Hukou and Graduates’ Job Search in China

I. Introduction

Since the turn of the new millenium, graduates in China have faced fierce competition when seeking employment, and graduate unemployment has been high, raising social and economic concern. Researchers have explored the reasons behind these high rates of graduate unemployment, and have focused on the determinants of job search success. Existing empirical studies in China, such as Zhou (2003), Yue et al. (2004) and Yan et al. (2005), report that graduates' search effort has no significant influence on the search outcome. This conclusion challenges both established search theory based on Stigler (1961Stigler ( , 1962, and also the results of numerous empirical studies from other countries (Saks and Ashforth, 1999;Van Hooft et al., 2004;Saks, 2006). It is this conflict of opinions among researchers that motivates our own research, in which we focus on the question: is job search effort beneficial in today's graduate labor market in China?

Data from the China Education Ministry convey an idea of the extent of the problems facing graduates searching for employment. In 2001, the number of graduating students was approximately 1 million, and, as a result of higher education expansion policy, this increased more than four-fold to 4.36 million in 2006. As seen in Table 1, the graduate employment rate has remained at around 70 percent over this period. Two important qualifications relate to the information in Table 1. First, the significant fall in the employment rate in 2003 was a direct result of the outbreak of Severe Acute Respiratory Syndrome (SARS), during which time face-to-face contact was strictly forbidden. Second, in 2003, as a consequence of SARS, the time of data collection was moved from July (straight after graduation) to September (3 months after graduation): hence the two figures for the employment rate for 2003. In the years since 2003, the data collection month has continued to be September. This change has had the effect of inflating the observed employment rate, because the job search period was effectively extended by 3 months. This means that the apparent stability in the employment rate between 2001 and 2005 is deceptive.

The statistics presented in Table 1 are only from state universities, of which there were 1792 in 2005. These statistics have been used to obtain a forecast of 25 million graduate job seekers during the period 2007-2010. 1 Some universities have allegedly been reporting exaggerated employment figures, because this has the potential to boosts their ratings. Employment rates are used to decide on the number of university admissions. The result is that the employment figures in Table 1 might be upward biased. Moreover, there are more than 1000 private higher education institutions in China, from which graduates find it even more difficult to gain employment.

1. Wang and Jiang (2005). Meanwhile, only a very small proportion (5.2 percent) of the total population has received tertiary education. 2 The higher education enrolment rate has risen steeply since 1999, reaching 19 percent in 2005. However, this is still low in comparison to the rates of around 50 percent seen in many developed countries, such as Australia, (Graduate Careers Council of Australia, 2006) the UK and Germany. 3 There is no evidence of large scale unemployment among graduates of these countries.

One important aspect of the graduate unemployment problem in China is the labor market inefficiency caused by Hukou policy. This policy was generated within the centrally planned economic system, and has been in operation since 1958. People inherit Hukou status from the previous generation. An important distinction is made between urban Hukou and agricultural Hukou (for details see Chan and Zhang, 1999). The essence of the policy is as follows. An urban Hukou holder is entitled to social welfare, such as education, medical care, pension rights and unemployment benefits, 4 and also has access to desirable positions in the labor market. For an agricultural Hukou holder, a plot of land loaned from the government is the only source of social welfare; when seeking employment in urban areas, they are excluded from urban social welfare and are constrained to inferior or temporary positions in the labor market. Also, they have no entitlements to unemployment benefits. Hukou is not transferable even between cities. For example, a Shanghai citizen is excluded from the Beijing municipal social welfare system even after working or living in Beijing for many years. 5 This institutional impediment to job seekers is expected to continue for at least another 15-20 years.

The disadvantaged position of rural Hukou holders is the central focus of this paper. The approach is analogous to that of Holzer (1987Holzer ( , 1988, who contrasts the effectiveness of job search between black and white youths in the USA, finding, for example, that informal search methods, such as through friends or relatives, and the direct application search method are more productive for whites than for blacks, in terms of the likelihood of gaining employment. However, our study is different in at least two ways. First, the concept of Hukou and its effect on job search are peculiar to China. Second, a key concept facilitating the route into employment in China, but less relevant in developed countries, is that of 'institutional embeddedness'. This is a concept borrowed from Brinton and Kayia (1998), representing the university-related activities that assist graduates in the job search process, such as: job fairs organized by the university; making CV templates available on the university website; university departments recommending graduates to potential employers; introduction by university faculties to external contacts who are potential employers; and alumni vouching for graduates from their institution in the hiring process. The university system in China has a long history as a job assignment agency (see Section II.2), with an important influence on graduates' job search outcomes. For these reasons, we consider it worthwhile to treat institutional embeddedness as a determinant of the search outcome in its own right.

