**3. Results and Discussion**

### *3.1. Basic Results*

Table 4 reports the estimated results of model (1) and model (2). Among them, column (1) exhibits the results from model (1), while column (2) represents the results from model (2). Specifically, the coefficient of the variable *PM2.5* in column (1) of Table 4 was significantly negative at the level of 5%, which indicated that with all else being equal, an increase in the PM2.5 concentration could decrease the settlement intentions of Chinese migrants. In other words, when increasing the annual average PM2.5 concentration of Chinese cities by 1 µg/m<sup>3</sup> , the odds ratio of the settlement intentions of Chinese migrants decreased by 0.30% (calculated using 1 − e <sup>−</sup>0.003). Simultaneously, the coefficient of the variable *income* was significantly positive at the level of 1%, which showed that the higher the incomes that the migrants earned from the city, the stronger the settlement intentions of the migrants. That is, with all else being equal, the odds ratio of the migrants to settle down in the city increased by 12.76% when the monthly incomes of the migrants increased by 10,000 yuan (calculated using 1 − e <sup>−</sup>0.755). Because poorer air quality can decrease the utility levels of the migrants, and higher incomes can increase the utility levels of the migrants, the signs of the coefficients of the above two variables were consistent with the theoretical expectations.


**Table 4.** Results of baseline regression for factors influencing the settlement intentions of Chinese migrants in cities.

Note: The robust standard errors are given in parentheses, \*\*, and \*\*\* indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The abbreviations are defined in Table 2.

After including the interaction term *PM2.5* × *income,* the regression result of column (2) in Table 4 shows that the coefficient of *PM2.5* × *income* was significantly negative, which confirmed the trade-off between a higher income and poorer air quality that was faced by the migrants when deciding whether to settle down in a city. In other words, the influence of air quality on the settlement intentions of Chinese migrants was conditional on their incomes. As income opportunities increased at a migrant's location, the negative influence of poorer air quality on the settlement intention at that location gradually decreased, which indicated that a higher income could make up for the lost utility caused by poorer air quality. Figure 1 further shows the average marginal effect of the variable *PM2.5* on

the settlement intentions of Chinese migrants. As shown in the figure, with the increasing income of the migrants, the average marginal effect of PM2.5 gradually increased from significantly negative to eventually being insignificant. Therefore, the results above indicated that a higher income weakened the negative influence of poor air quality on the settlement intentions of Chinese migrants. *Int. J. Environ. Res. Public Health* **2020**, *17*, x FOR PEER REVIEW 9 of 18

**Figure 1.** Average marginal effects of PM2.5 with 90% confidence intervals. **Figure 1.** Average marginal effects of PM2.5 with 90% confidence intervals.

*3.2. Heterogeneity Test*  The heterogeneity in the influence of both air quality and income on the settlement intentions of Chinese migrants was mainly tested in terms of the following two aspects: (1) Heterogeneity of the migrants' "hukou." Under the current "hukou" system in China, the "hukou" of the migrants can be divided into two types: non-agricultural "hukou" and agricultural "hukou." The migrants with a non-agricultural "hukou" are mainly city residents, while the migrants According to the results of the coefficients of the individual control variables, migrants with the following properties had stronger settlement intentions: non-Han ethnic group, female, married, CCP member or CCYL member, higher educational level, a non-agricultural "hukou," a longer duration of residence in migratory cities, a shorter migration distance, and a non-economic purpose. With the increase of age, the settlement intentions of Chinese migrants first increased and then decreased, which was basically consistent with the existing studies [65–67].

