Public Health Service and Migration Destinations among the Labor of Xinjiang Uygur Autonomous Region of China

Abstract

The relationship between public investment and population migration is a classic topic in developing regions. With social and economic development, the role of public health services is paid high attention. However, empirical studies on the relationship between public health services and labor migration are rare, especially for populations from minority areas. This study investigated the correlation between public health services and migration destinations and its heterogeneity among laborers from Xinjiang. Data of the annual Xinjiang Statistical Yearbook and the CMDS are used, and multinomial logit regressions are adopted. The results show that public health services in a county have a significant and negative correlation with the probability of migration with a long range among laborers from Xinjiang. This correlation is inclusive of gender, age, human capital, hukou, marital status, home region, and economic status. It implies that the improvement of public health services in Xinjiang can attract laborers to live and work there, which could contribute to the revitalization of Xinjiang and reduce the development gap between Xinjiang and other provinces.

1. Introduction

Labor migration in China plays an important role in promoting economic growth and reducing poverty. Laborers from rural areas and developing regions flow into cities to supply cheap labor resources and stimulate urbanization [1,2]. Meanwhile, many studies have shown that rural labor migration was historically met with increased rural incomes [3,4,5,6]. Remittances of migrants not only increased the accumulation of assets for farmers and promoted self-employment, but they also increased consumption among those farmers and had an effect on the alleviation of poverty [2,7,8,9].

Several theoretical frameworks of labor mobility have been proposed in the classic literature. Petersen (1958) believed that labor mobility was affected by both “push” and “pull” factors [10]. On this basis, Lee (1966) further classified the factors affecting labor migration into inflow forces, outflow forces, intermediate barriers, and individual factors of labor forces [11]. Tiebout (1956) proposed that residents would choose their living places by “voting with their feet”, determined by local housing prices and public services [12]. This view has been empirically tested [13].

Public services have been proven to be an important factor affecting labor migration. Previous studies have found that the improvement of regional public services would reduce population outflow and increase population inflow [14,15]. An empirical study on Appalachia showed that direct public service investments of government would not only affect household income and regional employment growth rates but also lead to intercounty population mobility [16]. Research on South Africa indicated that government policy intervention in public services would effectively reduce population migration [17]. Based on the panel data of Chinese cities from 2006 to 2014, a study found that the improvement of urban public services’ quality helped population inflow [14].

Public health services, a kind of crucial public service, have been paid high attention from the perspective of utilization and social integration among migrants living in cities. Most studies have found that migrants had low utilization rates of public health services, both in China and other countries [18,19,20,21,22,23,24]. Institutional factors, such as the household registration system (hukou), social integration, and social support, had a close relationship with the utilization of public health services among migrants [21,23,24]. Individual and household characteristics, including social capital, socioeconomic status, and migration destinations, also affected the probability of using public health services among migrants [22,25].

However, studies on the relationship between public health services and migration in China are rare, although there are large regional gaps in the accessibility and quality of public health service in China [26,27,28,29]. To our knowledge, there are four studies that have focused on the impact of public health services on migration. An empirical study using the data of China Migrants Dynamic Survey (CMDS) in 2014 found that access to basic health services in cities and the quality of urban medical services significantly increased the willingness of rural–urban migrants to live there, and there was no heterogeneity in gender, age cohort, educational attainment, or income level [14]. Public health file management played a significant role in improving the willingness of migrant workers to stay in cities [15,25].

Some gaps in this topic need to be narrowed. First, most studies have focused on the correlation between public health services in cities and their attraction of migrants, but the relationship between public health services of sending places and migration destinations is unclear. Under the background of common prosperity, rural revitalization, and border area vitalization in China, improving public health services is considered to be an important measure to achieve these goals by improving human capital and attracting labor inflow border areas. A study on the correlation between public health services and labor migration therefore has significant policy implications. Second, migrants from areas inhabited by ethnic minorities should be paid more attention since they are more vulnerable than Han due to culture and language differences and development stage.

Xinjiang Uygur Autonomous Region (hereinafter Xinjiang) is an ideal region to study the relationship between public service and minority population mobility. Xinjiang is the home of a large number of ethnic minorities, with 47 ethnic minorities living there, accounting for 65% of the population in 2017 [30]. Therefore, this study aims to assess the correlation between public health services of sending places and migration destinations among the laborers of Xinjiang. For the purposes of this goal, we first analyze the relationship of hardware facilities and software resources of public health services at home counties and migration destinations based on the descriptive analysis. Second, we test the inclusion of the role of public health services in migration destinations by conducting heterogeneities from multiple perspectives, including gender, birth cohort, educational attainment, hukou, marital status, geographical location, and economic status of home county.

There are three potential contributions of this study to the literature. First, it investigates the roles of hardware facilities and software resources of public health services in migration destinations from the perspective of sending places, which enriches empirical studies on migration theories. Second, this study focuses on the labor migration of Xinjiang, inhabited by ethnic minorities, which enriches the literature on minorities. Third, the findings have specific and targeted policy implications, which could be helpful to improve governance efficiency of the Xinjiang government.

The remainder of this paper is structured as follows: Section 2 presents the literature review, framework, and hypothesis. Section 3 introduces the materials and methods. Section 4 shows the results of descriptive analysis. Section 5 presents the results and discussion of econometric models. The conclusions and policy implications are included in Section 6.

2. Literature Review, Framework, and Hypothesis

Tiebout (1956) proposed that residents would choose their living places by “voting with their feet”, determined by local public services [12]. Following Maslow’s hierarchy of needs, public health services, one kind of important public service, is paid high attention by governments and scholars when absolute poverty have been addressed and people pursue quality of life in China.

