Skip to main content
NCBI home page
Scientific Reports logoLink to Scientific Reports
. 2025 Mar 22;15:9899. doi: 10.1038/s41598-025-95202-8

The impact of air pollution on employment mobility of college graduates in the Yangtze River Delta

Bing Zeng 1, Wenhui Gao 2,✉, Binyu Wei 2,✉, Xiaojie Shu 3
PMCID: PMC11929922  PMID: 40121367

Abstract

The group of college graduates is the top priority for stabilizing employment in China. This research explores the impact of air pollution on the employment flow of college graduates in the Yangtze River Delta by collecting and matching PM2.5 satellite data in grid form and the data from employment quality reports. The air flow coefficient is used as an instrumental variable of air pollution. Research has shown that there is a negative correlation between air pollution levels in a student’s city of study and the likelihood of college graduates staying in the region after graduation. Namely, higher levels of air pollution are associated with lower intraregional stickiness rates. Also, graduate students are more susceptible to the effects of air pollution than undergraduates. And the impact of air pollution on the employment mobility of graduates is greater for those who attended public colleges and universities compared to those who attended private institutions. The conclusions of this research demonstrate the importance of air pollution governance in relation to graduate employment mobility and provide both theoretical support and empirical evidence for local authorities to manage and improve environmental quality and graduate employment opportunities.

Keywords: Air pollution, College graduates, Employment stickiness rate, Instrumental variable method, Yangtze River Delta

Subject terms: Climate-change impacts, Socioeconomic scenarios

Introduction

Under the background of global knowledge economy competition, countries regard talent development as an important factor of national development1. Human capital is a necessary factor in urban development and innovation, and the geographical mobility of talents is reshaping the pattern of regional economic development2. Therefore, it is significant to regard talent as primary resource. Facilitating the smooth and orderly mobility of talent is crucial for stimulating creativity, innovation and entrepreneurship, as well as a key element in the workforce development strategy. Factors such as income, cost of living, and livability of the city can impact talent mobility3,4. Air quality, as one of the characteristics of urban livability, naturally affects the employment mobility of talents explicitly. Air quality is a key aspect of urban livability that can significantly impact employment mobility for talented individuals5. This research examines the correlation between air pollution and the employment mobility of college graduates. And this study focuses on this specific group of highly qualified individuals who are vital and creative in the labor market. To begin with, the exportation of these talents from higher education institutions plays a crucial role in promoting scientific and technological progress as well as industrial upgrading. The distribution and employment of college graduates is important for the optimal use of national human capital and the coordinated development of the region6. In China, the group of college graduates has become the second-largest group of migrants, following the group of migrant workers. Since the expansion of higher education in 1999, coupled with the market-based self-selection of graduates, there has been a chronic problem of difficult employment in China. Based on data from the Chinese Ministry of Education, the number of college graduates in the country has increased from 8.2 million in 2018 to 10.76 million in 2021, and is projected to reach 11.85 million in 2023. Currently, employment pressure and difficulty in China have reached unprecedented levels. On the one hand, studying the employment mobility behaviour of college graduates can assist local governments in formulating policies to promote high-quality employment opportunities for them. On the other hand, a scientific analysis of the impact of air pollution on the mobility of college graduates is conducive to a comprehensive understanding of the socio-economic value of environmental pollution governance. This research aims to present that the impact between air quality and employment mobility of college graduates and highlights the need for local government management to address air quality concerns. Meanwhile, it aims to provide empirical evidence for local authorities to manage and improve environmental quality and graduate employment opportunities.

The Yangtze River Delta is one of the regions with the most vibrant economic development, the highest degree of openness and the strongest innovation capacity in China. With the deepening of industrialization, urbanization and integration of the Yangtze River Delta, the Yangtze River Delta region is suffering from excessive emissions of pollutants such as sulfur dioxide, smoke and poeder, as well as carbon oxides. Further, it leads to the decline of urban livable environmental security and ecosystem services. This has also become a bottleneck restricting regional economic development, and it seriously affecting people’s health and life. According to the Ministry of Ecology and Environment of China in 2023, the emission of air pollutants per unit area in the Yangtze River Delta is three to five times that of the national average. The average PM2.5 concentration was 55 µg per cubic meter, an increase of 7.8% compared to the previous year (see: https://www.mee.gov.cn/ywdt/xwfb/202401/t20240125_1064784.shtml) At the same time, the Yangtze River Delta has a favorable endowment of higher education resources. The number of universities accounts for 17% of the total number of universities in China. College graduates have become the new force to promote scientific and technological innovation in the Yangtze River Delta region. Therefore, in the context of increasingly prominent air pollution, how to effectively treat the employment mobility of college graduates is a topic worth studying.

