The Role of Neighbourhood Social and Built Environments – Including Third Places – in Older Adults’ Social Interactions

ABSTRACT

Neighbourhood environments shape older adults’ social interactions. This research conceptualises a comprehensive set of perceived and objective measures of neighbourhood social and built environments, including third places, with older adults’ social interactions. In Melbourne, based on the person-environment fit framework, mediation analyses showed that the perceived social environment measures of community spirit, participation in community groups and belonging to suburb, were the strongest predictors of social interactions, followed by a few types of perceived third places. Our findings suggest that policymakers should focus on how the objective environment characteristics are interpreted as levers when planning for changes.

摘要

邻里环境能塑造老年人的社交互动。这项研究构思了一套全面的邻里社会和建成环境（包括第三场所）与老年人社会互动的感知和客观衡量标准。在墨尔本，基于个人-环境契合框架，中介分析表明，社区精神、社区团体参与度和郊区归属感等感知社会环境指标是社会互动的最强预测因素，其次是几种感知的第三场所因素。我们的研究结果表明，政策制定者在规划变革时应重点关注如何将客观环境特征理解为杠杆。

KEYWORDS:

• Urban environment
• neighbourhood planning
• third places
• aging-in-place
• healthy aging
• weak social ties

1. Introduction

The global population is aging (OECD 2021) and the WHO (2021) predicts that by 2050, the proportion of older adults will double, reaching 22% of the population. Australia, like other developed countries, is also predicted to undergo similar demographic changes (AIHW 2021). Population aging presents policymakers with several opportunities and challenges, including the provision of social care and services; ensuring social security and retirement benefits; facilitating accessible and affordable housing and transport and supporting multigenerational communities. It also poses questions about how to improve well-being for older people and promote healthy aging. This paper focuses on the latter and seeks to explore the role of neighbourhood social and built environments, including third places, in older adults’ social interactions, using Australia as a case.

Neighbourhood social interactions contribute to the well-being of older adults and are crucial to aging-in-place (Scharlach and Lehning 2012, Luo et al. 2020). Aging-in-place is a concept that has received attention from policymakers and academic researchers as a way to address the issues around aging society. Aging-in-place is where adults age in their homes, neighbourhoods and communities, despite their situational circumstances changing over time, such as children leaving home (WHO 2007). Aging-in-place has many benefits for older adults as they can maintain their connections, security and independence (Lager et al. 2015, Weijs-Perrée et al. 2015) while local governments benefit from reduced healthcare costs as independent living is preserved for as long as possible (AHURI 2021).

“Place” in aging-in-place refers to a home and its surrounding neighbourhood (Gardner 2011). Place attachment is an important contributor to older adults to be able to aging-in-place (van Hees et al. 2021). Meeting others and having casual interactions plays a key role in forming place attachment (van Hees et al. 2021), but retirement (Kobus-Matthews et al. 2010), reduced health, mobility and energy levels and other critical life events, such as death of partners and friends, (Pinquart and Sorensen 2001) can make it harder for older adults to maintain social contacts (Kemperman and Timmermans 2014, Lager et al. 2015). Our research investigates older adults' face-to-face neighbourhood social interactions. We note the importance and role of family; however, neighbourhood contacts are also important because these tend to be more frequent than family interactions (Buffel et al. 2012). Moreover, engagement outside the family unit is strongly related to overall life satisfaction and meaning of life in old age (Gallagher 2012).

We focus on neighbourhoods because older adults tend to spend much of their time in their homes and neighbourhoods (Wang and Lin 2013, Lager et al. 2015). In comparison to an individual's personal characteristics, the neighbourhood environment is also more easily regulated by policy-makers and urban planners. The interplay between neighbourhoods’ social and built environments in aging has long been recognised (Wahl and Lang 2003, WHO 2023) and is important for older adults’ healthy aging and well-being (Lui et al. 2009, Steels 2015), but, so far, there is no agreement on the nature and strength of these relationships. A review article by Padeiro et al. (2021) concluded that most studies investigating built, social and service environments with older adults’ well-being, failed to simultaneously incorporate these three environmental dimensions. Similarly, Lee and Tan (2019) highlighted that there was little empirical evidence investigating built and social environments with older adults’ social connectedness. Thus, we ask the main research question using the example of the metropolitan Melbourne, Australia:

Research Question: Which built and social neighbourhood environment features and socio-demographic attributes are associated with the frequency and satisfaction of social interactions of older adults?

We use the construct of third places to address the research question. The nomenclature comes from Oldenburg (1989), who divided places in neighbourhoods into first places as homes, second places as work and third places as community places. The role of third places for the aging population has received relatively little attention (Gardner 2011, Hickman 2013, Lane et al. 2020) even though their importance in fostering neighbourhood social interactions (Thompson 2018) and maintaining older adults’ well-being (Gardner 2011, Lee 2015, Lee and Tan 2019, Sugiyama et al. 2022) has been recognised.

This paper departs from established research by emphasising the importance of perceived social interactions in the well-being of older adults and argues that this should be more deeply considered when examining how neighbourhood environments affect well-being. Aligned with the WHO (2023) guidelines for healthy aging and community programmes, enhancing the living environments of older adults in urban communities not only improves their well-being but also yields multigenerational benefits for the broader community, fostering a healthier, more just and sustainable future for all.

This paper continues next by providing background highlighting the need for more comprehensive sets of urban environment measures; outlining a broad conceptual framework; reviewing the multidisciplinary literature on place-based social interaction for older adults and specifying the study’s hypotheses.

