Does Air Pollution Influence the Settlement Intention of the Floating Population in China? Individual Heterogeneity and City Characteristics

Abstract

Serious air pollution has caused widespread concern in Chinese society in recent years. China’s floating population plays an important role in China’s economic development, and the determinants of the floating population’s settlement intentions have attracted increasing attention. Using the 2017 China Migrants Dynamic Survey Data (CMDS) and the satellite grid data of global PM2.5 concentration as well as city-level data, this study investigated the influences of air quality on migrants’ settlement intention, in particular, individual heterogeneity and city characteristics. Using an instrumental variable to correct for endogeneity, we found that air pollution has a significant negative effect on the settlement intentions of China’s floating population. Migrants who were older, better educated and with poorer health are more sensitive to air pollution with regard to settlement intention. Meanwhile, settlement intentions are also influenced by individual adaptability: Respondents with better air quality in their hometown are more sensitive to air pollution. Poor air pollution has not weakened the attractiveness of Tier-1 cities to the floating population, nor has the administrative level of a city.

1. Introduction

In recent years, China has experienced rapid urbanization, with the urbanization level increased from 49.68% in 2010 to 63.89% in 2020 (China National Bureau of Statistics, 2021). In 2020, the number of floating population in the mainland reached 375.82 million, which rose from 221.43 million in 2010, an increase of 69.73% (National Bureau of Statistics of China, 2021). At the same time, with a series of traditional barriers restricting population mobility have been broken, attracting and retaining floating population has become an important way to improve regional competitiveness. As an important contributor to China’s rapid urbanization and economic development, how to design and implement appropriate policy measures to help the floating population settle down in cities is an urgent issue faced by policymakers and scholars.

For more than a decade now, the research results on the floating population’s settlement intention have increased rapidly. To the best of our knowledge, Zhu (2007) was one of the first scholars to pay attention to this subject. In a questionnaire survey based on Fujian Province in 2002, he found that only about 20.6% of floating population plan to permanently settle in cities [1]. Four years later, this figure increased to 35.8% in Fujian Province [2]. Prior research mainly takes the institutional, demographic, culture, social and economic factors as the influencing factors in floating population’s settlement intention [3,4,5,6,7]. However, air quality, as an important factor of urban livability characteristics, is often neglected in previous studies.

Dramatic economic development over the past few decades in China has also brought about serious environmental problems, especially air pollution. Serious air pollution has brought great pressure on people’s life and health [8,9]. On one hand, air pollution is directly harmful to physical health, easily causing respiratory diseases such as pneumonia. At the same time, it will also increase the death rate of hypertension, cardiovascular diseases [10,11]. Air pollution will also affect the weight of newborn babies. If pregnant women are overexposed to air pollution, they may face problems such as premature delivery and low birth weight [12,13]. On the other hand, air pollution will bring adverse effects on people’s mental health. Air pollution increases psychological pressure and makes people more prone to depression resulting in a lower level of life satisfaction and mental well-being [14,15]. The literature also shows that air pollution has a negative impact on fertility intention in China [16]. Concerns about the health of their own and their families may affect the settlement intention of migrants to settle in cities, and make them move to cities with better air quality. Therefore, air pollution has become an important factor in determining their intention to settle down.

With the improvement of people’s demand for environmental quality, taking environmental factors into account in the analysis has important theoretical significance for accurately understanding the behavior of floating population’s settlement intention. So, does air pollution affect the floating population’s settlement intention in cities? Furthermore, are there group differences in this effect? How does air pollution affect the settlement intention of floating population in different types cities? In order to answer the above questions, this study, based on the data from the 2017 China Migration Dynamics Survey (CMDS), investigated the impact of air quality on the settlement intention of the floating population’s, especially individual and urban characteristics.

Compared with previous studies, the contributions of this study are as follow, first, in terms of data, PM2.5 data is obtained by matching remote sensing satellite data and county-level maps, which is more accurate and objective. Second, in terms of the research design, this paper analyzes the impact of air pollution on the settlement intention of the floating population, and uses instrumental variables to verify the research results. Third, this paper also examines and discusses the differential impacts of different individuals and different types of cities. The effect of air pollution on the settlement intention of the floating population will be influenced by individual health status, education level, occupation type, and other factors, especially individual adaptability, which previous studies did not include. In addition, the settlement intention of the floating population is also affected by regional characteristics of place of destination. Such as the city types, economic development level and public services of the destination areas, which have not received enough attention by previous studies.

