects on one’s own wages. Empirically, the first condition is satisfied: when we augment the probit model for 1999 in Table 2 with two dummy variables for parental party membership, the variables were jointly significant at the 5% level. The second condition was tested using over­identifying tests (OID) to see whether the instruments were associated with the residuals from the wage functions 16 . These OID tests could not reject the null hypothesis of no association at the 1% level of significance, although it was rejected at the 5% level for non­members (only). Given these results, we have some grounds for regarding parental party membership as valid instruments to identify corrections for sample selectivity, although the case is stronger for CCP members than for non­members. It is noteworthy, therefore, that these corrections were wholly insignificant ­ with t­ratios less than one ­ in both the wage function for members and that for non­ members (Appendix Table 1 refers). These results imply that we can simply use OLS estimates of the wage functions for 1999. Unfortunately, neither the 1988 nor the 1995 surveys inquired about parental membership of the CCP. To try to circumvent this problem, we estimated wage functions for sub­samples of workers who lived with their parents since for each member of the household, there was a question about whether they were CCP members (see Appendix Table 1 for the estimated coefficients on the selectivity terms). The comparison of the results for 1999 provides some corroboration of the results of the sub­sample – neither method gives selectivity corrections that are significant at the 5% level. In the earlier years, working with sub­samples also leads to insignificant corrections for sample selectivity, with the exception of non­members in 1988 for whom the selectivity correction was significantly negative 17 . Finding that sample selectivity does not appear to be a significant problem in most cases gives us some confidence in relying on the results of the OLS estimates in Table 4. This approach seems preferable to working with just the sub­sample of workers still living with their parents, since 16 The over­identifying test is more commonly used in the context of two stage least squares (for an exposition, see Deaton, 1997, p112). However, the extension to a Heckit model is straightforward. We estimated separate wage functions for CCP and non­CCP members using the Heckit model. Then we took the residuals from these equations and regressed them on the two instruments identifying the sample selectivity correction and the observed determinants of wages. Under the null hypothesis, the number of observations multiplied by the uncentered R­squared from each auxiliary regression should be distributed as a chi­square with 2 degrees of freedom. The test statistic we obtained for the CCP wage function was 1.47, less than 5.99, the critical value of the chi­squared distribution at 5% significance with 2 degrees of freedom. The test statistic for the non­CCP wage function was 6.70, less than 9.21, the critical value of the chi­squared at the 1% level. 17 This result implies that there was a positive correlation between the unobservables determining wages for non­ members and the unobservables determining CCP membership.
10 that entails its own sample selectivity problem. Moreover members tend to be somewhat older whereas it is the younger workers who tend to still live with their parents. Hence the subsamples include only small numbers of CCP members. In what follows, we focus on the wage functions estimated separately for Party members and non­members by OLS. Table 4 uses these functions to estimate the wage premia, P, for Communist Party membership for different kinds of workers, using a simplified version of equation (5): P = E(lnY |M=1) ­ E(lnY |M=0) = (dp­ dn)´X (5’) We obtain baseline wage premia by evaluating the sample means of the explanatory variables X (CCP members and non­members combined) 18 . This shows that for a worker with average characteristics, they would have earned 10% more in 1988 if they were in the CCP, rising to 14% in 1995 and 1999. Then, in a manner analogous to that used to obtain the predicted probabilities of membership, we simulate the effect of altering a given explanatory variable (e.g. sex) while holding the value of all other variables at their sample means. Generally speaking, the personal characteristics which increase the probability of CCP membership – male sex, education and experience – are associated with a lower wage premium for membership. For example, although men are more likely to be party members, the CCP wage premium (evaluating at the mean of other variables) was only 11% for them in 1999 ­ compared to 16% for women. A worker with no experience and otherwise average characteristics would earn 32% more in 1999 if they were a CCP member whereas one with 20 years of experience would earn only 4% more. In the same year, a primary school graduate (with 6 years of education) with average characteristics would receive 17% higher wages if they were in the Party compared to only the 11% premium enjoyed by college graduates (15 years of education) 19 . Interestingly, all three personal characteristics ­ male sex, education and experience ­ are associated with higher wages. Thus, it appears that CCP membership may in some way substitute for such remunerative characteristics. This pattern is also suggested by the results of wage functions estimated by quantile regressions. The wage functions in Table 4 were augmented with a dummy variable for CCP membership and estimated using quantile regressions on pooled (members and non­members combined) samples of workers. In both 1995 and 1999, the implied CCP wage premium was higher for the lower quartile than for the median or upper quartile 20 . The lower quartile estimate applies to workers with unobserved characteristics that are less favorable for earnings. Hence it appears that CCP membership, by helping these kinds of workers more, is once again substituting for the other remunerative characteristics (in this case, unobserved ones such as “ability” or luck). It is also noteworthy, that the rise in the Party wage premium is more marked for the lower quartile than for the median or the upper quartile. Being a white collar worker ­ like being educated or experienced ­ is another remunerative characteristic that increases the chances of Party membership but is associated with a lower Party wage premium. In 1999, “blue collar” workers with otherwise average characteristics would earn 21% more if they were in the Party; “white collar” workers would earn only 7% more. Looking at the coefficient on the dummy variable for being a “blue collar” worker in 18 Strictly speaking, P is an approximation of the percentage wage differential; Table 4 uses the transformation exp (P)­1 (see Halvorsen and Palmquist, 1981). 18. The one exception to this generalization concerns the effect of ethnicity in 1999(only). Non­Han Chinese were less likely to be CCP members, and the wage functions imply no CCP premium for such workers. 19 We performed pair­wise Wald tests for whether coefficients on particular explanatory variables differed significantly between the CCP and non­CCP wage functions. The coefficient on the sex dummy variable differed significantly at the 5% level in 1988 and at the 10% in 1999. The coefficients on education differed significantly at the 5% level in 1988 and 1995; those on experience (and its square) differed significantly at the 5% level in 1988 and 1995. 20 The quantile results are not reported but are available upon request. The CCP premium for the lower quartile was 5% in 1988, rising to 8% in 1995 and 11% in 1999. For the upper quartile, the rise was from 5% through 3% to 7%. For the median quantile regression, the corresponding figures were 4%, 6% and 8%.