Local public health services affect migration destinations through two channels according to the literature. First, if people are not satisfied with the public health services of their residence county, they can choose to migrate to other places [31,32,33,34], which is “voting with their feet”. Some studies have defined moving out of county as migration [35,36]. There are three destinations for the migrants moving out of county, including across counties within a prefecture-level, across prefecture-level cities within a province, and across provinces. Considering that counties within a prefecture-level city face the similar financial constraints comparing with the countries of other prefecture-level cities, it means that they are more likely to have similar public health services.

There is household registration system in China. The hukou system affects the welfare of migrants, but the internal migration controls due to hukou have been loosened in recent years. Before the 1980s, rural people had little chance to move to urban areas, which caused a great loss of welfare for rural people, such as off-farm employment, public infrastructure, and public service. In the middle of 1980s, the government began to allow rural people to move to urban areas, but they still had no access to urban public service. After 2010, hukou discrimination is almost non-existent in some small and medium-sized cities [37]. The central government is also committed to eliminating hukou discrimination to promote urbanization and decrease the development gap among different areas.

Along with the reform of the hukou system, the reimbursement system of medical expenses has also been reforming since the 2010s. In particular, the reimbursement system in different places was implemented, which means that the people with medical insurance can claim refund of medical expenses when they visit doctors out of home county. At the end of 2013, Xinjiang has implemented the policy of reimbursement system in different places within Xinjiang for medical insurance among urban residents [38], and the rural people with medical insurance were also covered by 2014. It means all Xinjiang people nearly have access to the public service of reimbursement system in different places, because more than 95 percent of Xinjiang people have medical insurance [39].

However, the reimbursement rate of medical expenses varies with the location of medical treatment due to institutional differences, the imbalance between basic medical insurance funds among provinces, as well as complicated referral procedures across provinces. The reimbursement rate of medical expenses for medical treatment in households registered within the county is higher than that outside county but within the province. It was lowest if people seek medical treatment outside province.

Second, there is a high probability of having a good health status for people living in a county with good public health services. Health, as a kind of human capital, determines the radius of labor migration. The healthier the laborer, the longer the migration distance for pursuing a large wage premium [35,36].

Migration destinations are also highly related to individual characteristics, such as gender and educational attainment. Cultural differences, institutional barriers, and migration costs are also the factors affecting migration destinations. Generally, educational attainment and age are used to measure human capital and experience, which are the key elements in acquiring a job and a high wage in a labor market [40]. Cultural differences are reflected by the status of being an ethnic minority, due to different diets, customs, and cultures [41,42]. The household registration system is the leading institutional barrier for rural laborers’ migration in China [43]. Migration costs include opportunity cost and emotional cost [44,45]. Opportunity cost largely depends on the level of economic development and the land resources of a sending place, which are the basic conditions to gain income. Emotional cost is correlated with the structure of a household, such as marital status (Figure 1).

Based on above discussion, we have following hypotheses:

Hypothesis 1.

The labor from a county with poor public health services are more likely to migrate across prefecture-level cities within a province rather than across provinces.

Hypothesis 2.

The role of the public health services of home counties in migration destinations are heterogeneous in gender, human capital, ethnic minority status, marital status, hukou status, GDP, and land endowment.

3. Materials and Methods

3.1. Data

To achieve the goal, we used the data from two datasets, including China Migrants Dynamic Survey (CMDS) and Xinjiang Statistical Yearbooks. The data on migration destinations, individual characteristics, ethnic minorities, marital status, hukou status, and year of migration were from China Migrants Dynamic Survey (CMDS). CMDS is an annual national migrant survey since 2009. The survey was conducted in all 31 provinces, autonomous region, and municipalities of mainland China (except for Hong Kong, Macau, and Taiwan). In each province, almost all the cities, districts, and counties are included. Subsamples were obtained through stratified, PPS, and multistage sampling methods. The respondents were at least 15 years old and had stayed there for at least a month. The sample size of the 2017 CMDS was 169,989 households. The information covered the individual characteristics, family demographic characteristics, migration destinations, and migration reasons of each family member. Information on the employment, social security, income and expenditure, residence willing, utilization of basic public health services, and self-reported health of each migrant in the cities were collected.

However, the data of county of origin for migrants was only collected in the CMDS of 2017. According to the data of CMDS, we know that the years of the migration of laborers were during 2011–2017. We then obtain the information of public health service, economic development, and land resources endowment at the county level during 2011–2017 from annual Xinjiang Statistical Yearbook. Therefore, we can only merge these two datasets from 2011 to 2017.

According to World Health Organization (WHO), the labor is defined to be those aged between 16 and 64 years old, finishing schooling, no retirement, and no disability. We refer to this definition in this study. There are 2093 migrants who are labor, and the year of recent migration was between 2011 and 2017. They were merged with the data of public health services, economic development, and land resources endowment at the county level from the annual Xinjiang Statistical Yearbook during the period of 2011–2017. We then excluded the observations with missing values of variables, and 1967 observations were finally obtained.

