Transformations in China’s Internal Labor Migration and Hukou System


This paper examines China’s changing internal labor migration patterns between 1990 and 2005 as its household registration (hukou) system evolves. We document a drastic increase in the size of the migrant population, along with significant composition shifts in migrants’ characteristics, and geographic and employment distributions. Recent migrants are on average older, more educated, more likely to be female, more likely to be married, and more likely to have an urban hukou. Regression analysis shows that migration rates increased substantially during this period for all individuals regardless of their education, gender, age, marital or hukou status. By employing a simple migration location choice model, we investigate the relationship between hukou policy and migration behavior. We find that larger and more developed cities are more attractive to migrants but tend to set more stringent hukou restrictions. Rural migrants are significantly more deterred by hukou restrictions relative to urban migrants. These findings suggest that institutional factors, such as the hukou system, are important for understanding the observed patterns in China’s labor migration.


China has experienced a dramatic increase in internal labor migration since the early 1990s. While there were only 18 million migrants in 1990, the size of total migrants increased to 76 million in 2005 (Table 1).Footnote 1 China’s internal migration reflects the country’s fundamental demographic and social transformations and is an important driving force of China’s rapid urbanization and economic growth (Chan and Hu 2003; Bosworth and Collins 2008). It is widely believed that the evolving hukou (household registration) system has had a profound impact on China’s internal labor migration patterns. However, one major difficulty of assessing the impact of changes in the hukou system is that, concurrently with the institutional reform since the 1980s, other fundamental socioeconomic changes have occurred in China. For example, on the supply side, the rural reforms of the late 1970s and early 1980s increased agricultural productivity substantially and generated significant surplus of rural labor (Lin 1992; Zhao 2004). On the demand side, the development of township village enterprises (TVEs) and urban private and informal sectors increased the demand for migrant labor. Thus, a comprehensive assessment of the impact of hukou system reforms and socioeconomic changes on migration behavior remains a challenging research topic.

Table 1 Changes in numbers of migrants (millions), 1990–2005

In this paper, we first briefly describe the evolution of China’s hukou system from its establishment during the central planning era to its reforms since the mid 1980s and discuss the hukou system’s remaining influence today. We identify various government guidelines and regulations, focusing on the timing of policy changes and regional variation. With the policy environment as background, we document China’s changing patterns of internal labor migration between 1990 and 2005, by using micro data from the 1990 and 2000 Chinese censuses and the 2005 1% National Population Survey. More specifically, we first document a drastic increase in the size of migrant population, from 18 million in 1990 to 76 million in 2005, along with significant composition shifts as migrants have become more likely to be female, hold an urban hukou, and migrate for a longer duration. We compare the age, education, marriage, employment, and geographic distributions of migrants from each survey, and analyze the changes in migration patterns by gender, by hukou status, and by migration duration. Using a regression approach, we show that migration rates increased substantially between 1990 and 2005 for all individuals regardless of their education, gender, age, marital or hukou status.

By employing a simple migration location choice model, we investigate the relationship between hukou policy and migration behavior. We find that larger and more developed cities are more attractive to migrants but tend to set more stringent hukou restrictions. Rural migrants are significantly more deterred by hukou restrictions relative to urban migrants. Although our analysis cannot completely separate the effects of China’s hukou system on migration behavior from the effects of other socioeconomic factors, our findings suggest that the hukou system is important for understanding the observed patterns of internal labor migration in China.

Our study differs from the existing literature on China’s labor migration in that we are able to systematically document trends in labor migration between 1990 and 2005 by demographic characteristics and regions with comprehensive microdata. This paper also contributes to the literature that assesses the impact of China’s hukou system. The existing literature either focuses on the effects of the hukou system on income inequality (e.g., Liu 2005; Whalley and Zhang 2007) or on the aggregate implication of the hukou policy on labor allocation inefficiency and industrialization (e.g., Au and Vernon Henderson 2006; Brandt et al. 2008; Ngai et al. 2018). But there exists limited empirical evidence on the relationship between China’s hukou reform and individual migration decisions. Bao et al. (2011) is one of the few exceptions that directly investigates the link between the hukou control and labor migration at provincial level. Unlike Bao et al. (2011), we focus on migration choice at city level rather than at provincial level as there exist substantial variations in economic conditions and hukou restrictions within provinces. We also highlight how different demographic groups might respond differently to local hukou restrictions.

The rest of the paper is organized as follows. Section “China’s Hukou System” provides an overview of China’s hukou system and how it has evolved over time. Section “Descriptive Patterns of China’s Internal Migration” introduces the census data and documents the long run trends in China’s internal labor migration. Section “Location Choice of Migrants and Hukou Policy” investigates the relationship between the observed labor migration patterns and hukou restrictions. The last section presents the concluding remarks.

China’s Hukou System

The Basics of the Hukou System

China’s hukou system was first set up in cities in 1951 and expanded to cover both rural and urban areas in 1955 as a means to monitor population movements. The first set of hukou legislation was promulgated by the National People’s Congress in 1958, which granted state agencies powers in controlling people’s geographical mobility through a system of migration permits and other regulations (Chan and Li 1999). The rationale for establishing the hukou system was to keep most of the population in the countryside, to secure social and political order, and to ensure food provision and industrialization in cities.

Under the hukou system, each person is assigned a hukou status that classifies both hukou type and the residential location attached to the hukou. The hukou registration type is categorized into “agriculture (rural) hukou ” and “nonagriculture (urban) hukou,” a classification which originated from the occupation division in the 1950s. Loosely speaking, individuals in rural villages are born with an “agriculture hukou” whereas those born in cities are assigned a “nonagriculture hukou.” During the centrally planned regime, this dual classification determined one’s entitlements to state-provided goods and services, including rationed food grain. In addition to one’s hukou type, each person is attached to his or her place of hukou registration based on their presumed regular or permanent residence, such as a city, town, or village. Therefore, in addition to the “agriculture” versus “nonagriculture” hukou type, each person is also distinguished by whether he or she has a “local hukou” or “nonlocal hukou” with respect to a specific residential location. The local hukou registration ensures one’s rights to access services in a specific locality. A person’s hukou registration type and location were inherited from his or her mother before 1998, and after a policy change in 1998 children could inherit hukou status from either their mother or their father.

Change of hukou registration status (both type and location) is far from a matter of personal choice. In fact, the key to regulating population movements, especially formal rural-to-urban migration, under the hukou system is to strictly control hukou conversion. For an individual to officially move from a village to a particular town or city, he or she would have to receive approval from the state to convert his or her hukou type from agriculture to nonagriculture (a process known as nongzhuanfei) first, and then change the hukou registration locality. Nongzhuanfei was the more critical and most difficult part in this two-step process. The central government set an annual quota of nongzhuanfei in each locale and stipulated very stringent criteria for obtaining nongzhuanfei. Common nongzhuanfei channels included recruitment by a state-owned enterprise, enrollment in an institution of higher education, promotion to a senior administrative job, displacement due to state land expropriation and demobilization to cities for soldiers. In general, the criteria for nongzhuanfei were designed to serve the needs of the state (Chan 2009).

During the pre-reform period, a person with an agriculture hukou would have serious difficulties living in an urban area without an urban local hukou because employment and the allocation of food, housing, and other necessities were all contingent on urban registration, and “undocumented” migrants (without formal hukou conversion) would be repatriated by the police. As a result, the overall internal migration rate was very low. According to estimates by Duan et al. (2008), the migrant population size in 1982 was only 6.57 million, accounting for merely 0.66% of the total population.

The Hukou System Reform

After the inception of China’s economic reform in the late 1970s, national and local authorities began to loosen restrictions on physical movements of rural labor for work in the cities. The hukou system has undergone important changes over time, and one of the most significant changes is the shift of power of making hukou-related policies from the central government to local governments.

