Influence of the Built Environment on Elder Social Capital and Its Structure: An Empirical Study Based on Three Characteristic Communities in High-Density Cities of China

Abstract

In this study, we utilized survey data from three Shenzhen communities to investigate how the built community environment influences elderly physical activity and social capital in China’s high-density urban settings. Based on this, we analyzed how the spatial characteristics of the built environment affect the formation and structure of social capital. Structural equation modeling (SEM) was employed to assess the influencing factors and pathways of the built environment on social capital, with physical activity being a mediating factor. The results show that the built environment significantly affects physical activity, which in turn promotes social capital. Key attributes such as the green space ratio, building density, land-use mixture, and street density positively influence both physical activity and social capital. Additionally, the distribution of the surrounding environment and activity space within the community will influence the structural features of social capital by affecting the structure of social networks. Consequently, communities with different spatial layout features will develop social capital with distinct structures. This study also highlights the importance of community design in fostering social interactions and trust among the elderly. These findings offer valuable guidance for urban design and policy planning to support active aging and social inclusion in rapidly urbanizing societies.

1. Introduction

Population aging is a global challenge, particularly in China, where its speed and scale are particularly significant [1,2]. With declining fertility rates and extended life expectancy, the proportion of the elderly population is rapidly increasing, presenting significant economic and public health challenges. Older adults, especially retirees, often experience social isolation and mental health issues due to reduced social connections, which poses new challenges for the construction and sustainable development of community-based elderly care environments [3]. The World Health Organization (WHO) has proposed the “Aging-Friendly Cities” (AFC) initiative, which emphasizes the importance of social participation and communication rights for older adults in promoting sustainable community development [4]. Additionally, various policies highlight the significance of fostering social capital (SC) among older adults, particularly at the community level, as this is crucial for their health and well-being, especially given that they may spend a substantial amount of time in community settings [5].

SC, as an indicator of relationships, provides a comprehensive measure of the extent to which elderly individuals maintain connections with their communities. SC not only includes social relationships between individuals, but also informal norms such as trust, mutual assistance, and cooperation within the community [6]. Research has shown that elderly people with higher levels of SC tend to engage in more social and physical activities (PAs) and achieve better physical and mental health, a higher quality of life, and greater happiness [7,8]. The built environment (BE) influences the elderly’s PA not only through direct physical factors such as accessibility, safety, and convenience but also through indirect mechanisms, including social interactions and psychological factors [9]. For example, a well-designed BE, such as safe and accessible walking paths and a variety of public facilities, can promote outdoor activities among the elderly, thereby enhancing their social interactions and community participation, which in turn enhances their SC, including trust, belonging, and social networks [10]. Additionally, research has shown that the influence of the BE on the elderly’s PA is multifaceted, encompassing both direct physical environmental factors and indirect mediating effects of subjective perceptions and psychological factors [11]. For instance, a favorable BE may enhance the elderly’s confidence and motivation to participant in PA, indirectly promoting PA and social participation, thereby enhancing SC. Therefore, the BE serves not only as a support system for the elderly’s PA but also as an important factor in the formation of their SC.

However, despite the widespread recognition of the role of SC in the health and well-being of the elderly, the connection between SC and the BE, particularly in high-density urban settings in China, remains underexplored. The core characteristics of high-density cities include a high population density, a high building density, and intensive land use [12]. With the acceleration of urbanization in China, cities and regions with highly concentrated populations and buildings are gradually becoming more layered, and many elderly people are migrating from suburban and low-density areas to high-density areas with better development and higher urbanization, resulting in a particularly prominent aging situation in high-density cities [13]. As the most typical high-density cities in China, Shanghai, Beijing, and Guangzhou have aging rates of 36.8%, 21.6%, and 18.8%, respectively, all reflecting a transition into a moderately aging society [2]. Shenzhen is the “youngest high-density city in China”, and although only 14.1% of its inhabitants with a registered residence are over 60 years old, it has a large number of migrant elder adults; the permanent elderly population in Shenzhen is projected to exceed 1.5 million, thus entering an aging society [14]. Due to reasons such as immigration and retirement, the fragmented SC in high-density cities is further hindered by congested motor traffic, fragmented public spaces, and high-density building environments, which further impede interactions between people, community integration, and sustainable development, accelerating social isolation, the decline in SC, and the occurrence of mental health problems among the elderly [7,13]. Therefore, studying the interrelationships among the BE, SC, and PA, as well as their interaction mechanisms in high-density cities, is of great significance for understanding the promotion of health and well-being among the elderly.

While there is some research on the influence of the BE and PA on SC in general, relatively few studies have examined the complex interactions and impact pathways between community design and elderly PA and SC, particularly in the context of high-density Chinese cities. This gap is particularly important for urban planners and policymakers designing inclusive and sustainable communities. In this study, Shenzhen, a coastal city in southeastern China, was selected as a representative high-density urban sample, specifically exploring the following three issues: (1) how does the BE of a community influence the SC of elderly residents? (2) Does participation in outdoor PAs mediate this relationship? (3) What are the differential impacts of various types of community BEs on the structural characteristics of elderly SC? By exploring these relationships, this study aims to provide empirical evidence to support the development of BEs that promote both PA and social well-being among the elderly.

2. Literature Review

2.1. The Concepts of Social Capital, the Built Environment, and Physical Activity

Over the past 10–15 years, SC has emerged as a significant area of research and policy interest, yet a universally accepted definition remains elusive [15]. The earliest conceptualization of SC viewed it as the accumulation of resources by individuals or groups through sustained social networks [16], while other scholars characterized it as a productive social network [17]. Subsequently, SC has been defined by social attributes like trust, reciprocity norms, and group memberships that encourage collective action [18]. In the past decade of research, SC has been conceptualized as a characteristic of social organizations, such as trust between citizens, norms, and interactions of group members that promote collective action [19,20]. Research on measuring SC has also gradually matured, initially measuring the amount of SC held through three dimensions: participation, trust, and social networks. The five terms commonly used in existing studies to describe the components of SC are social network, social trust, interaction and participation, social cohesion, and neighborhood attachment [21,22].

The BE is usually defined as the environment created, designed, constructed, or maintained by humans, aimed at meeting human needs, activities, and values. It not only includes buildings, parks, cities, and infrastructure, but also covers the range from personal space to the entire urban area [23]. Specifically, a BE is a “place” for human activities, including living, working, and leisure spaces [24]. This concept has been applied in multiple fields such as architecture, sociology, and environmental science. Relevant research not only focuses on the physical space but also on its impact on population behavior and activities.

