Transport equity assessment based on accessibility disparities in terms of multi-job opportunities across Beijing

Abstract

Evaluating equal access to opportunities across social groups is an important aspect of transport equity research, but existing schemes rarely focus on accessibility to specific opportunities for specific groups. This study distinguishes between different education groups and job sectors which come from Nation Census Statistics and National Economic Census, respectively, with the aim of evaluating transport equity in Beijing by examining the disparity between high- and low-education groups in accessing job opportunities of specific sectors with different education requirements. Using the bivariate local spatial autocorrelation model to reveal the relationship between provision of accessibility and the population quantity of different groups in regions based on different transport modes. Regions with low accessibility level but high groups numbers may rise to inequity of transport. It was found that high-education groups had more equal accessibility to high-education job opportunities than low-education groups had in their access to low-education jobs. In terms of diversity of opportunity, the differences in job accessibility between low- and high-education groups tended to diminish, and equity in accessibility was also improved. Disadvantaged areas with high population demands for jobs exhibited lower accessibility levels by public transport compared to car travel, and more of the low-education groups were located in disadvantaged areas and obtained fewer opportunities, particularly for the groups without Hukou status. Diversity of opportunity helped to reduce the difference between car travel and public transport in disadvantaged areas, and also decreased the inequities in accessibility among different educational groups. To improve transport equity, it is recommended to prioritize public transport in fringe areas and disadvantaged areas, and the development of multi-job sectors in these areas should be undertaken to cater to the demands of different education groups.

Transport equity is a significant concern for researchers and transport agencies globally. What is actually assessed to determine the equity of transport has been the focus of many studies. The definition of equity is one of the primary problems in the evaluation of transport equity, and the core concept is focused on the distribution of benefits or costs/burdens among social members provided by transport systems¹. Even if it is difficult to conceptualize equity, scholars have examined transport equity from a variety of perspectives. A widely recognized framework distinguishes between horizontal equity and vertical equity: the horizontal perspective advocates the equity of transport required to provide the same level of services for all groups and different locations, it also reflects spatial equity². The vertical perspective stresses the resource distribution based on the needs and abilities of different sociodemographic groups, particularly for disadvantaged groups, and identifying who has disadvantaged status is a key^3,4. Since determining transport equity is complex and not straightforward⁵, the distributional effects of transport systems, particularly the equitable distribution of transport services and accessibility to key services or opportunities among specific groups, have been widely considered in the transport equity studies^6,7,8,9. Accessibility serves as an effective lens for understanding equity and acts as a common measure for revealing distributional differences in access to different opportunities among various groups in transport research^10,11,12.

Accessibility can be defined in various ways, including the potential of opportunities for interaction¹³, the ease of reaching land use given the transport system, the capacity to access opportunities¹⁴, the potential to reach spatially distributed opportunities¹⁵, the number of opportunities reachable within a given travel time or distance by specific modes¹⁶, or the feasibility of people reaching a specific activity or place¹⁷. As a result, it represents the feasibility for people to reach specific destinations or opportunities^18,19. Improving people’s accessibility to job opportunities is a crucial goal for achieving transport equity. Job accessibility can be defined as the ease of reaching potential job opportunities²⁰. Meanwhile, employment opportunities play a core role in human development, contribute to improving well-being and quality of life, and have an impact on social equity^21,22. Job accessibility also impacts the economic mobility of people; low job accessibility is associated with poor socioeconomic status and affects social members’ participation in social activities²³. Equity of accessibility to job opportunities is therefore an important aspect of transport equity analysis, focusing on the distribution of job opportunities for socially vulnerable members has become one of the primary contents of equity²⁴. To promote social equity in transport, it is essential to improve job accessibility across groups with disadvantaged characteristics such as physical, financial, cultural, gender, skill, or educational status, because these groups are apt to gain poorer job opportunities²⁵. Therefore, a focus on the job accessibility gains of disadvantaged groups has become a key issue in analyzing transport equity.

The matched relationship between accessibility provision and demands of different social groups has been particularly focused on evaluating transport equity²⁶. The mismatch between accessibility and groups has a significant influence on equity, which can be categorized as spatial and non-spatial mismatches. The former stresses physical separation between accessibility and social groups;transport is a key factor influencing spatial mismatch, and increasing distance to job opportunities will exacerbate transport inequity for disadvantaged groups²⁷. The non-spatial mismatch between job accessibility and certain groups also impacts equity, although residents are spatially close to jobs, they may not be able to secure employment due to differing skill requirements and educational qualifications²⁸. Moreover, not every job can provide an economic opportunity; only accessible and valued jobs for specific groups are really meaningful^29,30, and matched job accessibility for specific groups is strongly linked to equity³¹. Considering different groups with different locations and social characteristics have specific demand on jobs, comparing the accessibility gap between different groups, and different transport modes has become a commonly and straightforwardly used evaluation method in transport equity analysis¹⁹, revealing the underlying spatial and non-spatial matches.