3.2. Model Specification

The labor migration destinations were affected by multiple factors. The cross-tabulation only analyzes one of these factors. Therefore, we conduct a multivariate analysis to obtain the net correlation of public health services on labor migration destinations by controlling other factors. This study adopts the following multivariate analysis model:

Y_i = α + β₁Doctor_i + β₂Hospitalbed_i + β₃Hospital_i + β₄X_i + β₅Di + β₆Y_i + ε_i

(1)

where i represents the ith migrant. Y is the migration destinations, including across counties within a prefecture-level city, across prefecture-level cities within a province, and across provinces. Doctor indicates the number of doctors per 10,000 population of migrants’ home county. Hospitalbed and Hospital are the number of hospital beds per 10,000 population and hospitals of migrants’ home county, respectively. In this study, the doctor only includes medical personnel and excludes the management staff. The hospital includes those located in counties and towns. Doctor, Hospitalbed, and Hospital are the key independent variables in this study. We used the log of the number of doctors in the model to meet the requirement of a normal distribution of independent variables.

X is the vector of control variables, including individual characteristics and county characteristics. Individual characteristics contains gender (1 = male, 0 = female), age (years), minority status (1 = Han, 0 = minority), educational attainment (1 = primary, 2 = junior high, 3 = senior high, 4 = associate college, and 5 = college and above), hukou (1 = rural, 0 = otherwise), and marriage (1 = married, 0 = unmarried). County characteristics includes per capita GDP (yuan) and log of land size (square of kilometers) of migrant’s home county, which indicate economic development and land resources.

In order to control the effect of unobserved characteristics on migration destinations and the trends in migration destinations, we add the dummy variables of prefecture-level cities and the year of migration, respectively. The statistical description of independent variables is shown in

Table 1. Description of variables for migrants during 2011 to 2017.

Variable	Definition	Mean	Std. Dev.	Min	Max
Doctor_no	No. of doctors per 10,000 population in county	57.15	15.21	32.07	90.29
Doctor	Log of number of doctors per 10,000 population in county	4.01	0.27	3.47	4.50
Hospitalbed	No. of hospital beds per 10,000 population in county	59.68	8.04	35.80	71.15
Hospital	No. of hospitals in county	234.39	121.59	16.80	665.32
Gender	1 = male, 0 = female	0.46	0.50	0	1
Age	Years	31.4	9.89	16	64
Minority	1 = Han, 0 = minority	0.35	0.48	0	1
Education
Primary	1 = yes, 0 = no	0.19	0.39	0	1
Junior high	1 = yes, 0 = no	0.38	0.49	0	1
Senior high	1 = yes, 0 = no	0.19	0.39	0	1
Associate college	1 = yes, 0 = no	0.14	0.34	0	1
College and above	1 = yes, 0 = no	0.11	0.31	0	1
Hukou	1 = rural, 0 = otherwise	0.68	0.47	0	1
Marriage	1 = married, 0 = unmarried	0.16	0.37	0	1
GDP	Per capita GDP	30,916.07	21,868.21	4741	200,647
Land size	Land size of county	14,029.17	18,477.39	199.38	198,793.5
Land	Log of land size of county	9.02	1.15	5.3	12.2
Prefecture-level cities a
Kashgar	1 = yes, 0 = no	0.17	0.38	0	1
Kizilsu Kirghiz	1 = yes, 0 = no	0.02	0.15	0	1
Hotan	1 = yes, 0 = no	0.05	0.21	0	1
Aksu	1 = yes, 0 = no	0.21	0.41	0	1
Bayingol	1 = yes, 0 = no	0.10	0.3	0	1
IIi	1 = yes, 0 = no	0.18	0.39	0	1
Altay	1 = yes, 0 = no	0.02	0.16	0	1
Tarbagatay	1 = yes, 0 = no	0.07	0.26	0	1
Bortala	1 = yes, 0 = no	0.03	0.17	0	1
Turpan	1 = yes, 0 = no	0.03	0.17	0	1
Hami	1 = yes, 0 = no	0.02	0.15	0	1
Changji	1 = yes, 0 = no	0.09	0.29	0	1

Note: ^a The names of prefecture-level cities are from their official website, and we use the shorthand.

.

The model is estimated by maximum likelihood method (MLE) of multinomial logit regression because migration destinations is a multiple-choice variable.

4. Descriptive Analysis

4.1. Trends in Labor Migration of Xinjiang from 2011 to 2017

There are three choices of migration destinations, including across counties within a prefecture-level city, across prefecture-level cities within a province, and across provinces. Among the 1967 observations, the proportion of these three types of migration from 2011 to 2017 were 31.8 percent, 63.7 percent, and 4.5 percent, respectively. The proportion of migration across provinces was much lower than that in the study using the whole sample of CMDS and that in the Report on the Monitoring and Investigation of Migrant Workers in 2017 [24,46], which contains large proportion of Han. However, the proportions of migration across counties within a prefecture-level city and across prefecture-level cities within a province were much higher than them. These findings indicate that the laborers of Xinjiang mainly migrated within the province during 2011–2017.

Figure 2 shows the trends in migration destinations among the laborers of Xinjiang from 2011 to 2017. The proportion of migration across provinces during 2011–2017 fluctuated between two and four percent. The proportion of migration across prefecture-level cities within a province increased from 60 to 70 percent during this period. The proportion of migration across counties within a prefecture-level city declined by seven percentage point at the same time. These findings show that the migration range of the laborers of Xinjiang enlarged after 2011.

4.2. Trends in Public Health Services in Xinjiang from 2011 to 2017

According to the data from the annual Xinjiang Statistical Yearbook, public health services of Xinjiang was generally improved from 2011 to 2017 (

Table 2. Trends in public health services in Xinjiang from 2011 to 2017.