PA is typically characterized as a form of energy expenditure resulting from muscle contraction [25]. This definition emphasizes that energy expenditure is the core criterion for judging PA. In later research, PA can be divided into various types, including occupational activities, sports, household activities, and leisure activities. However, for the elderly, common PAs mostly refer to leisure activities, such as physical exercise, entertainment, and social activities [26]. The health benefits of PA are well established, especially for older adults. Engaging in regular PA supports physical and mental well-being, reduces the risk of cognitive decline, and decreases the likelihood of falls [27]. However, people who lack PA have a much higher probability of developing non-communicable diseases such as cardiovascular disease, depression, and dementia than those who exercise [28]. In addition, studies have shown that older individuals with regular PA habits can obtain more SC from family or friends [29], and their social isolation and loneliness are lower compared to the inactive group [30]. However, older adults without regular PA may experience a greater likelihood of loneliness and reduced SC [29,31].

2.2. The Measurement of Social Capital, the Built Environment, and Physical Activity

The SC in this study draws on the definitions provided by Fukuyama, stating that elderly SC refers to informal networks that promote connections between older adults and between older adults and other groups, as well as the norms and trust that these networks bring. SC is also measured using the above five dimensions [32].

Social networks represent the connections among participants within a group. Each participant acts as a node in the network, establishing connections through information exchange and interactions, thereby forming the social network [33,34]. In the context of SC, the structure between nodes is crucial, as it not only illustrates the relationship patterns among members but also highlights the functional utility of social relationships [35,36,37,38]. Consequently, existing research often employs concepts such as the number of acquaintances within a specific area, the contact frequency, and the depth of interactions to characterize the structure and scale of social networks [39,40].

Social trust refers to the expectations and beliefs individuals hold regarding their social partners in interpersonal communication [41]. Interpersonal trust within a specific region can influence the formation of SC, while broader societal trust can affect behavioral norms such as reciprocity, thereby impacting the generation of SC [17,42,43].

Interaction and participation are prerequisites for forming SC, representing the values or norms from which SC can develop and be used to promote further activities and itineraries of group members [22]. Formal or informal interactions and information exchanges, such as greetings and community assistance, fall under the category of interaction and participation [44]. At the same time, participating in an organization is another indicator of measuring SC. SC does not emerge spontaneously; its creation relies on social networks.

Indicators such as neighborhood relationships, community awareness, and cohesion—reflecting a sense of belonging—are commonly used to characterize SC [45]. For instance, an individual’s significance within a group can indicate their sense of belonging, and their pride in the group often characterizes its cohesion, which is closely linked to SC. Although there is some overlap in the indicators used to measure SC across existing studies due to variations in research scope and focus (

Table 1. A summary of literature reviews on measures of SC.

Dimensions of SC	Indicators	Measure	References
Social networks	Social network size	Count	[35]
	Types of social ties	Count per type of social tie	[46]
	Frequency of meeting	Categorical variable	[47]
Social participation	Memberships in community organizations	Binary variable	[46,47,48,49,50,51]
	Frequency	Categorical variable	[47]
	The sense of achievement from the organization	Likert scale	[46]
Social interaction and reciprocity	Frequency	Categorical variable	[52]
	Visit at home	Binary variable	[52]
	Easily getting help	Likert scale	[48]
	Receive/provide emotional support (any one or more)	Binary variable	[5]
Trust	Level of trust	Likert scale	[46,53,54]
	The general trust in society	Likert scale	[52]
	Trust in neighbors	Likert scale	[52,55]
	Community environment trust and sense of security	Likert scale	[52,56]
Cohesion	Neighborhood harmony level	Likert scale	[57]
	Unique identity of the neighborhood	Likert scale	[39,45]
	Mingling between neighborhood groups	Categorical variable	[45]
	Tolerance of diversity	Likert scale	[55,58]
Attachment and sense of belonging	Sense of belonging to neighborhood	Likert scale	[46,59]
	Community pride	Likert scale	[48,51,52]

), they predominantly concentrate on aspects such as social networks, trust, interaction and reciprocity, and community cohesion.

In studies examining the influence of the BE on human behavior, the indicators used to measure the BE are often categorized under three well-established principles, also known as “3Ds”: density, design, and diversity [60]. Density refers to the density of the population, buildings, or facilities within a certain spatial range. It reflects the compactness of space utilization and the concentration of resources. In studies examining the influence of the BE on PA and SC, building density and the density of facilities such as churches are often closely related to outdoor activity behavior [61,62], while residential and population density are significantly correlated with social interaction behavior [63,64,65,66].

The design aspect of the BE focuses on the spatial layout and physical features of the BE, including street networks, building layouts, spatial organization, and functional zoning. It emphasizes the connectivity, accessibility, and functionality of space [67]. Existing studies use indicators such as the street density, intersection density, and point line ratio to measure the rationality of street design. There are also studies that use the distance from potential destinations to evaluate street accessibility [68]. Relevant research has confirmed that the street density and distance to surrounding blue/green resources have a significant correlation with residents’ walking behavior and SC levels [67,69].

Diversity is characterized by the extent of mixing between different functions or land-use types within a given area. It reflects the diversity of space, the richness of functions, and the diversity of socio-economic activities [70]. The land-use mixture has been found to significantly influence PA, community security, and in turn, the SC of community residents [55,71]. At the same time, parks and green spaces can improve residents’ physical fitness and enhance SC [72,73,74,75]. The integration of public service facilities, such as libraries and sports stadiums, in residential areas can also enhance the SC of the community by promoting interactions among residents [76,77]. The measurement methods for the relevant indicators are shown in

Table 2. Measurement and indicator calculation method for BE.

3Ds	BE Indicators	Measurement	Reference
Density	Building density	The number of buildings or building area per unit area.	[78]
	Residential density	The number of residential units per unit area.	[79,80,81]
	Population density	The population per unit area.	[44,60,82]
Design	Street density	Street density is measured as the total length of linear kilometers of streets per one square kilometer of land.	[81,83]
	Intersection density	Street intersection density is measured as the number of intersections per one square kilometer of land.	[50,84,85]
	Accessibility Index/Pedestrian Route Directness Index	An Accessibility Index is calculated as actual travel distances divided by direct travel distances. It is also called the Pedestrian Route Directness Index (PRD). An index of 1.0 is the best possible rating, indicating that pedestrians can walk directly to a destination. An average value of 1.5 is considered acceptable.	[86]
	Link-to-node ratio	The link-to-node ratio is equal to the number of links divided by the number of nodes. Links are defined as street or pathway segments between two nodes. A higher link node ratio implies higher street connectivity (Actual Walking Distance/Direct Distance).	[87]
Diversity	Land-use mixture	The degree of mixing of different functional land uses (residential, commercial, industrial, green spaces, etc.). Land-use mixture degree = −1 ∗ ∑ (Pi ∗ ln (Pi)), where Pi is the proportion of different functional land uses.	[47,51,88,89]
	Distance to public facilities	The distance that residents can travel from a certain location to the target facility through a certain mode of transportation.	[90,91,92]
	Green space ratio	Proportion of green space area per unit area Green space ratio = Green area/Total area.	[93,94,95]

.

