How Does Urban Rail Transit Density Affect Jobs–Housing Balance? A Case Study of Beijing

Abstract

Jobs–housing balance is a critical concern in urban planning and sustainable economic development. Urban rail transit, as a key determinant of employment and residential location decisions, plays a pivotal role in shaping jobs–housing dynamics. Beijing, the first Chinese city to develop a subway system, offers a comprehensive rail network, making it an ideal case for exploring the effects of transit density on jobs–housing balance. This study utilizes medium-scale panel data from Beijing (2009–2022) and employs a fixed-effects model to systematically examine the impact of rail transit station density on jobs–housing balance and its underlying mechanisms. The results indicate that increasing transit station density tends to aggravate jobs–housing separation overall, with pronounced effects in central and outer suburban areas but negligible effects in near suburban areas. Mechanism analysis reveals two primary pathways: (1) improved accessibility draws employment toward transit-rich areas, reinforcing the attractiveness of central districts; (2) rising housing prices elevate residential thresholds, pushing lower-income populations toward outer suburbs. While enhanced transit density improves commuting convenience, it does not effectively reduce jobs–housing separation. These findings offer important policy implications for optimizing transit planning, improving jobs–housing alignment, and promoting sustainable urban development.

1. Introduction

Traffic congestion and excessive commuting are major challenges faced by Beijing in its urban development, with jobs–housing separation being an important contributing factor [1,2]. According to the 2023 Annual Commuting Monitoring Report of Major Chinese Cities, in 2022, the proportions of commutes within 45 min of Beijing, Shanghai, Guangzhou, and Shenzhen were 56%, 70%, 71%, and 77%, respectively, while the proportions of commutes exceeding 60 min were 28%, 18%, 15%, and 12%, respectively. Among them, Beijing had the lowest share of commutes within 45 min and the highest share of commutes exceeding 60 min, indicating a particularly severe problem of excessive commuting, low commuting efficiency, and prominent jobs–housing separation.

This situation is closely related to the classic Spatial Mismatch Hypothesis [3], which posits that the spatial separation between employment centers and residential areas increases commuting costs and limits employment opportunities for low-income populations. Similarly, the principles of Transit-Oriented Development [4] emphasize how well-integrated rail systems can promote a more balanced urban form and alleviate jobs–housing separation. Integrating these theoretical perspectives provides a stronger conceptual basis for examining how transit systems influence jobs–housing dynamics in Beijing.

Beijing’s urban planning has actively promoted polycentric development in recent years. Strategic decentralization policies have aimed to shift employment opportunities and public resources to sub-centers such as Tongzhou District and Yizhuang Economic Development Zone, in an effort to alleviate pressures on the urban core and improve the jobs–housing balance [5]. Similar polycentric and transit-integrated strategies have been observed in international cities. For example, Tokyo’s rail-oriented urban form and Paris’s efforts to coordinate transit development with urban regeneration provide useful references for understanding how large cities leverage transportation infrastructure to shape spatial structure and mitigate jobs–housing separation [4,6]. These comparative perspectives help situate Beijing’s experience within a broader global context while highlighting its unique local conditions.

As the first city in China to open a subway line, Beijing has continuously expanded its urban rail transit network since the launch of Line 1 in 1971. As of 2023, Beijing had 27 urban rail transit lines with a total length of 907 km and 404 stations, including 83 transfer stations (Source: China Association of Metros), as shown in Figure 1. Urban rail transit accounts for 48.99% of total public transport passenger volume in Beijing, highlighting its critical role in public transportation (Source: Ministry of Transport). Although the development of rail transit has shortened commuting times and expanded the travel radius, it has also led more workers to reside in suburban areas with lower housing prices, thereby exacerbating the jobs–housing separation. The density of the rail transit network directly affects commuting convenience and the extent of jobs–housing separation. As both a pioneer and a representative example of rapid and large-scale urban rail transit development in China, Beijing is chosen in this study to examine the impact of rail transit density on the jobs–housing relationship, focusing on regional heterogeneity and the underlying mechanisms.

This study selects urban rail transit density as the core explanatory variable, measured by the ratio of the number of urban rail transit stations to the administrative area size at the district level. In this study, the term urban rail transit refers to the entire urban rail system of Beijing, including both subway lines and suburban railways, as defined by the official statistics of the Beijing Municipal Commission of Transport. Urban rail transit density reflects the coverage and development level of a city’s transit system. By enhancing population mobility, rail transit enables residents to live farther from their workplaces without being overly concerned about commuting, which, to some extent, may also exacerbate jobs–housing separation.

The jobs–housing relationship is influenced by a variety of factors. Existing literature primarily explores this relationship from the perspectives of spatial structure, housing systems, income levels, and other socioeconomic dimensions [7,8,9], while relatively few studies have focused on the role of rail transit density. This paper innovatively utilizes meso-level panel data from Beijing and constructs a dynamic model to address the limitations of previous studies that relied on cross-sectional data, which are unable to capture dynamic effects. Furthermore, by analyzing high-quality data, this study investigates the mediating effects of employment and housing prices, thereby enriching the research framework on the jobs–housing relationship and providing practical insights for urban rail transit planning and improving jobs–housing balance.