2. Background

No universally accepted theory offers a comprehensive set of social and built environment measures that contribute towards older adults’ social interactions. Therefore, literature from different disciplines was reviewed to identify relevant measures that could be added to this study.

Within each discipline, the research may be generally viewed as mature – but as we argue, more comprehensive empirical models spanning disciplines are less common. The purpose of the literature review is to demonstrate the lack of comprehensive studies that capture the broad spectrum of environmental features related to older adults’ social interactions and to justify the need for a large set of measures in combination with a broad conceptual model.

2.1. A Person-Environment Fit Framework

We used a broad conceptual framework to guide our empirical analysis - the Person-Environment fit (P-E) framework, developed by Lawton (1982). The framework is foundational in environmental gerontology, and it suggests that the behaviour and well-being of older people are influenced by the fit between their personal attributes (for example, physical and mental health) and environment (for example, home and neighbourhood) (Wahl 2006, Baum et al. 2016). The P-E framework is associated with people’s ability to age-in-place (Park et al. 2017).

Figure 1 shows a P-E framework for the residential satisfaction and well-being of older adults (adopted from Kahana et al. 2003: psychological characteristics removed). Box E includes both objective and perceived characteristics of the environment for both physical and social domains. This study focuses on predicting social interactions of older people in the social domain using both objective and perceived measures of the built and social urban environments while controlling for personal characteristics. However, we also recognise that the effects of objective environments on social interaction frequency and satisfaction are commonly mediated by subjective perceptions of those environments.

Figure 1. Person-Environment fit model, adapted from Kahana et al. (2003).

Our aim is to identify important aspects of urban environments for enhancing social interaction frequency and social interaction satisfaction of older adults, both directly and indirectly by improving the fit between older people and their urban environments.

2.2. Local Environments and Social Interactions of Older Adults

The P-E framework can be used to evaluate objective factors of neighbourhoods and residents’ perceptions of these factors, noting that the two evaluations may differ based on peoples’ personal characteristics, such as age, gender and health condition (Moser 2009). Moreover, it is found that people’s perceptions of the environment impact their behaviour, but perceptions, at least to some degree, are influenced by objective environments (Hou et al. 2020). According to Sallis et al. (2006), the perceived environment refers to an individual's perception of the built environment. However, planning strategies that impact the objective environment may not yield the intended outcomes if older residents are unable to see the advantages of these strategies (Yue et al. 2022). Thus, it is crucial to comprehend the connection between objective and perceived measures in the context of older residents (Yue et al. 2022). To our knowledge, only a few studies (Bowling and Stafford 2007, Levasseur et al. 2020, Zhong et al. 2022) have delved into the investigation of older adults’ social interactions concerning perceived and objective measures of built and social environments, with only a limited number of measures being considered. The findings of these studies are not consistent, suggesting that it is necessary to take into account a broad range of measures. Bowling and Stafford (2007) investigated the associations between neighbourhood features and social and physical functioning, including frequency of social interactions, finding no associations between the built environment and social interactions in the UK. They found that lower levels of neighbourliness were associated with fewer social interactions (Bowling and Stafford 2007). Levasseur et al. (2020) studied neighbourhood characterises with social participation in Quebec, Canada, considering metropolitan, urban and rural areas. They concluded that social participation, which considered the frequency of involvement with family and friends outside the household, was associated with the concentration of older adults in the area, travelling by paratransit, medical clinics, home adaptions, density of road intersections and social deprivation (Levasseur et al. 2020). Zhong et al. (2022) investigated the relationship between the perceived social and built environments, along with their objective built environments, for older adults’ intergenerational and peer social interactions, focussing on ways to measure and develop intergenerational communities in Austin, US. They concluded that the diversity of age groups, digital communication with older adults, social places in neighbourhoods and objective distance to train stations contributed to all older adults’ social interactions (Zhong et al. 2022).

2.3. The Role of Neighbourhood Social Environment in Older Adults’ Social Interactions

The social environment refers to the relationships, norms and cultural practices that influence social behaviour within a community (Buffel et al. 2012). In a study conducted in the Australian Capital Territory, researchers have found that trust and neighbourliness enhance older adults' feelings of comfort and support (Windsor et al. 2012). Comfort and support, in turn, encourage social participation and increase the exchange of information (Cattell 2001). Social participation, participation in community activities and volunteering are social environment features that have been found to increase the frequency of older adults' social interactions as found in Singapore (Aw et al. 2017), and provide a sense of meaning and support active aging in the US (Black et al. 2015). Sense of community (Morris 2012, Van Hees et al. 2017), sense of belonging (Lager et al. 2015) and place attachment (Buffel et al. 2012) are related to older adults’ social interactions (Windsor et al. 2012). Various safety factors, including objective elements such as crime and traffic hazards, have been linked to access and participation in using neighbourhood spaces (Bowling and Stafford 2007). Additionally, the perception of neighbourhood safety has been found to influence residents’ interactions with their neighbours in the US (Hong and Frank 2018).

2.4. Built Environment Contributions to Social Interactions

The literature review confirmed the neighbourhood built environment influences peoples' walking behaviour (Leslie et al. 2007, Villanueva et al. 2014), providing opportunities for neighbourhood social interactions (i.e. Van Cauwenberg et al. 2011). However, the strength of evidence for other neighbourhood-built environment features is mixed.

