Although the two concepts of FP and MOFP are familiar to the public in China and used extensively in academic literature and government documents, they are not commonly used in international migration research. “Floating population”, the English translation of “liudong renkou”, could hardly be found in English language literature before the 1990s. It has appeared more frequently only in the last 20 to 30 years (Solinger, 1995; Chang, 1996; Fan, 1999), and has been used mainly in studies on migration in China, with rare use in studies of migration outside of China. In the textbook Population Geography: Tools and Issues by K. Bruce Newbold, now in its fourth edition and translated into Chinese, the term “the floating population” cannot be found at all (Newbold, 2021); in the International Handbook of Migration and Population Distribution edited by Michael J. White and published by Springer, the term “floating population” is only used in chapters and statements relating to China, with particular emphasis on its close relationship to China’s household registration (or hukou) system (White, 2016).

In international migration literature, which is mainly published in English, the concepts most closely corresponding to FP and MOFP are “migrant” and “migration”, with the latter referring to an event of spatial mobility leading to a change in one’s usual residence across certain administrative boundaries, and the former denoting the actor in such an event (Duan et al., 2006; Bilsborrow, 2016; Song, 2019; Newbold, 2021). Based on these definitions, for a migration event to happen or for a person to be a migrant, two conditions must be met, in temporal and spatial terms respectively: first, there must be a movement across certain administrative boundaries; and second, the person who moves must remain in the destination long enough to effect a change in residence (UNIPI, 1992). Not all spatial movements of human beings meet the above temporal and spatial criteria, and therefore there are concepts of “mobility” and “movers” to capture all spatial movements and their actors (Zelinsky, 1971; Skeldon, 1990; Bilsborrow, 2016). In addition, the word “mobility” is often used in combination with other adjectives to refer to the movements of human beings not meeting the spatial and temporal criteria of migration, an example of which is “residential mobility” (Bilsborrow, 2016; Newbold, 2021). From these concepts, their definitions, and associated measurements, one can see that international migration studies are focused on spatial movements, and the definition and measurement of migration are based on changes in usual residence caused by migration events. In studies on China’s population movements, scholars outside China mostly use the terms “permanent migration” and “temporary migration” (Sun and Fan, 2011), “hukou migration” vs. “non-hukou migration” (Chan, 2009), and “formal migration” vs. “informal migration” (Fan, 1999) to reflect the characteristics of migration in China while maintaining consistency and continuity with international literature in the use of relevant terms.

Based on the above concepts, there are two main ways of measuring migration that are used internationally, which correspond to two types of data in migration research (Figure 1). The first type is event data, which are obtained by using relevant population registers and administrative records to identify and produce statistics on migration events. Some fifty countries possess and use this kind of migration data; they are mainly European (especially Scandinavian) and Asian countries with well-established traditions of civil registration. In some European countries, population registration data (including migration data) is relatively complete, such that in the 2010 census round, eight European countries replaced traditional censuses with register-based censuses (Bell et al., 2015a; Newbold, 2021). The second type is transition data, which is based on changes of residence. Such data identify migration events indirectly by comparing a person’s current place of residence and their place of residence several years ago, and inferring migration information accordingly. Internationally, this is the most common way of measuring migration, and it can be further divided into three categories, with respect to whether it is based on information on place of birth, place of previous residence at a fixed interval (usually five years or one year), or place of last residence. In the 2000 census round, migration information was derived from answers to a question on place of birth in 122 countries, by comparing current residence with place of residence at some previous date in 113 countries, and through a question on last residence and duration in current residence in 55 countries (Bell et al., 2015a). It is noticeable that in the international literature, comparing current residence and place of residence five years ago has become a standard way of measuring migration (Newbold, 2021). In the recent IMAGE project funded by the Australian Research Council, Bell et al. used information on current place of residence and place of residence five years ago to compare migration intensities in 61 countries (Bell et al., 2015b; Zhu et al., 2016). In the mainstream international migration literature, whether based on direct measurement of migration events or indirect measurement of residence transition, the definition of migration and the collection of migration data have never deviated from the purpose of reflecting spatial movements.

At face value, MOFP refers to a population’s events of spatial movement, while FP denotes the actors in such events. They can be identified and confirmed through the events that per se cause spatial changes for the population, or changes in the spatial location of the population (usually place of residence) caused by such events. In the Chinese language, it is not difficult to find the literal use of the term “renkou liudong” (MOFP) to indicate spatial movements of a population, such as MOFP during the spring festival or in urban shopping malls and other public places. In such contexts, the meaning of “liudong” in “renkou liudong” is much the same as “mobility”. However, in China’s unique social and historical context, especially its hukou system, FP and MOFP have unique connotations (Zhu and Lin, 2019), and their identification and measurement are therefore not entirely based on the spatial movements of a population.

