Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Oct 6.
Published before final editing as: Cities Health. 2026 Sep 11:10.1080/23748834.2026.2722874. doi: 10.1080/23748834.2026.2722874

Observed and perceived walkability among young adults in San Francisco: spatial divergence, neighborhood inequality, and self-rated health

Benjamin C Townsend a,*, Suraiya Parvin a, Nafisa Anjum a, Louisa M Holmes a,b
PMCID: PMC13637480  NIHMSID: NIHMS2211768  PMID: 42836065

Abstract

Walkability is a common built environment measure for guiding health-supportive planning. However, built environment metrics do not necessarily capture resident experiences of neighborhood walkability. This study examines observed audit-based walkability and resident-perceived walkability in San Francisco as complementary but distinct measures of neighborhood walkability and examines their spatial relationships with neighborhood sociodemographic characteristics and young adult self-rated health. Data come from the 2014–2020 Bay Area Young Adult Health Panel Study, a probabilistic stratified sample of young adults aged 18–32 and census block-level neighborhood audits. We constructed composite observed and perceived walkability measures, generated continuous spatial surfaces using kriging, and estimated both global and local associations using multiscale geographically weighted regression. Observed and perceived walkability showed divergent spatial patterns and relationships with self-rated health, socioeconomic conditions, and racial/ethnic composition. The observed walkability (R2 = 0.60) model explained more variance than the perceived model (R2 = 0.24), suggesting that structural neighborhood inequalities are more directly captured by observable built environment measures. Perceived walkability, however, reflected broader lived experiences, including everyday mobility. Our findings suggest that walkable built environments in infrastructural terms do not consistently align with pedestrian experience or better health, underscoring the need to integrate experiential data into equitable urban health and planning research.

Keywords: Urban health, spatial analysis, built environment, health equity, urban planning, Young Adult Health

Introduction

Walkability, broadly defined for this study as the extent to which built environments support and encourage walking, has been shown over the past few decades to be an important predictor of chronic disease, including obesity, diabetes, and cardiovascular disease (Brown et al. 2009; Sallis et al. 2012; Braun et al. 2016; Creatore et al. 2016; Howell et al. 2019). Walkable features such as well-maintained sidewalks, trees and vegetation, and mixed land use have also been shown to influence these disease outcomes through walking and physical activity (Sallis et al. 2012; Malambo et al. 2016; Tabatabaie et al. 2023).

Walkability research has proliferated over the past 20 years with studies using both ‘objective’ and perceived measures to evaluate walking behavior and walkable features of local environments (Saelens et al. 2003; Leslie et al. 2005; Wood et al. 2010; Cain et al. 2014; Chiang et al. 2017; Deng et al. 2020; Rollings et al. 2024). Objective measures, such as GIS-derived indices and item inventories, have quantified features such as tree canopy, street connectivity, traffic density, and proximity to destinations, such as parks (McCormack et al. 2010; Wang and Yang 2019; Venerandi et al. 2024). While these objective measures provide standardized, scalable assessments, they may overlook compositional, social, and psychosocial factors that shape how residents experience their local environments. Additionally, many of them collect data at a coarse scale, without attention to less observable aspects like tree roots causing cracked sidewalks or safe crossing or bicycle infrastructure (Oreskovic et al. 2014; De Vos et al. 2023). Those that do measure at finer scales can be long and burdensome to complete and do not necessarily account for a person’s reason for walking, whether for transport, leisure, or social interaction (Sugiyama et al. 2008; Alidoust et al. 2018; De Vos et al. 2023). Objective measures may also be biased in terms of what researchers value when assessing walkability versus what residents value. This divergence may be particularly pronounced when researchers and area residents differ on demographic characteristics and lived experiences of the areas under study.

Alternatively, perceived walkability measures focus on how area residents view ease, aesthetics, and safety of walking in their neighborhoods. Perceived walkability has been linked to lower likelihood of obesity, more physical activity, greater satisfaction, and better quality of life among adults (Stronegger et al. 2010; Sallis et al. 2011; Van Dyck et al. 2011; De Vos et al. 2023). Perceived measures can better capture the contextual dynamics that objective measures may miss. However, documenting the direct experience of residents is also resource- and time-intensive (Consoli et al. 2020; Meng et al. 2024); thus, few studies have directly compared observed and perceived walkability within the same spatial analytic framework despite growing recognition that perceived walkability can be measured through a variety of approaches including self-report surveys, mental mapping exercises, and walking or stationary interviews (De Vos et al. 2023; Tabatabaie et al. 2023).

The imbalance in measurement is important because the few studies that have examined observed and perceived measures together have found that there is substantial discordance in how researchers view an environment for walking versus how residents experience it, and in some cases, observed measures have shown no relationship to resident well-being at all where perceived measures did (Gebel et al. 2009, 2011; Lee et al. 2017). It may, therefore, be the case that perceived and observed measures of walkability are capturing distinct domains of human activity that need not be juxtaposed with each other but that instead serve different purposes in our attempts to understand how to promote active populations and activity-friendly environments. This suggests that observed and perceived walkability capture related but distinct dimensions of neighborhood features and may serve different analytic purposes when seeking to understand and promote activity-supportive environments. That distinction is especially important in racially and ethnically unequal urban contexts and when observational data collectors differ sociodemographically from neighborhood residents (Axinn et al. 2011).