Walkability, land use mix and density: Walkable-built environments are embodied in urban morphologies, such as the “urban DMA” classification by Dovey and Pafka (2020) (1) Density of people and buildings, (2) Mix of uses and attractions and (3) Access to destinations. However, the nature of the urban DMA and older adults’ social interactions is far from established. Researchers agree that higher land-use mix contributes towards positive social interactions (Leyden 2003, Farber and Li 2013, Neutens et al. 2013, Easthope and McNamara 2015) and overall social well-being (Mouratidis 2018). We note that none of these studies focussed on older adults. There is little agreement on whether, and to what extent, residential density is associated with social interactions. Delmelle et al. (2013), for example, concluded that higher residential densities increase neighbourhood social contact in Vienna, Austria and increase social well-being in Oslo, Norway (Mouratidis 2018). Similarly, Dempsey et al. (2012) found that medium residential densities support neighbourhood social contacts in the UK. However, Brueckner and Largey (2008), Wood et al. (2010) and Feng et al. (2017) did not find any associations between density and social capital, which included social interactions in the US, sense of community in Perth, Australia and older adults’ quality of life in Nanjing, China, respectively. It is noteworthy that all these studies have investigated urban areas with different densities; however, studies in Europe revealed positive relationships, while studies conducted in the US, Australia and China yielded non-significant results. Different contexts may thus yield different results: European cities tend to be more compact, whereas the cities in the US and Australia are typically more car-dependent. Additionally, most of these studies focused solely on objective density measures and covered participants of all ages. Given our specific focus on older adults and to address the knowledge gaps, perceived and objective density measures are added to this study.

Access and destinations: Mobility and accessibility in neighbourhoods have been long discussed, particularly concerning aspects, such as the economy, efficiency, land use planning and health and well-being (Johnson et al. 2017). Closer proximity and better access to destinations are associated with increased social interactions for participants of all ages in the Netherlands (Sharmeen et al. 2014) and social participation for older adults in Santiago, Chile (Krellenberg et al. 2014). There are ongoing debates, however, as to what is a comfortable walking distance and hence good proximity for older adults (Sugiyama et al. 2022). Distances tested have ranged from (Ribeiro et al. 2015) 200 to 800 m (Yang and Diez-Roux 2012). To investigate this further, objective built environment measures were studied in five differently sized neighbourhoods in this study.

Third places: Third places are important for fostering social interactions for adults of all ages (Thompson 2018) and older adults’ well-being (Gardner 2011). However, the role of third places as facilitators of social interactions of older adults has received little attention (Lane et al. 2020). Quantitative studies have mostly considered one or a very limited number of third places with social interactions. For instance, urban green areas in the Netherlands (Kemperman and Timmermans 2014) and Santiago, Chile (Krellenberg et al. 2014), urban parks (Peters et al. 2010) and neighbourhood open spaces in Copenhagen, Denmark (Schmidt et al. 2019), public spaces in Perth, Australia (Francis et al. 2012) and libraries in the US (Lawson 2004). The exceptions are Lee (2015) who investigated third places with older adults’ social connectedness in Texas, the US and Richard et al. (2013) studied neighbourhood resources with older adults’ social participation in Montreal, Canada. Lee and colleagues found residential facilities with food and retail outlets enhanced older adults’ perception of social connectedness (Lee 2015, Lee and Tan 2019). Richard et al. (2013) concluded that access to neighbourhood resources, such as food, businesses and services and leisure and physical activities increased social participation in communities. The perceived importance of eight types of third places for older adults’ social interactions is examined in this paper as part of the perceived neighbourhood environment. Objectively measured DMA attributes of seven types of third places are studied as part of an objective built environment.

2.5. Types of Social Interactions

Social interactions occur with friends, family, colleagues, acquaintances and strangers (Farber and Li 2013). These may be spontaneous and brief conversations that are positive in nature create trust and connection among people and can offer practical and emotional support (Du Toit et al. 2007). Often these casual conversations happen in third places (Leyden 2003).

The vast majority of social interactions research in the urban planning discipline focuses on social interaction frequency (Brueckner and Largey 2008, Morita et al. 2010, Sharmeen et al. 2014, Easthope and McNamara 2015, Van den Berg et al. 2016), number of social contacts (Wang and Lin 2013) or social network size (Kalmijn 2012), while some qualitative studies have investigated social interactions more generally (Gardner 2011, Hickman 2013). Satisfaction with neighbourhood social interactions has been investigated in a limited number of studies (in Bonsang and van Soest 2012, Delmelle et al. 2013, Weijs-Perrée et al. 2015). Gibson et al. (2010), Pinquart and Sorensen (2001) and Van den Berg et al. (2016) suggested that the quality of social interactions may be more important than the quantity regarding older adults’ quality of life and life satisfaction. This paper considers both frequency and satisfaction of social interactions in older adults.

2.6. Hypotheses

Based on the literature review presented, we present three hypotheses:

Hypothesis 1 (Perceived and objective built environment): Higher perceived and objective density, land use mix, and accessibility scores will have a significant positive relationship with social interaction frequency and social interaction satisfaction.

Hypothesis 2 (Perceived third places): Higher satisfaction with social interactions in third places will have a significant positive relationship with social interaction frequency and social interaction satisfaction.

Hypothesis 3 (Perceived social environment): Higher perceived social environment scores will have a significant positive relationship with social interaction frequency and social interaction satisfaction.

3. Materials and Methods

The P-E fit framework focuses on the capacities of a person, for example, health and mobility and characteristics of the environment (Lawton and Nahemov 1973, Lawton 1982, Thomése and Broese van Groenou 2006). Personal characteristics, social environment and social interactions were measured using perceptions. The built environment was measured using perceived and objective data. Perceptions were collected using survey and objective data via Geographic Information Systems (GIS) software. The selection of the measures and data sources was informed by an extensive literature review and is introduced further in the following sections.