Measuring accessibility is the key step in the evaluation of transport equity. Accessibility measurements have been defined in terms of infrastructure-based, person-based, utility-based, or location-based approaches³². Specific approaches, such as cumulative or opportunity-based accessibility, space-time based accessibility, relative accessibility, and gravity-based accessibility, have also been proposed³³. When evaluating transport equity, it is essential to select an appropriate accessibility method. Gravity-based and cumulative opportunity measures are commonly used accessibility evaluation methods for assessing transport equity, both types of methods are specific methods of location-based accessibility, evaluating disparities in accessibility across different locations and under different travel modes. Some studies have suggested that gravity measures suffer from drawbacks such as computational complexity, difficulty in interpretation, and a tendency to overestimate accessibility scores in relatively isolated areas³⁴, while cumulative opportunity accessibility is easier to interpret and easier to use in the discussion of policy with decision-makers³⁵. Meanwhile, accessibility measurement involves a wide range of destinations and opportunities being made available to individuals or groups. The incorporation of a diverse range of activities has gained attention in accessibility studies, but it has seldom been the focus of a transport equity analysis^36,37,38. Some theories have pointed out the importance of diversity for people. Maslow and ERG theories have classified human needs into many categories, with each category representing different opportunities in the real world, which can then be applied to categorize various destinations, services, or activities in accessibility analysis³⁶. Nussbaum expanded on Sen’s capability approach (CA) by providing a list of diverse central human capabilities to assess transport equity, such as life, bodily health, and bodily integrity, and each capability represents an activity or destination³⁹. In reality, both the demand and supply sides have an inherent diversity, with the demand side originating from individuals, while the supply side(accessibility) encompasses diverse opportunities or destinations⁴⁰. People’s diverse needs mean they have multiple options, and different options will produce different values for individuals or groups, with the value of these options having been investigated⁴¹. Previous studies have analyzed the equity of accessibility, focusing on the differences in accessibility to various types of destinations(employment, groceries, parks, schools, and health care) or accessibility to various types of destinations in a specific activity among groups^12,42,43. Researchers have analyzed the accessibility to different types of jobs for residents and evaluated the equity of job accessibility for unemployed workers⁴⁴, and also incorporated geographical skill matches into the accessibility metric to assess the equalities of job accessibility⁴⁵. However, few studies focus on the specific job demands for specific groups by classifying job types to assess the equity of job accessibility, and existing accessibility measures also neglect diverse destinations or opportunities for specific groups’ needs^13,46,47.

As work commuting constitutes a high proportion of urban residents’ trips, for example, it has reached 47.4% in Beijing’s urban area (2021), and job accessibility serves as a prerequisite for enabling and safeguarding other social activities. Consequently, equity is typically evaluated by assessing the distribution of job accessibility across different geographical spaces and demographic groups. However, prior research has predominantly focused on the distribution of job accessibility, that is, the distribution of all the job opportunities among groups, with limited differentiation of employment types. Research cared about the equity of access to all jobs and assumed that everyone can access all the jobs without distinguishing among sectors²⁸. In reality, different employment sectors have specific demands for workers with varying backgrounds, such as education and skill levels, and the economic opportunities represented by different sectors vary significantly. Since different demographic groups are matched with specific employment sectors, analyzing the distribution of accessibility to specific industries among specific groups is essential to reveal equity implications. Yet, this aspect is rarely addressed or neglected in transport equity analyses. Furthermore, while the impact of destination diversity on equity has received attention^48,49, the influence of diversity within job opportunities on transport equity has been less explored^16,46.

This study considered job as a specific desirable opportunity, and recognized that the diversity of opportunities in job accessibility has not been thoroughly analyzed in previous evaluations of equity⁵⁰. All jobs were categorized into various sectors, and job accessibility was considered in different job sectors. Considering the matching relationship between different education groups and related job types in accessibility analysis with a vertical equity perspective, the disparities of different job accessibilities by different transport modes to their matched education groups were compared to reveal the equity of job accessibility in Beijing, China. Furthermore, the study also measured job accessibility with diversity and examined the influence of diversity of job opportunity on the equitable distribution of job accessibility among different educational groups, and followed the rationale that emphasizing socially disadvantaged groups would lead to job accessibility gains. Through analyzing the disparity of specific job sector accessibility and its diversity among specific educational groups to reveal transport equity, this research would provide a reference for transport equity-related policies in cities.

Study context

Transport equity has been a specific focus in developing countries in recent years, and these studies have used job accessibility to identify the equity of transport^50,51. As in other developing countries, the rapid urban expansion experienced during the development of China has increased the motorization of people’s mobility⁵². In the context of urban motorization with transport construction, transport infrastructure and services have improved the ease of movement in Chinese cities. Consequently, the nexus between transport and mobility has received widespread attention.

