Patterns of onwards migration within the urban hierarchy of China: Who moves up and who moves down?

Abstract

Chinese

This paper examines onwards migration patterns and the factors that influence them in urban China. In contradiction to Ravenstein’s laws of step migration, we argue that movement up the urban hierarchy does not dominate onwards migration in the Chinese context. Using the 2017 China Migrants Dynamic Survey, we model how past migratory experiences intersect with migrants’ demographic and socio-economic characteristics to explain the stay-versus-move decision and who moves up as well as who moves down within China’s urban hierarchy. The general pattern of China’s domestic migration is a movement up the urban hierarchy in the first migration and a movement down in onwards migration. We identify a transition-levels effect whereby the larger the difference between the hukou-registered place and the first migration city in the urban hierarchy, the greater the probability of onwards migration, with the effect manifesting most strongly for less-educated, mid-age (the 1970s and 1980s birth cohorts) migrants with rural origins. As to onwards migration direction, mixed results were found for migrants’ demographic and socio-economic characteristics whereas educational level and the transition-levels effect are both highly predictive of movement up and down the urban hierarchy. Our findings suggest that ‘jumping too high’ in previous migration(s) predicts a greater likelihood of moving down the urban hierarchy in response to insurmountable obstacles to settling in cities classified as large and above. Our research advances migration scholarship by considering the multipolar nature of onwards migration and offering an integrated approach to analysis that foregrounds the causes and effects of multiple-step migration.

Keywords

onwards migration past migratory experiences selectivity transition-levels effect urban hierarchy

Introduction

Migrants move for household betterment. The critical questions in migration studies have always been: What is the migration pattern? And: Who moves and what explains the decision-making that leads to each move? (Clark and Lisowski, 2017). The first question concerns the sequence of geographical moves, whether one-way migration, that is leaving home and settling in a host destination, or repeat migration, that is making return, circular and/or onwards moves. Whatever the migration trajectory, it is generally assumed that onwards migration is dominated by upward migration because moving to larger cities is expected to bring greater returns (Yap, 1977). The more educated and better-resourced migrants are deemed to be more likely to undertake onwards migration (Della Puppa et al., 2021). Accordingly, Ravenstein’s (1885) age-old ‘laws of migration’ present migration as characterised by ‘step migration’– that is a movement up the urban hierarchy from a rural area to a nearby town and then to a larger commercial and industrial centre. A similar pattern is reported in recent literature on stepwise international migration whereby migrants from the ‘Global South’ accumulate economic and social capital as well as legal rights (e.g. European citizenship) during their stay in the first host country only to make their way to a final destination usually in Western Europe or North America (Ahrens et al., 2016; Della Puppa and King, 2019; Della Puppa et al., 2021; Jeffery and Murison, 2011; King and Newbold, 2007; Ortensi and di Belgiojoso, 2018).

This study focuses on substantiating an alternative onwards migration pattern based on Chinese survey data. Counter to the step migration model, we consider migration patterns in China relative to moving up and down the urban hierarchy. On this basis, we present a panorama of migration patterns within the urban hierarchy to capture in a coherent way the dynamics of the primary migration and subsequent ‘corrective’ moves. We then advance the field’s theoretical understanding of onwards migration by examining the impact of individual characteristics, household characteristics, and city level of residence on stay-versus-move decisions and patterns of onwards migration.

Onwards migration is usually attributed to factors relating to the migrant’s job/income and/or to life course events (Clark, 2013) or access to urban amenities (i.e. neighbourhood and lifestyle) (Clark and Maas, 2015). We add a related perspective that foregrounds previous migratory experience as a causative factor. Further, our focus on the explanatory role of migratory experiences is broader than the ‘learning-by-doing’ effect (DaVanzo, 1981) or the idea of ‘migration seniority’ (Della Puppa, 2018) as presented in earlier studies, which highlight knowledge, skills, and social capital accumulated during previous migration(s) and posit migratory experiences as conducive to further moves. We agree with this view in part, yet present a different scenario wherein the experience of ‘jumping too high’ may increase the likelihood of downward movement in the urban hierarchy. Moreover, onwards migration varies in terms of the destination level targeted, such that migrants may move down to their original level or to some intervening level between it and the first migration level.

In line with the call for greater emphasis on the multipolar nature of migration (Fan, 2011), this study constitutes a significant step towards a comprehensive examination of diverse onwards migration patterns relative to varying destinations, including upward, horizontal, and downward movements. The findings of this study contribute to migration scholarship by moving beyond the well-established dichotomous ‘settlement versus return’ idea of migration, which has long had the effect of limiting inquiries into migration patterns in China.

Theoretical context

The multipolar nature of onwards migration

For a migrant originating from the bottom of the urban hierarchy, Ravenstein’s (1885)‘laws of migration’ predict his/her migratory trajectory as most likely to move upward to the ideal city level in the urban hierarchy. Disparities in urban employment opportunities, wage levels, and living standards between small and large cities encourage migrants to move up the urban hierarchy. Inasmuch as larger cities tend to offer higher wages and more urban amenities, migrants may move from smaller to larger cities over time (Yap, 1977). Although upward urban mobility has been predominant in most countries, especially during the industrialisation era, migration down the urban hierarchy has also been found in some circumstances (Plane et al., 2005). In the US, adults aged 25–29 are more likely to move upward in the urban hierarchy, whereas people at mid-career and during their child-rearing years tend to move downward and people at pre-retirement and retirement age (late 50s and 60s) tend to prefer non-metropolitan destinations (Plane and Jurjevich, 2009; Plane et al., 2005). Similar patterns are found in the UK, the Netherlands and Australia (Atkins, 2018; De Jong et al., 2016; Dennett and Stillwell, 2010).