In view of this, The research begins by illustrating the impact of air pollution on the employment mobility behaviour of college graduates from a theoretical perspective, based on the utility preference theory. At the empirical level, PM2.5 concentration data were calculated using vector layer masks in the Yangtze River Delta, which served as metrics of urban air pollution levels. The employment quality report data for graduates from various universities in the Yangtze River Delta from 2019–2021 was collected manually annotated, and the dataset of employment mobility of college graduates was constructed. The study is based on data on air quality and employment mobility, as well as urban economic and geographical attribute data. The analysis investigates the effect of air pollution on the employment mobility of college graduates. And the air flow coefficient is used as an instrumental variable for air quality, while considering the escape elasticity and heterogeneity characteristics of air pollution. The empirical evidence was shown at the university and college level, regarding the impact of air quality on the employment mobility behaviors of college graduates. Also, the study enriches research on air pollution by examining the mobility of college graduates. It highlights the practical significance of improving environmental governance capacity and promoting high-quality employment opportunities for college graduates. It is also significant in terms of the practical implications of brain drain caused by air pollution. This study makes the following marginal contributions. Firstly, this study utilizes precise PM2.5 satellite data in grid form and data from the employment quality report to investigate the impact of air pollution on the employment behavior of this group. The results provide new empirical evidence for studying graduate labor mobility behavior in China. Secondly, air flow coefficients were used as instrumental variables. This method aims to exclude problems of endogeneity in the regressions, ensuring that the causal effects are clarified, and the estimation results are relatively unbiased and reliable for drawing conclusions.

Literature review and theoretical analysis

Literature review

Population mobility and employment location choices are individual decisions made to maximize utility based on costs and benefits. The behavioral impacts of population mobility have been effectively portrayed by the push–pull theory, the Tiebout model, as well as the new economic geography model7–9. Previous research has primarily investigated population mobility in relation to economic factors, such as income, taxes, and house prices, as well as non-economic factors, such as public services, dialects, and geographic conditions10–16. In recent years, scholars have increasingly focused on the impact of air pollution on population mobility. Banzhaf and Walsh17 discovered that enhanced air quality in communities encourages population migration. Using mobile population survey data and PM2.5 satellite data, Sun et al.14 examined that air pollution reduces the population mobility in cities. Chang et al.18 and Zivin and Neidell19 also confirmed that air pollution in urban areas worsens population loss, and meanwhile found that air pollution has a negative influence on worker productivity. Population mobility involves group differences. Therefore, some researchers have started to investigate the relationship between air pollution and the mobility behavior of talented groups. Lin et al.20 investigated the impact of air pollution on the working place choice by using regression discontinuity design, which have shown the labor mobility and mismatch caused by air pollution. Lan et al.21 studied the impact between pollution emissions and the agglomeration of high-level human capital from the perspective of different level of environmental effects. On the other hand, there is research focused on the impact of environmental features on the issue of international students’ mobility22. Based on the reviewed literature, few scholars combined the issue about air pollution and the employment mobility of college graduates.

On the other hand, there is a significant body of literature on the spatial patterns and influencing factors of college graduates’ mobility. Regarding spatial patterns, Comunian and Jewell23 used higher education student micro-data from UK, to illustrate the migration behavior of graduates. There is also studies focused on ‘first-class’ university graduates in China, to explore the human capital migration patterns and trends and its impacts on reginal development with the method of negative binomial model24. Most studies on the aspect of factors influencing the mobility of college graduates focus on economic and urban comfort factors25–27. Liu et al.28 analyzed data from the nationwide population census to determine the impact of education and healthcare on graduates’ employment mobility in China. Scholars have shown that individual factors, such as intergenerational mobility, gender differences, and social stratification, significantly impact graduate mobility29–31. Zheng et al.32 discovered the crowd-out effect of air pollution on graduate mobility. However, the data was limited to a sample from Tsinghua University, which lacks the breadth of the region and the population. To extrapolate the research findings to other colleges, universities and regions, extensive sample size will be used in this research.

This study focuses on the group of college graduates from the aspect of university, rather than previously on the impact of air pollution on talent mobility. This study adopts comprehensive employment data from college graduates, making it more representative and generalizable. Meanwhile, considering the complexity and dynamics of the employment city choices of college graduates, this study uses the instrumental variable method to address the endogeneity problem stochastically and effectively.

Theoretical analysis and research hypotheses

The group of college graduates may face cost constraints when it comes to employment mobility and seeking better opportunities to maximize the expected utility. Referring to Lai et al.33, research on the relationship between air pollution and employment mobility in higher education, the theoretical analysis is as follows.

Assume that graduates with the level of ability Inline graphic are employed in region Inline graphic at a wage of Inline graphic. Graduates with higher ability normally tend to be paid relatively more, namely Inline graphic. Inline graphic is the migration cost of college graduates from place of study Inline graphic to employment area Inline graphic due to air pollution. The other additional impacts are Inline graphic, with mutually independent homogeneous distributions as well as the Gumbel distribution. The revenue function of college graduates for employment mobility behaviors is:

graphic file with name d33e376.gif 1

The Inline graphic reflects the coefficient of influence degree of Inline graphic. Inline graphic is the utility function of a college graduate at level Inline graphic in region Inline graphic, which is defined as Inline graphic. The utility function is related to wage Inline graphic, price level Inline graphic, and air pollution disutility Inline graphic. Therefore, college graduates choose the place of employment and decide whether to leave the place of study, based on the influence of wage levels, price level, air pollution, and migration cost.

It is further assumed that Inline graphic is the ratio of college graduates (who with level of ability Inline graphic), leave Inline graphic to Inline graphic. This ratio is also expressing the meaning of the possibility of higher earnings from working in area Inline graphic than in other areas (including Inline graphic).

graphic file with name d33e478.gif 2

In the above formula, Inline graphic. And the Inline graphic shows the migration elasticity. So, it is obvious that areas with a high level of utility (e.g., higher wage levels and less air pollution) attract more graduates to relocate.