The paper is organized as follows. Section II provides relevant institutional details; Section III presents a literature review; Section IV outlines a theoretical job search model; Section V describes the data collection; and Section VI presents econometric results. Finally, Section VII concludes.

II. Institutional Detail

II.1 The transition of university students' Hukou status

In the present paper, a graduate whose family is registered in an urban area is labeled as an urban Hukou holder. In contrast, a graduate who reports that his family registry place is in a rural area is labeled as a non-urban Hukou holder. This is because his current Hukou status is neither agricultural nor permanent urban.

One way to change Hukou status is to invest in higher education. Once students have been admitted to a state university, they temporarily acquire urban Hukou status, equivalent to citizenship in the location of the university. After graduation, which normally takes place by the end of June each year, there are three possible outcomes regarding a student's Hukou status; these outcomes are depicted in Figure 1. The first is where the graduate finds a position with a company that has an urban Hukou quota (such as a Beijing Hukou ). The graduate's Hukou can then be transferred to the company, and he or she gains urban Hukou status from this time on. The graduate and their offspring (but typically not their spouse) will then be entitled to urban welfare. 6 The second possible outcome is where the graduate manages to find a formal job in a company without an urban Hukou quota. The graduate's Hukou will be kept at the local labor market office provided an administrative fee is paid; the graduate is not given the right to urban welfare. This is called mobile Hukou : as long as the jobhopper keeps formal jobs, the mobile Hukou can be transferred between labor market offices; if the job-hopper is fortunate and receives an offer with urban Hukou , the mobile Hukou turns into urban Hukou . If the job-hopper is less fortunate and becomes unemployed, his or her Hukou status becomes pending. Although the final outcome in this case depends on individual circumstances, 6. If the graduate is considered highly desirable, the company might offer one of their quota positions to the spouse, endowing the spouse also with urban Hukou status.

there is a possibility that Hukou is returned to the original residential place. The third possibility arises if the student takes a temporary job. In this situation, the Hukou will be kept at the University for 1 year, again subject to a fee. Thereafter, if the graduate still does not have a formal job, the worst outcome comes into force: his or her Hukou will be sent back to his or her original residential place. This institutional feature is highly relevant to the objectives of the present study. Clearly, non-urban Hukou holders have just 1 year to secure formal employment. If they fail, the penalty is an almost irredeemable relegation to their previous Hukou status. This explains why the pressure on non-urban Hukou holders in the search for employment is so great.

II.2 History of graduate job search in China

Job search is a relatively new activity for graduates in China. This is because since 1949, people were guaranteed a job from the day they received an offer from a state university. To conform to the centrally planned economic structure, the Chinese higher education system was highly centralized; the number of admissions to universities was set in advance by the state. Only the small number of students who achieved good marks in the National Higher Education Institution Entrance Examination would gain access to higher education. These fortunate students would be provided with free education and living expenses during their period of study, normally 4 years. Universities, following guidelines from the National Planning Commission, performed the role of job assignment agencies, individually assigning each graduate to a job with a particular company. This naturally led to a situation in which each university had a close relationship with certain companies.

In 1997, the Chinese Education Ministry initiated a process of reform: a graduate employment system was to be established in which universities would have no obligations in job assignment, graduates compete for positions, and employers may recruit graduates from any university, basing recruitment decisions on an individual's ability. Hence, although job search was practiced in a minor way before 1997, it is only in the past decade that job search has become the dominant route into employment. However, the close relationships formed between universities and employers under the previous regime have persisted, and still play an important role in the job-matching process.

