Floating Choices: A Generational Perspective on Intentions of Rural-Urban Migrants in China

Abstract

Using data from a 2005 Survey of rural-urban migrants in Shenzhen, this paper investigates intentions of two groups of migrants. We use the birth years from 1970 to 1980 as a reasonable range of dividing lines to separate the two groups. For each year we divide the sample into those born before that year and those born in or after that year. These are referred to as the old and the new generation, respectively. Three possible development trajectories are considered: settling in cities, returning home to seek a nonagricultural job, and returning home to farm. We find that members of the new generation have stronger desires to do non-farm work, and returning to seek a nonagricultural job has become the most important planned trajectory for this generation. Sharp differences exist between the two generations in the reasons that underlie their intentions. For the old generation, conditions such as age, family responsibility, and type of job are important determinants of intentions, while other conditions such as initial migration motives, social capital, and socioeconomic conditions of origin areas are important for the intentions of the new generation. Thus the new generation is more likely to view migration as a form of investment with the accumulation of human capital and social capital. Those migrants from the old generation who have higher education levels also intend to seek non-farm jobs. However, because of the combined effects of life cycle and the market transition in China, these intentions are not as strong as those of the new generation. We discuss economic and policy implications of our findings.

Introduction

Massive rural-urban migration has been occurring for more than 20 years in China. In recent studies on migration in China, a distinction between the old generation¹ and the new generation has been commonly used to capture the heterogeneity among migrants (e.g., Wang, 2001; Luo and Wang, 2003; Wang, 2008). The old generation is operationally defined as those migrants who were born before the early 1970s and the new generation is those born after the late 1970s (the dividing line between these two groups is still problematic, and will be discussed later). In the 1980s, hundreds of thousands of villagers from rural areas began to leave their home to search for new jobs in cities. They were/are the old generation of rural-urban migrants and were/are generally target earners and were motivated by the idea of earning cash income to subsidize their farming activities (Huang et al, 1996; Huang, 1997). This old generation retains a very strong attachment to rural areas and the land; most of them still think of themselves as farmers rather than workers. Therefore, most of them migrate into cities for limited periods of paid labor and eventually return home (Wang, 2001). After 1990, large numbers of migrants, who were born after the late 1970s and educated in the 1980s, began to migrate into cities (Wang, 2001; Luo and Wang, 2003). These younger migrants are the new generation. Migrants of the new generation are younger, more ambitious, more educated, more skilled and know little about farming. They do not like life in rural areas and are intent on securing jobs in cities. “Earning money” is not the only incentive to migrate for the new generation; the opportunities and life style in cities attract them strongly; “learning skills”, “broadening experience” and “preference for city life” have become primary reasons for their migration (Wang, 2001; Luo and Wang, 2003; Luo, 2007).

The main arguments for such a generational division are based on two theoretical perspectives. First, according to life-cycle theory, the younger and older generations differ in behaviors and attitudes even though they live within the same socio-economic context. Human life constitutes an age-related sequence of stages (preadulthood, early adulthood, middle adulthood and late adulthood) with different developmental tasks (Levinson et al, 1978; Levinson, 1986). Biologically, the 20s and 30s are the peak years of the life cycle; early adulthood (about age 17 to 45) is socially and psychologically the period for forming and pursuing youthful aspirations, establishing a niche in society, raising a family. In economics, Fuchs (1988) argues that 25–44 is a period of investment in human capital (“a time to sow”) while 45–64 is a period of consumption from human capital (“a time to reap”). Also, research on vocational behavior shows that there are variations in human capital investment activity by age; from the perspective of neoclassical human capital theory, older adults are less likely than younger adults to make human capital investments (Simpson, 2002). Second, the profound reform from the redistributive economy to a market-oriented economy has produced a deep social transition in China. According to Nee’s theory of market transition, unlike redistributive economies, markets provide powerful incentives for immediate producers (migrants are among these). The market transition results in new opportunity structures which open alternative paths of socioeconomic mobility (Nee, 1989). In a market-like economy, peasants, who are transformed into petty entrepreneurs, tend to pursue a strategy of maximizing profits and developing new, more profitable lines of activity (Nee, 1989; Huang, 1997). This socioeconomic transition can widen the gap between these two generations of migrants who are likely to have substantially different characteristics in their behaviors and attitudes. The combined effects of life cycle and market transition suggest that the new generation’s intentions should be different from the older generation.

However, the dividing line between the old generation and the new generation for rural-urban migrants in China is somewhat arbitrary. Such generational terms as the “first generation”, the “one-and-a-half generation”, and the “second generation” have been used in studies on international immigration or internal migration in the US: the “first generation” refers to those immigrants who arrived in the host areas after adulthood, immigrants from the one-and-a-half generation are those who arrived before they reached adulthood, and the second generation is those host-area-born children with at least one origin-area-born parent (e.g., Zhou, 1997; Ellis and Goodwin-White, 2006). In this sense, the size of the true second generation in China is extremely limited because only a few migrants have been able to settle down to a comfortable life in cities. Most of the current migrants in China, either young or old, are from the first generation, which has led to development of a new generational nomenclature (i.e., the old generation and the new generation). Previous studies have usually based their generational division of migrant workers on natural age or a specific birth year. Using data from a 2000 survey, Luo and Wang (2003) defined the migrant group aged 25 or below as the new generation. In Wang’s study, the post-80s (i.e., those who were born from 1980 on) are defined as the new generation (Wang, 2008). We argue that migrants who were born in or before the 1960s are of the old generation and those born in or after the 1980s belong to the new generation, while those born in the 1970s are a transitional generation. We will use the birth years from 1970 to 1980 as a reasonable range for separating the old from the new generation in our study.