Although rural decollectivization generated a large surplus of rural labor, rural-to-urban migration was tightly controlled until the mid-1980s. In 1984, a major policy reform initiated by the State Council’s “Directive Concerning Issues Related to Peasants Settling Down in Market Towns” (Guanyu nongmin jinru jizhen luohu wenti de tongzhi) allowed farmers to move to (urban) market towns if they could make their own food grain arrangements, and if they were able to either run businesses or be officially employed and have their own accommodation in the market towns (Chan and Li 1999). In 1988, the government further relaxed the control over labor mobility and began to encourage farmers to move towards the coastal cities. These policies were initially responses to the rapid growth of rural surplus labor and the demand for cheap labor in the Special Economic Zones. Despite these policies, rural-to-urban migration remained extremely restrictive. The migration control policy was adjusted between 1989 and 1991 due to the perceived negative effects of migration on transportation, public security and urban employment. Migrants were classified as “blind flows” (mangliu), and the government deemed it necessary to tighten migration controls. In the 1980s, the adjustments to hukou system were overall very limited.

In the 1990s, the central government gradually delegated the power of setting and implementing hukou regulations to local governments and issued policies to ease rural hukou conversions to small towns and small county-level cities. A policy document “Circular on Implementation of Locally-valid Urban Hukou Registration ” (Guanyu shixing dangdi youxiao chengzhen jumin hukou zhidu de tongzhi) issued by the Ministry of Public Security in 1992 endorsed some local governments’ practice to grant “blue-stamp” urban hukou to eligible migrants in small cities and towns and special zones in some big cities.Footnote 2 The blue-stamp hukou program’s specific design and implementation were left to local governments. To acquire a blue-stamp hukou, a migrant would pay a large lump sum fee of “urban infrastructure construction,” which varied greatly across locations. In 1997, the State Council further approved a policy document titled “Pilot Schemes for Reforming Hukou Administration System in Small Towns” (Xiaochengzhen huji guanli zhidu gaige shidian fang’an) that allowed qualified individuals with a rural hukou to receive an urban hukou in 450 pilot towns and small cities. Rural hukou holders who had stable non-agricultural jobs and regular accommodation in those selected towns and small cities and lived there for more than two years were qualified to apply for urban hukou conversion or nongzhuanfei. The reform was restricted to small towns and cities where state-provided welfare was very limited. The introduction of the blue-stamp hukou and the relaxation of hukou control in small towns presented new directions in the management of internal migration by the Chinese government.

A few provinces, including Guangdong, Zhejiang and Jiangsu, began to eliminate nongzhuanfei quotas in some cities and all towns and abolish the distinction between “agriculture” and “nonagriculture” hukou in the early 2000s. The two hukou types were unified into one single category called resident household (jumin) hukou (Chan and Buckingham 2008). It is important to note that the program to eliminate the agriculture and nonagriculture distinction only applies to local population in a number of cities, extending urban benefits to former local agricultural population.Footnote 3 The hukou management was further localized in the 2000s, and many local governments (towns and cities) have gained full discretion to set their own criteria upon which local hukou would be granted. Currently, migrants’ qualifications for local hukou conversion are assessed based on their contributions to investment, tax payment, urban housing purchase, employment status, among others. City governments stipulate their own hukou conversion requirements, and cities generally tend to grant local hukou status to those who are either rich (able to purchase urban housing or make large investments) or have higher skills (usually with a education degree or professional qualification).

The Current State of the Hukou System

In the post-reform era, the most substantive change of the hukou system has been the relaxation of physical controls on migration. These policy changes were at least in part responses to the increasing demand for unskilled labor in China’s export-oriented manufacturing sector and urban informal sector since the 1990s and especially after China’s accession to the World Trade Organization (WTO) in 2001. Having an agriculture or nonlocal hukou no longer directly restricts people to move. In coastal cites specializing in export-processing industries (such as Shenzhen), migrant workers easily account for greater majority of the labor force. Since the mid 1980s, the hukou system has been decentralized and localized. Overall migration control through the hukou system, measured by local hukou conversion requirements, has decreased in most Chinese cites in the 2000s (Zhang and Lu 2018).

Currently, hukou policies tend to be more lenient in small cities relative to large cities. The level and availability of state-provide services (such as education, health care and urban infrastructure) and welfare are highly correlated with the administrative hierarchical rank of a city in China (Cai et al. 2001). Higher-rank (typically also larger) cities generally have higher quantity and quality of services. Thus local hukou in big cities is much more desirable, and they are also the major migration destinations. However, big cities tend to impose much more stringent local hukou qualification threshold.Footnote 4 In recent years, local governments in some large Chinese cities, such as Beijing and Shanghai, are tightening instead of loosening their entry barriers (Zhang and Lu 2018).

Under the current hukou system, it is still a person’s hukou status, rather than where he/she lives and works, that determines his/her access to public service and social welfare. The most important benefit of an agriculture hukou is the land-use rights on the piece of land allocated to the hukou holder. An agriculture hukou is also associated with access to agricultural subsidies and some sort of rural social security programs in recent years (Song 2014). Due to China’s urban-biased development strategy, government provision of social services and social welfare is at a very low level in rural areas. For example, rural public schools are usually of low quality and short of funding and staff, and most rural people have no access to retirement pensions or unemployment benefits. A nonagriculture hukou is associated with access to local public schools, health insurance, retirement pensions, housing subsidies, unemployment benefits and other social welfare and social services in urban areas. Migrants without local hukou often have very limited access to these public service and social welfare. This is especially true for rural migrants as they are less educated and more likely to work in low-paying jobs compared to urban migrants. The benefits associated with nonagriculture hukou is considered much larger than those associated with agriculture hukou in most places.

The hukou system continues to be a major obstacle in preventing migrants from settling in the city, especially in large cities. It divides the Chinese population to segments (rural versus urban, local versus nonlocal) and discriminates based on that. As one of China’s fundamental institutional arrangements, the hukou system still imposes significant restrictions on population movements and social mobility.

Descriptive Patterns of China’s Internal Migration

The Census Data

To examine China’s internal migration patterns, we use micro-level data from the 1990 and 2000 censuses and the 2005 1% National Population Survey (also known as the 2005 mini-census) collected by China’s National Bureau of Statistics.Footnote 5 These data are considered to be the best data sources for identifying migrants in China. They are well suited to describing broad country-wide patterns of internal migration and its changes over time. The censuses and population survey contain basic demographic information, such as an individual’s age, gender, education, and marital status as well as information on employment. Therefore we are able to measure changes in the distribution of migrants at disaggregated levels. More importantly, each survey includes a number of questions on hukou and migration status. In all of the 1990–2005 censuses and mini-census, each person in a household was asked whether they had hukou registration at their current location. Individuals also reported when they moved to their current location, their reason for migration, and whether they had an agriculture or nonagriculture hukou. In this study, a “migrant” is defined as a person who reports living in a county that differs from his/her place of hukou registration. Thus we focus on cross-county non-hukou migration. Each of the three surveys contains data on whether a respondent lives in the county of their hukou registration. Therefore, our definition of migrant is consistent across the surveys.Footnote 6 For all our analysis, we restrict the sample to individuals between 16 and 65 years old.

Trends of Migration Size in China

Internal labor migration has grown rapidly as hukou restrictions are gradually relaxed since the mid 1980s. Table 1 describes the growth of the total number of non-hukou, cross-county migrants between the ages of 16 and 65 from 1990 to 2005, as well as changes in the distributions of migrants by gender, hukou status and migration duration. In 1990, the number of migrants was 18 million, comprising 2.3% of the total population of this age group. The size then increased more than threefold in the ten years from 1990 to 2000 to 58 million, which made up 7.2% of the total population in this age range. The 2005 mini-census shows a continual increase in the number of migrants to 76 million.Footnote 7

The second and third columns of Table 1 show changes in the gender composition of migrants. In 1990, less than 44% of migrants were female. Over time, the number of female migrants grew at a higher rate than that of male migrants. Thus female migrants have grown steadily as a fraction of total migrants. Between 2000 and 2005, about half of the total migrants were female. The increasing labor demand in urban manufacturing and service sectors may have contributed to the rise in female migrants.