Typically, a student's 'search period' is the whole of their final year of study. Officially, the student leaves university at the end of June. Beggs and Hurlbert (1997) and Granovetter (1973) suggested that job search methods be divided into three groups: informal, formal and direct application. Informal methods rely on interpersonal networking, or 'social contacts', such as introduction to employers by parents, relatives or friends. Such means of securing employment have been widely explored by both economists and sociologists (Marsden and Hurlbert, 1988;Boxman et al., 1991;Montgomery, 1992). Formal job search methods depend on non-personal labor market intermediaries, such as job postings, newspaper advertisements and employment agencies (Bridges and Villemes, 1986). Methods of direct application include the sending of a CV to potential employers, or visiting the office of an employer to enquire about job openings.

III. Literature Review

The particular problems faced by graduates seeking employment have been identified in the context of other countries. Saks and Ashforth (1999) conducted a survey of 384 graduating students in one major city, in Canada, to find that active job search behavior and higher job search intensity both have a positive effect on the search outcome. Saks (2006) analyzes data on 225 recent graduates in Canada, and reports evidence that active job search intensity impacts positively on the number of job interviews and offers. Try (2005) used data from the Norwegian Graduate Surveys of 1995-2000 to investigate the use of different job search strategies, to find that 70 percent of graduates respond to employers' advertisements directly, and that the use of informal search methods is affected by the possession of social capital. Another finding from this study is that graduates with poor job prospects (either in terms of poor initial quality or lower probability of employability) are more likely to seek employment through the 'Public Employment Service' (job center).

There are few empirical studies on graduate job search using data from China. This is because job search has only become a widespread phenomenon in China in the past decade. Furthermore, there are well-known problems with obtaining data in China. However, a small number of empirical studies have recently been carried out by researchers at Beijing University, using a nationwide survey of graduate employment status conducted in June 2003. A stratified sampling procedure was adopted, ex ante . The stratification was firstly by region: three provinces were selected from each region: east, middle and west. 7 Then, within each selected province, one or two institutions were selected from each of the three types: key university, regular university and college. Then, from each of the selected institutions, a sample of 600 graduates would be taken in a way that ensured appropriate coverage of level (Bachelors, Masters or Research degree) and subjects studied. The planned stratification was followed with only partial success. The western region was significantly overrepresented, with 22 of the 45 participating institutions being from this region, of which 17 were from a single province: Yunnan. The middle region was underrepresented, with only 6 participating institutions, all from the same province: Hunan. There were 18 722 valid responses in the survey, implying a response rate of approximately 60 percent.

One study using this data is that by Yue et al. (2004), who use logistic regression to investigate the determinants of the job search outcome, with 'success' defined as: receiving a job offer; waiting for a job offer; choosing self-employment; undertaking further education; or remaining unemployed for a short time. We consider this definition of success to be too wide, partly because the last three groups in the list include people who have not engaged in job search. Their results show that both academic performance and job information provided by the university have beneficial effects on search outcome; but search intensity, search skill training courses, and graduates' original residential place have no noticeable effect. A second study using the same data is that by Li (2004), who investigates the interaction of job search channels (formal versus informal) and educational level. He uses a binary probit, with search outcome as the dependent variable. However, his definition of informal job search is questionable: it includes 'recommended by University', which we consider to be a formal channel. Zhou (2003) defines an ordinal (0-9) variable to represent the self-reported amount of help the graduate receives from their university, and uses this as an explanatory variable in logistic regression analysis of the search outcome, although the definition of the dependent variable is not clear. It is found that university help does have a positive effect on the probability of employment, whereas pecuniary investment on search by individual students has little benefit. 7. Mainland China is divided into three regions for the purposes of the nation's economic plan: the east, the middle and the west. For other purposes, it is divided into 22 provinces, 5 autonomous regions and 4 municipalities.

The main concern of the survey is that it was conducted in June 2003. 8 As already mentioned in Section I, the outbreak of SARS meant that face-to-face contact was forbidden between April and July of that year. Over the same period, all kinds of recruitment activity on campus, for example job fairs, were cancelled. 9 Students were physically confined to the campus, and so were unable to engage in direct search activities. For these reasons, results from studies of graduates' job search based on this survey are likely to be biased.