Current research often uses self-reported and objective observation methods to evaluate PA [96,97]. Some scholars use metabolic equivalents (METs) based on energy expenditure during quiet sitting (1 MET), while the intensity of other activities is expressed as multiples to evaluate the intensity of PA [98,99]. Some scholars use the heart rate reserve (HRR) to evaluate the intensity of PA, such as the maximum heart rate (HRmax) and HRR used in sports medicine-related studies in the United States to evaluate the intensity and endurance of aerobic exercise [100]. For research on the elderly, self-reported methods mainly rely on questionnaire surveys, which often include the PA content, frequency, duration, and diversity of exercise types. The International Physical Activity Questionnaire (IPAQ) has become one of the most popular assessment methods in researching PA intensity [101]. When assessing the PA intensity among older adults, both the frequency and duration of different activity patterns are taken into account [102]. The relevant measurement methods for PA are shown in

Table 3. Indicators and measurement for PA.

	Indicators	Measurement	Reference
PA Frequency	Frequency	Categorical variable	[96,103]
PA Duration	Duration	Minutes	[104]
Intensity	Metabolic Equivalent of Task (MET)	Based on the energy consumption during quiet sitting (1 MET); the intensity of other activities is expressed in multiples	[98,99]
	Heart rate reserve (HRR)	The difference between the maximum heart rate (HRmax) and resting heart rate (HRrest)	[100]
	PA level	Categorical variable	[105]
Diversity	Diversity of PA types	Entropy value of PA types	[106]

.

2.3. The Relationship Among Social Capital, the Built Environment, and Physical Activity

Existing research has confirmed the important role of PA in multiple dimensions, including social participation, social trust, and community belonging. PA, especially collective activities such as group sports, community activities, and physical exercise, can promote interactions and communication among the elderly. Research has shown that sports activities not only enhance the individual SC of the elderly but also strengthen collective SC through collective participation (such as community engagement, social trust, and collective cohesion). For example, one study found that participating in outdoor activities significantly boosts the collective SC of older adults, indicating that sports activities enhance their SC by promoting collective interaction and cooperation among participants [107]. Meanwhile, PA can enhance social trust and support among the elderly, helping to enhance their SC by providing social opportunities and emotional support [108,109,110]. PA not only promotes individual participation among the elderly but also strengthens their sense of community belonging [111]. PA will also directly affect SC, with cross-sectional survey data from Japan showing that a decrease in PA among Japanese adults is associated with a decrease in SC [112].

Existing research has indicated that the BE influences SC through both direct and indirect pathways [113,114,115,116]. The dominance of motor vehicle transportation, driven by new urbanization, has diminished opportunities for social interactions among individuals due to inconveniently located destinations and congested street spaces [113]. This impact is particularly severe for elderly individuals who spend a significant portion of their time in the community settings, directly leading to increased social isolation and a reduction in SC [117,118]. At the level of BE, improving the street network and community pedestrian system in high-density areas, freeing up space for more social activities, or increasing the accessibility of entertainment venues can all increase residents’ social opportunities, thereby enhancing SC [119,120]. At the same time, it has been found that the SC in the urban center of the BE is also affected by the building density, population density, and street density [62,63,64,65,66], but the conclusions of the impact are not consistent, which may be related to the spatial layout form in the community. Research on Orenco Station in Portland, Oregon, has shown that residents living in communities with traditional spatial layouts have higher levels of SC, which may be related to the complex street grids of suburban sprawl and new urbanization areas, which can affect residents’ mobility [71,121]. Similar findings have also been reported in other studies, indicating that suburban sprawl inhibits the formation of community SC, while an increase in lonely commuting is associated with a decrease in citizen participation [122,123]. Conversely, there are also studies indicating that SC was higher in New Urbanist neighborhoods [45,122].

The BE indirectly affects SC by influencing residents’ perception and PA. For example, studies have found that the density of street networks and land-use mixture significantly affected the perception of community safety, which in turn affects the SC of community residents [55,71]. Research has also examined the relationship between residents’ SC and the availability of neighborhood sports venues, parks, and green spaces. Findings indicated that open areas, such as parks and green areas within the community, can improve residents’ physical fitness while enhancing SC [72,73,74,75]. Beyond external environmental features such as parks and roads, other venues for collective activities, such as libraries, schools, and community centers, can also bolster a community’s SC by fostering increased interaction among residents. The presence of liquor stores and signs of neighborhood disorganization may negatively affect residents’ sense of belonging to the community, thereby adversely affecting the formation of SC [76,77]. SC is linked to PA, mental health, and life satisfaction outcomes in older adults. Research suggests that higher levels of SC are associated with increased PA, fewer mental health issues, and reduced psychological distress, and more PA is often closely related to the BE [124]. In Asian studies, significant correlations have also been found among park accessibility, PA frequency, and SC [125]. Similarly, several studies have employed content analysis of focus group discussions in Singapore to examine how the community environment influences elderly PA and social interactions, showing that the SC of the elderly is influenced by outdoor activity facilities, travel frequency, and patterns [126]. According to the existing research mentioned above, the pathways through which the BE and PA impact SC are summarized in

Table 4. A summary of the literature on the indicators of the BE and SC.