In all three surveys, respondents have reported whether they had an agriculture (rural) or nonagriculture (urban) hukou. The fourth and fifth columns of Table 1 show the composition changes of migrants by hukou status between 1990 and 2005. In all survey years, there were substantially more migrants with a rural hukou than migrants with an urban hukou. In 1990, 14.1 million out of the 18 million migrants, or 79% of the total migrants, had rural hukou. Most of the migrants with rural hukou migrated to cities. This is the main reason why China’s internal migration is typically characterized as rural-to-urban migration. However, the number of migrants with an urban hukou has increased significantly over time, from 3.9 million in 1990 to 22.6 million in 2005. By the early 2000s, the percent of urban migrants increased to nearly 30% of the total migrant population. To paint a more complete picture of China’s internal migration, we need to study migrants with both rural and urban hukou. Because hukou status plays an important role in China, we will analyze migration patterns for migrants with rural and urban hukou separately for the rest of our analysis whenever it is possible.

Based on the information on when migrants moved to their current location in the surveys, the next two columns of Table 1 describe the compositions of migrants by migration duration. We differentiate between “long-term” and “short-term” migrants, where “long-term” migrants are defined as migrants who have resided in their current location for more than five years and “short-term” migrant as those who have resided in this location for less than five years. Migrants are often referred as the “floating population” in China, as migration without local hukou is implicitly regarded as temporary. However, along with the rising number of migrants, migrants are also increasingly more likely to stay in their destination locations for longer durations. Less than a quarter of all migrants were long-term migrants in 1990, who stayed in their current location for more than five years, whereas by 2005, the fraction of total migrants who were long-term migrants had expanded to nearly one third.Footnote 8

Changes in Migrants’ Characteristics

Although the general surge in China’s internal migration has been well documented (e.g., Chan 2008; Cai et al. 2008), it is less well known how migrants’ characteristics have changed over time and how migration decisions respond to socioeconomic changes and institutional reforms (e.g., hukou system reform). Next, we conduct a comprehensive analysis of the evolution of migrants’ characteristics on age, education and marital status between 1990 and 2005. We present sample statistics for all migrants and for migrant groups divided by gender, by hukou status, and by migration duration.

Age Distribution

Figure 1 shows how the age distribution of migrants has changed over time. In all three survey years, the largest fraction of migrants were roughly in their late teens and early twenties. Migrants between the ages 16 to 25 comprised 49%, 35% and 32% of total migrants in 1990, 2000 and 2005, respectively. The average age of migrants was 29.4 in 1990 and then increased to 30.7 in 2000 and 32.5 in 2005. Migration at younger ages may generate higher lifetime returns to migration. Individuals’ age at migration can also influence their adaptation process in host region (Friedberg 1992). There is evidence that young migrants can better integrate in the urban labor market and earn higher wages than older migrants (Ge 2018). After 1990, migrants became slightly older, with a larger fraction of migrants in their thirties. In 1990, only 18.7% of migrants were between the ages 31 and 40. This proportion grew to 23.0% in 2000 and to 29.6% in 2005. The fraction of migrants above age 40 grew as well, from 16.1% in 1990, to 17.0% in 2000, and to 21.8% in 2005. Over time older people have become more likely to move, and those who have already moved are more likely to stay longer in destination regions. The overall population ageing may also have contributed to the observed shift in migrants’ age distribution.

Fig. 1

Age distribution of migrants, 1990–2005

Figure 2 shows how the age density of migrants across gender, hukou status and migration duration has changed over time. The top two panels (Panels a and b) present the age densities over time for male and female migrants. In 1990 the age distributions of male and female migrants were very similar, with a large number of both male and female migrants in their late teens and twenties. The age distributions of male and female migrants diverged in 2000 and 2005, with a much larger fraction of females migrating in their twenties compared to male migrants. In 2000 (2005), 40.1% (35.7%) of female migrants were between the ages 16 and 25 as compared to 29.9% (28.4%) of males, and female migrants were on average 1.8 years younger than male migrants (31.6 versus 33.4 years old). A recent study by Kuhn and Shen (2013) shows that employers’ relative preferences for female versus male workers are strongly related to the preferred (young) age in China. The gender-specific age preferences may be related to the observed gender differences in migrants’ age distribution. Both male and female migrants’ age densities have shifted to the right between 1990 and 2005, but male migrants’ average age has grown faster.

Fig. 2

Age distribution of migrants by characteristics, 1990–2005

In the middle two panels (Panels c and d) of Fig. 2 we examine the age distributions of migrants conditional on hukou status. In all three survey years, the age densities of urban migrants are considerably flatter than those for rural migrants. Migrants with both rural and urban hukou are most likely to migrate when they are young adults in their twenties and thirties. However, urban migrants on average are 4 to 6 years older than rural migrants. Older migrants make up a much larger fraction of urban migrants, with migrants over age 40 comprising 31%, 29% and 33% of urban migrants, compared to 12%, 14% and 17% of rural migrants in 1990, 2000 and 2005, respectively. The differences in age distributions of migrants with different hukou status suggest that the hukou system still has a significant influence on migration decisions. For those with an urban hukou, the costs of migrating to a different place and to settle down in a new region may be lower, and they may have better access to jobs with less strict age requirements. Therefore, they are more likely to move at an older age and stay longer in their destination region.

The bottom two panels (Panels e and f) of Fig. 2 compare the age distributions of short-term and long-term migrants. Not surprisingly, short-term migrants are much younger than long-term migrants in all three surveys between 1990 and 2005. Individuals in their late teens make up the largest fraction of short-term migrants. In 1990, long-term migrants were relatively evenly distributed across ages. In 2000 and 2005, long-term migrants were disproportionately in their thirties, with migrants aged 30 to 40 making up 33% and 38% of long-term migrants in 2000 and 2005, respectively, compared to 20% and 26% of short-term migrants in 2000 and 2005.

Education Distribution

Next we analyze how the education distribution of migrants has changed over time. We group individuals into four education groups of “primary school,” “middle school,” “high school,” and “college.” The category “primary school” includes people with primary school and below education; “high school” refers to both regular and vocational high school; and “college” consists of three-year specialized colleges, four-year universities, and post graduate education.

The results are displayed in Table 2. The first column shows the percent of migrants with each level of education in each of the three survey years. For comparison, the final two columns of the table present the corresponding percentages for nonmigrants. Generally speaking, migrants are more educated than nonmigrants with a rural hukou, but less educated than nonmigrants with an urban hukou. Over time, education levels have increased substantially for migrants. In 1990, 38% of migrants had a primary school education or less, and by 2005, this number dropped to 21%. The proportion of migrants with a college education or more has grown rapidly from 2.3% in 1990 to 11.5% in 2005.

Table 2 Changes in education distribution (%), 1990–2005

The second and third columns of Table 2 show the distribution of schooling by gender. Between 1990 and 2005 male migrants were substantially better educated than female migrants. The next two columns compare the education distributions of migrants with a rural hukou and those with an urban hukou. Migrants with a rural hukou have much lower education than migrants with an urban hukou. For example, only 2% of migrants with a rural hukou attended college in 2005, whereas 33% of urban migrants received some college education in the same year. Rural migrants generally are much better educated than nonmigrants with a rural hukou, consistent with previous findings on the interaction between education and migration (e.g., Zhao1999; De Brauw and Giles 2008). In 1990, urban migrants had slightly lower education compared to urban nonmigrants, but since 2000 urban migrants have become increasingly more educated than urban nonmigrants. As we have discussed in the previous section, hukou policy has become more localized and leaning towards more educated migrants over time. The sixth and seventh columns of Table 2 show the education distributions of long-term migrants and short-term migrants. Short-term migrants are slightly more educated than long-term migrants on average as they tend to be younger.