Racial/ethnic disparities in walkability and built environment perceptions

Research on walkability reveals persistent racial and ethnic disparities in access to supportive built environments (Sallis et al. 2011; Rollings et al. 2024). Communities with higher concentrations of racial and ethnic minorities often face underinvestment in infrastructure, leading to fewer and less well-maintained sidewalks, and fewer quality green spaces and pedestrian amenities (Joassart-Marcelli 2010; Rollings et al. 2024; Osei Owusu and Rigolon 2025). Historical segregation, discriminatory planning, and disinvestment have compounded these inequities, leaving many majority Black and Latino neighborhoods less walkable and more exposed to traffic hazards, pollution, and crime or feelings of unsafety (Scott et al. 2020; Dyer et al. 2023; Riley et al. 2024). Observable indicators can capture some of these differences; however, objective measures alone fail to capture how race, ethnicity, and socioeconomic status shape personal experiences of walkability.

Perceived safety, cultural relevance of amenities, and experiences of discrimination heavily influence whether residents feel comfortable walking or using nearby amenities, such as parks (Byrne and Wolch 2009; Gras-García and Ruiz-Azarola 2025). Existing studies show systematic differences in perceptions across racial and ethnic groups, with disinvested communities often reporting their neighborhoods to be less supportive of walking (Whitfield et al. 2018). Relying on indices like Walk Score® can obscure localized barriers such as poor lighting, inadequate maintenance, or threats from policing and surveillance (Shashank and Schuurman 2019; Bereitschaft 2023). These findings underscore the need to integrate perceived and observed assessments of walkability, particularly for racially minoritized populations who already bear disproportionate health burdens. Despite growing recognition of these inequities, relatively few studies have examined how racial and ethnic differences in both objective and perceived walkability relate to health outcomes. Addressing this gap is crucial for designing equitable, culturally responsive interventions that improve walkability across diverse communities.

Young adults and self-rated health in walkability research

Much of the walkability literature focuses on middle-aged and older adults. Young adults remain an understudied population despite the importance of this life stage for establishing long-term health trajectories (Bartshe et al. 2018; Shareck et al. 2023). Self-rated health (SRH), shown to be a robust predictor of morbidity, mortality, and health behaviors, offers a useful indicator for examining how neighborhood environments shape health in early adulthood. Young adult reports of self-rated health can provide insight into the cumulative exposures that lead to chronic diseases before they fully develop (Vie et al. 2019). Longitudinal research shows that self-rated health in adolescence and early adulthood predicts later-life health outcomes, with trajectories differing by sex, race/ethnicity, and socioeconomic background (Bauldry et al. 2012; Sokol et al. 2017). At the same time, young adulthood is characterized by transitions in education, employment, and living arrangements that influence daily routines and health behaviors (Holmes et al. 2016; Lawrence et al. 2017). Neighborhood characteristics, including walkability, aesthetics, and access to resources like healthy food and green spaces, may, therefore, shape health behaviors and early risk profiles during this formative period (Reuben et al. 2020; Christie-Mizell 2022).

Yet relatively few studies explicitly examine young adults, and those that do often rely on objective measures of walkability while overlooking how young people perceive their environments. Because young adults tend to be highly mobile and interact with neighborhoods differently than older residents, perceptions of safety, accessibility, and aesthetics may be especially consequential for shaping health behaviors and well-being (Christie-Mizell 2022). Examining both observed and perceived walkability in relation to self-rated health can, therefore, help clarify the mechanisms through which built environments influence health at a critical life stage. Notably, self-rated health among adults aged 20–29 has declined in recent decades even as older groups report improvements, suggesting growing vulnerability in this younger cohort, and highlighting the need for place-based interventions that support long-term well-being (Lysberg et al. 2021).

Despite evidence that observed and perceived walkability may capture distinct dimensions of neighborhood environments, little research has examined how these measures diverge spatially among young adults or whether they relate differently to self-rated health and neighborhood sociodemographic inequality. In this study, we, therefore, investigate both observed and self-reported measures of walkability among young adults (18–32 years) in San Francisco City & County with attention to the spatial patterns of each and their differential associations with self-rated health and sociodemographic characteristics, including race/ethnicity and socioeconomic status. We derived the observed walkability measure via systematic neighborhood audits conducted by trained observers, while we generated the perceived walkability index from a probabilistic household survey of young adults, stratified by race and ethnicity, using validated self-reported items. The goal of this study was not to evaluate observed and perceived walkability as interchangeable measures of the same construct because valid measurement strategies differ for audit-based and self-reported assessments. The measures were intentionally designed to meet the neighborhood assessment and household survey’s distinct objectives. Thus, in this study, we sought to explore how well these different measures capture similar spatial patterns to identify any gaps between more commonly used built environment measures and resident and pedestrian experiences of walking, and to assess whether they relate differently to self-rated health and neighborhood sociodemographic contexts. Our goal was not to achieve measurement equivalence, but to examine spatial patterns and divergence between two approaches to understanding walkability.