As used in China, FP refers to those people who do not have local hukou status in a certain administrative area (usually a township or a town, a subdistrict of a city, or a county-level administrative area), but have lived in this area or left their place of hukou registration for a certain period of time (usually a half year or one year), while MOFP refers to events of spatial movement leading to the separation of these peoples’ places of current residence and their hukou registration (Duan et al., 2006; Song, 2019). According to these definitions, the identification of FP or MOFP is not based on a change in an individual’s residence, but rather on the difference between the individual’s current residence and his or her place of hukou registration, which is commonly referred to as “the separation of residence from hukou registration (SORFHR)” (Song, 2019). All the data on China’s floating population and their mobility used in the last several decades have been conceptualized and used according to the definitions that refer to places of hukou registration. It is in this sense that “liudong renkou (FP)” and “renkou liudong (MOFP)” are phenomena unique to China (Duan et al., 2006; Zhang and Yang, 2013; Song, 2019), with special connotations beyond their literal meanings. It is important to realize this so that problems in the data on FP and MOFP examined below can be better understood¹(1Two other concepts in the literature, namely nongmingong (rural migrants) and wailai renkou (non-natives), are closely related to the concept of the floating population. However, the floating population (FP) is the most commonly used among the three concepts, including its use in census results by the State Statistical Bureau, and it is therefore the focus of our analysis in this article. For more details on the differences and similarities among these three concepts, see Zhang and Yang, 2013 and Song, 2019.).

The results of censuses in China since 1982, and of China’s Migrant Dynamic Surveys (CMDS) conducted between 2009 and 2018, provide rich information on the floating population and its mobility as defined above. Taking the 2020 census as an example (Figure 2), it is possible to assess whether a respondent experienced mobility as defined above for the floating population, and then determine whether he or she is a member of the floating population, by comparing his or her answers to census question C7 (about place of residence at the census time, i.e. zero hour, November 1, 2020) and question 8 (about place of hukou registration). In addition, answers to the three questions C9, C10, and C11 provide further information about the events of spatial movement of the floating population, namely the dates of and reasons for leaving place of hukou registration and the types of places of hukou registration. In the large scale China Migrant Dynamic Monitoring Surveys conducted nation-wide and annually by the National Health Commission during the 2009-2018 period, FP and MOFP are also identified by comparing the respondent’s current place of residence and place of hukou registration to see whether there has been a change, although the temporal and spatial criteria are different from those used in the census; in the CMDS, the criteria are being away from the place of hukou registration for more than one month and crossing the boundaries of city-district, county or county-level-city administrative areas. The data on China’s floating population provided by these two data sets, namely China’s censuses and CMDS, have been the main data sources for national-level studies of migration in China in the last 40 years, especially the last 10 years, and these data have played important roles in migration research and policy making.

To a certain extent, it is reasonable to identify FP and MOFP according to whether a person’s place of residence is separated from his or her place of hukou registration, and this was especially the case in the early years of China’s reform and opening up. First, in that period, most members of the floating population were still in the stage of their initial migration: the events of spatial movement leading to the SORFHR had just occurred, and events of subsequent movement had rarely taken place. In such a context, it was feasible and generally reliable to judge whether a migration event had occurred by looking at whether a person’s place of current residence was consistent with his or her place of hukou registration. Second, in the Chinese language, the phrase “liudong” connotes a temporary or nonpermanent move, and the related change in place of residence is also provisional (Zhang and Yang, 2013). Relatedly, the floating population’s stay outside their places of hukou registration has long been considered an informal and unstable arrangement in research and policy making in China. In addition, the non-permanence of the spatial movements implied by the concept of FP makes it similar to the concept of circular migrants (Zhu, 2004), giving it some international significance. Third, in the period of the planned economy and the initial stage of reform and opening up, hukou status was a main criterion for entitlements to social welfare and social security benefits and various public services, and therefore data on the population with local hukou status were important and basic information; correspondingly, producing statistics on the local “floating population”—people who did not have local hukou status—was of great significance for socioeconomic and development planning of an area (Qiao, 1995). For example, local hukou status has been used in many areas as a criterion for entitlement to purchase a house. Relevance to policymaking is one important reason that the practice of measuring FP and MOFP with reference to hukou registration has continued.