Marital Status Distribution

Marital status and family structure are found to have a significant effect on migration decision and the labor market outcomes of migrants (Mincer 1978; Baker and Benjamin 1997). We analyze how the marital status distribution of migrants has changed from 1990 to 2005.

Table 3 shows the proportions of migrants who had never been married, who were currently married, and who were divorced or widowed in each year. Migrants (column 1) are much more likely to be single than nonmigrants (columns 8 and 9), indicating that migration cost may be higher for married people. For example, migrant children are usually without entitlements for free public education in their destination cities (Chen and Feng 2013). The proportion of married migrants has increased over time, from 61% in 1990 to 69% in 2005. This trend may reflect that the relaxation of hukou control has made family migration easier over time. However, millions of these migrants left their children behind.Footnote 9

Table 3 Changes in marital status distribution (%), 1990–2005

There were stark differences in the marital status distributions between male and female migrants (columns 2 and 3) in 1990. Male migrants were more likely to be single than female migrants, whereas female migrants were more likely to be married, widowed or divorced. The gender differences in marital status are consistent with the fact that women are much more likely than men to migrate for family reasons, and men are more likely to migrate for work.Footnote 10 The gap between men and women has shrunk over time: in 1990 male migrants were 40% more likely as female migrants to be single, whereas in 2000 and 2005 male migrants were only slightly more likely than female migrants to be single.

There are also some differences between the marital status of migrants with a rural hukou and those with an urban hukou (columns 4 and 5). In all years, urban migrants were much more likely to be married, divorced or widowed than rural migrants. Cities may be more accommodating to urban migrants, and thus family migration is less costly for them. Furthermore, long-term migrants are considerably more likely than short-term migrants to be married (columns 6 and 7), as long-term migrants tend to be older and migrants are more likely to settle down if moving with their family.

Trends in Geographic Distribution

Regional disparity in wages is one important driving force of internal migration in China (Cai 1999; Zhu 2002; Ge and Yang 2011). Migrants move across different geographic areas, sometimes traveling thousands of miles. The census and mini-census data allow us to examine the national geographic distributions of migrants over time.

The existing studies on the geographic distribution of China’s migrants typically focus on inter-provincial migration flows (e.g., Liang and Ma 2004; Chan2008). However intra-provincial migration accounted for close to half of total internal migration in 2000 and 2005. We go beyond inter-provincial migration and examine how the geographic distribution of migrants across Chinese sub-provincial regions has changed over time in Fig. 3. Prefecture is an administrative division that ranks below a province and above a county in China’s administrative structure. One challenge to construct prefecture-level migration statistics is that prefecture boundaries have changed over time. In this study, we use consistently defined regions constructed by IPUMS International (Minnesota Population Center 2017) as our geographic unit. Changes in prefecture administrative boundaries over time are handled by combining the affected geographic units to create larger, stable units comprised of two or more prefectures. There are 199 consistently defined regions after aggregation, and they do not cross province boundaries and fully nest within provinces.

Fig. 3

Geographic distribution of migrants, 1990–2005

Figure 3 shows how the number of migrants in each consistently defined Chinese regions evolved between 1990 and 2005. In 1990, most regions had less than 200 thousand migrants. The cities of Shenzhen, Shanghai, Tianjin, and Guangzhou had the highest levels of migration in 1990, but many other coastal cities were not yet major migration destinations. In 2000 and 2005, after China’s accession to WTO and becoming “factory of the world,” along China’s east coast the Pearl River delta and the Lower Yangtze River delta became the prime destinations of migration and many cities in both regions had over one million migrants. Shenzhen, Shanghai, Guangzhou, and Beijing were among the regions with the highest levels of migration these years. In addition, a few major cities or provincial capitals in West and Central China have also become prime destination for migrants. For example, Wuhan, Chengdu, and Kunming had over 500 thousand migrants in 2000 and 2005. Appendix Tables 1112 and 13 provide the lists of top 30 cities with the largest number of migrants between 1990 and 2005.

Figure 4 shows the geographic distribution of migrants by hukou status between 1990 and 2005. The size of urban migrants was much smaller than that of rural migrants across all regions in 1990, but urban migration size increased more relative to rural migration between 1990 and 2005. In terms of spatial distribution, rural migrants are more concentrated in coastal cities, with cities like Wenzhou, Dongguan, and Zhongshan having the highest rural to urban migrant ratio. In contrast, urban migrants seem to prefer mega-cities like Beijing and Shanghai as well as provincial capital cities such as Wuhan, Nanjing and Chengdu. We have also examined the geographic distribution of migrants by other characteristics such as gender and age but found no significant difference.

Fig. 4

Geographic distribution of migrants by Hukou Status, 1990–2005

Trends in Employment Distribution

Most migrants leave their hometowns for the booming cities in search of more lucrative employment. To investigate how migration interacts with employment, we analyze changes in the employment patterns of migrants from 1990 to 2005. We first examine the employment rates of migrants and then examine how the distribution of migrants across industries has changed over time. For comparison, we also look at the employment rates and industry distributions of rural and urban nonmigrants.

Table 4 shows the percentages for migrants and nonmigrants that were employed in each year between 1990 and 2005. In 1990, we classify those who answered “no” to a question on whether the respondent was nonemployed (i.e., being a student, a homemaker, unemployed, or disabled) as being employed. In 2000 and 2005, those who worked for pay for at least one hour in the week before the survey were classified as being employed. Migrants’ average employment rate has declined continuously over time, from 81% in 1990, to 80% in 2000 and 78% in 2005. Migrants’ employment rates were lower than rural nonmigrants’ employment rates, but they were much higher than urban nonmigrants’ employment rates.

Table 4 Changes in employment rate (%), 1990–2005

The second and third columns of Table 4 show that male migrants have much higher employment rates than female migrants. Gender differences in employment rates among migrants are much larger than those among nonmigrants. In addition, migrants with a rural hukou have much higher employment rates than migrants with an urban hukou (Table 4, columns 4 and 5). In particular, more than 80% of rural migrants were employed in all three years, compared to between 63 to 71% of employed urban migrants. Similarly, short-term migrants are more likely to be employed than long-term migrants.

The census have no information on earnings. Hence we use the distribution of migrants across industries to further investigate their labor market outcomes. Table 5 presents the changes in industry distribution of migrants across surveys. Industries are reported in five broad categories: 1) agriculture, fishing and forestry; 2) mining and construction; 3) manufacturing; 4) basic services, and 5) advanced services.Footnote 11 From 1990 to 2005, the proportions of migrants in agriculture, fishing and forestry and mining and construction decreased substantially while the proportion of migrants in the manufacturing and service sectors increased steadily. The fraction of migrants in manufacturing was 29% in 1990 but increased rapidly between 1990 and 2005 as China entered WTO during the time period and manufacturing employment increased dramatically to 46% in 2000 and 42% in 2005. Migrants’ employment share in basic service sectors increased continuously, from 19% in 1990, to 25% in 2000 and 28% in 2005, whereas their employment in advanced service sectors fluctuated and changed little over time. In the meantime, rural nonmigrants’ industry distribution has been relatively stable (column 8). By 2005, almost 80% of rural nonmigrants still concentrated in agriculture. On the other hand, urban nonmigrants have been moving out of manufacturing sector into advanced service sectors since 1990. Over time, an increasingly larger proportion of migrants worked in the manufacturing sector than urban nonmigrants. The changes in industry distribution for migrants and nonmigrants presented in Table 5 suggest that internal migration has played a crucial role in China’s structural transformation.