11 Table 4, it appears as if such members have been protected from the widening occupational wage differentials since 1988 experienced by non­members. Thus pair­wise Wald tests show that, at the 5% level, the variable has significantly different coefficients for members and non­members in both 1995 and 1999. The coefficients on the dummy variables for the ownership and industrial sector of the worker’s enterprise do not generally differ significantly as between Party members and non­ members. This is perhaps surprising, since it implies that membership does not bring significantly greater returns in the SOE sector compared to the collective or private sectors. The main exception is the dummy variable for being in government administration. In 1988, this variable had a significantly smaller coefficient for CCP members than for non­members 21 . Thus there was a smaller CCP premium for civil servants than for other kinds of workers; for example, a 5% premium for workers with otherwise average characteristics compared. However, by 1999, the reverse was true ­ the variable had a significantly larger coefficient for CCP members and the CCP wage premium for government administration workers stood at 20%. In summary, we find little evidence to support the hypothesis that characteristics that make it likely that workers join the CCP also tend to raise the wage premium for membership. It is true that Party membership appears to pay more for those working in government administration and such workers are more likely to join the Party. However, the balance of the evidence demonstrates contrary patterns. Ceteris paribus, men are more likely to join, but the premium is lower for them. Similarly, more educated and experienced workers are more likely to be members but the returns to education and experience are lower for members. Blue­collar workers are the least likely to be members, but are the occupational category that appears to benefit the most from Party membership. Some characteristics that raise the probability of membership – such as employment in an SOE – appear to have no independent effect on the wage premium for Party membership. As suggested in Section 2, this pattern seems consistent with the notion that membership of the CCP is primarily determined by screening, rather than by individual demand for membership. The Party prefers to recruit people who will keep the system going; with marketisation, these are increasingly those with characteristics that are rewarded through higher wages. However, such people do not necessarily benefit more than others from being Party members. If the value of Party membership depends partly on scarcity, then among groups of workers where membership is relatively common, the average benefit of membership may be lower. For example, among particular groups of workers, there may be a given number of posts of leadership or responsibility that attract higher wages and which Party members are more likely to be given. The more CCP members within a group of workers competing for those posts, the less the likelihood that anyone member will be successful and hence the lower the average wage premium for CCP membership. 5. The CCP wage premium during retrenchment So far, we have relied on cross­sectional estimates of the effects of CCP membership on wages. The main potential methodological problem with this approach is that the endogeneity of membership may give rise to biases in ordinary least squares estimates. Using techniques to control for sample selectivity, we have argued that such biases are not a significant problem with our data. However, an alternative approach is to use fixed effects estimates derived from panel data on individuals. If CCP members have unobserved characteristics that make them more productive and hence higher paid, then these characteristics should be controlled for as part of the individual­level fixed effect. The 1999 survey included recall questions on wages in the previous four years. 21 Interestingly, in 1988, working in government administration led to significantly lower wages than working in manufacturing but by 1999 this had been reversed.