2. Literature Review and Research Hypotheses

Jobs–housing separation can trigger a series of urban problems in large cities, such as environmental pollution and traffic congestion [10]. At the same time, excessive commuting increases the physical and mental burden on workers, negatively affecting their productivity and overall well-being [11]. Meanwhile, in pursuit of a better quality of life, people may choose to live farther from their workplaces despite the commuting burden. The reasons are as follows. With rising living standards, people are placing higher demands on their quality of life and living environment. This may lead them to prefer living in areas farther from the city center that offer better environmental conditions, thereby contributing to the separation of work and residence [12]. Urban rail transit systems, capable of traversing cities quickly, offer residents a faster and more efficient commuting option, allowing them to live farther from their workplaces while still maintaining relatively short commuting times [13]. Moreover, the opening of subway lines can lead to a reconfiguration of residential and commercial areas. Increased commercial development around subway stations may reduce the availability of residential land, forcing original residents to relocate to more distant areas, while job centers remain concentrated in the urban core, thereby exacerbating jobs–housing separation [14]. Thus, as living standards improve, residents are more inclined to choose housing farther from city centers, further intensifying the separation between work and living locations. Although urban rail transit improves commuting efficiency and may partially ease this problem, it may also prompt residential and commercial reallocation, pushing some residents even farther out and deepening jobs–housing separation. Based on this, Hypothesis 1 is proposed in this paper.

Hypothesis 1. 

Increasing urban rail transit density exacerbates jobs–housing separation.

The characteristics of urban residential space reflect the overall layout and population distribution density of a city, shaped by the interaction of natural, social, economic, cultural, and historical factors. The degree to which different functional zones are influenced by the rail transit system varies depending on the urban spatial structure. Huang et al. [15], using smart card data from Beijing and Shanghai, conducted a comparative study on the spatial characteristics of jobs–housing patterns in the two cities. They found that Beijing exhibits a concentric spatial structure, while Shanghai displays a fan-shaped layout. Compared with Beijing, Shanghai has a higher degree of jobs–housing mix, and the two cities are affected differently by the rail transit system. Blumenberg and Siddiq [16] found an apparent jobs–housing imbalance through the study of Los Angeles communities in the United States, which is more prominent in the coastal areas and high-income areas of Orange. Zhou et al. [17] found that when large spatial units are used, the uniformity of spatial analysis units affects the impact of jobs–housing balance on commuting distance. Due to different sociodemographic characteristics across regions, the relationship between jobs–housing balance and commuting distance also varies. In Beijing, there is a significant gap in economic development between central urban areas and outer suburbs. The central areas are commercially developed and concentrated with employment opportunities, while lower housing prices in the outer suburbs attract many workers to live there. In contrast, the jobs–housing relationship in nearby suburban areas tends to be more balanced. These prior studies and observations highlight the existence of regional heterogeneity in jobs–housing patterns and how urban rail transit may have different effects depending on spatial structure and regional characteristics. This provides the theoretical and empirical basis for proposing the following hypothesis. Therefore, this paper proposes Hypothesis 2.

Hypothesis 2. 

Urban rail transit density has a significant impact on the jobs–housing relationship in central urban areas and outer suburbs, but shows no obvious effect in nearby suburban areas.

Urban rail transit enhances travel accessibility and influences people’s residential and employment choices [18,19]. The improvement of urban rail transit significantly enhances accessibility, which has a substantial impact on employment decisions. Increasing transit density can attract more working populations to transit-convenient areas by improving regional accessibility, reducing commuting costs, fostering industrial agglomeration, promoting the development of service industries, improving the urban environment, and lowering recruitment difficulties for enterprises, thereby raising overall employment levels [20,21]. The effect is particularly pronounced in areas with a dense rail transit network. Bautista-Hernandez [22] studied the relationship between jobs–housing balance and commuting in Mexico City and found that the jobs–housing ratio positively correlates with the number of locally employed residents. Kwon and Gil [23] found a U-shaped relationship between commuting time and jobs–housing balance in Seoul, with the shortest commutes in balanced areas and longer times as employment or residential agglomeration increases. Zheng et al. [24] found that POI (point of interest, POI) mixing has divergent effects on different types of jobs–housing balance, and a moderate job-to-housing ratio does not guarantee achieving balance. Ge and Wang [25], using Guangzhou as a case study, established a grey model to predict the city’s investment attraction and industrial recruitment trends over the next five years and found that urban rail transit construction plays an important role in attracting employment. This concentration of employment reinforces the attractiveness of employment centers but may also further increase the spatial separation between residential and work locations, intensifying jobs–housing separation. Therefore, this paper proposes Hypothesis 3.

Hypothesis 3. 

Increasing urban rail transit density induces employment concentration, thereby exacerbating jobs–housing separation.

Urban rail transit construction also significantly affects residential choices. The convenience of transit not only improves the accessibility of nearby properties and facilitates commuting but also stimulates economic prosperity and development by increasing land development intensity, changing land use patterns, adjusting industrial layouts, and accelerating urbanization. However, this process is often accompanied by a sharp rise in housing prices around transit stations, raising the residential threshold [26]. Zhou et al. [27] found that in Suzhou Industrial Park, high housing prices contribute to a qualitative jobs–housing imbalance, forcing workers to live far from their workplaces despite planned spatial balance. Zhou et al. [28] found that while residential/employment mixed use can improve jobs–housing balance in suburban industrial areas, high housing prices and other factors limit its effectiveness in achieving employment self-containment in central commercial areas. As housing prices near stations rise, low- and middle-income households may find it unaffordable to live in these areas. Consequently, they may opt to reside farther from transit stations, thereby intensifying jobs–housing separation [10]. Although many people desire to live near subway stations, high housing prices often make such locations unaffordable, resulting in a mismatch between desired and actual residence locations. This further contributes to the separation between workplace and residence [29]. Based on this, the paper proposes Hypothesis 4.

Hypothesis 4. 

Increasing urban rail transit density drives up housing prices and raises the residential threshold, thereby exacerbating jobs–housing separation.