Although attitudes and behaviors of the new generation have changed a lot, they are not much closer to integration into the core society of the city than their predecessors were. The Hukou System (household registration system) has perpetuated the inferior institutional and social status of rural-urban migrants, and still restricts migrants’ access to public health care, a pension system, legal aid, and social services (e.g., Ke and Li, 2001; Liu and Zhou, 2004). Migrants’ institutional and social inferiority undermines their chance of success in the host cities and reinforces their desire to return home. Finally, most of them can not settle down in cities, and only a few, who are very successful in business or in securing a career, remain in the cities. Urban-rural return migration has been a relatively prevalent phenomenon since 2000 in China, with many migrants returning to continue subsistence farming, and only a few managing to find nonagricultural jobs (Ma, 2001; Zhao, 2002; Bai and He, 2003; Wang and Fan, 2006). With regard to their intentions, both the old and new generations of migrants are confronted with three alternatives: first, settling in cities, “leaving the land as well as the village”, which can accelerate the modernization process of China; second, returning home to seek a nonagricultural job, “leaving land but not village”, to become either self-employed, or employed in the nonagricultural sector near home, which can be regarded as a grass-roots path to modernization; the last choice is returning to farm. From the perspectives of urbanization and modernization, both settling in cities and returning home to seek a nonagricultural job can relieve surplus labor pressure in rural areas, reduce the growing regional disparity between coastal/rural and interior/urban regions, and increase the pace of urbanization in China (Ma, 1999).

Despite sharp differences between the old and new generations, little is known about whether the determinants of intentions of the old generation differ from those of the new generation. Previous literature has focused either on the behavior of returning migrants without specifying that returnees have two choices, or has only paid attention to the intention of settling in cities. As discussed above, there are three trajectories of intentions for rural-urban migrants, not just two (i.e., either returning or remaining). Under the “three trajectories” framework, analysis of determinants of rural-urban migrants’ intentions provides a more comprehensive understanding of the impact of rural-urban migration on social and economic development in China. The major objectives of the present study are to examine the intentions of both old and new generations, and to explore the differences in determinants of these intentions between the two generations.

Theoretical and Empirical Background

Remain or return: theories and determinants

For both international migrants and internal migrants, return migration has been an integral part of the migration process (Gmelch, 1980; Zhao, 2002). The determinants of migration and return migration have attracted substantial attention in economics and sociology and several theoretical perspectives have been developed which relate to the present study. Neoclassical economics (NE), for example, views migration as a cost-benefit decision, with actors deciding to settle or return in order to maximize expected net lifetime earnings (Sjaastad, 1962; Todaro, 1976). Migration is conceptualized as a form of investment in human capital. People who choose to migrate to capture higher wages and a higher living standard, must make certain investments, “which include the material costs of traveling, the costs of maintenance while moving and looking for work, the effort involved in learning a new skill or language, the difficulty experienced in adapting to a new society and culture, and the psychological costs of cutting old ties and forging new ones” (Massey et al, 1993, page 434).

A second economic perspective on migration is the new economics of labor migration (NELM), which views migration as a response to market failure at the source community rather than as an adjustment to disequilibrium in labor markets (Stark, 1991). According to this theory, “people seek to migrate temporarily for limited periods of paid labor, either to remit earnings or accumulate savings in anticipation of an eventual return home” (Constant and Massey, 2002, page 10). Migrants are generally considered as target earners, and once their earning targets have been met, they return (Piore, 1979).

In sociology, network theory argues that migrant networks, “the sets of interpersonal ties that connect migrants, former migrants, and non-migrants in origin and host areas through ties of kinship, friendship, and shared community origin”, constitute a form of social capital the migrants can rely on to gain access to employment (Massey et al, 1993, page 448). Migrant networks can also facilitate the adjustment and settlement of newcomers, reduce the costs and risks of migration, and raise the probability of new migration (Massey, 1990; Massey et al, 1993). On the other hand, a deficiency of social capital can cause return migration (Orrenius, 1999).

However, neither economic theories nor network theory can perfectly interpret the causes of migration and return migration, and many researchers argue that decisions about migration and return migration are made based on a combination of both “pull” and “push” factors (Gmelch, 1980; Hare, 1999; Constant and Massey, 2002; Zhao, 2002). Drawing upon previous theoretical and empirical research, we divide the determinants that might affect decisions to remain or return into four categories: first, individual factors, including gender, age, marital status, human capital, social capital, and migration motives; second, familial factors such as spousal separation and parent’s health; third, migrant’s working and living conditions in the host city; fourth are the social and economic conditions in the origin and host societies. Among individual factors, human capital such as education, language proficiency, and work experience have always been emphasized (Borjas, 1989; Newbold, 2001; Wang and Fan, 2006). How migrant’s human capital is rewarded at both origin and destination can affect migrants’ decisions to return or to remain. Skills and education acquired at home, for example, are usually difficult to transfer and thus are rewarded more at home than at the destination, which predicts that migrants with higher levels of human capital tend to return (Constant and Massey, 2002). Since social capital can facilitate the adjustment and settlement of newcomers, its deficiency predicts that return migration is more likely to occur (Orrenius, 1999). Constant and Massey (2002) argue that the motivations of immigrants can also influence the decision to remain or return.

Among familial factors, frequently mentioned reasons for return migration are strong family ties and a desire to be in the company of one’s own kin and longtime friends (Gmelch, 1980; Constant and Massey, 2002). From the NE perspective, attachments to family members in the place of origin lower the costs of returning home, both psychological and monetary, and raise the costs of remaining in the host society. It has been found that some migrants, particularly eldest children, are obligated by their ailing or elderly parents to return (Gmelch, 1980). Spousal separation is also a major reason for migrants to return (Constant and Massey, 2002; Zhao, 2002).

Migrants’ living and working conditions also affect decision-making. One of the most important predictors is income (Gmelch, 1980; Constant and Massey, 2002). A migrant without enough income to afford to live in a city is unlikely to decide to remain. In addition, work is not only a matter of money, but also of status. While NE generally considers occupational prestige to be a non-monetary benefit in the cost-benefit calculus, NELM views prestige as irrelevant (Constant and Massey, 2002). Temporary migrants are only there for the money and do not care if they have low social status; what is important is the status at home that foreign earnings can buy (Piore, 1979).