12 Consequently, we can use these data to construct a retrospective panel of observations on wages from 1995 to 1999. Unfortunately, most of the potential explanatory variables (excluding experience) can appear only as time­invariant variables in the analysis. For example, we do not have data on when people joined the CCP and so cannot estimate the overall membership premium based on wages before and after joining the Party. However, we can still use a fixed effects model to explore whether there are changes in the effects of these variables over time. Specifically, we can use the panel to see how the Party premium has changed during the years 1995­1999, which was a period of large scale retrenchment of labor. Table 5 reports the results of the fixed effects estimates, estimated separately for the non­ retrenched and the retrenched workers 22 . We classify workers as retrenched if they were retrenched at any time between 1992 and 1999/2000; hence this classification is time­ invariant 23 . To estimate the possible effects of re­employment on wages, we include a time­varying dummy variable for being re­employed 24 . Since the wage structure may be different for the re­employed, we interact this dummy variable with variables for personal characteristics and also the year dummies. The results of this analysis are reported in detail in a companion paper (Appleton et al., 2004). Here we focus on the results about CCP membership. For the majority of workers, those who have not been retrenched, the interaction term between the dummy variable for 1999 and that for Party membership is statistically significant with a coefficient of 0.059. This implies that, compared to the base year of 1995, the premium for such workers has risen by 6 percentage points. Inspection of the interactions between the CCP dummy and dummies for intervening years indicates that the rise of the premium was sustained and incremental during the period. Since this result is generated by a fixed effects estimate, it is not possible to argue that the rise in the premium is due to a chance in the composition of the party membership, arising, for example, from the CCP trying to recruit more entrepreneurial members. What is particularly interesting is how the premium for Party membership varies with retrenchment and re­employment. It has previously been shown using the 1999 survey that being a member of the CCP reduces the probability of being retrenched, ceteris paribus, although it has no effect on the conditional probability of re­employment (Appleton et al, 2002). In the fixed effects wage function for “retrenched” workers, the interactions between the dummy for Party membership and the year dummies are near zero and statistically insignificant. That is to say, for those workers who were retrenched in the period 1995 to 1999, there was no tendency for the premium to rise prior to their retrenchment. Moreover, the interaction between the dummy for Party membership and that for re­employment is significantly negative. The overall effect of re­employment on wages cannot be readily evaluated since there are several interaction terms between re­ employment and other determinants of wages. At the mean of all explanatory variables, the overall effect is positive: that is to say, retrenched workers earn more than they did in their previous jobs if they are re­employed. However, a Party member with other personal characteristics equal to the mean for the re­employed would earn 21% less when re­ employed in a new job than they would have earned if they had remained in their old job. 22 We do not correct for the selectivity of these two groups of workers because this can be viewed as giving rise to differences in the time­invariant unobserved characteristics of the individual which are eliminated by the fixed effects estimation. 23 Such “retrenched” workers are likely to have spent some of the period of the panel in their previous jobs prior to retrenchment, some of the time unemployed and possibly some of the time re­employed in new jobs. Where workers are unemployed for a whole year, they have no wage and are thus not included in the panel analysis for that year (hence our panel for the retrenched workers is unbalanced). 24 Where a worker works in both their pre­retrenchment job and their re­employed job in the same year, we use the re­employed wage rate for that year.
13 6. Conclusions The economics of Communist Party membership in China is curious in several respects. A first paradox is that both the number of members and the size of the wage premium for membership appear to have risen during economic transition. With the move away from a command economy, one might expect Communist Party membership to have become less important in determining a worker’s wages and general welfare. Such a reduction in the economic benefits of Party membership might reduce the demand for membership and lead to a fall in recruitment. Much as Marx had envisaged the state withering away under communism, so might one anticipate the same fate for the CCP as China moves from planning towards the market. However, the reverse has been observed in China during the transition. The membership of the Party has risen and the rising wage premium for members estimated in this paper helps us understand this phenomenon. A comparison of urban household surveys shows that the “pure” wage premium for Party membership – that part of the wage differential that cannot be explained by other factors that we observe – has risen from 10% in 1988 to 14% in 1999. The two trends of a rising wage premium and a rising membership are consistent with a demand­side explanation, whereby individuals invest in Party membership as a form of political capital. In terms of the aggregate time series evidence, a rise in the monetary benefits might be said to have induced an increase in membership. To some extent, the rise in the wage premium may be a disequilibrium phenomenon. If the Party fully accommodates the increase in demand, the greater initial rents will induce more people to join the Party, lowering the average return. However, such a simple demand side explanation seems inadequate to explain a second paradox observed in our cross­sections ­ that the people who are the most likely to join the Party receive the lowest benefit. It is true that Party membership is higher, ceteris paribus, among workers in the government administration sector and that the wage premium for membership is also higher in that sector. In most other cases, worker characteristics that raise the likelihood of a worker being in the Party also reduce the wage premium for membership. Education, experience, male sex and being in a white­ collar occupation all significantly raise the probability of a worker being a Party member. These characteristics are also rewarded with higher wages, for both members and non­ members. However, they appear to be more rewarded for non­members than for members. In general, therefore, one cannot explain the higher rates of membership among certain groups – such as men or the educated – by a higher expected benefit. If anything, membership rates are higher when the premia are lower. This second paradox suggests that Party membership is supply­constrained rather than demand­constrained: the cross­sectional patterns are determined by the Party’s screening decisions, not by individuals’ demand