2. Literature Review and Research Hypotheses

2.1. Air Pollution and Migrants’ Settlement Intentions

As an important component of urban environmental quality, the impact of air pollution on migrants has been widely concerned, and studies have found that people do "vote with their feet" on air quality [17]. Some studies have shown that air pollution can cause respiratory diseases such as bronchitis, pneumonia, and asthma, and cause people with a history of respiratory diseases to get sick or relapse [18]. Air pollution reduces the supply of labor [19], while the removal of smog can result in population inflow [20]. Air pollution is one of the important reasons for China’s population migration. A research by Hua & Liao (2016) found that after controlling other factors, the rural–urban migration is strong negatively association with urban air quality [21]. Qin et al. (2018) used Baidu search of the keyword "emigration" to study the impact of air pollution on people’s interest in immigration. It was found that the searches on “emigration” would increase by approximately 2.3–4.8% if today’s air quality index (AQI) is increased by 100 points [22]. Based on the data of overseas students in Chinese, Li et al. (2019) found that international students in China are highly sensitive to urban air pollution in China. For 1% increase of the average annual air pollution index, the expected number of international students would decrease by 3.85% [23]. Further, based on the Global Annual PM2.5 Grids from satellite data and National Migrant Dynamics Monitoring Survey from 2011 to 2015, Sun et al. (2019) reveal that air pollution has a significant negative impact on the employment location of floating population. For every increase by one quartile deviation of the concentrations of PM2.5 would lead to a 0.39 percentage point decrease in the probability of migrants moving to the city [24]. Li et al. (2020) found that settlement decisions of Chinese migrants involved a trade-off between income and air quality, poor air quality will significantly decrease the Chinese floating population’s settlement intention, while a higher income will significantly increase their settlement intention [25]. Based on the research reviewed above, we propose Hypothesis 1 as below:

Hypothesis 1 (H1).

Air pollution has a negative effect on the settlement intention of the floating population.

2.2. Individual Characteristics and Migrants’ Settlement Intentions

The literature show that the floating population’s settlement intention is closely related to demographic factors such as age, gender, marital status, and education [26,27]. Generally, young floating population are more able to adapt to the challenges of working and living in cities, and they are more likely to settle in cities [28,29]. In terms of education, the floating population with higher education are more inclined to settle in cities due to they have more knowledge and skills to live in cities [30]. The research of Yao et al. shows that the air pollution perception has a significant impact on young talent urban settlement intentions [31]. Others also show that unmarried and female migrants are more likely to settle in cities, possibly because they have less need to take care of family members left behind in their hometowns, while female migrants may have more employment opportunities in the service industry which provides more stable employment [32].

In terms of the economic and housing conditions of the floating population, this type of research is mainly based on neo-classical economic theory, emphasizing the economic incentive behind the migrant settlement decisions [33]. A research by Cao et al. (2015) found that self-employed migrant workers have a stronger settlement intention [26]. Recently, more and more attention has been paid to the influence of the floating population’s housing status on their settlement intention in cities [34]. The research results show that the better housing conditions of the floating population in the place of destination, especially property ownership, have a significant positive impact on their settlement intention in cities [35,36].

Migration characteristics have also been proved to have an impact on the settlement intentions of the floating population. Migration duration and geographic range are also important factors affecting their settlement intention. Related studies have shown that the longer migrant is on the move, the stronger their settlement intention [37,38]. Compared to the inter-provincial floating population, the settlement intention of intra-provincial floating population is stronger [39]. The migration of family members also has an impact on the migrants’ settlement intention, and research has suggested that parents with migrant children are more likely than those without to settle permanently in the destination cities [40].

In recent years, with the gradual reform of the hukou system, scholars recognized that settlement intention of the Chinese floating population involves more complicated factors over hukou regulation. There is no doubt that the hukou system has a significant impact on the floating population’s settlement intention, but this is only one of the many factors [41,42]. In addition, the hukou system cannot be neglected. Due to the degree of air pollution in their hometown (hukou registration area), the floating population from different regions will have different feelings about the sensitivity of the air quality in the destination area. Accordingly, this study hypothesizes the following:

Hypothesis 2 (H2).

There are individual differences in the impact of air pollution on the settlement intentions of floating population, and those from hukou registration places with better air quality are more sensitive to air pollution.

2.3. City Characteristics and Migrants’ Settlement Intentions

Some scholars have found that the administrative level of the destination city will affect the settlement intention of the floating population. Those cities with high administrative levels, or large populations are more likely to attract migrants to settle [43]. Compared with developing regions, the developed areas are more attractive to the floating population, and their settlement intention is stronger [44], whether it is to make money or earn a living [45]. The study by Hao Pu et al. also verified the above view. They found that the willingness to settle by the floating population in southern Jiangsu with a higher level of socioeconomic development was higher than that in northern Jiangsu with a lower level of socioeconomic development [46]. The spatial location of cities also has a significant impact on the settlement intention of the floating population, and show obvious regional differences [47].

Regarding urban public services, some studies have found that floating population’ settlement intention is positively associated with urban social welfare schemes [48,49]. Liu et al. explored the impact of urban public services on the willingness of permanent migration, using data from the China Migrants Dynamic Survey, and they found that urban public services significantly affect the permanent settlement intention of floating population. The greater the public services in a city, the stronger the permanent settlement intention of the floating population [50]. Yu et al. found that urban medical resources have a significant positive impact on the settlement intention of the floating population, and the floating population is more inclined to settle in cities with rich medical resources [51]. Song et al. found that China’s migrant workers preferred to permanently settle in relatively large cities, rather than in first-tier megacities such as Beijing and Shanghai. There was an inverted U-shaped relationship between the floating population’s willingness to permanently settle and the size of China’s cities [52]. Hypothesis 3 is therefore proposed:

Hypothesis 3 (H3).