Some researchers point to unfavorable economic conditions in the host society, such as recession or layoffs and unemployment within a single industry, as causes of return migration (Gmelch, 1980). The decision to return or settle is also influenced by ethnic prejudice and discrimination in the host society (Gmelch, 1980; Constant and Massey, 2002). The social and economic conditions at the community of origin also play an important role in decision-making. International migrants, who come from the poorer countries where the home economy cannot provide returnees with adequate employment and a comfortable standard of living, seldom return (Gmelch, 1980).

Remain or return: internal migration in China

Although no study has directly explored the determinants of intentions among the two generations of Chinese rural-urban migrants under the “three trajectories” framework, two strands of research have shed some light on this issue. Many studies have examined the determinants of return migration and the determinants behind settling in cities. Zhao (2002) finds returnees tend to be older, married, better educated, and with a spouse who has not migrated. Based on a survey in Henan Province, Hare (1999) finds that the length of the observed migration spell is greatly influenced by household labor ratio and land endowments. Wang et al. (2006) argue that the institutional context of the transitional economy in China complicates our understanding of return migration. Migrants’ institutional and social inferiority in cities undermines their likelihood to succeed at the destination and reinforces their return migration, especially when family needs arise. They also find that age is an important predictor of return migration; migrants are more likely to be returnees as they get older. Whether rural-urban migrants want to settle in cities is another perspective for research on return migration. Zeng and Qin (2003) find that both educational attainment and spousal separation play an important role in the decision to settle down in cities. According to Xiong and Shi (2007), occupation, income level, and housing conditions have significant effects on the decision to settle in cities, while gender, age and education have no significant influence.

Rural-urban migrants are not a random sample of the rural population; they tend to already have more human capital, i.e., they are on average younger, more skilled, and more educated than the non-migrants; this has been referred to as “migrants are positively selected” (e.g., Wang and Fan, 2006). Although researchers believe that return migration is also selective, there is no consensus on how urban-rural returnees are selected, especially in terms of education (e.g., Bai and He, 2003; Zhao, 2002). Inconsistent findings in earlier literature lead to different evaluations of the impacts on economic development of rural areas. Some researchers argue that, compared with the settlers, returnees are somewhat negatively selected, and most of them in fact return home to resume farming with a tiny number able to start their own businesses (Bai and He, 2003; Liang and Wu, 2003). Accordingly, they have limited influence on local economic development. In contrast, Zhao (2002) argues that they are better educated and invest significantly more in productive farm assets. Return migration reverses the brain-drain process, and the businesses and enterprises set up by returnees provide more employment opportunities for the home society, which can diversify the local economy in rural areas (Ma, 1999).

The generational perspective on the study of intentions has received scant attention in the literature, although it has been found that age of rural-urban migrants has a significant influence on the decision to remain or return. The “three trajectories” framework used here may also shed some light on this issue.

Data and Methods

Data for this study come from the Shenzhen Survey of rural-urban migrants conducted by Institute for Population and Development Studies at Xi’an Jiaotong University in April 2005. Shenzhen is in southern China’s Guangdong province, and is situated immediately north of Hong Kong. After China’s opening policy and economic reform (since 1980), this area rapidly became China’s first — and ultimately most successful — Special Economic Zone. Shenzhen’s population was 10,350,000 in 2005, of whom only 16.5% have Shenzhen Hukou (Yang, 2005). Shenzhen has the highest proportion of immigrants (including other kinds of migrants besides rural-urban migrants) among Chinese cities, which makes it an excellent example of rural-urban migrants’ destination cities.

Our survey subjects were rural-urban migrants aged above 15 without Shenzhen Hukou. Since there are two types of rural migrants according to their housing arrangement, concentrated-housing migrants and scattered-housing migrants, two different sampling methods were employed. Among the former who live in dormitories or work sheds provided by factories, about 550 respondents were interviewed by cluster sampling in three electronic companies and two construction sites situated in Baoan, Nanshan, and Longgang districts. Among the latter who live in a community of urbanites or in a mixed community of both urbanites and rural-urban migrants, and whose places of residence are mostly rented and rarely owned by themselves, about 1,200 respondents were interviewed by simple random sampling in five streets of Nanshan, Luohu, and Yantian districts. In the survey, individual information, migration history, family conditions, and the working and living conditions in the city were collected. Additionally, based on the categories made by Van del Poel (1993), information on social support networks including instrumental support network, emotional support network, and social contact network was also collected. In total there were 1,739 eligible questionnaires left after excluding the ineligible ones.

Dependant variable

In the survey, each respondent was asked “where are you willing to develop your career or settle down permanently”. According to respondents’ answers, our sample can be divided into three groups: agricultural returnees, those migrants who plan to return to farm; nonagricultural returnees, who intend to return to seek a nonagricultural job; and settlers, who intend to settle in cities with a non-farm career plan.

Independent variables

With the four categories of determinants discussed above, we group the possible determinants (except the generation variable) into five categories. Table 1 presents the independent variables and their definitions.

Table 1.