The first problem with the concepts of FP and MOFP as defined above, and with data on SORFHR based on these concepts, lies in the fact that they have gradually become detached from migration events in real life. As mentioned earlier, in the early stage of reform and opening up, judging whether a migration event had taken place by looking at whether one’s current place of residence was the same as place of hukou registration was not only feasible, but also basically reliable. However, as time has passed, it has become less common for this separation to reflect a real migration event. The long-term trend of migration and increasing trend of family migration in China are the first set of factors causing this problem (Zhu et al., 2016). Some members of the “floating population”, who were identified by judging whether their places of residence were separated from their places of hukou registration, may have never moved again after they came to their current destinations. If they have remained in that destination for 5 years or more, they would not be identified as migrants according to international practices, since they have experienced neither any migration event, nor any change of usual residence within 5 years. However, they would still be counted as members of the floating population, or more specifically, the “nonmobile floating population”. As the 2020 census results indicate, members of this nonmobile floating population account for a significant and increasing proportion of the floating population (Table 1).

In the 2020 census, those who had been away from their place of hukou registration for at least five years accounted for 36.89% of the population with a current residence different from their place of hukou registration, a 9 percentage point increase from the 2010 census. Such a situation is particularly prominent among the inter-provincial FP, a population in which the proportion of those who had left their places of hukou registration for at least five years increased faster than it did among the intra-provincial FP. More importantly, as can be seen from Figure 3, the CMDS results show that the proportions of FP in their current places of residence for more than five years and more than ten years both demonstrate a trend of fluctuating increase, and they accounted for 44.37% and 20.34% of FP respectively by 2018. This indirectly confirms the above speculation about the high and increasing proportion of non-mobile members in the FP.

Obviously, including this nonmobile “floating population” in relevant statistics artificially exaggerates the real volume and intensity of China’s migration, as compared to how that migration would be measured by commonly accepted international standards. This is confirmed by comparing inter-provincial FP (as defined in China) to inter-provincial migrants (as defined according to international practice) in terms of volumes and trends. The inter-provincial FP, as judged by the separation of current residence from place of hukou registration, increased rapidly throughout the entire 2000-2020 period (Figure 4), with its volume rising from 42.42 million in 2000 to 124.84 million in 2020, accounting for 8.86% of China’s total population. In contrast, while the volume of inter-provincial migrants, identified by a change in residence as compared to five years ago, increased in the 2000-2010 period, its growth rate decreased and even demonstrated a negative trend in the 2010-2020 period, and its volume stood at only 50.24 million at the end of the period, accounting for 3.56% of China’s total population. Clearly, in the recent three censuses, the inter-provincial floating population identified by the separation of current residence from place of hukou registration not only exaggerates the volume and intensity of China’s migration, but is also misleading as to the direction of changes in volume and intensity. Moreover, as time goes by, the gaps between both the volumes and the intensities defined by these two different criteria will continue to grow, suggesting that the volume of FP identified by SORFHR is increasingly divergent from the volume of people to whom real migration events occur, and that such exaggerations and misleading effects will become increasingly serious.

Identifying FP by SORFHR can lead to another completely different result, namely underestimating the real volume and intensity of migration in China. A major shortcoming of this way of identifying and measuring migration is that it cannot detect migration events subsequent to an initial migration from the place of hukou registration, thereby omitting a large number of migration events meeting commonly accepted international criteria. As mentioned earlier, in the early stage of reform and opening up, when simple initial migration dominated, this was not a prominent issue; however, movements subsequent to the initial migration have become increasingly common (Lin and Zhu, 2015; Tian et al., 2016; Tian et al., 2020), and this problem will become increasingly serious. Among the moves systematically omitted are the urban-urban moves undertaken by a large proportion of the FP whose places of hukou registration are in rural areas. Because these members of the FP still have their hukou registration in rural areas, they continue to be identified as rural-urban FP, while they have actually become urban-urban migrants. Thus, important changes in migration form and direction are obscured.