Further, the outflow of graduates from Inline graphic is as following:

graphic file with name d33e506.gif 3

The derivation of air pollution disutility Inline graphic from (3) is as following:

graphic file with name d33e520.gif 4

Therefore, Hypothesis 1: Higher levels of air pollution in area of study increase the probability of college graduates leaving for employment in other areas.

The derivation of graduates with the level of ability Inline graphic from (4) is as following:

graphic file with name d33e536.gif 5

Therefore, Hypothesis 2: Graduates with higher levels of ability are more likely to be affected by air pollution and consider leaving the place of study than graduates with lower levels of competence.

Empirical design

Causal recognition strategy

The econometric model can be used to examine the impact of air pollution on the employment mobility of college graduates.

graphic file with name d33e550.gif 1

In formula, GRA represents the employment mobility, c represents city, i represents colleges and universities, t represents year. Air pollution is represented by the variable AIR, and X represents the corresponding control variable. ε is the random error. The city fixed effect λc controls for unobservable factors that do not change over time in the city. The year fixed effect ηt controls for macroeconomic policy shocks in each year. The regression coefficient β indicates the impact of air pollution on the employment mobility of college graduates. A negative coefficient indicates that as air pollution levels increase in the city of study, the rate of college graduates remaining in the area decreases. So, college graduates in the city are more likely to leave the city of study and seek employment in cities with better air quality.

Direct estimation of the relationship between air pollution and employment mobility for college graduates is prone to estimated deviation and misinterpretation. There is an endogeneity problem with the model. On the one hand, college graduates may choose their employment cities based on air quality, which can affect the supply and demand relationship in the city’s talent market. The supply and demand relationship will in turn influence the employment site selection of college graduates through the wage level. On the other hand, there may be economic co-drivers factors between air pollution and employment mobility. For instance, the growth of relevant industries not only leads to air pollution, but also attracts local college graduates for employment. The regression coefficients for variables of air pollution and industrial expansion would be numerically different. Endogeneity problems can also arise from missing variables or unobservable latent variables. Therefore, this study chose the instrumental variables method to minimize estimation bias. Instrumental variables are typically selected based on two main criteria. One, it must be associated with the endogenous explanatory variable, air pollution. The other is that the instrumental variable is exogenous.

Meteorological conditions are often used as instrumental variables for air pollution in some literature. Arceo et al.34 used the intensity of inversion as an instrumental variable on the grounds that pollutants are less likely to disperse with temperature inversion, which increasing air pollution. Some scholars have also used wind direction as an instrumental variable because it increases air pollution levels in downwind areas33. Hering and Poncet35 analyzed the interaction between wind speeds and heights of planetary boundary layer for each city and year to obtain air flow coefficients, which is named LIU in this study. The coefficients were then used as instrumental variables for air pollution. Due to the increase in wind speeds and boundary layer, pollutants are more likely to disperse both transversely and longitudinally. Compared with variables such as intensity of inversion and wind direction, air flow coefficient has both horizontal and vertical explanatory power. The air flow coefficient more fully conform to the condition for correlation of instrumental variables. And both wind speeds and heights of planetary boundary layer are determined by complex meteorological systems and geographic conditions. These factors are not related to economic activities such as economic growth, capital, and technology, which effectively satisfies the exogeneity assumption of instrumental variables36. Therefore, air flow coefficients (namely LIU) are also used as instrumental variables in this study.

Two-stage least squares (2SLS) was used to carry out further regression analysis with instrumental variables. In the first stage, the regression model of instrumental variable (LIU) and endogenous variable (air pollution) is constructed. A fitted value for air pollution is obtained that reflects the exogenous part of the variable air pollution. The model is as follows:

graphic file with name d33e612.gif 2

In the second stage, the fitted values of air pollution are used in the following regression model. The coefficient Inline graphic indicates that it is a reasonable assessment of the impact of air pollution on the employment mobility of college graduates.

graphic file with name d33e626.gif 3

Sample selection and data source

  1. Core explained variable: employment mobility of college graduates. The graduates in this study focus mainly on Chinese graduates with an undergraduate degree or above. They are from full-time colleges and universities accredited by the Chinese Ministry of Education. Data on employment mobility are taken from the employment quality annual reports of graduates, published on the official websites of higher education institutions. The report presents an objective reflection of the basic situation, main characteristics, and development trend in graduates’ employment. It provides valuable data for the study of employment mobility among college graduates. In this study, employment mobility refers to the movement of graduates from their place of study to the first place of employment. The data does not include other types of mobility, such as pursuing further education, going abroad, or waiting for employment. Here in this study, use the employment stickiness rate to measure the mobility of college graduates. The indicator reflects the attractiveness of the city to local college graduates. The calculation formula is as follows:
    graphic file with name d33e642.gif 4
    Inline graphic is the number of graduates from college located at city i who employed in the city of study and Inline graphic is the total number of employment.

  2. Core explanatory variable: air pollution. In this study, the annual average concentration of PM2.5 is used to represent air pollution. It is an important indicator not only for national ambient air quality standards but also for relevant studies worldwide. The website (http://fizz.phys.dal.ca/~atmos/martin/?page_id=140) provides corresponding data in grid form. In this study, PM2.5 concentration data (mg/m3) were calculated using vector layer masks in the Yangtze River Delta by using ArcGIS10.2

  3. Instrumental variables: Using wind speeds and heights of planetary boundary layer represent air flow indicator, which is released by European Centre for Medium-Range Weather Forecasts (ECMWF: http://apps.ecmwf.int/datasets/data/interim-full-moda/levtype=sfc/). The vector map data of Yangtze River Delta were masked. Tools such as grid counter and zonal statistics were used to extract the mean values of annual indicators at the prefecture-level city level.