3.1. Study Sample

The study was conducted in six Local Government Areas (LGAs) in Greater Melbourne, Australia, see Figure 2. As identified in the literature review, thus far, only a few studies examining social interactions and the environment with older adults have been conducted in Australia (e.g. Windsor et al. 2012). Melbourne provides an interesting study context as it is often characterised as a vibrant and liveable city (DELWP 2017). The LGAs chosen represented a range of built and social environment features, urban DMA, across Melbourne and exemplified different urban areas. Three inner and three outer suburbs were selected. To be eligible, survey participants needed to be 18 years or older, live in one of the six LGAs, and be able to speak English. A third-party company selected the survey participants randomly. Participants were invited to complete a computer-assisted telephone interview or an online survey. The survey took approximately 30 min to complete. Three quotas were imposed: age – 18–34 years, 35–54 and 55 years or above; sex – male or female and employment status – working or not working. Older adults in this research were considered people aged 55 years or above. The research was approved by the ethical review process of the CSIRO (ethics clearance no 015/5) and the University of Melbourne (ID 1544874.1).

Figure 2. Study areas in Greater Melbourne.

3.1.1. Independent Perceived Variables

Perceived measures were obtained from the CSIRO Community Functioning and Wellbeing Survey was conducted in 2015. The survey measured different aspects of quality of life. Thirteen perceived measures were adopted. These are shown in Table 2 and Figure 3. The survey items were designed to be treated as sets, therefore, mostly composite measures instead of single-item measures were used. A composite score was calculated by taking the arithmetic means of each multiple-item set. The exact survey measures and their reliability are shown in Appendix A.

Figure 3. Models of social interaction frequency and social interaction satisfaction. NOTE: Social interactions frequency model is shown using bold arrows, while the social interactions satisfaction model is shown using fine arrows, TP= third places.

3.1.2. Independent Objective Variables

Geographic Information Software (GIS) analyses were used to obtain objective built environments and third places measure in the survey participants’ neighbourhoods. Survey participants’ addresses were converted into geographic coordinates in QGIS Desktop 2.16.2. The objective built environment was measured with 13 measures using the ArcGIS software. These are shown in Figure 3. Formulas of the objective measures are shown in Appendix B.

Considering the inconsistency in previous research of “walkable distances” for older adults, street network distance buffers of 100, 200, 400, 600, 800 and 1000 m were created from the survey participants’ homes to obtain six differently sized “neighbourhoods” using ArcGIS 10.3.1 Network Analyst New Service Area tool.

3.1.3. Dependent Variables

Two outcome measures were used, survey respondents’ perception of social interaction frequency and satisfaction with social interaction frequency, referred to as social interaction satisfaction hereinafter.

Social interaction frequency was measured by combining five survey items into one composite measure by calculating their arithmetic mean. Social interaction satisfaction was a single-item measure. Detailed social interaction items are shown in Appendix A.

3.2. Analysis

The importance of the social and built environment measures in predicting social interaction frequency and social interaction satisfaction (hypotheses 1–3), was tested with mediation analysis. Two models were developed and tested, one for social interaction frequency, shown by bold arrows in Figure 3, and the other for social interaction satisfaction, shown by fine arrows.

Social interaction frequency and satisfaction can also reasonably be expected to influence each other in feedback loops. However, we did not include these as predictors of each other in our models as the focus of this paper was on establishing the relative importance of built and social environment predictors on social interaction frequency and social interaction satisfaction.

In these models, the objective built environment was hypothesised to be an indirect predictor of perceived social interactions. It was hypothesised to influence perceptions of the built environment, which, in turn, affects social interaction frequency and social interaction satisfaction. A similar conceptual framework has been previously used when predicting older adults’ mental well-being (i.e. Guo et al. 2021).

Multiple regression analyses were carried out to test the mediation. Social interaction satisfaction was a single-item measure in the survey and, therefore, an ordinal variable. Normally, ordered logistic regression analysis is used to model an ordinal dependent variable, but its main assumption, the proportional odds assumption, was violated. The assumption was violated even after the social interaction satisfaction measure categories were changed. However, linearity was tested before the analysis, the relationships were found to be linear, and social interaction satisfaction was treated as a continuous variable in this study. Data analyses were carried out using Stata14 software.

4. Results

4.1. Socio-demographic Attributes of the Sample

Overall, 476 of this sample were aged 55 years or above and therefore included in this analysis. The socio-demographic attributes of the participants are shown in Table 1. Over half of participants were older than 65 years (56.7%) and female (52.5%). Overall, 61.4% of the sample were retired. The most common household type was a couple with no children (50.9%) or single-person (28.9%) households. Most participants were satisfied or very satisfied with their physical mobility (81.3%) and health (73.9%). On average, respondents spent 5.3 days a week in and around their suburbs.

Table 1. Socio-demographic attributes.

Socio- demographic attributes	n	%
Age (years):
55-65	206	43.3
65+	270	56.7
Sex:
Female	250	52.5
Male	226	47.5
Satisfaction with physical mobility:
5 (very satisfied)	224	47.1
4	163	34.2
3	59	12.4
2	21	4.4
1 (very dissatisfied)	9	1.9
Satisfaction with health:
5 (very satisfied)	173	36.3
4	179	37.6
3	84	17.6
2	28	5.9
1 (very dissatisfied)	12	2.5
Employment:
Retired	274	61.4
Working part time	93	20.9
Working full time	79	17.7
House type:
Separate house	323	68.6
Semi-detached house, townhouse	59	12.5
Flat, unit or apartment	89	18.9
Tenure:
Own	423	90.5
Rent	44	9.5
Household type:
Couple with children	89	20.2
Couple with no children	224	50.9
Single person household	127	28.9

4.2. Descriptive Statistics

The means of the perceived measures are presented in Table 2. The means show that most respondents were very satisfied or satisfied with their perceived social and built environments. Only the social environment measure for participation in community groups had a mean of 2.5, showing that fewer participants agreed that they participate in community groups. The built environment measures for access and feelings towards urban growth had means < 3.0. Social interaction frequency had a mean of 3.5 and social interaction satisfaction had a mean of 3.8, suggesting that participants were satisfied with their local social interactions. The built environment measure density had a mean of 4.1, reflecting that 75.0% of respondents strongly agreed or agreed that their suburb was becoming denser. However, 47.4% agreed or strongly agreed that urban growth represents problems for their suburb.

