1. Introduction
Commercial sites and residences are the two dynamic and vital components of urban spaces. Owing to socioeconomic development, commercial sites and residences have become spatially dependent and closely connected physical spaces [
1,
2]. Traditionally, commercial sites and residences always present specific spatial associations. In early feudal society in China, the spatial association between commercial sites and residences manifested as the separation of “neighborhood” and “market” in the “
Lifang system” [
3]. After the middle of the Tang Dynasty, the spatial association between commercial sites and residences appeared as a combination of “neighborhood” and “market” in the “
Jiefang system” [
3]. In modern society, commercial sites are always concentrated around residences, and the development of commerce is dominated by the relocation of residences during suburbanization. All these phenomena reflect shifts in spatial associations between commercial sites and residences in the form of geospatial associations, such as distribution, distance, and topological associations. China’s rapid urbanization and population growth are causing polycentricity, decentrality, and even recentrality in its urban commercial and residential spaces [
4]. Specifically, its residential spaces have been restructured as a result of the marketization of land and housing as well as increasing suburbanization [
5]. Moreover, the rise of e-commerce and online purchasing has substantially altered the original urban commercial formats, which refer to the business forms or sales forms of commerce [
4], and has shifted the focus from shopping to experiential consumption. These changes have resulted in a mismatch between commercial sites and residences in terms of scale, function, format, and distribution [
4,
5,
6,
7]. As the spatial association between commercial sites and residences has a significant impact on people’s daily lives and social fairness [
8,
9,
10], investigating and assessing the spatial associations between commercial sites and residences is essential to streamlining commercial facilities’ and residences’ organization, increasing the efficiency of urban function allocation, and realizing scientific planning and management [
11,
12].
According to Tobler’s first law of geography, “all things and phenomena on the earth surface are spatially associated” [
13]. Spatial association is a fundamental characteristic of geospatial phenomena and processes [
14]. The spatial association between commercial sites and residences comprises three elements: commercial sites, residences, and the flow of consumption between them [
15]. Current research has focused on two aspects: the physical spatial association between the distributions of the two types of spaces from a static perspective [
4,
16,
17,
18] and the functional spatial association based on the consumption behavior of the two types of spaces from a dynamic perspective [
15,
19,
20,
21]. Studies based on the dynamic perspective are often based on questionnaire survey data [
22]. Owing to the number of samples in residential areas and the accuracy of data, determining the characteristics of the spatial association between commercial sites and residences in a whole city is difficult. By employing the concept of “self-containment” in the urban planning field, static research on this topic has mostly been conducted from a mixed-land-use perspective, referring to an appropriate commercial facility–housing ratio [
23] or the spatial correlation between the distributions of commercial facilities and housing [
4,
24]. These studies have mainly focused on the static spatial correlation and mismatch between commercial sites and residences by employing land-use or point-of-interest (POI) data [
4,
25]. These two types of functional spaces affect each other’s spatial distributions in different ways and to varying degrees, and their spatial association is heterogeneous across a city. Moreover, people with varying sociodemographic traits often reside in residences of different levels in a city. The spatial association between each commercial format and residences of different levels also varies [
24]. However, current analysis methods cannot capture the process and mechanism underlying the spatial association between commercial sites and residences, creating a research gap that needs to be filled.
With the advancement of big data collection and analysis technologies, POI data have become key data sources for explaining urban spatial organization. By reflecting the locations and attributes of geographic entities [
26], POI data have expanded the research possibilities for urban spatial organization [
27]. With the rapid expansion of urban commercial sites and residences, POI data with full-sample characteristics provide a new perspective for quantitative research on the spatial association between the two types of spaces. By adopting the concept of location quotient in economic geography, Leslie et al. [
28,
29] introduced a quantitative analysis method called the colocation quotient, which measures the directed spatial dependence among different elements. This method allows for the analysis of the spatial association among spatial elements from the perspective of asymmetric spatial dependence and of the spatial heterogeneity of spatial association at the local scale [
30]. The application of this method has been widely recognized in research on the spatial association of a variety of situations, such as crime, urban planning, transportation, and public health, and it has become valuable for discovering the patterns of spatial interaction, dependence, causality, or symbiosis among spatial entities [
30,
31,
32,
33,
34,
35,
36].