In the present study, we capture the complex dynamics of migratory behaviours in the urban hierarchy in an integrated way with the goal of achieving a comprehensive picture of migration patterns in the urban Chinese context. We offer a thorough examination of these patterns in which we account for the multipolar nature of onwards migration with the possibility of transitioning between any two urban levels, whether from any higher to any lower level including to the ground state (hometown) or from any lower to any higher level. Previous studies have shown that the majority of Chinese migrants show a preference for large cities (Mu et al., 2022; Xing and Zhang, 2017). However, a ‘great leap forward’ from a lower-level household registration system (hereafter referred to as the hukou system)-registered birthplace to a higher-level destination city is hardly risk-free, with rural migrants the largest group to account for in this regard. A substantial literature describes rural migrants as a ‘floating population’ (Goodkind and West, 2002) or ‘economic sojourners’ (Wu, 2012) subjected to far-reaching institutional discrimination and social marginalisation in urban China (Chan and Wei, 2019; Fan, 2004; He et al., 2019; Mai and Wang, 2022). Consequently, researchers have generally argued that rural migrants have relatively low settlement intentions relative to their host cities (Cao et al., 2015; Hao and Tang, 2015; Mohabir et al., 2017; Zhu and Chen, 2010). This brings a new question into focus: if migrants are unsettled, where do they move? By offering the conclusion that migrants either circulate between their hometowns and host destinations or that they eventually return to the former, the literature on circular and return migration offers only partial answers to this question (Chen and Fan, 2018; Fan, 2011; Hu et al., 2011; Wang and Fan, 2006; Zhao, 2002). Emphasising institutional constraints (i.e. hukou) and attachment to the hometown, studies in this vein proceed from the implicit assumption that migrants cannot escape the ‘gravity’ of the latter. This bipolar assumption of migratory behaviours, however, is problematic as it excludes any possibility of onwards migration to a new place, one that is neither the hometown nor the first destination city.

Although the onwards migration phenomenon in urban China has been identified by a few scholars, its pattern in relation to the urban hierarchy has remained unclear. Based on field research in the Anhui and Sichuan provinces in 1998, Zhao (2002) reported that a return home often constitutes a stopover between two migrations. When leaving a given city, migrants often go to a different city, other than their hometown, even if they spend time in the hometown first. Supporting Zhao’s assertion, Zhu and Chen (2010) concluded both that migration is not always a one-way process of leaving the hometown (often in a rural area) and settling in a city and that the final destinations of the floating population have become increasingly diverse. However, the literature falls short of producing the overall patterns of China’s onwards migration and offers few attempts to identify the factors influencing them.

Onwards migration as an integrated process of multiple moves

A synthetic theorisation drawing on multiple disciplines and theoretical mechanisms comes closer to explaining migration than does any single model (Fussell and Massey, 2004; Massey et al., 1993). Economic perspectives dominated earlier explanations of out-migration, principally neoclassical economics (Sjaastad, 1970; Todaro, 1969), dual labour market theory (Piore, 1979), and the new economics of labour migration (Stark, 1991). In subsequent studies, researchers shifted away from the economic equilibrium approach by taking non-monetary concerns into consideration (Clark and Maas, 2015; Liu et al., 2017). In this regard, studies on the cumulative causation of international migration (Massey et al., 1993), that is life course events (Clark, 2013) and prospects (Kahneman and Tversky, 2013), provide a valuable foundation for apprehending the role of multiple additional factors, such as social network, changing family needs, communities and lifestyles, in migration decision-making.

However, most of these studies adopt a static, rather than a dynamic, relational approach to explaining migration. Life course theory and prospect theory have advanced our understanding by contextualising migration within a process of change and presenting it as an age-selective process with migration propensities and location choices varying greatly over the life course (Bailey, 2009; Bernard et al., 2014; Geist and McManus, 2008). Migrants’ changing priorities and preferences in regard to urban size at different ages and life stages are often considered the main cause of migration patterns across the urban hierarchy (Clark, 2013; Plane et al., 2005). According to prospect theory, the migrant’s expectations in regard to gains and losses, relative to what his/her household already has, drive the evaluative migration decision process (Clark and Lisowski, 2017). The evaluation considers not only the concrete endowments associated with staying, such as income, housing tenure and urban amenities, but also the abstract endowments of doing so, such as social ties, sense of belonging, and attitude to risk (Clark and Lisowski, 2017; Mohabir et al., 2017). Endowments are positively related to duration of residence: the more endowments a migrant has, the greater the value of staying.

Given a stay-versus-move framework, the research to date on theorising migration reveals little about how to interpret an integrated process of multiple migrations. In terms of onwards migration, most analyses focus on the demographic and socio-economic characteristics and conditions of the hometowns and the current residences (see, Clark and Maas, 2015; Plane et al., 2005) while downplaying the role of past migratory experiences in shaping migratory decisions. In China, given the deprived and marginalised living experiences of rural migrants in their first destination cities (Afridi et al., 2015; Wang and Fan, 2006), migrating up the urban hierarchy in subsequent moves is unlikely to be a reasonable choice. Simply put, larger cities often present higher barriers to settlement. Those who have failed to become well-established in the first destination are more likely to move down rather than up the urban hierarchy. Another line of explanation draws on the life course approach, foregrounding family demand as an important driver of subsequent moves (Wang and Fan, 2006). From this perspective, onwards migrants may be either positively or negatively selected. Our focus on extended onwards migration, therefore, emphasises that migration is not a one-time behaviour but a successive process in which past experiences are surely fundamentally important (DaVanzo, 1983; Della Puppa, 2018; Paul, 2011; Tian et al., 2016).

To summarise, we argue that movement up the urban hierarchy does not dominate onwards migration in China. In other words, migrants are more likely to relocate down the urban hierarchy if unsettled in their initial destination. Although we propose downward movement as most likely, migration direction can also be horizontal or even upward. Drawing on a synthesis of migration theories, we explore the causes and effects of the first two successive migrations. In particular, we seek to answer three specific questions: What factors influence migrants’ decisions over whether or not to undertake onwards migration from the first migration city? When migrants do undertake onwards migration, which of them tend to move up the urban hierarchy and which tend to move down, and what factors contribute to these respective directions? In what ways do past migration experiences influence onwards migration decisions and destination choices?

Data and methodology

Data source

The dataset used in the present study is the 2017 China Migrants Dynamic Survey (CMDS) – a national survey conducted by the National Health and Family Planning Commission of China based on multistage stratified probability proportional to size sampling. Township-level units and village or neighbourhood committees are the primary sampling units (PSUs) and secondary sampling units (SSUs), respectively. The survey provides rich information on the migrant population, including hukou registration place (usually hometown), current place of residence, date and destination of first migration, and personal and household demographic and socio-economic characteristics. The sample used in this study was randomly selected from all the migrants who had been living in each selected SSU for at least one month without local hukou at the time of the survey.