	Indicators	Relationships Explored	Methods	Findings	Reference
Density	Is the community location in the center of town?	BE → SC	Generalized linear model	Negative correlation except for churches	[61]
	Residential density	BE → SC	Block regression	Negative correlation	[127]
Design	Neighborhood walkability	BE → SC	Logistic regression	Positive correlation	[69,128]
		BE → SC	Multilevel	Positive correlation	[129]
	Retail access only	BE → SC	regression	Negative correlation	[130]
	Distance to nearest shops	BE → SC	Multilevel	Positive correlation	[128]
	New Urbanist neighborhood	BE → SC	Regression	Significant positive correlation	[131]
	Distance to nearest bus stop	BE → SC	Generalized linear model	Negative correlation	[128]
	New Urbanist neighborhood	BE → SC	Linear regression	Negative correlation	[131]
	Distance to park	BE → SC	SEM, multilevel regression	Positive correlation	[72,73,75,116]
	Library use	BE → SC	Multilevel regression	Positive correlation	[132]
	River length in neighborhood	BE → SC	Correlation coefficients	Positive correlation	[133]
PA	Collective activity	PA → social interaction → SC	Linear regression	Positive correlation	[107]
	Individual activity	PA → social trust and support → SC	Linear regression	Positive correlation	[108,109,110]
		PA → sense of community belonging → SC	Multilevel regression	Positive correlation	[111]
	PA duration	PA → social interaction → SC	Linear regression	Positive correlation	[104]
	PA frequency	PA → social interaction → SC	SEM	Positive correlation	[103,134]
Density	Residential density	BE → PA → attachment	Content analysis of focus group discussions	Residential density is negatively associated with place attachment	[62,135]
	Population density	BE → outdoor interactions → SC	Hierarchical linear regression model	Fewer interactions	[136]
		BE → social interaction and social networks → SC	Content analysis of focus group discussions	More social networks and interactions in the downtown area	[62,135]
		BE → PA → social cohesion	OLS Regression and logistic regression	Significant negative correlation	[64]
Design	Street connectivity and walking/cycling infrastructure	BE → PA → social cohesion	SEM	Positive correlation	[66]
		BE → PA diversity → SC	Generalized linear model, multilevel regression	Positive correlation	[106,137]
	Street connectivity in a 600 m buffer	BE → PA → social cohesion	Multilevel	Positive correlation	[138]
Diversity	Land-use mix	BE → PA → social cohesion	SEM, generalized linear model	Negative correlation	[55,66]
	Distance to park	BE → PA → social cohesion	Multilevel regression	Positive correlation	[133]
	Park access only	BE → PA → social cohesion	Multilevel regression	Negative correlation	[130]

.

Research increasingly emphasizes the influence of the BE on social outcomes, prompting urban planners to pursue designs that yield favorable social results. For the study of the relationship between the three factors, existing research has confirmed that both the BE and PA have a significant impact on SC. However, the impact pathway of the BE on SC is not yet clear, and the mediating role of PA in it also lacks in-depth exploration.

In terms of research methods, existing studies frequently employ a combination of qualitative and quantitative approaches, such as Social Network Analysis and Geographic Information System (GIS) technology, to investigate the relationship between the BE and PA. In quantitative analysis, regression analysis is commonly used to study influencing factors, while methods such as structural equation modeling (SEM) are used to investigate the relationship among the BE, PA, and social outcome, to explore how environmental factors affect SC, health, and other well-being through mediating variables such as PA and social participation.

Despite this growing emphasis, few studies have examined the interactions among the BE, PA, and the SC of the elderly, a topic particularly relevant to the concept of “aging at place” from the perspective of urban planners. Compared to other age groups, older adults are more susceptible to local conditions due to their limited mobility, which increases the time they spend at home. They depend more on community-based social connections and resources to sustain their health and stay in the community [139]. While neighborhood characteristics have been found to influence individual health status, there is a limited understanding of how neighborhood environments contribute to the SC of older adults [140,141]. Additionally, the structure of SC deserves further exploration. The spatial patterns of communities significantly affect the structure of residents’ SC, particularly regarding neighborhood trust and network structure, which are crucial for senior capital [121,142].

3. Data and Methods

3.1. Measurements

SC: In this study, we use the SC scale to collect and investigate the SC retention of the elderly in the community. The indicators in the scale are primarily based on the earlier-reviewed literature on SC in China [143,144]. The scale includes 17 indicators selected from five dimensions to assess the SC of the elderly in the community. These dimensions are social network, participation, social interaction, support and trust, and belonging and cohesion. The specific indicators are detailed in

Table 5. The indicators used to measure individual SC and PA.

Latent Variable	Indicators
SC
Social network	N1 How many elderly people do you know in your daily activities? a
	N2 How many young people and children do you know in your daily activities? a
	N3 How many neighborhoods can you visit? a
Participation	P1 Are you a member of some organization in your community? c
	P2 Are your family members involved in any organizations of your community? c
	P3 I’m an important part of the neighborhood. d
Social interaction	S1 How many times have you contacted your neighbors by telephone or online in the past week? b
	S2 How many times have you visited your neighbors in the past week? b
	S3 Can you borrow daily necessities from your neighbors? c
	S4 Have you asked your neighbors for help in the last months? c
Support and trust	T1 Most people in the community are willing to help each other. d
	T2 Do you trust most people in your community? d
	T3 Do you trust most people in society? d
	T4 You trust the neighborhood committee very much. d
Belonging and cohesion	B1 I like my community. d
	B2 I’m proud I live in this community. d
	B3 Neighborhood harmony in my community. d
PA
Frequency of PA	PAF1 Typical frequency of personal activities (walking, running, shopping). e
	PAF2 Typical frequency of partnering activities (table tennis, chess, and cards). e
	PAF3 Typical frequency of collective activities (square dancing, collective gymnastics). e
	PAF4 Typical frequency of intergenerational activities (childcare). e
Mode of PA	PAM1 I engaged in personal activities in the last week. c
	PAM2 I engaged in partnering activities in the last week. c
	PAM3 I engaged in collective activities in the last week. c
	PAM4 I engaged in intergenerational activities in the last week. c

^a Five categories: “0–5 persons”, “5–10 persons”, “10–15 persons”, “15–20 persons”, and “more than 20 persons”. ^b Five categories: “0 times”, “1–2 times”, “3–5 times”, “5–7 times”, and “more than 7 times”. ^c The mute variable is used to represent this; “Yes” is represented by “1” and “No” is represented by “0”. ^d Indicators rated on 5-point Likert scale—from completely disagree to completely agree. ^e Five categories: everyday, a few times per week, once or twice per month, once in a while, never.

, which also includes the measurement methods for each variable.

Elderly PA: In this study, we focus on examining two aspects of PA: its pattern and frequency. These aspects were selected because they are more likely to influence the relationship among PA, the BE, and SC, as supported by previous research [103,107,108,109,110]. The pattern of PA is assessed by asking respondents about different forms of PAs, including personal activities (e.g., walking, running, and shopping), partnering activities (e.g., table tennis, chess, and cards), collective activities (square dancing, collective gymnastics, etc.), and intergenerational activities (e.g., childcare). The frequency of PA is measured by asking respondents about the typical frequency of extensive PAs, such as how often they participate in such activities, as detailed in Table 5.

Participants were recruited randomly using a quota sampling method from the three communities. To qualify for participation, individuals had to be 60 years of age or older, reside in the selected communities, be able to communicate in Mandarin, and provide informed consent. The community committee and property management were involved in the recruitment process, and volunteers were assessed to ensure that they met the criteria for participation. Well-trained investigators conducted one-on-one questionnaire surveys and semi-structured interviews, each lasting approximately 30 min. A small gift was given to respondents who completed the questionnaire and interview as a token of appreciation. A structured questionnaire was carefully developed and pilot-tested before the main data collection to ensure both validity and reliability. The questionnaire was designed to collect quantitative data on individual SC and other relevant factors, including PA frequency and mode, demographics, and family structure information. The pilot testing involved a small sample of respondents to identify and correct any ambiguities or inconsistencies in the instrument.