  4. Control variables: The main three levels of control variables are included: university, physical geography and city.

Regarding the control variables at the university level: (1) university’s strength (STR)33, characterized by whether it is a ‘Double First-Class’ initiative university; (2) location advantage (LOC)37, characterized by whether it is in the capital of the province; (3) historical heritage (HIS)37,38: represented by length of establishment; ④scale of operation (SCA), represented by the number of teachers and students in the university37,38.

As for control variables at the topographic features, it was considered that the air coefficients are influenced by natural geographical condition factors. The data for three natural geographical conditions were used: (1) the degree of topographic relief (TOP)39; (2) the precipitation (PRE)40; (3) the temperature (TEM)41.

As for the control variables at the city characteristics, referring to the relevant literature and taking into account the availability of data, the following variables were used: (1) level of economic development (ECO)11, represented by the logarithm of real GDP per capita and deflated by prices; (2) industrial structure (IND)42, represented by the proportion of secondary and tertiary output in total output; (3) level of innovative (INN)43, represented by the logarithm of the number of patents granted for inventions; (4) wage level (WAG)43,44, represented in the form of the logarithm of the average wage of active employees; (5) level of public services (PUB)43,44, represented by the logarithm of per capita expenditure on medical and health care.

  • (5)

    Data Source: This study focuses on a sample of colleges and universities located in 41 cities within the Yangtze River Delta region. The Yangtze River Delta is a region of economic dynamism in China, and as such, with generous endowment of higher education resources. It is important to consider the representativeness of universities in this region. Secondly, data accessibility is a concern. The sample data for Yangtze River Delta colleges and universities is comprehensive. Thirdly, to reduce endogeneity. Compared to using data from a National sample of cities, there is a risk of omitted variable bias due to factors such as the complex and diverse natural and economic elements.

This study collects and sorts of data from 219 colleges and universities (excluding data from Sino-foreign Cooperative Education) to provide accessible and consistent information. Among them, 39 are in Shanghai, 76 are in Jiangsu, 58 are in Zhejiang, and 46 are in Anhui, spanning from 2019 to 2021. To ensure the timeliness of employment quality reports, this study uses one-period-ahead data for processing and matching. The approach avoids the issue of reverse causation of the variables. The data is sourced from the China City Statistical Yearbook. Missing values in individual samples are filled by combining local statistical yearbooks, statistical bulletins, or linear interpolation. Data at the college and university level is sourced from employment quality annual reports, which are published on official websites. Data at the physical geography level is sources from the Chinese Relief Factor Degree of Land Surface (Rdls) and the China Integrated Meteorological Information Service System (CIMISS). Nested data sets are prepared after processing and matching the data.

Empirical results analysis

Status analysis

Firstly, ArcGIS 10.2 is used to visually analyze the air pollution and the situation of universities in the Yangtze River Delta in 2021 (Fig. 1). From the perspective of air pollution, the overall haze pollution in the Yangtze River Delta urban agglomeration is generally high in the northwest and low in the southeast. From the perspective of universities, undergraduate universities in the Yangtze River Delta region basically cover the whole region of the Yangtze River Delta. However, there are significant differences in the number and the school running system of universities and colleges in different provinces. ‘Double First-Class’ initiative universities are concentrated in provincial capitals, municipalities or regional central cities.

Fig. 1.

Fig. 1

Open in a new tab

Distribution of air pollution and universities in the Yangtze River Delta in 2021. Match location, situation of universities and colleges and PM2.5. This map shows the 41 cities in the Yangtze River Delta region. The map is based on the standard map produced under the supervision of the Ministry of Natural Resoures of the Peoples Republic of China: Approval number GS (2020)3183 (https://bzdt.ch.mnr.gov.cn/), made with ArcGIS 10.2 software (https://www.esri.com/).

Regression results and analysis

OLS regression is used to preliminarily verify the influence of air pollution (AIR) on stickiness rates (GRA). The regression results, including control variables, are presented in Table 1. The coefficients of PM2.5 are all found to be less than 0, which is significant at the 1% level. These results indicate that PM2.5 has a significant negative effect on stickiness rates. College graduates are more willing to pay for clean air as air pollution becomes more severe. The cost of employment migration is relatively lower, and individuals are more likely to leave their place of study and move to other cities with less air pollution.

Table 1.

OLS regression results.