IV. Theoretical Model of Job Search

Standard search models (McCall, 1970;Mortensen, 1970;Kiefer and Neumann, 1979) assume that in any period of time, a job-seeker might, according to some probability distribution, receive one or more job offers, each with an offered wage. The job-seeker accepts an offer if the offered wage is at least as high as his or her reservation wage; otherwise, the offer is rejected. The search outcome, that is to say, the receipt and subsequent acceptance of an offer, is positively related to the number of search methods available and to the level of search effort, and negatively to the job-seeker's reservation wage. The reservation wage is determined by the level of unemployment benefits, the state of the labor market, search costs, the job-seeker's rate of time preference, and the job-seeker's initial wealth.

Applying these ideas to the situation of graduates seeking employment in China, we first see, in Figure 2, that initial wealth has the potential to vary considerably according to Hukou status. This graph, constructed from data provided by Li (2006), shows that per capita income is 2-3 times higher in urban households than in rural households, and, moreover, has been rising much more rapidly in the former. In the same graph we see that university tuition fees have been rising steeply, and, since 1999, have exceeded per capita income in rural households. This means that it would be very hard for a typical rural household to finance a university education without incurring significant debt. In contrast, in urban households, because it is typical for both parents to work and for there to be only one child, 10 with this child likely to be receiving financial support from grandparents on both sides, the financial constraints of educating the child are considerably less severe. This is one difference that leads to the prediction that non-urban Hukou holders are likely to have a lower reservation wage.

Another difference leading to the same prediction is in entitlement to unemployment benefits. As explained in Section II, this is an automatic entitlement 8. The survey was repeated in 2005 by Yan et al. (2005), who performed the analysis using the same definition of search success as Yue et al. (2004). However, Yan et al. (2005) do not distinguish between graduates with different Hukou status; instead they simply divide graduates into two groups: those from large cities, and others. Hence, it is hard to draw conclusions from Yan et al. (2005) with respect to the effect of Hukou status. 9. Mo and Liu (2003). 10. The one-child policy commenced in 1978, but applies strictly in urban areas. to urban Hukou holders, but denied to those with rural Hukou. Some cities recently implemented a system of providing subsistence expenses to all graduates seeking employment. However, because the subsidy is distributed by local urban community administration centers, administrative barriers prevent rural graduates from claiming it.

In terms of job search methods available, graduates from rural areas are believed to have less search opportunities than those from urban areas because the former have fewer work-related social ties. They are also less likely to benefit from information and referrals from their close relatives, who are likely to be working in the fields or in inferior jobs in cities.

It might be predicted that graduates with non-urban Hukou put greater effort in their job search than their urban counterparts, in an effort to secure their attained Hukou status and to compensate for their disadvantaged position in terms of available search means.

With the preceding discussion in mind, we advance the following three hypotheses:

Hypothesis 1: Graduates with non-urban Hukou conduct more intensive job search than those with urban Hukou, ceteris paribus. Hypothesis 2: Graduates with non-urban Hukou have a higher probability of gaining an offer than those with urban Hukou, ceteris paribus. Hypothesis 3: Graduates with non-urban Hukou have a greater propensity to accept a given offer than those with urban Hukou, ceteris paribus.

Note that these three hypotheses are interrelated. In Section II.1, we explained the institutional reasons why there is more pressure on non-urban Hukou job-seekers to gain employment in the period leading up to graduation, than on

Figure 2 Higher education tuition fee and per capita income in China

Source: Data is sourced from Li (2006) urban Hukou holders in the same position. This pressure derives from the fact that confirmation of non-agricultural Hukou status is conditional on gaining formal employment within 1 year of graduation. In Section IV, we describe a theoretical model of job search, and advance several reasons why these institutional factors cause non-urban Hukou holders to have a lower reservation wage. Hypothesis 1 is easily derived from the first of these differences: a greater incentive to secure formal employment must lead to more intensive job search. Hypothesis 2 derives from both differences: employers know from experience that non-urban Hukou holders are, on average, more dedicated, purely because their will to secure and retain employment is so great; employers also know that, on average, non-urban Hukou holders have a lower reservation wage, and can, therefore, be hired at lower cost. Hypothesis 3 also derives from both differences: when offered a position at a given starting salary, the non-urban Hukou holder is more likely to accept it: first, because they are generally more eager to secure employment, and second, because their reservation wage is lower than that of their urban counterparts. Each of the three hypotheses is tested econometrically in Section VI.