Year	Total			Southern			Non-Southern
	Doctor	Hospitalbed	Hospital	Doctor	Hospitalbed	Hospital	Doctor	Hospitalbed	Hospital
2011	52.58	51.31	205.54	45.46	51.25	189.06	62.71	51.30	228.99
2012	53.79	53.79	226.64	46.82	52.37	223.32	62.39	53.83	230.72
2013	56.65	55.09	244.49	47.77	53.61	218.89	67.21	56.84	274.91
2014	57.65	55.67	247.80	46.70	53.20	242.57	69.61	58.38	253.52
2015	62.02	59.23	248.99	52.52	57.41	239.56	73.17	61.38	260.06
2016	56.21	66.04	225.08	48.46	66.04	226.64	65.80	66.04	223.14
2017	56.02	69.39	221.38	49.14	69.39	218.29	66.02	69.39	225.86

). First, the number of doctors per 10,000 population increased from 52.58 in 2011 to 62.02 in 2015, with an annual growth rate of 4.49 percentage points. It then decreased to 56.02 in 2017, due to the increase of population and the reform of the healthcare system in Xinjiang since 2015. During this reform, the maternal and child healthcare system and family planning service system were integrated, and many doctors were transferred from medical personnel to other positions. Second, the number of hospital beds per 10,000 population continuously increased from 51.3 to 69.39 during 2011–2017, at a growth rate of 5.9 percentage points in each year. Consistent with the change in doctors, the number of hospitals increased by 21.1 percent from 2011 to 2015 and then decreased by 11.1 percent in next two years. According to the Plan for the Medical and Health Service System of Xinjiang (2016–2020), the number of doctors per 10,000 population and the number of hospital beds per 10,000 population have achieved their goals [47].

In order to objectively evaluate public health services in Xinjiang, we compare it with the national level. The number of doctors per 10,000 population of Xinjiang was more than that of the national level by 2015, but its growth rate was lower. For example, the number of doctors per 10,000 population of national level was 46 in 2011 and 58 in 2015 with annual growth rate of 6.5 percent, which was much larger than that of Xinjiang [48]. Since 2016, the number of doctors per 10,000 population at national level continuously increased to 65, but it decreased to 56.02 in Xinjiang. It indicates that the gap of the number of doctors between Xinjiang and other provinces enlarged. The number of hospital beds per 10,000 population in Xinjiang was much larger than that of the national level, but with lower growth rate from 2011 to 2017. The national level of number of hospital beds per 10,000 population increased from 38.4 in 2011 to 57.22 in 2017, with an annual growth rate of 8.2 percentage points, which was much larger than that of Xinjiang [48]. However, the number of hospitals in Xinjiang was much less than that of the national level. The gap between the number of hospitals in Xinjiang and national level unchanged from 2011 to 2017. These findings indicate that public health services in Xinjiang have developed since 2011, but the gap between Xinjiang and other provinces enlarged, which reflects the regional disparities in public services in China.

Xinjiang, similar to other provinces, has regional differences in the development of its public health services. The southern area of Xinjiang is an area inhabited by ethnic minorities, and is less developed. In this study, we divided Xinjiang into southern and non-southern areas to test the inclusion of the role of public health services in migration destinations. The southern area contains Kashgar, Kizilsu Kirghiz, Hotan, Aksu, and Bayingol. The non-southern area includes the northern and eastern parts: the northern part contains Iii, Altay, Tarbagatay, and Bortala; the eastern part includes Hami, Turpan, and Changji. We excluded Urumchi, the capital of Xinjiang, as it is quite different from other prefecture-level cities. For the number of doctors per 10,000 population, it was much less in southern counties, but it increased faster than that of non-southern area during 2011–2017. The number of hospital beds in southern counties was a little less than that of non-southern counties at the early stage of this period, but it exceeded the latter in the years of 2016 and 2017. With the development of hospitals in southern counties, the difference in the number of hospitals between southern and non-southern counties decreased during 2011–2017. These findings imply that the public health services in the southern counties of Xinjiang were less developed than that of non-southern counties, but the gaps between the southern and non-southern areas were narrowed.

The southern area of Xinjiang was one of the areas in China with severe poverty before 2020. Since the implementation of national targeted poverty reduction in 2015, the southern area of Xinjiang has become a key poverty alleviation region. Poor health is an important cause of poverty [49], and the improvement of public health services is an effective measure of poverty alleviation. The development of public health services in the southern counties of Xinjiang may also benefit from national poverty alleviation efforts.

4.3. Public Health Services and Migration Destinations for Laborers from Xinjiang

In order to find the correlation of public health services and the migration destinations among the laborers of Xinjiang, we first test the relationship between them by conducting a cross-tabulation analysis. Due to the fact that the number of doctors per 10,000 population, the number of hospital beds per 10,000 population, and the number of hospitals is a continuous variable, we classified each of them into three equal groups according to ascending order.

The laborers in the counties with top 1/3 of the number of doctors were more likely to migrate across provinces and across counties within a prefecture-level city. There were 7.45 percent and 39.91 percent of laborers migrating across provinces and across counties within a prefecture-level city during 2011–2017 in the counties with top 1/3 of the number of doctors, respectively. These figures were 1.29 percent and 26.29 percent for laborers in the counties with bottom 1/3 of the number of doctors. With the increase in the number of doctors per 10,000 population, the proportion of the laborers migrating across provinces and across counties within a prefecture-level city rose (Figure 3, Panel A).