(1) (2) (3) (4)
AIR

− 0.0112***

(0.0014)

− 0.0109***

(0.0014)

− 0.0108***

(0.0025)

− 0.0057***

(0.0016)
STR

0.1535***

(0.0257)

0.0906***

(0.0245)

0.0303**

(0.0142)
LOC

0.0899***

(0.0252)

0.0667**

(0.0304)

0.0936***

(0.0169)
HIS

0.0012

(0.0011)

0.0009

(0.0008)

0.0003

(0.0004)
SCA

− 0.1116***

(0.0255)

− 0.1594***

(0.0272)

− 0.1424***

(0.0159)
TOP

− 0.4146***

(0.0446)

− 0.1200***

(0.0322)
PRE

0.0021***

(0.0005)

0.0006**

(0.0003)
TEM

− 0.0490***

(0.0178)

− 0.0520***

(0.0129)
ECO

0.0985***

(0.0276)
IND

0.0025

(0.0039)
INN

0.0338**

(0.0154)
WAG

0.2610***

(0.0446)
PUB

0.1274***

(0.0191)
Fixed effects Yes Yes Yes Yes
R2 0.1229 0.1895 0.3343 0.8022
N 439 439 439 435
Open in a new tab

***p < 0.01, **p < 0.05, *p < 0.1. Robust standard errors in brackets parenthesis. The same below.

Further, the air flow coefficient is used as an instrumental variable for air pollution. Comparing the results of the regressions in column (4) of Table 1 with those in Table 2 it is evident that PM2.5 has a significant impact on the employment stickiness of local college graduates in the 2SLS regression. The coefficients in 2SLS are over twice the size of those in column (4) of the benchmark regression. It is obviously that the benchmark regression estimates will significantly underestimate the negative impact of air pollution on the employment stickiness rate of graduates due to the issue of endogeneity. Therefore, the use of the air flow coefficient as an instrumental variable is scientifically sound. Using instrumental variables ensures consistent and valid regression results, providing a scientific view of the impact of air pollution on college graduates employment mobility.

Table 2.

2SLS regression results.

Variable First stage Second stage
LIU

− 0.0060***

(0.0007)
AIR

− 0.0124***

(0.0040)
University characteristic controls Yes Yes
Topographic features controls Yes Yes
City characteristics controls Yes Yes
Fixed effects Yes Yes
Underidentification test LM value 69.645***
Weak identification test F value 80.252***
N 435
Open in a new tab

***p < 0.01, **p < 0.05, *p < 0.1. Constant terms and control variables are not shown in this table, without affecting the overall results. The same below.

With the air flow coefficient (LIU) as an instrumental variable, the results of the first stage regression shows that the air flow coefficient is negative at the 1% significance level. The air pollution level (AIR) decreases as the air flow coefficient increases. Namely, the correlation assumption of instrumental variables is satisfied. The under identification test result indicates a value of 69.645, which significantly rejects the null hypothesis, which the instrumental variables are not identifiable. The weak identification test results indicate that the F-statistic value of 80.252 is significantly higher than the critical value of 16.38, indicating the absence of weak instruments. In summary, these test results support the use of instrumental variables in this study. The exogeneity of the instrumental variables is further discussed. By including both air flow coefficient and air pollution in the regression model, Table 3 shows that the air flow coefficient (LIU) is no longer significant and tends towards 0. The results indicates that the air flow coefficient has no significant effect on the stickiness rate of college graduates. Thus, using the air flow coefficient as an instrumental variable satisfies the exogeneity assumption.

Table 3.

Exogeneity test results.

Variable Regression results
AIR

− 0.0044**

(0.0017)
LIU

0.0000

(0.0000)
University characteristic controls Yes
Topographic features controls Yes
City characteristics controls Yes
Fixed effects Yes
R2 0.8038
N 435
Open in a new tab

***p < 0.01, **p < 0.05, *p < 0.1.

Robustness test

Firstly, the 1% and 5% winsorization of dependent variable is used to account for the impact of extreme values. After excluding extreme values, the regression results (Table 4) show that air pollution continues to significantly reduce the employment stickiness rate of college graduates. The regression results show that the coefficients are consistent with Table 3, which is significant at the 1% level. It further illustrates the robustness of the results of this study. Secondly, this study replaces the core explanatory variables and re-regresses the data. In addition to PM2.5, other common indicators to measure air pollution include PM10 and AQI. PM10 is an indicator for respirable particulate matter with a diameter of 10 microns or less. The Air Quality Index (AQI) is a non-dimensional parameter that quantitatively describes the air quality situation. It includes six pollutants: SO2, NO2, PM10, PM2.5, O3, CO. Even though the corresponding air pollution indicators were replaced, the results show that air pollution still significantly reduces the employment stickiness rate of college graduates. Therefore, the regression results are relatively robust. It should be noted that the regression coefficients for PM10 and AQI are significantly smaller than those for pm2.5. Because of the small particle size, PM2.5 is easier to be retained in the bronchioli terminales and pulmonary alveoli. PM2.5 is a more significant contributor to air pollution than PM10 and is more likely to affect the employment rate of college graduates. The Air Quality Index (AQI), on the other hand, has a smaller coefficient due to different standards of measurement.

Table 4.

Rubustness test results.

1% level winsorization 5% level winsorization PM10 AQI
AIR

− 0.0122***

(0.0039)

− 0.0113***

(0.0038)

− 0.0019***

(0.0006)

− 0.0039***

(0.0013)
University characteristic controls Yes Yes Yes Yes
Topographic features controls Yes Yes Yes Yes
City characteristics controls Yes Yes Yes Yes
Fixed effects Yes Yes Yes Yes
Under identification test LM value 69.645*** 69.645*** 162.247*** 106.954***
Weak identification test F value 80.252*** 80.252*** 250.432*** 137.260***
N 435 435 435 435
Open in a new tab

***p < 0.01, **p < 0.05, *p < 0.1. The results presented only include the regression results from the second stage. The same below.