This study employs POI data of commercial sites and residences in Beijing and the colocation quotient to analyze the characteristics and patterns of the spatial association between commercial sites and residences. First, the global colocation quotient (GCLQ) is applied to explore the spatial association between different commercial formats and different levels of residences at the city scale. Second, the local colocation quotient (LCLQ) is applied to explore the local characteristics of the spatial associations between commercial sites and residences. The research aims to answer the following questions: (1) what are the spatial association patterns between commercial sites and residences across an urban space, (2) how do the spatial association patterns of each commercial format and different levels of residences vary, which commercial format in an area contributes to the development of residence, and which commercial format in an area is attracted by the development of residence? The findings of this research can help reveal the spatial development mechanisms of urban commercial sites and residences and create a commerce–residence balanced region, providing insights for urban planners to improve commercial accessibility, reduce consumption inequality, and achieve self-contained goals in the context of urban spatial restructuring.
4. Global Characteristics of the Spatial Associations between Commercial Sites and Residences
GCLQ analysis of residences and various commercial formats (
Table 3) shows that
GCLQR→Ci and
GCLQCi→R are less than 1, indicating that the spatial association between commercial sites and residences in Beijing is weak and presents the characteristic of relatively independent distribution. Considerable differences were observed in the
GCLQR→Ci of the residences attracted by each commercial format; the following are the values in descending order: catering service > life service > specialty store > recreation and leisure service > convenience store > farmers’ market > supermarket > shopping mall > home building material market > home appliance and electronics store. Catering service and life service are the commercial formats with high frequency consumption and high demand. The highest
GCLQR→Ci with values close to 1 are recorded for the residences attracted by these two commercial formats, indicating that residences are preferentially distributed near catering and life services. Recreation and leisure service covers a wide range of segmented commercial formats, and convenience stores, farmers’ markets, and supermarkets target daily family consumption. These four commercial formats are widely distributed, and the
GCLQR→Ci values of the residences attracted by them are relatively high. The shopping mall, home appliance and electronics store, and home building material market are the non-daily consumption commercial formats characterized by a large market, small number, and fragmented distribution. Thus, the
GCLQR→Ci values of the residences attracted by them are relatively low.
The
GCLQCi→R of each commercial format attracted by residences differs from the
GCLQR→Ci, and the values in descending order are as follows: life service > catering service > recreation and leisure service > convenience store > supermarket > farmers’ market > specialty store > shopping mall > home building material market > home appliance and electronics store (
Table 3). The dense population distribution of the residential area has a strong attraction to the site selection of recreation and leisure service. Thus, the
GCLQCi→R value of recreation and leisure service attracted by residences is higher than the
GCLQR→Ci value. The
GCLQCi→R value of convenience store and supermarket attracted by residences is also higher than the
GCLQR→Ci value, showing that both have a strong attachment to residences as daily consumption commercial formats. The
GCLQCi→R value of specialty stores attracted by residences is lower than the
GCLQR→Ci value; the main reason is that some specialty stores rely on the distribution of large commercial centers and have a low degree of attachment to residential areas. Large-scale commercial formats, such as shopping malls, home appliance and electronics stores, and home building material markets occupy a relatively large area, and their location selection mostly considers factors, such as agglomeration and land price. Thus, these commercial formats are less attracted by residences, leading to their relatively low
GCLQCi→R. The life service, catering service, recreation and leisure service, and convenience store are commercial formats with high daily consumption frequencies and small footprints. The
GCLQCi→R values of these four commercial formats attracted by residences are close to 1, indicating that their location tends to be dependent on residence distribution.