This study included several objective BE attributes, such as residential density, availability of destinations, land-use mixture, ratio of greening, and street connectivity. These attributes have been repeatedly demonstrated to influence the social interactions and outdoor activities of the elderly, which are closely related to SC indicators [97,98,99,100]. Objective BE attributes were quantified using geographic information systems (GIS) 10.6 software (Figure 1). An 800 m circular buffer was applied to participants’ geocoded residential addresses to assess neighborhood BE attributes. The buffer size is consistent with a previous study [101] and refers to the previous survey on the travel distance of the elderly in Shenzhen (the travel distance of individual PAs of 63.52% and companion PAs of 76.68% is between 500 and 1000 m). To enable comparisons across variables, each objective BE attribute was standardized (i.e., z-scores).

Control variables encompassed respondents’ age, gender, educational background, and income level.

3.2. Study Site Selection

Shenzhen, situated in the Pearl River Delta region of southern China, is a core city in the Guangdong–Hong Kong–Macao Greater Bay Area and was the first city to open up during China’s reform and opening-up period. According to 2023 statistics, the city has an administrative area of 1997.47 km², with 975.5 km² designated for construction, and a service population of 20 million people. It is the smallest in terms of land area and the largest in terms of population among all Chinese metropolises.

Shenzhen was established in 1980, starting from the Shenzhen Special Economic Zone (SEZ), with an area of approximately 327.5 square kilometers, commonly referred to as the “Shenzhen internal”. The SEZ incorporates top-down urban development as part of the overall urban planning, creating well-developed areas with mixed land use and high-density street networks. At the same time, around the SEZ, the “outside of Shenzhen”, covering an area of over 1600 square kilometers, is being constructed in a disorderly manner from bottom to top, and the city is rapidly spreading along the external transportation corridor, forming a sub-developed area [137]. According to the 2020 Shenzhen statistics, the proportion of elderly residents in developed areas (Nanshan District, Luohu District, Futian District, and Yantian District) is 16.8%, 16.3%, 16.2%, and 15.6%, respectively, which is much higher than the sub-developed area aging level in Shenzhen [145], and the degree of aging has continued to rapidly increase in the past five years [14]. Therefore, the data for this study came from a community survey carried out in three districts with high levels of the aging population in well-developed areas in Shenzhen between April and November 2020. These districts were selected to represent the diversity of community types and the social characteristics of elderly residents in Shenzhen.

The three selected communities were the following:

• Lianhua North Community (LNC): A government-developed community located in the core area of Futian District, characterized by low-rise apartment complexes (primarily 5–7 stories). There are 6052 households and approximately 18,000 people in the entire community, of which 16.2% are permanent elderly residents.
• Yuanling New Community (YNC): Located in the old city district of Luohu, it is a mixed residential community with a combination of multi-story and high-rise buildings. There are 5861 households and approximately 17,000 people in the entire community, of which 16.3% are permanent elderly residents.
• Haiyin Community (HYC): Situated in the emerging development zone of Nanshan District, this community is primarily composed of closed high-rise residential quarters. There are 4900 households and approximately 14,500 people in the entire community, of which 15.8% are permanent elderly residents.

3.3. Data Analysis Methods and Instruments

In this study, we used structural equation modeling (SEM) to construct potential variables of SC and test the relationship among SC, the BE, and PA [39,66], following related studies [103,134]. SEM is a multivariate statistical technique widely employed across various disciplines, including social sciences, health sciences, and natural sciences, to examine direct and indirect causal relationships between variables. SEM is particularly useful for examining the relationships between latent constructs and observed indicators, as well as for accounting for measurement errors in the model [146].

For this study, we used the Analysis of Moment Structures (AMOS) 24.0 software package, which supports SEM model estimation, for analysis. AMOS is a statistical software used for SEM, widely applied in fields such as social sciences, education, psychology, and medicine. It helps users to construct and analyze complex relationships between variables through a graphical interface and supports path analysis, factor analysis, causal relationship modeling, and more, as well as the analysis of factors and influencing pathways in this study.

Before SEM, it is necessary to first perform confirmatory factor analysis (CFA) on the measurement model to validate its effectiveness, which is a key component of the subsequent SEM analysis. Meanwhile, when conducting parameter estimation, given that the indicators in this study are categorical in nature, traditional estimation methods such as maximum likelihood (ML) may not be applicable; thus, the Diagonally Weighted Least Squares (DWLS) estimation method was selected to ensure appropriate model fitting and parameter estimation [147]. Compared with ML, DWLS performs more robustly in processing small sample classification data and is more suitable for the data in this study. The model fit was evaluated using several standard indices, including the Chi-square test, Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), Standardized Root Mean Square Residual (SRMR), and Weighted Root Mean Square Residual (WRMR) [146,148]. These indices were chosen to provide a comprehensive assessment of the model’s fit to the data. Among them, the values of CFI, TLI, and RMSEA range from 0 to 1. The closer CFI and TLI are to 1, the better the model fit; usually, a value greater than 0.9 indicates good model fit. The closer the RMSEA is to 0, the better the model fit; an RMSEA less than 0.05 is generally considered to indicate good model fit and that less than 0.08 to indicate acceptable model fit. The smaller the SRMR, the better the fit between the model and the data; an SRMR less than 0.08 is generally considered to indicate good model fit [149]. Following the evaluation of the measurement model, two structural models were constructed to test the hypothesized relationships. The first model examined the direct effects of the BE on SC, while the second model investigated the mediating role of PA in the relationship between the BE and SC. This approach allowed for a more nuanced understanding of how the BE influences SC, both directly and indirectly through the promotion of PA.

4. Results

4.1. Descriptive Statistics

The data was collected through one-on-one interviews conducted by professionally trained researchers. Based on the optimal sample size calculated based on G * power, for this study, we recruited 200 participants from each community. After excluding invalid data, a total of 582 valid questionnaires were completed. The sample sizes for each community are 196 from LNC, 192 from YNC, and 194 from HYC, which are in line with the optimal sample size. As mentioned earlier, the sample size is proportional to the population distribution of each community, ensuring that samples from different age and gender groups within each community are representative. To identify anomalies in the data, the normality of the data is first evaluated using skewness < ±2 and kurtosis statistics < ±10 standards.

Table 6. The sample size and demographic characteristics of the respondents.