Heterogeneity analysis

According to Hypothesis 2, there is heterogeneity in the degree of sensitivity to air pollution among college graduates with varying levels of ability in employment mobility. This study examines the heterogeneous effect of ability level from two perspectives. Firstly, the impact of individual educational background heterogeneity is considered. The sample is divided into undergraduates and postgraduates for employment mobility, then analyzed separately. Secondly, consider the heterogeneity effects from the nature of education. The sample of graduates is divided into those from public universities and those from private universities for the analysis of employment mobility.

As shown in Table 5 of the regression results, air pollution has a significant inhibitory effect on the employment stickiness rate of both undergraduate and postgraduate graduates. Air pollution has a greater impact on the stickiness rate of postgraduate graduates. Also, air pollution is more likely to prompt local postgraduate graduates to leave their place of study. Postgraduate graduates have higher levels of education and more employment options. Because of higher level of educational background, they are less impacted by changes in labor market demand and can earn higher wages. Meanwhile, they are also more conscious of the health impacts of air pollution. The increase in wages results in a relatively lower marginal utility compared to the improvement in air quality. Thus, if the level of air pollution is high in their place of study, they are more motivated to relocate and seek employment in areas with better air quality.

Table 5.

Heterogeneity regression.

Educational background Nature of education
Undergraduates Postgraduates From public universities From private universities
AIR

− 0.0136***

(0.0047)

− 0.0488***

(0.0179)

− 0,0125***

(0.0035)

0.0571

(0.1735)
University characteristic controls Yes Yes Yes Yes
Topographic features controls Yes Yes Yes Yes
City characteristics controls Yes Yes Yes Yes
Fixed effects Yes Yes Yes Yes
Underidentification test LM value 54.902*** 10.589*** 71.393*** 0.216
Weak identification test F value 63.569*** 10.314*** 86.591*** 0.186
N 314 142 341 94
Open in a new tab

***p < 0.01, **p < 0.05, *p < 0.1.

As for the perspective of universities, air pollution still has a significant inhibitory effect on the employment stickiness rate of graduates from public colleges and universities. However, the regression results do not provide sufficient evidence to demonstrate that increased air pollution inhibits the employment stickiness rate of graduates from private colleges and universities. Instead, there is a non-significant positive coefficient, which is not contrary to general. It is not difficult to understand that private colleges and universities are established by law for individuals or enterprises and operated independently. The schooling philosophy and management mode tend to be more market-oriented and localized, focusing on the input–output ratio and economic benefits. As a result, graduates of private colleges and universities are often preferred in the local employment market. The structural differences between private and public colleges and universities may result in hierarchical distinctions among graduates in terms of employment opportunities and bargaining power for wages3. In general, graduates from private colleges have lower employment prospects in other regions compared to graduates from public colleges and universities. Therefore, in the same air pollution situation, private college graduates with weaker employment opportunities and wage bargaining power are less likely to avoid the negative effects of air pollution. Together with the attraction of the labor market in the place of study, employment mobility of graduates from private college is less sensitive to air pollution.

Conclusions and enlightenment

In recent years, air pollution has become a significant concern for all sectors of society. Air quality has become a core factor in cities’ ability to attract and retain human capital45. Graduates are a dynamic and creative group as part of human capital. The issue of employment mobility has received significant attention. Therefore, it is important to scientifically understand the impact of air pollution on the employment mobility of college graduates. This study is based on data such as PM2.5 satellite data in grid form and the employment quality report of college graduates in the Yangtze River Delta. The 2SLS method uses the air flow coefficient as an instrumental variable for air quality. The study analyzed the impact of air pollution on the employment mobility of college graduates. In conclusion, firstly, air pollution has a significant negative effect on the employment stickiness of college graduates. Increasing air pollution levels will repel local college graduates from employment opportunities. Secondly, under the same air pollution conditions, postgraduate graduates have a higher ‘escape elasticity’ of employment mobility compared to undergraduate graduates. Graduates from public universities exhibit a significant ‘escape elasticity’ to air pollution, whereas private university graduates do not show significant ‘escape elasticity’.

The study’s findings suggest that improving environmental quality is crucial for the Yangtze River Delta and even Chinese cities to attract talent. It is also valuable for promoting high-quality employment for graduates and for comprehensively assessing the socio-economic value of investing in environmental management. On the one hand, talents are regarded as primary resource. To compete for talent resources, various policies on income and welfare incentives have been introduced. At the same time, it is also necessary to focus on the air quality for the Yangtze River Delta region. It is important to consider negative impact of air pollution to acknowledge the complexity of the talent flow phenomenon and respond positively to the increasingly competitive talent environment46. On the other hand, certain employment groups, such as college graduates, play a crucial role in ensuring stability in employment. Improving air quality could potentially alleviate employment problems faced by college graduates. So, the improvement of air quality can be taken into consideration in the performance evaluation system of local governments in the Yangtze River Delta. The Yangtze River Delta region should jointly deal with and control air pollution issues, such as by establishing a cross-provincial and cross-municipal air quality monitoring network, which will contribute to jointly improving the air quality in the Yangtze River Delta, and make the Yangtze River Delta an important choice for college graduates’ employment. Furthermore, the heterogeneous findings presented in this study provide a more accurate understanding of the employment behavior among graduates at various levels. When implementing employment policies, it is important to focus on classification management for the Yangtze River Delta region, fully releasing the employment potential of college graduates, and establishing a long-term incentive employment service and security system policy47.