Table 2. Descriptive statistics of the perceived measures.

Measure	Observations	Mean	SD	Reliability (Cronbach alpha)	Spearman Correlation coefficients
					SI frequency	SI satisfaction
	PERCEIVED SOCIAL ENVIRONMENT
Social interaction (SI) frequency	454	3.50	0.89	0.81	N/A	0.66
Social interaction (SI) satisfaction	454	3.80	1.03	N/A	0.66	N/A
Belonging to suburb	454	4.29	0.82		0.44	0.47
Safety	474	3.84	0.88	0.82	0.25	0.22
Participation in community groups	473	2.53	1.28	0.87	0.41	0.33
Community wellbeing	474	4.28	0.73	0.87	0.37	0.34
Community spirit	473	3.67	0.86	0.93	0.57	0.61
Community inclusion	469	3.68	0.84	0.88	0.35	0.38
Levels of trust	467	3.50	0.82	0.83	0.44	0.42
	PERCEIVED BUILT ENVIRONMENT
General appearance	474	3.97	0.74	0.88	0.33	0.32
General environment	474	3.76	0.92	0.87	0.25	0.24
Density	472	4.06	1.11	N/A	0.05	0.10
Mix	470	3.37	1.05	N/A	0.19	0.21
Access	473	2.90	1.29	N/A	−0.09	−0.07
Feelings towards urban growth	451	2.79	0.49	0.88	0.12	0.08
	PERCEIVED THIRD PLACES^3
Cafes/bars/restaurants	472	3.61	1.27	N/A	0.32	0.26
Shops	471	3.42	1.23	N/A	0.31	0.23
Services	468	3.24	1.33	N/A	0.27	0.11
Public transport stops	462	3.10	1.37	N/A	0.27	0.09
Natural environments	471	3.57	1.22	N/A	0.38	0.23
Footpaths	473	3.48	1.16	N/A	0.47	0.32
Open public spaces	473	3.44	1.14	N/A	0.42	0.24
Community places	464	3.59	1.21	N/A	0.32	0.27

Third places, such as cafes, bars and restaurants (mean of 3.61), community places (mean of 3.59) and natural environments (mean of 3.57), had the highest means when participants considered the importance of third places for their social interactions.

Table 2 shows that correlations between community spirit, social interaction frequency and social interaction satisfaction were moderately strong with coefficients of 0.57 and 0.61, respectively. The remaining correlations between social interaction frequency and social and built environment measures were weaker with coefficients < 0.50. Similarly, social interaction satisfaction had weaker correlations with these measures.

The means of the objective measures in all neighbourhood sizes are presented in Appendix B.

4.3. Mediation Analysis

1. The mediation was tested with the four steps presented in Baron and Kenny (1986) and Kenny (2018). The steps were as follows: (1) Test that the causal variable, the objective built environment in our study, is correlated with the outcome variables, social interactions frequency and satisfaction in our study;

2. Test that the causal variable (objective built environment) is correlated with the mediator, the perceived built environment in our study;

3. Test that the mediator (perceived built environment) affects the outcome variables (social interactions frequency and satisfaction); and

4. Test that the effect of the objective built environment on social interactions frequency and satisfaction controlling for the perceived built environment is not significantly different from zero.

These mediation conditions were not met. In particular, the perceived built environment, including third places, did not mediate the objective built environment, including third places, nor did the objective built environment directly predict social interactions. This was similar for all neighbourhood buffer sizes for both social interaction frequency and social interaction satisfaction. This means that hypothesis 1 was partially rejected. Thus, since the objective measures were deemed not important for predicting social interaction frequency and satisfaction, they were excluded from further analyses.

4.4. Social Interaction Frequency

The perceived social and built environments were used to model social interaction frequency. All the perceived social and built environment and perceived third places measures were entered into a multiple regression model predicting social interaction frequency, while statistically controlling for socio-demographic attributes. As this research aimed to distinguish the most important social interaction frequency predictors and to develop an easily replicable model, a decision was made to include only the significant variables in the final model, shown in Table 4. First, each independent variable group, see Table 3, was modelled separately with social interaction frequency to capture the relative importance of these groups. Data in Table 3 show that the perceived social environment measures and the perceived third places had the highest R² values; therefore, they were the most explanatory predictors of social interaction frequency in the model. Therefore, our study supported hypotheses 2 and 3.

Table 3. Variation explained by different independent variable groups when predicting social interactions.