3. Research Design

3.1. The Measurement Model

Figure 2 shows the jobs–housing ratio of various towns and streets in Beijing for the year 2020. Figure 3 displays the changes in the jobs–housing ratio for various towns and streets in Beijing from 2010 to 2020. Negative values in Figure 3 indicate areas where the jobs–housing ratio decreased compared to 2010, while positive values indicate areas where the ratio increased. It can be observed that the jobs–housing separation in Beijing has generally intensified. For each district, the opening of the metro line has a certain exogeneity. To further explore the underlying causes of jobs–housing separation, this study conducted model selection tests before constructing the two-way fixed-effects model. Specifically, F-tests and Hausman tests were employed to ensure the scientific validity of the model choice. The results are shown in

Table 1. Statistical test results for the selection of two-way fixed-effects model.

Test Type	Statistic Value	p-Value
F-test	15.34	0.001
Hausman test	22.17	0.001
Time fixed-effects test	8.45	0.003

. As indicated in Table 1, the F-test comparing the pooled regression model with the fixed-effects model yielded an F-statistic of 15.34 (p < 0.05), which significantly rejects the null hypothesis of the pooled model, indicating the presence of individual fixed effects. Furthermore, the Hausman test comparing the random effects model with the fixed-effects model produced a Hausman statistic of 22.17 (p < 0.05), significantly rejecting the null hypothesis of the random effects model, thereby supporting the use of a fixed-effects model. In addition, to verify whether time effects are significant, a test for time fixed effects was conducted, yielding a statistic of 8.45 (p < 0.05), indicating that time fixed effects are also significant. Based on the above results, this study ultimately adopts a two-way fixed-effects model for the analysis.

The specifics are presented as Formula (1).

$${RJHB}_{it}={\alpha }_0+{\beta }_1{Rail}_{it}+{\gamma }_j\sum Z_{jit}+{\eta }_t+{\mu }_i+{\epsilon }_{it}$$

(1)

In Formula (1), $i$ represents the number of individuals in districts and counties and $t$ represents the number of periods. The explained variable ${RJHB}_{it}$ represents the revised residence/employment deviation index in district i in year t. ${Rail}_{it}$ represents the dummy variable of the year when the metro was opened, and its value is between 0 and 1. If the metro is opened in district i in year t, then ${Rail}_{it}$ takes 1, and if the metro is not opened in district i in year t, then ${Rail}_{it}$ takes 0. ${\alpha }_0$ is the intercept item; ${\beta }_1$ is an estimated coefficient, indicating the impact of the metro’s opening on the jobs–housing balance. $Z$ represents the $j$ control variables selected in this paper, and $\gamma$ is the corresponding estimated coefficient. ${\eta }_t$ is the time-fixed effect, ${\mu }_i$ is the individual fixed effect, and ${\epsilon }_{it}$ is the random error item.

3.2. Indicator Explanation

3.2.1. Dependent Variable

There are many ways to measure the jobs–housing balance level in a particular area, including the entropy value description method and the Gini coefficient description method based on the Lorenz curve. In measuring the employment/residential balance, the employment/residential deviation index is often used to deform the jobs–housing ratio. The specific calculation formula of JHB (jobs–housing balance, JHB) is:

$$JH{B}_{it}=\frac{E_{it}/E_t}{P_{it}/Pt}$$

(2)

Among them, ${JHB}_{it}$ is the JHB index of district $i$ in year $t$. $E_{it}$ is the number of the employed population in district $i$ in year $t$, and $E_t$ is the total number of the employed population in Beijing in year $t$. $P_{it}$ is the number of the resident population in district $i$ in year $t$. The resident population aged 15–64 in year t is expressed. $P_t$ is the total resident population of Beijing in year $t$, represented by the resident population aged 15–64 in Beijing in year $t$. According to the definition and nature of the jobs–housing employment deviation index, when $JHB$ is less than 1, it means that the resident population in administrative region $i$ is relatively large, while the employed population is relatively small; when $JHB$ is greater than 1, it means that the resident population in administrative region i is relatively tiny and the employed population more; when $JHB$ is equal to 1, it means that the resident population and the employed population of administrative region $i$ have reached an equilibrium state. It can be seen from the formula that the value range of the residence/employment deviation index is (0, +∞), with one as the center, and the values in the intervals of (0, 1) and (1, 2) cannot be accurately sorted. In order to better measure the impact of the metro opening on the jobs–housing balance, this paper uses the revised jobs–housing deviation index as the explained variable. $RJHB$ represents the revised jobs–housing deviation index. The smaller the $RJHB$, the better the match between living and employment spaces. The more significant the $RJHB$, the worse the match between living space and employment.

$${RJHB}_{it}=|{JHB}_{it}-1|$$

(3)

3.2.2. Control Variable

This study draws on previous research [30] and selects the following control variables: (1) Level of economic development (lnPgdp): Regions with different levels of economic development exhibit significant differences in housing supply and demand, industrial layout, and transportation networks. This variable is measured by the logarithm of real per capita GDP, adjusted to eliminate price factors using 2009 as the base year. (2) Per capita disposable income (lnPcdi): Income level influences residents’ housing choices and commuting behavior. High-income groups are more likely to live in central or nearby suburban areas, while low-income groups may be constrained by housing prices and opt for outer suburban residences. This variable is measured by the logarithm of per capita disposable income. (3) Industrial agglomeration level (Ind): Areas with high industrial concentration tend to attract large numbers of workers, which may lead to jobs–housing imbalance. This is measured by the share of tertiary industry GDP in total GDP. (4) Healthcare level (lnHos): Reflecting the level of public service provision, healthcare facilities significantly influence residential location choices. This is measured by the logarithm of the number of health institutions in each administrative district. (5) Education level (lnStu): Educational attainment directly determines individuals’ occupations and income levels. People with different education levels show significantly different preferences for work and residential locations. This is measured by the logarithm of the number of full-time teachers per 100 students. (6) Urbanization rate (Urb): Urbanization is an important contextual factor for jobs–housing balance. A higher level of urbanization may be associated with more developed urban functional zoning and transportation networks. This is measured by the ratio of the urban population to the total resident population in each district. (7) Commercial housing sales (lnAre): This reflects the scale of housing supply in a region. The supply and demand of commercial housing directly affect the jobs–housing balance. It is measured by the logarithm of the total floor area of commercial housing sold. (8) Level of social consumption (lnRet): The total retail sales of consumer goods reflect the degree of commercial prosperity and living convenience in a region, which may influence jobs–housing balance. This is measured by the logarithm of the actual total retail sales.