The entire CMDS dataset comprises 169,989 migrants residing in 1290 county-level units (urban districts, county-level cities, and counties and their equivalents) in 31 provincial-level units in mainland China. We performed a data-cleaning process based on two concerns. First, as the survey identified only the migrants’ first destination city and current city of residence, we excluded migrants who had migrated more than twice in our sample. Second, any migration undertaken for a special purpose such as marriage, child care, and college attendance were excluded, as these migrants are to some extent passively selected. Our final sample comprised 81,038 observations.

Urban hierarchy

In China, ‘cities’ (shi) are defined in relation to administrative status: county level (xianji shi), prefectural level (diji shi), and centrally administered (zhixia shi). Almost all China’s cities at and above prefecture level have an urban district (shi xia qu),¹ counties, and county-level cities within their administrative purview. Urban districts are generally the continuous built-up areas and areas which can be regarded as urban areas in a true sense given the nature of their land use, population density, and urban functions (Ma, 2005). The counties and county-level units administered by the prefecture-level cities and the prefectures are regarded as separate units. We treat these as three distinct types of geographical units and refer to all as ‘cities’: the urban districts of prefecture-level cities and above, county-level cities, and counties and equivalents.

We adopted a classification based on population size, which is well-suited to capturing the hierarchical features of the urban system and, therefore, widely used in the literature (De Jong et al., 2016; Korpi et al., 2011; Mu et al., 2022; Plane and Jurjevich, 2009; Plane et al., 2005). By taking The Notice on Adjusting the Standard of City Size Division as the reference,² we divide Chinese cities into five categories based on the urban population of each city:³ (1) super-large (higher than 10 million); (2) extra-large (5–10 million); (3) large (1– 5 million); (4) medium (500,000 to 1 million); (5) small (less than 500,000); and (6) counties and equivalent county-level units. Approximately 85% of these counties had an urban population of less than 0.2 million in 2010, with an average urban population of 0.12 million. The numbers of cities of the six levels identified in the 2017 CMDS were 4, 10, 66, 111, 250 and 440, respectively.

Measurement of onwards migration and migratory directions

On the basis of the six levels comprising the urban hierarchy, we identify onwards migration by comparing three locations of record: the place where hukou is registered, the first-migration city, and the current city of residence. When the current city of residence differs from the place where hukou is registered and from the first migration city, onwards migration has occurred – that is the migrant has moved to another city rather than either remaining in the first migration city or returning home. Of the 81,038 valid observations, 62,379 observations (approximately 77%) stayed in the first migration city, and the remaining 18,659 observations (approximately 23%) undertook onwards migration. Similarly, through a comparison of the first migration city and the current city of residence in terms of urban level, three possible migratory directions can be observed: upward, downward, and horizontal. Moving upward means relocating from a lower-level to a higher-level city, whereas moving downward means a move in the the opposite direction. Migration between two cities at the same urban level is termed horizontal migration.

Variables and methods

Based on our three research questions, we model (1) the impact of a set of individual migrant characteristics and the difference in urban level between the hukou city and the first migration city in stay-versus-move decisions; (2) the direction of onwards migration relative to the urban hierarchy for migrants who undertake onwards migration; and (3) past migratory experiences and their association with downward, horizontal, and upward movement.

The regression models in this study include two dependent variables. First, a dichotomous nominal variable is used to indicate whether or not migrants undertook onwards migration or stayed in the first migration city after moving from their hometown (onward = 1, staying = 0). The second dependent variable is multinomial, denoting the three possible onwards migration directions of downward, horizontal, and upward, set to 1, 2 and 3, respectively.

In addition, several variables are included to control the effects of individual and family attributes – gender, ethnicity, marital status, children, hukou status, birth cohort, and educational level – on decisions as to whether or not to undertake onwards migration and the directions of onwards migration undertaken. Hukou status is a dummy variable indicating whether or not a migrant holds non-agricultural hukou. We use four dummies to capture respondents belonging to the cohorts born in the pre-1970, 1970–1979, 1980–1989 and post-1990 periods. Four dummies are employed to measure educational level, ranging from elementary school and below, junior high school and senior high school, to college.

Additionally, variables related to migratory experience are included in the models. A dummy is included to reflect whether or not either or both of a migrant’s parents had a migratory experience. Two sets of dummies are used to indicate the period of first migration and the period of onwards migration (i.e. pre-1990, 1990s, 2000s, 2010s). We also consider the effects of the direction of the first migration on subsequent migration. Variables measuring the urban levels of the hukou city and the first migration city are also included in the models. Further, we constructed a continuous variable measuring the difference in urban level between the first migration city and the hukou city: the six city categories are coded as follows: county-level unit = 1, small city = 2, medium city = 3, large city = 4, extra-large city = 5 and super-large city = 6. The value of this variable ranges from −5 to 5, where the negative number indicates downward migration, the positive number indicates upward migration, and 0 implies horizontal migration. The higher the absolute value, the greater the difference in level between the first migration city and the hukou city. For example, if a migrant whose hukou city is a small city (coded 2) moved to an extra-large city (coded 5) in his/her first migration, the value of the variable equals 3. The descriptive statistics of all the variables are presented in Table 1.

Table 1.

Descriptive statistics of characteristics of migrants of different groups.