Researchers have focused on transport equity in Chinese cities, with analyzing accessibility distribution, travel satisfaction, fare for specific transport modes, physical and mental health, commuting, and the gap between transport supply and demand^53,54,55. Zhang and Zhao⁶reported that the research focus has shifted from mobility-based equity to accessibility distribution in China, and the equity effect in accessibility distribution under transport project, policy, and planning is one of the important contents, and transport equity means equity of accessibility distribution and especially improves disadvantaged groups’ accessibility. To determine the distribution of accessibility, researchers have studied the inequality of accessibility in Chinese cities with consideration of the accessibility of disadvantaged groups to different destinations or services⁵⁶, it also pays attention to the distribution of job accessibility under different modes of transport, however, there is less attention paid to the diversity and population matching of job accessibility in the equity evaluation.

With the advancement of urban motorization in China, cities face challenges such as transport congestion, environmental pollution, spatial constraints (particularly growing parking demands stemming from surging private car ownership), and social equity concerns. In order to address these issues, promoting public transport priority has become one of the crucial strategies in China. A key transport policy within this strategy involves reducing private car usage by encouraging shifts to public transport. Meanwhile, governments have significantly increased investment in urban public transport infrastructure, but it remains unclear whether residents in different locations have equitable access to these services. Comparing the modal disparities between public transport and private car travel helps reveal which locations are disadvantaged in terms of public transport provision under the priority development framework. This understanding is thus vital for comprehending the equity implications of urban transport planning. Moreover, Beijing’s commuting data reveals that in 2023, the average commuting distance within the central city was 13.2 km(Beijing Transport Institute, 2024), and over 30% of work commutes exceeded 60 min in duration. This distance is suitable for motorized transport modes such as buses, subways, and private cars. A comparative analysis of public buses and private cars better conforms to the practical realities of work commuting in Beijing. In particular, Beijing has a typical disparity between public and private transport provision⁵⁰, the scale of increasing public transport infrastructure didn’t quell the demand for private cars, and it had to limit the number of private cars, so it is necessary to discuss equity among these two modes.

Methods and data

Study area

Beijing city is the study area, which is the capital city of China, and covering a total area of 16,410 km². The subdistrict administrative units (Streets within urban areas and villages or townships in suburban and peripheral districts) of the Beijing division were from the sixth population census³². This study focuses on transport equity at the street/township scale. The fundamental administrative unit of Chinese cities is the street/township, and government socioeconomic statistics are mainly reported at this spatial level^50,57. Because analyzing social equity requires social statistics indicators, townships are often applied as the research unit in China. In this study, to meet the research requirements while considering the overlap of areas’ boundaries, some townships were integrated into one area in Beijing (e.g., Huaxiang and Xincun streets were merged into one area, and Lugouqiao town and Lugouqiao street were merged into one area), and job and population data from two areas are also aggregated into the merged area. Consequently, the basic administrative units of Beijing consisted of 321 township-scale units. We take these units as demand locations and their social groups as the measure of demand sides (Fig. 1).

Fig. 1

The area of study.

Accessibility metric for transport equity

As discussed above, we investigate the gap of job accessibility based on different jobs and different transport modes across different groups. The difference between the supply of accessible opportunities and the demand of social groups for these opportunities was assessed to determine the equity of accessibility distribution via different transport modes²⁵. To assess the diversity of job opportunities for the needs of specific groups, the gaps of diverse job opportunities between groups, modes were also determined. The cumulative measure was adopted in this study and was easier to use to assess the difference between accessibility and population than gravity methods. Accessibility was measured as follows:

$$\:{\text{A}}_{\text{i}(\text{p}/\text{c})}={\sum\:}_{\text{j}=1}^{\text{j}}\text{O}\text{T}\text{i}\text{j}(\text{p}/\text{c})$$

(1)

where if Tij(p/c) ≤ T, O = Oj, otherwise, O = 0. O represents the number of jobs, i and j represent township i and township j, respectively. T represents the threshold of travel time, p and c represent public transport and car travel, respectively. To consider different job sectors, accessibility types were categorized into total accessibility(TA), low education job accessibility(LA), and high education job accessibility(HA), and the accessibility of these job categories by car and public transport wasmeasured.

The travel time threshold is a key factor for cumulative accessibility measurement. This parameter can be selected from transport planning standards/goals or obtained from real travel data. In previous studies, thresholds of 15, 30, 45, and 60 minutes have been applied for different global cities. A single time threshold cannot capture the real inequities or reflect the differences in the social, economic, and historical contexts of different cities^6,16. Therefore, travel time thresholds of 15, 30, 45, and 60 minutes were set in this study. The reasons for setting these thresholds were as follows: the 60 min commuting circle is a future goal in the planning of transport systems for Beijing, and is closer to the current average travel time (51 min) according to the Beijing Commuting Characteristics Annual Report from Beijing Transport Institute in 2023. A 45 min travel time to employment destinations was suggested as the general commuting time in Chinese cities (including public transport and car) by the Commute Monitoring Report of Major Cities in China. The 30 min threshold is widely used in accessibility studies, while 15 min is typically the threshold of the community life circle. The 15 min threshold mainly depends on walking within the community life circle, and this parameter was used for a sensitivity analysis in this study.