GCLQ analysis of high-, medium-, and low-grade residences with each commercial format (
Table 3) shows that the spatial associations between low-grade residences and commercial formats are relatively strong, whereas those between high-grade residences and commercial formats are relatively weak. High-grade residences are mainly clustered in the core and central areas with a relatively high land price (
Figure 2). In recent years, the number of commercial formats in the core area has been reduced because of high operating costs and the policy of optimizing Beijing’s urban functions [
44,
49]. As a consequence, the spatial association between high-grade residences and various commercial formats is low. Low-grade residences present a spatial distribution pattern of clustering from the center area to the surrounding areas along major traffic arteries (
Figure 2). The areas where they are located have relatively low land prices. Thus, the colocation characteristics of low-grade residences with commercial sectors are highly evident. Moreover, people living in low-grade residences mostly have a relatively low consumption level. Convenience is an important factor in choice of residence. The distribution of commercial formats with high frequency consumption has a stronger attraction to the distribution of low-grade residence than medium- and high-grade residences.
6. Conclusions and Discussion
The colocation quotient is used to measure the spatial association characteristic and pattern of commercial sites and residences in Beijing on the basis of the POI data of commercial sites and residences. The results demonstrated wide variations in the spatial associations of each commercial format and residences, exhibiting the characteristic of integrated high-frequency consumption and segregated low-frequency consumption. In addition, the degrees of various commercial formats attracted by residences are significantly higher than those of residences attracted by various commercial formats, reflecting the differentiated distribution mechanism of various commercial formats and residences. In terms of residences, high-grade residences and commercial formats have relatively weak spatial associations, whereas low-grade residences and commercial formats have relatively strong ones. These results are different from those obtained in North America, where the spatial associations of commercial formats with high-grade residences are strong and those in low-grade residences are relatively weak [
41]. This discrepancy may due to the fact that the high-grade residences in Beijing are mainly clustered in the core and central areas with relatively high land prices. The number of commercial formats in the core area has decreased recently because of high operating costs and the policy of optimizing Beijing’s urban functions [
44,
49]. As a consequence, the spatial association between high-grade residences and various commercial formats is low.
LCLQ can effectively reflect the spatial heterogeneity of the spatial associations of various commercial formats and residences. The asymmetric spatial dependence of commercial sites and residences was found. The proportion of various commercial formats attracted by residences is significantly higher than that of residences attracted by various commercial formats. Within the Fourth Ring Road, the commercial formats are mainly attracted by residences, showing a spatial association pattern of “distribute commercial sites according to the location of residences”. However, the proportions of residences attracted by commercial formats increase outside the Fourth Ring Road, presenting a spatial association pattern of “commercial formats attracting residences”. The findings help reveal the spatial development of commercial sites and residences in Beijing. The spatial association between commercial sites and residences considerably varies between the north and south regions bound by Chang’an Street. The spatial association between commercial sites and residences is strong in the northern part of the city where the commercial facilities in the residential areas are relatively well developed. This finding partially explains residential preferences in the northern part of the city in Beijing. On the basis of the spatial association pattern of commercial sites and residences, strengthening the allocation of commercial formats of daily consumption and physical experience in residential areas can promote the balanced development of commerce and residence.
As a method to measure the spatial association between point elements, a colocation quotient is adopted to explore the spatial association between commercial sites and residences from the perspective of asymmetric spatial dependence and to investigate the spatial heterogeneity of the spatial association across the city. On the basis of the LCLQ analysis results, this study draws a spatial association pattern of commercial sites and residences, that is, “distribute commercial sites according to the location of residences” and “commercial formats attracting residences” from the city center to the outer parts of Beijing. As the political and cultural center of China, Beijing is experiencing a unique form of suburbanization of commercial and residential spaces because of population agglomeration and the urban function optimization policy. Whether the spatial association pattern of commercial sites and residence from the city center outward is applicable to other cities remains to be further verified.
Metropolitan areas in China are confronted with issues arising from the spatial imbalance between commerce and residence as urban space is restructured and residential segregation intensifies [
15]. The distinctiveness of Beijing, as the capital of China, can cause the findings to be different from those obtained from other cities. This phenomenon is not the same as the inadequate commercial facilities of low-income neighborhoods in the city core of a North American metropolis, where decline in commercial services and the decentralization of mid- and high-income residents and commercial functions turn the city core into a habitat for low-income residents [
39]. In Chinese cities, the city core is more developed than the suburban areas. The disparities between Beijing and North American metropolises are mostly due to differences in urban spatial restructuring processes. Therefore, the spatial association between commercial sites and residences in cities of different scales and urban structures should be further studied.