with an agricultural "hukou" are mainly rural workers who moved from rural areas to cities. The preferences of the migrants of these two types in terms of both income and air quality may be significantly different. Based on model (2), the following model could be further estimated: = + ଵ2. 5 +ℎ2. 5 × × ℎ ℎ ℎୀ,ଵ + + + ௧ (3) where the coefficients *αh* (*h* = 0 or 1) of the interaction term *PM2.5 × income × hukouh* represents the difference in the preferences of income and air quality in the settlement decisions of Chinese migrants According to the results of the city-level control variables, Chinese migrants were more willing to settle down in the cities with higher proportions of tertiary industries that accommodate the major part of the employed migrants. In our samples, 59.3% of the migrants were employed in a tertiary industry. Meanwhile, the settlement intentions of the migrants were stronger in the cities with a higher level of openness to international trade, as international trade creates a lot of employment opportunities [68]. In addition, Chinese migrants tended to settle down in the cities with a higher per capita GDP (not significant) and a higher GDP growth rate.

#### with either an agricultural "hukou" or a non-agricultural "hukou." Column (1) in Table 5 reports the *3.2. Heterogeneity Test*

estimated results of the coefficients of the variables in model (3), which showed that the coefficients of the interaction term *PM2.5 × income × hukouh* were significantly positive. On this basis, Figure 2 reports the average marginal effects of PM2.5 on the settlement intentions of the migrants with The heterogeneity in the influence of both air quality and income on the settlement intentions of Chinese migrants was mainly tested in terms of the following two aspects:

different types of "hukou." (1) Heterogeneity of the migrants' "hukou." Under the current "hukou" system in China, the "hukou" of the migrants can be divided into two types: non-agricultural "hukou" and agricultural "hukou." The migrants with a non-agricultural "hukou" are mainly city residents, while the migrants with an agricultural "hukou" are mainly rural workers who moved from rural areas to cities. The preferences of the migrants of these two types in terms of both income and air quality may be significantly different. Based on model (2), the following model could be further estimated:

$$\text{SI}\_{\text{ij}} = \alpha\_0 + \alpha\_1 \text{PM2.5}\_{\text{ij}} + \sum\_{h=0,1} \alpha\_h \text{PM2.5}\_{\text{ij}} \times \text{income}\_{\text{ij}} \times \text{hukol}\_{\text{ij}}^h + \lambda \text{X} + \rho \text{Z} + \mu\_{\text{il}} \tag{3}$$

where the coefficients <sup>α</sup>*<sup>h</sup>* (*<sup>h</sup>* <sup>=</sup> 0 or 1) of the interaction term *PM2.5* <sup>×</sup> *income* <sup>×</sup> *hukou<sup>h</sup>* represents the difference in the preferences of income and air quality in the settlement decisions of Chinese migrants with either an agricultural "hukou" or a non-agricultural "hukou." Column (1) in Table 5 reports the estimated results of the coefficients of the variables in model (3), which showed that the coefficients of the interaction term *PM2.5* <sup>×</sup> *income* <sup>×</sup> *hukou<sup>h</sup>* were significantly positive. On this basis, Figure <sup>2</sup> reports the average marginal effects of PM2.5 on the settlement intentions of the migrants with different types of "hukou." *Int. J. Environ. Res. Public Health* **2020**, *17*, x FOR PEER REVIEW 10 of 18

**Figure 2.** Average marginal effects of PM2.5 with 90% confidence intervals for the migrants of different "hukous." **Figure 2.** Average marginal effects of PM2.5 with 90% confidence intervals for the migrants of different "hukous."