To measure accessibility, calculating travel time between origins and destinations (O–D) is one of the key steps, the study uses ArcGIS software to generate centroids as origins and destinations based on townships’ administrative boundaries. The travel time data between townships’ centroids by two different transport modes (public transport and car travel) was computed using the Gaode Map(Amap) API at peak time. Gaode Map is a premier navigation and location-based service platform owned by Alibaba. It dominates the Chinese market with precise real-time driving, walking, and public transit navigation, leveraging massive user data and AI for optimal routing and traffic avoidance^22,48. In this work, our travel time is calculated through the Gaode map API (https://lbs.amap.com/api/webservice/guide/api/newroute). This map platform can provide an optimized route to reach destinations based on user demand through calling API, and provides a set of algorithms to plan routes between origins and destinations by car, public transport, walking, and other transport modes, and it can calculate real travel time from origins and destinations under one transport mode or multiple transport modes.

To measure the diverse opportunities in job accessibility, the Shannon entropy index was used³⁸. The formula is as follows:

$$D = \sum\limits_{{i = 1}} { - p_{i} \ln (p_{i} )}$$

(2)

where pi equals the proportion of job accessibility at specific location occupied by specific job type i based on accessibility.

To identify the difference between two transport modes, the diversity index was converted to the relative Shannon entropy index. The Shannon’s entropy-based index is as follows:

$$H = \frac{{\sum\nolimits_{{i = 1}} { - p_{i} \ln (p_{i} )} }}{m}$$

(3)

where m equals the number of job sectors. The value of H ranges from 0 to 1.

Spatial correlation analysis

A geographical spatial autocorrelation model is typically used to identify spatial correlations between populations and accessibility in the assessment of transport equity⁵⁷. Global bivariate Moran’s I scatter plots and a bivariate local indicator of spatial autocorrelation (LISA) can be used to analyze the global and local spatial correlations between accessibility and populations, respectively. The former is represented by Moran’s I value, with a strong spatial correlation producing a high value of Moran’s I; the latter is expressed as LISA maps that reveal the local clusters of the spatial associations between accessibilities and populations^58,59. Using GeoDA software to analyze spatial autocorrelation characteristics through these two indicators, this study identifies accessibility disparities among social groups and examines their spatial clustering patterns (Fig. 2).

Global spatial autocorrelation:

$$GMI = \left( {n/\sum\limits_{{i = 1}}^{n} {\sum\limits_{{j = 1}}^{n} {Wij} } } \right) \times \left[ {\sum\limits_{{i = 1}}^{n} {\sum\limits_{{j = 1}}^{n} {Wij} } (X_{i} - \mu )(X_{j} - \mu )/\sum\limits_{{i = 1}}^{n} {(X_{i} - \mu )^{2} } } \right]$$

(4)

where n represents the number of administrative areas, and W represents the spatial weight matrix. X_i and X_j represent the accessibility values/the number of populations within each administrative areas, and µ represents the mean of observations. When GMI > 0, it indicates that accessibility and populations have a positive spatial correlation; whereas a negative indicator (GMI < 0) indicates a negative spatial correlation between accessibility and populations.

Local spatial autocorrelation:

$$I = \frac{{X_{i} - \mu }}{{\sum\limits_{{i = 1}}^{n} {(X_{i} - \mu )^{2} } }}\sum\limits_{{i \ne j}}^{n} {W_{{ij}} (X_{i} - \mu )}$$

(5)

where I represents the clustering correlation between the area and its neighboring areas. Local spatial autocorrelation analysis generated LISA cluster maps to reveal the spatial patterns of spatial relationships between accessibility and different population groups. The LISA analysis identified four distinct cluster types reflecting correlation between accessibility levels and population groups: H–H denotes areas with high accessibility and large population groups; L–L indicates areas with both low accessibility and small population groups; H–L signifies areas of high accessibility but small population groups; L–H identifies areas of low accessibility despite large population groups. The L–H pattern reveals a critical spatial mismatch between accessibility and population demand. These disadvantaged areas should be prioritized for accessibility improvements. Consequently, identifying L–H areas is important as they are the key target areas for advancing transport equity.

Fig. 2

Technical flow chart of the study.