That 2020 census data underestimates urban-urban migration in China’s overall migration process (as mentioned in the introduction section) is a prominent manifestation of the limits of identifying FP in terms of SORFHR. In light of this, we have redefined the statistical criteria for urban-urban FP, so that the subsequent migrations of those members of the FP who have their hukou registration in rural areas and whose first migration destination was in an urban area are included in statistics on urban-urban migration¹(1The new criteria for urban-urban FP are as follows: ① Those members of FP who have their hukou registration in township or town seats, county seats or cities, and whose current places of residence are in neighborhood committees; ② Those members of FP who have their hukou registration in rural areas and have migrated to at least two cities with the destination of the first one being a city at the county-level or above, and whose current places of residence are associated with neighborhood committees.). Based on the new criteria and using the data from 2017 CMDS, we recalculated the volume of urban-urban migrants and their proportion of the total FP. As can be seen in Table 2, the proportion of urban-urban FP in the total FP in China calculated according to the new criteria is 42.07%, an increase of 23.22 percentage points compared to the result based on the old criteria, which defined urban-urban FP as those who have non-agricultural hukou status and have migrated to an urban area. If we apply this proportion of 42.07% to estimate the volume of urban-urban FP for the 2020 census, the result is 158 million, almost double the original figure of 82 million. This clearly demonstrates that data on the urban-urban FP, which are based on their members’ place and rural-urban category of hukou registration and current place of residence, greatly underestimate the real volume and intensity of urban-urban migration. In addition, identifying FP by SORFHR results in another kind of distortion in the data of FP, because many counties have become districts of cities through administrative changes since the reform and opening up (especially between the 2010 and 2020 censuses), and this has turned many members of the FP into those whose current places of residence and places of hukou registration are different but in the districts of the same city, and they are therefore excluded from FP statistics, following the way of identifying FP demonstrated in Figure 2. This artificially reduces the volume of FP in the original administrative areas of the city.

A third problem is related to the two discussed above: a serious flaw of identifying FP by SORFHR is that data obtained in this way is based on migrant stocks rather than flows; however, what are measured in the international migration literature, whether according to events or residence transitions, are migration flows. Hence the FP data based on SORFHR is not internationally comparable, a problem we elaborate further below. More importantly, such data cannot be used to calculate demographic rates to measure the intensity of migration events. This is because in demographic statistics, the numerator in the calculation of the rate is the number of certain events occurring in a certain period of time (usually one year), which are often represented by the number of people who have experienced these events (e.g., the number of migrants) or the number of people existing as a result of such events (e.g., the number of births), and the denominator is the number of person-years determined by the people who have had the risk of experiencing such events in the same period of time (Rowland, 2003; Song, 2017). In such a context one can see that the calculation of FP based on SORFHR is not conditioned on the period of time, and therefore cannot be used to calculate the demographic rates. The ratio derived by dividing FP by the total population cannot reflect the mobility intensity of China’s population; it only represents the proportion of the population that has a place of current residence different from their place of hukou registration. Therefore, strictly speaking, the volume of FP and its proportion of the total population derived from the 2020 census cannot reflect the volume and intensity of migration. As already demonstrated by previous research, China’s migration intensity is not high by international standards (Bell et al., 2015b; Zhu et al., 2016). Whether China has really entered the “age of migration” and can be said to have a “highly mobile population” still need to be verified by internationally comparable migration rates.

FP data based on SORFHR are not only invalid and unreliable in reflecting migration volume and intensity, but also very likely to mask or distort the current migration directions and their changes. As can be seen from Figure 5a, data based on the SORFHR in the 2020 census suggest that the main places of origin for China’s inter-provincial FP were still Anhui, Henan, Sichuan, Hunan, Jiangxi, Guangxi and Guizhou, which are all conventional sending areas for FP in China. These data further suggest that the places of destination were highly concentrated in southeastern provincial-level regions such as Guangdong, Zhejiang, Shanghai, Jiangsu, Beijing and Fujian and mega cities in these areas, which are all conventional receiving areas for FP in China. However, as demonstrated by Figure 5b, if a different statistical criterion is used—based on the change of current usual residence as compared to five years ago—some noticeable changes in the spatial patterns of China’s migration are immediately visible. This is mainly reflected in the facts that return migration from the eastern to the central and western regions has become a clear trend, with migrant places of destination no longer highly concentrated in the coastal areas. Some central and western provincial-level regions with a strong development momentum, such as Anhui, Sichuan, Hunan and Chongqing, have become important migrant destinations; in contrast, Guangdong, Jiangsu, Shanghai, and Zhejiang have become important places of origin, with some new migration flows, including those from Guangdong to Sichuan, Hunan, Hubei, Guangxi and Jiangxi and those from Jiangsu, Shanghai, and Zhejiang to Anhui. It is important to note that such new migration patterns can only be revealed by comparing migrants’ current place of usual residence and place of usual residence five years ago; if observing the inter-provincial MOFP from the perspective of SORFHR, one will still see the conventional pattern of enormous flows from the central and western regions to the eastern region. In fact, as demonstrated in Table 3, taking the volume of migrants and its proportion of the total based on the change in the current place of usual residence compared to that of five years ago as references, one can see clearly that the FP’s volume and its proportion of the total FP based on SORFHR significantly exaggerate the status of the eastern region, and underestimate the status of the central and western regions, as places of migrant destination. Furthermore, the migrants’ volume and proportion of the eastern region based on the change in the current place of usual residence compared to that five years ago at the 2020 census not only were significantly smaller than those based on SORFHR, but also markedly decreased, compared to the corresponding figures from the 2010 census. This dynamic change, which is of great importance in understanding China’s macro spatial patterns of migration, can be easily overlooked if one only looks at the figures suggesting the continued increase in the volume of FP in the eastern region based on SORFHR.