Table 5 Changes in industry distribution (%), 1990–2005

There exist substantial differences in the industry distributions of male and female migrants (columns 2 and 3). In 1990, female migrants were much more likely than men to be engaged in agriculture, fishing and forestry, with 29% of female migrants compared to 13% of male migrants. As migrants in general have become less likely to be employed in these industries, the difference between men and women has shrunk substantially as well. Male migrants are much more likely to work in the mining and construction industries than female migrants. In addition, migrants with an urban hukou are more likely to work in the service sectors, especially advanced service industries, and less likely to work in manufacturing compared to migrants with a rural hukou (columns 4 and 5). Similarly, long-term migrants are also more likely to work in the service industries and less likely to work in manufacturing compared to short-term migrants (columns 6 and 7). Many jobs in the manufacturing sector, such as some assembly jobs, are so-called “3D jobs” (dangerous, dirty, and demeaning jobs). These jobs are considered inferior and undesirable relative to most service jobs. This suggests that rural migrants and short-term migrants are at a more disadvantaged position in the labor market compared to urban and long-term migrants.

Estimates of Conditional Migration Propensity

The aggregate migration trends reported in Table 1 do not control for changes in migration propensity arising from shifts in the characteristics of the population such as gender, age, education, or marital status. A more informative documentation of the migration patterns would show relative migration propensity changes over time, holding the distribution of individual attributes fixed. Thus we specify the following function:

$$ {{M}_{i}^{t}}={{\beta}_{0}^{t}}+\underset{k}{\sum} {{\beta}_{k}^{t}} {S}_{ik}^{t}+\underset{a}{\sum} {\beta}_{a}^{t} {A}_{ia}^{t}+ {{\beta}_{f}^{t}}{{G}_{i}^{t}}+{{\beta}_{m}^{t}}{{D}_{i}^{t}}+{\beta}_{u}^{t}{{U}_{i}^{t}}+{{\varepsilon}_{i}^{t}}, $$

where \({{M}_{i}^{t}}\) is a dummy variable on migration status for individual i in survey t, with \({{M}_{i}^{t}}\) equal to 1 if individual i is a migrant and equal to 0 otherwise; \({S}_{ik}^{t}\) are dummy variables for schooling levels with k ∈ {midsch, highsch, col} corresponding to middle school, high school and college education; \({A}_{ia}^{t}\) are age dummies; \({{G}_{i}^{t}}\) is a dummy variable for female; \({{D}_{i}^{t}}\) is a dummy variable for being married; and \({{U}_{i}^{t}}\) is a dummy for having an urban hukou. The regression coefficients are allowed to vary by survey year in Eq. 1, thus we can analyze whether individual characteristics have become more or less correlated with migration probability over time.

The regression results for each of the three survey years are displayed in Table 6.Footnote 12 We find that migration rates increased substantially between 1990 and 2005 for all individuals regardless of their education, gender, age, marital or hukou status. The constant term in Eq. 1 is the average migration propensity of the reference group in each year, which refers to single males with a primary school education and rural hukou at age 16. It rose from 2.67% in 1990, to 8.16% in 2000 and 7.93% in 2005. Migration probability (conditional on individual characteristics) increased significantly in the 1990s, but became more stable between 2000 and 2005.

Table 6 Estimates of migration propensity

Table 6 indicates that the influence of education on an individual’s migration decision has grown significantly over time. In 1990, relative to having a primary school or less education, having a middle school education was associated with 1.4 percentage points increase in migration propensity, high school education was associated with 0.8 percentage points increase in migration propensity, and college education was associated with 0.7 percentage points increase in migration propensity. Between 2000 and 2005, those with middle school, high school, and college education were 1.8–2.5, 2.6–3.3, and 1.4–4.9 percentage points more likely to migrate compared to those with primary school education.

The coefficients on the female dummy are negative but very small in size in 1990 and 2005. Females were slightly less likely to migrate than males in 1990 and 2005, but men and women with similar characteristics had similar migration propensities in 2000. Being married is negatively associated with migration probability. Conditional on individual characteristics, urban hukou holders are less likely to migrate compared to rural hukou holders.

Figure 5 plots the estimated coefficients on age dummies, which present the migration propensity over the life-cycle conditional on other characteristics, between 1990 and 2005. All the age coefficients profiles are hump-shaped, indicating that migration propensity is lower for young people below age 20 and older people above 40. In all years, migration rates were highest for people in their 20s. The migration propensity increased dramatically for all ages between 1990 and 2000, whereas the change between 2000 and 2005 was relatively small.

Fig. 5

Estimated coefficients on age dummies

In Table 7, we present estimates of migration propensity by hukou type. We run separate regressions for individuals with an urban hukou and those with a rural hukou while controlling for individual education, age, gender and marital status. For both urban and rural samples, the average migration propensity of the reference group (the constant term) increased substantially between 1990 and 2005, and females and married individuals tend to be less likely to migrate. One notable difference between the estimates from the urban sample and the rural sample is that the positive influence of education on an individual’s migration decision is much larger for rural individuals.

Table 7 Estimates of migration propensity by Hukou type

The estimated time effects, captured by the constant terms in Table 6, and the estimated age effects shown in Fig. 5 may be driven by unobserved cohort effects. We use the following age-period-cohort model to estimate age, time and cohort effects on changes in migration probability over time,

$$ {{M}_{i}^{t}}={{A}_{i}^{t}}+{{C}_{i}^{t}}+{{P}_{i}^{t}}+\beta {{X}_{i}^{t}}+{{e}_{i}^{t}}, $$

where \({{M}_{i}^{t}}\) is a dummy variable on migration status for individual i in survey t, \({{A}_{i}^{t}}\) is a vector of age effects, \({{C}_{i}^{t}}\) is a vector of birth cohorts, and \({{P}_{i}^{t}}\) is a vector of years, and \( {{X}_{i}^{t}} \) is a vector of other control variables including gender, education, marital status and hukou status. Because of the linear dependency of age, period and cohort as calendar year equals age plus birth year, we have an identification problem.

We use the intrinsic estimator (IE) described in Yang et al. (2004) to estimate the model with age, time and cohort effects in Eq. 2.Footnote 13 The estimation results are shown in Fig. 6. In Panel (a), we show the age effects, which are hump-shaped and have a plateau in the mid to late 30s. Note that in each given year migration rates were highest for people in their 20s (Fig. 5). The difference is driven by the fact that young people in each year belong to more recent birth cohorts, who tend to have higher migration propensity. Panel (b) of Fig. 6 shows the cohort effects. There is a clear upward trend in cohort effects for those born after 1955. The cohorts born between 1960 and 1970 have experienced the fastest increase in cohort effects on migration. In Panel (c), we show the year effects, which increase over time, with a slight slowdown after 2000.

Fig. 6

Age, cohort and year effects

Figure 7 shows the age, cohort and year effects on changes in migration probability by hukou status. The age effects for both the urban and rural samples are hump-shaped, but the profile for the urban sample is much flatter. The urban age effects (Panel b) do not drop as fast as the rural age effects (Panel a) after age 40. The overall trends of rural and urban cohort effects are similar in Panels (c) and (d) of Fig. 7, but the increase in cohort effects for the rural sample is faster. There also exists notable difference between the time effects by hukou status (Panels e and f). The rural time effects slowed down after 2000, whereas the urban time effects accelerated.

Fig. 7

Age, cohort and year effects by Hukou status

Location Choice of Migrants and Hukou Policy

The Model

The observed changes in migration behavior over time and by successive cohort are consistent with the anticipated effects of the relaxation of the hukou control, but the fundamental socioeconomic changes that occurred in China could also have made important contribution to China’s labor migration pattern. Therefore, determining whether the hukou policy has an effect on migration behavior is difficult from a simple time-series study. As is described in Section “China’s Hukou System”, the hukou policy has become increasingly more localized, and city government stipulate their own hukou conversion requirements. In this section, we investigate whether city-specific hukou policy at destination affects migration behavior.