However, the relationship between the number of hospital beds and migration destinations was not obvious. It seemed that the laborers from the counties with middle 1/3 of the number of hospital beds were more likely to migrate across provinces (5.13 percent) and across counties within a prefecture-level city (35.44 percent), and less likely to migrate across prefecture-level cities within a province (59.43 percent). The proportions of migration destinations of the laborers from the counties with bottom 1/3 and top 1/3 of the number of hospital beds were similar (Figure 3, Panel B).

The relationship between the number of hospitals and migration destinations varied according to the increase of the number of hospitals. Of the laborers in the counties with bottom 1/3 of the number of hospitals, 3.27 percent migrated across provinces, which was much less than those from the counties with top 1/3. The laborers from the counties with top 1/3 of the number of hospitals were more likely to migrate across prefecture-level cities within a province, whereas those from the counties with bottom 1/3 of the number of hospitals were more likely to migrate across counties within a city (Figure 3, Panel C).

5. Empirical Results and Discussion

5.1. The Correlation of Public Health Services and Migration Destinations

Table 3. The correlation between public health service and migration destinations.

	Across Provinces			Across Prefecture-Level Cities within a Province
Variables	(1)	(3)	(5)	(2)	(4)	(6)
Doctor	−11.135 **	−9.274 *	−9.398 *	−1.887	−1.467	−1.226
	(4.410) a	(4.832)	(4.861)	(1.907)	(1.934)	(1.959)
Hospitalbed	−0.040	−0.064	−0.058	−0.003	−0.006	−0.001
	(0.049)	(0.054)	(0.054)	(0.020)	(0.020)	(0.021)
Hospital	0.010 ***	0.010 ***	0.009 ***	0.007 ***	0.007 ***	0.006 ***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
Gender		0.107	0.144		−0.175	−0.146
		(0.260)	(0.261)		(0.122)	(0.124)
Age		−0.007	−0.007		0.013 *	0.012 *
		(0.016)	(0.016)		(0.007)	(0.007)
Minority		2.194 ***	2.191 ***		0.169	0.176
		(0.375)	(0.377)		(0.169)	(0.174)
Junior high		−1.017 **	−1.028 **		−0.268	−0.263
		(0.451)	(0.452)		(0.168)	(0.170)
Senior high		−0.851 *	−0.869 *		0.082	0.053
		(0.494)	(0.495)		(0.220)	(0.222)
Associate college		−1.058 **	−1.077 **		−0.174	−0.214
		(0.526)	(0.529)		(0.254)	(0.261)
College and above		−0.616	−0.631		−0.471 *	−0.506 *
		(0.530)	(0.533)		(0.276)	(0.282)
Hukou		−0.514 *	−0.418		−0.129	−0.015
		(0.295)	(0.297)		(0.155)	(0.158)
Marriage		0.561	0.584		0.507 ***	0.549 ***
		(0.353)	(0.355)		(0.197)	(0.202)
GDP			0.000 **			0.000 ***
			(0.000)			(0.000)
Land			−0.033			−0.189 **
			(0.183)			(0.081)
City dummies	Yes	Yes	Yes	Yes	Yes	Yes
Migration year dummies	Yes	Yes	Yes	Yes	Yes	Yes
Constant	40.453 **	34.239 *	34.543 *	5.113	3.423	3.525
	(16.645)	(18.267)	(18.390)	(7.193)	(7.303)	(7.450)
LR chi2	784.64	869.35	910.57	784.64	869.35	910.57
Prob > chi2	0.000	0.000	0.000	0.000	0.000	0.000
Pseudo R2	0.252	0.280	0.293	0.252	0.280	0.293
Mean VIF	1.28	1.35	1.50	1.28	1.35	1.50
Observations	1967	1967	1967	1967	1967	1967

Note: ^a Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, and * p < 0.1.

shows the results of multinomial logit regression.

Table 4. The correlation between public health service and migration destinations (Marginal effects).

	Across Provinces			Across Prefecture−Level Cities within a Province
Variables	(1)	(2)	(3)	(4)	(5)	(6)
Doctor	−0.423 ***	−0.381 **	−0.389 **	0.101	0.114	0.146
	(0.162) a	(0.163)	(0.163)	(0.288)	(0.289)	(0.286)
Hospitalbed	−0.003 **	−0.003 **	−0.003 **	−0.001	−0.001	0.001
	(0.001)	(0.001)	(0.001)	(0.003)	(0.003)	(0.003)
Hospital	0.000 ***	0.000 ***	0.000 ***	0.001 ***	0.001 ***	0.001 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Gender		0.009	0.009		−0.032 *	−0.027
		(0.009)	(0.009)		(0.019)	(0.019)
Age		−0.001	−0.001		0.002 **	0.002 *
		(0.001)	(0.001)		(0.001)	(0.001)
Minority		0.075 ***	0.075 ***		−0.032	−0.031
		(0.014)	(0.014)		(0.026)	(0.026)
Junior high		−0.035 *	−0.035 *		−0.012	−0.010
		(0.021)	(0.021)		(0.028)	(0.028)
Senior high		−0.036	−0.036		0.039	0.035
		(0.022)	(0.022)		(0.034)	(0.034)
Associate college		−0.037	−0.037		0.003	−0.002
		(0.023)	(0.023)		(0.040)	(0.040)
College and above		−0.011	−0.010		−0.059	−0.064
		(0.026)	(0.026)		(0.044)	(0.044)
Hukou		−0.014	−0.014		−0.007	0.009
		(0.010)	(0.010)		(0.023)	(0.023)
Marriage		0.009	0.009		0.066 **	0.070 **
		(0.012)	(0.012)		(0.029)	(0.029)
GDP			0.000			0.000 ***
			(0.000)			(0.000)
Land			0.004			−0.029 **
			(0.006)			(0.012)
City dummies	Yes	Yes	Yes	Yes	Yes	Yes
Migration year dummies	Yes	Yes	Yes	Yes	Yes	Yes
LR chi2	784.64	869.35	910.57	784.64	869.35	910.57
Prob > chi2	0.000	0.000	0.000	0.000	0.000	0.000
Pseudo R2	0.252	0.280	0.293	0.252	0.280	0.293
Mean VIF	1.28	1.35	1.50	1.28	1.35	1.50
Observations	1967	1967	1967	1967	1967	1967