We hypothesize that (1) spatial patterns of observed and perceived walkability differ significantly across the city, (2) spatial patterns of self-rated health among resident young adults vary significantly when measured against observed versus perceived walkability, and (3) block group level racial, ethnic, and sociodemographic composition impacts how residents perceive neighborhood walkability distinctly compared to observations of walkability.

Methods

Sample characteristics

The 2014–2020 Bay Area Young Adult Health Panel Study (BAYAHS) was a probabilistic longitudinal survey of young adults (18–26 at baseline; 23–32 at follow-up) in Alameda and San Francisco Counties in California that included a replenishment sample of young adults aged 18–26 (N = 481), and census block-level observational data collection in a subset of participant neighborhoods (Holmes and Ling 2016; Nguyen et al. 2023). We wished to capture contextual characteristics of the areas in which young adults resided concurrently with their self-reported health indicators. Although BAYAHS is a longitudinal cohort study, the present analysis is cross-sectional and uses contemporaneous neighborhood and health measures. From 577 total census blocks where at least one survey participant resided, we randomly selected 25% of blocks in neighborhoods with the highest concentrations of young adults and 15% of all remaining blocks in San Francisco County (N = 246 total).

Ten research assistants (RAs), who resided in the two counties, were in the participants’ age range and were racially diverse, were sent in pairs to walk each selected block in its entirety and complete an assessment instrument, programmed in ESRI Survey123. They each independently recorded various features of the block, including aesthetic items, access to amenities, sidewalk safety and maintenance, and other factors. Each block, therefore, had two assessments, and we conducted inter-rater reliability tests to ensure general agreement between the two RAs for each block. We then randomly selected one report for each block to retain for analysis of walkability. For this study, we chose to focus on San Francisco County (N = 138 of 7348 blocks) because the topographical and architectural environment differs substantially from that in Alameda County, as does the scale of the county – 7 versus 821 square miles, respectively, and San Francisco City and County are the same entity. We linked to 2015–2019 American Community Survey period data, accessed via the National Historical GIS website (Manson et al. 2021), to account for area-level differences in sociodemographic characteristics. This study was deemed exempt by The Pennsylvania State University IRB.

Measures

Dependent variables

Observed walkability was measured using items from the Survey123 neighborhood assessment instrument noted above. These final six items used 5-point Likert response scales and assessed: (1) attractiveness of buildings and homes; (2) amount of green or open space, such as parks, cemeteries, or sports fields; (3) number of nearby amenities, such as coffee shops, dry cleaners, and beauty shops; (4) number of vacant lots; (5) sidewalk safety, defined as the presence of barriers or spaces separating pedestrians from traffic; and (6) presence of streetlights. A seventh item assessed street hilliness on a 7-point scale ranging from extremely hilly to extremely flat and allowed us to account for San Francisco’s distinctive topography. We coded all items so that higher values reflected more walkable conditions. We used principal components analysis (PCA) to select the final observed walkability indicators from an initial set of 43 measures. Items with an eigenvalue > 1 were retained. We then standardized the items prior to aggregation as they were initially measured on different scales and exhibited different variances. Finally, we summed the standardized values with equal weighting, as there was no theoretical basis for assigning differential weights, to create a composite observed walkability score for each census block.

Next, we linked block-level observed walkability scores to participants’ geocoded residential address points and used ordinary kriging to estimate observed walkability values across all census blocks in San Francisco County, including blocks that did not receive direct observational audits. We used ordinary kriging to estimate continuous spatial surfaces of observed and perceived walkability because neighborhood environments tend to exhibit spatial continuity. Rather than assigning the same walkability value to all residents within a census block group or restricting analyses only to audited blocks, kriging allowed us to estimate walkability continuously across space while accounting for the spatial structure present in the observed data. We then assigned walkability values from the kriged surface to participant residential locations and used those interpolated values as the dependent variable in regression analyses.

Perceived walkability was measured using survey responses from young adult participants in the BAYAHS study. Participants rated several aspects of their neighborhood environment using Likert-scale items, including perceptions of neighborhood safety during the day and night, the attractiveness of the neighborhood, the quality of sidewalks, the ease of walking to nearby destinations, and the presence of people engaging in physical activity outdoors. These measures came from commonly used scales related to perceived neighborhood walkability, disorder, and cohesion and were not meant to map onto the observed measures directly, although safety was measured the same way in both instruments (Sampson and Raudenbush 1999; Gebel et al. 2011; Holmes and Marcelli 2020). Thus, comparisons between the two measures focus on their spatial distributions and relationships with neighborhood characteristics rather than item-by-item measurement equivalence. We coded responses so that higher values indicated more favorable perceptions of walkability and averaged the items to create a composite perceived walkability score for each participant. Internal consistency of the six-item perceived walkability scale was high (Cronbach’s α = 0.87). We then used participants’ geocoded residential address points and ordinary kriging to estimate perceived walkability values across San Francisco. Finally, we assigned values from the kriged perceived-walkability surface to participant residential locations and used those interpolated values as the dependent variable in regression analyses.