Independent variable groups	R^2
	Social interaction frequency	Social interaction satisfaction
Socio-demographic characteristics	0.12	0.14
Perceived social environment	0.42	0.51
Perceived built environment (DMA)	0.13	0.12
Perceived third places	0.24	0.13

Table 4 presents the final model, which was statistically significant. Approximately half (47%) of the variance in social interaction frequency was predicted by five independent variables. The perceived social environment measures: (1) belonging to the suburb, (2) community spirit and (3) participation in community groups, alongside the perceived social interactions in third places such as in (4) cafes, bars and restaurants and on (5) footpaths were significant social interactions frequency predictors (p < 0.05). Thus, hypotheses 2 and 3 were accepted. While the mean score for participation in community groups was lower compared to other social environment measures scores, the model indicates that individuals who participated in community groups tended to have higher levels of social interaction frequency. None of the socio-demographic variables were significant in predicting social interaction frequency. Based on the P-E framework, we assumed that personal characteristics, such as satisfaction with health, age group or employment status, contribute to older adults’ social interaction frequency.

Table 4. Social interaction frequency and satisfaction models.

Independent variables	Social interaction frequency	Social interaction satisfaction
Belonging to suburb	0.15*** (0.04)	0.26*** (0.05)
Community spirit	0.33*** (0.04)	0.49*** (0.05)
Participation in community groups	0.19*** (0.03)	0.12*** (0.03)
Importance of cafes, bars and restaurants (TPs)	0.08*** (0.03)	–
Importance of footpaths (TPs)	0.18*** (0.03)	-
Importance of shops (TPs)	-	0.10*** (0.03)
Importance of services (TPs)	-	−0.09*** (0.03)
Age	-	-
Sex	-	-
Employment	-	-
Satisfaction with health	-	-
Constant	0.34* (0.19)	0.39** (0.22)
Observations	466	455
R-squared	0.48	0.45
Adjusted R-squared	0.47	0.43
Prob > F	<0.001	<0.001

Notes: Standardised beta coefficients followed by a standard error in brackets; *** p < 0.01, ** p < 0.05, * p < 0.1; TPs = Third places.

4.5. Social Interaction Satisfaction

Social interaction satisfaction was modelled similarly to social interaction frequency. As before, the objective built environment measures were excluded from the modelling. The significant model variables for social interaction satisfaction are presented in Table 4.

The model was significant, with approximately 44% of the model variance ascribed to the independent variables. Five variables were significant for predicting social interaction satisfaction: belonging to the suburb, community spirit, participation in community groups and perceived importance of social interactions in shops and services, such as medical practices, banks and hairdressers, p < 0.05. The small negative coefficient for the importance of services suggests that older residents who rely on services for their social interactions tend to have less overall satisfaction with social interactions in and around their suburbs. Thus, hypothesis 3 was accepted and hypothesis 2 was partially accepted.

5. Discussion

This research adds to the existing literature by conceptualising and considering the most comprehensive set of perceived and objective measures to date that captured the social and built environments, including third places, with older adults’ neighbourhood social interactions based on the P-E framework. Based on the mixed results in previous research and by simultaneously adding as many model variables as possible, our aim was to identify which specific environmental factors are most important for older adults’ social interactions, thereby providing more precise and accurate insights for policy-makers and urban planners to create age-friendly environments.

Our findings highlight the importance of promoting community spirit and participation in community activities among older adults for the P-E fit. Lee and Kim (2017) discuss that policy-makers could achieve this by offering senior employment, classes, workshops and events. Melbourne is located in the state of Victoria. The Victorian Government (2022) has initiated several programmes and policies that support healthy aging. The Age-Friendly Victoria Declaration is working to create more age-friendly communities, services and local environments. The Seniors Card Age-Friendly Partners Programme is fostering partnerships with businesses to develop age-friendly products and services, further enhancing the lives of older citizens (The Victorian Government 2022). Internationally, the WHO (2023) has taken a leading role in the United Nations Decade of Healthy Ageing 2021–2030, a global collaborative effort aimed at enhancing the quality of life for older adults, highlighting the importance of the issue. The WHO (2023) has also recently released a publication titled “National Programs for Age-Friendly Cities and Communities: A Guide”, which underscores the significance of social and physical environments in shaping the well-being of older individuals. The mediation analysis did not find that the objective built environment measures contributed to either social interaction frequency or satisfaction, directly or indirectly. This was unexpected, partially rejecting hypothesis 1, and suggests how older adults perceive their built environments as more important for social interactions than aspects of their objective built environment. In contrast to our findings, the P-E framework identified the importance of considering both objective and perceived measures in well-being studies (Moser 2009). This study did not support this aspect of the P-E framework. While it would be premature to draw a conclusive statement about the relevance of incorporating objective built environment measures into the P-E framework, our finding that perceived measures were more important than objective measures aligns with previous behavioural research. For example, researchers have found that physical activity (Koohsari et al. 2015), walking (Chan et al. 2021) and perception of well-being and community cohesion (Kent et al. 2017) were more strongly associated with people’s perception of an environment than by the actual environment. Hence, further research is warranted to determine whether, to what extent, and how objective measures should be measured and included within the P-E framework.

Built environment measures might not be strongly associated with neighbourhood assessments because people choose neighbourhoods that already have attributes present that are important to them (McCrea et al. 2014) and older adults tend to be attached to their neighbourhoods (van Hees et al. 2021). Neighbourhood self-selection is based on a range of factors, including economic means and personal preferences, suggesting a possible bias towards the particular environment (McCrea et al. 2014). While the survey did not ask questions about residential preferences, socio-demographic factors were not significant predictors of social interactions.

There are several limitations to this study. The survey was only carried out across six LGAs in Melbourne, Australia. While these LGAs were chosen to represent different types of urban areas and socio-demographic groups, the results were limited to the Melbourne context and might differ in cities with different densities, populations, urban environment configurations and features and socio-cultural and economic contexts. Furthermore, the survey was not designed specifically for older adults; they were one demographic group in a broader survey of the general adult population, and thus the sample size was relatively small. It is also worthwhile mentioning that the objective measures of third places weighted different types of third places equally in this study. The importance of different third places will likely differ across the population.