Note: ^a Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, and * p < 0.1.

shows its marginal results. Individual characteristics and county characteristics are gradually added as control variables in columns two, three, five, and six to conduct a robustness check. For brevity, we explain the magnitude of the coefficients in columns three and six of Table 4 since they are the most robust. The reference group is migration across counties within a prefecture-level city. The values of Prob > chi2 were 0.000, which indicates that the models were successful when conducting multiple hypothesis testing. The values of VIF were less than 2, which means that there were no multicollinearities in the models.

The results show that the number of doctors per 10,000 population was significant and negative correlation with the probability of migrating across provinces comparing with that of migrating across counties within a prefecture-level city. The results were robust when we gradually added individual and county characteristics in the model. The magnitude of the coefficient of doctors was 0.389, which means that the laborers were 38.9 percentage points less likely to migrate across provinces than across counties within a prefecture-level city with a one percent-increase in the number of doctors in the county (about 5.5 doctors). The number of doctors per 10,000 population had no correlation with the probability of migrating across prefecture-level cities within a province comparing with that of migrating across counties within a prefecture-level city. These findings are consistent with previous studies [14]; that is, a large number of doctors will reduce the outflow of the population.

The number of doctors per 10,000 population is an important indicator with which to measure the quality of public health services in a county, and it is directly related to the availability of healthcare resources as well as to the efficiency of medical treatment [50]. An area with many doctors can provide timely and sufficient health care services, thus it has a strong attraction to migrants. Increasing the number of doctors will significantly weaken the local thrust on migration, making people more inclined to stay in their home county.

The number of hospital beds had a significant and negative correlation with the probability of migrating across provinces comparing with that of migrating across counties within a prefecture-level city. However, it had no correlation with the probability of migrating across cities in a province comparing with that of migrating across counties within a prefecture-level city. With a one-percent increase in the number of hospital beds, the probability of migrating across provinces decreased by 0.3 percent comparing with that of migrating across counties within a prefecture-level city. This could be yielded by the medical insurance policy in Xinjiang. Similar to other provinces of China, migrants face many institutional obstacles to access public health services across provinces, such as the reimbursement rate of medical expenses due to institutional differences, the imbalance between basic medical insurance funds among provinces, as well as complicated referral procedures across provinces. These obstacles make people more inclined to choose local hospitals, which have a high reimbursement rate and simple procedures, especially for hospitalized patients. Therefore, an increase in the number of local hospital beds will significantly reduce the probability of migration across provinces. For cities within a province, they are less affected by reimbursement rates, thus the number of hospital beds has no correlation with the probability of migration across cities within a province.

The number of hospitals is positively correlated with the probabilities of migration destinations. The probabilities of migrating across provinces and across prefecture-level cities increased by 0 and 0.1 percent comparing with that of migrating across counties within a prefecture-level city with a one-percent increase in the number of hospitals, respectively. This indicates that the magnitudes of the coefficients were much smaller, which means that the correlations were negligible. This could have resulted due to most of the role of hospitals being decomposed by the number of doctors and hospital beds.

Xinjiang is one of the very important frontier autonomous regions. Encouraging local laborers or attracting migrants to build Xinjiang is crucial to narrow the development gap between Xinjiang and other provinces, as well as to maintain social stability. Improvement of the quality of public health services in counties is expected to be one way to reduce the outflow of people according to these findings. Additionally, the ongoing reform of Xinjiang’s medical and health service system will make precise efforts to optimize medical resource allocation by promoting comprehensive reform at the county level as the key, which will make efforts to bring high-quality medical resources to the counties. There are software and hardware in the high-quality medical resources. These findings imply that the correlation of software medical resources (doctors) is much more vital than that of hardware medical resources when promoting local prosperity by attracting people to stay.

The coefficients of control variables are expected. Gender had no correlation with the migration destinations, which is not consistent with a previous study [51]. One possible explanation for this is the difference in migration modes between Han and minorities among Xinjiang laborers. Minorities are more likely to migrate with their partners. In the sample of this study, 78.35 percent of minorities migrated with their partners, but this was only the case for 67.38 percent of Han migrants. With a one-year increase of age, the labor was 0.2 more likely to migrate across prefecture-level cities than across counties within a prefecture-level city. Han people had a higher probability of migrating across provinces than ethnic minorities did. It is difficult to integrate into the cities outside the province where there are mainly Han, because there are huge gaps in customs, living habits, and languages between Han and ethnic minorities. However, there was no significant difference in migration across prefecture-level cities in a province between Han and minorities, which is consistent with the results based on the data on migrants in Urumqi [52]. There are 24.45 million population in Xinjiang in 2017, and 65 percent of them are ethnic minorities [30]. There may be two reasons to yield this result. On the one hand, it is easy for ethnic minorities to migrate to cities that are suitable for them in terms of customs, living habits, languages, and so on. On the other hand, ethnic minorities and Han ethnic groups have coexisted for a long time within Xinjiang, so they well know how to respect each other in culture and customs. All of these environmental conditions benefits both Han and ethnic minorities to migrate within Xinjiang.