Independent variables and their definitions

Independent variables	Definition
Generation
The old generation	Reference (Respondent born before the dividing year)
The new generation	Respondent born in or after the dividing year
Individual factors
Age	Respondent’s age, in years
Male	Respondent is male
Married	Respondent is married
Human Capital
Educational attainment
Primary & below	Respondent attended primary school or never attended school
Junior high	Reference (Respondent attended junior high school)
Senior high & above	Respondent finished senior high school, technical secondary school, college or university
Work experience	The common logarithm of months since respondent began to work in cities
Dialect proficiency	Respondent can speak Cantonese
Social capital	1 more than the mean of the average size of social support network
Initial migration motives
Economy/family-oriented	Reference(Respondent migrated to make money, to marry or to take care of family members at the very beginning of migration)
Skill-driven	Respondent migrated to gain studying/learning skills, broaden experience in the first place
Preference for city life	Respondent migrated because of preference for life in cities
Familial factors
Spouse at hometown	Respondent’s spouse is left behind at home
Children at hometown	Respondent has at least one children <=16 at hometown
Parents’ health status	Respondent’s father or mother is not able to do housework, fieldwork or work outside
Current working and living conditions
Job nature
Manual	Reference (Respondent is an industrial worker or manages the household, or works in commerce or service industries)
Non-manual/semi-manual	Respondent is an administrator, a manager, a professional or technical employee, a clerk or an owner of a private enterprise
Self-employed	Respondent is self-employed
Number of jobs taken	The number of jobs respondent has taken since beginning to work in cities
Income	The common logarithm of monthly income plus 1(Yuan), i.e., log(income+1)
Housing condition
Concentrated-housing	Reference (Respondent lives in a relatively isolated community that is full of rural-urban migrants)
Scattered-housing	Respondent lives in a community of urbanites, or a mixed community of urbanites and rural-urban migrants
Social conditions of host areas
Discrimination	Respondent reports being frequently discriminated against by urbanites
Social and economic conditions of origin areas
Western	Reference(Respondent comes from Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Ningxia, Qinghai or Xinjiang)
Central	Respondent comes from Shanxi, Inner Mongolia, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, Hunan, Guangxi, or Hainan
Eastern	Respondent comes from Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shangdong, or Guangdong

With respect to the definition of the new generation, Wang’s (2001) seminal study claims that the new generation of rural-urban migrants are those who were born after the late 1970’s, educated in the 1980’s, and first migrated into cities after 1990. In previous empirical studies, a certain birth year has been used as a dividing line between the two generations (e.g., 1975 in Luo and Wang’s (2003) research, 1980 in Wang (2008)). As mentioned above, this division is somewhat arbitrary. In order to reduce the results’ sensitivity to using a specific birth year as a dividing line, we use the transitional period (from 1970 to 1980) between the two generations as 11 possible dividing lines in our study. The younger respondents who were born after the dividing line are defined as the new generation, and all the other older respondents are referred to as the old generation.

The first category of determinants is termed individual factors, and includes age, gender, marital status, human capital, social capital, and initial migration motives. In our study, human capital not only includes educational attainment but also work experience and dialect proficiency, which are accumulated or obtained during migration and may affect a migrant’s productivity. In this paper social capital refers to resources embedded in social networks, and is measured by the size of social support networks. Because members of the three types of social support networks overlap substantially, the average size of the three networks is taken as an indicator of a migrant’s social capital. Regarding initial migration motives, we define those respondents who migrated to make money, to marry, or to take care of family members as “economy/family-oriented” migrants, those who migrated for the purpose of studying, gaining skills, or broadening experience as “skill-driven” migrants, and those who migrated because they preferred life in the city as the third group.

The second category of determinants is familial factors and includes whether the migrant has a spouse or children at home and the health status of the migrant’s parents. These variables can reflect the psychological cost to migrants of migration; they can also indicate the family responsibilities of a migrant.

The third category includes factors that concern current working and living conditions. Three groups (manual, non-manual/semi-manual, and self-employed) are identified in terms of the respondent’s job. Job nature is used as an indirect indicator of occupational status. In our study, a migrant who is an industrial worker or manages the household or works in commerce or service industries is defined as a manual laborer. Owners of a private enterprise (only ten) are incorporated into the non-manual or semi-manual group. The number of jobs the respondent has taken is also regarded as an indicator of that migrant’s working conditions. Monthly income is used as an indicator of economic status. Respondents who live in a community of urbanites or a mixed community of urbanites and rural-urban migrants are defined as scattered-housing migrants. Compared with the scattered-housing group, concentrated-housing migrants are often housed in more cramped areas with poorer ventilation and sanitation and usually spend less or even no money on housing.

The last two categories are indicators of social and economic conditions at the areas of origin and destination (also referred to as host areas). Social conditions in host areas are indicated by whether migrants are frequently discriminated against. Because our respondents all come from Shenzhen, the economic conditions in the host city cannot be addressed in our study. Social and economic conditions at the place of origin are indicated by the region in which the respondent’s hometown is situated. According to the standard established by the National Bureau of Statistics, China can be geographically divided into three regions: eastern, central, and western (National Bureau of Statistics of China, 2003). Consequently, Guangxi and Hainan are part of the Eastern region. However, their per capita GDPs in 2005 (data source: National Bureau of Statistics of China, 2006) are so low that they cannot compare economically with other eastern provinces. Thus we group them with the central region. The eastern region is very fertile, and also enjoys the highest levels of economy and marketization. The central region also has good conditions for agriculture, but although the levels of its economy and marketization are better than the western, they are far behind the eastern (National Bureau of Statistics of China, 2003). Thus we use western, central, and eastern as categories representing the social and economic conditions in the rural area of origin.

Of the total of 1,739 participants in the survey, the 6% of the respondents who reported having no idea about intentions are excluded from our study. Because of the “temporary” nature of rural-urban migration, our cross-sectional data are inevitably biased to some extent. Thus the migrants who remain in the city at any point in time are not a representative sample of the cohort that originally entered. Specifically, the earlier migrants, either failures who could not secure a job or adapt to city life, or target earners who accomplished their earning targets, have returned home, while some very successful migrants have settled down in the city. This cumulative effect causes bias in sample. Since few migrants are permitted to settle in cities, the bias is limited and not fatal. At the same time, in order to remedy the problem in our data, we excluded from our analysis 43 interviewees who already owned their own house in Shenzhen. Because they are potential or even actual city settlers, most of them have already decided to settle in Shenzhen for the rest of their life. The remaining 1,598 respondents are included in the present study.