to join. Although the Party has allowed the number of members to rise, it has nevertheless rationed places. The rationing criteria reflect the personal characteristics that are valuable for the CCP’s political and economic objectives. As a result of there may be a scarcity value to membership for workers with less favored characteristics. For example, Party membership leads to higher wages partly through providing increased access to a certain number of responsible posts. Securing these posts may provide more of an increase in wages for a person with otherwise low­return characteristics than for one with high­return characteristics. Hence, Party membership may partly substitute for experience or education in obtaining certain posts. This effect may be reinforced if competition for some posts is limited to people of similar characteristics. For example, a factory floor supervisor may have to be a blue­collar rather than a white­collar worker. If there are relatively few Party members among the blue­ collar workers, then the likelihood of any one member obtaining a desired supervisory post will be higher and thus the benefits of Party membership greater. What explains the rise in the wage premium for Party members? We reject the explanation that it reflects a change in the composition of the CCP – that is to say, the
14 Party recruiting members with higher unobserved productivity (“ability”). Our recall data on wages prior to 1999 showed increases in the wage premium even after controlling for unobserved time­invariant individual characteristics. Our evidence also contradicts the hypothesis that the rise in the premium reflects an increase in the returns to “ability”. When we used parental party membership as an instrument to correct for the selectivity of Party membership, we found no evidence that those recruited into the CCP tend to have higher unobserved productivity. Instead, the rise in the wage premium for members – much like the rise in the gender gap in wages ­ may be a by­product of the increase in wage differentials during the transition from planning. Whereas before income inequalities were compressed for political reasons, under reform, enterprises have more discretion in setting wages. This may give more room for discriminatory, as well as productive, factors to work in determining wages. CCP members may more be able to secure personal benefits from their political status during the transition from planning, much as managers in Eastern Europe and Russia were accused of using their positions to benefit from the privatization of assets in the 1990s. Although the rise in the wage premium may have been fortuitous rather than planned,, the Party may well have had an interest in allowing it to happen. If the non­economic benefits of membership diminish – as ideology and status become less important ­ a higher premium may be necessary to keep up recruitment and so maintain Party control. The prospect of increased economic rewards to activities other than rent­seeking entry to the Party may also require that the return to membership be increased. Nonetheless, it is interesting that retrenchment and re­employment has a particularly negative effect on the wages of Party members. When turned into “outsiders” from their enterprises, Party members seem ill­equipped to flourish in a harsher, more competitive environment. What are the implications of the rise in the wage premium for party members during reform? At face value, it contradicts Nee’s market transition hypothesis of a rise in the return to productive characteristics and a fall in the returns to unproductive ones. This need not be a concern on efficiency grounds if the rise in the wage premium could actually be explained in terms of the rewards to unobserved productivity of Party members. However, the evidence given in this paper that the rising premium does not appear to be explained by either an increase in the average “ability” of Party members or an increase in the returns to ability casts some doubt. The rise in the premium may indicate a limitation in the extent to which the Chinese labour market has become competitive. In distributional terms, an increasing wage premium would seem to be undesirable. Not only is it an unjustified horizontal inequality but it is also likely to worsen vertical inequalities since CCP members typically have characteristics ­ such as education and experience ­ which are economically rewarded in their own right. However, the political implications of these distributional changes are less clear­cut. One could take a sanguine view and argue that growing benefits for Party members are desirable because they help to ensure that the reforms are self­sustaining. If CCP members do well during the transition, they will continue to support the reforms. Nonetheless, if these benefits are viewed as unjustified, they risk discrediting both the Party and the reforms.
15 References Appleton, Simon; Knight, John; Song, Lina; and Xia, Qingjie. “Labour retrenchment in China: Determinants and consequences”, China Economic Review, 14 (2­3) (2002): 252­ 75. Appleton, Simon; Knight, John; Song, Lina; and Xia, Qingjie. “Contrasting paradigms: Segmentation and competitiveness in the formation of the Chinese labour market”, Journal of Chinese Economic and Business Studies, 2 (3), (2004) 185­205. Becker, Gary S. The economics of discrimination, Chicago: University of Chicago Press (1957). Bian, Yanjie; Shu, Xiaoling; and Logan, John R. "Communist Party membership and regime dynamics in China", Social Forces, 79 (2001): 805­41. Bird, Edward J.; Joachim, Frick R.; and Gert Wagner G. “The income of socialist upper classes during the transition to capitalism: Evidence from longitudinal East German data”, Journal of Comparative Economics 26 (2) (1998): 211­25. Deaton, Angus. The analysis of household surveys, Johns Hopkins University Press: Baltimore (1997). Djilas, Milovan. The new class: an analysis of the Communist system in power New York: Praeger (1957). Economist Intelligence Unit. China: country profile (2004). http://www.economist.com/countries/China/profile.cfm?folder=Profile­Political%20Forces Frye, T.; and Shleifer, A. “The invisible hand and the grabbing hand”, American Economic Review 87(2) (1997): 354­58. Griffin, Keith; and Zhao, Renwei (Eds.). The distribution of income in China. Macmillan: London (1993). Groves, Theodore; Hong, Yongmiao; McMillan, John; and Naughton, Barry. “China’s evolving managerial labour market” Journal of Political Economy 103 (4) (1995): 873­92. Guo, Gang. Party Recruitment and Political Participation in Mainland China PhD, Political Science, University of Rochester (2002). Halvorsen, Robert; and Raymond, Palmquist. “The interpretation of dummy variables in semilogarithmic equations”, American Economic Review 70 (3) (1981): 474­75. Heckman, J. “Sample selection bias as a specification error”, Econometrica, 47 (1979): 153­ 61. Konrád, George; and Szelenyi, Iván. The intellectuals on the road to class power: A sociological study of the role of the intelligentsia in socialism New York: Harcourt Brace Jovanovich (1979). Morduch, Jonathan; and Sicular, Terry. "Politics, growth, and inequality in rural China: Does it pay to join the Party?" Journal of Public Economics 77 (2000): 331­56. Naughton, Barry. Growing out of the plan: Chinese economic reform, 1978­1993 Cambridge University Press, (1995).