	Full sample	Staying migrants	Onward migrants
No. of individuals	81,038	62,379	18,659
Age
Age at the survey time (2017)	36.3 (10.3)	36.5 (10.6)	35.7 (9.1)
Age at the first migration	27.8 (9.6)	29.4 (9.7)	22.6 (7.1)
Duration since the first migration	10.1 (7.3)	8.7 (6.7)	14.8 (7.1)
Male	43,555 (53.7%)	32,973 (52.9%)	10,582 (56.7%)
Non-agricultural migrants	16,201 (20.0%)	13,370 (21.4%)	2831 (15.1%)
Educational level
Elementary school and below	14,181 (17.5%)	11,117 (17.8%)	3064 (16.4%)
Junior high school	37,674 (46.5%)	28,234 (45.3%)	9440 (50.6%)
Senior high school	18,160 (22.4%)	13,920 (22.3%)	4240 (22.7%)
College	11,023 (13.6%)	9108 (14.6%)	1915 (10.3%)
Migration alone in first migration	50,403 (62.2%)	41,301 (66.2%)	9102 (48.8%)
Married in first migration	46,361 (57.2%)	40,393 (64.8%)	5968 (32.0%)
Having child(ren) in first migration	27,467 (33.9%)	24,537 (39.3%)	2930 (15.7%)
Urban level of first-migration destination
Counties and equivalent	20,964 (25.9%)	16,916 (27.1%)	4048 (21.7%)
Small cities	5475 (6.8%)	4801 (7.7%)	674 (3.6%)
Medium-size cities	6250 (7.7%)	5152 (8.3%)	1098 (5.9%)
Large cities	25,530 (31.5%)	20,431 (32.8%)	5099 (27.3%)
Extra-large cities	12,368 (15.2%)	8012 (12.8%)	4356 (23.4%)
Super-large cities	10,451 (12.9%)	7067 (11.3%)	3384 (18.1%)
Urban level of origin (hukou registered place)
Counties and equivalent	52,851 (65.2%)	40,035 (64.2%)	12,816 (68.6%)
Small cities	14,922 (18.4%)	11,762 (18.9%)	3160 (16.9%)
Medium-size cities	7143 (8.8%)	5522 (8.9%)	1621 (8.7%)
Large cities	4039 (5.0%)	3397 (5.5%)	642 (3.4%)
Extra-large cities	896 (1.1%)	753 (1.2%)	143 (0.8%)
Super-large cities	1187 (1.5%)	910 (1.5%)	277 (1.5%)

Note: For age, age at the survey time and at the first migration, and for duration since the first migration, the numbers in a cell are in the format of ‘mean (standard deviation)’. For other variables, the numbers are in the format of ‘counts (proportion)’.

We apply the binary logistic regression model to examine the effects of the variables described above on onwards migration. In addition, the multinomial logistic regression model is used to examine the factors that influence the direction of onwards migration. In the multinomial logistic regression model, the dependent variable values for the three possible choices are set as integer values: j = 1 for downward migration, 2 for horizontal migration, and 3 for upward migration. The probability of the ith observation making choice j is $P_{ij}$ , given as

P_{ij} = \Pr (U_{ij} > U_{ik}), \forall k \neq j, j = 1, 2, 3

(1)

where $U_{ij}$ is the utility of observation i when a migrant makes choice j, which takes the form

U_{ij} = β_{j} x'_{i} + ε_{ij}

(2)

where $x'_{i}$ is the vector of the explanatory variables of making choice j. The multinomial logit model is

P_{ij} = P (Y_{i} = j | x_{i}) = \frac{\exp (β_{j} x'_{i})}{\sum_{k = 1}^{J} \exp (β_{k} x'_{i})}

(3)

Similarly, the parameters $β_{j}$ are estimated via maximum likelihood estimation.

Results

Migration patterns within the urban hierarchy

In this section, we delineate migration patterns within the urban hierarchy relative to the ‘transition levels effect’, that is making a transition between urban levels from the first migration to onwards migration. A Sankey diagram shows a panoramic view of the migration flow from various origin levels to the first destination cities and discrepancies in onwards migration patterns (Figure 1a). Using demographic effectiveness (DE) (Plane and Jurjevich, 2009) as the indicator, we present a quantitative illustration of the migratory direction patterns within the urban hierarchy. DE is calculated based on migration flows $m_{ij}$ and $m_{ji}$ between cities classified as any pair of levels i and j of the hierarchy. The indicator is calculated as

D E_{ij} = 100 \times \frac{n_{ij}}{t_{ij}}

(4)

where $n_{ij} = | m_{ij} - m_{ji} |$ and $t_{ij} = m_{ij} + m_{ji}$ refer to the net exchange and the total exchange of migration flows between level i and level j, respectively. $DE$ is, therefore, the value of net migration divided by gross migration between any pair of levels. The value of DE ranges from 0 to 100, where 0 means the flows in both directions are balanced and 100 means all migrants are moving in one direction. Figure 1b and c shows the DE results for first-time and onwards migration.

Figure 1.

Panorama of migration patterns within the urban hierarchy: (a) Sankey diagram of direction and choice of first and onwards migration, (b) demographic effectiveness between urban hierarchical levels of migration flows from hukou cities to first migration cities and (c) demographic effectiveness between urban hierarchical levels of migration flows from first migration cities to current cities of residence.

We found a primarily upward pattern for cumulative and first-time migration and a primarily downward pattern for onwards migration (Figure 1). These results suggest that migrants tend to move upward in the first migration and that a large proportion move from larger to smaller cities when undertaking onwards migration to a new city. Migrants whose first migration destination is a higher-level city than their hometown are much more likely to undertake onwards migration than to remain in the first destination, probably because it is more difficult to settle in higher-level than lower-level cities (Figure 1a). However, there is an exception to this pattern: a small number of migrants relocate from counties or equivalent units to small and medium cities (Figure 1c). Overall, the general pattern of China’s domestic migration is an upward movement in the urban hierarchy in the first migration and a downward movement in onwards migration. Contrary to the step migration model, large cities and above are the most attractive destinations for an initial move without stopovers in smaller cities and towns. Moreover, onwards migration destinations are very diverse with small and medium cities being more attractive than other cities.

Effects on onwards migration decisions

To advance understanding of onwards migration decisions and patterns, we model past migratory experiences in relation to the decision to undertake onwards migration. Specifically, in Model 1, we explore the explanatory roles by a set of individual characteristics and the urban levels of the hukou-registered cities and the first migration cities. Then, in Model 2, we add the transition-levels effect – the difference in urban level between the hukou-registered place and the first migration city. Models 1 and 2 report the results of the baseline models. Next, we add a set of interaction terms to Model 2 to examine the interaction effects between the transition-levels effect and each of three important individual characteristics, that is, hukou status, birth cohort, and educational level, in Models 3, 4 and 5, respectively.