3.2.3. Mediator Variable

An increase in rail transit density enhances the accessibility of a given area, and this improved convenience attracts more enterprises and commercial establishments, thereby creating additional employment opportunities. As the working population flows into these well-connected areas, the demand for housing rises significantly. Given the limited supply of housing, this surge in demand drives up local housing prices and rents. The rise in housing costs raises the residential threshold for entering these areas, particularly affecting low-income groups and the middle class [31,32]. As housing prices and rents in these areas continue to increase, many residents who could previously afford to live there are forced to relocate. The rising cost of living compels low-income groups to move to areas farther from transit-convenient locations where housing is more affordable, thereby subjecting them to longer commuting times and greater commuting pressure [33]. Therefore, this study selects employment population and housing prices as mediating variables and employs a mediation effect model to test whether employment and housing prices serve as the mechanisms through which urban rail transit density affects jobs–housing balance.

3.3. Data

This study focuses on the 16 administrative districts of Beijing, covering the period from 2009 to 2022. The data used in this paper come from three main sources: (1) Beijing urban rail transit density data: The number and distribution of rail transit stations were compiled from resources available on the official website of Beijing Subway. (2) Meso-level data for Beijing: This includes GDP, resident population, number of health institutions, and the number of full-time teachers in regular secondary schools across Beijing’s 16 districts from 2009 to 2022. The data are sourced from the Beijing Statistical Yearbook and the Beijing Regional Statistical Yearbook. (3) Data compiled from public sources: Housing prices in each district of Beijing were collected from third-party data platforms such as Anjuke and Lianjia. Descriptive statistics for the above variables are presented in

Table 2. Summary statistics of variables.

Variable	Unit	Obs	Mean	Std. Dev.	Min	Max
Rail	stations/km2	224	0.512	0.399	0	1.868
RJHB	ratio	224	0.182	0.281	0	1.197
RJHR	ratio	224	11.069	0.709	9.939	12.844
lnPgdp	log(CNY/person)	224	10.635	0.391	9.948	11.502
lnPcdi	log(CNY/person)	224	0.684	0.419	0.116	5.419
Ind	%	224	6.319	0.637	4.424	7.521
lnHos	log(count)	224	2.64	0.234	1.999	3.164
lnStu	log(teachers/100 students)	224	13.156	1.266	9.153	15.847
Urb	%	224	5.811	1.177	3.025	8.198
lnSub	log(stations)	224	0.832	0.187	0.436	1.03
lnRet	log(CNY 100 million)	224	10.313	0.696	8.347	11.797
lnPri	log(CNY)	224	12.875	1.07	10.841	14.859

.

4. Empirical Result

4.1. Basic Analysis

This study conducts regression analysis based on the fixed-effects model specified in Equation (1) to examine the impact of urban rail transit density on the jobs–housing relationship in Beijing (see

Table 3. Baseline model results.

Variables	(1)	(2)
	RJHBit	RJHBit
Railit	0.214 ***	0.079 ***
	(0.032)	(0.032)
lnPgdp	1.5381 ***	1.3575 ***
	(0.097)	(0.159)
lnPcdi	−0.2024 ***	−0.1493 **
	(0.037)	(0.087)
Ind	0.1382 **	0.0336 *
	(0.063)	(0.068)
lnHos	0.0437	0.0363
	(0.046)	(0.028)
lnStu	0.1915 ***	−0.0116
	(0.035)	(0.039)
Urb	−0.1698 **	−0.4505 ***
	(0.069)	(0.085)
lnAre	−0.3029 ***	−0.8739 ***
	(0.123)	(0.332)
lnRet	−0.0497 ***	−0.0111 *
	(0.026)	(0.012)
Constant	−0.0860 ***	−0.0238 *
	(0.021)	(0.023)
Controls	yes	yes
Zone F.E.	no	yes
Year F.E.	no	yes
R2	0.614	0.784
Observations	224	224

Note: ***, **, * mean that the coefficient is significant at the level of 1%, 5%, and 10%.

). Table 3 shows that, regardless of whether time fixed effects and individual fixed effects are controlled for, the impact of urban rail transit density on the jobs–housing relationship remains significantly positive at the 1% significance level. This indicates that the denser the distribution of urban rail transit stations, the greater the deviation between residential and employment locations, and the more severe the jobs–housing separation.

The possible reasons are as follows: First, with the development of urban rail transit, the spatial structure of cities may change, leading to spatial separation between residential and employment areas. The speed and convenience of rail transit allow residents to tolerate longer commuting distances, which in turn promotes urban expansion into suburban areas. Second, rail transit stations—especially transfer hubs—often become employment centers. While the concentration of job opportunities in these areas attracts a large number of workers, the supply of nearby housing may be insufficient, resulting in jobs–housing imbalance. Lastly, residents’ commuting behavior and preferences also influence jobs–housing separation. Some individuals may prefer longer commutes in exchange for better living environments or lower housing costs.