There are limitations as well. Limited by the availability of data, on the one hand, the time dimension of the data is relatively short. On the other hand, acquiring additional data on the employment mobility of graduates at the doctoral and junior college level is not feasible. Further study in the future is about more detailed analysis of group behavior. This study still has some limitations in terms of the reasons for how air pollution affects graduate employment. Although this study has made a preliminary attempt. For example, the influence of utility preference and university strength is found. However, limited by the setting of the data collection section of the university department, the variable data of the cause level cannot go deeper. There are research topics that can be further expanded in the future. For example, further researches about under the influence of air pollution, how do graduates stay in the city where their universities are located and how do they choose their working and living places within the city.

Author contributions

Bing Zeng has made a contribution to the concept and design of the paper in the writing,empirical analysis and drafted the paper. Wenhui Gao and Binyu Wei collected data, analyzed data and carefully revised and polished the language of paper. Xiaojie Shu collected relevant literature and data, adjusted article format.

Funding

This work is supported by Anhui Office of Philosophy and Social Science, China(AHSKY2022D094) and the National Natural Science Fund Project of China(72163010).

Data availability

All data generated or analyzed during this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Contributor Information

Wenhui Gao, Email: 106595946@qq.com.

Binyu Wei, Email: 492521634@qq.com.

References

  • 1.Glaeser, E. L., Ponzetto, G. A. M. & Tobio, K. Cities, skills and regional change. Reg. Stud.48, 7 (2012). [Google Scholar]
  • 2.Lawton Smith, H., Glasson, J. & Chadwick, A. The geography of talent: Entrepreneurship and local economic development in Oxfordshire. Entrep. Reg. Dev.17(6), 449–478 (2005). [Google Scholar]
  • 3.Faggian, A. & McCann, P. Universities, agglomerations and graduate human capital mobility. Tijdschr. Econ. Soc. Geogr.100(2), 210–223 (2009). [Google Scholar]
  • 4.Starr, E., Ganco, M. & Campbell, B. A. Strategic human capital management in the context of cross-industry and within-industry mobility frictions. Strateg. Manag. J.39(8), 2226–2254 (2018). [Google Scholar]
  • 5.Park, Y. M. & Kwan, M. P. Individual exposure estimates may be erroneous when spatiotemporal variability of air pollution and human mobility are ignored. Health Place43, 85–94 (2017). [DOI] [PubMed] [Google Scholar]
  • 6.Glaeser, E. L. & Gottlieb, J. D. Urban resurgence and the consumer city. Urban Stud.43(8), 1275–1299 (2006). [Google Scholar]
  • 7.Bougue, D. J. Principle of demography (Willey, 1969). [Google Scholar]
  • 8.Tiebout, C. M. A pure theory of local expenditures. J. Polit. Econ.64(5), 416–424 (1956). [Google Scholar]
  • 9.Krugman, P. Increasing returns and economic geography. J. Polit. Econ.99(3), 483–499 (1991). [Google Scholar]
  • 10.Lei, W., Jiao, L., Xu, Z., Zhou, Z. & Xu, G. Scaling of urban economic outputs: Insights both from urban population size and population mobility. Comput. Environ. Urban Syst.88, 101657 (2021). [Google Scholar]
  • 11.Akcigit, U., Baslandze, S. & Stantcheva, S. Taxation and the international mobility of inventors. Am. Econ. Rev.106(10), 2930–2981 (2016). [Google Scholar]
  • 12.Hunt, G. L. Equilibrium and disequilibrium in migration modelling. Reg. Stud.27(4), 341–3 (1993). [DOI] [PubMed] [Google Scholar]
  • 13.Dehaan, E., Madsen, J. & Piotroski, J. D. Do weather-induced moods affect the processing of earnings news?. J. Account. Res.55(3), 509–550 (2017). [Google Scholar]
  • 14.Sun, W. Z., Zhang, X. N. & Zheng, S. Q. Air pollution and spatial mobility of labor force: Study on the migrants’ job location choice. Econ. Res. J54, 102–117 (2019). [Google Scholar]
  • 15.Liu, Y., Xu, X. & Xiao, Z. The pattern of labor cross-dialects migration. Econ. Res. J.10, 134–162 (2015). [Google Scholar]
  • 16.Zhou, Q. & Qi, Z. Urban economic resilience and human capital: An exploration of heterogeneity and mechanism in the context of spatial population mobility. Sustain. Cities Soc.99, 104983 (2023). [Google Scholar]
  • 17.Banzhaf, H. S. & Walsh, R. P. Do people vote with their feet? An empirical test of Tiebout’s mechanism. Am. Econ. Rev.98(3), 843–863 (2008). [Google Scholar]
  • 18.Chang, T. Y., Graff Zivin, J., Gross, T. & Neidell, M. The effect of pollution on worker productivity: Evidence from call center workers in China. Am. Econ. J. Appl. Econ.11(1), 151–172 (2019). [Google Scholar]
  • 19.Zivin, J. G. & Neidell, M. The impact of pollution on worker productivity. Am. Econ. Rev.102(7), 3652–3673 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Lin, T., Qian, W., Wang, H. & Feng, Y. Air pollution and workplace choice: Evidence from China. Int. J. Environ. Res. Public Health19(14), 8732 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Lan, J., Kakinaka, M. & Huang, X. Foreign direct investment, human capital and environmental pollution in China. Environ. Resour. Econ.51, 255–275 (2012). [Google Scholar]