Data source and processing

The data sources used in the study included employment and population data at the town level, which were obtained from the third national economic censuses and the sixth national population censuses, respectively. Considering the matching relationship between different educational groups and industrial sectors, this study focused on low-education groups (without college degrees) and high-education groups (with college degrees and above)⁶⁰. We assume that members of specific educational groups have a high possible of reaching nearby specific sectors if they live near these sectors, but if they have no qualifications for these sectors, a mismatch may occur (e.g., skill/education mismatch)⁴⁵. Hukou is a system of household registration used in mainland China, it officially identifies a person as a permanent resident of a place, and this system is an important mechanism for population management. China’s urban hukou system grants registered residents access to urban welfare benefits including elementary education, housing, and healthcare, but the non-registered and floating populations who do not hold a local household/hukou registration do not receive equal treatment. This study employed national population census data to compare job accessibility disparities between non-registered and registered populations across different education levels. Finally, the population data were divided into four distinct groups of people based on registration status and educational attainment: registered low-education groups (RLG), registered high-education groups (RHG), non-registered low-education groups (NLG), and registered high-education groups (NHG).

Employment data was available from different enterprises based on the third National Economic Censuses, with each enterprise having a specific code related to one industrial sector. The data included information about each enterprise’s number of employees, industrial sector (industrial sector code), address, and enterprise name. The number of employees of each enterprise was used as job data, and these employees are at the real places where they work. Enterprises were classified into 18 major sectors based on their industrial sector code, which is from the criterion of Industrial Classification for National Economic Activities (GB/T4754-2017), and then the number of jobs in different sectors was aggregated based on each enterprise’s location at the town level. To distinguish the educational requirements of particular jobs, this study classified employment data into different sectors with specific education demands and measured their accessibility. Referring to previous studies⁶⁰, each industrial sector could be referred to as low-education and high-education level requirements, and the educational level required for jobs in specific sectors was expressed as the percentage of high- or low-education workers in the total workforce. If the percentage of low-education workers was lower than that of high-education workers in one sector, then all jobs in this sector were classified as high-education jobs. Table 1 shows the employment classification with different educational requirements.

Table 1 Employment classification.

Results and analysis

Spatial differences in accessibility between private car and public transport

Spatial difference in job accessibility

The results as Figs. 3 and 4 shown, the spatial distribution of accessibility at the town level, which has revealed differences in the provision of job accessibility between car travel and public transport. The distribution of towns with below average accessibility was mapped to identify areas where there were accessibility advantages when using car travel or public transport. With public transport, for 230 areas the TA was below the average accessibility level within the 15 min travel time threshold, and 76% of the areas had below-average job opportunities. For LA, 221 areas were below the average accessibility level, and the LA distribution had the same trend as the TA distribution. For HA, 253 areas had a below-average accessibility level. When the travel time threshold was 60 min, the number of areas below the average level was 218, 217, and 219 for TA, LA, and HA, respectively. With car travel, within the 15 min travel time threshold, the number of areas with a below-average accessibility level was 230, 225, and 240 for TA, LA, and HA, respectively. For the 60 min threshold, there were similar characteristics for public transport, with the number of areas below the average accessibility level reduced to 129, 134, and 124 for TA, LA, and HA, respectively. It was evident that more areas had access to potential job opportunities through car travel than public transport, but many areas consistently had below-average accessibility levels, resulting in fewer job opportunities. There were 178 areas within all travel time thresholds (15, 30, 45, and 60 min) by public transport that were below the average level of job accessibility, and this number was higher than that for car travel (119).

There was a significant difference in access to job opportunities between the two transport modes. When the travel time threshold increased, the number of areas with above-average accessibility levels by car was greater than that of public transport. For public transport, areas with above-average accessibility levels had a higher concentration in the central area compared to private car. Moreover, different types of job accessibility have a similar spatial distribution, and accessibility to high-education job opportunities by public transport was spatially distributed more unequally than total job opportunities or low-education job opportunities.

Fig. 3

The areas with above and below average accessibility (TA, LA, HA) by car and public transport (within 15 min).

Fig. 4

The areas with above and below average accessibility (TA, LA, HA) by car and public transport (within 60 min).

Spatial differences in the diversity of accessible job opportunities

Figures 5, 6, 7, and 8 show the spatial distribution of the diversity of accessible job opportunities for low- and high-education jobs. The highest diversity of accessible job opportunities was observed in the inner areas of the 6th ring road of Beijing. The average diversity level of accessible low-education jobs is higher than that of high-education jobs. There were small differences between the two transport modes within different travel thresholds, and a similar distribution of high-diversity accessible job opportunities. When the travel time increased to 60 min, the difference in the diversity of accessible job opportunities between car travel and public transport decreased. Car travel had a clear advantage in providing better access to a diversity of accessible jobs in the areas, especially for low-education jobs. Car travel improved the diversity of accessible job opportunities in urban fringe areas compared to public transport, while public transport tended to improve the diversity of accessible job opportunities in the central area.

Fig. 5

Diversity of accessibility (HA) by car and public transport (within 15 min).