Variables	Total		LNC		YNC		HYC
Total sample size of the community	-		18,000		17,000		14,500
Optimal sample size	-		150		151		149
Sample size	582		196		192		194
Error margin	±4.5%
Categorical Variables	Count	%	Count	%	Count	%	Count	%
Age
60–69	279	47.94	96	48.98	92	47.92	103	53.09
69–79	169	29.04	57	29.08	52	27.08	50	25.77
80 and above	134	23.02	43	21.94	48	25.00	40	20.62
Gender
Male	279	47.94	96	48.98	92	47.92	91	46.91
Female	303	52.06	100	51.02	100	52.08	103	53.09
Education
Lower—Primary, secondary	174	29.90	59	30.10	61	31.77	54	27.84
Middle—Diploma	321	55.15	115	58.67	106	55.21	100	51.55
High—University graduate	87	14.95	22	11.22	25	13.02	40	20.62
Income
Below 36,000 RMB/year	128	21.99	43	21.94	46	23.96	39	20.10
3000–8000 RMB/year	326	56.01	110	56.12	106	55.21	110	56.70
Above 8000 RMB/year	128	21.99	43	21.94	40	20.83	45	23.20
Employment
Employed	430	73.88	149	76.02	138	71.88	143	73.71
Unemployed	105	18.04	24	12.24	38	19.79	43	22.16
Others	47	8.08	23	11.73	16	8.33	8	4.12
Hometown city
Local	227	39.00	88	44.90	77	40.10	62	31.96
Other city	355	61.00	108	55.10	115	59.90	132	68.04
Family structure
Living alone	35	6.01	12	6.12	17	8.85	6	3.09
Living with spouse	140	24.05	47	23.98	54	28.13	39	20.10
Living with children	128	21.99	41	20.92	38	19.79	49	25.26
Living with children and grandchildren	262	45.02	90	45.92	75	39.06	97	50.00
Others	17	2.90	6	3.06	8	4.17	3	1.55

shows the population and sample size, including the counts and corresponding percentages for each category of categorical variables. The elderly in HYC are relatively young, followed by those in LNC. This is because HYC is a new type of residential area, with more elderly immigrants bringing their children to Shenzhen, and data from “hometown cities” can also reflect this situation. Meanwhile, HYC has the highest proportion of higher education, while LNC has the highest proportion of elderly people with intermediate diplomas, which is related to its status as a community of civil servants. This impact on demographic statistics is also reflected in other variables such as income and employment.

A comparative analysis of BE characteristics within the 800 m buffer zones around the three communities is presented in

Table 7. Average comparison of BE within 800 m between three communities.

Community	Green Space Ratio (%)	Residential Building Ratio (%)	Land-Use Mixture (Average Value)	Street Density (Unit: km/km2)	Street Network Integration
LNC	26.01%	36.00%	0.031	9.48	1.28
YNC	8.00%	19.57%	0.042	8.77	1.48
HYC	4.06%	37.10%	0.046	14.6	1.35

. LNC demonstrates the highest green space ratio, which is largely attributed to its proximity to a large-scale urban park. In contrast, HYC exhibits the lowest green space ratio, as it is predominantly surrounded by commercial plazas and office buildings. Regarding the residential building density, both LNC and HYC display relatively high values, reflecting their locations in core development zones of Shenzhen. In contrast, YNC shows a lower residential building ratio, likely due to the extensive use of land for transportation infrastructure and commercial activities in its surrounding area. In terms of land-use mixture, YNC and HYC show higher levels, while LNC demonstrates a lower level of land-use mixture, primarily due to its adjacency to large urban parks and residential areas. With respect to accessibility, HYC achieves the highest street density, indicating a well-connected urban fabric that supports pedestrian mobility. Meanwhile, YNC demonstrates superior street network integration, which may enhance the accessibility of community services and public spaces. This spatial configuration suggests that YNC maintains the strongest centrality within the 800 m buffer zone among the three studied communities.

A comparative analysis of the SC ratings of the elderly in three communities reveals notable differences in the levels of social network, participation, support and trust, and belonging and cohesion (

Table 8. Statistical results of dimensions and total amounts of SC, PA mode, and frequency.

SC Dimension	SC Ownership
Social Network	LHB		YLX		HYC
	Mean	Bate	Mean	Bate	Mean	Bate
N1 How many elderly people do you know in your daily activities?	3.37	0.93	3.14	0.58	3.57	0.79
N2 How many young people and children do you know in your daily activities?	1.59	0.72	1.41	0.52	1.11	0.58
N3 How many neighborhoods can you visit?	4.01	0.10	3.98	0.09	3.34	0.15
Average value	2.99		2.84		2.67
Participation
P1 Are you a member of some organization in your community?	0.52	0.50	0.47	0.50	0.39	0.49
P2 Are your family members involved in any organizations of your community?	0.34	0.48	0.22	0.41	0.20	0.40
P3 I’m an important part of the neighborhood.	3.13	0.34	2.82	0.26	2.56	0.19
Average value	1.33		1.17		1.05
Social interaction
S1 How many times have you contacted your neighbors by telephone or online in the past week?	3.90	0.10	3.97	0.11	3.09	0.15
S2 How many times have you visited your neighbors in the past week?	2.82	0.12	3.13	0.12	2.56	0.14
S3 Can you borrow daily necessities from your neighbors?	0.77	0.04	0.55	0.05	0.20	0.05
S4 Have you asked your neighbors for help in the last months?	0.70	0.46	0.91	0.29	0.53	0.50
Average value	1.43		1.53		1.10
Support and trust
T1 Most people in the community are willing to help each other.	3.34	0.12	3.30	0.13	2.83	0.17
T2 Do you trust most people in your community?	3.97	1.11	3.90	1.13	3.89	1.25
T3 Do you trust most people in society?	3.85	1.03	3.62	1.06	2.06	1.21
T4 You trust the neighborhood committee very much.	2.49	0.11	2.72	0.12	3.36	0.13
Average value	3.44		3.41		3.10
Belonging and cohesion
B1 I like my community.	3.90	0.66	3.67	0.56	3.83	0.52
B2 I’m proud I live in this community.	4.33	0.68	4.16	0.82	4.01	0.71
B3 Neighborhood harmony in my community.	3.30	0.95	3.34	0.66	2.83	0.81
Average value	3.85		3.72		3.56
Total SC (Sum of the average values)	13.04		12.67		11.48
Frequency of PA
PAF1 Typical frequency of personal activities.	4.82	0.82	4.75	0.71	4.68	0.66
PAF2 Typical frequency of partnering activities.	4.61	0.67	4.70	0.69	3.89	0.54
PAF3 Typical frequency of collective activities.	4.32	0.61	4.01	0.57	3.00	0.32
PAF4 Typical frequency of intergenerational activities.	3.22	0.16	3.00	0.09	2.86	0.06
Mode of PA	Proportion of participants in the activity
PAM1 I engaged in personal activities in the last week.	87.23%		85.62%		87.12%
PAM2 I engaged in partnering activities in the last week.	76.88%		82.56%		59.22%
PAM3 I engaged in collective activities in the last week.	79.04%		78.21%		82.00%
PAM4 I engaged in intergenerational activities in the last week.	61.21%		58.62%		55.19%

). The elderly in LNC exhibit the highest scores in all four dimensions, followed by those in YNC, while the overall average score in HYC is the lowest. In terms of the social interaction dimension, HYC shows the highest ownership, followed by LNC, while YNC has the lowest score. This indicates that despite the lower overall SC scores in HYC, the elderly there engage in more frequent and diverse forms of social interaction compared to the other two communities. This finding highlights the complexity of SC, which is not only reflected in the breadth of social networks but also in the depth and quality of interactions.