V. Data Collection

The data to be used is from our own survey, entitled Graduates' Job Search and Outcomes, conducted in June 2005 by the careers centers of 3 universities: China University of Geosciences (CUG), Huazhong Agricultural University and Zhongnan Finance and Law University. All 3 universities are situated in Wuhan, the capital city of Hubei province, where there are 85 higher education institutions, turning out more than 143 000 graduates in 2005.

In March 2005, a research proposal was presented to these careers centers. The officials at the careers centers showed an interest in the survey and were pleased to cooperate, partly because they themselves face great pressure to improve their own graduate employment rates. They assisted by releasing 1500 questionnaires and collecting 512 responses in total.

The survey questionnaire elicited detailed information on the respondent's characteristics and background (e.g. gender, Hukou status and household income), their academic performance, and their recent search activities and outcomes, or expected search outcomes. An important part of the survey is on employment expectations, including questions about anticipated annual remuneration, desired type of employment, and desired location of employment. Equally important is the job-search section, which asks about labor market experience, search methods used, vacancy information channels used, intensity of search, numbers of interviews attended, and offer acceptance criteria. Respondents who indicate that they have had an offer are asked about the number of offers received, and whether an offer has been accepted. Those who claim to have accepted an offer are then asked about the type and size of their employing company, their position within the company, and their starting salary.

Qualitative information is extracted on reasons for rejecting offers, reasons for setting up one's own business, and the respondent's assessment of employment services provided by the careers centers. Graduates who have decided to pursue further study or who are not seeking employment are excluded from the survey.

As Figure 3 illustrates, the majority of graduates in the sample (60 percent) are from rural areas. The proportion of the agricultural population in China has decreased from 82 percent in 1978 to 57 percent in 2005. 11 Therefore, it appears that our sample is representative of the population. However, the distribution of university students by Hukou background does not always appear representative. Tong (2006) reports that students with non-urban Hukou only accounted for 17.7 percent of the prestigious university student population in 1999 in Beijing. 12 In complete contrast, it is reported by an official website, Xinhuanet (www.xinhuanet.com), that the vast majority (approximately 80 percent) of students in local higher education institutions in Qinghai (an impoverished western province) in 2005 had non-urban Hukou status. This suggests that there are significant regional imbalances in the proportion of rural students receiving higher education.

Each of the 3 universities in our survey enjoys a national reputation and all have students from all over the country, hence the balanced distribution of students in the sample across Hukou status. We consider this to be a virtue of our dataset, given that the purpose of our study is to analyze differences of job-search behavior between graduates with non-urban and urban Hukou. Two of the three surveyed universities, CUG and Huazhong, were originally founded as single discipline universities: respectively, agriculture and geology. However, since 1999 as a result of the decline in importance of the centrally-planned curriculum, and also in response to popular demand, teaching at these institutions has become much more diverse, including, for example, degree studies in economics, law and computer sciences, fields which are booming at both universities.

VI. Data Analysis

VI.1 Exploratory data analysis

In line with the objective of the paper, Table 2 presents summary statistics on search choices, and other variables, for graduates with non-urban and urban Hukou separately. The results of independent sample t-tests for differences between these two groups are also shown in each row. The key conclusions from this table are that the non-urban Hukou job-seekers have a significantly higher chance of gaining employment (as represented by receipt of at least one offer; p = 0.003), and also appear more likely to have accepted an offer (p = 0.003). These results could be attributed to greater job-search effort among non-urban Hukou graduates: we also see in Table 2 that non-urban respondents use a wider variety of search methods and contact more employers, than do their urban counterparts, although neither of these tests is statistically significant.

Another significant difference between the two groups is in standard of living. Average household income is almost twice as high for urban respondents as for non-urban respondents (p = 0.000), and this confirms the wealth disparity between rural and urban areas discussed in Section II.