Independent variables

Self-rated health was based on the survey item ‘Would you say your health in general is (1) excellent, (2) very good, (3) good, (4) fair, (5) poor?’ We dichotomized the variable so that values of excellent or very good were set equal to ‘1’ and values of good, fair, or poor were set equal to ‘0’.

We used 2015–2019 American Community Survey (ACS) period estimates to construct a composite block group index to account for socioeconomic vulnerability. Measures included per capita income, housing costs relative to income, housing values, unemployment, poverty (below 200%), health insurance coverage, internet access, educational attainment, and population density. We divided each variable into quartiles across block groups and assigned ordinal scores, summed the scores, and then divided the resulting composite measure into quartiles to produce the index. Higher index values corresponded to lower neighborhood socioeconomic vulnerability.

Lastly, we used 2015–2019 ACS data to measure racial and ethnic composition as the percentage of the population in each census block group that was non-Latino Black and non-Latino Asian. Ethnic composition was measured as the percentage of the population in the census block group that was Latino. We omitted non-Latino White composition as the comparison group.

Analysis plan

We first constructed a neighborhood audit instrument and collected observational data at the census block level using items that captured domains of walkability identified in prior literature, such as land use, aesthetics, safety, distance to amenities, and streetscape (Frank et al. 2006; Adkins et al. 2012; Suarez-Balcazar et al. 2020; Wali and Frank 2024). We then conducted principal components analysis in R (version 1.4.1106) with the original 43 items in the built environment instruments using the prcomp function in the stats package. Variables with eigenvalues > 1 were retained for the measure. Because the items used different response scales, we standardized them before combining them into a single index. We summed the standardized item values with equal weighting, producing a continuous observed walkability score in which higher values reflect more walkable neighborhood conditions. Next, we used kriging to spatially interpolate both observed and perceived walkability across San Francisco City/County and assigned interpolated values from the resulting countywide surfaces to participant residential locations for analysis. Rather than assigning the same walkability value to all residents within a census block group or restricting analyses only to audited blocks, this approach allowed us to estimate both measures continuously across space while reducing reliance on administrative boundaries.

We first estimated global regression models in Stata version 17 to assess overall associations between walkability and the individual- and neighborhood-level covariates. We then tested for spatial autocorrelation in the dependent variables using Moran’s I in R with the spdep package. This test indicated significant spatial autocorrelation (p < .001), so we next estimated multiscale geographically weighted regression (MGWR) models in MGWR 2.2 with an adaptive kernel. Bandwidths were selected automatically using the MGWR calibration procedure with AICc optimization. We used MGWR instead of standard GWR because the factors associated with walkability likely operate at different spatial scales; MGWR accommodates spatial heterogeneity by estimating a separate bandwidth for each predictor. We fit separate Gaussian MGWR models for observed walkability and perceived walkability. The MGWR models included self-rated health, neighborhood SES, and neighborhood racial/ethnic composition as independent variables. MGWR allowed us to assess how the magnitude and direction of associations varied across space within San Francisco. We mapped local parameter estimates for locations where associations reached statistical significance at p < 0.05. A purple color scale was used for positive associations with the dependent variable and green for negative associations. For ease of interpretation and to maintain confidentiality, we displayed model results as smoothed kernel density surfaces, visualized as heat maps, rather than as individual residential points.

Results

Table 1 shows the sample characteristics for the 577 block groups included in the analysis. The block group index average across the County was 2.6 / 4, i.e. leaning middle-to-higher SES. Average self-rated health was 3.5 out of 5, leaning good to fair. The composition of selected block groups was 43% non-Latino White (compared to 36.8% in the county), 14.1% Latino (compared to 16.2%), 4.7% non-Latino Black (same as the county), and 33% non-Latino Asian (compared to 35.3% in the county).

Table 1.

Sample characteristics, bay area young adult health study (N = 577).

Characteristics μ or % / (S.D.)
Block group index (1-lowest SES; 4-highest) 2.6 (1.1)
Self-rated health 3.5 (0.57)
Racial/ethnic composition
NL White 43.0% (0.22)
Latino 14.1% (0.13)
NL Black  4.7% (0.08)
NL Asian 33.0% (0.20)
NL = non-Latino
Open in a new tab

Table 2 presents the global parameter estimates from the multiscale geographically weighted regression (MGWR) models predicting observed and perceived walkability. The model explaining observed walkability demonstrated substantially stronger explanatory power (R2 = 0.60; AICc = 1114.76) than the model predicting perceived walkability (R2 = 0.24; AICc = 1490.34).

Table 2.

Global regression results of walkability on independent variables.