Table 2 presents the results of Models 1–5. In general, the coefficients of the individual characteristics are highly consistent across all the models. The hukou effect persists as agricultural hukou migrants and those with parental migratory experience are most likely to undertake onwards migration. Household arrangement also matters for migration. Migrants who are alone on their first migration are more likely than those who are accompanied to undertake onwards migration. Having at least one child is a deterrent to onwards migration, as the costs of migration for those with a child are greater than for those without. In terms of birth cohort, the younger the migrant, the more likely he/she is to undertake onwards migration. On the one hand, younger migrants are at the peak age of their mobility and expect to maximise benefits by migrating to a new location. On the other hand, as late comers, younger migrants may face more obstacles to settling down than their predecessors. Educational level is another key factor, as well-educated migrants (i.e. senior high school and above) tend to stay in the first migration city. A higher educational level may improve a migrant’s competitiveness in the labour market, thereby facilitating settlement in the first migration city. Furthermore, the period during which first migration took place also has a significant effect on stay-versus-move outcomes. Compared to those who first migrated before or during the 1980s, more recent migrants, those who migrated during the 1990s and 2000s, are more likely to relocate to a new city. The post-1990 group has little reference value, as it is too early to determine the onwards migration propensity of the migrants who belong to it.

Table 2.

Regression results for effects on onwards migration.

	Model 1	Model 2	Model 3	Model 4	Model 5
Intercept	−0.863***	−0.981***	−1.007***	−1.030***	−0.919***
Gender (ref.: female)	0.004	−0.009	−0.010	−0.011	−0.009
Minority	−0.032	−0.070*	−0.067*	−0.065	−0.071*
Parents having migration experiences	0.248***	0.263***	0.263***	0.265***	0.263***
Non-agricultural hukou holder	−0.311***	−0.333***	−0.189***	−0.346***	−0.333***
Married in first migration	0.455***	0.449***	0.448***	0.448***	0.450***
Having child(ren) in first migration	−0.940***	−0.953***	−0.953***	−0.954***	−0.950***
Educational attainment (ref.: elementary school and below)
Junior high school	−0.001	0.002	−0.003	0.018	0.000
Senior high school	−0.158***	−0.160***	−0.167***	−0.068	−0.162***
College	−0.541***	−0.535***	−0.543***	−0.282***	−0.537***
Birth cohort (ref.: pre-1970 cohort)
1970–1979 cohort	0.121***	0.121***	0.124***	0.120***	0.019
1980–1989 cohort	0.284***	0.279***	0.281***	0.277***	0.179***
Post-1990 cohort	0.300***	0.302***	0.304***	0.299***	0.310***
Migration alone in first migration	0.455***	0.449***	0.448***	0.448***	0.450***
Year of first outmigration (ref.: pre-1990)
1990s	0.808***	0.830***	0.830***	0.828***	0.828***
2000s	0.288***	0.295***	0.294***	0.294***	0.294***
2010s	−1.479***	−1.469***	−1.469***	−1.470***	−1.469***
Level of hukou city (ref.: counties and equivalent)
Small cities	−0.157***
Medium-size cities	−0.081*
Large cities	−0.478***
Extra-large cities	−0.471***
Super-large cities	−0.026
Level of first migration city (ref.: counties and equivalent)
Small cities	−0.578***
Medium-size cities	−0.131**
Large cities	0.038
Extra-large cities	0.705***
Super-large cities	0.441***
Difference in urban level between first migration city and hukou city		0.117***	0.131***	0.149***	0.079***
Non-agricultural hukou holders × Difference in urban level between first migration city and hukou city			−0.075***
Junior high school × Difference in urban level between first migration city and hukou city				−0.013
Senior high school × Difference in urban level between first migration city and hukou city				−0.047**
College × Difference in urban level between first migration city and hukou city				−0.121***
1970–1979 cohort × Difference in urban level between first migration city and hukou city					0.057***
1980–1989 cohort × Difference in urban level between first migration city and hukou city					0.055***
Post-1990 cohort × Difference in urban level between first migration city and hukou city					0.006
Observations	81,038	81,038	81,038	81,038	81,038
Log Likelihood	−34870.3	−35247.4	−35229.3	−35217.8	−35234.0
Pseudo R²	0.203	0.194	0.194	0.195	0.194

Note: ***p < 0.001. **p < 0.01. *p < 0.05.

The results of Model 1 indicate that the respective urban levels of the hukou city and the first migration city are associated with onwards migration decisions. Migrants with hukou registered in small, medium, large, or extra-large cities are significantly more likely to stay in the first migration city as compared to those with hukou registered in counties or equivalents and super-large cities. The level of the first migration city is associated with the likelihood of onwards migration: those who first migrated to small and medium cities are more likely to remain, whereas those who first migrated to extra- and super-large cities are more likely to undertake onwards migration. In Model 2, we substantiate the transition-levels effect – the larger the difference between the hukou-registered city and the first migration city, the greater the probability of onwards migration. Movement up the urban hierarchy is a double-edged sword: although they may benefit from diversified labour markets and economies of scale, migrants may face significant difficulties in settling there. The extent to which both the benefits and difficulties come into play differs between migrant groups.

To determine the selectivity of onwards migration, we examined the interaction effects between the transition-levels effect and three individual characteristics (Models 3–5). We found that the transition-levels effect is stronger for agricultural hukou migrants than for non-agricultural hukou migrants (Figure 2a). In terms of educational level, the effect is stronger for less-educated than for better-educated migrants. Notably, for migrants with a college education, the likelihood of onwards migration has little relevance for the change in urban level accruing from first-time migration (Figure 2b). For birth cohort, the effects of difference in urban level between first migration and hukou cities are stronger for the 1970–1979 and 1980–1989 groups than for the pre-1970 and post-1990 groups (Figure 2c). In summary, compared with migrants in the opposite conditions, non-agricultural hukou migrants, better-educated migrants, and those in the pre-1970 and post-1990 cohorts are more capable of overcoming barriers associated with moving upward and settling in the first migration cities.

Figure 2.

Interaction effects between difference in urban level of first migration city and hukou city and (a) hukou status, (b) educational level and (c) birth cohort.