16 Nee, V. “A theory of market transition: from redistribution to markets in state socialism”, American Sociological Review 54 (5) (1989): 663­81. Pei, Minxin. “China's governance crisis” Foreign Affairs 81(5) (2002): 96­109. Riskin, Carl; Zhao, Renwei; and Li, Shi. China’s retreat from equality: income distribution and economic transition. Armonk, New York: M.E. Sharpe (2001). Rona­Tas, A. “The first shall be last? Entrepreneurship and communist cadres in the transition from socialism”, American Journal of Sociology 100(1) (1994): 40­69. Rosenthal, Elisabeth. “China's communists try to decide what they stand for”. New York Times, (May 1, 2002). Sen, A. Rationality and freedom, Cambridge, MA: Belknap Press (2002). Solotaroff, Jennifer. “Gender inequalities in urban China’s career mobility patterns”, mimeo, Department of Sociology. Stanford: Stanford University (2003). Szelenyi, Iván. “The prospects and limits of the East European new class project: An auto­ critical reflection on The Intellectuals on the Road to Class Power”, Politics and Society 15 (1986): 103­44. Walder, Andrew G. “Career mobility and the communist political order”, American Sociological Review 57 (3) (1995):524­39.
17 Table 1 Characteristics of CCP members and non­members Descriptive Statistics of Workers: 1988, 1995 and 1999 Surveys Observations Male Female 1988 Non­CP CP 13,571 4,162 (76.53) (23.47) 55.12 76.19 1995 Non­CP CP 9244 3001 (75.49) (24.51) 46.48 71.34 1999 Non­CP CP 4592 1689 (73.11) (26.89) 48.78 68.80 44.88 23.81 53.52 28.66 51.22 31.20 3.55 4.47 4.45 3.87 4.44 3.43 16.54 8.81 15.52 2.69 0.50 2.74 34.99 (9.97) 35.00 19.24 (10.72) 19.00 21.26 11.12 19.67 2.98 0.38 2.98 43.95 (8.60) 44.00 26.97 (9.41) 27.00 25.43 23.05 21.81 2.99 0.77 3.08 36.87 (9.43) 37.00 20.68 (10.44) 21.00 34.04 25.52 29.45 3.36 0.60 3.38 43.76 (8.38) 44.00 26.05 (9.20) 26.00 31.26 57.24 24.80 3.15 0.74 3.21 37.92 (8.90) 38.00 21.03 (9.97) 22.00 41.50 66.73 35.08 3.52 0.61 3.56 43.62 (8.03) 45.00 25.42 (9.06) 26.00 7.07 27.85 36.27 21.78 6.76 0.27 9.75 (2.48) 9.00 0.12 5.10 32.27 36.33 25.35 0.84 10.97 (2.81) 11.00 2.97 23.99 36.28 29.10 7.31 0.34 10.19 (3.12) 10.00 0.00 6.23 28.72 41.92 22.16 0.97 11.71 (3.23) 12.00 1.72 21.36 35.74 34.17 6.86 0.15 10.89 (2.63) 11.00 0.00 7.10 26.11 47.19 19.36 0.24 12.19 (2.75) 12.00 College and above 8.51 26.29 17.68 40.36 21.37 50.44 Professional high school 9.44 16.12 15.61 19.33 13.09 14.51 Senior middle school 26.42 19.39 26.57 17.49 29.03 18.18 Lower middle school 41.63 28.76 33.71 20.03 33.62 15.57 Primary school and below 13.35 8.89 6.44 2.80 2.90 1.30 0.46 0.55 0.00 0.00 0.00 0.00 73.46 24.12 0.98 0.41 91.42 7.74 0.07 0.22 75.42 17.63 2.13 1.54 90.17 7.13 0.17 0.47 72.91 15.68 5.99 2.20 89.82 6.45 0.77 1.42 1.03 0.55 3.28 2.07 3.22 Minority Mean of daily wages Standard deviation Median Mean of log daily wages Standard deviation Median Mean of age Standard deviation Median Mean of experience Standard deviation Median Age group: 16­20 (%) 21­30 (%) 31­40 (%) 41­50%) 51­60%) 61­65 (%) Education in year Standard deviation Median Education level: Not reported Ownership: State­owned Urban collective Urban private Joint venture and Foreign investment Others 1.54