The difference in academic performance between the two groups, as measured by English proficiency, success in degree, average grade, and major, is not statistically significant. It appears, therefore, that the difference in search success between these two groups cannot be explained by any differences in academic performance. This is in disagreement with Yue et al. (2004) and Yan et al. (2005), who both provide evidence that graduates' academic performance greatly improves employment prospects.

When looking for differences in choice of job search method between the two groups, we focus on the method reported by the respondent as being the most important, in terms of potential to result in an offer. Here, the one significant difference is in institutional embeddedness. 13 Non-urban respondents appear to attach significantly more importance to this type of search method (particularly recommendation by lecturers, and attendance at job fairs organized by universities) than do their urban counterparts (p = 0.07). In contrast, urban respondents appear to attach more importance to social contacts (p = 0.11). These patterns are consistent with the idea that social contacts are a privilege that only urban Hukou holders are in a position to enjoy, and non-urban Hukou holders therefore need to rely on any such assistance as is provided by their institution.

Meanwhile, there is a telling difference in the starting salary between the two groups. Graduates with urban Hukou, on average, take higher-paid positions; this is in agreement with the findings of Yan et al. (2005). However, we also notice that there are no obvious differences in the expected annual income between these 13. Institutional embeddedness was defined in Section I. In the survey, respondents were not asked directly about the importance of institutional embeddedness. The variable used here is a combination of variables obtained from the survey responses. For details, see the notes below Table 2. Notes: *, ** and *** represent 10%, 5% and 1% levels of significance, respectively. †In most Chinese universities, to procure a degree, students have to meet two criteria. One is no failure in course work; the other is to pass the College English Test Band 4 (CET4), the latter is a national exam and taken twice a year. ‡Expected annual income in 1000 RMB is obtained from this survey question: 'what annual income do you expect to receive in your first job?'. Respondents are invited to select an income range as their reply. §The actual annual income in RMB1000 is collected by asking respondents who declared to accept an offer to choose an income range, there are 305 observations. These variables are generated from responses by indicating one of the above listed search methods as the most important, at least producing a job offer. ¶The variable used to represent 'institutional embeddedness' is one that indicates that the most important search method is one of the following: job fair organized by university; recommendation by faculty members; use of CV template on career center website; and networking while studying.

© 2008 The Authors Journal compilation © 2008 East Asian Economic Association and Blackwell Publishing Ltd.

Table 2 Comparison of job search behavior between graduates with non-urban and urban Hukou (independent samples t-test for each variable individually)

Variables

Mean t Significance (two-tailed) Graduates with non-urban Hukou two groups. This suggests that graduates with non-urban Hukou, despite being as optimistic about their future as their urban counterparts, are willing to accept a lower starting salary when the offer actually comes (they have a lower reservation wage); we conjecture that this is because of the pressure resulting from their Hukou status.

Graduates with urban Hukou

VI.2 Econometric analysis

In Section IV, we described a theoretical model of job search in which greater intensity of job search results in more offers being received. An obvious way to measure search intensity is using the number of employers contacted within a certain period (Blau and Robins, 1990). Analysis of such a variable is made easy in the case of Chinese graduates due to the fact that they are engaging in search subject to the unofficial but generally accepted deadline of 30 June each year, so the period of search is effectively the same for all individuals in the population of interest. Because the number of companies contacted within this period is a count variable, taking only non-negative integer values, we apply the Poisson regression model.

The results are shown in Table 3. The central result is that individuals with non-urban Hukou conduct job-search significantly more intensively, ceteris paribus, than their urban counterparts, reflecting the greater pressure that they are under to secure a position. This provides evidence in favor of our Hypothesis 1.

The significant coefficient of the variable 'succeed in degree' indicates that the prospect of academic success is associated with more intensive job search. 14 This is expected: students who fail their degree are required to retake the failed component in the following year, clearly a hindrance to the employment process. Expected annual income has the expected positive effect: the higher the status to which the individual aspires, the more effort they are prepared to invest to attain it.

Social science students appear to search harder than students in other disciplines. This is consistent with recent reports that social science graduates have difficulties finding employment. 15 In particular, those majoring in law, administration and economics face the problem that they are part of a significant excess supply, because almost every university has set up these degrees as part of the higher education expansion initiated in 1999.