Fig. 6

Diversity of accessibility (HA) by car and public transport(within 60 min).

Fig. 7

Diversity of accessibility (LA) by car and public transport (within 15 min).

Fig. 8

Diversity of accessibility (LA) by car and public transport (within 60 min).

The gaps between the supply of accessibility and different groups

Spatial relationships between accessibility and different groups

The bivariate spatial autocorrelation model was employed to examine the spatial relationship between accessibility and each population group based on Moran’s index. According to the classifications of job sectors and population groups, low-education groups were associated with low-education requirement jobs, while high-education groups corresponded to high-education requirement jobs. As shown in Table 2, Moran’s I values between car-based accessibility and population groups were higher than those between public-transport-based accessibility and population groups. A higher Moran’s I indicates that the spatial distribution of accessibility and groups had a high spatial correlation. This suggests that the provision of public-transport-based job accessibility is less well matched to the distribution of populations with different educational levels than that based on car travel.

Comparing low education groups with high-education groups, Table 3 shows that Moran’s I values between high-education-requirement job accessibility and high-education groups were higher than those between low-education-requirement job accessibility and low-education groups. This indicates a better spatial match between high-education groups and job accessibility compared to low-education groups. Hukou status also significantly influenced the spatial correlation between job accessibility and different population groups. A higher Moran’s I indicated that registered residents (with Hukou) exhibited a better spatial match with relevant job opportunities than non-registered residents (without Hukou).

Table 2 The Spatial correlation between total accessibility and different educational groups.

Table 3 The Spatial correlation between accessibility by relative job sectors and different educational groups.

Spatial clusters of the disparities between accessibility and different groups

Figures 9, 10, 11, 12, 13, and 14 present statistically significant local clusters (p < 0.1) of accessibility and population groups (p < 0.1), indicating the H–H, L–L, H–L, and L–H associations between the different groups and accessibility levels across neighborhoods. The cluster maps exhibited similar spatial patterns across the different travel time thresholds. Among all the cluster modes, “L–L” was the most common, and these areas were located in the urban fringe district far from the urban center, with notable concentrations in the southwest and north regions. In “L–L” areas, both supply of accessibility and population demand were low, and the relative demand for job opportunities from both low-and high-education groups was also significantly low. The other cluster modes were located outside of the central area (outside the 3rd ring road), and significant H–H and L–H cluster modes were consistently observed adjacent to each other across all travel time thresholds.

Table 4 indicates variation in the number of L–H areas across transport modes and population groups. Specifically, the spatial association analysis of public transport accessibility (TA) showed that low-education groups without hukou status had the highest number of L–H areas (32), followed by low-education groups with hukou status (19). For registered high-education groups, there was a smaller number of L–H areas (14) than for low-education groups. When considering car travel, the number of L–H areas for registered low-education groups and TA was higher than for other groups, but it was lower at most travel times when considering public transport. For LA and low-education groups, those without Hukou status had more L–H areas compared to registered low-education groups. For HA and high-education groups, the registered high-education groups had an advantage in terms of accessibility, with fewer L–H areas than for low-education groups.

As the travel time threshold increased, the disparity between the two transport modes became more pronounced. Car travel consistently achieved better alignment between accessibility and population groups compared to public transport. For low-education groups, there were more areas with low accessibility, with higher educated groups generally disadvantaged less than low-education groups. Low-education groups face heightened transport disadvantage, characterized by inferior job accessibility in their residential locations. Populations without Hukou status experienced a generally low accessibility level and were at a disadvantage in seeking job opportunities, especially in areas with low-education groups.

Table 4 The number of areas of “L–H” mode between high/low educational groups and accessibilities.

Fig. 9

Local spatial clustering maps of TA and groups (within 15 min).

Fig. 10

Local spatial clustering maps of LA and low-education groups (within 15 min).

Fig. 11

Local spatial clustering maps of HA and high-education groups (within 15 min).

Fig. 12

Local spatial clustering maps of TA and groups (within 60 min).

Fig. 13

Local spatial clustering maps of LA and low-education groups (within 60 min).

Fig. 14

Local spatial clustering maps of HA and high-education groups (within 60 min).

We identified disadvantaged areas characterized by the L-H cluster mode across three job accessibility metrics (TA, LA, HA) for each education group. For public transport, the number of such areas increased from four at the 15-minute travel time threshold to seven at the 60-minute threshold (Table 5). The number of disadvantaged areas for car travel was less than for public transport at the 15 to 60 min thresholds. Some specific areas, such as Beiqijia, Dongxiaokou, and Wanpingcheng, had disadvantaged status for both transport modes. These findings suggested that car travel could reduce the number of inequitable areas and improve the accessibility and equity of disadvantaged areas with large population agglomerations. However, for public transport, the number of unequal areas with low accessibility increased as travel time increased. Although public transport broadened the geographical reach of employment opportunities, disparities in access to potential jobs persisted in some areas.

