Impact of PM2.5 on Second Birth Intentions of China’s Floating Population in a Low Fertility Context

The total fertility rate of the Chinese population has declined dramatically over the last three decades. Research has substantiated the causal link between particulate matter (PM) and adverse health effects. However, the impact of PM on the birth intentions or fertility behavior of the childbearing population remains understudied. The paper analyzes the impact of PM2.5 concentration (a mixture of extremely small solid particles and liquid droplets found in the air) on the second birth intentions of the Chinese floating population. We used urban migrant population matching data at the prefectural level for the analysis. The unique datasets were derived from the Chinese Floating Population Dynamic Survey in 2014 administered by the National Health Commission, the National Prefecture-level City Matching Data administered by the National Bureau of Statistics of China, and the air pollution index PM2.5 collected by the Green Peace Organization. The results show that PM2.5 concentration has a negative impact on the second birth intentions of the floating population. This impact exhibits marked regional heterogeneity: the desire for a second birth across migrant groups living in south China decreases if PM2.5 concentration goes up, while migrants coming from, and living in, north China show strong intentions to have a second birth despite an increase in PM2.5 concentration in northern cities. The results have direct implications for the Chinese government at various levels to play a vital role in making and implementing environmental policies on the mitigation of smog to effectively safeguard the health of individuals and communities and potentially raise China’s fertility rate.


Introduction
The total fertility rate of the population in China has experienced a remarkable decline since the 1970s and has been well below replacement since 1990. The rate is even lower for the floating population (0.94) than for their counterparts living in rural areas (1.28) or urban destinations (1.01) [1]. The floating population refers to people who have lived away from the place where they registered household status for six months or more [2]. The floating population consists of two groups of migrants: the rural-to-urban migrants and the urban-to-urban migrants (e.g., university graduates seeking jobs in other cities). The stock of the floating population has increased over time, rising from 6.57 million in 1982 to 221 million in 2010, and further to 247 million in 2016, accounting for 0.7%, 16.5% and