Table 2 presents descriptive statistics of the independent variables used in our analysis, separately for the old and new generations (taking the models with a dividing line of 1975 as an example). On average, work experience of the new generation is less than the old generation. Significant differences exist between the two generations in initial migration motives: only 2% of the old generation migrated for the purpose of learning skills or broadening experience and 4% because they preferred city life, compared to 17% and 20% of the new generation, respectively. Because of their earlier stage of life, the new generation does not have the same extent of psychological costs and family responsibilities as the old. A slightly larger proportion of the new generation is engaged in non-manual/semi-manual work, while many more migrants of the old generation are self-employed than of the new. The new generation’s monthly income is not as high as the old generation. Discrimination against migrants is not very severe in Shenzhen. Most of the migrants of both generations come from the central region.

Table 2.

Descriptive statistics of independent variables

Variables	Whole sample		The old generation		The new generation
	Mean	SD	Mean	SD	Mean	SD
Generation
The new generation	0.52	—	/	/	/	/
Individual factors
Age	/	/	37.89	5.87	24.31	3.62
Male	0.51	—	0.66	—	0.37	—
Married	0.66	—	0.94	—	0.39	—
Human capital
Educational attainment
Primary & below	0.12	—	0.20	—	0.05	—
Senior high & above	0.29	—	0.25	—	0.34	—
Work experience	1.76	0.43	1.88	0.43	1.64	0.40
Dialect proficiency	0.24	—	0.22	—	0.27	—
Social capital	0.11	—	0.12	—	0.10	—
Initial migration motives
Skill-driven	0.10	—	0.02	—	0.17	—
Preference for city life	0.12	—	0.04	—	0.20	—
Familial factors
Spouse at hometown	0.15	—	0.24	—	0.06	—
Children at hometown	0.34	—	0.50	—	0.19	—
Parents’ health status	0.17	—	0.26	—	0.08	—
Working and living conditions
Job nature
Non-manual/semi-manual	0.10	—	0.09	—	0.11	—
Self-employed	0.18	—	0.27	—	0.10	—
Number of jobs taken	1.84	1.93	1.99	2.28	1.71	1.53
Income	2.94	0.65	2.98	0.68	2.90	0.63
Housing condition
Scattered-housing	0.67	—	0.77	—	0.57	—
Social conditions of host area
Discrimination	0.02	—	0.04	—	0.01	—
Social and economic conditions of origin areas
Central	0.51	—	0.50	—	0.52	—
Eastern	0.25	—	0.25	—	0.26	—
Number of cases	1598		762		836

Note: the dividing year is 1975.

Source: Data from the 2005 Shenzhen Survey of rural-urban migrants.

Analytical strategy

According to the intentions they reported, we divide the respondents into three groups: agricultural returnees, nonagricultural returnees, and settlers. Multinomial logistic regression is well suited for analyzing the relationship between a multiple category dependent variable and metric or dichotomous independent variables (Powers and Xie, 2000). Taking the agricultural returnees as the reference group, for each possible dividing line (from 1970 to 1980) we carried out three multinomial logistic regressions using STATA 10.0: first, we ran a whole sample model with all the other variables controlled for; second, we did the analysis separately for the old and new generations. In total there are 33 regressions. Only results that were consistent in terms of direction and significance (at least significant at the 0.1 level, i.e., the absolute values of t statistics are not less than 1.65) in all 11 sets of models will be interpreted and discussed in our study. However, for the purpose of saving space, we present only the results with the dividing line of 1975 as an example.

Results

Intention comparison between the two generations

Table 3 presents the distribution of intentions by birth year and generation (taking the model with a dividing line of 1975 as an example). The patterns of intentions differ by birth year: the younger the migrants, the more likely they are to intend to be nonagricultural returnees and settlers, rather than to be agricultural returnees. The same patterns hold by generation, substantial differences in intentions can be seen between the two generations. The proportion of the old generation that intends to settle in cities is a little higher than that of the new one, but because of the bias mentioned above, it is not safe to conclude that the old generation is more willing to settle down in cities. Further analysis of the differences in intentions between the generations will be presented in the multivariate models in the next section.

Table 3.

Intentions of migrants by birth year and generation (%)

	Agricultural Returnees	Nonagricultural Returnees	Settlers
By birth year
Before 1970 (n=452)	49.8	7.3	42.9
1970 (n=54)	42.6	25.9	31.5
1971 (n=65)	38.5	21.5	40.0
1972 (n=63)	38.1	11.1	50.8
1973 (n=62)	21.0	22.5	56.5
1974 (n=66)	31.8	21.2	47.0
1975 (n=62)	25.8	40.3	33.9
1976 (n=92)	33.7	28.3	38.0
1977 (n=58)	20.6	39.7	39.7
1978 (n=52)	23.1	30.8	46.1
1979 (n=66)	16.7	37.9	45.4
From 1980 (n=506)	12.9	48.6	38.5
Whole sample (n=1598)	29.9	28.6	41.5
By generation (the dividing line is 1975)
The old generation (n=762)	43.4	12.6	44.0
The new generation (n=836)	17.6	43.2	39.2

Source: As for Table 2.

Determinants of intentions

Taking the agricultural returnees as the baseline group, we present the results of three multinomial logistic analyses of the whole sample, the old generation and the second generation in Table 4. We take here the models with a dividing line of 1975 as examples. In the whole sample models, the variable of generation is defined by age, so age is excluded from the whole sample analysis.

Table 4.