From the perspective of the control variables, the level of economic development (lnPgdp) and industrial agglomeration (Ind) exhibit significant positive effects on the jobs–housing deviation index, indicating that regions with greater economic and industrial concentration tend to experience more pronounced jobs–housing separation. This may be because these areas have more concentrated employment opportunities and higher housing prices, forcing some residents to live farther from their workplaces, which exacerbates jobs–housing separation. Per capita disposable income (lnPcdi), urbanization rate (Urb), commercial housing sales (lnAre), and social consumption level (lnRet) show significant negative effects, suggesting that improvements in these factors help promote jobs–housing balance. Higher income levels enhance residents’ flexibility in housing choices, while better urbanization and housing supply help shorten jobs–housing distances. In addition, greater commercial prosperity and living convenience increase the willingness of people to live near their places of employment. The coefficients for healthcare level (lnHos) and education level (lnStu) are not statistically significant, indicating relatively limited impacts, possibly due to the diverse distribution of public services and varied preferences across groups.

4.2. Robustness Test

4.2.1. Replace the Explanatory Variable

An increase in the number of rail transit stations can enhance residents’ accessibility to employment opportunities, thereby influencing their commuting choices. At the same time, the expansion in the number of stations may reshape the urban spatial structure, prompting city development to extend along transit lines and forming residential and employment corridors along the rail network. This could have complex effects on the jobs–housing relationship. In this robustness test, we replace rail transit density with a new explanatory variable: lnSub, which represents the logarithm of the total number of urban rail transit stations within each administrative district. This variable serves as an alternative measure of transit system scale (Source: Beijing Municipal Commission of Transport). Therefore, to further explore the impact of the number of rail transit stations on the jobs–housing relationship, this study replaces rail transit density with the number of rail transit stations across Beijing’s 16 administrative districts and conducts a new regression analysis (see

Table 4. Replace the variables.

Variables	(3)	(4)	(5)	(6)
	RJHBit	RJHBit	RJHRit	RJHRit
Railit	0.5038 ***		1.5287 ***	1.7083 ***
	(0.073)		(0.098)	(0.114)
lnSub		0.1125 ***
		(0.034)
lnPgdp	0.3155 ***	0.0981 **	0.2023 ***	0.2251 ***
	(0.027)	(0.043)	(0.045)	(0.044)
lnPcdi	−0.1341 ***	−0.0352	−0.1334 **	−0.1257 **
	(0.045)	(0.079)	(0.063)	(0.061)
Ind	0.0258	0.0478	0.0432 *	0.0453 *
	(0.029)	(0.061)	(0.037)	(0.036)
lnHos	0.1061	0.1571	0.1887 *	0.1721
	(0.024)	(0.052)	(0.032)	(0.031)
lnStu	−0.0821 *	−0.2813 ***	−0.1711 **	−0.1224
	(0.052)	(0.098)	(0.069)	(0.068)
Urb	−0.1265	0.0421	−0.3041 ***	−0.3713 ***
	(0.083)	(0.165)	(0.119)	(0.114)
lnAre	−0.0223 *	−0.1939 ***	−0.0494 ***	−0.0383 **
	(0.014)	(0.023)	(0.018)	(0.018)
lnRet	−0.0638 ***	−0.1056 ***	−0.0833 ***	−0.0772 ***
	(0.015)	(0.032)	(0.016)	(0.028)
Constant	2.3581 ***	2.3135 ***	1.6768 ***	1.8396 ***
	(0.379)	(0.794)	(0.529)	(0.501)
Controls	no	yes	no	yes
Zone F.E.	yes	yes	yes	yes
Year F.E.	yes	yes	yes	yes
R2	0.615	0.575	0.789	0.778
Observations	224	224	224	224

Note: ***, **, * means that the coefficient is significant at the level of 1%, 5%, and 10%.

).

4.2.2. Replace the Dependent Variable

There are many ways to express jobs–housing balance. In order to increase the robustness of the empirical results, this paper replaces the explained variables and performs regression again. The job-to-resident ratio refers to the ratio of employed persons to the resident population in a geographic unit, directly reflecting the balance in the number of jobs in a region [34]. The higher the employment-to-residential ratio of an area, the greater the proportion of employment functions. The lower the job-to-resident ratio, the greater the proportion of residential functions. The job-to-resident ratio of a region can also directly reflect the employment-to-population ratio of the resident population. The job-to-resident ratio is calculated as follows:

$${JHR}_{it}={EMP}_{it}/{POP}_{it}$$

(4)

Among them, ${JHR}_{it}$ is the job/resident ratio in district $i$ in year $t$. ${EMP}_{it}$ is the number of jobs in district $i$ in year $t$, and is generally replaced by the total employed population, and ${POP}_{it}$ is the resident population in district $i$ in year $t$. The population aged 15–64 is defined as producers, and this paper only retains the resident population aged 15–64 who need employment as producers. Previous studies believe that when the ratio of work to housing is between 0.8 and 1.2, the region has achieved jobs–housing balance. In order to better measure the effectiveness of the metro opening on the jobs–housing balance, this paper uses the revised job-resident ratio instead of the revised jobs–housing deviation index (RJHR) for regression.

$$RJHR=|JHR-1|$$

(5)

See Table 4 for the regression results after replacing the explanatory variable with the revised job/resident ratio. From columns (4) and (5), we can see that regardless of whether control variables are added, the regression results are all significantly positive at the 1% confidence level. The regression coefficients are 0.112 and 0.110, respectively, indicating that the intervention of medical care, education, and other factors will weaken the impact of the metro’s opening on the separation between employment and housing. To sum up, the empirical results of this study are robust and reliable, further verifying that the metro opening exacerbates jobs–housing separation.