It was found that the negative influences of PM2.5 on the settlement intentions of Chinese migrants gradually became insignificant with the increase of their incomes, no matter whether they had an agricultural "hukou" or a non-agricultural "hukou," which was consistent with the conclusion drawn from Figure 1. However, with the increasing incomes of Chinese migrants, the negative influence of PM2.5 on the settlement intentions of Chinese migrants with a non-agricultural "hukou" decreased at a faster speed, which indicated that the migrants with a non-agricultural "hukou" paid more attention to the air quality than those with an agricultural "hukou" when deciding whether to settle down at a location in the city. In other words, in order to get higher incomes, Chinese migrants with an agricultural "hukou" were more tolerant of poorer air quality than those with a nonagricultural "hukou." The main reason for this was that rural workers with an agricultural "hukou" tended to have lower environmental awareness [69] and paid more attention to incomes in the settlement decision. Gu and Ma [46] also found that Chinese migrants show an indifferent attitude toward the environmental problems in their immigratory cities in their case study of Shenzhen, which is one of the most developed cities in China. It was found that the negative influences of PM2.5 on the settlement intentions of Chinese migrants gradually became insignificant with the increase of their incomes, no matter whether they had an agricultural "hukou" or a non-agricultural "hukou," which was consistent with the conclusion drawn from Figure 1. However, with the increasing incomes of Chinese migrants, the negative influence of PM2.5 on the settlement intentions of Chinese migrants with a non-agricultural "hukou" decreased at a faster speed, which indicated that the migrants with a non-agricultural "hukou" paid more attention to the air quality than those with an agricultural "hukou" when deciding whether to settle down at a location in the city. In other words, in order to get higher incomes, Chinese migrants with an agricultural "hukou" were more tolerant of poorer air quality than those with a non-agricultural "hukou." The main reason for this was that rural workers with an agricultural "hukou" tended to have lower environmental awareness [69] and paid more attention to incomes in the settlement decision. Gu and Ma [46] also found that Chinese migrants show an indifferent attitude toward the environmental problems in their immigratory cities in their case study of Shenzhen, which is one of the most developed cities in China.

where the coefficient *γd* (*d* = 1, 2, 3) of the interaction term *PM2.5 × income × distanced* represents the preferences for income and air quality in the settlement decision of migrants with different migration distances. Column (2) in Table 5 reports the estimated results of model (4), which shows that the coefficient of the interaction term *PM2.5 × income × distanced* was significantly positive. The average marginal effects of PM2.5 for migrants with different types of migration distances are shown in Figure

migration distances, the following model based on model (2) was estimated:

ଷ

ௗୀଵ

3.

= + ଵ2. 5 +ௗ2. 5 × ×

(2) Heterogeneity of the migrants' migration distance. According to the migration distances, the sample was divided into three subsamples of intercounty migration (within the same city), intercity

ௗ

+ + + ௧ (4)


**Table 5.** Heterogeneity test for factors influencing the settlement intentions of Chinese migrants in cities with interaction terms included.

Note: The robust standard errors are given in parentheses. \*, \*\*, and \*\*\* indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The abbreviations are defined in Table 2.

(2) Heterogeneity of the migrants' migration distance. According to the migration distances, the sample was divided into three subsamples of intercounty migration (within the same city), intercity migration (within the same province), and interprovince migration. In order to test the heterogeneous

effects of both income and air quality on the settlement intentions of the migrants with different migration distances, the following model based on model (2) was estimated:

$$\text{SI}\_{\text{ij}} = \gamma\_0 + \gamma\_1 \text{PM2.5}\_{\text{ij}} + \sum\_{d=1}^{3} \gamma\_d \text{PM2.5}\_{\text{ij}} \times income\_{\text{ij}} \times distanc\_{\text{ij}}^d + \lambda X + \rho Z + \mu\_{\text{il}} \tag{4}$$

where the coefficient <sup>γ</sup>*<sup>d</sup>* (*<sup>d</sup>* <sup>=</sup> 1, 2, 3) of the interaction term *PM2.5* <sup>×</sup> *income* <sup>×</sup> *distance<sup>d</sup>* represents the preferences for income and air quality in the settlement decision of migrants with different migration distances. Column (2) in Table 5 reports the estimated results of model (4), which shows that the coefficient of the interaction term *PM2.5* <sup>×</sup> *income* <sup>×</sup> *distance<sup>d</sup>* was significantly positive. The average marginal effects of PM2.5 for migrants with different types of migration distances are shown in Figure 3. *Int. J. Environ. Res. Public Health* **2020**, *17*, x FOR PEER REVIEW 11 of 18

**Figure 3.** Average marginal effects of PM2.5 with 90% confidence intervals for migrants migrating over different migration distances. **Figure 3.** Average marginal effects of PM2.5 with 90% confidence intervals for migrants migrating over different migration distances.