The laborers with junior and senior high school education were less likely to migrate across provinces than those with primary and below education. Previous studies found that migration was the second-best choice for laborers [53]. It is difficult for laborers with primary and below education to find off-farm jobs in their locales. They have to migrate to acquire off-farm earnings. Those with tertiary education, including associate college as well as college and above, have the ability to find stable and high-wage off-farm jobs at their locale or migrate long distances to acquire a wage premium [54]. Hukou had no correlation with the probability of migration destination, which is related to the efforts to eliminate hukou discrimination by the Chinese government. There is a household registration system in China. Under this system, there are rural people and urban people. Before the 1980s, rural people had little chance to move to the urban areas, which caused a great loss of welfare for rural people, such as off-farm employment, public infrastructure, and public service. In the middle of 1980s, the government began to allow rural people to move to the urban areas, but they still had no access to urban’s public service. After 2010, hukou discrimination became almost non-existent in some small and medium-sized cities. The central government is also committed to eliminating hukou discrimination to promote urbanization and decrease urban-rural development gap. With this background, people with rural or urban hukou can move freely. Married people are more inclined to migrate across prefecture-level cities in a province, which benefits them in terms of earning money and taking care of their family.

With the growth in the GDP of a county, the probability of migrating across prefecture-level cities within a province increased a little faster than that across counties within a prefecture-level city. However, the laborers in a county with a large land size were less likely to migrate across prefecture-level cities within a province than across counties within a prefecture-level city.

5.2. Heterogeneity of the Correlation of Public Health Service and Migration Destinations

In order to test the inclusion role of public health services in migration destinations, we conducted heterogeneous analyses from the perspective of gender, birth cohorts, ethnic minority status, educational attainment, hukou, marital status, region, and GDP by adding cross-items between these variables and public health services in multinomial logit regressions. According to the results in Table 4, we found that the number of doctors and the number of hospital beds had correlations with the probability of migrating across provinces comparing with that of migrating across counties within a prefecture-level city, and that the number of hospitals had correlations with migration both across provinces and across prefecture-level cities within a province comparing with that of migrating across counties within a prefecture-level city. We only conducted heterogeneous analyses for those statuses with significant correlations in Table 4. For brevity, we summarize the coefficients of the cross-items in

Table 5. Heterogeneities of the correlation between the number of doctors and the probability of migration across provinces.

Variables a	Coefficients b	Standard Errors c
Doctor * gender	0.009	0.035
Doctor * 1970s	0.134 *	0.075
Doctor * 1980s	0.059	0.063
Doctor * 1990s	0.033	0.064
Doctor * minority	0.002	0.048
Doctor * junior high	0.014	0.058
Doctor * senior high	−0.065	0.059
Doctor * associate college	−0.082	0.063
Doctor * college and above	0.033	0.063
Doctor * hukou	−0.033	0.039
Doctor * marriage	−0.048	0.042
Doctor * southern	0.062	0.060
Doctor * bottom 1/3 of GDP	0.032	0.082
Observations	1967

Note: ^a * indicated the cross term of the variable of doctor and individual characteristics. ^b Marginal effects are reported, and * p < 0.1. ^c Robust standard errors are reported.

,

Table 6. Heterogeneities of the correlation between the number of hospital beds and the probability of migration across provinces.

Variables a	Coefficients b	Standard Errors c
Hospitalbed * gender	−0.000	(0.001)
Hospitalbed * 1970s	0.003	(0.002)
Hospitalbed * 1980s	0.003	(0.002)
Hospitalbed * 1990s	0.002	(0.002)
Hospitalbed * minority	−0.001	(0.001)
Hospitalbed * junior high	−0.001	(0.002)
Hospitalbed * senior high	−0.002	(0.002)
Hospitalbed * associate college	−0.001	(0.002)
Hospitalbed * college and above	0.001	(0.002)
Hospitalbed * hukou	0.000	(0.001)
Hospitalbed * marriage	−0.001	(0.001)
Hospitalbed * southern	0.003 *	(0.001)
Hospitalbed * bottom 1/3 of GDP	−0.003	(0.003)
Observations	1967

Note: ^a * indicated the cross term of the variable of hospitalbed and individual characteristics. ^b Marginal effects are reported, and * p < 0.1. ^c Robust standard errors are reported.

and

Table 7. Heterogeneities of the correlation between the number of hospitals and migration destinations.

Variables a	Across Provinces	Among Prefecture−Level Cities within a Province
	(1)	(2)
Hospital * gender	0.000	0.000
	(0.000) b	(0.000)
Hospital * 1970s	0.000 **	−0.000
	(0.000)	(0.000)
Hospital * 1980s	0.000	−0.000
	(0.000)	(0.000)
Hospital * 1990s	0.000	−0.000
	(0.000)	(0.000)
Hospital * minority	−0.000	0.000
	(0.000)	(0.000)
Hospital * junior high	−0.000	0.001 ***
	(0.000)	(0.000)
Hospital * senior high	−0.000	0.000
	(0.000)	(0.000)
Hospital *associate college	−0.000	−0.000
	(0.000)	(0.000)
Hospital * college and above	−0.000	0.000
	(0.000)	(0.000)
Hospital * hukou	−0.000	0.000 *
	(0.000)	(0.000)
Hospital * marriage	0.000	0.000
	(0.000)	(0.000)
Hospital * southern	−0.000	0.001 ***
	(0.000)	(0.000)
Hospital * bottom 1/3 of GDP	−0.000	−0.000
	(0.000)	(0.000)
Observations	1967	1967

Note: ^a * indicated the cross term of the variable of hospital and individual characteristics. ^b Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, and * p < 0.1.