Odds-ratios and t-values of the multinomial logistic regression of intentions (agricultural returnees constitute the reference group)

Variables	Whole sample		The old generation		The new generation
	Nonagricultural returnees	Settlers	Nonagricultural returnees	Settlers	Nonagricultural returnees	Settlers
Generation
The new generation	3.295 (5.97)	1.433 (2.03)	/	/	/	/
Individual factors
Age	/	/	0.896 (−3.62)	0.988 (−0.78)	0.938 (−1.45)	0.945 (−1.28)
Male	0.775 (−1.50)	0.783 (−1.65)	0.796 (−0.80)	0.881 (−0.64)	0.898 (−0.44)	0.825 (−0.77)
Married	0.675 (−1.69)	0.815 (−0.95)	0.618 (−0.89)	0.900 (−0.25)	1.094 (0.27)	1.182 (0.51)
Human capital
Educational attainment
Primary & below	0.314 (−3.93)	0.649 (−0.95)	0.392 (−2.23)	0.710 (−1.51)	0.245 (−3.11)	0.359 (−2.41)
Senior high & above	2.127 (4.00)	2.164 (4.46)	2.546 (3.12)	1.553 (1.95)	2.593 (3.29)	3.338 (4.13)
Work experience	0.913 (−0.47)	0.953 (−0.28)	1.327 (0.84)	0.932 (−0.32)	0.921 (−0.27)	1.135 (0.40)
Dialect proficiency	1.782 (2.36)	1.949 (3.17)	2.021 (1.73)	2.256 (2.86)	1.552 (1.33)	1.602 (1.47)
Social capital	1.369 (1.28)	0.938 (−0.28)	1.130 (0.33)	0.651 (−1.48)	2.503 (2.06)	2.019 (1.53)
Initial migration motives
Skill-driven	3.757 (4.29)	1.641 (1.53)	1.117 (0.11)	2.484 (1.31)	3.565 (3.63)	1.374 (0.85)
Preference for city life	2.477 (3.04)	2.768 (3.43)	3.946 (1.85)	2.532 (1.36)	2.073 (2.16)	2.409 (2.57)
Familial factors
Spouse at hometown	0.581 (−2.21)	0.547 (−2.92)	0.686 (−1.09)	0.642 (−1.84)	0.568 (−1.36)	0.368 (−2.19)
Children at hometown	1.103 (0.49)	0.785 (−1.51)	1.042 (0.14)	0.806 (−1.08)	0.867 (−0.43)	0.669 (−1.21)
Parents’ health status	0.680 (−1.82)	0.681 (−2.17)	0.911 (−0.32)	0.635 (−2.21)	0.638 (−1.25)	0.854 (−0.42)
Working and living conditions
Job nature
Non-manual/semi-manual	1.656 (1.69)	1.538 (1.58)	1.227 (0.44)	1.616 (1.39)	1.444 (0.78)	1.302 (0.56)
Self-employed	1.502 (1.59)	2.248 (4.23)	1.763 (1.60)	2.239 (3.62)	0.918 (−0.20)	1.821 (1.56)
Number of jobs taken	1.059 (1.37)	1.048 (1.25)	1.023 (0.42)	1.033 (0.75)	1.196 (1.84)	1.209 (1.97)
Income	0.957 (−0.37)	1.128 (1.18)	0.938 (−0.32)	1.130 (0.91)	0.922 (−0.49)	1.048 (0.29)
Housing condition
Scattered-housing	0.582 (−2.97)	2.101 (4.16)	0.589 (−1.62)	3.279 (4.29)	0.516 (−2.57)	1.690 (1.94)
Social conditions of host areas
Discrimination	0.533 (−1.21)	0.604 (−1.25)	0.474 (−0.92)	0.641 (−0.93)	0.650 (−0.55)	0.647 (−0.56)
Social and economic conditions of origin areas
Central	0.889 (−0.66)	1.085 (0.49)	0.695 (−1.25)	0.886 (−0.56)	1.262 (0.93)	1.562 (1.65)
Eastern	0.968 (−0.12)	1.876 (2.73)	1.020 (0.05)	1.409 (1.14)	1.162 (0.40)	2.792 (2.75)
Chi-square	657.38		241.48		246.24
Significance level	<0.001		<0.001		<0.001
Number of cases	1598		762		836

Note: the dividing year is 1975. The numbers in parentheses are the t-values. Items in bold and underlined are consistently significant results in terms of direction and significance (at least significant at the 0.1 level on two-tailed tests, i.e., the absolute values of t-values are not less than 1.65) in all 11 sets of models.

Source: as for Table 2.

In the whole sample analysis with the other variables controlled for, we find that the new generation is more likely to plan to be nonagricultural returnees and settlers rather than to be agricultural returnees. In the models with the dividing line of 1975, the odds-ratios associated with the new generation dummy are 3.295 and 1.433, respectively. Since the same effect holds in all 11 models, it is safe to conclude that the new generation is more willing to secure an off-farm job, either to be settlers in cities or to be nonagricultural returnees, than to be agricultural returnees. We also find that the new generation has a higher likelihood of being nonagricultural returnees rather than settlers by taking the nonagricultural returnees as the baseline group (when we use 1980 as the dividing line the effect of generation is not significant, the t-value is −0.88). Because of space limitation, we do not show these results.

In the 11 models for the old generation, several consistent findings are worth noting. Age and education have effects on migrants’ intention of being nonagricultural or agricultural returnees. Increasing age significantly reduces the old generation’s likelihood of being nonagricultural returnees. As age goes up by one year, the old generation is 0.896 time less likely to be nonagricultural returnees in the model with the dividing line of 1975. Education has a positive effect on it, with those who enjoy a higher level of education more likely to be nonagricultural returnees. For example, in the model with the dividing line of 1975, migrants who finished junior high school are 2.551 (i.e., 1/0.392) times more likely to be nonagricultural returnees than those who finished primary school & below, and those who finished senor high school & above are 2.546 times more likely to be nonagricultural returnees than those with a educational level of junior high school. Education, dialect proficiency, familial factors, job nature and housing condition have effects on migrants’ intention of being settlers in cities. Education’s positive effect holds for being settlers as it does for being nonagricultural returnees. Those migrants of the old generation who can speak Cantonese are more likely to be settlers rather than agricultural returnees. For example, the odds-ratio associated with dialect proficiency is 2.256 in the model with the dividing line of 1975. Having a spouse at home significantly reduces the likelihood to settle in cities. The odds-ratio associated with it is 0.642 in the model presented in Table 4. Self-employed migrants tend to be settlers rather than agricultural returnees. In the model with the dividing line of 1975, for example, the self-employed migrants are 2.239 times more likely to be settlers than the manual labors. Those scattered-housing migrants who have better housing conditions are 3.279 times more likely to plan to settle in cities in the model with the dividing line of 1975. It is also worth noting that increasing age significantly increases the old generation’s likelihood of being settlers rather than being nonagricultural returnees by taking the nonagricultural returnees as the baseline group. The odds ratio associated with age is 1.103 (i.e., 0.988/0.896) when we use 1975 as the dividing line.