Moving to the final section of Table 3, 16 it is evident that individuals who wish to work in smaller cities in the middle or the west region engage in less intensive job search than those who wish to work in large cities, or in coastal 14. See Table 2 for the definition of this variable. 15. See Guomin and Yuejin (2006), unpublished, 'The employment difficulty faced by social science students', on Dean's Forum, at http://www.chinalawinfo.com/Fxyluntan/gq1.html. 16. This categorization is based on patterns observed in surveys (such as 'Graduates' ideal employer survey ' in 2001 and 2004) conducted by the magazine Chinese Student Career, which is published by China Education Ministry, and in the study of Lai (2001). areas. This is consistent with the idea that it is harder to get a job in coastal areas, or in large cities, because of higher competition in these locations. This high competition is partly a result of Hukou status in large cities in the coastal areas being more valuable, in terms of the various benefits for the holder and for the holder's offspring. Hence, those who are prepared to work in central or western areas do not need to search so intensively.

Next we turn to an analysis of the number of offers received. This also is a count variable, with an added complication. The variable is the response to the question 'how many offers have you received so far?' and the possible responses are: none; 1 offer; 2 offers; and 3 or more offers. This means that the count variable is upper-censored at 3, and to identify its determinants, we require the upper censored Poisson regression model, which will now be described.

Let be the true count for respondent i, and let Y i be the censored counterpart. The censoring of the variable is such that: (1)

The probability of each possible occurrence is given by:

where is the Poisson mean, x i is a vector of explanatory variables relevant to the number of offers received, and β is a corresponding vector of parameters to be estimated.

Let y i be the sample realization of the random variable Y i . Furthermore, we let d i be a censoring indicator, defined such that:

(3)

The sample log-likelihood can then be written: (4) The log-likelihood function presented in Equation (4) has been programed using the maximum likelihood routine in STATA to estimate the model.

The results are presented in Table 4. First of all, note that we are including the number of employers contacted as an explanatory variable, because this is obviously a key determinant of the number of offers received. As expected, it has a significantly positive effect, confirming that job search is richly rewarded. 17 We further see that, controlling for the number of employers contacted, a non-urban Hukou holder is expected to receive more offers than an urban Hukou holder with otherwise the same characteristics. This provides evidence in favor of Hypothesis 2: employers favor non-urban Hukou holders because of their dedication and their willingness to work for lower remuneration. Another interpretation of this effect is that urban Hukou holders have less success as a result of the ineffectiveness of social networks, on which they tend to rely (see t-test for 'rely on other relationships' in Table 2). The majority of urban Hukou holders in the sample are from medium or small cities (see Figure 3), but they inevitably prefer to work in large cities. Social networking is only likely to be effective in the job-searcher's home-city.

Employers' apparent favoring of non-urban Hukou holders could also be interpreted in terms of sample selection. A well-known inequity in the Chinese higher education system is that entrance requirements are higher for non-urban 17. However, Yue et al. (2004) and Yan et al. (2005) disagree with us on this point.

Hukou holders ( Wang, 2004). It follows that it is reasonable to expect non-urban graduates to be academically stronger, on average, than urban graduates. Employers may thus be using a form of 'statistical discrimination' (Elliot, 1991) in favoring the former. The effect of institutional embeddedness (defined in Section I) is statistically significant. Its positive coefficient shows that it is a more effective job search method than any of the others. On this point, we agree with Zhou (2003), Yue et al. (2004) and Yan et al. (2005).

Next, we turn to the decision by the job searcher on whether or not to accept an offer. Because this is a dichotomous choice, the probit model is used. For Note: *, ** and *** represent 10%, 5% and 1% levels of significance, respectively. The method used to estimate this model is described in Section VI.2. †After the second year at university, students can apply for a second degree course. Whether the application is approved is based on their grade achievement during the 2 years of their first major. Once approved, students are required to take 6-7 extra courses for the second degree, in addition to those for the first degree. These extra courses for the second degree are mainly attended at weekends, in order to avoid timetable clashes. The most capable students tend to have two degrees after 4 years of study, and they use this as a positive signaling device in the labor market. For some students it takes an additional year (i.e. a fifth year) to complete the second degree.

this purpose, we only use data on the 330 respondents who received at least one offer. The model can be represented as follows:

where P Ai represents the probability of offer acceptance, conditional upon receiving an offer, by individual i. SM denotes the search method they consider to be most useful; is the individual's reservation wage. Hukou i is, as in previous models, a dummy variable taking the value 1 if individual i is non-urban, zero if urban. X i is a vector of individual characteristics, such as gender, grade achievement, household income and English proficiency. Y i is a vector of variables representing behavior affecting the probability of gaining an offer, such as search intensity and labor market experience.

