Examining the impact of inter-provincial migration on environmental health in China

China has a large volume of inter-provincial migrants, accounting for more than 11% of the total population. The economic benefits of inter-provincial migration have been well studied, whereas the health impacts related to environmental factors are generally ignored. In this study, the exposure-response function was used to estimate the health effects and the corresponding economic value was calculated by the adjusted-human-capital and cost-of-illness methods. Considering a scenario without personal protection, inter-provincial migration resulted in a reduction of 6114 premature deaths, 233.4 thousand hospitalization cases, and 1.5 million asthma attacks due to the PM2.5 population-weighted exposure decreasing by 3.94 μg m−3 in 2015. The avoided economic value of these health benefits was 0.02% of the national GDP in 2015. However, two-fifths of inter-provincial migrants whose destinations are areas with heavier pollution suffered greater health losses at the regional level. Additionally, personal protection does not significantly reduce the health loss. Thus, national pollution control measures are required to curb air pollution.


Introduction
Rapid industrialization contributed significantly to serious environmental pollution. The WHO (2016) showed that 92% of the global population lives in areas where the concentration of particulates exceeds the WHO threshold. Most of the particles larger than 10 μm stay in the nose and throat, while the particles smaller than 2 μm can reach the lungs through the nose and throat, and finally deposit in the terminal bronchioles and alveoli. The small particles may also enter the blood circulation system through the alveoli. In addition, compared with PM10 and TSP, PM2.5 has a larger specific surface area and stronger adsorption, which can enrich more toxic substances (Li et al 2016). Therefore, this kind of inhalable particles with a smaller size will cause greater harm to the human body, which causes serious health loss including the incidence of respiratory, cardiovascular, circulatory, and other diseases (Pope et al 2011, Behrooz et al 2019, Nattavudh and Andrew 2020. Around 3 million people die due to outdoor air pollution every year, hence, the benefits of economic development are partially offset by the huge costs in health reduction. Air pollution has become the fourth in the list of factors contributing the most to disease in China , Zhou et al 2019. To disclose the specific effects of air pollution, a growing literature has evaluated the nationwide and regional health effects associated with PM2.5 in China. The total premature mortality related to PM2.5 was 1 331 000 in 2013 , 916 000 in 2014 (Archer-Nicholls et al 2016), and 2.47 million in 2015 (Burnett et al 2018). Xie et al (2019) found that for every 10 μg m −3 increase of PM2.5 concentration, the total mortality increased by 5.67% in the Beijing-Tianjin-Hebei region. In view of the serious effects of air pollution, the Chinese government has implemented some clean air policies including more stringent emission standards, more powerful administrative measures, and the substation of gas and/or electricity for coal in the industrial and residential sectors. Reduction of fossil fuel consumption (Shea et al 2020), the penetration of cleaner technologies in the power sector (Huang et al 2019), the electrification of the vehicle fleet (Pan et al 2019), the clean energy use due to urbanization (Aunan and Wang 2014) have been proved as effective measures for improving health benefit. From the perspective of personal protection, some families or individuals buy a large number of masks and air purification products, however, few studies have analyzed the air pollution effect in the human health of personal protection measures.
Although methods using by the above-mentioned references are very different, the common thing is the prior studies considered the effect of air pollution on all residents at the national or regional level, and studies exploring the health effects of migration remain relatively scarce. Under the Chinese Hukou registration system, migrants are defined as citizens who have resided away from their registered address for more than six months (Qiao and Huang 2013). Although the Hukou registration system has certain restrictions on migration, the regional disparity in income and employment opportunities has attracted a large number of migrants in China, reaching 245 million in 2016 (Migrant Population Service Center 2020). Inter-provincial migrants accounted for 63.5% of the migrants, and 11.25% of the total population. The inter-provincial migrants mainly come from the central and western provinces with a large population and flow into the eastern economically developed areas. The proportion of the working-age population aged 15-59 is greater than 90% and the proportion of children aged 0-14 years and the elderly over 60 years old is on the rise, which of children increased from 2.4% in 2011 to 3.5% in 2016, and of elderly people increased from 1.3% to 3.2%. Inter-provincial migration likely changes the exposure to PM2.5 due to the great difference of regional PM2.5 concentrations level. However, the current studies pay more attention to the economic motivation of migration, including income disparity (Todaro 1977, Oded 1984, regional economic development (Liang andLin 2008, Shen andZhang 2019), social resources (Zhang and Cen 2014), employment (Ma et al 2018), and market potential (Yu and Gao 2018), and the environmental health is less concerned, which can't support the policy-making of migration to enhance welfare.
A limited number of studies have investigated the relationship between air pollution and the health of migrants in China. For example, Aunan and Wang (2014) estimated population exposure change to PM2.5 pollution caused by rural-to-urban migrants and indicated that the net exposure effects of migrants led by positive indoor clean fuel use and negative outdoor working and living in the polluted environment should be further studied. Shen et al (2017) studied the health impact of migration due to the great change of direct residential and transportation energy consumption. Shen et al (2018) evaluated the impacts of rural migrant workers' migration on residential energy use mix pollutant emissions, ambient PM2.5, and subsequent premature deaths across China. Lin et al (2020) provided an assessment of changes in the environmental health burden of different types of migrants to facilitate migration policy development. However, the above studies focus on indoor health effect and urbanization process and ignore the effect of outdoor PM2.5 pollution in different regions. Thus, this paper aims to estimate the health effects of inter-migrants' outdoor exposure to PM2.5 due to the great difference in pollution levels in different regions of China. To our knowledge, this is the first study aimed to answer the following questions: whether are the health effects mitigated or exacerbated at the national level due to the huge inter-migration?Why?What are the economic effects?What are the effects of personal protective measures?
The structure of this paper is as follows: section 2 introduces the research methods; section 3 shows the results; section 4 is the discussion, and section 5 is the main conclusion and policy suggestions.