  • 22.McCollum, D. & Nicholson, H. Internationalisation, sustainability and the contested environmental impacts of international student mobility. Int. J. Sustain. High. Educ.24(7), 1561–1575 (2023). [Google Scholar]
  • 23.Comunian, R. & Jewell, S. ‘Young, Talented and Highly Mobile’: exploring creative human capital and graduates mobility in the UK. In New Frontiers in Interregional Migration Research 205–230 (Springer, 2018). [Google Scholar]
  • 24.Cui, C., Wang, Y. & Wang, Q. The interregional migration of human capital: The case of “First-Class” university graduates in China. Appl. Spat. Anal. Policy15(2), 397–419 (2022). [Google Scholar]
  • 25.Bound, J., Groen, J., Kezdi, G. & Turner, S. Trade in university training: cross-state variation in the production and stock of college-educated labor. J. Econom.121(1–2), 143–173 (2004). [Google Scholar]
  • 26.Krabel, S. & Flöther, C. Here today, gone tomorrow? Regional labour mobility of German university graduates. Reg. Stud.48(10), 1609–1627 (2014). [Google Scholar]
  • 27.Venhorst, V., Van Dijk, J. & Van Wissen, L. An analysis of trends in spatial mobility of Dutch graduates. Spat. Econ. Anal.6(1), 57–82 (2011). [Google Scholar]
  • 28.Liu, Y., Shen, J., Xu, W. & Wang, G. From school to university to work: Migration of highly educated youths in China. Ann. Reg. Sci.59, 651–676 (2017). [Google Scholar]
  • 29.Torche, F. Intergenerational mobility at the top of the educational distribution. Sociol. Educ.91(4), 266–289 (2018). [Google Scholar]
  • 30.Moskal, M. Gendered differences in international graduates’ mobility, identity and career development. Soc. Cult. Geogr.21(3), 421–440 (2020). [Google Scholar]
  • 31.Jacob, M. & Klein, M. Social origin, field of study and graduates’ career progression: Does social inequality vary across fields?. Br. J. Sociol.70(5), 1850–1873 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Zheng, S., Zhang, X., Sun, W. & Lin, C. Air pollution and elite college graduates’ job location choice: Evidence from China. Ann. Reg. Sci.63, 295–316 (2019). [Google Scholar]
  • 33.Lai, W., Song, H., Wang, C. & Wang, H. Air pollution and brain drain: Evidence from college graduates in China. China Econ. Rev.68, 101624 (2021). [Google Scholar]
  • 34.Arceo, E., Hanna, R. & Oliva, P. Does the effect of pollution on infant mortality differ between developing and developed countries? Evidence from Mexico City. Econ. J.126(591), 257–280 (2016). [Google Scholar]
  • 35.Hering, L. & Poncet, S. Environmental policy and exports: Evidence from Chinese cities. J. Environ. Econ. Manag.68(2), 296–318 (2014). [Google Scholar]
  • 36.Broner, F., Bustos, P., & Carvalho, V. M. Sources of comparative advantage in polluting industries (No. w18337). National Bureau of Economic Research. (2012).
  • 37.Wang, H. & Guo, F. City-level socioeconomic divergence, air pollution differentials and internal migration in China: Migrants vs talent migrants. Cities133, 104116 (2023). [Google Scholar]
  • 38.Zhu, M. & Heyes, A. Dreaming of blue skies: Evidence on air pollution and the mobility aspirations of young people in Beijng from online search behavior. Environ. Resour. Econ.87(11), 2889–2933 (2024). [Google Scholar]
  • 39.Liang, Y. & Lu, P. Effect of occupational mobility and health status on life satisfaction of Chinese residents of different occupations: Logistic diagonal mobility models analysis of cross-sectional data on eight Chinese provinces. Int. J. Equity Health13, 1–14 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Chen, J. & Long, X. The impact of air pollution on career changes among Chinese workers. Sci. Rep.15(1), 3782 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Zhang, N. et al. The impact of transient air pollution exposure on worker performance in Chinese soccer players. Sci. Rep.14(1), 31093 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Moretti, E. Local multipliers. Am. Econ. Rev.100(2), 373–377 (2010). [Google Scholar]
  • 43.Poncet, S. Provincial migration dynamics in China: Borders, costs and economic motivations. Reg. Sci. Urban Econ.36(3), 385–398 (2006). [Google Scholar]
  • 44.Fujita, M., Mori, T., Henderson, J. V. & Kanemoto, Y. Spatial distribution of economic activities in Japan and China. In Handbook of regional and urban economics Vol. 4 2911–2977 (Elsevier, 2004). [Google Scholar]
  • 45.Baek, J. & Gweisah, G. Does income inequality harm the environment?: Empirical evidence from the United States. Energy Policy62, 1434–1437 (2013). [Google Scholar]
  • 46.Fu, Y. & Gabriel, S. A. Labor migration, human capital agglomeration and regional development in China. Reg. Sci. Urban Econ.42(3), 473–484 (2012). [Google Scholar]
  • 47.Darchen, S. & Tremblay, D. G. What attracts and retains knowledge workers/students: The quality of place or career opportunities? The cases of Montreal and Ottawa. Cities27(4), 225–233 (2010). [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

All data generated or analyzed during this study are available from the corresponding author upon reasonable request.

Articles from Scientific Reports are provided here courtesy of Nature Publishing Group

RESOURCES