Table 5 The distribution of areas in the L–H cluster mode between accessibility and population groups.

Spatial clusters of the disparities between diversity of job accessibility and each group

For the diversity of job accessibility and different population groups, the global Moran’s I presented a small gap between the two transport modes (Table 6). With increasing travel time, the degree of spatial correlation between the diversity of job accessibility and different population groups increased significantly. The correlation between the non-registered high-education group and HA was lower than for the other groups, and the spatial correlation between accessibility and population groups by public transport was higher than by car travel, and this characteristic differed from the above situation in which diversity was not taken into account.

Figures 15 , 16, 17, and 18 show maps of the local clusters of spatial association between diversity of opportunity and population groups. The L–L areas were concentrated in the northern and southwestern districts, and the H–L areas were located adjacent to the L–L areas. Most of the H–H areas were located in the district between the 3rd and the 6th ring road, and the L–H areas had the same location as the H–H areas. The number of L–H areas by public transport was higher than by car travel at all travel time thresholds, and there were no L–H areas for car travel at the 60 min threshold (Table 7).With public transport, for the diversity of opportunity among low-education groups, the number of L–H areas for the non-registered low-education population group was higher than for other groups. Similarly, with car travel, the number of L–H areas for non-registered groups was higher than that for registered groups. This means that for all population groups, the Hukou status had the advantage of improving access to a diversity of job opportunities that match their educational level. For the specific L–H areas, these problem/disadvantaged areas with a low diversity of opportunity and large populations, including Qinghe, Shahe, and Xisanqi, were distant from the urban center. There was a small gap between car travel and public transport in the number of L–H areas. The number of disadvantaged areas was less than in the situation where diversity of opportunity was not considered, and diversity of opportunity tended to provide a more equitable distribution of accessibility.

Table 6 The spatial correlation between diversity of job opportunity accessibility and different population groups.

Fig. 15

Local spatial clustering maps of diversity of LA and low-education groups (15-min).

Fig. 16

Local spatial clustering maps of diversity of HA and high-education groups (15-min).

Fig. 17

Local spatial clustering maps of diversity of LA and low-education groups (60-min).

Fig. 18

Local spatial clustering maps of diversity of HA and high-education groups (60-min).

Table 7 The number of areas of “L–H” mode between specific groups and diversity.

Distribution of transport disadvantage areas

A further analysis of the equity revealed that by focusing on the number of different groups located in L–H areas, i.e., disadvantaged areas with high population sizes and low accessibility levels, the social groups that faced inequities with different transport modes could be identified. According to Table 8, at the 15 min travel time threshold, the proportion of the non-registered low-education group was higher than that of the other groups, and this trend was observed for both car travel and public transport. For car travel, the proportion of low-education groups located in L–H areas was higher than that for public transport, and low-education groups had higher levels of inequity than high-education groups, and the proportion of specific groups was also higher for public transport than car travel at the 60 min travel time threshold. The non-registered low-education group had the highest proportion living in L–H areas, followed by the registered low-education group and the non-registered high-education group. This indicates that the low-education group was more vulnerable to inequitable job accessibility. While public transport exacerbated this disparity more than car travel, their Hukou status could notably mitigate such inequities.

Considering the diversity of job accessibility, with public transport, there was a larger proportion of high-education groups in L–H areas than low-education groups, and more non-registered groups than registered groups faced low levels of diversity of opportunity. Similar trends were observed for car travel. Furthermore, the number of individuals from different educational backgrounds in disadvantaged areas was lower than in scenarios where diversity of opportunity was not considered.

Table 8 Differences in the different education population (%) in L–H areas.