Methods
The change in health effects from air pollution caused by inter-provincial migration is determined using an exposure-response function, and the economic value of health effects is evaluated using the adjusted-humancapital and cost-of-illness method.
2.1. PM2.5 population-weighted exposure By calculating the change in PM2.5 exposure after migration, we determined whether the inter-provincial migrants benefited from improved air quality or suffered from more serious air pollution. After migration, the population-weighted exposure (PWE) of inter-provincial migrants is: Where, P is the number of inter-provincial migrants; i is the migration origin; j is the migration destination; and ADE proj is the PM2.5 average exposure of the destination province in 2015. PWE P,N is the population-weighted exposure of the inter-provincial migrants in the current residential area in 2015.
The PWE before migration (PWE P,B ) was obtained by the counterfactual assumption that the interprovincial migrants always stayed in the original location in 2015, using equation (2), where ADE proi is the PM2.5 average exposure of the origin province in 2015. The counterfactual assumption is widely used to determine causality by setting conditions opposite to facts. For example, Wang (2016) estimated the health impacts of food safety interventions based on counterfactual selections. Aunan and Wang (2014) used the counterfactual assumption of 100% urbanization and 100% clean fuel. Perez and Garcia-Rendon (2021) studied the integration of non-conventional renewable energy and the spot price of electricity based on a counterfactual analysis for Colombia. We refer to Shen et al (2017) and assume that no migration happened. Here, the counterfactual scenario aims to compare the health effects caused by exposure in different provincial of PM2.5 level.
proi j k n l l k 1 1 365 , * ADE proi/j is the average exposure of PM2.5 in province i or j in 2015; n is the number of cities in the province i or j. TR is the ratio of exposure duration; C l,k is the daily concentration of PM2.5 in each city in 2015. Two scenarios were investigated here. First, a full-exposure scenario assuming exposure was constant over 24 h and TR is 1; and second, the protective scenario assuming that the migrants used personal protective measures, which is equivalent to reducing the exposure time. Hence, in the protected scenario, we reduced the exposure time by certain proportions for different pollution conditions. As shown in table 1, the proportion of exposure time used here refers to our national household haze protection survey from 2019, which included 1129 households. 84.23% of the questionnaires came from areas with severe air pollution, representing the level of family haze protection in China. The purpose of using the survey was to obtain the duration of use of protective measures (including air purifiers and ventilation systems) under different pollution conditions. The exposure duration was calculated by deducting the duration of the use of protective measures from 24 h.
Here, DPWE p is the change of population-weighted exposure after migration.