18 Occupation: Private enterprise owner or Private enterprise owner and manager White Collar Blue Collar Others Industries sector: 1. Primary sector 2. Manufacturing 3. Construction 4. Transportation and communication 5. Retail, catering, and wholesales 6. Real estate and Individual service 7. Healthcare, sports and social welfare 8. Education, culture, art quango 9. Scientific research technology services 10. Finance and Insurance 11. Government and quango 12. Others 1.30 0.91 1.62 1.00 1.68 0.71 33.76 64.27 0.67 83.47 15.23 0.38 43.55 45.05 11.24 81.41 14.00 4.40 39.09 55.73 3.51 80.34 17.64 1.30 4.34 46.84 3.57 6.73 3.44 29.31 2.88 6.80 15.53 2.42 42.99 3.06 4.99 3.33 30.22 2.30 4.47 3.22 34.54 4.57 8.36 4.38 24.10 3.73 11.66 10.76 15.66 9.83 12.26 6.81 2.68 1.71 4.01 3.17 11.61 7.99 4.27 5.45 4.29 4.67 3.96 5.68 6.43 9.78 6.67 8.46 6.60 9.12 2.56 3.96 2.07 2.90 2.09 2.43 1.37 2.04 1.88 2.03 2.05 2.13 4.05 22.71 7.20 23.99 5.40 18.12 1.64 1.15 4.74 4.63 5.34 3.85
19 Table 2 Binomial Probit Regressions: the Determinants of Being a CCP Member 1988, 1995 and 1999 Surveys Constant Male sex Experience Experience squared Full­time education in years Ethnic minority Ownership (default variable is state­ ownership) Occupation (default variable is white collar) Industry (default variable is manufacturi ng) Urban collective Private enterprises Foreign­owned or joint venture Other ownership Private enterprise owner Blue collar Other occupations Primary industries Construction Transportation and communication Commerce Real estate Social welfare Education Sciences and research Financial sectors Government Other industries 1988 ­2.883 (25.16)*** 0.628 (23.29)*** 9.62E­02 (18.14)*** ­1.06E­03 (10.04)*** 0.069 (11.48)*** ­0.003 (0.04) ­0.263 (6.67)*** ­1.309 (4.62)*** 0.082 (0.33) ­0.161 (0.86) ­0.205 (1.46) 1995 ­3.082 (24.74)*** 0.493 (16.66)*** 0.089 (14.79)*** ­0.001 (7.55)*** 0.090 (15.58)*** ­0.062 (0.87) ­0.246 (5.18)*** ­1.021 (3.96)*** ­0.278 (1.74)* ­0.109 (1.16) 0.262 (1.63)* 1999 ­3.458 (18.51)*** 0.361 (9.01)*** 6.96E­02 (8.21)*** ­4.22E­04 (2.33)** 0.135 (14.03)*** ­0.235 (2.22)** ­0.280 (4.13)*** ­0.771 (4.73)*** ­0.164 (1.11) ­0.393 (2.73)*** ­0.508 (2.73)*** ­0.963 (29.84)*** ­0.504 (2.16)** 0.010 (0.14) ­0.061 (0.85) 0.031 (0.60) ­0.733 (19.13)*** ­0.516 (8.35)*** 0.105 (1.25) ­0.132 (1.51) 0.006 (0.09) ­0.796 (16.17)*** ­0.567 (3.72)*** 0.217 (2.04)** ­0.145 (1.44) 0.307 (4.42)*** 0.000 (0.01) ­0.115 (1.29) ­0.137 (2.36)** ­0.233 (4.76)*** ­0.068 (0.96) 0.033 (0.34) 0.674 (15.06)*** ­0.030 (0.23) ­0.059 (1.23) 0.026 (0.34) ­0.091 (1.30) ­0.130 (2.24)** ­0.206 (2.18)** 0.185 (1.90)* 0.552 (11.99)*** 0.059 (0.80) 0.182 (2.35)** 0.072 (0.98) 0.006 (0.07) ­0.160 (1.98)** ­0.098 (0.74) 0.072 (0.53) 0.448 (6.23)*** 0.214 (2.07)**
20 Number of observations Log­likelihood Restricted log­likelihood Pseudo R­squared 1988 Actual 0 1 Total 1995 Actual 0 1 Total 1999 Actual 0 1 Total 17733 ­6563.519 ­9662.670 0.3207 0 12477 1919 14396 0 8581 1770 10351 0 4189 905 5094 12245 ­5210.770 ­6818.808 0.2358 Predicted 1 1094 2243 3337 Predicted 1 663 1231 1894 Predicted 1 403 784 1187 6281 ­2776.994 ­3656.597 0.2406 Total 13571 4162 17733 Total 9244 3001 12245 Total 4592 1689 6281 Notes: (1) Regional dummy variables are controlled for in all models. For the brevity, the coefficients are not reported here. (2) T­ratios are in brackets. *** denotes statistical significance at 1% level and below, ** at 5%, and * at 1% level.