Effects on onwards migration direction

For migrants who did not remain in the first migration city, we developed a multinomial logistic regression model to explain the directions of their onwards migration – downward, horizontal or upward. The results are presented in Table 3. Downward migrants are the reference group for the dependent variable. We report the coefficients and the odd ratios to indicate the likelihood of upward movement and horizontal movement versus downward movement in onwards migration.

Table 3.

Regression results for effects on direction of onwards migration.

	Model 6 (ref.: downward migrants)		Model 7 (ref.: downward migrants)
	Horizontal	Upward	Horizontal	Upward
Intercept	0.376 [0.687]	0.281 [1.325]	−0.477 [0.621]	0.481 [2.303]
Gender (ref.: female)	−0.032 [0.969]	−0.033 [0.968]	−0.025 [0.976]	−0.029 [0.971]
Minority	0.060 [1.062]	−0.125 [0.882]	0.076 [1.079]	−0.125 [0.882]
Parents having migratory experience	0.106 [1.111]	0.141 [1.151]*	0.094 [1.099]	0.140 [1.151]*
Non-agricultural hukou holder	−0.065 [0.937]	−0.306 [0.736]***	−0.089 [0.914]	−0.319 [0.726]***
Married at onwards migration	−0.142 [0.868]*	−0.221 [0.802]***	−0.154 [0.857]*	−0.219 [0.803]***
Having child(ren) at onwards migration	0.154 [1.167]**	0.078 [1.082]	0.167 [1.181]**	0.089 [1.093]
Educational attainment (ref.: elementary school and below)
Junior high school	0.041 [1.042]	0.118 [1.125]	0.042 [1.042]	0.109 [1.115]
Senior high school	0.243 [1.275]**	0.331 [1.392]***	0.250 [1.284]**	0.329 [1.389]***
College	0.301 [1.351]**	0.343 [1.410]***	0.235 [1.264]*	0.315 [1.370]***
Birth cohort (ref.: pre-1970 cohort)
1970–1979 cohort	−0.065 [0.937]	0.016 [1.016]	−0.044 [0.957]	0.021 [1.021]
1980–1989 cohort	0.040 [1.041]	0.172 [1.187]*	0.086 [1.090]	0.190 [1.210]*
Post-1990 cohort	0.172 [1.187]	0.199 [1.221]*	0.225 [1.253]*	0.230 [1.258]*
Year of first migration (ref.: pre-1990)
1990s	0.157 [1.169]	0.134 [1.143]	0.133 [1.142]	0.120 [1.127]
2000s	0.070 [1.072]	0.039 [1.040]	0.038 [1.038]	0.023 [1.023]
2010s	0.061 [1.063]	0.001 [1.001]	0.031 [1.032]	−0.015 [0.985]
Duration in first migration city (year)	0.013 [1.013]	0.013 [1.013]	0.007 [1.007]	0.011 [1.011]
Year of onwards migration (ref.: pre-1990)
1990s	0.735 [2.086]	0.973 [2.645]	0.837 [2.309]	0.743 [2.103]
2000s	0.737 [2.090]	1.011 [2.749]	0.791 [2.206]	0.769 [2.158]
2010s	0.663 [1.940]	1.043 [2.838]	0.720 [2.054]	0.797 [2.219]
Difference in urban level between first migration city and hukou city	−0.620 [0.538]***	−0.977 [0.376]***	−0.581 [0.559]***	−0.929 [0.395]***
Observations	18,659	16,362
Log Likelihood	−14748.3	−14067.5
Pseudo R²	0.217	0.174

Note: ***p < 0.001, **p < 0.01, *p < 0.05. Odds ratio in brackets. In Model 7, we excluded migrants whose first migration was from a lowest-level to a highest-level city or from a highest-level to a lowest-level city.

In general, the results show that hukou status, marital status, children, educational level, birth cohort, first migration direction, and parental migratory experience are significantly associated with onwards migration direction. Migrants with a parent with a migratory experience are 1.141 times more likely to undertake upward rather than downward or horizontal onwards migration. Non-agricultural hukou migrants and married migrants tend to move downward rather than upward, but there is no significant difference between the likelihood of downward and the likelihood of horizontal movement for these groups. Migrants with at least one child tend to move horizontally rather than downward, but show no significant difference in their likelihood of moving downward as compared to upward. Birth cohort is not an important variable for any of the groups in predicting migratory direction with the exception of the 1980–1989 cohort, which showed a significantly higher probability of moving upward than either downward or horizontally. We also found that for migrants who undertake onwards migration, the time of their first migration is not significantly associated with the destination choice of onwards migration.

Of these factors, educational level and the transition-levels effect from the first migration experience exert the most influence on onwards migration. On the one hand, the decision to move up or down the urban hierarchy greatly depends on educational level. Better-educated migrants (senior high school and above) are more likely to move horizontally and upward rather than downward in the urban hierarchy. In other words, migratory direction is positively selected by educational level. Synthesising the results of Models 2 and 6 suggests that better-educated migrants undertake onwards migration only when there are opportunities to move to higher- or same-level cities; otherwise, they are more likely to stay in the first migration city than to move to a new city. In contrast, less-educated migrants tend to undertake onwards migration with a clear preference for the downward direction. It is reasonable to infer that less-educated migrants are negatively selected to undertake onwards migration given the daunting obstacles they face in settling in any new city.

On the other hand, the effect of the difference in urban level between the hukou city and the first migration city − the transition-levels effect − is not negligible in decision-making pertinent to onwards migration. Our results indicate that migrants who settle in large cities and above in their first migration are more likely to move downward than upward in onwards migration. As the difference increases by one, migrants are 0.538 and 0.376 times less likely to move horizontally and upward than to move downward in onwards migration. Obviously, migrants who move to either the lowest-level cities or the highest-level cities cannot move any further downward or upward, respectively. Therefore, to check the robustness of Model 6, we dropped those cases and performed the regression model again (Model 7). The results of Model 7 are highly consistent with those of Model 6 and reinforce our earlier observation that the trajectory of China’s domestic migration does not follow the step migration model. In fact, among the migrants who moved from a low- to a high-level city in the first migration, only a small portion (about 4%) migrated to cities larger than their first migration destinations.