We start with a simple location choice model of migrants. Let a potential migrant be indexed by i. Each individual is endowed with a vector of demographic characteristics, which consists of the individual’s hukou status, gender, and age group. Let r, f, o be the indicators for rural hukou, female, and above age 30, respectively; then each individual’s characteristics will be determined by a triple (r, f, o), where r ∈ {0, 1}, f ∈ {0, 1}, and o ∈ {0, 1}.Footnote 14 Each individual can choose to migrate to any of the j ∈ {1, 2, … , J} locations/cities or may choose the outside option of not migrating, j = 0.

Consider an individual i with demographic characteristics (ri, fi, oi). Let Vij be is indirect utility from migration to a location j, and Vij is given by:

$$ V_{ij}=F_{i}\left( I_{j},X_{j}\right) +\varepsilon_{ij}=F^{\left( r_{i},f_{i},o_{i}\right) }\left( I_{j},X_{j}\right) +\varepsilon_{ij}, $$

where Ij measures the strictness of hukou policy in location j, Xj is a vector of characteristics or amenities in location j (for example, the average wage), and εij measures individual i ’s idiosyncratic preference for living in location j. In this specification, we allow the function mapping location js characteristics to mean utility of living in j to vary by the individual’s demographics. For example, one might expect that the importance of hukou restrictions on migration behavior may depend on the individual’s hukou status.

We assume that the idiosyncratic preference shock, εij, is distributed as extreme-value type 1. Therefore, the probability an individual i chooses to live in location j will be given by:

$$ Pr(choice_{i}=j)=\frac{e^{F_{i}\left( I_{j},X_{j}\right) }}{{\sum}_{j^{\prime}= 0}^{J}e^{F_{i}\left( I_{j^{\prime }},X_{j^{\prime }}\right)}}. $$

The total number of individuals of group (r, f, o) who move to location j is given by:

$$ {M}_{j}^{\left( r,f,o\right) }=N^{\left( r,f,o\right)}\frac{e^{F^{\left( r,f,o\right) }\left( I_{j},X_{j}\right) }}{{\sum}_{j^{\prime}= 0}^{J}e^{F^{\left( r,f,o\right) }\left( I_{j^{\prime }},X_{j^{\prime}}\right)}}, $$

where N(r, f, o) is the total number of individuals from group (r, f, o). Taking logs yields

$$ {m}_{j}^{\left( r,f,o\right) }=F^{\left( r,f,o\right) }\left( I_{j},X_{j}\right) +\gamma^{\left( r,f,o\right) } $$

where \({m}_{j}^{\left (r,f,o\right ) }=\log ({M}_{j}^{\left (r,f,o\right ) })\) and \(\gamma ^{\left (r,f,o\right )}=\log (N^{\left (r,f,o\right ) })-\log \left ({{\sum }_{j^{\prime }= 0}^{J}e^{F^{\left (r,f,o\right ) }\left (I_{j^{\prime }},X_{j^{\prime }}\right ) }}\right ) \).

Motivated by this model, we consider an empirical specification of the following form:

$$ \begin{array}{@{}rcl@{}} {m}_{jt}^{\left( r,f,o\right) } &=&\alpha_{0}+\alpha_{1}I_{jt}+\alpha^{r}rI_{jt}+\alpha^{f}fI_{jt}+\alpha^{o}oI_{jt} \\ &&+\beta X_{jt}+\beta^{r}rX_{jt}+\beta^{f}fX_{jt}+\beta^{o}oX_{jt}+\gamma_{t}^{\left( r,f,o\right)}+\epsilon_{jt}^{\left( r,f,o\right)}. \end{array} $$

The dependent variable is the log of the number of migrants with characteristics (r, f, o) in location j at time t. In this model, the geographic distribution of migrants depend on the city-specific hukou policy at time t(Ijt) , and city characteristics (Xjt). We include the interaction terms between Ijt and Xjt and individual characteristics (r, f, o) to allow individuals with different characteristics to respond differently to hukou restrictions and city characteristics. The variable \({\gamma }_{t}^{\left (r,f,o\right ) }\) is a group-specific time trend, and \({\epsilon }_{jt}^{\left (r,f,o\right ) }\) is the error term. Therefore, the parameter α1 measures the correlation between the number of young urban male migrants and the hukou index at the city level. The coefficients on the interaction terms αr, αf, and αo measure the differential effects of local hukou restrictions on individuals with a rural hukou, on females, and on individuals over age 30. Similarly, the coefficients βr, βf, and βo measure the differential effects of city characteristics on individuals with different hukou status, gender, and age.

We use 2000 census and 2005 mini-census to estimate the model in Eq. 3. Our goal is to examine how hukou policy affects migration behavior by exploring the cross-sectional variations in the stringency of hukou control at city level. We use the hukou index for 120 Chinese cities constructed by Zhang and Lu (2018) to measure the stringency of hukou control at migration destination and investigate how the city-level number of migrants respond to the hukou index.Footnote 15 The hukou index used in this study is constructed based on city-level policy documents on requirements for local urban hukou conversion between 2000 and 2014, thus there is no time variation in the hukou index between 2000 and 2005 in our data. Other city-level controls (Xjt) include city average wage, size of non-migrant population, foreign direct investment (FDI) per capita, GDP per capita, and fixed asset investment per capita in each year.

The main challenge to study the effect of hukou policy to migration behavior is that the stringency of hukou control is not set randomly and may be correlated with city characteristics. Larger and more developed cities, where demand for local hukou is the highest, tend to set higher entry barriers. In our city-level data, the hukou index is positively correlated with city average wage, GDP per capital, population size and other amenities. While we control for a wide range of local variables that might affect migration decision in Eq. 3, Ijt may still be correlated with some unobserved city characteristics. Suppose the true model is given by

$$ \begin{array}{@{}rcl@{}} {m}_{jt}^{\left( r,f,o\right)} &=&\alpha_{0}^{\prime}+{\alpha}_{1}^{\prime}I_{jt}+\alpha^{r^{\prime }}rI_{jt}+\alpha^{f^{\prime }}fI_{jt}+\alpha^{o^{\prime}}oI_{jt} \\ &&+\beta^{\prime}X_{jt}\! + \!\beta^{r^{\prime}}rX_{jt}+\!\!\beta^{f^{\prime}}fX_{jt}+\beta^{o^{\prime }}oX_{jt}+{\sigma}_{t}^{\left( r,f,o\right)}\!\!+\eta_{jt}+\!{e}_{jt}^{\left( r,f,o\right)}, \end{array} $$

where ηjt is an omitted city amenity that influences migration choice. If the omitted variable ηjt is positively correlated with the hukou index Ijt such that

$$ \eta_{jt}=a+cI_{jt}+u_{jt}, $$

where c > 0, then our estimates of α1 in Eq. 3 will be upward biased and give an estimate of \({\alpha }_{1}^{\prime }+c\). However, as long as the unobserved factor ηjt influence the migration decisions of all individuals regardless of their demographics, the coefficient estimates on the hukou index interaction terms (αr, αf, αo) will still be unbiased estimates of the true parameters \(\left (\alpha ^{r^{\prime }},\alpha ^{f^{\prime }},\alpha ^{o^{\prime }}\right )\). In this case, the coefficients on the interaction terms recover the differential effects of hukou restrictions on various demographic groups relative to young men with an urban hukou.

Furthermore, we also explore the role that hukou restrictions play on the growth of migrant population, while we control the changes in city characteristics. We estimate the following equation:

$$ \begin{array}{@{}rcl@{}} {\Delta} {m}_{jt}^{\left( r,f,o\right)} &=&\alpha_{0}+\alpha_{1}I_{jt}+\alpha^{r}rI_{jt}+\alpha^{f}fI_{jt}+\alpha^{o}oI_{jt} \\ &&+\beta {\Delta} X_{jt}+\!\beta^{r}r{\Delta} X_{jt}\! + \!\beta^{f}f{\Delta} X_{jt}\! + \!\beta^{o}o{\Delta} X_{jt} + \gamma_{t}^{\left( r,f,o\right) }\! + \epsilon_{jt}^{\left( r,f,o\right)}. \end{array} $$

In this specification, the coefficient α1 measures the relationship between hukou restrictions and changes in migrant population, the coefficients αr, αf, and αo measure the differential effects of hukou restrictions on changes in the number of migrants of different hukou status, gender and age, whereas the coefficients βr, βf, and βo measure the differential effects of changes in city characteristics on migration growth.