According to Figure 3, with the increase of income, the negative influence of PM2.5 on the migrants' settlement intentions decreased the fastest in the intercounty (within the same city) samples, followed by the intercity (within the same province) samples, and finally the interprovince samples. That is to say, migrants with a longer migration distance preferred income and were less likely to substitute income for air quality in their settlement decisions. The main reason for this was that 92.73% of Chinese migrants in the sample made their migration decisions for economic reasons, such as working or doing business, to improve their incomes. For all migrants with economic reasons for migrating, the migrants that undertook interprovince migration, intercity migration, and intercounty migration accounted for 50.14%, 32.83%, and 17.03%, respectively. Hence, migrants with a longer migration distance paid more attention to income than air quality. According to Figure 3, with the increase of income, the negative influence of PM2.5 on the migrants' settlement intentions decreased the fastest in the intercounty (within the same city) samples, followed by the intercity (within the same province) samples, and finally the interprovince samples. That is to say, migrants with a longer migration distance preferred income and were less likely to substitute income for air quality in their settlement decisions. The main reason for this was that 92.73% of Chinese migrants in the sample made their migration decisions for economic reasons, such as working or doing business, to improve their incomes. For all migrants with economic reasons for migrating, the migrants that undertook interprovince migration, intercity migration, and intercounty migration accounted for 50.14%, 32.83%, and 17.03%, respectively. Hence, migrants with a longer migration distance paid more attention to income than air quality.

#### **Table 5.** Heterogeneity test for factors influencing the settlement intentions of Chinese migrants in *3.3. Robustness Test*

cities with interaction terms included. **Variables (1) (2)**  *PM2.5* −0.004 \*\*\* −0.005 \*\*\* (0.001) (0.001) This study used different estimation methods. In particular, the logistic regression model was used to estimate model (2), the results of which are reported in the previous tables. Next, this study further used the probit regression model to re-estimate the model. Column (1) in Table 6 reports the estimated results. It was found that the coefficient of the variable *PM2.5* was significantly negative

*PM2.5 × income × distance1* 0.003 \*

*PM2.5 × income × distance2* 0.005 \*\*\*

*PM2.5 × income × distance3* 0.006 \*\*\*

(0.002)

*income* 0.563 \*\*\* 0.552 \*\*\*

*nation* −0.102 \*\*\* −0.104 \*\*\*

*gender* −0.238 \*\*\* −0.240 \*\*\*

(0.002)

(0.002)

(0.002)

(0.090) (0.090)

(0.029) (0.029)

(0.015) (0.015)

*PM2.5 × income × hukou0* 0.0040 \*\*

at the level of 1%, the coefficient of the variable *income* was significantly positive at the level of 1%, and the coefficient of the interaction term *PM2.5* × *income* was significantly positive at the level of 1%. As for the average marginal effect of the variable *PM2.5*, it is clearly shown in Figure 4 that the negative influences of PM2.5 on the settlement intentions of the migrants gradually became insignificant with the increase of income, which was consistent with the results presented in Table 4.


**Table 6.** Robustness test for the factors influencing the settlement intentions of Chinese migrants in cities using a probit regression model and the instrumental variables.

Note: The robust standard errors are given in parentheses. \*\*, and \*\*\* indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The abbreviations are defined in Table 2.

*Int. J. Environ. Res. Public Health* **2020**, *17*, x FOR PEER REVIEW 13 of 18

**Figure 4.** Average marginal effects of PM2.5 with 90% confidence intervals using the probit regression model. **Figure 4.** Average marginal effects of PM2.5 with 90% confidence intervals using the probit regression model.