Conclusion and discussion

Onwards migration is an integral part of human mobility. Using nationally representative data for China, we have shown that domestic migrants in China are more likely to relocate down the urban hierarchy if unsettled in their initial destination, where the coexistence of individual- and urban-level forces complicates decision-making over onwards migration. The literature focuses on human capital, life course events, and location-specific capital to explain migration behaviours (Chen and Fan, 2018; Clark, 2013; Clark and Maas, 2015; DaVanzo, 1981). However, we argued that this approach tends to privilege prospective utility vis-à-vis the endowments at status quo without giving due attention to past migratory experiences.

The results of our two-stage modelling on stay-versus-move decisions and then the direction of onwards migration (upward, horizontal, and downward differences) support our view that past migration experiences are surely at the heart of the explanation. We found that the difference in urban level between the hukou-registered city and the first migration city – the transition-levels effect – has important effects on onwards migratory choices. The larger the difference between the urban levels of these cities, the greater the likelihood of migrants undertaking onwards migration – an effect that is particularly strong for less-educated, mid-age (1970s and 1980s birth cohorts) migrants with rural origins. In terms of the onwards migration direction, hukou, marital status, children, birth cohort, and parental migratory experience yielded mixed results without a clear-cut orientation. However, we found educational level and the transition-levels effect to be highly predictive of movement up and down the urban hierarchy: well-educated migrants tend to move upward whereas less-educated migrants tend to move downward. Additionally, upward migration from the hometown to the first migration city is more likely to predict downward than onwards migration.

Our findings support the position that migration is highly selective: female, more-educated, and urban-origin migrants are more likely to arrive at their ‘dream destination’ in their first migration and, therefore, tend not to undertake onwards migration, whereas male, less-educated, and rural migrants with little human capital have a higher propensity for onwards migration but a lower propensity for remaining in and moving onto cities classified as large and above such that they tend to relocate down the urban hierarchy. However, these findings for onwards migration in China contradict findings in intra-European contexts, where the most highly educated and those with more social and cultural resources and legal rights undertake onwards migration (Della Puppa et al., 2021).

Our findings regarding Chinese migratory dynamics offer a basis for rethinking the conventional assumption of a linear association between mobility and migration. Previous studies connect migration with the capacity to move, but tend to privilege human capital and household needs over social context. The patterns of China’s onwards migration reveal an enormous ‘pull’ effect of the most rapidly urbanising cities on absorbing migrants. In that ongoing economic transition and urbanisation processes are polarised towards large cities, human capital plays a minor role, if any, in the mobility of rural migrants: instead, it is the availability of unstable, low-paid manual jobs and the obstacles to staying in the cities classified as large and above that are most determinative. Here, the socio-economic contexts driving changes in mobility (Findlay et al., 2015) are unarguably important in explaining why China’s rural migrants can achieve a ‘great leap forward’ by skipping intervening urban levels in their first migration. In regard to decision-making over whether to stay or move on, market mechanisms cannot speak to this issue alone. For the most deprived migrant labourers who have consistently been subject to uneven access to social welfare, moving down the urban hierarchy means they no longer have to negotiate the obstacles to settling presented by large cities. Our results, therefore, call for serious reconsideration of current migration policies. Beyond the often-criticised institutional constraints imposed by the hukou policy, we have offered solid evidence reflecting the challenges faced by the floating population in terms of pursuing upward mobility, both in spatial and social terms.

This study has several limitations. First, the dataset does not include the permanent migrant population who have obtained local hukou status in the cities to which they have migrated. Thus, we were not able to examine either their onwards migration patterns or the potential impacts of past hukou migration experiences on their subsequent onwards migration patterns. Further studies are needed to incorporate migrations with hukou into the analytical framework to provide a more comprehensive picture of migration patterns within urban China. Second, although we think population size is a comprehensive index for classifying the urban hierarchy, we also call for further research to use alternative indicators to classify urban hierarchy and re-examine onwards migration patterns and directions in China.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This paper was supported by the National Natural Science Foundation of China (Grant No. 72274200,42001197) and the Natural Science Foundation of Guangdong Province (Grant No. 2022A1515011062).

ORCID iDs

Jiejing Wang

Xin Mai

Notes

References

Afridi

Ren

(2015) Social identity and inequality: The impact of China’s hukou system. Journal of Public Economics 123(C): 17–29.

Ahrens

Kelly

Van Liempt

(2016) Free movement? The onward migration of EU citizens born in Somalia, Iran, and Nigeria. Population Space and Place 22(1): 84–98.

Atkins

(2018) On the move, or staying put? An analysis of intrametropolitan residential mobility and ageing in place. Population Space and Place 24(3): e2096.

Bailey

(2009) Population geography: Lifecourse matters. Progress in Human Geography 33(3): 407–418.

Bernard

Bell

Charles-Edwards

(2014) Life-course transitions and the age profile of internal migration. Population and Development Review 40(2): 213–239.

Cao

, et al. (2015) Self-employment and intention of permanent urban settlement: Evidence from a survey of migrants in China’s four major urbanising areas. Urban Studies 52(4): 639–664.

Chan

Wei

(2019) Two systems in one country: The origin, functions, and mechanisms of the rural–urban dual system in China. Eurasian Geography and Economics 60(4): 422–454.

Chen

Fan

(2018) Gender and generational differences in first outward- and first inward-moves: An event-history analysis of rural migrants in China. Environment and Planning A: Economy and Space 50(8): 1646–1669.

Clark

WAV

(2013) Life course events and residential change: Unpacking age effects on the probability of moving. Journal of Population Research 30(4): 319–334.

10.

Clark

WAV

Lisowski

(2017) Prospect theory and the decision to move or stay. Proceedings of the National Academy of Sciences of the United States of America 114(36): E7432–E7440.

11.

Clark

WAV

Maas

(2015) Interpreting migration through the prism of reasons for moves. Population Space and Place 21(1): 54–67.

12.

DaVanzo

(1981) Repeat migration, information costs, and location-specific capital. Population and Environment 4(1): 45–73.

13.

DaVanzo

(1983) Repeat migration in the United States: Who moves back and who moves on? The Review of Economics and Statistics 65(4): 552–559.

14.

De Jong

Brouwer

McCann

(2016) Moving up and down the urban hierarchy: Age-articulated interregional migration flows in the Netherlands. The Annals of Regional Science 57(1): 145–164.