Observed walkability
 Perceived walkability
β SE β SE
Block group index 0.10 0.04* 0.08 0.05
Self-rated health 0.03 0.03 0.03 0.04
Racial/ethnic composition
NH White (referent) – – – –
Latino −0.40 0.03*** −0.29 0.04***
NL Black −0.15 0.03*** −0.29 0.05***
NL Asian −0.63 0.04*** −0.17 0.05***
Model fit
R2 0.60 0.24
AICc 1114.76 1490.34
Open in a new tab
*

Note: p < .05;

**

p < .01;

***

p < .001.

The block group socioeconomic vulnerability index was positively and significantly associated with observed walkability (lower vulnerability = better walkability) but not with perceived walkability. Areas with higher proportions of Latino, non-Latino Black, and non-Latino Asian residents all showed significant negative associations with both observed and perceived walkability, but the strength of the associations differed with higher percentage Latino and non-Latino Asian block groups showing a stronger negative association with observed walkability and non-Latino Black composition a stronger negative association with perceived walkability.

Spatial patterns in the local MGWR coefficients reveal notable geographic heterogeneity across San Francisco (Figure 1). Observed walkability was most highly rated in downtown San Francisco between the Financial District, Nob Hill, Chinatown, and North Beach areas with strong values in the Mission-Noe Valley area as well. In contrast, perceived walkability was lower in much of this area, except around Cow Hollow. Instead, the Mission-Castro-Haight and Sunset and Richmond areas were deemed more walkable by young adult residents, although the values were not as strong anywhere in the county as for observed walkability.

Figure 1.

Figure 1.

Open in a new tab

a. Multiscale GWR predicted observed walkability values; b. MGWR predicted perceived walkability values. Sparser geographies correspond to lower walkability scores; denser geographies correspond to higher walkability scores.

For observed walkability, positive associations with self-rated health (Figure 2) were concentrated in Presidio and Pacific Heights and Chinatown/South of Market (SOMA) neighborhoods, while negative associations concentrated in the Lower Haight/Hayes Valley/Japantown, Mission to Excelsior, and the Inner Richmond neighborhoods. Interestingly, associations between perceived walkability and self-rated health were inverted for Presidio and Pacific Heights, Hayes Valley, and the Mission District. Negative associations also concentrated in the southern part of the county, centering on Ingleside and Bayview, while the Outer Richmond and Outer Sunset showed positive associations. Downtown areas encompassing Chinatown were regarded as walkable using both measures.

Figure 2.

Figure 2.

Open in a new tab

a. Significant local associations between observed walkability and self-rated health; b. perceived walkability and self-rated health. Negative associations with walkability = green; positive associations with walkability = purple. Sparser geographies correspond to lower beta coefficients; denser geographies correspond to higher beta coefficients.

The association between the block group index and walkability also displayed substantial spatial heterogeneity (Figure 3). In the observed walkability model, stronger positive associations were concentrated in the Noe Valley/Diamond Heights area in central SF, the southern part of the county and Cow Hollow and SOMA. The perceived walkability model showed a broader pattern of negative associations in the northeastern part of the county (higher SES-lower walkability) and positive relationships in the south, suggesting that objective measures of SES do not necessarily map onto perceptions of walkability in a consistent manner.

Figure 3.

Figure 3.

Open in a new tab

a. Significant local associations between observed walkability and block group index; b. perceived walkability and block group index.

Figures 4–6 illustrate the spatial distribution of local MGWR coefficients for neighborhood racial and ethnic composition. There were significant negative associations between proportion of non-Latino Black residents and observed walkability across the southern part of the county, but the strongest effect sizes were in the Civic Center-Tenderloin-Hayes Valley area extending north to the Marina. The pattern for perceived walkability was similar in the South and around Hayes Valley, but the strongest effect was in the Russian Hill area.

Figure 4.

Figure 4.

Open in a new tab

a. Significant local associations between observed walkability and proportion of the population that was non-Latino black; b. perceived walkability and proportion of the population that was non-Latino Black.

Figure 6.

Figure 6.

Open in a new tab

a. Significant local associations between observed walkability and proportion of the population that was non-Latino Asian; b. perceived walkability and proportion of the population that was non-Latino Asian.

Similarly, neighborhoods with higher proportions of Latino residents exhibited predominantly negative associations with both observed and perceived walkability, but the spatial pattern was relatively similar. In both cases, southern San Francisco exhibited significant negative associations between walkability and proportion of Latino residents; however, the effect was stronger for observed walkability. Similarly, there were negative associations in the northeast part of the city, but for observed walkability, these patterns were less extensive and centered around Pacific Heights and the Marina compared again to the Civic Center and Chinatown areas for perceived walkability. There was also one area of positive association for observed walkability in the northeast neighborhoods along the Bay, which contrasted with perceived walkability.