21 Table 3 Probabilities of CCP Membership and Wage Premium (percentages) Baseline Male Female Experience of 0 years Experience of 10 years Experience of 20 years Experience of 30 years Education of 0 years Education of 6 years Education of 9 years Education of 12 years Education of 15 years Han Chinese Ethnic minority Ownership State­owned Urban collective Private enterprises Foreign­owned/ joint venture Other ownership Occupation Private enterprise owner White collar Blue collar Other occupations Industry Primary industry Manufacturing Construction Transportation and communication Commerce Real estate Social welfare Education Sciences and research Financial sectors Government Other industries (a) Predicted probability CCP member 1988 1995 1999 14 18 21 22 25 26 8 12 16 1 1 2 5 6 7 17 18 19 30 31 34 4 3 1 8 9 6 12 15 13 17 22 24 22 30 38 14 18 21 14 17 15 (b) Predicted CCP wage premium 1988 1995 1999 10 14 14 9 11 11 12 17 16 34 25 32 15 23 10 6 12 4 5 12 6 23 28 21 15 20 17 12 16 15 9 12 13 5 8 11 10 13 14 14 23 ­1 16 10 1 20 14 3 23 16 7 9 13 ­20 13 15 50 15 5 9 18 12 13 17 19 13 38 99 6 23 7 17 23 29 7 15 37 28 9 13 17 33 11 16 3 9 11 11 34 9 20 14 ­27 7 21 39 14 14 12 20 17 14 24 18 15 11 12 14 13 14 18 1 9 12 14 14 11 11 9 12 15 34 13 17 15 18 15 14 12 22 34 18 27 23 20 18 14 16 20 32 24 8 8 11 13 10 8 12 5 12 23 15 13 4 9 20 14 8 25 20 20 11 12 11 2 13 20 36 Notes: (1) All results evaluated at the mean of the other explanatory variables. (2) Probabilities generated from Table 2, wage premia from Table 4.
22 Table 4 Wage functions for 1988, 1995 and 1999 for CCP members and non­members 1988 1995 Non­CP CP Non­CP CP Male 0.102 0.073 0.144 0.087 (14.54) (5.16)*** (10.26)** (3.84)*** *** * Experience 4.96E­02 3.01E­02 5.987E­ 5.10E­02 (36.54) (9.58)*** 02 (9.29)*** *** (20.91)** * Experience ­7.30E­04 ­3.56E­04 ­1.052E­ ­8.82E­04 squared term (23.05) (5.82)*** 03 (7.71)*** *** (15.21)** * Full­time 0.033 0.024 0.036 0.025 education in (16.32) (9.53)*** (12.42)** (7.33)*** years *** * Minority ­0.007 0.032 ­0.118 ­0.033 (0.39) (1.38) (3.39)*** (0.70) Ownership (default variable is state­owned) Urban ­0.144 ­0.110 ­0.253 ­0.231 collective (17.6) (4.85)*** (13.15)** (5.13)*** * Private ­0.328 ­0.642 ­0.460 ­0.171 enterprises (2.51) (1.25) (6.34)*** (0.91) Foreign­owned 0.027 0.260 0.155 0.094 or joint venture (0.26) (2.2)** (2.65)*** (1.07) Other ­0.575 0.021 ­0.303 ­0.217 ownership (5.7) *** (0.17) (5.61)*** (2.18)** Occupation (default variable is white collar) Private 0.055 ­0.008 0.053 0.258 enterprise (0.69) (0.16) (0.63) (2.15)** owner Blue collar ­0.057 ­0.041 ­0.165 ­0.066 (6.47) (2.21)** (9.74)*** (1.99)** *** Other ­0.205 ­0.186 ­0.219 ­0.171 occupations (1.76) * (0.73) (7.79)*** (2.46)*** Industry (default variable is manufacturing) Primary 0.062 0.058 0.058 0.050 industries (3.72) (2.45)*** (1.24) (0.70) *** Construction 0.018 0.033 0.001 0.032 (0.98) (1.29) (0.02) (0.62) Transportation 0.027 ­0.005 0.046 0.125 and (1.79) * (0.22) (1.25) (3.01)*** communication Commerce ­0.004 ­0.043 ­0.056 ­0.043 (0.34) (2.07)** (2.61)*** (1.18) Real estate ­0.073 ­0.083 ­0.047 ­0.051 (2.69) (2.18)** (1.29) (1.01) *** Social welfare ­0.030 ­0.017 0.110 0.022 (2.13) ** (0.82) (3.62)*** (0.47) 1999 Non­CP CP 0.195 0.153 (10.35)** (4.86)*** * 4.71E­02 3.016E­ (12.29)** 02 * (4.67)*** ­8.79E­04 ­4.986E­ (10.08)** 04 * (3.66)*** 0.041 0.035 (8.96)*** (6.18)*** 0.019 (0.40) 0.019 (0.26) ­0.166 ­0.260 (5.66)*** (3.53)*** ­0.028 (0.61) 0.344 (5.11)*** ­0.283 (3.65)*** ­0.082 (0.54) 0.269 (2.37)** ­0.272 (1.88)* 0.187 (2.06)** ­0.193 (1.09) ­0.171 (7.45)*** ­0.044 (1.02) ­0.306 (3.91)*** ­0.040 (0.32) 0.130 (2.68)*** 0.060 (0.97) 0.094 0.127 (2.03)** (2.09)** 0.283 0.377 (7.79)*** (8.09)*** 0.070 0.167 (1.98)** (2.30)** 0.213 0.235 (6.23)*** (3.73)*** 0.334 0.360 (9.31)*** (5.44)***
23 Education Sciences and research Financial sectors Government Other industries Constant term No. of observations Adjusted R­ squared ­0.050 ­0.070 0.150 0.106 0.300 0.323 (3.56) (3.86)*** (6.39)*** (3.24)*** (8.89)*** (6.48)*** *** ­0.009 ­0.042 0.156 0.207 0.325 0.264 (0.45) (1.81)* (3.74)*** (3.90)*** (6.26)*** (3.64)*** ­0.038 ­0.038 0.290 0.292 0.412 0.446 (1.32) (1.11) (6.02)*** (4.52)*** (6.72)*** (6.29)*** ­0.060 ­0.120 0.082 0.027 0.241 0.338 (3.72) (7.71)*** (3.35)*** (1.10) (6.37)*** (8.05)*** *** ­0.144 ­0.140 ­0.311 ­0.222 ­0.029 0.197 (2.58) (2.14)** (5.97)*** (2.68)*** (0.44) (2.71)*** *** 1.874 2.239 2.364 2.650 2.519 2.734 (54.28)** (39.89)** (39.83)** (28.87)** (28.77)** (22.30)** * * * * * * 13571 4162 9244 3001 4592 1689 0.364 0.247 0.273 0.298 0.283 0.264 Notes: Dependent variable is log hourly wage; T­ratios are in brackets. *** denotes statistical significance at 1% level, ** at 5% and * at 1% level.