Determinants of Birth Intentions of the Chinese Floating Population
Fertility policy. Since the implementation of the National Family Planning (one child) policy in the early 1970s, there has been a dramatic decline in the total fertility rate (TFR) [25,26]. Since 2000, despite a slight growth (changing from 1.58 in 2001 to 1.70 in 2016), the TFR has been much lower than the replacement level, resulting in the rapid ageing of the Chinese population. To respond to the ageing population and related aged care and other social issues, in 2013, China started to implement a conditional two-child policy, permitting a family in which at least one of the married couple was the only child of their parents to have two children [27,28]. Since October 2015, China has abolished the iconic one-child policy and started to implement a universal two-child policy, under which a married couple can have two children [29]. Some studies argue that the two-child policy does not boost the TFR because many of the new generation of migrants are unwilling to have a second child. The key reasons include their disadvantaged position in the competitive labor market and high costs of raising a child in cities [30]. Low birth intentions have become a new norm of the floating population [31]. The era of policy driven childbearing behavior has gone, and the present two-child policy plays a limited role in influencing fertility behavior [32].
Education. Education is one of the most established socioeconomic determinants of fertility behavior or birth intentions [24]; it is inversely associated with fertility behavior [33]. Highly educated women tend to substitute child numbers with childrearing quality [34]. Since childrearing is time-intensive, an increase in wage rates induces a negative substitution effect on the demand for more children [35]. Previous studies in other developing countries emphasize that the educational composition of a population is a key determinant of a country's TFR and that women's fertility intentions tend to decline with educational attainment [36,37]. In the Chinese context, the migrant group holding a relatively high level of education has a strong capability to factor social environments into their fertility plans, leading to a sharp contrast between fertility intentions and actual fertility behavior [38]. The birth intentions of the childbearing age females of the floating population drop by their education level [31].
Economic status. This is a complex factor that reflects an individual's occupation and income, and thus determines a family's affordability to raise children. Many migrants work more hours than local urban workers and work in low-end jobs that local urban workers do not want to take [39]. Generally, the current economic status of migrants is much better than that in the 1980-1990s [38]. Studies show that annual household income positively affects both actual fertility and preferred fertility: the higher the economic status, the greater the concern about the quality of childrearing and the higher the unwillingness to have children [40,41]. Female workers at childbearing ages tend to have fewer children than unemployed ones because the opportunity cost of women engaging in socialized careers is higher than that of those not in the labor market [42]. The higher the occupational level and wages, the greater the job opportunity cost and consequently the lower their second birth intentions. However, as highlighted by some Chinese researchers, the longer the floating population live in urban areas and the more wages they earn, the greater inclination to form a family [43].
Cultural environment. People's cultural attitudes towards birth differs according to an individual's social culture, social class, and social resources [44]. For the floating population, the geographical environment and cultural differences between the origin and destination affect fertility intentions in different ways. Traditional culture in rural areas motivates people to prefer sons over daughters, and such a cultural mindset makes it difficult for rural migrants to integrate into urban societies [45]. There is a significant difference in gender preference [46]. Literature shows that the occupation of migrants influences their cultural inclusion in the destination [47]. For example, compared with farmers, migrants working in business and service sectors have a greater propensity to accept local urbanites' view on fertility and thus cater to the fertility trends in the destination. Also, a friendly, accepting urban environment is associated with positive fertility behavior [48]. A comparative study of migrants' living conditions in two megacities (Shanghai and Shenzhen) shows a stronger willingness of forming families and giving birth for migrants living in Shenzhen than in Shanghai [38]. This is because the former city has a more inclusive culture than the latter; it is more accepting of diverse linguistic and cultural diversities caused by a massive influx of migrants.
Hukou-based social exclusion and floating experience. China's unique household registration system (hukou), established in 1958, has been used as a demographic mechanism for managing population mobility and as social engineering for distributing social security and social welfare schemes that disproportionately benefit urban citizens [49]. Hukou status has two related parts: residential location (cities or towns and rural residential locations like villages or state-owned farms) and socioeconomic status (agricultural or non-agricultural). It classifies Chinese citizens into two groups (agricultural/rural and non-agricultural/urban) and endows each with vastly different opportunities and socioeconomic status, producing social segregation and disparity. Migrants cannot enjoy the same social security benefits (age pension, medical care, unemployment, maternity, and work-related injury insurance), legal aid and children's education as the local urban residents due to the constraints of the long-lasting hukou system [50]. This discriminatory hukou system affects migrants' fertility desire. Many migrants live in an unstable state and experience multiple risks including unemployment, frequent movement, poor housing, and low income, potentially causing the fertility decline [51]. Some researchers further note that the rural-urban divide caused by the hukou system, and then hukou-or location-based social security system, has not yet been fully removed in China [7]. The hukou-based social exclusion hinders migrants' integration into urban life and culture.
Due to the restriction of the hukou system, it is very difficult for the floating population to obtain permanent residency in the cities [50]. Therefore, unmarried migrants are more likely to choose permanent residency in cities and the migrants who have a family left behind often choose to take temporary or circular migration [51]. Frequent movements and unstable living conditions reduce their intentions and realization of fertility [52]. Research into the relationship between migration and fertility in other nations also show that migrants have lower fertility rates than non-migrants in their homelands but higher fertility rates than local residents in the destination [53]. The separation of migrant spouses may have a significant effect on the timing of childbearing. Adaptation to an urban lifestyle plays a lesser role in reorienting their fertility intentions because many migrants stay in the destination for a limited time [54].