By comparing the scores across the four dimensions, it is evident that, although HYC has the lowest overall average score, it performs significantly better than the other two communities in certain aspects, such as the rating of N1 (How many elderly people do you know in your daily activities?) in the social network dimension. This suggests that the elderly in HYC have a wider social network compared to those in LNC and YNC. This finding underscores the importance of considering both the quantity and structure of social networks when evaluating SC.

In terms of individual PAs, there is little variation among the three communities, with PA frequency scores above 4.5. Over 85% of the elderly in each community reported engaging in individual PAs within the past week, indicating a consistent level of individual PA across the communities. However, significant differences are observed in group activities. Fewer than 60% of the elderly in HYC participated in group activities in the past week, with a score of 3.89, which is significantly lower than the other two communities. YNC scored the highest in group activities, followed by LNC. In contrast, collective activities showed higher scores in both HYC and LNC, with the proportion of elderly participants in HYC being 82%, higher than in LNC.

The intergenerational engagement rate is the lowest among all activities. LNC had the most elderly participation, followed by YLX, and HYC had the least. This may be related to the fact that there are more schools around LHB.

4.2. Factor Analysis Results

The direct impact of the BE on elderly SC: The first structural model tested the direct effect of the BE on SC. The model’s fit indices indicated a good overall fit: x²(32) = 28.366; p = 0.700; RMSEA = 0.003 (90% CI: 0.003, 0.045); CFI = 1.030; TLI = 1.006; WRMR = 0.617. The results demonstrated a significant positive association between the BE and SC, with a standardized path coefficient of β = 0.311 (SD = 0.110; p < 0.05; 95% CI [0.095, 0.527]). Prior to model estimation, multicollinearity was assessed, and the correlation coefficients between all independent variables were below 0.8, ensuring model stability.

The mediating role of PA: The final structural model included both the BE and PA as predictors. The model fit was adequate, with x²(76) = 62.325; p = 0.467; RMSEA = 0.000 (90% CI: 0.000, 0.049); CFI = 1.069; TLI = 1.013; WRMR = 0.788. The results indicated that PA was a significant predictor of SC (β = 0.602; SD = 0.354; p < 0.05). Additionally, the BE significantly predicated PA (β = 0.165; SD = 0.132; p < 0.001). Notably, when controlling for PA, the direct effect of the BE on SC was no longer significant (β = 0.299; SD = 0.351; p = 0.411), suggesting that PA fully mediated the relationship between the BE and SC (β = 0.099; SD = 0.050; p < 0.05). These findings support the hypothesis that the BE influences SC through the promotion of PA. The results of the final structural model are presented in Figure 2.

Regarding covariates, the variable “hometown city” was significantly associated with SC (β = 0.203; SD = 0.070; p < 0.003). Age and family structure were also significant predictors of SC (age: β = −0.012; SD = 0.010; p < 0.05; family structure: β = 0.042; SD = 0.014; p < 0.05).

4.3. The Impact of the Built Environment and Physical Activities on Social Capital Structure

In this study, we use the evaluation value comparison method to analyze the amount of SC held by the elderly in three communities. To account for the varying measurement scales across different dimensions of SC, the evaluation value is calculated as the ratio of the measurement value to the maximum value for each dimension, multiplied by 100. This results in evaluation values ranging from 1 to 100, where a higher value indicates a greater level of SC retention among the elderly. The evaluation values are used to assess both the total SC and the performance of each dimension within the communities. A higher evaluation value indicates a stronger SC base in the community. The evaluation values are divided into five equal intervals (0–20, 20.01–40, 40.01–60, 60.01–80, and 80.01–100) to categorize the levels of SC retention among the elderly in the three communities. These categories are used to identify the relative strength of SC in each community.

Table 9. Assessment value of SC retention of elderly.

The Dimensions and Total Score of SC Composition	LHB		YLX		HYC		Total
	MV	EV	MV	EV	MV	EV	MV	EV
1. Social network	5.73	38.22	4.64	36.49	5.09	40.11	5.16	34.34
2. Participation	3.99	57.00	3.50	50.01	3.14	44.90	3.55	50.64
3. Social interaction	7.81	65.10	7.61	63.42	6.30	52.52	7.23	60.34
4. Support and trust	15.49	77.46	15.33	76.59	13.13	60.66	14.31	71.57
5. Belonging and cohesion	11.53	76.87	11.17	74.47	10.67	71.15	11.13	74.16

EV: measurement value; MV: evaluation value. 0–20: lower level; 20.01–40: lower middle level; 40.01–60: medium level; 60.01–80: upper middle level; 80.01–100: higher level.

presents the assessment values of SC retention among the elderly in the three communities. The dimensions and total scores of SC composition are listed, along with the measurement values (MVs) and evaluation values (EVs) for each community.

The distribution of evaluation values across the dimensions of SC among the elderly in the three communities shows a similar trend, with higher evaluation values in the dimensions of support and trust and belonging and cohesion and lower evaluation values in the social network dimension. This suggests that the elderly in all three communities exhibit relatively stable SC, a strong sense of belonging, and a high level of trust among those they interact with daily.

Among the three communities, LNC and YNC have lower evaluation values in the social network dimension but higher evaluation values in other dimensions compared to HYC. Although the evaluation values for support and trust and participation are relatively low in HYC, its social network evaluation value is the highest among the three communities. These differences may reflect variations in the internal structure of SC among the elderly in different community types.