Finally, for neighborhoods with higher proportions of non-Hispanic Asian residents, the spatial pattern differed somewhat from the other racial and ethnic groups, but there was common ground in large effect sizes concentrated downtown for both measures. Negative associations with observed walkability were widespread across much of the city, particularly between downtown, Hayes Valley and the Mission. In the perceived walkability model, however, negative relationships were more spatially concentrated, with clusters appearing, in addition to downtown/Russian Hill, primarily from the upper Haight down through Visitacion Valley. In contrast to observed walkability, there were no negative associations with perceived walkability in the Richmond or Sunset neighborhoods.

Discussion

This study examined spatial relationships between observed and perceived walkability, neighborhood sociodemographic characteristics, and young adults’ self-rated health across San Francisco City and County. The results indicate that observational and perception-based measures capture different dimensions of walkability and demonstrate different spatial patterns. In our models, observed walkability was more tightly associated with the neighborhood factors included in the study, whereas perceived walkability appeared to reflect additional experiential and contextual processes not fully captured by the models. In other words, it appears that structural neighborhood characteristics, such as socioeconomic status and racial and ethnic composition, are more directly reflected in measurable features of the built environment in San Francisco than in residents’ evaluations of their neighborhood environments. This is not surprising given distinctions between observing a neighborhood in which you do not live versus residing in that neighborhood, but it does underline the importance of not conflating observational walkability measures with resident experiences of walkability.

The results largely support our three hypotheses and reinforce the idea that observed and perceived walkability operate as related but distinct spatial processes. We found clear evidence for our first hypothesis as observed and perceived walkability displayed different spatial distributions across San Francisco and different levels of explanatory power in the MGWR models. We found partial support for our second hypothesis because self-rated health did not show strong global associations with either outcome, yet its local associations varied across the city and often differed sharply between observed and perceived walkability. The absence of strong global relationships is noteworthy, as it suggests that walkability-health associations among young adults are spatially heterogeneous and analyses relying solely on global models may miss important contextual differences. We also found strong support for our third hypothesis, as neighborhood socioeconomic vulnerability and racial/ethnic composition related differently to the two walkability measures in both strength and spatial extent. In the global models, SES aligned more clearly with observed walkability, whereas in the local models perceived walkability showed a stronger spatially varying relationship, including negative associations in the northeast and positive associations in the south. These patterns indicate that infrastructural conditions and lived experience do not map neatly onto one another, and that both are necessary for understanding how urban inequality shapes walkability.

The observed spatial patterns were not randomly distributed across the city. The strongest positive observed walkability values concentrated in northeastern neighborhoods such as Chinatown, North Beach, Nob Hill, and the Financial District, while lower observed walkability values were concentrated in more residential and industrialized portions of southern San Francisco, as well as neighborhoods like Civic Center and the Tenderloin, which face many challenges, including drug use, poorly maintained housing and litter. Similarly, the MGWR models demonstrated significant negative associations between walkability and Latino, Black, and Asian population composition across multiple parts of the city, although the magnitude and spatial extent of these relationships varied between the two measures. This suggests that contemporary patterns of walkability are linked to broader spatial inequalities in neighborhood development and investment.

Like most U.S. cities, San Francisco has a history of residential segregation, uneven infrastructure investment, and land-use regulation that have shaped its neighborhood environments (O’Hern 1973; Katz and Mitchell 2006; Mirabal 2009; Gao et al. 2024). San Francisco’s original post-colonial settlements were in the Northern part of the county, including the Presidio, and the gridded neighborhoods in the northeast, including Chinatown, North Beach, and the port area. These areas not only developed earlier but have since experienced sustained public and private investment, for example in the Financial District and Fisherman’s Wharf. The street networks in this area are denser, there is more mixed land use, and there are greater concentrations of pedestrian- and tourist-oriented amenities. The stronger observed walkability scores in these neighborhoods reflect this history (Figure 1). Alternatively, the southern part of the county developed later, experiencing greater industrial zoning, urban renewal, and residential expansion. Thus, these neighborhoods tend to be lower density, more residential, and have comparatively limited pedestrian infrastructure (Bayview Hunters Point Mothers Environmental Health & Justice Committee, Huntersview Tenants Association, Greenaction for Health & Environmental Justice 2004, Brandi and LaBounty 2010). Figure 1, for example, demonstrates these lower walkability scores in the southeast in particular, and Figure 2 shows inverse associations between walkability and self-rated health.

However, the results also tell a more nuanced story. Although the observers evaluated walkability to be lower quality in the southern part of the city compared to the northeast (Figure 1), they also rated it better across the city than did residents. Observed walkability was also most highly rated in the downtown (northeast) area, whereas this was generally not a highly rated area for perceived walkability. Perceived walkability was greatest in the western and central neighborhoods like the Mission-Glen Park corridor, the Sunset, and the Richmond, where streets are wider, more residential, and experience less traffic, both foot and vehicle, than downtown. Similarly, while there was overlap between walkability measures and where people rated their health as very good/excellent (Figure 2), self-rated health was inversely related to observed walkability in neighborhoods like the Richmond and Mission-Glen Park-Excelsior while perceived walkability was positively related to resident self-rated health in these areas. Equally, although the Stonestown-Ingleside neighborhoods saw a strong inverse relationship between very good/excellent self-rated health and perceived walkability, there was no relationship in this area for observed walkability.