Impact of Air Pollution on the Birth Intentions of China's Floating Population
General literature in population and environment research shows that people's experience in air pollution differs by their social strata [55]. Low-income people such as migrant workers are more likely to be exposed to toxic and hazardous work environments. Smog is extremely harmful to human health and has claimed tremendous loss of life and property of Chinese people. The number of deaths caused by PM 2.5 in Beijing in 2010 was 2349, accounting for 1.9% of the total deaths in that year; the direct economic loss amounted to nearly US$18.6 billion [56]. Medical studies establish a significant correlation between air pollution and residents' respiratory disease, cardiovascular disease, and asthma [57,58]. The health conditions for people with respiratory diseases such as asthma, emphysema, or bronchitis become even worse if they work in smog-affected situations [59]. Smog also affects the mental health of people, increasing depression and suicidal events [15].
Severe smog can lead to an increase in the death rate of the population, and removal of smog can result in population growth. In Los Angeles, US, a large-scale smog removal process from the 1970s to the beginning of the 21st century demonstrates that when the severity of the smog problem is reduced, a population (especially in relatively poor areas) grows significantly [60]. However, there is no study directly showing how smog reduces an individual's birth intention or human reproductive behaviors. A relevant study in Barcelona identifies a significant association between a decrease in fertility rates and an increase in traffic-related air pollution (particularly for the coarse fraction of PM) [61]. Medical scientists find that smog can damage the health of newborn babies. A longitudinal study of 100 newborn babies aged till 7 years old in New York finds that children growing up in areas with heavy pollution are more likely to have distractions, anxiety disorders, and depression than their counterparts growing in areas with little pollution [62]. Using multinational data, Dadvand et al. [17] find that if pregnant women live in conditions with severe air pollution, their babies are likely to be underweight. For an increase by every 10 micrograms per cubic meter of PM 10 , the risk of underweight babies grows by 3%. Further, taking advantage of the natural experiment that occurred during the 2008 Beijing Summer Olympics, Rich et al. [18] reveal that an increase in air pollutants reduces the baby's weight at birth. For every increase by one quartile deviation of the concentrations of PM 2.5 , CO, SO 2 , and NO 2 in air, the neonatal birth weight would reduce by 18, 17, 23, and 34 g, respectively. Clearly, smog can influence the health and development of the fetus. Fertility is an important life domain and individuals have to make important decisions in this domain [43]. In an era when people are under heavy pressure, one of the major motivations for fertility is not to produce one more offspring to carry on the family line in industrialized economies. Therefore, environmental quality must be taken into account as an important factor in individuals' decision-making process including fertility intentions. From this point of view, it is a rational choice for people to reduce or postpone baby plans and keep the next generation from being harmed by air pollution. The data shows that over one in five (21.3%) of eligible respondents indicated intentions to have a second child (Table 1). This rate is slightly higher for those living in the southern provinces than those living in the northern regions. The average concentration of PM 2.5 in the north reached 68.45 µg/m 3 , which was approximately 20% higher than that in the south (56.85 µg/m 3 ). Over 90% of the respondents indicated that they are aware of the conditional second-child policy. Approximately 13% of the respondents were eligible for having a second child.

The Outcome Variable
In the 2014 CMDS, migrants were asked "Whether or not you intend to have a second child?". The response to this question is the main dependent variable (second birth intention) in this paper. Considering marriage status and the actual age of childbearing, we classified the sample into three groups: "married", "primary child-bearing ages (20-49 years old)", and "having one child who is under 20 years old". The valid sample size includes 34,391 individuals-of which, 38.5% were females and 61.5% were males. The mean age of the sample was 34 years.

The Key Explanatory Variable
The core explanatory variable in this study is PM 2.5 concentration. In general, air pollution is more severe in northern regions than in the southern regions of the country [63]. The most polluted areas encompass the Beijing-Tianjin-Hebei region and its neighbouring provinces including Shanxi and Shaanxi ( Figure 1). These areas are the largest concentrations of China's heavy industries, that is, coal mining, iron and steel, metallurgy, and chemical industries [64]. These areas are drought-prone and have relatively low rates of vegetation coverage. The second highly polluted areas in China include the provinces of Liaoning, Shandong, Jiangsu, and Henan. These areas are densely populated and host massive manufacturing industries emitting hazardous pollutants. The third highly polluted areas are located in the southwestern region centered in the provinces of Sichuan and Chongqing. Due to their geographical location, i.e., situated in Sichuan Basin and surrounded by highlands in west and south and mountains in north and east, tremendous pollutants generated by coal, steel, petroleum processing and other chemical industries in the region cannot be with atmospheric evacuation. Accordingly, we hypothesize that the impact of smog on birth intentions exhibits a geographical discrepancy between the north and south parts of China.