6. Conclusions

The main contribution we make is assembling a broad range of perceived and objective social and built environment measures to predict their importance for older adults’ social interactions in an empirical enquiry of metropolitan Melbourne. By incorporating both objective and perceived neighbourhood measures, this analysis provides a more comprehensive understanding of how place and our perceptions of place affect social interactions, thus enhancing our understanding of the complex relationship between neighbourhood environment and social interactions in older age.

Based on the P-E framework we found that, in Melbourne, the perceived social environment was more important for older adults’ neighbourhood social interaction frequency and satisfaction than the perceived and objective built environments and personal socio-demographic attributes, such as age and satisfaction with health and mobility. Community spirit, participation in community groups and belonging to suburb, were the strongest social environment predictors of social interaction frequency and satisfaction together. Perceptions of certain types of third places, like cafes, shops, services and footpaths, were relatively less important in predicting neighbourhood social interactions.

We would not have been able to reach the above conclusions without considering such a comprehensive set of measures. The results contribute to the body of knowledge supported by the P-E framework by highlighting that the fit between older adults’ personal preferences and their perceived evaluations of environments are much more important for their social interactions and well-being than the fit between their personal preferences and objective environments. This suggests that researchers, policy-makers and practitioners using the P-E framework should focus on older adults’ perceptions of environmental characteristics, that is, how the objective characteristics are interpreted as levers when planning for changes in objective aspects of the environments. The conclusions drawn from this study have significant policy implications for those interested in improving older adults’ well-being.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Dr Piret Veeroja was supported by the University of Melbourne MIRS and MIFRS scholarships and CSIRO Postgraduate Studentship. Professor Hannah Badland is supported by an RMIT Vice-Chancellor’s Senior Research Fellowship.

Appendices

Appendix A. Perceived social and built environment measures

Table A1. Perceived social and built environment measures and their reliability.

Measure (no of items)	Reliability (Cronbach alpha)
PERCEPTION OF SOCIAL ENVIRONMENT
1) Social interaction frequency^1: I stop and talk to my neighbours in my suburb I stop and talk to strangers in my suburb I say hello to people in my suburb I go out together socially with others in my suburb I enjoy simply seeing others when out in my suburb	0.81
2) Social interaction satisfaction^1:    Overall, I am satisfied with the amount of my social interaction in and around my suburb	n.a.
3) Safety (4)^1: It is safe to walk alone in the street during the day It is safe to walk alone outside at night It is safe to leave the car on the side of the road at night Overall, I feel safe living in this suburb	0.82
4) Participation in community groups (4)^1: You regularly help out a local group as a volunteer You have attended several community events in your suburb in the past year You are a very active member of a local organisation or club Overall, you participate regularly in a variety of community activities in the suburb	0.87
5) Community wellbeing (3)^1: This suburb is suitable for seniors Overall, this suburb offers a good quality of life Overall, I am happy living in this suburb	0.87
6) Community spirit (4)^1: People can rely upon one another for help People have friendly relationships People can work together if there is a serious problem Overall, I am satisfied with the community spirit in my suburb	0.93
7) Community inclusion (3)^1: Suburb is welcoming newcomers Suburb is welcoming of people of different cultures Overall, suburb includes everyone no matter who they are.	0.88
8) Level of trust (2)^1: People that you see around can generally be trusted Overall, I am satisfied with the levels of trust in my suburb	0.83
PERCEPTION OF BUILT ENVIRONMENT
9) General appearance (6)^2: Cleanliness Level of graffiti Greenery and parks Walkways and paths Neighbourhood character Overall satisfaction with the general appearance	0.88
10) General environment (4)^2: Amount of traffic Quality of the air Level of noise The overall quality of the general environment	0.87
11) Feelings towards urban growth (3)^1: Overall, urban growth is changing my suburb Urban growth represents problems for my suburb	0.88
12) Density^1:  My suburb is becoming denser	n.a.
13) Mix^1:  My suburb is becoming more varied, mixed and interesting	n.a.
14) Access^1:  My suburb is becoming harder to get around	n.a.
15) Third places (each item was an individual measure)^3: The importance of social interactions in the following places:   Cafes/bars/restaurants   Shops (shopping malls and speciality shops)   Services (like the doctor)   Public transport stops and hubs   Natural environments (like local parks and nature reserves)   Footpaths   Local streets and squares   In community places (like libraries, churches and sporting clubs)	n.a.

NOTES: ¹ – 5 point scale: strongly agree – strongly disagree; ² – 5 point scale: very satisfied – very dissatisfied; ³ – 5 point scale: very important – not important at all.

Appendix B. Objective-built environment measures

Table B1. Objective built environment measures.