Next, this study controlled for the endogeneity in the variable *PM2.5*. The baseline regression used in this study was the microlevel cross-sectional data model. However, the variable *PM2.5* was a variable at the city level. The PM2.5 concentration of a city may be correlated with other unobservable factors affecting migrants' settlement intentions in the residuals *μij*, thus causing potential endogeneity problems. In order to control for this potential endogeneity, instrumental variables were used to re-estimate the probit regression model. In terms of the selection of the instrumental variables, this study calculated the average PM2.5 concentrations 2.5 of each city during 2015–2016 as the instrumental variable of *PM2.5*, and used 2.5 × as the instrumental variable of *PM2.5 × income* accordingly. The PM2.5 concentration of a city tends to be relatively stable in the short term, where the correlation coefficient of *PM2.5* and 2.5 in the sample was 0.96, and thus the instrumental variables were highly correlated with the endogenous variables. Furthermore, since 2.5 was a lagging variable, the correlation of 2.5 and the city level unobservable factors in the residual term could be theoretically excluded; therefore, 2.5 satisfied the condition for an instrumental variable. The column (2) in Table 6 shows the estimation results using the instrumental variables, where it was found that the results were basically consistent with the results of column (1) in Table 6. Therefore, the robustness of the relationship between the influences of air quality and income on migrants' settlement intentions was confirmed. Next, this study controlled for the endogeneity in the variable *PM2.5*. The baseline regression used in this study was the microlevel cross-sectional data model. However, the variable *PM2.5* was a variable at the city level. The PM2.5 concentration of a city may be correlated with other unobservable factors affecting migrants' settlement intentions in the residuals µ*ij*, thus causing potential endogeneity problems. In order to control for this potential endogeneity, instrumental variables were used to re-estimate the probit regression model. In terms of the selection of the instrumental variables, this study calculated the average PM2.5 concentrations *PM*2.5 of each city during 2015–2016 as the instrumental variable of *PM2.5*, and used *PM*2.5 × *income* as the instrumental variable of *PM2.5* ×*income* accordingly. The PM2.5 concentration of a city tends to be relatively stable in the short term, where the correlation coefficient of *PM2.5* and *PM*2.5 in the sample was 0.96, and thus the instrumental variables were highly correlated with the endogenous variables. Furthermore, since *PM*2.5 was alagging variable, the correlation of *PM*2.5 and the city level unobservable factors in the residual term µ*ij* could be theoretically excluded; therefore, *PM*2.5 satisfied the condition for an instrumental variable. The column (2) in Table 6 shows the estimation results using the instrumental variables, where it was found that the results were basically consistent with the results of column (1) in Table 6. Therefore, the robustness of the relationship between the influences of air quality and income on migrants' settlement intentions was confirmed.

#### **Table 6.** Robustness test for the factors influencing the settlement intentions of Chinese migrants in *3.4. Discussion*

cities using a probit regression model and the instrumental variables. **Variables (1) (2)**  *PM2.5* −0.002 \*\*\* −0.005 \*\*\* (0.001) (0.001) *PM2.5 × income* 0.003 \*\*\* 0.002 \*\* (0.001) (0.001) *income* 0.320 \*\*\* 0.357 \*\*\* (0.053) (0.054) *nation* −0.058 \*\*\* −0.061 \*\*\* With the gradual progress of the Chinese urbanization process, it is crucial to advance the smooth integration of the migrants into cities to further achieve sustainable development in terms of both modernization and urbanization. For emerging countries with a relatively low urbanization ratio, improving the living environment and income level of the migrants can increase their settlement intentions and integration into city lives. The empirical research based on the Chinese dataset used in this study indicated that the migrants often faced a trade-off between income and air quality in their settlement intentions. In varying degrees, the migrants tended to tolerate poorer air quality to obtain higher incomes. In order to break through this dilemma, it is necessary to find a way out of

*gender* −0.140 \*\*\* −0.141 \*\*\*

(0.017) (0.017)

(0.009) (0.009)

the traditional development pattern indicated by the left side of the environmental Kuznets curve, in which economic growth tends to aggravate environmental pollution.