Control Variables
Our regression models considered a number of demographic and socioeconomic factors as control variables. These include individual-level variables (sex, hukou status, age, education, income, and social status); fertility policy; sex and age of the first child; contextual variables at the prefectural city level, e.g., gross domestic product (GDP), per capita GDP, population density, and the share of secondary industry. Controlling economic and industrial variables can avoid the interference of contextual factors at the city (macro) level on the estimation.

Method
Considering the sampling nature of the floating population survey, we used a weighted regression method for data analysis. The outcome variable of this research is second birth intention, i.e., whether or not migrants intend to have a second child under the premise of already having a child. This is a dichotomous variable (1 = 'yes', 0 = 'no'), and so we used a Probit model to examine the relationship. Greene [65] (pp. 772-775) provided the detailed econometric specification for Probit models. In the Probit model, the inverse standard normal distribution of the probability is modeled as a linear combination of the predictors. We also tested the robustness of the results using a subsample that includes responses from migrants in 40 cities/districts in the 2014 survey (see Figure  1). These 40 cities/districts absorbed the largest floating populations and thus can be used as a

Control Variables
Our regression models considered a number of demographic and socioeconomic factors as control variables. These include individual-level variables (sex, hukou status, age, education, income, and social status); fertility policy; sex and age of the first child; contextual variables at the prefectural city level, e.g., gross domestic product (GDP), per capita GDP, population density, and the share of secondary industry. Controlling economic and industrial variables can avoid the interference of contextual factors at the city (macro) level on the estimation.

Method
Considering the sampling nature of the floating population survey, we used a weighted regression method for data analysis. The outcome variable of this research is second birth intention, i.e., whether or not migrants intend to have a second child under the premise of already having a child. This is a dichotomous variable (1 = 'yes', 0 = 'no'), and so we used a Probit model to examine the relationship. Greene [65] (pp. 772-775) provided the detailed econometric specification for Probit models. In the Probit model, the inverse standard normal distribution of the probability is modeled as a linear combination of the predictors. We also tested the robustness of the results using a subsample that includes responses from migrants in 40 cities/districts in the 2014 survey (see Figure 1). These 40 cities/districts absorbed the largest floating populations and thus can be used as a representative subsample of the floating population survey to infer the situation of the entire country. To examine the individual differences in the impact of PM 2.5 on second birth intentions, we performed the regression with interaction terms between smog and selected individual variables. To further examine any regional difference in the impact of PM 2.5 on second birth intentions, we performed separate regressions for four categories of the floating population: "living in the south and from the south", "living in the south but from the north", "living in the north but from the south", and "living in the north and from the north". These four categories were classified in terms of the origin and current residency of migrants.

Results
Tables 2-6 present the results. Overall the models were a good fit as suggested by the significant Wald Chi-square statistics. The prediction accuracy of the models was quite satisfactory, being within 76% and 82% of the correct prediction for the impact models. The models did not suffer from serious multicollinearity (Mean Variance Inflation Factor/VIF = 1.61) and the robust standard errors were used to control for any heteroskedasticity.   Note: *** p < 0.01; ** p < 0.05; * p < 0.1. Numbers in brackets are robust standard errors.