Estimation Results

The estimation results of Eq. 3 are reported in Table 8.Footnote 16 The first column shows the results when we do not include any city characteristics. We find that the number of migrants in each city is strongly associated with the hukou index, confirming that the hukou index is correlated with city characteristics that make a location more attractive. In the second column, we include log city average wage and its interactions with demographics as additional controls. As expected, a city’s migration level is strongly correlated with its average wage. The coefficient on hukou index remains positive and statistically significant after including average wage, but much smaller in size. The coefficient on the hukou index interacted with rural migrants is significantly negative and of a significant magnitude. This result suggests that individuals with a rural hukou are relatively more deterred by more stringent hukou restrictions. We find no evidence that hukou restrictions have differential effects on migration patterns by gender or age as indicated by the insignificant coefficients on the gender and age interaction terms.

Table 8 Effects of Hukou on migrant location choice

In the next four columns, we sequentially include log local population, log FDI per capita, log GDP per capital and log fixed asset investment per capita as additional controls. While all these variables have positive effects on migration, the estimated effects of hukou index are not sensitive to the inclusion of these additional controls. Taken together, we find that the cities with more stringent hukou restrictions tend to be more attractive to migrants, even after we control for observed city characteristics. In addition, hukou restrictions generate more deterrence to individuals with a rural hukou relative to those with an urban hukou.

The estimation results of Eq. 4 are presented in Table 9. The first column shows the estimates with no additional controls. The results show that cities with more strict hukou restrictions experience faster migration growth as implied by the statistically positive coefficient on the hukou index. At the same time, the coefficient on the interaction term between the index and rural hukou is strongly negative. We find that more stringent restrictions deter the growth of rural migrant population. We include additional controls on changes in city characteristics in columns 2 to 6 in Table 9, and we find that the effects of hukou on migration growth are not sensitive to the additional controls.

Table 9 Effects of Hukou on migration growth

Concluding Remarks

This paper uses micro-level data to analyze China’s internal labor migration patterns between 1990 and 2005. Migration has increased rapidly since the 1990s. By 2005, the size of China’s cross-county working-age migrant population (without local hukou) has grown to 76 million. Along with the rise in size, the composition of the migrant population has shifted significantly over time. Recent migrants are on average older, more educated, more likely to be female, more likely to be married, and more likely to have an urban hukou. The increasing proportion of urban migrants shows the importance of considering both countryside and cities as places of origins for migrants. Migrants are also increasingly more likely to stay in their destinations for longer period of time and show intentions to settle in the cities permanently instead of being seasonal (or being the “floating population” ). China’s coastal regions attracted the largest number of migrants, although several major cities and provincial capitals in western and central China have also become prime destinations for migrants. Migrants’ average employment rate has declined continuously over time, but the proportion of migrants in the service sectors has increased steadily.

China’s internal migration is closely tied with the reforms of the hukou system. Although various changes and new initiatives have been made to the hukou system, the current system continues to restrain migrants from settling in their destination cities. We find that larger and more developed cities are more attractive to migrants but tend to set more stringent hukou restrictions. Rural migrants are significantly more deterred by hukou restrictions relative to urban migrants.

We must point out that the reforms in the hukou system represent only one set of factors that influence labor migration. The fundamental socioeconomic changes that occurred in China since the 1990s could have made important contribution to China’s labor migration patterns. Further investigation into the interactive relationship between socioeconomic changes and hukou policies is essential to understand China’s changing internal migration patterns and to provide guidance for future policy reforms.


  1. 1.

    All references to migrants in this study include individuals between the ages of 16 and 65, who live in counties other than where they register their hukou.

  2. 2.

    “Blue-stamp” urban hukou is granted by local authorities and is valid only locally. It is distinguished by a blue stamp, as compared to a red stamp that is carried by a formal urban hukou on the hukou registration book.

  3. 3.

    To obtain urban hukou, farmers have to give up their land-use rights, which may provide more financial benefits in an urbanizing region. Critics of this program argue that the government unfairly appropriates farmers’ property during the process.

  4. 4.

    Several recent papers (Sun et al. 2011; Zhang and Tao 2012; Zhang and Lu 2018) compare rules and regulations from local policy documents to analyze the stringency of hukou restrictions across different provinces and cities.

  5. 5.

    We use a 1% sample of the 1990 census, a 0.095% sample of the 2000 census, and a 20% sample of the 2005 population survey. In the 2005 mini-census, certain provinces (such as Guangdong) were oversampled, and therefore the provincial population shares differ from the aggregate population statistics. To correct for this, we reweight the 2005 sample such that the population distribution across provinces after reweighting is consistent with the national and provincial statistics reported in the statistical yearbook for the year of 2005. More specifically, let \({n}_{sample}^{j}\) be province j’s population share implied by the 2005 population survey and \({n}_{agg}^{j}\) be province j’s population share reported in the statistical yearbook, we assign a weight \(w^{j}={n}_{agg}^{j}/{n}_{sample}^{j}\) to all individuals in province j in the 2005 sample. The reweighting does not change the descriptive patterns or empirical results in any significant way.

  6. 6.

    Migration is also commonly defined as the separation of hukou and residence across township boundaries. The 2000 and 2005 surveys kept track of whether a respondent registered his/her hukou in the township and county where they lived, but the 1990 survey only had information on whether a respondent lived in the county of their hukou registration. Therefore cross-county non-hukou migration can be defined most consistently over time. In addition, for those moving within a county or district, especially urban residents, some of them may be just living in a different neighborhood for work or for school and they may have nothing to do with migration.

  7. 7.

    In Appendix Table 10, we show changes in numbers of non-hukou, cross-township migrants over time. In 1990, only cross-county migration information was available, but the number of cross-township migrants was 106 million in 2000 and 119 million in 2005, much higher than the cross-county migration size presented in Table 1.

  8. 8.

    Using a cross-sectional migrant survey data in 2008, Ge (2018) shows that rural-to-urban migrants in the sample on average resided in cities for 8 years.

  9. 9.

    According to a 2013 report from the All-China Women’s Federation (2013), 30 million children were left behind and living in the coutryside without their parents. In the censuses and population survey, only women report the number of children they have. There is also no direct information on whether a respondent lives with his/her spouse or children in the data. We have attempted to analyze whether married migrants were traveling with their spouses by using the information on each respondent’s relation to the household head. As such, we limit this analysis to respondents who list themselves as the household head or the spouse. Furthermore, only a subset of all household members were interviewed for a large fraction of households in 2005, and thus we limit our analysis to households in which all household members were interviewed for the 2005 sample. For this sample of households, the proportion of married migrants living with their spouses has been roughly constant at 85% between 1990 and 2005, and female married migrants are more likely to live with their spouses than male married migrants.

  10. 10.

    The 1990 to 2005 surveys ask migrants about their reasons for coming to their current location. In all three years, most migrants migrate for work related reasons. However female migrants are much more likely than male migrants to move for family reasons such as moving with family or to be with relatives or moving through marriage, with 47%, 22% and 30% of female migrants moved for family reasons in 1990, 2000, and 2005, respectively, compared to 11%, 6% and 11% of men in those three years.

  11. 11.

    We group electricity, gas and water in the manufacturing sector. The basic service consists of wholesale and retail trade, hotels and restaurants, transportation, storage and communications. The advanced service sector consists of financial services and insurance, public administration and defense, real estate and business services, education, health and social work and other services.