4.2.3. Winsorized Regression

In regression analysis and other models, extreme values may distort the estimation of model parameters. Winsorization can improve model fit and prediction accuracy. Compared with the method of directly deleting outliers, winsorized regression retains most of the original data in the dataset, avoiding potential information loss caused by data removal. It enhances the robustness of statistical analysis by reducing the influence of extreme values on statistical indicators such as mean, variance, and correlation coefficients. In this study, 1% and 5% winsorized regressions were conducted (see Table 4). As shown in Table 4, all four robustness tests—including replacing the dependent variable, replacing the explanatory variable, and the 1% and 5% winsorized regressions—produce results that are statistically significant at the 1% level. These findings further confirm the conclusion of the baseline regression: increasing urban rail transit density exacerbates jobs–housing separation.

4.3. Heterogeneity Analysis

Based on the level of regional development, Beijing’s 16 districts can be categorized into central urban areas, nearby suburban areas, and outer suburban areas (Figure 4). The impact of urban rail transit network density on the jobs–housing relationship varies across these regions. As shown in

Table 5. Heterogeneity analysis.

Variables	(7)	(8)	(9)	(10)	(11)	(12)
	Central Urban Areas		Nearby Suburban Areas		Outer Suburban Areas
	RJHBit	AJHRit	RJHBit	AJHRit	RJHBit	AJHRit
Railit	1.0059 ***	0.5013 ***	0.7357	1.9371	5.6985 ***	2.3584 **
	(0.125)	(0.144)	(2.269)	(1.411)	(28.249)	(16.788)
lnSub	0.2894 ***	−0.3772 ***	−0.3436 ***	−0.2735 ***	0.3331 *	0.0173
	(0.079)	(0.091)	(0.035)	(0.023)	(0.191)	(0.121)
lnPgdp	−0.0318	0.1767 *	0.2781 ***	0.1570 ***	−0.1782 **	−0.0459
	(0.085)	(0.096)	(0.069)	(0.041)	(0.082)	(0.044)
lnPcdi	−0.1165	−0.2213	0.0227	−0.0112	0.3078 ***	0.0727
	(0.257)	(0.287)	(0.027)	(0.016)	(0.122)	(0.076)
Ind	−0.0425	0.1190 *	0.0639 *	0.0748 ***	0.0455	0.0432 **
	(0.058)	(0.057)	(0.038)	(0.026)	(0.034)	(0.027)
lnHos	−0.2586 ***	−0.0269	−0.0973	−0.1518 **	−0.4169 ***	−0.2733 ***
	(0.084)	(0.108)	(0.109)	(0.068)	(0.116)	(0.073)
lnStu	−4.0727 **	2.6857	−0.1477	0.5092 ***	0.1031	0.3280
	(2.038)	(2.388)	(0.218)	(0.138)	(0.389)	(0.225)
Urb	−0.0625 ***	−0.0667 **	−0.0229	0.0025	−0.0021	−0.0215
	(0.028)	(0.029)	(0.025)	(0.018)	(0.023)	(0.016)
lnAre	−0.0396 *	−0.0357	0.0116	−0.0036	−0.0416	−0.0018
	(0.027)	(0.032)	(0.028)	(0.019)	(0.058)	(0.029)
lnRet	3.3542	0.6125	1.4063 **	1.4465 ***	−0.5524	1.3788 *
	(2.144)	(2.514)	(0.569)	(0.356)	(1.257)	(0.743)
Constant	1.0059 ***	0.5013 ***	0.7357	1.9371	5.6985 ***	2.3584 **
	(0.125)	(0.144)	(2.269)	(1.411)	(28.249)	(16.788)
Controls	yes	yes	yes	yes	yes	yes
Zone F.E.	yes	yes	yes	yes	yes	yes
Year F.E.	yes	yes	yes	yes	yes	yes
R2	0.954	0.463	0.789	0.911	0.715	0.697
Observations	84	84	84	84	56	56

Note: ***, **, * mean that the coefficient is significant at the level of 1%, 5%, and 10%. In this paper, the central urban areas refer to Dongcheng, Xicheng, Chaoyang, Haidian, Fengtai, and Shijingshan districts; the nearby suburban areas include Daxing, Changping, Tongzhou, Shunyi, Mentougou, and Fangshan districts; and the outer suburban areas refer to Pinggu, Miyun, Huairou, and Yanqing districts.

, rail transit density exacerbates jobs–housing separation in both central urban and outer suburban areas, with the effect being more pronounced in the outer suburbs, while no significant impact is observed in the nearby suburbs.

This can be explained as follows: the central urban areas of Beijing have high employment density and expensive housing. The development of rail transit extends people’s commuting radius, encouraging more residents to live in suburban areas with lower housing prices. Outer suburban areas are located farther from the city center and are mainly served by suburban railways, which operate at higher speeds, make fewer stops, and offer greater efficiency. These railways shorten commuting times between outer suburbs and the city center, allowing more residents of outer suburbs to work in the employment-rich central areas. In contrast, the nearby suburbs already maintain a relatively balanced jobs–housing relationship, and rail transit stations are evenly distributed, so the effect of transit density on jobs–housing separation in these areas is not significant.