24 Table 5 Fixed effects estimates of changes in wages function coefficients, 1995­1999 Male*99 Male*98 Male*97 Male*96 Experience*99 Experience*98 Experience*97 Experience*96 Experience squared*99 Experience squared*98 Experience squared*97 Experience squared*96 Education in years*99 Education in years*98 Education in years*97 Education in years*96 Minority ethnicity*99 Minority ethnicity*98 Minority ethnicity*97 Minority ethnicity*96 Party member*99 Party member*98 Party member*97 Party member*96 Non­retrenched Retrenched Coeffici T­ Coefficie T­ ent ratio nt ratio ­0.003 ­ 0.113 1.52 0.26 ­0.008 ­ 0.037 0.94 0.65 ­0.005 ­ 0.035 0.91 0.45 ­0.005 ­ 0.021 0.56 0.38 ­2.07E­ ­ *** 1.41E­02 0.69 02 7.48 ­1.92E­ ­ *** ­7.92E­ ­0.75 02 7.19 03 ­1.13E­ ­ *** ­7.49E­ ­0.08 02 4.34 04 ­1.65E­ ­ ­8.78E­ ­0.94 03 0.65 03 3.60E­ 6.00 *** ­4.43E­ ­0.96 04 04 3.76E­ 6.22 *** ­2.83E­ ­0.12 04 05 2.33E­ *** ­1.46E­ ­0.64 04 3.80 04 4.32E­ 0.69 1.38E­04 0.61 05 0.012 4.61 *** ­0.012 ­0.69 0.006 2.21 ** ­0.012 ­1.25 0.003 1.11 ­0.014 ­1.50 0.000 0.11 ­0.001 ­0.09 0.088 2.84 *** 0.020 0.11 0.047 1.50 ­0.032 ­0.32 0.043 1.38 0.024 0.25 0.028 0.88 ­0.002 ­0.03 0.059 *** 0.122 0.94 4.25 0.055 3.95 *** ­0.001 ­0.02 0.039 2.76 *** 0.066 1.01 0.010 0.70 ­0.013 ­0.21 Year dummy for 1999 0.370 7.84 *** 0.037 0.12 Year dummy for 1998 Year dummy for 1997 Year dummy for 1996 0.283 0.129 0.000 6.20 *** 2.91 *** ­ 0.00 0.177 0.113 0.055 1.06 0.73 0.37 Dummy variable for re­ 0.842 2.93 employment (time varying) Interactions with a time varying dummy variable for re­employment: Male*re­employment 0.169 2.62 ** * **
25 experience*re­employment ­3.35E­ ­1.75 02 5.79E­04 1.34 Experience squared*re­ employment School years*re­employment ­0.041 ­2.54 Minority*re­employment CP member*re­employment 0.150 1.06 ­0.319 ­2.83 Year dummy for 1999*re­ employment Year dummy for 1998*re­ employment Year dummy for 1997*re­ employment Year dummy for 1996*re­ employment Constant ­0.131 ­1.19 2.16 0.066 0.70 ** * ** * ** ­0.018 ­0.19 3.089 *** 724.5 4 Number of observations R­squared across individuals 0.204 * * 26938 0.1421 2.710 206.89 ** * 4639 0.0675
26 Appendix table 1: Coefficients on sample selectivity corrections for CCP and non­ CCP wage functions (a) Full sample 1988 Non­CP CP N.A. N.A. 1995 Non­CP CP N.A. N.A. 1999 Non­CP CP ­0.094 ­0.068 (­0.91) (­0.57) 4592 1689 No. of observations (b) Sub­ sample (living with parents) No. of observations ­0.388 (2.55)*** 0.095 (0.33) ­0.331 (1.03) 1.087 (0.66) 0.098 (0.37) ­0.812 (1.79)* 2900 98 1707 97 882 87 Notes: T­ratios are in brackets. *** denotes statistical significance at 1% level, ** at 5% and * at 1% level.
27