Health effects of change in migration exposure
To estimate the health effects of changes in migrant's exposure to PM2.5, we use exposure-response functions, which describe the relationship between environmental pollution and the health of the residents in the area. It is widely used in environmental epidemiological studies, such as the causal relationship between air pollution or noise pollution and human health ( Here, the risk of premature death and disease caused by PM2.5 exposure was calculated using this model, which describes the relationship between the change in ambient air quality and the health endpoint of the residents. According to the relevant investigation, the expected duration of migrants in their new destination has increased from 5.4 years in 2000 to 7.9 years in 2005, and further to 10.8 years in 2010 (Chen 2013). Due to longterm employment contracts and the gradual formation of stable social relations in the migration destination, migrant workers rarely change their destinations. Besides, due to the industrial agglomeration, the migration destination of workers with specific skills is also fixed. For example, a worker who has mastered the technology of loading and unloading machinery usually goes to work in the port. For the migrants whose purpose of migration is to receive higher education, they shall stay in the destination for at least three to five years to complete their studies. Therefore, the destination is relatively fixed, and the PM2.5 concentration in the destination had an impact on the health of migrants.
The studies on the E-R relationship have been more abundant in Europe and the United States, where PM2.5 concentrations are low (Apte et al 2015). At present, more studies begin to pay attention to the Asian countries, where PM2.5 concentration is much higher. Du et al (2016) and Yang et al (2020) studied the health impact of ambient PM2.5 in China. Anenberg et al (2019) analyzed particulate matter-attributable mortality in 250 urban cities and highlighted the characteristics of Asian cities. In this study, we obtained E-R coefficient data from existing epidemiological studies, as shown in table 2. Health endpoints include respiratory disease mortality (RM), cardiovascular disease mortality (CM), respiratory disease hospital admission (RHA), cardiovascular disease hospital admission (CHA), and asthma attack (AA).
At a given PM2.5 exposure dose, the attributable cases of premature mortality or disease were calculated, and then the changes in attributable cases due to this change in exposure dose is determined (Aunan and Wang 2014). The changes in attributable cases were assumed to reflect health changes caused by inter-provincial migration.
Here, p and p 0 are the mortality or morbidity of each health endpoint for people exposed to the polluted and clean environments, respectively, where the latter is the baseline rate. In addition, β is the exposure-response coefficient; C is the actual concentration of PM2.5; and C 0 is the baseline concentration of PM2.5, taken as 15 μg/m 3 following China's National Primary Standard for Ambient Air Quality. PWE is used as the actual concentration of air pollution exposure, we got: The number of attributable cases (AC), including deaths, hospital admissions, or asthma attacks caused by risk factors (PM2.5 exposure) were calculated using equation (7), where P is the size of the exposed population, i.e., inter-provincial migrants.
The change in the number of attributable cases after migration ΔAC was calculated using equation (8), where AC N and AC B are the number of attributable cases after and before migration, respectively.

Economic value evaluation
Here, the adjusted-human-capital method and cost-of-illness method were used to evaluate the economic value of health benefits, which was used to analyze the contribution of human capital to economic growth from the perspective of the whole society. Since the adjusted-human-capital method uses the per capita GDP as the value of a single individual's statistical life year, there was no difference in individual value (Li et al 2016).