Air pollution weakens the settlement intentions of floating population to settle in first-tier cities.

3. Data and Methods

3.1. Data Source

Based on three datasets from multiple sources, this study formulated a ‘air quality-migrant population matching dataset’ to answer the core question of how air quality influence the resident intention of the floating population. The first dataset is the 2017 China Migrants Dynamic Survey (CMDS) data. The survey is a large-scale survey conducted by the China Population and Development Research Centre under the direction of the China National Health and Family Planning Commission. So far, the CMDS is the most detailed micro-level survey data about China’s floating population. The migrants in the survey were 15–85 years old migrants who live in the inflow area for one month or longer and have no registered permanent residence in the area. The sample was selected from the Chinese mainland’s 31 provinces (except for Hong Kong, Macau, and Taiwan) and the annual report data of the floating population in the first year of the Xinjiang production and Construction Corps as the basic sampling framework, and then stratified. The survey applied a multistage, cluster, stratified, probability--proportional-to-size (PPS) sampling technique to select migrant respondents. The survey data were carried out in 1325 county-level administrative units (including Xinjiang production and Construction Corps), while the source of floating population was extensive, including all county-level administrative units in China.

The second source provides PM2.5 concentration data, which comes from the Atmospheric Composition Analysis Group at Washington University (https://sites.wustl.edu/acag/datasets/surface-pm2-5/ (accessed on 7 September 2021)). Compared with other data, the geographic coverage of satellite observation data is wide, and can match our floating population data. At the same time, the satellite monitoring data is more objective and accurate, which can avoid the measurement errors caused by human factors. Third, some macro statistical data are also used, which are mainly from China’s Urban Statistical Yearbook and China’s County Statistical Yearbook for 2017, including indicators such as per capita GDP, the share of tertiary industry, and the number of doctors per 100 people. In addition, we extracted the mean wind speed at the county level in 2017 as instrumental variable to overcome the endogeneity of air pollution. The data came from the European Center for medium range weather forecasts (https://apps.ecmwf.int/datasets/data/interim-full-moda/levtype%3Dsfc/ (accessed on 15 December 2021)).

3.2. Variable Selection

3.2.1. Dependent Variable

We examined respondents’ settlement intention based on the question “In the future, do you intend to live in the current city? (Yes or No)”. This question was used to identify those who have settlement intentions. Specifically, respondents who answered “Yes” could be considered to have settlement intentions, and the answer was coded “1”, “0” otherwise. The study does not consider respondents who are still considering or have no answers.

3.2.2. Independent Variables

The core explanatory variable of this study is air quality, which is the PM2.5 concentration value calculated based on raster data format. There are several reasons for choosing PM2.5 to represent air pollution. First, PM2.5 data obtained from county-level administrative units are more objective and accurate. Second, after the nationwide smog in 2013, residents paid more attention to air pollution, especially PM2.5. Third, the study of the relationship between air quality and economic and social development has been widely recognized in the academic community by taking PM2.5 concentration as a proxy variable for air pollution. After downloading the global PM2.5 map of raster data format, we matched the vector map of China’s county-level administrative divisions in 2015 (The data comes from the resource and environment science data center of the Chinese Academy of Sciences, http://www.resdc.cn/data.aspx?DATAID=202 (accessed on 30 June 2022)) with the administrative codes of China Migrants Dynamic Survey (CMDS) data in 2017. With the help of ArcGIS (East China Normal University, Shanghai, China), we calculated the PM2.5 concentration values of each respondent’s outflow area (hukou registration area or area of origin) and inflow area (Destination area) at the county level in 2017.

3.2.3. Control Variables

Following some previous studies [53,54,55,56], we included a set of covariates in the models, including age (continuous variable), gender ( female = 1, male = 0), hukou (agricultural hukou = 0, nonagricultural hukou = 1), marital status (single, divorced and widowed = 0, married = 1), education (categorical variable, primary school or below = 1, junior high school = 2, senior high school = 3, college or above = 4), general self-rated health (categorical variable, not good = 0, good = 1), income (continuous variable measured in yuan), flow range (categorical variable, inter-province migration = 1, inter-city migration = 2, Inter-county migration = 3), occupation type (categorical variable, employee = 1, employer = 2, own business = 3, other = 4), housing situation (categorical variable, renting = 1, commercial housing = 2, self built house = 3). To control the impact of different city types, according to the administrative level, the cities where the migrants were located are divided into four categories: provincial capital city (including municipalities directly under the central government), prefecture level city, county-level city and county. Additionally, it has been demonstrated that economic development differences between inflow city (destination city) and outflow area (hukou registration area or area of origin) will affect the settlement intention of the floating population. According to the response, we obtained the county-level administrative unit of the outflow area and the inflow city of the respondent. By matching the data of the statistical yearbook, the variable of economic gap is obtained, which is composed of the ratio of the per capita GDP of the inflow city and the outflow area of the floating population. In addition, we also selected the share of the tertiary industry representing the city industrial structure, and the number of medical doctors per 100 people, which represents the level of public services in the city. The results show the definitions and descriptive statistics of variables involved in the empirical analysis(

Table 1. Descriptive statistics of variables.