Conclusion and discussion

Conclusion

Taking Beijing city as a study area, this study assessed transport equity by analyzing job accessibility disparities between population groups with different education levels based on car travel and public transport mode for different travel time thresholds, and considering specific population groups’ needs for certain job sectors, with a focus on the diversity of specific job opportunities obtained. The main conclusions are as follows: (1) there was a significant difference in the spatial distribution of job accessibility and diversity of job opportunity. High accessibility levels were mainly concentrated in the central area, while areas outside the central region had low accessibility, and there was a significant core-periphery pattern. Moreover, in terms of absolute number, most job opportunities were located in a few areas. When considering diversity of job opportunity, towns outside of the central area had low accessibility levels but high levels of diversity of job accessibility, and more areas could potentially share this high diversity of job opportunities, which reduced the difference of job accessibility between areas, consistent with the findings by Dai et al.³⁸, who measured the diversity of job accessibility in Beijing. (2) When analyzing the spatial relationship between job accessibility and different education groups, consistent with other research, it was observed that disadvantage areas with low transport equity were located in suburban areas(outside 3rd ring road)⁵⁰, and the areas that exhibit a low accessibility but high demand were centered in suburbans⁵⁶. With the increase of travel time, the status of transport equity improved significantly; car travel could improve the equity of job opportunities accessible among groups more effectively than public transport. This echoes the findings of Sun and Zacharia⁵⁰, who noticed that the job accessibility gap between car and public transport would be reduced with an increasing in travel time thresholds, and it tended to expand expands equitable spatial coverage in Beijing. We also found that there were fewer areas with low diversity and high demand than areas with low accessibility and high demand. Additionally, there were more disadvantaged areas due to low accessibility levels than low levels of diversity of opportunity. This indicated that improving job diversity could effectively reduce the inequity that arises from accessibility. (3) Low-education groups had a lower average accessibility level than high-education groups at the same location. These low-education groups faced unequal low-education job accessibility and were at a disadvantage compared to high-education groups, especially non-registered low-education groups. These findings were similar to Xiao et al.’s work⁴⁵, who evaluated job accessibility between different educational groups. High education/skill people hade significantly higher job accessibility, and registered population had the advantage of job opportunities over the non-registered population. This study further revealed that although low-education groups consistently remained disadvantaged in job accessibility and its diversity, diversity of opportunity could improve the equity status of low-education groups in the employment market and also benefit populations without Hukou status.

Discussion

To assess transport equity, this study took Beijing as a case to identify disparities in job accessibility between public transport and car travel across different population groups, with a focus on differences between potential demand from certain population groups and the supply of job opportunities, and providing a reference case and guidance for future transport equity policies. Suggestions are as follows: firstly, under the transit-priority strategy, megacities like Beijing had a higher level of investment in public transport than other cities, and the difference in accessibility obtained between central areas and peripheral areas is huge. Peripheral areas spatially disconnected from urban cores or employment centers necessitate enhanced new public transport infrastructure provision, such as new stations, routes, and roads to improve commuting efficiency and equity. Secondly, reducing the modal disparity in employment accessibility between public transport and car travel is critical for advancing spatial equity. Planning prioritization should target areas exhibiting adequate car access but deficient public transit accessibility. Specifically, when areas have disadvantaged status with low job opportunities via both public transport and car travel, these areas need to prioritize public transport plans to enhance accessibility. Thirdly, efforts should be made to enhance the accessibility of diverse employment opportunities. The accessibility of diverse employment opportunities exhibits a comparatively small gap among different demographic groups, indicating that promoting mixed activities helps reduce inequity⁴⁹. Therefore, spatial planning of industries should place adequate emphasis on diversity in specific locations to improve residents’ access to a variety of job opportunities. Low- and high-education groups may share the same locations, there is a need to develop multiple job sectors to meet these low- and high-education requirements in these locations and narrow the gaps between different groups and job opportunities in the territorial spatial planning and urban industrial planning. Fourthly, cities of varying sizes and locations should set appropriate travel time thresholds as planning standards to achieve transport equity. Research revealed significant disparities in accessibility between central and peripheral areas under different travel time thresholds, leading to different equity outcomes in accessibility. While Beijing has committed to planning 45-min and 60-min commuting catchment areas, it has not specified the transport modes for these targets. Under a transit-priority strategy, planning suitable public transit commuting catchment areas is particularly essential. Furthermore, the setting of commuting time thresholds varies significantly across different urban locations. Megacities require a larger commuting scope to encompass a greater population farther away from the city center, under the expansion of the urban region.

However, there are still some limitations in this article. (1) It is important to categorize job opportunities based on different job types, because different job sectors have different education requirements for groups. We used the Industrial Classification for National Economic Activities to classify the major sectors and assumed that most workers in each sector were attributed to the same education level. However, our classifications of job sectors could be further refined, and the same sector’s employment could also have both low- and high-education group requirements among its workforce. It is difficult to fully define which sector needs high or low-education workers. In the employment market, one industry sector may have diverse workforce needs; the workers could be divided into more refined categories with specific socio-economic attributes or needs to match their specific employment. There is a need to set a relative weight for different types of workers in the accessibility measure. Additionally, the informal employment sector—a significant contributor to urban economies—was excluded due to gaps in official statistical coverage. Future studies should integrate multi-source datasets (including informal employment) to address this gap. (2) Regarding the diversity of job accessibility, this study set the same weight of accessibility for each job type to calculate diversity separately, and how best to obtain a weight for each opportunity and incorporate diversity into the accessibility measure has become a critical issue³⁶, which needs to be supported by individual-level data. (3) The selection of the accessibility measure directly impacts transport equity evaluations. This study used a cumulative measurement to assess spatial accessibility without considering the complex characteristics of job seekers and job opportunities. The actual demand and job preferences of different groups should be incorporated in accessibility measures via specific transport modes, as Liu stressed that the spatial accessibility cannot capture actual access to opportunities for a specific individual’s job and transport service⁴⁶.