( ) = DÉL
AC HC 9 ij mt j pc EL ij mt is the economic loss caused by the premature death of the migrants from province i to province j; and HC j pc is the human capital per capita in province j, as defined by equation (10). GDP i pc is the per capita GDP of province j; α is the annual growth rate of per capita GDP; γ is the social discount rate; and t is the average remaining life expectancy of those who died prematurely due to exposure to high PM2.5 concentration. According to the report of the World Bank (2010), we used γ=8%, and t=18 year. The per capita GDP and growth rate of each province were obtained from the China Statistical Yearbook (National Bureau of Statistics of China 2016).
The health benefits calculated by the cost-of-illness method were quantified using the economic loss caused by diseases of the migrants from province i to province j, EL :

Data sources
This study used the national 1% population sample survey data from 2015 (Department of Population and Employment Statistics of China, National Bureau of Statistics), which had the largest sample size, widest coverage, and strongest representation in recent years. The survey takes 0:00 on November 1, 2015, as the standard time point, takes the whole country as the whole, and takes cities (regions) as the sub-population adopting the methods of stratified, two-stage, probability proportion and cluster sampling. According to the Chinese Hukou registration system, migrants are defined as citizens who have resided away from their registered address for more than six months. Therefore, the migration duration of all objects in this study is more than half a year, which will be affected by air pollution in the destination. The sample size of inter-provincial migrants is 1 507 018. Samples are selected according to the random principle, and sampling error is calculated and controlled in advance. Following Aunan and Wang (2014), we assumed that the 1% sample survey fully reflected the Note: β is the increase in mortality or hospitalization rate when PM2.5 concentration increases by 10 μg m −3 ; P 0 is the baseline incidence of each health endpoint.
characteristics of all inter-provincial migrants, and scaled the 1% results to 100% of the migrants. It should be noted that migration did not necessarily occur in 2015, but may occur earlier. The PM2.5 daily concentration data from 360 cities were obtained from the China Air Quality Online Monitoring and Analysis Platform.

Environmental health impact on inter-provincial migrants
The number of total inter-provincial migrants was 150.7 million in 2015. Due to inter-provincial migration, the number of premature deaths related to air pollution decreased by 6 114, hospitalizations decreased by 233.4 thousand, and asthma attacks decreased by 1.5 million. The economic value of these health benefits was about 10.44 billion yuan, accounting for 0.02% of the national GDP in 2015.
As shown in figure 1, among the origin provinces, the largest positive effect on health occurred in migrants from Henan, Hubei, Hunan, Hebei, Chongqing, and Anhui; among the destination provinces, the largest positive effect on health occurred in migrants migrating to Guangdong, Fujian, Zhejiang, Shanghai, Inner Mongolia, and Jiangsu. The severity of harm caused by air pollution to these inter-provincial migrants was greatly reduced after migration, resulting in health benefits.
Among the origin provinces, the largest negative effect on health occurred in migrants from Heilongjiang, Guizhou, Fujian, Yunnan, Inner Mongolia, and Gansu; among the destination provinces, the largest negative effect on health occurred in migrants migrating to Beijing, Tianjin, Hebei, Henan, Shandong, and Hubei. Although these migrants may have obtained economic benefits after migration, they were exposed to increased risk of disease due to the poorer air quality.
The health effect is due to the change of PWE after migration. Figure 2 shows the direction of interprovincial migration. Three-fifths of which migrated from heavily polluted regions such as the central east provinces to areas with relatively mild pollution. Therefore, at the national level, the PWE of inter-provincial migrants was 49.36 μg m −3 after migration, which decreased by 3.94 μg m −3 compared to the counterfactual situation (no migration).
The PM2.5 concentration varies considerably across different regions of China. Due to the concentration of heavily polluting industries and the associated use of fossil fuels, the central east provinces of China are the most seriously affected by PM2.5 pollution, as shown in figure 3. The provinces of Henan, Beijing, Hebei, Tianjin, Shandong, Hubei, Jiangsu, and Anhui are heavily polluted, and 76.94% of the migrants from these provinces moved to regions with lower pollution and hence, reduced their health risks. Among the migrants who moved to more polluted regions, most migrated to Beijing. As the capital city, Beijing attracts numerous migrants, but suffers from severe air pollution, as shown in figure 4.
The main migration origins were concentrated in the central region of China, and the top provinces included Anhui, Henan, Sichuan, Hunan, and Jiangxi. The main migration destinations were concentrated in the southeast coastal provinces and three economic zones, where the top provinces included Guangdong, Zhejiang, Shanghai, Jiangsu, and Beijing. Guangdong attracted the most migrants, accounting for 25% of the total interprovincial migrants. Although most of the motivation for migration from the heavily polluted central provinces    to the lightly polluted southeast coastal provinces was economic, the migrants benefited from improved air quality.