Variables	Description	Mean	Std. Dev.
Settlement intention	In the future, do you intend to live in current city? Yes = 1, No = 0	97.234	0.164
PM2.5	Air pollution concentration in the destination area (μg/m3)	39.423	16.380
Age	Age of floating population	35.760	9.570
Gender	male = 0; female = 1	42.427	0.494
Hukou	Agricultural hukou = 0; Nonagricultural hukou = 1	22.592	0.418
Marital status	Single, divorced and widowed = 0; Married = 1	81.287	0.390
Education	Primary school or below = 1; Junior high school = 2; Senior high school = 3; College or above = 4	2.487	0.962
General self-rated health	Not good = 0; Good = 1	98.84	0.107
Income	Average monthly income of floating population (10,000 yuan)	0.443	0.284
Flow time	Flow time (year)	6.349	5.995
Flow range	Inter-province Mobilization = 1; Inter-city Mobilization = 2; Inter-county Mobilization = 3	1.673	0.752
Occupation	Employee = 1; Employer = 2; Own business = 3; Other = 4	1.753	0.968
house	Renting = 1; Commercial housing = 2; Self built house = 3	1.289	0.509
Economic Gap	Per capita GDP of inflow city (Destination city)/per capita GDP of outflow area (Hukou registration area)	3.351	2.880
City type	Provincial capital city = 1; Prefecture level city = 2; County level city = 3; County = 4	1.925	1.018
Doctor_Num	Number of doctors per hundred (Per 100 people)	0.471	0.160
Third	Proportion of tertiary industry in GDP (%)	53.26	11.360

Note: The data are taken from 2017CMDS, with 109,969 observations.

).

3.3. Empirical Strategy

In order to verify our research hypothesis, we use probit models to evaluate the impact of air pollution on floating population’s settlement intentions. The model expression is shown in Equation (1).

$$Probit\left(Settl{e}_{ij}=1\right)=\varphi \left({\beta }_0+{\beta }_{1}PM2.5+{\beta }_2{X}_1+{\beta }_3{X}_2\right)$$

(1)

In the model, a migrant’s settlement intention $Settl{e}_{ij}$ is a binary choice variable, defined as either 1 or 0. $PM2.5$ refers to air pollution, expressed by the average concentration of PM2.5 in the place where migrant resided. ${\beta }_2$ is a vector coefficients for the a set of control variables $X_1$, i.e., age, gender, hukou, marital status, education, personal income, health status, flow time, flow range, occupation, housing conditions, economic gap, city types, the number of medical doctors per 100 people, proportion of tertiary industry in GDP. ${\beta }_3$ is a vector of coefficients for the city’s characteristics, i.e., city types, economic development level, industrial structure, and medical resources.

In the analysis, we included an instrumental variable(IV) to control the endogeneity of air pollution, and focus on the heterogeneity of air pollution in different individuals and cities. In addition, we discussed the limitation of research and summarized the results.

4. Results

4.1. Basic Results

Table 2. Estimated marginal effects on probability from the probit model on floating population’s settlement intentions.

	Model 1	Model 2	Model 3
	Effect	Effect	Effect
Settlement intention
PM2.5		−0.000 ***	−0.012 ***
		(0.000)	(0.002)
Income	0.006 ***	0.006 **	0.089 **
	(0.002)	(0.002)	(0.037)
Age	0.002 ***	0.002 ***	0.030 ***
	(0.000)	(0.000)	(0.006)
${\mathrm{Age}}^2$	−0.000 ***	−0.000 ***	−0.001 ***
	(0.000)	(0.000)	(0.000)
Gender (male)	−0.000	−0.000	−0.000
	(0.001)	(0.001)	(0.018)
Education (primary school or below)
Junior high school	0.008 ***	0.008 ***	0.154 ***
	(0.001)	(0.001)	(0.024)
Senior high school	0.011 ***	0.011 ***	0.207 ***
	(0.002)	(0.002)	(0.029)
College or above	0.009 ***	0.009 ***	0.180 ***
	(0.002)	(0.002)	(0.035)
Nation (Ethnicity)	−0.002	−0.001	0.037
	(0.002)	(0.002)	(0.033)
Marital status (no spouse)	0.010 ***	0.010 ***	0.171 ***
	(0.001)	(0.001)	(0.024)
Hukou (Agricultural hukou)	0.001	0.001	0.021
	(0.001)	(0.001)	(0.023)
Employment status (employee)
Employer	0.005 **	0.005 **	0.091 **
	(0.003)	(0.003)	(0.044)
Own business	0.000	0.000	0.010
	(0.001)	(0.001)	(0.019)
Other	−0.003	−0.003	−0.070
	(0.004)	(0.004)	(0.061)
Housing info (Renting)
Commercial housing	0.005 **	0.005 **	0.084 **
	(0.002)	(0.002)	(0.043)
Self-built house	0.023 ***	0.023 ***	0.402 ***
	(0.003)	(0.003)	(0.049)
Type of migration (Inter-county migration)
Inter-province migration	−0.006 ***	−0.007 ***	−0.155 ***
	(0.001)	(0.001)	(0.028)
Inter-city migration	−0.001	−0.001	−0.049 *
	(0.002)	(0.002)	(0.028)
Migration duration	0.001 ***	0.001 ***	0.015 ***
	(0.000)	(0.000)	(0.002)
Self-integration	0.049 ***	0.049 ***	0.850 ***
	(0.001)	(0.001)	(0.022)
Self-evaluation health	0.016 ***	0.017 ***	0.298 ***
	(0.003)	(0.003)	(0.058)
Economic Gap	0.001 ***	0.001 ***	0.011 ***
	(0.000)	(0.000)	(0.003)
Destination city type (county)
Provincial capital city	0.013 ***	0.015 ***	0.351 ***
	(0.002)	(0.002)	(0.036)
Prefecture level city	0.011 ***	0.012 ***	0.228 ***
	(0.002)	(0.002)	(0.027)
County level city	0.007 ***	0.007 ***	0.115 ***
	(0.002)	(0.002)	(0.030)
Number of doctor s/100 residents_	0.003	0.003	0.031
	(0.003)	(0.003)	(0.058)
Third	0.000 **	0.000 **	0.003 ***
	(0.000)	(0.000)	(0.001)
Observations	109,969	109,969	109,969