On the one hand, green and environmentally friendly industries should be encouraged in the cities of emerging industrializing countries. The coordinated development model of both economic growth and environmental protection needs to be further explored. At the same time, the government should strengthen environmental regulations, reinforce the prevention and control of environmental pollution, and keep engaging in the continuous improvement of the environment during economic development. Furthermore, it is of great importance to take environmental governance as one of the most important goals of urban development to further improve the livability of cities for both local residents and migrants.

On the other hand, regional coordinated development models should be established to reduce the gaps in both income inequality and environmental awareness between Chinese cities. At present, there exist significant differences in economic development and the level of the livability of cities in China. Therefore, migrants tend to settle down in the cities with lower air quality but higher incomes. Therefore, it is necessary to further promote coordinated economic development between cities, innovate the existing household registration system, and use the development of urban agglomerations or city clusters to drive economic development and employment in the small- and medium-sized cities surrounding metropolitans. Furthermore, the economic development and awareness of environmental health in less developed regions of emerging countries, such as China, should be promoted and encouraged. It is critical to promote the optimization of the regional industrial layout and spatial agglomeration and nurture the environmental awareness of residents to achieve an optimized balance between economic development and environmental health in emerging countries.

### **4. Conclusions**

Using a microlevel sample from the China Migrants Dynamic Survey data from 2017 and the annual average concentration of PM2.5 to measure city air quality, this study investigated the influences of both air quality and income on the settlement intentions of Chinese migrants. The results of this study showed that when making a settlement decision, Chinese migrants were faced with a trade-off between poorer air quality and higher income. Poorer air quality could significantly decrease the settlement intentions of the migrants, while a higher income could significantly increase the settlement intentions. However, the negative influences of poorer air quality on the settlement intentions of the migrants gradually decreased with the increasing income opportunities of the migrants at that location. This seems to be a bit of an unreasonable choice that was made by the migrants, as generally, people tend to live in an environmentally safer place with the rise of income. The findings implied that the Chinese rural migrants moving into cities were still financially weak and sacrificed their wellbeing for higher income opportunities. They had not yet crossed the income threshold from where they would prioritize settling down in an environmentally safer place over a location with a higher income opportunity but was environmentally less safe.

Furthermore, there existed an apparent heterogeneity in the influences of both air quality and income on the settlement intentions of the migrants with different "hukous" and migration distances. Specifically, when deciding whether to settle down in a city, Chinese migrants with a non-agricultural "hukou" paid more attention to air quality than the migrants with an agricultural "hukou," while the migrants with an agricultural "hukou" were more tolerant of poorer air quality than migrants with a non-agricultural "hukou." Furthermore, the longer the migration distance of the Chinese migrants, the more emphasis that was put on the income when making settlement decisions. In addition, this study also used different estimation methods in the robustness test and controlled for potential endogeneity using instrumental variables. The robustness of the relationship between air quality and income regarding their influences on the settlement intentions of the migrants was confirmed.

**Author Contributions:** Conceptualization, B.L., M.M., and Q.C.; methodology, B.L. and Q.C.; software, B.L. and Q.C.; validation, B.L., M.M., and Q.C.; formal analysis, B.L., M.M., and Q.C.; investigation, B.L. and M.M.; resources, B.L. and Q.C.; data curation, B.L. and Q.C.; writing—original draft preparation, B.L. and Q.C.; writing—review and editing, B.L., M.M., and Q.C.; visualization, B.L., M.M., and Q.C.; supervision, M.M.; project administration, B.L. and Q.C.; funding acquisition, B.L. and Q.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by The National Social Science Fund of China, grant number 18BJY061.

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