In the models for the new generation, education’s role holds as for the old generation: the higher the level of education, the more likely the migrants are to secure an off-farm occupation. Social capital, initial migration motives and housing condition have significant effects on the intention of being a nonagricultural returnee. Migrants from the new generation who have more social capital are 2.503 times more likely to plan to be nonagricultural returnees in the model with the dividing line of 1975. The skill-driven migrants of the new generation and those migrants who prefer city life are more likely to be nonagricultural returnees. In the model with the dividing line of 1975, the odds-ratios associated with them are 3.565 and 2.073, respectively. Scattered-housing condition reduces their likelihood to be nonagricultural returnees. The odds-ratio associated with housing condition is 0.516 in the model presented in Table 4. Initial migration motives and social & economic conditions exert significant effects on migrants’ intentions of being a settler. Those migrants who prefer city life have a higher likelihood of staying in cities. In the model presented in Table 4, for example, this is 2.409 times more likely to occur. Migrants from the eastern region are more likely to plan to settle in cities. The odds ratio associated with the eastern region is 2.792 in the model shown in Table 4. Familial factors have no consistent significant effect on migrants’ intentions.

Education has the same effects on intentions of rural-urban migrants: those with higher education are more likely to intend to be non-farm workers, either settlers or nonagricultural returnees. But when compared with the nonagricultural returnees and taking the “primary & below” as the reference group, those migrants with higher education are less likely to intend to settle down in cities, i.e., returnees are not necessarily negatively selected by education, nonagricultural returnees are positively selected (this effect is not significant for the new generation). With respect to the housing condition, we find that those scattered-housing migrants are more willing to settle down in cities rather than to return to farm or to secure a nonagricultural job (this effect is not consistently significant for the new generation). Among returnees, the scattered-housing migrants of the new generation are less likely to intend to be nonagricultural laborers. A possible explanation for this is that scattered housing is of a higher standard but costs more, which would force failed migrants with lower income to return. Since securing a nonagricultural job requires more financial and human capital, they are more likely to intend to return to agriculture.

There are sharp differences in determinants between the two generations. Age, dialect proficiency, having a spouse at the hometown, and job nature only have effects on the old generations’ intentions. The effect of age cannot be interpreted straightforwardly. On one hand, increasing age decreases the physical capability to meet the demands of migrant work, but on the other, age also reflects the stage of life; for example, after children have grown up and married, their demand on family income would decline (Wang and Fan, 2006). To some extent this provides evidence that migrants from the old generation tend to be target earners. They will return to continue subsistence farming having accomplished their earning targets. At the same time, older migrants have usually accumulated more resources than the younger ones; thus the former are more likely to become successful and to afford life in cities. It is no longer necessary for them to return to secure a non-farm job. This possibly explains why the older migrants of the old generation are more willing to be settlers than nonagricultural returnees (see Table 1). Dialect proficiency, a typical measure of acculturation, also has an effect on intentions of the old generation, although Mandarin is also the main language in Shenzhen. This effect is not significant for the new generation, which might reflect greater adaptability and tolerance among the new generation. The effect of having a spouse at the hometown reflects the importance of family responsibility in determining intentions of being returnees or settlers for the old generation who generally carry more responsibilities than the new generation. In Table 4, we also see that parents’ health status plays a role in intentions of the old generation, although its effect is not consistently significant. To some extent, type of job determines migrants’ living conditions; inasmuch as the self-employed migrants, who usually earn higher income, are more likely to live a relatively decent life in cities than the manual laborers, these successful migrants are more likely to plan to settle down. However, this effect is not significant for the new generation.

Social capital, migration motives, and social & economic conditions in areas of origin all have significant effects on intentions of the new generation. For the new generation, not only human capital but also social capital affects migrants’ intentions. Migrants’ networks are mainly made up of relatives and fellow villagers from the same origin areas, and few network members come from the host city (Li et al, 2007). Thus, social capital only affects migrants’ intentions of being nonagricultural returnees rather than being settlers. A possible interpretation of the role of initial migration motives in intentions for the new generation is that, living in a transitional era, members of the new generation are more eager to take action to pursue their ambitions, which could be inferred from life-cycle theory. The socioeconomic situation at the origin areas plays an important role in migrations’ intentions. Migrants of the new generation from the eastern region where levels of economy and marketization are highest are more likely to plan to settle down in cities, suggesting that market transition does have an effect on migrants’ intentions. However, probably because of their later stage of life, this effect is not significant for the old generation.

Possible explanations for such differences between the two generations are the following. Because of their different stage of life plus growing up within a different socioeconomic context, i.e., the combined effect of life cycle and market transition, for the old generation the hard conditions such as age, family responsibility, and type of job play more important roles in their intentions, while for the new generation such conditions as migration motives and social capital matter much more. The old generation‘s intentions are strongly based on whether the current situation can provide a satisfying life, whereas intentions of the new generation are not sensitive to this. For the latter, migration is more likely to be conceptualized as a process of accumulating human and social capital, and they are more likely to adjust their attitudes and behavior when experiencing the profound socioeconomic changes occurring in contemporary China. Therefore, through learning new skills and broadening experience during migration, they want to realize their dream of attaining a non-farm job, either in a city or at their hometown.