Effect of personal protective measures
Due to the severe haze in China, many people take personal protective measures, and even some schools, hospitals, shopping malls, and other public areas have protective systems in place. For individuals and families, masks are often used for outdoor activities. According to Ali Health Research Center, during red alerts for haze in Beijing, the sales of masks on their retail platform increased significantly (9.3 times the usual quantity). Many families also install air purifiers and ventilation systems. According to the Wind Database, cumulative sales of air purifiers reached 64.57 million from 2013 to 2018. Therefore, we also considered the effect of such personal protective measures.
In this scenario, PM2.5 exposure of these migrants decreased by 3.60 μg m −3 after inter-provincial migration, resulting in a decrease of 5 519 premature deaths, 204.5 thousand hospitalizations, and 1.24 million asthma attacks, corresponding to health benefits worth about 9.10 billion yuan. These effects were slightly lower than the results for the full-exposure scenario, as shown in figure 5. Therefore, even if migrants adopt air pollution protection measures, it is difficult to significantly reduce the health risks of air pollution. According to the Chinese Air Purifier Industry Alliance and Chinese Mechanical Ventilation Industry Alliance, the average selling price of an air purifier is 2 319 yuan, and the average selling price of a ventilation system is 10 750 yuan in 2015. When ongoing expenses, such as the costs of electricity, maintenance, and repair are taken into account, the cost of using protective measures is large, but the resulting health benefits are not significant.