Note: Robust standard errors in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01.

presents the results of the probit regressions analyzing the factors influencing the settlement intention of Chinese floating population. In model 1 only the individual effects and city effect are included. Model 2 incorporates the individual characteristics, city characteristics, and air pollution. The results of model 2 show that the coefficient of the variable PM2.5 was significantly negative at the level of 1%, which indicates that an increase of PM2.5 concentration could decrease a Chinese migrant’s settlement intentions.

Considering that air pollution is affected by city economic activities and population aggregation, in order to control the endogeneity of the variable PM2.5, referred to prior research [54], we use the annual mean wind speed at county level as the instrumental variable. There are two reasons for this. On the one hand, the instrumental variable depends on the current wind speed of the county-level administrative unit and is determined by the meteorological conditions and geographical conditions. After controlling the variables at the urban level, it has no direct relationship with the population inflow in the region, and can meet the exogenous hypothesis of an instrumental variable. On the other hand, our research unit is a county-level administrative unit, and it is difficult to obtain the temperature, humidity and other meteorological data at this scale.

In the first stage, the coefficient of wind speed (w10) is significantly negative at the 1% significance level, which means that the higher the wind speed, the lower the PM2.5 concentration. At the same time, in the exogeneity test of the IV, the F values (17.28) of the first stage is greater than 10 and all passed the Wald exogeneity test [54]. Therefore, the selection of instrumental variables is reasonable and effective. The IV-probit model results are shown in Model 3. After the IV method is used to correct any exogeneity bias, we found that air pollution is still one of the reasons that hinder the settlement intention of floating population. In other words, every one-unit increase (1 μg/m³) in PM2.5 concentration, the probability of the settlement intention s of a migrant’s will fall by 1.2%.

According to estimates for the individual-level control variables, married, highly educated, employers with high-income floating migrants have stronger willingness to settle in the destination city. With the increase of age, Chinese floating population’s settlement intention first increases and then decreases, which is similar to the findings in the existing studies [7,26]. The self-assessment health and integration status of the floating population also influence their settlement intention. The floating population with good health has a strong willingness to settle down. In addition, floating population with longer migration duration and shorter migration distance have stronger settlement intention.

In terms of city characteristics, the migrantis more likely to stay in the city with high administrative level. The greater the difference of economic development level between the inflow city (destination city) and outflow area (hukou registration area), the stronger a migrant’s settlement intention. The results also show that the migrant preferred to settle in cities with a high proportion of tertiary industries, which can accommodate most of the employed migrants.

4.2. Impact of PM2.5 on Settlement Intention: Individual Differences

We further investigated whether the effect of air pollution on floating population’s settlement intention has age cohort, education, and air-quality gap differences. Samples will be grouped from multiple perspectives to analyze the impact of air pollution on individual heterogeneity.

(1) Heterogeneity of the floating population’s age cohort. Air pollution may have different effects on the settlement intention of floating population of different ages. According to the age structure, the samples were divided into three subsamples: under 30 years old, 30–45 years old and over 45 years old. The regressions were performed separately for the three age groups categories. The results show that the impact of air pollution on the settlement intention of floating population gradually increases with the increase of age. For the floating population in the 30–45 age group, as the PM2.5 concentration in the destination city rises by 1 unit (1 μg/m³), the settlement intention will decrease by 1%. However, for over 45 years old age group, when the concentration of PM2.5 increases by 1 μg/m³, the probability of the floating population’s settlement intention will fall by 1.8%. This means that oler migrants are more sensitive to air pollution than the young(

Table 3. Estimated marginal effects on probability from the IV-probit model on floating pouplation’s settlement intentions by age group.

Variables	Model 4	Model 5	Model 6
	under 30 Age Group	30–45 Age Group	above 45 Age Group
	Effect	Effect	Effect
PM2.5	−0.006	−0.010 ***	−018 ***
	(0.005)	(0.003)	(0.005)
Other variables	Control	Control	Control
Observations	33,096	54,556	22,317

Note: Robust standard errors in parentheses, *** p < 0.01.