4.4. Mechanism Test

4.4.1. Mechanism Analysis Using Employment Population as a Mediating Variable

After the metro’s opening, the supply and demand of surrounding land will usually change. Due to the improvement of transportation convenience and development potential, the demand for land may increase while the supply is relatively small. This supply/demand imbalance could lead to higher land and housing prices. High housing prices will increase the cost of living and commuting, limit individual employment options, lead to socio-economic differences, and affect the balance of work and housing. In order to test how the opening of the metro affects jobs–housing balance through housing prices, this paper constructs the following mediation effect model.

$${RJHB}_{it}={\alpha }_0+{\alpha }_1{Rail}_{it}+{\gamma }_j\sum X_{jit}+{\eta }_t+{\mu }_i+{\epsilon }_{it}$$

(6)

$${lnEmp}_{it}={\alpha }_2+{\alpha }_3{Rail}_{it}+{\gamma }_j\sum X_{jit}+{\eta }_t+{\mu }_i+{\epsilon }_{it}$$

(7)

$${RJHB}_{it}={\alpha }_4+{\alpha }_5{Rail}_{it}+{\alpha }_6{lnEmp}_{it}+{\gamma }_j\sum X_{jit}+{\eta }_t+{\mu }_i+{\epsilon }_{it}$$

(8)

As shown in

Table 6. Testing employment as a mediating mechanism.

Variables	(13)	(14)	(15)
	lnEmpit	RJHBit	RJHBit
lnEmpit		−0.1423 ***
		(0.042)
Railit	0.7695 ***	1.5223 ***	1.3575 ***
	(0.241)	(0.129)	(0.159)
lnPgdp	0.6298 ***	−0.0784	−0.1493 *
	(0.081)	(0.039)	(0.087)
lnPcdi	0.1887 ***	0.1382 ***	−0.0336
	(0.083)	(0.059)	(0.068)
Ind	−0.0779 ***	−0.0622 ***	0.0363 *
	(0.079)	(0.044)	(0.028)
lnHos	0.2769 ***	0.0436 ***	−0.0116
	(0.085)	(0.042)	(0.039)
lnStu	0.4749 ***	−0.0158 ***	−0.4505 ***
	(0.153)	(0.070)	(0.085)
Urb	0.3997 ***	−0.0578 **	−0.8739 ***
	(0.628)	(0.065)	(0.332)
lnAre	0.2379 **	0.0219 **	−0.0111
	(0.035)	(0.021)	(0.012)
lnRet	0.4492	0.0778 ***	−0.0238
	(0.031)	(0.029)	(0.023)
Constant	0.5658 ***	1.6238 ***	4.2918 ***
	(1.109)	(0.560)	(1.313)
Sobel	0.189 *** (z = 2.278)
Goodman-1	0.189 *** (z = 2.225)
Goodman-2	0.189 *** (z = 2.335)
Indirect effect	0.189 *** (z = 2.278)
Direct effect	1.169 *** (z = 12.416)
Total effect	1.358 *** (z = 11.742)
Mediating effect ratio	13.92%
Controls	yes	yes	yes
R2	0.872	0.742	0.729
Observations	224	224	224

Note: ***, **, * mean that the coefficient is significant at the level of 1%, 5%, and 10%.

, the mediating effect of employment population passes the significance test, with the mediation effect accounting for 13.92% of the total effect. The dense transportation network has turned certain areas into high-density clusters of commerce and employment, concentrating job opportunities in a limited number of regions. Areas with relatively poor transit accessibility, in contrast, see a reduction in job opportunities, leading to a more uneven distribution of jobs and residences. People living in more distant areas increasingly rely on rail transit for commuting, which further exacerbates jobs–housing separation. These findings indicate that employment population serves as a mediating mechanism through which rail transit density affects the jobs–housing relationship.

4.4.2. Mechanism Analysis Using Housing Prices as a Mediating Variable

Increasing rail transit density enhances transportation convenience. As accessibility and development potential improve, the demand for land may increase. This growing demand can disrupt the original supply/demand balance, leading to rising land and housing prices. High housing prices raise the cost of living and commuting, restrict individuals’ employment choices, widen socioeconomic disparities, and negatively affect the balance between jobs and housing. To examine how urban rail transit density affects the jobs–housing relationship through housing prices, this study constructs a mediation effect model with housing prices as the mediating variable, as shown in Equations (9)–(11). The regression results are presented in

Table 7. Testing housing prices as a mediating mechanism.

Variables	(16)	(17)	(18)
	lnPriit	RJHBit	RJHBit
lnPriit		0.2364 ***
		(0.002)
Railit	1.3559 ***	0.5828 ***	1.3575 ***
	(0.598)	(0.029)	(0.159)
lnPgdp	1.0612 ***	−0.0126	−0.1493 *
	(0.913)	(0.032)	(0.087)
lnPcdi	0.5859 ***	1.4053 ***	−0.0336
	(0.327)	(0.176)	(0.068)
Ind	0.3625 ***	0.2884 ***	0.0363 *
	(0.213)	(0.018)	(0.028)
lnHos	0.1862 ***	1.3519 ***	−0.0116
	(0.082)	(0.209)	(0.039)
lnStu	0.2092 ***	−0.3432 ***	−0.4505 ***
	(0.061)	(0.082)	(0.085)
Urb	0.9922 ***	−0.5818 ***	−0.8739 ***
	(0.628)	(0.045)	(0.332)
lnAre	0.5242 **	−0.4625 **	−0.0111
	(0.263)	(0.091)	(0.012)
lnRet	−0.8442	0.6748 ***	−0.0238
	(0.579)	(0.025)	(0.023)
Constant	−0.9213 ***	3.5813 ***	4.2918 ***
	(0.496)	(0.653)	(1.313)
Sobel	0.199 *** (z = 2.514)
Goodman-1	0.199 *** (z = 2.475)
Goodman-2	0.199 *** (z = 2.556)
Indirect effect	0.199 *** (z = 2.514)
Direct effect	1.159 *** (z = 15.204)
Total effect	1.358 *** (z = 15.705)
Mediating effect ratio	14.56%
Controls	yes	yes	yes
R2	0.858	0.819	0.729
Observations	224	224	224

Note: ***, **, * mean that the coefficient is significant at the level of 1%, 5%, and 10%.