Impact of PM 2.5 on Second Birth Intentions: Reducing Intentions
A central finding drawn from Models 1-4 ( Table 2) is that an increase in PM 2.5 concentration would reduce the second birth intentions of the floating population. Model 1, shown in the first column of Table 2, considered only one independent variable, PM 2.5 . This factor has a negative association with migrants' second birth intentions. For every one-unit increase (10 µg/m 3 ) in PM 2.5 concentration, the rate of second birth intentions of the population under study would decrease by 6.4% (p < 0.01), assuming other factors remain constant. After controlling for individual, institutional (policy)and contextual variables, there are very few changes in the regression coefficients and standard errors across Models 2, 3, and 4.
As expected, these results of the subsample, as presented in Model 5 of Table 2, are consistent with the results of Model 4 that used the complete sample.
Another interesting finding, as shown in Models 2-4, is that women's second birth intention is significantly lower than that of men. When age increases, people's second birth intentions decline. This is especially the case for the group who hold urban (or non-agricultural) hukou status, and also for the males. These findings are consistent with the findings of previous research on fertility desire and fertility behavior [66,67]. Moreover, the higher the education level, the lower the second birth intention. The rural household registers have higher second birth intentions than those urban counterparts. In terms of the first child's sex, if a couple already has a son, their second birth intention is significantly lower than those who have a daughter. This suggests that boy preference still influences the reproductive ideology of some migrants. This result supports the findings of other researchers [31]. Employers and self-employed workers are more willing to have a second child than wage-earning migrants. Importantly, the commencement of the conditional two-child policy in 2013 exhibits a significant and substantial effect on people's intentions to have a second child. The group who are eligible to have a second child has a much higher propensity (by 57.8%) of reproductive intentions than their counterparts who were ineligible in 2014. As the first child grows older, people's intentions to have a second child decline. A couple of contextual factors at the city level have a significant association with second birth intentions (Model 4). Migrants residing in relatively developed and rich cities (measured as per capita GDP) have fewer intentions to have a second child than those living in less developed cities. This result is consistent with the literature [67].

Impact of PM 2.5 on Second Birth Intentions: Individual Differences
Using Model 4 in Table 2 as the baseline model, we conducted a further analysis of the interactions between smog and selected individual variables: gender, educational level, hukou status, income, and sex of the first child. Such analysis provides nuanced understanding of individuals' sensitivity to smog. Interestingly, the interaction terms between PM 2.5 concentration and these individual variables are not statistically significant across Models 6-12 presented in Table 3. This result indicates no significant difference between diverse migrant groups differentiated by important characteristics when they experience smog. However, PM 2.5 concentration remains to be a significant force driving the declines in second-birth intentions.  Note: *** p < 0.01; ** p < 0.05; * p < 0.1. The reproductive age was categorized into three groups: young age: <30 years old (reference group); middle age: 30-39 years; old age: 40-49 years. Monthly income was grouped into four scales: <3000 yuan (reference group); 3000-4000 yuan; 5000-7000 yuan; >7000 yuan. The results for other independent variables in regression are not included in the table due to word limit. Such variables include age, sex, education, social class, hukou status, income, the first child's sex and age, eligibility for a second child, knowing the second-child policy or not, total GDP and per capita GDP in a city, urban population density, and proportion of the secondary industry.