The results are presented in Table 5. Most importantly, there is evidence that graduates with non-urban Hukou have a higher probability of accepting an offer, providing support for our Hypothesis 2.

In addition to the number of offers received, another outcome of job search that we consider is the level of starting salary. Here, we only consider the 305 Note: *, ** and *** represent 10%, 5% and 1% levels of significance, respectively.

, = w i r respondents who accepted offers. The starting salary was recorded as interval data; that is, all that is known about a respondent's starting salary is that it is in a particular range of values. Accordingly, we use the interval regression model, and the results are presented in Table 6. Two specifications are adopted: the Note: *, ** and *** represent 10%, 5% and 1% levels of significance, respectively. †Starting salary is recorded in the following intervals: less than RMB15 000 per annum; RMB15 001-20 000; RMB20 001-30 000; RMB30,001-50 000; more than RMB50 000. ‡The excluded category for type of employer is 'other', such as joining the army.

second specification includes additional variables representing the type of employer. Surprisingly, type of employer is found to be insignificant, although the significance of company size does indicate that large companies tend to pay higher starting salaries.

Turning to the focus of the analysis, both specifications indicate that non-urban Hukou status significantly and negatively relates to starting salary. In particular, from specification 1, we see that the starting salary for a non-urban Hukou holder is RMB3355 lower each year than for an urban Hukou holder with otherwise the same characteristics. This result is consistent with Hypothesis 3. Non-urban Hukou holders appear willing to accept lower offers, and this is in turn consistent with the idea that they have a lower reservation wage. 18 Students who have passed College English Test Band 6 and with one or two degrees are usually offered higher starting salaries, confirming that academic achievement is highly rewarded in pecuniary terms in the labor market.

Graduates with higher expected annual income, unsurprisingly, receive higher starting salaries, suggesting that individuals are a good judge of their own potential.

VII. Conclusion

In Section IV, we advanced three hypotheses, each concerning the difference between graduates of non-urban and urban Hukou status. In Section VI, we reported econometric evidence in favor of all three hypotheses, using data from job-seekers in Wuhan. The first was that non-urban Hukou holders engage more intensively in job search, and this was confirmed using a count data model of the number of employers contacted. The result was interpreted in terms of the higher pressure placed on non-urban Hukou holders to secure employment; this pressure comes as a direct consequence of certain features of the institutional framework, discussed in Section II.1. The second hypothesis was that non-urban Hukou holders have a higher probability of receiving an offer, and we found evidence for this using a count data model of the number of offers received. This was interpreted in terms of various reasons why we believe that non-urban Hukou holders are more desirable to employers than urban Hukou holders. The third and final hypothesis stated that non-urban Hukou holders are more likely to accept an offer, and we found evidence of this using a binary data model of offer acceptance. The interpretation of this result is in terms of the lower reservation wage of non-urban Hukou holders, which is a consequence of a number of institutional factors connected with Hukou status; these were explained in Section IV. Further confirmation of the difference in reservation wages was found in a model of starting salaries, in which we found that non-urban Hukou holders accept significantly lower starting salaries, on average. The paper has established the result that effort put into job search is rewarded in China. This is important because job search is a relatively new phenomenon among graduates in China, and the small amount of previous research in this area has, for a variety of reasons, not provided a definitive answer on the effectiveness of job search. This paper is, to our knowledge, the first to show empirical evidence of the positive relationship between search effort and search outcome, although the Chinese labor market is still under development.

The paper has also demonstrated that university activities play an important role in assisting graduates in the transition from university into the labor market, suggesting that universities themselves have the potential to increase the graduate employment rate.

Table 2 (continued)