Discussion
To our limited knowledge, this is the first study to disclose the health effects of inter-migrants' outdoor exposure to PM2.5, which is very different from the study about the indoor heath effect of migration. In this study, the adjusted-human-capital and cost-of-illness methods used here to evaluate the economic value of health benefits only considered the direct loss of premature death and disease caused by air pollution. The psychological loss borne by the inter-provincial migrants and the burden on the social medical security system were not considered, resulting in the underestimation of the results.
Among the total inter-provincial migrants, the labor force aged 15-59 accounts for more than 90%, where the main purpose of migration is to seek jobs and do business (Migrant Population Service Management Department of China Family Planning 2018). In general, these migrants obtained both economic and health benefits after migration. However, air pollution is expected to lead to a further decline in the attractiveness of the central provinces to the labor force, which aggravates the 'brain drain' and population loss in these areas. Most of the inter-provincial migrants are young adults, while those who stay in their hometown are often young children and the elderly. Hence, these regions lose a lot of human capital and are facing the pressures of an aging population, which poses huge challenges to their economic development, and is not conducive to China's regional balance.
Considering the main economic motivation for inter-provincial migration, the population flows from lowincome to high-income provinces, as shown in figure 6. High-income provinces can be divided into highpollution group and low-pollution group. The low-pollution group includes Fujian, Guangdong, Zhejiang, and Inner Mongolia. The migrants flowing into these provinces obtain the dual benefits of increased income and improved health. Shanghai, Jiangsu, Beijing, Tianjin, and Shandong with high levels of pollution are among the top ten provinces with the largest inflow of migrants, which may receive higher income and better employment prospects, but are subjected to increased risk of disease caused by air pollution. For low-income provinces, such as Hebei, Henan, and Shaanxi, the brain drain problem is already very serious, and migration motivated by air pollution will aggravate this issue. These provinces will be in a very unfavorable position to attract talented workers. To effectively reduce the environmental health burden of the migrants, achieving a win-win situation with health and the economy is key (Lin et al 2020). For the population migrating to highly polluted provinces, air pollution control should be a focus in their destinations, especially in some high-income provinces which attracted numerous migrants by their economic advantages, such as Shanghai, Jiangsu, Beijing, Tianjin. At the national level, air pollution control should be unified nationwide to avoid health loss caused by regional differences.
The outdoor PM2.5 standard recommended by the US Environmental Protection Agency is 15 μg m −3 and 10 μg m −3 according to the WHO Global Air Quality Guideline. Most urban observation points still fail to meet these standards (Yang et al 2019). In Beijing, the annual average PM2.5 concentration was 80 μg m −3 in 2015. High outdoor PM2.5 concentrations have posed a serious health risk. In the calculation of this paper, it is assumed that exposure in outdoor concentration is constant for 24 h, which may lead to overestimation of the health hazards of air pollution. According to a relevant survey, the average indoor PM concentration in Chinese cities is about 80% of the outdoor concentration (Chen and Zhao 2011), but there is little difference between the results deducting part of the exposure time in the protective measures scenario and that in the full exposure scenario. Therefore, PM2.5 will cause significant health effects even if migrants stay indoors. In this paper, the base concentration of PM2.5 is taken as 15 μg m −3 according to China's National Primary Standard for Ambient Air Quality. When the threshold is set at 35 μg m −3 (China's National Second Standard for Ambient Air Quality), the health benefits caused by migration are reduced. It has little effect on deaths, but has a greater impact on hospitalization cases and asthma attacks (RM: −5.64%, CM: −5.26%, RHA: −19.59%, CHA: −12.72%, AA: −34.30%).
When calculating the health effects, we selected the death rate, the hospitalization rate of respiratory diseases and cardiovascular diseases, and asthma attacks to investigate the health effects caused by the change of PM2.5 average exposure due to migration. The chronic health effects of long-term cumulative exposure to air pollution are usually cohort studies, which require a lot of manpower and material resources and long-term follow-up investigation. The effect of chronic exposure on premature death was not considered in this paper. For the lag effect of the health risk of air pollution, most medical research results are that the lag days are several days. For example, in the study of Shenyang, the lag days of PM2.5 on respiratory diseases is 1 day (Wei 2018); the total risk of air pollutants in Wuhan is the highest when the cumulative lag is 4 days (Liu 2018). Another study showed that PM2.5 exposure significantly increased the risk of hospitalization for cardiovascular disease at 2-5 days and 8-13 days lag (Zhang et al 2019). Since the migrants involved in this study are all those who have migrated for more than half a year, and their residence time is relatively long, which is much longer than lag days, the risk of air pollution at destination to these migrants will be obvious.

Conclusion
This study is the first attempt to provide information for improving the environmental health benefits of inter-provincial migrants by evaluating the impact of air pollution on the health of migrants in 31 provinces of China. At the national level, the overall reduced exposure to PM2.5 pollution of inter-provincial migrants greatly reduced the number of premature deaths and hospitalizations, which was worth millions of yuan to the economy. At the regional level, most of these benefits were a result of the migrant flow from the central region to the southeast coastal region of the country, which is expected to further reduce the attractiveness of the central provinces to the labor force. However, migrants flowing into large cities only obtained economic benefits, while exposing themselves to greater health risks. Personal protective measures did not significantly increase the health benefits of the migrants. Hence, national pollution control measures are required to curb air pollution. According to the results of this study, the following suggestions are given: (a) We recommend strengthening air pollution control in the central provinces to alleviate the brain drain. The central provinces should take measures to improve air quality and promote livability to attract the labor force. Such measures could include: optimizing the industrial and energy structures; rationalizing industrial distribution; popularizing air pollution prevention technology and equipment; and promoting the concept of low-carbon consumption to residents.
(b) High-income provinces focus on improving air quality and reducing the health risk of air pollution. Such provinces have a large population inflow, and the benefits of pollution control measures in the high-income provinces with high population density will be more significant. In particular, for provinces with high pollution concentration, pollution control can effectively reduce the health costs caused by migration.
(c) Household protective equipment, such as air purifiers and ventilation systems, does not significantly reduce health risks, so improving the air quality should be a priority.
The results of this study did not consider intra-provincial migration. The health impact of migration on specific populations can be further analyzed, such as migrant workers and highly educated migrants. In future research, these topics will be considered to extract further knowledge from the available data.