).

(2) Heterogeneity of the floating population’s education. According to the educational attainments of the floating population, the sample is divided into two subsamples: the below senior high school group and the high senior school or above group. The model results show that the probability of settlement intention of the floating population with an education level below high school will decrease by 1.1% for every 1 μg/m³ increase of PM2.5 concentration in the inflow city. while it has a greater influence on the floating population with high school education and above, the effect is 1.3% (

Table 4. Estimated marginal effects on probability from the IV-probit model on floating population’s settlement intentions by education.

Variables	Model 7	Model 8
	Below Senior High School	High Senior School or Above
	Effect	Effect
PM2.5	−0.011 *	−0.013 ***
	(0.001)	(0.001)
Other variables	Control	Control
Observations	62,797	47,172

Note: Robust standard errors in parentheses, * p < 0.1, *** p < 0.01.

).

(3) Heterogeneity of the air quality between the inflow city and outflow area. Previous studies have shown that the intention of environmental migration will be affected by personal adaptability [55]. Therefore, we consider the impact of the difference in air quality between the inflow city and outflow area. The samples were divided into two groups by comparing the air quality of outflow area and inflow city. In the first group, samples were come from places with poor air quality to cities with good air quality, while the other group was from places with good air quality to cities with poor air quality.

As expected, we find that the influence of air pollution on the floating population’s settlement intention is affected by their adaptability to air pollution. According to the results, the PM2.5 concentration coefficient of model 9 is −0.017, which is significant at 1% level. This means that when concentration of PM2.5 increases by 1 μg/m³, the probability of settlement intention will fall by 1.7%. While, the coefficient of PM2.5 concentration in the model 10 was significantly negative (5%). With an increase in the PM2.5 concentration could decrease the probability of settlement intention by 1.1%. The comparison of the coefficient changes in models 9 and 10 demonstrates that the floating population from areas with better air quality is more sensitive to air pollution (

Table 5. Estimated marginal effects on probability from the IV-probit model on floating population’s settlement intentions by local air quality.

Variables	Model 9	Model 10
	Inflow City PM2.5 ≥ Outflow Area PM2.5	Inflow City PM2.5 < Outflow Area PM2.5
	Effect	Effect
PM2.5	−0.017 ***	−0.011 **
	(0.004)	(0.005)
Observations	58,702	51,267

Note: Robust standard errors in parentheses, ** p < 0.05, *** p < 0.01.

).

(4) Heterogeneity of the floating population’s health status. The impact of air pollution on the settlement intention of the floating population may also be affected by the health status of the floating population According to the reports of the respondents, we divided the floating population into two groups according to their health status: healthy and unhealthy. The results of the model show that air pollution has a significant negative impact on the settlement intention of both groups. When the concentration of PM2.5 increases by 1 μg/m³, the settlement intention of the unhealthy floating population will decrease by 3%, while that of the healthy floating population will only decrease by 1.2%. It can be seen that the unhealthy floating population is more sensitive to air pollution (

Table 6. Estimated marginal effects on probability from the IV-probit model on floating population’s settlement intentions by health status.

Variables	Model 11	Model 12
	Healthy	Unhealthy
	Effect	Effect
PM2.5	−0.012 ***	−0.030 ***
	(0.000)	(0.015)
Other variables	Control	Control
Observations	108,679	1290

Note: Robust standard errors in parentheses, *** p < 0.01.

).

4.3. Impact of PM2.5 on Settlement Intention: City Characteristics

In China, there is a strict and multi-level administrative hierarchy between cities, which afffects the distribution of resources. Some studies suggest that city administrative hierarchy may be more important than city geographical location, economic level and industrial structure [56]. Therefore, we have reason to believe that the impact of air pollution on the settlement intention of the floating population may vary with the administrative level, geographical location and economic development level of the city.

(1) Heterogeneity of Tier-1 city and others cities. Beijing, Shanghai, Guangzhou, and Shenzhen are classified as first tier cities according to the factors such as the economic development level, population size, public service, and other factors [57]. It can be seen from results that the PM2.5 concentration coefficient of Model 13 is significantly negative (p < 0.01), indicating that in Non-Tier 1 Cities, air pollution could decrease the settlement intentions of floating population. While, the coefficient of PM2.5 concentration in the model 14 was significantly positive (p < 0.01), which indicated that Tier-1 cities’ respondents were more tolerant of air pollution. Tier-1 cities, which are regarded as the engine of China’s economic growth, can provide more employment opportunities, higher income and better public services, and have remained a strong attraction for China’s floating population (

Table 7. Estimated marginal effects on probability from the IV-probit model on floating population’s settlement intentions bycity type.