These divergences suggest that young adult perceptions of neighborhood walkability are complex and speak not only to observable features of the surrounding built environments but also to contemporary social dynamics, neighborhood change, and everyday patterns of mobility. Nevertheless, these results also suggest that historical planning and policy processes that have embedded more or less walkable neighborhood features in different parts of the city remain important for assessing how to mitigate distributional inequity in the built environment.

Limitations

This study has several limitations. Although the BAYAHS data were longitudinal, the analyses presented here were cross-sectional and do not support causal claims about the relationships between walkability and self-rated health. In addition, the observational audit was conducted on a subset of census blocks and then spatially interpolated across San Francisco; while this approach allowed us to estimate walkability continuously across the county, it may not fully capture highly localized variation or short-term neighborhood change. The study also took place in San Francisco, a geographically small, dense, and topographically distinctive city, which may limit generalizability. Finally, there are likely other measures not included in our models that contribute to neighborhood walkability.

Conclusion

These findings have important implications for urban health and planning. Observational measures remain important for identifying structural inequities in pedestrian-supportive infrastructure and targeting investments in safe and healthy streetscapes. However, the divergence between observed and perceived walkability shows that material improvements alone may not be sufficient to create environments residents experience as supportive of walking. For cities seeking to advance health equity, walkability assessment should, therefore, move beyond approaches that focus only on the built environment to incorporate resident perspectives, participatory assessment, and other mixed-methods strategies. Municipal agencies and public health departments should consider incorporating neighborhood-level walkability assessments that combine systematic audits with resident feedback to identify areas where investments in pedestrian infrastructure may not align with residents’ lived experiences of safety, accessibility, or comfort. Doing so can help identify neighborhoods where built environment features and lived experience diverge, allowing planners and public health practitioners to design more contextually appropriate interventions that address both the physical and social dimensions of walkability.

Figure 5.

Figure 5.

Open in a new tab

a. Significant local associations between observed walkability and proportion of the population that was Latino; b. perceived walkability and proportion of the population that was Latino.

Funding

The work was supported by the California Tobacco-Related Disease Research Program; National Cancer Institute [U01-154240]; National Institute on Minority Health and Health Disparities [5P60MD006902].

Biography

Benjamin C. Townsend is a doctoral student and graduate research assistant at the University of Georgia’s Warnell School of Forestry and Natural Resources. He began this research while completing his MS in Geography at Penn State as a member of Dr. Holmes’ GEOHealth Lab. His work uses GIS, remote sensing, and other spatial methods to examine relationships among environments, ecosystem services, land management, conservation, and human health.

Suraiya Parvin, Nafisa Anjum and Louisa M. Holmes are part of the GEOHealth Lab, which is described in the Author’s note.

Footnotes

Disclosure statement

No potential conflict of interest was reported by the authors.