5. Discussion

In this section, we discuss the structural relationships between latent variables and their implications for understanding the formation of SC among the elderly in high-density Chinese communities. Although mixed results have been reported in the literature on the relationship among the BE, PA, and SC [39,55,71], the results of this study indicate a positive correlation between the BE promoting PA and elderly SC in the context of Chinese communities. As illustrated in Figure 2, the BE significantly influences SC through its impact on PA. This is consistent with existing research findings in other regions such as Europe and the Americas, which emphasize the role of the BE in promoting PA and enhancing SC in older adults [62,66,138]. However, this study found that key BE attributes, such as the green space ratio, residential building density, land-use mixture, and street density, have a significant positive impact on the frequency and pattern of PA, especially in individual and collective activities, as well as intergenerational activities. For example, a high green space ratio can help support more outdoor interactions, thereby promoting the frequency of neighborhood interactions, which is consistent with existing research on the effectiveness of urban green spaces [150]. In addition, this study found that in high-density cities in China, a high green space ratio also has a significant effect on supporting collective activities such as square dancing and collective fitness exercises. In addition, higher street density and land-use mixture can promote intergenerational activities by improving the accessibility of surrounding children and elderly businesses. This finding confirms its promoting effect on elderly SC and compensates for the inconsistency in existing research.

This study also demonstrates the mediating role of PA in the influence of the BE on SC, which is consistent with existing research conclusions [39,107,108,109,110]. Based on this, this study reveals the impact of PA frequency and PA mode on supporting the formation of SC, where the individual and collective activity frequency can significantly promote the formation of SC, and intergenerational activity patterns and frequency can also do the same. Specifically, individual or collective activities with higher participation frequency can enhance interaction and trust between individuals, thereby promoting the accumulation of SC. For example, the higher the participation of LNC elderly in collective activities, the higher the amount of SC they possess. As intergenerational activities that can connect different groups, they play a significant role in the formation of intergenerational social networks, thus promoting the formation of SC. The existing research on the impact of group activities on the formation of SC has not yet shown a significant promoting effect. Although group activities may enhance interaction between individuals in certain contexts, their promotion of SC is not significant. This may be related to the nature and participation methods of group activities; for example, group activities may emphasize individual interactions more than collective or social interactions, thus failing to effectively promote the formation of SC.

On the other hand, this study found that the BE can influence the structure of SC, especially the structure of social networks, by shaping PA. The characteristics of the BE, such as accessibility and spatial design, can affect the number of neighbors that elderly people interact with, the frequency of outdoor activities, and the nature of their interactions with others, which is consistent with previous research [103,134]. On this basis, in this study, we conducted a comparative analysis of the structure of SC among the elderly in three communities and found that spatial design affects the strength and breadth of social networks, thereby affecting the structure of SC. For example, LNC includes a central community park, two squares, and a nearby urban park. A centripetal spatial environment is more likely to promote the gathering of elderly people, support a wider and more active social network, and help improve the level of SC. In contrast, despite the lack of large public spaces, YNC benefits from a networked street system and green spaces between residences, which facilitate activities for individuals and partners. These features promote deeper and more intimate interactions among residents, especially among the elderly, thereby enhancing the depth and trust in their social networks. This type of social network is relatively small in scale, but the quality of interaction is high, which can form a relatively stable SC. As an open and closely connected community, HYC integrates various urban public spaces, increasing the possibility of interactions between elderly residents and people of different ages and social roles. This open space design supports the expansion of social networks (N1), enhancing the scale and diversity of social interactions. Based on this discovery, guidance can be provided for the construction of aging friendly communities in the future. For communities with a central mixed layout, enhancing the support of residential green spaces for outdoor activities is essential. This can be achieved by incorporating small parks, fitness walking trails, and other recreational facilities within the community. These amenities not only improve the accessibility and availability of activity venues but also encourage more frequent and diverse interactions among residents, particularly the elderly. In contrast, clustered layout communities often lack the spatial conditions to support a broad and diverse social network. To address this, improving the pedestrian system and enhancing the accessibility of various venues within the community is crucial. Additionally, differentiating the types of activity facilities in residential green spaces can promote more varied and meaningful interactions among the elderly, thereby supporting the development of a wider and more inclusive social network. To address this, improving the pedestrian system and enhancing the accessibility of various venues within the community is crucial. Additionally, differentiating the types of activity facilities in residential green spaces can promote more varied and meaningful interactions among the elderly, thereby supporting the development of a wider and more inclusive social network. For open-layout communities (e.g., HYC), while the high accessibility of urban public spaces facilitates the expansion of social networks, the strength of these networks may be relatively weak. To enhance the depth and quality of social interactions, it is recommended to incorporate rest and relaxation facilities in areas where elderly residents gather. Furthermore, the strategic placement of small communication venues, such as ground-floor elevated spaces and around express delivery cabinets, can foster stronger neighborhood relationships and encourage more frequent and meaningful exchanges among residents. The tailored design of green spaces, pedestrian systems, and community facilities based on the specific layout and needs of each community type is crucial for promoting sustainability and sustainability among the elderly population, which has not been extensively explored in the existing literature. This finding provides new directions for exploring the SC structure of the elderly in the future.

Overall, the formation of SC in elderly people is influenced by various factors, including the green space ratio around the community, street density, land-use mixture, spatial layout within the community, and the availability of activity supporting facilities. These elements collectively shape the opportunities for social interaction and the resulting structural and functional characteristics of SC, particularly in terms of the scale and structure of social networks. Among them, PA acts as an intermediary, and more frequent individual, collective, and intergenerational activities are related to a broader and more supportive network, which in turn increases the social support and mutual assistance level of the elderly. The depth of communication between neighbors significantly affects the stability of their social networks and the level of trust within the community. The frequency and intensity of interactions among residents are key factors determining their cohesion and participation in community life. This study supports the development of SC theory in China and confirms that the characteristics of PA and the BE in elderly people and the structure of their social networks are important determinants of SC formation in high-density areas of China.

6. Conclusions

This study aimed to explore the impact of the BE on the formation and structure of SC for the elderly in high-density cities in China. Considering the role of PA, an intermediary model was constructed based on the SEM framework. The results showed that the BE played an important role in the formation of SC for the elderly, and PA had a full mediation effect on the relationship between the BE and SC. The green space ratio, residential building density, land-use mixture, and street density of the BE have a significant positive effect on the frequency and mode of PA, particularly in individual and collective activities, and intergenerational interactions. The BE can affect the institutional characteristics of SC by affecting the frequency and mode of PA, thus affecting social networks, trust, participation, and other factors that can form SC. The findings provide empirical evidence for the role of environmental factors in SC formation, offering practical implications for urban planning and community development.

While the study provides valuable insights, it is limited by the cross-sectional nature of the data and the reliance on self-reported PA measures. Future research could explore the longitudinal designs and objective measures of PA to further validate the findings. Additionally, comparative studies across different urban contexts could enhance the generalizability of the results. In addition, some data were collected during the pandemic period; although the elderly are the main users of community activity spaces, in the post-pandemic era, young and middle-aged people and children also play an important role in the formation of SC in the community. This study also proves the important role of intergenerational activities in the formation of SC. In further research, the relationship between the BE and SC should be discussed from the perspective of user diversity.