.

The results in Table 5 show that there were not heterogeneities in the correlation between the number of doctors and the probability of migrating across provinces in terms of gender, ethnic minority status, educational attainment, hukou, marital status, and economic status. In other words, there were no significant differences in the correlation between the number of doctors and the migration destinations among laborers with different genders, ethnic minorities, human capital, hukou, marital statuses, and from different regions. It indicated that there were inclusions of the relationship between the number of doctors and the migration destinations at these perspectives.

The coefficient of the cross-item between doctors and the birth cohort of the 1970s was positive and significant at a 10-percent level. It indicates that the number of doctors had a relatively larger and positive correlation with the probability of migrating across province among the labor birthed during 1970s than those birthed during 1960s (reference group) comparing with that of migrating across counties within a prefecture-level city. The coefficient of southern counties was positive and significant at a 10-percent level in promoting migration across provinces comparing with that of migrating across counties within a prefecture-level city. This indicates that the number of doctors had a relatively larger and positive correlation with the probability of migrating across provinces among the laborers from South Xinjiang comparing with that of migrating across counties within a prefecture-level city.

The results in Table 6 show that there were not heterogeneities in the correlation between the number of hospital beds and the probability of migrating across provinces in terms of gender, birth cohort, ethnic minority status, educational attainment, hukou, marital status, and economic status. The coefficient of the cross-item between hospital beds and North Xinjiang was positive and significant at one-percent level. This indicates that the number of hospital beds had a relatively larger and positive correlation with the probability of migrating across provinces among the laborers from North Xinjiang than those from East Xinjiang (reference group) comparing with that of migrating across counties within a prefecture-level city.

The results in Table 7 show that there were not heterogeneities in the correlation between the number of hospitals and the probabilities of migrating across provinces and across prefecture-level cities within a province in terms of gender, ethnic minority status, educational attainment, hukou, marital status, and economic status. However, the correlation between the number of hospitals and the probability of migrating across provinces for the 1970s cohort was much larger than that for the 1960s cohort, but it was much smaller for laborers from South Xinjiang than for those from East Xinjiang. The correlation of the number of hospitals and the probability of migrating across prefecture-level cities were larger for those with an educational attainment of junior high as well as for those from East Xinjiang.

In sum, the heterogeneous analysis find that the laborers birthed during 1970s were more likely to migrate across provinces than those birthed after 1990s with a one percent increase of the number of doctors per 10,000 population or one increase of hospital. The laborers from southern counties of Xinjiang were more likely to migrate across provinces with one increase of hospital beds and migrate across prefecture-level cities within Xinjiang with one increase of hospital than those from non-southern counties. Although public health service at county level generally attracted laborers to stay within province, its roles were different among the vulnerable groups, such as older and minority laborers. For these laborers, it is easy to gain higher wage by migrating into other provinces than working within Xinjiang, where there is limited off-farm employment opportunities.

Due to data limitation, we have to use the data collected in 2017 to conduct this study. The trends of the development of public health service in these counties of Xinjiang seems to be consistent with some of our findings after 2017, which partially proves the practical significance of this study. According to the data of 2021 Xinjiang Statistic Yearbook, the average number of doctors per 10,000 population in these counties increased to 61.47 by 2020. The number of hospital beds per 10,000 population in these counties slightly declined to 60.24, and the average number of hospitals in these counties decreased to 154.73 by 2020.

6. Conclusions and Implications

This study investigates the correlation between public health services and migration destinations, as well as its heterogeneity, among laborers from Xinjiang using the data of annual Xinjiang Statistical Yearbook and CMDS. The results show that public health services in Xinjiang have developed since 2011, but the gap between Xinjiang and other provinces enlarged, which reflects the regional disparities of public services in China. It is gratifying that public health services in South Xinjiang are improving and that the gap within Xinjiang was narrowed. The results from multivariate regressions show that the number of doctors and hospital beds in a county have a significant and negative correlation with the probability of migration out of Xinjiang comparing with that of migrating across counties within a prefecture-level city, especially the number of doctors. The number of hospitals in counties promoted labor outflow, but the magnitudes of coefficients were negligible. These correlations were inclusive of gender, ethnic minority status, hukou, marital status, and economic status, but they varied among laborers with different birth cohorts and home regions.

These findings have important implications to promote vitalization and common propensity of border areas inhabited by ethnic minorities. First, there have been obvious differences in regional development of public health service in China. The areas inhabited by ethnic minorities are the focal points of future development. Second, the improvement of public health services can contribute to attract people to work and live in border areas, which is helpful to provide human capital to promote vitalization of border areas with social inclusion. Third, the government should pay more attention to the software of public health service when achieving the target of next five-year plan, which plays a much more important role in abstracting people to live and work in Xinjiang than that of hardware of public health service.

This study focused on the correlation between public health services and migration destinations and its heterogeneities, which contributes to the research on public investment and labor migration. However, we acknowledge the shortcomings of our study due to data limitations. First, the causal relationship between public health services and migration destinations should be further investigated by adopting advanced methods with data including the individuals without migration. Second, information on the accessibility of public health services at the household or individual level should be used to yield the average effect of the treatment on the treated. Third, the mechanisms under these impacts need to be explored to support detailed interventions to improve local public health services.