Built environment measures		Formula
Density	Number of third places in the neighbourhood	${\displaystyle \frac{\mathrm{No}\mathrm{of}\mathrm{one}\mathrm{type}\mathrm{of}\mathrm{third}\mathrm{place}}{\mathrm{NH}\mathrm{area}(\mathrm{ha})}}$ ${\displaystyle \frac{\sum \mathrm{No}\mathrm{of}\mathrm{all}\mathrm{types}\mathrm{of}\mathrm{third}\mathrm{places}}{\mathrm{NH}\mathrm{area}(\mathrm{ha})}}$
	Population density	${\displaystyle \frac{\mathrm{No}\mathrm{of}\mathrm{persons}\mathrm{resident}\mathrm{in}\mathrm{NH}}{\mathrm{NH}\mathrm{area}(\mathrm{ha})}}$
	Residential density	${\displaystyle \frac{\mathrm{No}\mathrm{of}\mathrm{dwellings}\mathrm{in}\mathrm{NH}}{\mathrm{NH}\mathrm{area}(\mathrm{ha})}}$
	Lot coverage	${\displaystyle \frac{\sum \mathrm{Building}\mathrm{footprints}\mathrm{area}\mathrm{in}\mathrm{NH}(\mathrm{ha})}{\sum \mathrm{Parcels}\mathrm{area}\mathrm{in}\mathrm{NH}(\mathrm{ha})}}$
	Location	Inner/outer suburb
Land-use mix	Percentage of types of third places	${\displaystyle \frac{\sum \mathrm{type}\mathrm{of}\mathrm{third}\mathrm{places}\mathrm{in}\mathrm{NH}\ast 100}{\mathrm{Nr}\mathrm{of}\mathrm{all}\mathrm{types}\mathrm{of}\mathrm{third}\mathrm{places}\mathrm{in}\mathrm{NH}}}$
	Entropy index	$\sum _{i=1}^n{\displaystyle \frac{\mathrm{Pi}\ast \mathrm{lnPi}}{\ln (n)}}$ Pi – the proportion of m2 of land use i n – number of land uses
Accessibility	Distance from home to third place (m)	–
	Number of three- and more-way intersections per NH	${\displaystyle \frac{\mathrm{No}\mathrm{of}3+\mathrm{way}\mathrm{intersections}\mathrm{in}\mathrm{NH}}{\mathrm{NH}\mathrm{area}(\mathrm{ha})}}$
	Block size	${\displaystyle \frac{\sum \mathrm{Block}\mathrm{sizes}\mathrm{in}\mathrm{NH}}{\mathrm{Nr}\mathrm{of}\mathrm{blocks}\mathrm{in}\mathrm{NH}}}$
	a) Cycle path length by road length b) Cycle path length in NH	${\displaystyle \frac{\sum \mathrm{Bike}\mathrm{path}\mathrm{lenght}\mathrm{in}\mathrm{NH}(m)}{\sum \mathrm{Road}\mathrm{lengthin}\mathrm{NH}(m)}}$ ${\displaystyle \frac{\sum \mathrm{Bike}\mathrm{path}\mathrm{length}\mathrm{in}\mathrm{NH}(m)}{\mathrm{NH}\mathrm{area}(\mathrm{ha})}}$
	a) Footpath length by road length b) Footpath length in NH	${\displaystyle \frac{\sum \mathrm{Footpath}\mathrm{length}\mathrm{in}\mathrm{NH}(m)}{\sum \mathrm{Road}\mathrm{length}\mathrm{in}\mathrm{NH}(m)}}$ ${\displaystyle \frac{\sum \mathrm{Footpath}\mathrm{length}\mathrm{in}\mathrm{NH}(m)}{\mathrm{NH}\mathrm{area}(\mathrm{ha})}}$
	Closest public transport stop (m)	-
	Socio-economic status	The Index of Relative Socio-economic Advantage and Disadvantage (IRSAD) decile

NH is neighbourhood.

Table B2. Descriptive statistics of the objective measures.

Objective built environment	NH 100		NH 200		NH 400		NH 600		NH 800		NH 1000
	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Population density	21.88	20.77	19.09	14.63	17.10	9.81	16.22	9.04	15.59	7.84	15.25	7.42
Dwelling density	10.52	13.35	8.91	9.26	7.81	6.12	7.36	5.55	7.03	4.66	6.86	4.37
Lot coverage	.18	.110	.16	.10	.14	.09	.13	.08	.12	.07	.12	.07
Cafes, bars, restaurants	.07	.31	.07	.27	.08	.24	.09	.22	.09	.21	.09	.19
Community places	.01	.07	.01	.04	.01	.02	.02	.02	.02	.02	.02	.02
Public transport stops	.06	.16	.06	.12	.07	.08	.08	.07	.08	.06	.08	.06
Services	.07	.25	.07	.20	.09	.19	.09	.16	.09	.14	.09	.12
Shops	.05	.27	.06	.24	.07	.20	.07	.19	.08	.17	.08	.15
Natural environments	.07	.14	.08	.22	.08	.16	.08	.12	.09	.10	.10	.10
All types of third places	.33	.85	.35	.74	.41	.67	.42	.62	.42	.57	.43	.51
% of third places	10.64	14.77	22.94	21.52	47.77	27.50	63.78	25.35	72.56	22.19	76.53	2.09
Entropy index	.16	.13	.22	.13	.26	.13	.29	.13	.30	.13	.31	.13
3-way intersections	.53	.47	.53	.36	.52	.29	.51	.28	.51	.27	.50	.26
Cycle path by road	.14	.21	.15	.17	.16	.13	.17	.11	.17	.09	.17	.08
Cycle path by neighbourhood	0.02	28.77	0.02	22.68	0.02	17.67	0.02	15.61	0.02	13.76	.00	.00
Footpath by road	1.94	.69	1.79	.49	1.74	.42	1.71	.40	1.69	.38	1.67	.36
Footpath by neighbourhood	0.2	95.50	0.2	83.96	0.2	74.93	0.2	71.30	0.2	68.97	.02	.01
Closest public transport stop	6.67	20.30	34.82	61.52	136.43	13.97	214.00	166.89	282.00	202.72	295.95	213.89
Block size	.94	.91	1.45	1.89	1.86	2.74	1.99	2.97	2.09	3.36	2.02	3.58