Variables	Model 13	Model 14	Model 15	Model 16	Model 17	Model 18
	Other cities	Tier-1 city	Beijing	Shanghai	Guangzhou	Shenzhen
	Effect	Effect	Effect	Effect	Effect	Effect
PM2.5	−0.014 ***	0.064 ***	−0.010	0.075 ***	−0.064	−0.047
	(0.002)	(0.019)	(0.017)	(0.021)	(0.070)	(0.064)
Other variables	Control	Control	Control	Control	Control	Control
Observations	96,821	13,148	5212	5280	1256	1400

Note: Robust standard errors in parentheses, *** p < 0.01.

).

(2) Heterogeneity of city administrative hierarchy. According to the destination city’s administrative hierarchy, the sample was divided into four subsamples of provincial capital cities, prefecture level cities, county level cities, and county-level cities. The results show that PM2.5 has the greatest negative impact on the settlement intention of the floating population in county-level cities, followed by prefecture level cities (

Table 8. Estimated marginal effects on probability from the IV-probit model on floating population’s settlement intentions by city hierarchy.

Variables	Model 8	Model 9	Model 10	Model 11
	County	County level city	Prefecture level city	Provincial capital city
	Effect	Effect	Effect	Effect
PM2.5	−0.000	−0.039 ***	−0.017 ***	−0.002
	(0.006)	(0.012)	(0.004)	(0.004)
Other variables	Control	Control	Control	Control
Observations	12,767	15,761	32,750	48,691

Note: Robust standard errors in parentheses, *** p < 0.01.

), which means that the higher the administrative level of the city, the less negative impact of air pollution on the settlement intention of the floating population. In prefecture level city, when the increasing of PM2.5 concentration by 1 μg/m³, the probability of floating population’ settlement intention would decrease by 1.7% (p < 0.001). For county level cities, every one-unit increase (10 μg/m³) in PM2.5 concentration, the probability of floating population‘settlement intention would decrease by 3.9% (p < 0.001).

(3) Heterogeneity of different regional. To look into differences of city characteristics in more detail, we divided China into four regions (The eastern region includes Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong and Hainan. The central region includes Shanxi, Anhui, Jiangxi, Henan, Hubei and Hunan. The western region includes inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia and Xinjiang. northeastern region includes Liaoning, Jilin and Heilongjiang. http://www.stats.gov.cn/ztjc/zthd/sjtjr/dejtjkfr/tjkp/201106/t20110613_71947.htm (accessed on 5 December 2022)). And then conducted separate regression analyses of the microdata on the floating populations in different types of cities in the four regions. The results show that there are regional differences in the impact of air pollution on the settlement intention of the floating population.

In the western region, PM2.5 significantly decreases the settlement intention in the provincial capital cities. The coefficient of PM2.5 concentration is positive in the eastern provincial capital cities. The capital cities in the eastern region have developed economy, most of the capital cities in the eastern are economically developed, with many employment opportunities and high salaries, which has a strong attraction for the floating population. In the centralregion, with 1 μg/m³ of PM2.5 concentration increasing, the probability of floating population’ settlement intention increases by 2%. This is a puzzling phenomenon. As we all know, the economy of the central region is not developed, and it is still facing serious air pollution. Monitored data shows that, most of the floating population in the central region comes from the interior of the central provinces, accounting for 80.55%, which is much higher than the other regions. In addition, the central faces the problem of “one city dominating” in the provincial capitals. These cities (such as Wuhan, Zhengzhou and Changsha) have a high degree of industrial concentration, which provides a lot of employment opportunities for the floating population. Therefore, many floating populations may ignore air pollution and choose employment opportunities nearby (

Table 9. Estimated marginal effects on probability from the IV-probit model on floating population’s settlement intentionsby region.

Variables	Provincial Capital City
	East	Middle	West	Northeast
PM2.5	0.006 *	0.020 *	−0.006 *	0.121
	(0.003)	(0.012)	(0.003)	(0.170)
Observations	23,256	7409	15,838	2455
	Prefecture level city
	East	Middle	West	Northeast
PM2.5	−0.020 ***	−0.001	0.007	−0.027 *
	(0.007)	(0.009)	(0.009)	(0.016)
Observations	15,489	8657	7868	1963
	County level city
	East	Middle	West	Northeast
PM2.5	−0.025 ***	−0.022 *	−0.119	−0.055
	(0.008)	(0.010)	(0.139)	(0.069)
Observations	8230	1687	3826	867
	County
	East	Middle	West	Northeast
PM2.5	−0.017	0.008	−0.006	−563.598
	(0.011)	(0.011)	(0.009)	(1013.173)
Observations	2907	2648	6766	103
Other variables	Control	Control	Control	Control

Note: Robust standard errors in parentheses, * p < 0.1, *** p < 0.01.

).

In prefecture-level cities, when the concentration of PM2.5 increases by 1 μg/m³, the probability of the floating population’s settlement intention in the eastern will fall by 2%, while in the northeast will decrease by 2.7%.For county-level cities, the negative impact of PM2.5 on the settlement intention is also different. The negative impact of PM2.5 in the eastern region is significantly greater than that in the central region, which may be related to the composition of the floating population in the two places. In the central county-level cities, the floating population mostly came from the rural areas within each province. While many of the floating population in the eastern county-level cities are from other regions, as those cities have certain requirements for the city’s environmental quality.