Discussion and Summary

After the two decades of rural-urban migration, the new generation is gradually constituting the majority of rural-urban migrants. Attitudes and behaviors of the immigrants have changed over this period. Using data from the 2005 Shenzhen Survey, we examined the determinants of intentions of rural-urban migrants between the two generations while taking into account three possible development trajectories. Several findings are worth summarizing.

First, generation has a significant effect on intentions of rural-urban migrants. Compared to the old, the new generation is more eager to change occupation from farming to non-farm work. Returning to seek a nonagricultural job has become the most common choice for this generation. Unlike the old, the new generation no longer has a strong attachment to the original village and the land. Returning to continue farming has become their least-favored option. They do want to settle in cities by taking nonagricultural jobs. But because of their institutional and social inferiority, the new generation is no closer to integration into the city society than their predecessors were. Therefore, making a compromise to reality, becoming a nonagricultural returnee has become the new generation’s suboptimal but primary choice.

Only education plays the same role in intentions of old and new generation rural urban-migrants. The more education they have, the more likely they are to intend to be non-farm workers. Sharp differences in determinants of intentions exist between the two generations. For the old generation, the consistently significant determinants include age, dialect proficiency, spouse at hometown, and job nature. But only social capital, initial migration motives, and socioeconomic conditions have effects on intentions of the new generation. These differences suggest that the new generation is more likely to view migration as a form of investment with the accumulation of human capital and social capital, through which they hope to eventually realize an occupational transformation into non-farm workers.

At the same time, we also find that those rural-urban migrants from the old generation with higher education level are also more likely to plan to seek an off-farm career, which seems inconsistent with the target-earner arguments in previous studies. This suggests that their attitudes and behaviors also evolve. According to the structuration theory of Giddens (1984), Huang et al. (1996) provide a possible explanation. At the very beginning rural-urban migrants are stimulated by the idea of earning cash income. However, no matter how reasonable their initial motivation was and their practical purposes might be, as time passes they learn not only to re-adjust their previous aims, but also to make some changes in their motivations and actions (Huang et al, 1996). This is supported by the finding that initial migration motives have no effect on current intentions of the old generation. However, because of their stage of life and their insensitivity to market transition, their intentions of realizing the transformation to a nonagricultural occupation appear to be not as strong as those of the new generation.

Our findings also settle the dispute concerning whether returnees are negatively or positively selected in terms of education. According to our study, returnees are not necessarily negatively or positively selected when we take the “primary & below” as the reference group: nonagricultural returnees from the old generation are positively selected even compared with migrants who intend to settle in cities, while the agricultural returnees are negatively selected. However, this effect is not significant for the new generation, which probably reflects that this selectivity pattern is experiencing some changes in China’s transitional era. A possible explanation for this is that because of the segmentation of the urban and rural labor markets in China, most migrants’ jobs tend to be menial and do not reward education (Zhao, 2002). So compared to settling in cities, returning to seek a nonagricultural job seems to reward education more. Since human capital has a great influence on the development of the economy, we suggest that “nonagricultural returnees” will play an important role in the future development of rural areas. This also has an important policy implication for migrant children. In the migration process, education is a principal concern for migrant families and for China’s Government. Despite government efforts to promote equal rights in the nine-year compulsory education for all of China’s children, including offspring of current migrants (whether they are among the 6.5 million taken to urban areas or the 22.9 million “left behind” in rural areas), nearly half of all migrant children cannot go to school and 9.3% of them drop out (China Daily, 2004). There are cheaper schools specifically for migrant children in cities, but the teaching and facilities are of a much lower quality. Considering the role of education in transforming domestic agricultural farmers into industrial workers, more measures need to be taken to improve migrant children’s education. The gap in education quality between sending and receiving areas, and between children of migrants and children with urban Hukou, should be narrowed. For adult migrants, vocational education and training programs should be developed according to the needs of employers. More communication between the local authorities in sending areas, vocational training institutions and employers from state-owned and private enterprises should be encouraged.

The tremendous abundance of labor in rural areas is still one of the most challenging issues currently facing policy-makers in China. Rural-urban migrants, especially the new generation, strongly desire to settle in cities. However, the institutional and social discrimination they experience pushes them into a “marginal man” dilemma; many of them are reluctantly forced to return with a city-settler dream. Our findings reveal that, after having enhanced their human, social and financial capital, taking into account their experience of institutional and social discrimination, the new generation plans to return to seek a nonagricultural job, and this should be a driving force for the economy of the sending areas. Returnees will play an important role in developing local economies by making use of their physical, human, financial, and social capital that was gained during migration, which can help to reduce the growing regional disparity between coastal/rural and inferior/urban areas. Return migration also facilitates the ongoing diffusion of labor-intensive industries from the coastal region to inland areas, which provides a great opportunity for these less developed areas. This “grass-roots” path of modernization can accelerate the pace of urbanization and industrialization, and bridge urban-rural and inter-regional inequality.

Our study has several limitations. First, the Shenzhen survey of rural-urban migrants was not specifically designed to study the intentions of the two generations. As a result, some important variables likely to affect migrants’ intentions, such as the per capita arable land at the area of origin, vocational training experience, farming experience, and the labor ratio in the family, were not included in our questionnaire and therefore could not be addressed in the present article. Second, our results are based on cross-sectional, rather than longitudinal data. As we discussed previously, the cumulative effect causes bias in our data. Despite measures taken to remedy this problem, our findings are probably somewhat affected; for example, it may lead us to overestimate migrants’ (especially the old generation) rate of planning to settle down in cities because failures and some target earners have returned home. Fortunately, for determinant analysis, these influences are likely to be limited. Finally, the intentions of rural-urban migrants will not necessarily translate into actions. However, the gap between intentions and eventual actions can enhance our understanding of rural-urban migrants and is an important issue.