15.

Della Puppa

(2018) Ambivalent mobilities and survival strategies of Moroccan and Bangladeshi families in Italy in times of crisis. Sociology 52(3): 464–479.

16.

Della Puppa

King

(2019) The new ‘Twice Migrants’: Motivations, experiences and disillusionments of Italian-Bangladeshis relocating to London. Journal of Ethnic and Migration Studies 45(11): 1936–1952.

17.

Della Puppa

Montagna

Kofman

(2021) Onward migration and intra-European mobilities: A critical and theoretical overview. International Migration 59(6): 16–28.

18.

Dennett

Stillwell

(2010) Internal migration in Britain, 2000–01, examined through an area classification framework. Population Space and Place 16(6): 517–538.

19.

Fan

(2004) The state, the migrant labor regime, and maiden workers in China. Political Geography 23(3): 283–305.

20.

Fan

(2011) Settlement intention and split households: Findings from a survey of migrants in Beijing’s urban villages. The China Review 11(2): 11–41.

21.

Findlay

McCollum

Coulter

, et al. (2015) New mobilities across the life course: A framework for analysing demographically linked drivers of migration. Population Space and Place 21(4): 390–402.

22.

Fussell

Massey

(2004) The limits to cumulative causation: International migration from Mexican urban areas. Demography 41(1): 151–171.

23.

Geist

McManus

(2008) Geographical mobility over the life course: Motivations and implications. Population Space and Place 14(4): 283–303.

24.

Goodkind

West

(2002) China’s floating population: Definitions, data and recent findings. Urban Studies 39(12): 2237–2250.

25.

Hao

Tang

(2015) Floating or settling down: The effect of rural landholdings on the settlement intention of rural migrants in urban China. Environment and Planning A 47(9): 1979–1999.

26.

Qian

Wang

(2019) Rural migrants in transitional urban China: Marginality, agency and social justice. Journal of Urban Affairs 41(3): 281–281.

27.

Chen

(2011) Circular migration, or permanent stay? Evidence from China’s rural–urban migration. China Economic Review 22(1): 64–74.

28.

Jeffery

Murison

(2011) The temporal, social, spatial, and legal dimensions of return and onward migration. Population Space and Place 17(2): 131–139.

29.

Kahneman

Tversky

(2013) Prospect theory: An analysis of decision under risk. In: MacLean

Ziemba

(eds) Handbook of the Fundamentals of Financial Decision Making: Part I. Singapore: World Scientific, pp.99–127.

30.

King

Newbold

(2007) Onward emigration to the United States by Canadian immigrants between 1995 and 2000. International Migration Review 41(4): 909–929.

31.

Korpi

Clark

WAV

Malmberg

(2011) The urban hierarchy and domestic migration: The interaction of internal migration, disposable income and the cost of living, Sweden 1993–2002. Journal of Economic Geography 11(6): 1051–1077.

32.

Liu

Wang

Chen

(2017) Does formal housing encourage settlement intention of rural migrants in Chinese cities? A structural equation model analysis. Urban Studies 54(8): 1834–1850.

33.

LJC

(2005) Urban administrative restructuring, changing scale relations and local economic development in China. Political Geography 24(4): 477–497.

34.

Mai

Wang

(2022) Situational differences, migratory duration, and social integration of internal migrants in urban China. Cities 125: 103596.

35.

Massey

Arango

Hugo

, et al. (1993) Theories of international migration: A review and appraisal. Population and Development Review 19(3): 431–466.

36.

Mohabir

Jiang

(2017) Chinese floating migrants: Rural–urban migrant labourers’ intentions to stay or return. Habitat International 60: 101–110.

37.

Yeh

AGO

Zhang

, et al. (2022) Moving down the urban hierarchy: Turning point of China’s internal migration caused by age structure and hukou system. Urban Studies 59(7): 1389–1405.

38.

Ortensi

di Belgiojoso

(2018) Moving on? Gender, education, and citizenship as key factors among short-term onward migration planners. Population Space and Place 24(5): e2135.

39.

Paul

(2011) Stepwise international migration: A multistage migration pattern for the aspiring migrant. American Journal of Sociology 116(6): 1842–1886.

40.

Piore

(1979) Birds of Passage. Cambridge and New York: Cambridge University Press.

41.

Plane

Henrie

Perry

(2005) Migration up and down the urban hierarchy and across the life course. Proceedings of the National Academy of Sciences of the United States of America 102(43): 15313–15318.

42.

Plane

Jurjevich

(2009) Ties that no longer bind? The patterns and repercussions of age-articulated migration. The Professional Geographer 61(1): 4–20.

43.

Ravenstein

(1885) The laws of migration. Journal of the Statistical Society of London 48(2): 167–235.

44.

Sjaastad

(1970) The costs and returns of human migration. Journal of Political Economy 70(5, Part 2): 80–93.

45.

Stark

((1991)) The Migration of Labor. Cambridge: Basil Blackwell.

46.

Tian

Cushing

(2016) Inter-city migration in China: A recurrent-event duration analysis of repeat migration. Cambridge Journal of Regions Economy and Society 9(3): 551–569.

47.

Todaro

(1969) A model of labor migration and urban unemployment in less developed countries. The American Economic Review 59(1): 138–148.

48.

Wang

Fan

(2006) Success or failure: Selectivity and reasons of return migration in Sichuan and Anhui, China. Environment and Planning A 38(5): 939–958.

49.

(2012) Neighborhood attachment, social participation, and willingness to stay in China’s low-income communities. Urban Affairs Review 48(4): 547–570.

50.

Xing

Zhang

(2017) The preference for larger cities in China: Evidence from rural-urban migrants. China Economic Review 43: 72–90.

51.

Yap

LYL

(1977) The attraction of cities: A review of the migration literature. Journal of Development Economics 4(3): 239–264.

52.

Zhao

(2002) Causes and consequences of return migration: Recent evidence from China. Journal of Comparative Economics 30(2): 376–394.

53.

Zhu

Chen

(2010) The settlement intention of China’s floating population in the cities: Recent changes and multifaceted individual-level determinants. Population Space and Place 16(4): 253–267.