5. Discussion

Researchers has conducted a large number of comprehensive studies on the factors that affect the floating population’s settlement intention, including economic factors (e.g., income level) [58], personal factors (e.g., age, gender, education, occupation, etc.) [59], public services (such as educational resources, medical resources, cultural facilities) [60,61], policy (hukou system) [62], social factors (e.g., social network and social environment) [63], and environmental factors (air) [64]. In the existing literature, there are few empirical studies on the impact of air pollution on the heterogeneity of the floating population’s settlement intentions. In this study, based on the 2017 China Migrants Dynamic Survey (CMDS) data plus two other data sources, more than 100 thousand Chinese floating population were selected for the analysis the impact of air pollution on the floating population’s settlement intentions, especially the exploration of individual heterogeneity and differences in city characteristics. We constructed a probit analysis, and used the instrumental variable method to deal with the potential endogeneity problem. China is not only the largest developing country in the world, but also has the largest number of migrant workers. Therefore, the results from the current study have strong theoretical and practical significance.

Our study found that air pollution has a statistically significant and negative effect on the settlement intention of floating population. This is consistent with the results of Yue et al. [65], who found that the concentration of PM2.5 increases by 1 unit, the probability of migrants settling down in the city in which they currently resided for work or business will significantly decrease. The results of heterogeneity analysis in the current study show that older, higher education levels and poor health are more sensitive to air pollution. The effect of air pollution on the settlement intentions of the floating population is also affected by the migrants’ adaptability to air pollution. That is, the intensification of air pollution will reduce the attraction of destination cities to those migrants from areas with better air quality. In first-tier cities, air pollution has not weakened their attraction to the floating population. While in the higher administrative level of the city, the negative impact of air pollution on the floating population’s settlement intention is smaller, and the role of air quality also varies among different regions. Our findings are also consistent with those Wang et al. [47], who found that city administrative level and air quality play an important role in shaping the willingness of hukou conversion for migrants with settlement intention. The findings of this paper provide new empirical evidence for research on the settlement intention of floating population.

This study has several limitations that can be addressed in future studies. First, as a cross-sectional study, the study time span was limited only to 2017, unable to establish a temporal relationship between the air pollution and the migrants’ settlement intention. Because, the impact of air pollution on the settlement intention of floating population is a dynamic process. future research should use panel data, when available, to further expand and verify this relationship in more detail. Second, air pollution indicators also include NO₂, SO₂, etc., and PM2.5 is just one of them. Individual sensitivity to different air pollution may be different. Thus, it will be necessary to use multiple indicators for a more comprehensive analysis. Finally, the study of the relationship between air pollution and the floating population’s settlement intention also involves many omitted variables, such as urban population size, urban climate and natural environment [52,61,66] Due to data availability, an analysis including these variables is beyond the scope of the current article.

6. Conclusions

Based on the China Migrants Dynamic Survey Data from 2017 the satellite grid data of global PM2.5 concentration, and additional area-level data, this study investigated the influences of air quality on China’s domestic migrants’ settlement intention of the floating population, and analyzed the individual heterogeneity and city characteristics and their effects on the relation between air pollution and migrants’ settlement intention. The main findings are summarized as follows:

Air pollution could significantly decrease the settlement intention of Chinese floating population, when concentration of PM2.5 increases by 1 μg/m³, the probability of settlement intention will fall by 1.2%. Our individual heterogeneity analysis shows that the influences of air pollution on different groups of migrants has significant heterogeneity. Those migrants who were older, better educatated levels and with poorer health are more sensitive to air pollution when it came to settlement intention. It is also worth mentioning that the influences of air pollution on settlement intention is influenced by the adaptability of individual to air pollution, that is, respondents with better air quality in their hometown were more sensitive to air pollution. Furthermore, the study found that there were significant city differences in the impact of air pollution on settlement intention. The higher the administrative level of a city, the smaller the negative impact of air pollution. And there are also regional differences in the effects ofair pollution: Its effect on settlement intention in the western provincial capital cities was negative, though it did not reduce settlement intention in the eastern and central provincial capital cities. Unsurprisingly, air pollution has not weakened the attractiveness of Tier-1 cities to the floating population. It is clear that air quality is not a priority in those cities, which can provide more employment opportunities, higher salaries and better public services for the floating population. Based on the conclusions above, we suggest the following policy recommendations,. First, local government should pay more attention to the role of environmental factors in forming their talent attraction strategies. The results of this study show that environmental quality indicated by air quality became a significant influencing factor of floating population’s settlement intention, and the groups with high human capital are more sensitive to air pollution. Therefore, in the context of the "talent competition", new policies for attracting and retaining talents should highlight the advantages of environmental quality, strengthen the synergy between environmental policies and talent policies. Second, local governments must strengthen the research on their impact of environmental pollution on health and increase public awareness in this respect. While this study found that air pollution did not weaken the attraction of large cities to migrants, and the potential increase in income could offset the negative impact of environmental pollution on settlement intention, this temporary solution at the cost of health is not sustainable in the long run. It will be a much more viable solution to reduce the research results of environmental pollution in all localities and improve local citizens’ understanding of the importance of air quality its impact on their lived environment.