Impact of PM 2.5 on Second Birth Intention: Regional Differences
The regressions were performed separately for the four migrant categories. People living in, and coming from, different regions have different levels of resilience to smog. For the group living in the south, regardless of their hometowns, their second birth intentions significantly decline with the increase in PM 2.5 concentrations (Models 13 and 14 in Table 4). However, the duration of migration might be an important factor influencing the sensitivity to PM 2.5 concentrations of the category of people living in the south but originating from the north. Although this conjecture cannot be confirmed from the regression results here, with the increase in the duration of stay, increased adaptive capacity to local climate and change in living habits and social-cultural norms would be affected more or less by local urbanites. Therefore, for this category of migrants, the way by which smog influences their fertility desire cannot be simply generalized. For migrants living in the north and originating from the south, their second birth intentions are not significantly affected by smog (Model 15 in Table 4). If their origin is in the north, their second birth intentions would increase accordingly when PM 2.5 concentration goes up (Model 16 in Table 4). Clearly, migrants living in the south are more sensitive to smog when it occurs as there is relatively light air pollution in the south. The northerners have a greater resilience to smog, probably because they have become customized to smog and see it as a normal phenomenon in their lives.  Note: * p < 0.05; ** p < 0.01; *** p < 0.001. Numbers in brackets are robust standard errors.
To look into this issue in more detail, we divided China into seven subregions, and then conducted separate regression analyses of the microdata on the floating populations in 20 major provinces (Tables 5 and 6). The results show that the regression coefficient of PM 2.5 concentration is positive only in north China. Especially for migrants in Hebei province, the higher the PM 2.5 concentration, the greater their second birth intentions. This is a puzzling phenomenon. Monitored data shows that smog and air pollution in Hebei has been the top area in China in recent years. The floating population there is dominated by intra-provincial rural migrants, and thus their resilience to smog is relatively high. Many cities there (e.g., Handan, Shijiazhuang) have high PM 2.5 concentrations due to massive heavy industries. These cities absorb young male workers who generally have strong fertility intentions. Migrants there also have a relatively high income, so they can usually afford the cost of raising children, and then births increase. Note: *** p < 0.01; ** p < 0.05. Numbers in brackets are robust standard errors. The results for other independent variables in regression are not included in the table due to the word limit. Such variables include age, sex, education, social class, hukou status, income, the first child's sex and age, eligibility for a second child, knowing the second-child policy or not, total GDP and per capita GDP in a city, urban population density, and proportion of the secondary industry. Table 6. The impact of smog on the second birth intention of the floating population in major provinces in China. Note: *** p < 0.01; ** p < 0.05; * p < 0.1. Numbers in brackets are robust standard errors. The results for other independent variables in regression are not included in the table due to the word limit. Such variables include age, sex, education, social class, hukou status, income, the first child's sex and age, eligibility for a second child, knowing the second-child policy or not, total GDP and per capita GDP in a city, urban population density, and proportion of the secondary industry.

Discussion
Both the fertility and natural population growth rates in China have remained at a decline over the past 20 years. Differing from previous research on the declining fertility rate of the Chinese population, which addressed the effects of demographic factors (especially age and education), Family Planning policy and rapid economic development, this study provides an environmental lens to understand the probable demographic consequence of smog risks. The findings of the study fill in part of the knowledge gap by demonstrating that the prevalence of smog significantly reduces intentions to have a second child across the Chinese floating population, and that the impact of smog on birth intentions exhibits a clear geographical discrepancy between north and south regions of China. A simple linear fit of PM 2.5 concentration and second birth intentions indicates a significant negative but weak correlation between these two variables (with a coefficient of −0.064, at the 0.001 significance level) ( ways. Firstly, existing research has shown that if pregnant women are exposed to smog for a long time, it will damage the physical and mental development of the fetus or the newborn, and this effect will last until the childhood phase [17]. For the sake of the health of newborn babies, women at childbearing ages usually reduce or postpone their fertility intentions. Secondly, smog affects the physical and mental health status of child-bearing women [68,69]. Long-term exposure to smog causes not only a variety of respiratory diseases, but also mental health problems such as anxiety, depression and stress, thereby reducing their fertility intentions [14]. In the Chinese context, some floating populations in the northern areas (e.g., Hebei) have shown strong resilience to smog and have strong intentions to have a second child even with the increased frequency and severity of smog. This phenomenon is complex and will need further quantitative and/or qualitative research in the future to fully explain it.
(SX), Jiangsu (JS), Sichuan (SC), and Chongqing (CQ) is more serious than in other provinces. At the prefectural city level, affected cities are mainly distributed in the north, northwest, central, and southwest regions of the country, which are dominated by heavy industries and arid or semi-arid climate. Overall, there are higher percentages of migrants with have second birth intentions in southern provinces (e.g., Hainan (HAN), Yunnan (YN), Fujian (FJ), Jiangxi (JX), Henan (HEN), Zhejiang (ZJ), Shandong (SD) and Guangdong (GD)) than there are of their counterparts living in northern regions. Smog can reduce fertility desire in at least two ways. Firstly, existing research has shown that if pregnant women are exposed to smog for a long time, it will damage the physical and mental development of the fetus or the newborn, and this effect will last until the childhood phase [17]. For the sake of the health of newborn babies, women at childbearing ages usually reduce or postpone their fertility intentions. Secondly, smog affects the physical and mental health status of child-bearing women [68,69]. Long-term exposure to smog causes not only a variety of respiratory diseases, but also mental health problems such as anxiety, depression and stress, thereby reducing their fertility intentions [14]. In the Chinese context, some floating populations in the northern areas (e.g., Hebei) have shown strong resilience to smog and have strong intentions to have a second child even with the increased frequency and severity of smog. This phenomenon is complex and will need further quantitative and/or qualitative research in the future to fully explain it. Since the 18th National Party Congress in 2012, the Chinese central government has made great efforts to reduce air pollution. Recent developments in environmental policy and programs include setting up strict rules to control motor vehicle pollution, coal-fired pollution, key industries emitting massive pollutants, and fugitive dust pollution, and to increase ecological rehabilitation and apply new technologies to combat smog. The implementations of these programs have achieved certain plausible outcomes [63]. However, large-scale and severe smog still frequently occurs across vast parts of the country in the autumn and winter seasons. Continuing to combat air pollution is an important environmental policy to protect the health of urban and rural residents, potentially enhancing the fertility rate of the Chinese population. y=-1.662x+30.363 R 2 =0.120 Since the 18th National Party Congress in 2012, the Chinese central government has made great efforts to reduce air pollution. Recent developments in environmental policy and programs include setting up strict rules to control motor vehicle pollution, coal-fired pollution, key industries emitting massive pollutants, and fugitive dust pollution, and to increase ecological rehabilitation and apply new technologies to combat smog. The implementations of these programs have achieved certain plausible outcomes [63]. However, large-scale and severe smog still frequently occurs across vast parts of the country in the autumn and winter seasons. Continuing to combat air pollution is an important environmental policy to protect the health of urban and rural residents, potentially enhancing the fertility rate of the Chinese population.