The above problems of masking or distorting the current migration directions and their changes by using FP data based on SORFHR exist not only at the national level, but also at the level of metropolitan regions. Taking Fuzhou, the core area of Fuzhou metropolitan region as an example (Figure 6), if the criterion based on SORFHR are adopted, Cangshan District and Jinan District were among the main destinations, and Gulou District, Taijiang District and Minhou county were among the main places of origin. However, if the criterion based on the change of current place of usual residence compared to five years ago is adopted, Minhou county in the suburbs becomes the second most attractive destination, receiving not only a large number of migrants from Cangshan District, Jinan District and Gulou District, but also some migrants from neighboring counties or county-level cities, including Fuqing City, Minqing County and Yongtai county. Clearly, the role of Minhou county in the latter situation is more consistent with the reality in the metropolitan area of Fuzhou, reflecting Minhou as a place with the dual functions of absorbing both migrants from the city districts of Fuzhou as a result of suburbanization on the one hand, and migrants from neighboring counties and county-level cities as a result of urbanization on the other hand (Figure 6). Similarly, examining the top 20 migration flows within the metropolitan area of Xiamen-Zhangzhou-Quanzhou based on the criterion of SORFHR, one will find that there are no migration flows from Huli District on Xiamen Island to Jimei District outside the island, and from the districts of Huli, Siming and Haicang in Xiamen Municipality to Longhai City in Zhangzhou Municipality; however, these migration flows are clearly visible under the criterion of change in the current place of residence compared to five years ago (Figure 7).

International incomparability of internal migration data has been a longstanding problem in migration studies. Data on internal migration from different countries, even though they are collected according to the aforementioned common international practices on migration definitions and measurements, are still not directly comparable, because the spatial and temporal criteria of migration, the ways of collecting migration data, and especially the spatial units and the approaches to their sub-division used in producing migration statistics, vary significantly among different countries. Fortunately, the problems concerning spatial criteria, spatial units, and approaches to sub-division have recently been basically resolved, and it is now possible to conduct strict international comparison of internal migration intensities using data from different countries (Bell et al., 2015b; Zhu et al., 2016). However, because members of China’s FP are identified by SORFHR rather than changes in place of usual residence, data on China’s FP and MOFP cannot be used directly in international comparisons, and this significantly restricts their usefulness for improving our understanding of the regularity of internal migration and relevant policy making. Even though there are some similarities between China’s concept of FP and that of circular migrants in other developing countries due to the former’s implicit non-permanent characteristic, the way FP is identified—based on SORFHR—makes it different from circular migrants, which are identified by changes in their places of usual residence. Further, the mobility of FP does not imply multiple round trips between the origin and destination, which is an important feature of circular migrants¹(1For international practices on defining and measuring circular migrants, see United Nations Economic Commission for Europe.). In the meantime, a whole set of concepts and data collection methods for FP and MOFP, which are different from relevant international practices, lead to some further confusions and misunderstandings. One example of a way confusion arises is when migration is defined as a movement accompanied by a hukou change (Duan et al., 2006; Song, 2019), regardless of when the movement occurred. If the move was recent, it is possible that the person has not been in the new location long enough for the move to be considered migration, as normally defined. In contrast, those people who have non-local hukou status, but have lived in an area or have left their place of hukou registration for more than half a year, are included in the statistics for “FP” (Qiao, 1995), though they may no longer be migrating. This way of distinguishing migration and mobility based on the mover’s hukou status is not only inconsistent with relevant international practices, it is also problematic for communication and understanding of scholarship. Terms defined with reference to hukou are often mixed up with terms based on commonly accepted international practices, often resulting in ambiguity in relevant literature.