  12. 12.

    We should interpret the regression results in Table 6 as correlations between individual characteristics and the propensity to migrate, instead of the causal effects of demographic characteristics on the decision to migrate. There may exist omitted variables that are correlated with both migration decision and individual characteristics, and migration decision may affect individual education and marriage outcomes.

  13. 13.

    More specifically, we use the apc-ie command in stata to estimate the model in Eq. 2 by using the intrinsic estimator method. An alternative method to estimate the age, time and cohort effects is to impose some explicit restrictions on one of the effects. For example, Deaton and Paxson (1994) identify age and cohort effects by imposing the constraint that the year effects must add up to zero and be orthogonal to a time trend. Yang et al. (2004) compares parameter estimates and model fit statistics produced by the two methods.

  14. 14.

    For example, (r, f, o) = (0, 0, 0) corresponds to males with urban hukou and below age 30.

  15. 15.

    Bao et al. (2011) use two alternative measures of hukou restrictions: (1) the fraction of previous migrants who secured a local hukou; and (2) the fraction of previous migrants who secured a local hukou and local employment. We believe our measure is more likely to reflect the actual policy differences across cities, as we do not rely on realized (migration) outcomes that respond to policy differences.

  16. 16.

    All regressions are weighted by city population in 1995. The unweighted results are qualitatively similar.


  1. All-China Women’s Federation (2013) A study of the situation of left-behind children in rural China

  2. Au C-C, Vernon Henderson J (2006) How migration restrictions limit agglomeration and productivity in china. J Dev Econ 80:350–388

    Article  Google Scholar 

  3. Baker M, Benjamin D (1997) The role of the family in immigrants labor-market activity: an evaluation of alternative explanations. Am Econ Rev 87(4):705–727

    Google Scholar 

  4. Bao S, Bodvarsson ÖB, Hou JW, Zhao Y (2011) The regulation of migration in a transition economy: China’s Hukou system. Contemp Econ Pol 29 (4):564–579

    Article  Google Scholar 

  5. Bosworth B, Collins SM (2008) Accounting for growth: comparing China and India. J Econ Perspect 22(1):45–66

    Article  Google Scholar 

  6. Brandt L, Hsieh C-T, Zhu X (2008) Growth and structural transformation in China. In: Brandt L, Rawski TG (eds) China’s great economic transformation. Cambridge University Press

  7. Cai F (1999) Spatial patterns of migration under China’s reform period. Asian Pac Migr J 8(3):313–327

    Article  Google Scholar 

  8. Cai F, Du Y, Wang M (2001) Household registration system and labor market protection. Economic Research (Jingji Yanjiu) 12:41–49

    Google Scholar 

  9. Cai F, Park A, Zhao Y (2008) The Chinese labor market in the reform era. In: Brandt L, Rawski T (eds) China’s great economic transformation. (Cambridge, UK: Cambridge University Press), pp 167–214

  10. Chan KW (2008) Internal labour migration in China: trends, geographical distribution and policies. In: United Nations expert group meeting on population distribution, urbanization, internal migration and development: New York: United Nations, pp 93–122

  11. Chan KW (2009) The Chinese Hukou system at 50. Eurasian Geography and Economics 50(2):197–221

    Article  Google Scholar 

  12. Chan KW, Buckingham W (2008) Is China abolishing the Hukou system. The China Quarterly 195:582–606

    Article  Google Scholar 

  13. Chan KW, Hu Y (2003) Urbanization in China in the 1990s: new definition, different series, and revised trends. China Rev 3(2):49–71

    Google Scholar 

  14. Chan KW, Li Z (1999) The Hukou system and rural-urban migration in china: processes and changes. The China Quarterly 160:818–855

    Article  Google Scholar 

  15. Chen Y, Feng S (2013) Access to public schools and the education of migrant children in China. China Econ Rev 26(75):75–88

    Article  Google Scholar 

  16. De Brauw A, Giles J (2008) Migration opportunity and the educational attainment of youth in rural China. World bank policy research working Paper No. 4526

  17. Deaton A, Paxson C (1994) Saving, growth, and aging in Taiwan, studies in the economics of aging. National bureau of economic research project report series, 331–57, University of Chicago Press: Chicago and London

  18. Duan C, Ke Y, Zhang F, Lu X (2008) Nine trends in changes of floating population in China since the economic reform. Population Research (Ren Kou Yan Jiu,) 32(6):30–43

    Google Scholar 

  19. Friedberg R (1992) The labor market assimilation of immigrants in the United States: the role of age at arrival. Unpublished manuscript. Brown University

  20. Ge S (2018) Assimilation and the wage growth of rural-to-urban migrants in China. Working paper. Virginia Tech.

  21. Ge S, Yang DT (2011) Labor market developments in China: a neoclassical view. China Econ Rev 22(4):611–625

    Article  Google Scholar 

  22. Kuhn P, Shen K (2013) Gender discrimination in job ads: evidence from China. Q J Econ 128(1):287–336

    Article  Google Scholar 

  23. Liang Z, Ma Z (2004) China’s floating population: new evidence from the 2000 census. Popul Dev Rev 30(3):467–488

    Article  Google Scholar 

  24. Lin JY (1992) Rural reforms and agricultural growth in China. Am Econ Rev 82(1):34–51

    Google Scholar 

  25. Liu Z (2005) Institution and inequality: the Hukou system in China. J Comp Econ 33(1):133–157

    Article  Google Scholar 

  26. Mincer J (1978) Family migration decisions. J Polit Econ 86(5):749–773

    Article  Google Scholar 

  27. Minnesota Population Center (2017) Integrated public use microdata series international: version 6.5 [dataset]. Minneapolis: University of Minnesota

  28. Ngai R, Pissarides CA, Wang J (2018) China’s mobility barriers and employment allocations. J Eur Econ Assoc.

  29. Song Y (2014) What should economists know about the current Chinese Hukou system? China Econ Rev 29:200–212

    Article  Google Scholar 

  30. Sun W, Bai C, Xie P (2011) The impact of household registration reform on rural labor mobility in China. Economic Research (Jingji Yanjiu) 1:28–41

    Google Scholar 

  31. Whalley J, Zhang S (2007) A numerical simulation analysis of (Hukou) labour mobility restrictions in China. J Dev Econ 83:392–410

    Article  Google Scholar 

  32. Yang Y, Wenjiang J, Land KC (2004) A methodological comparison of age-period-cohort models: the intrinsic estimator and conventional generalized linear models. Sociol Methodol 34(1):75–110

    Article  Google Scholar 

  33. Zhang J, Lu C (2018) A quantitative analysis on Hukou reform in chinese cities: 2000–2016. Working paper. Southwestern University of Finance and Economics

  34. Zhang L, Tao L (2012) Barriers to the acquisition of urban Hukou in Chinese cities. Environ Plan A 44:2883–2900

    Article  Google Scholar 

  35. Zhao Y (1999) Labor migration and earnings differences: the case of rural China. Econ Dev Cult Chang 47(4):767–782

    Article  Google Scholar 

  36. Zhao Z (2004) Rural-urban migration in China–what do we know and what do we need to know? China Econ Quarterly 3:517–536

    Google Scholar 

  37. Zhu N (2002) The impacts of income gaps on migration decisions in China. China Econ Rev 13:213–230

    Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Suqin Ge.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.



Table 10 Changes in numbers of migrants (millions), 1990–2005
Table 11 Prefectures with the 30 highest numbers of migrants, 1990
Table 12 Prefectures with the 30 highest numbers of migrants, 2000
Table 13 Prefectures with the 30 highest numbers of migrants, 2005

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Colas, M., Ge, S. Transformations in China’s Internal Labor Migration and Hukou System. J Labor Res 40, 296–331 (2019).

Download citation


  • Labor migration
  • Hukou
  • China