Conclusions and Limitations
Coupled with rapid urbanization, socioeconomic development and heightened air pollution in China, dramatic fertility decline has occurred since the mid-1990s. This study finds that the higher the PM 2.5 concentration, the lower the willingness of the floating population to give birth to a second child, after controlling for the first child's sex and age, and individual-, policy-and city-level factors. There exists significant regional disparity between PM 2.5 concentrations and the second birth intentions of the floating population. Migrants living in south China (especially central-south and southwestern regions) are more sensitive to smog and thus have a lower level of resilience to air pollution and resultantly fewer intentions to have a second child, compared to the group coming from the north and still living in the north, who have a relatively higher tolerance to smog and thus more intentions for second births. China's policy response to low fertility should, firstly, take environmental factors into account, and secondly, adopt effective methods and strategies to eliminate smog (air population broadly) not only in the urban areas of north China but also in the southern parts of the country. PM 2.5 is a key air pollutant causing not only adverse human health problems but also a significant impact on birth intentions. Policies and practical schemes designed for alleviating smog hazards and for raising population fertility should factor in regional geospatial differences in PM 2.5 concentration (or severity of smog) and the birth intentions of diverse groups of the floating population.
Two limitations of this study deserve mention. First, this research used only the 2014 floating population data to assess the impact of smog on fertility intentions. If large-scale data covering all residents can be collected in future studies, through longitudinal observations or quasi-experimental methods, a more comprehensive understanding of the causal effects of smog on fertility intentions (and fertility behavior) will become possible. Second, this study found that an increase in smog in the provinces of the north region (especially Hebei and Shaanxi) facilitates migrants' birth intentions. This result signals out that apart from the northerners' higher levels of resilience, other environmental and non-environmental factors might have mediated people's birth intentions in China. More studies will be needed in this research field.
Author Contributions: W.G. and Z.S., conceptualization, methodology, analysis, writing-original draft preparation; Y.T., writing, revising-reviewing and finalizing the manuscript; X.Y., data collection, writing-original draft preparation. All authors read and approved the final manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.