.

$${RJHB}_{it}={\alpha }_0+{\alpha }_1{Rail}_{it}+{\gamma }_j\sum X_{jit}+{\eta }_t+{\mu }_i+{\epsilon }_{it}$$

(9)

$${lnPri}_{it}={\alpha }_2+{\alpha }_3{Rail}_{it}+{\gamma }_j\sum X_{jit}+{\eta }_t+{\mu }_i+{\epsilon }_{it}$$

(10)

$${RJHB}_{it}={\alpha }_4+{\alpha }_5{Rail}_{it}+{\alpha }_6{lnPri}_{it}+{\gamma }_j\sum X_{jit}+{\eta }_t+{\mu }_i+{\epsilon }_{it}$$

(11)

As shown in Table 7, the mediating effect of housing prices passes the significance test, with the mediation effect accounting for 14.65% of the total effect. The denser the rail transit stations, the higher the prices of commercial housing, and the higher the housing prices, the more severe the jobs–housing separation. This is because denser station distribution brings greater transportation convenience and more developed infrastructure, which in turn drives up housing prices. Higher housing prices raise the threshold for residential access, pushing more workers to live in more distant areas with lower housing costs. Therefore, housing prices serve as a transmission mechanism through which rail transit density affects the jobs–housing relationship.

These results confirm that housing prices serve as a significant channel through which rail transit density affects jobs–housing balance. The increase in rail station density raises housing costs near transit hubs, which, while improving accessibility, can displace low- and middle-income residents to more remote areas, thereby exacerbating spatial mismatch and commuting burdens. This highlights the need for coordinated planning of transit and housing policies to ensure that the benefits of rail investments are inclusive and do not unintentionally reinforce social or spatial inequalities.

5. Discussion

This study proposed four hypotheses regarding the impact of urban rail transit density on jobs–housing balance and its underlying mechanisms. The empirical analysis provides strong support for Hypothesis 1, increasing urban rail transit density tends to exacerbate jobs–housing separation, particularly in central urban areas and outer suburbs. This finding aligns with prior studies such as Zhou et al. [14] and Huang et al. [15], who noted that rail transit development in concentric urban structures can intensify jobs–housing separation by reshaping land use and residential patterns. Hypothesis 2 is also supported, as the impact of rail transit density shows regional heterogeneity, with no significant effect found in nearby suburban areas. This is consistent with the conclusions of Blumenberg and Siddiq [16] and Zhou et al. [17], who emphasized that the relationship between jobs–housing balance and commuting patterns varies across regions due to differences in urban form and sociodemographic characteristics. Hypothesis 3, regarding employment agglomeration as a mediating mechanism, is validated by the mediation analysis results. This complements findings by Ge and Wang [25] and Bautista-Hernandez [22], who highlighted that transit construction fosters employment concentration, potentially increasing jobs–housing separation. Finally, Hypothesis 4 is supported: housing prices play a significant mediating role in the relationship between rail transit density and jobs–housing separation. This confirms and extends observations by Zhou et al. [28] that rising housing costs near transit lines force low- and middle-income residents to relocate farther from job centers.

These findings are broadly consistent with and extend prior studies cited in our literature review. In particular, our study contributes new empirical evidence by quantifying the mediation effects of employment concentration and housing prices and explicitly confirming regional heterogeneity in the impacts of rail transit density. These results provide practical insights for urban transit and housing policy planning.

6. Conclusions and Enlightenment

Based on meso-level panel data from Beijing from 2009 to 2022, this study draws the following conclusions: ① Increasing urban rail transit density generally exacerbates jobs–housing separation, with regional heterogeneity in its effects. In both outer suburbs and central urban areas, high-density rail transit has not alleviated but instead intensified jobs–housing imbalance. In nearby suburban areas, however, no significant relationship is observed. ② Mechanism tests reveal that employment population and housing prices are important pathways through which rail transit density influences the jobs–housing relationship. While high-density transit improves commuting convenience and increases employment opportunities, it also encourages longer commutes and contributes to jobs–housing separation. Additionally, the associated rise in housing prices raises the residential threshold, further intensifying jobs–housing separation.

Policy implications of this study include the following: ① Avoid blind expansion and promote rational planning of rail transit. The findings suggest that blindly pursuing high-density rail transit construction may aggravate jobs–housing separation. Urban rail transit planning should be scientifically designed based on the characteristics of local jobs–housing distribution and actual demand to avoid resource waste and further imbalance due to overexpansion. In particular, for central and outer suburban areas, efforts should focus on optimizing the operational efficiency of the existing network rather than merely increasing station density. For nearby suburbs where rail transit is not yet well developed, the goal should be to balance job and housing distribution by reasonably planning station locations and routes while minimizing the stimulation of housing prices. ② Optimize employment spatial distribution and reduce excessive job concentration. This study reveals that rail transit density influences jobs–housing balance through employment population agglomeration. Urban development planning should guide the reasonable spatial distribution of job opportunities and avoid overconcentration in central areas. For example, through policy support and resource allocation, industries can be encouraged to shift toward suburban areas, forming a polycentric and multi-nodal employment pattern. This would reduce commuting pressure along rail corridors and shorten jobs–housing distances. At the same time, public services in suburban and outer suburban areas should be improved to attract more businesses and talent. ③ Strengthen affordable housing supply and lower residential thresholds. The study shows that increased rail transit density significantly drives up housing prices along transit lines, raising the cost of living and pushing low-income groups toward outer suburbs. To address this issue, housing security policies should be strengthened in transit-accessible areas by increasing the supply of affordable housing and public rental units, providing more viable options for low- and middle-income families. Additionally, appropriate price control measures should be implemented in areas with rapidly rising housing prices to prevent housing costs from further undermining the jobs–housing balance and to promote both spatial equity and social fairness.