References

  1. Adkins A, et al. , 2012. Unpacking walkability: testing the influence of urban design features on perceptions of walking environment attractiveness. Journal of urban design, 17 (4), 499–510. doi: 10.1080/13574809.2012.706365. [DOI] [Google Scholar]
  2. Alidoust S, Bosman C, and Holden G, 2018. Talking while walking: an investigation of perceived neighbourhood walkability and its implications for the social life of older people. Journal of housing and the built environment, 33 (1), 133–150. doi: 10.1007/s10901-017-9558-1. [DOI] [Google Scholar]
  3. Axinn WG, Link CF, and Groves RM, 2011. Responsive survey design, demographic data collection, and models of demographic behavior. Demography, 48 (3), 1127–1149. doi: 10.1007/s13524-011-0044-1. [DOI] [PubMed] [Google Scholar]
  4. Bartshe M, Coughenour C, and Pharr J, 2018. Perceived walkability, social capital, and self-reported physical activity in Las Vegas college students. Sustainability 10. doi: 10.3390/su10093023. [DOI] [Google Scholar]
  5. Bauldry S, et al. , 2012. A life course model of self-rated health through adolescence and young adulthood. Social science and medicine, 75 (7), 1311–1320. doi: 10.1016/j.socscimed.2012.05.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bayview Hunters Point Mothers Environmental Health & Justice Committee, Huntersview Tenants Association, Greenaction for Health & Environmental Justice, 2004. Pollution, health, environmental racism and injustice: a toxic inventory of Bayview Hunters Point, San Francisco. San Francisco, CA: Greenaction for Health & Environmental Justice. [Google Scholar]
  7. Bereitschaft B, 2023. The changing ethno-racial profile of ‘very walkable’ urban neighbourhoods in the US (2010–2020): are minorities under-represented? Urban Studies, 60 (4), 638–654. doi: 10.1177/00420980221110829. [DOI] [Google Scholar]
  8. Brandi R and LaBounty W, 2010. San Francisco’s Ocean view, Merced Heights, and Ingleside (OMI) neighborhoods 1862–1959 (historical context statement). San Francisco, CA: San Francisco Historic Preservation Fund Committee. [Google Scholar]
  9. Braun LM, et al. , 2016. Walkability and cardiometabolic risk factors: cross-sectional and longitudinal associations from the multi-ethnic study of atherosclerosis. Health & place, 39, 9–17. doi: 10.1016/j.healthplace.2016.02.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Brown BB, et al. , 2009. Mixed land use and walkability: variations in land use measures and relationships with BMI, overweight, and obesity. Health & place, 15 (4), 1130–1141. doi: 10.1016/j.healthplace.2009.06.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Byrne J and Wolch J, 2009. Nature, race, and Parks: past research and future directions for geographic research. Progress in human Geography, 33 (6), 0309132509103156. doi: 10.1177/0309132509103156. [DOI] [Google Scholar]
  12. Cain KL, et al. , 2014. Contribution of streetscape audits to explanation of physical activity in four age groups based on the microscale audit of pedestrian streetscapes (MAPS). Social science and medicine, 116, 82–92. doi: 10.1016/j.socscimed.2014.06.042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Chiang Y-C, Sullivan W, and Larsen L, 2017. Measuring neighborhood walkable environments: a comparison of three approaches. International journal of environmental research and public health, 14 (6), 593. doi: 10.3390/ijerph14060593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Christie-Mizell CA, 2022. Neighborhood disadvantage and Poor Health: the consequences of race, gender, and age among young adults. International journal of environmental research and public health, 19 (13), 8107. doi: 10.3390/ijerph19138107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Consoli A, et al. , 2020. Associations between objectively-measured and self-reported neighbourhood walkability on adherence and steps during an internet-delivered pedometer intervention. PLOS ONE, 15 (12), e0242999. doi: 10.1371/journal.pone.0242999. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Creatore MI, et al. , 2016. Association of neighborhood walkability with change in overweight, obesity, and diabetes. JAMA, 315 (20), 2211–2220. doi: 10.1001/jama.2016.5898. [DOI] [PubMed] [Google Scholar]
  17. Deng C, et al. , 2020. A data-driven framework for walkability measurement with open data: a case study of triple cities, New York. ISPRS international journal of geo-information, 9 (1), 36. doi: 10.3390/ijgi9010036. [DOI] [Google Scholar]
  18. De Vos J, et al. , 2023. Determinants and effects of perceived walkability: a literature review, conceptual model and research agenda. Transport reviews, 43 (2), 303–324. doi: 10.1080/01441647.2022.2101072. [DOI] [Google Scholar]
  19. Dyer Z, et al. , 2023. Measuring the enduring imprint of structural racism on American neighborhoods. Health affairs, 42 (10), 1374–1382. doi: 10.1377/hlthaff.2023.00659. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Frank LD, et al. , 2006. Many pathways from land use to health: associations between neighborhood walkability and active transportation, body mass index, and air quality. Journal of the American Planning Association, 72 (1), 75–87. doi: 10.1080/01944360608976725. [DOI] [Google Scholar]
  21. Gao X, et al. , 2024. Historical redlining, contemporary gentrification, and severe maternal morbidity in California, 2005–2018. JAMA network open, 7 (9), e2429428. doi: 10.1001/jamanetworkopen.2024.29428. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Gebel K, et al. , 2011. Mismatch between perceived and objectively assessed neighborhood walkability attributes: prospective relationships with walking and weight gain. Health & place, geographies of care, 17 (2), 519–524. doi: 10.1016/j.healthplace.2010.12.008. [DOI] [PubMed] [Google Scholar]
  23. Gebel K, Bauman A, and Owen N, 2009. Correlates of non-concordance between perceived and objective measures of walkability. Annals of behavioral medicine, 37 (2), 228–238. doi: 10.1007/s12160-009-9098-3. [DOI] [PubMed] [Google Scholar]
  24. Gras-García EM and Ruiz-Azarola A, 2025. Walkability, right to the city and right to health: pathways toward equity and justice – a systematic mixed-methods review. Cities & health, (3), 1–19. doi: 10.1080/23748834.2025.2535240. [DOI] [Google Scholar]
  25. Holmes LM and Ling PM, 2016. Workplace secondhand smoke exposure: a lingering hazard for young adults in California. Tobacco control, 26 (e1), e79–384. doi: 10.1136/tobaccocontrol-2016-052921. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Holmes LM and Marcelli EA, 2020. Neighborhood social cohesion and serious psychological distress among Brazilian immigrants in Boston. Community ment health J, 56 (1), 149–156. doi: 10.1007/s10597-019-00468-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Holmes LM, Popova L, and Ling PM, 2016. State of transition: marijuana use among young adults in the San Francisco Bay area. Preventive medicine, 90, 11–16. doi: 10.1016/j.ypmed.2016.06.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Howell NA, et al. , 2019. Association between neighborhood walkability and predicted 10-year cardiovascular disease risk: the CANHEART (Cardiovascular Health in Ambulatory Care Research Team) cohort. Journal of the American Heart Association, 8 (21), e013146. doi: 10.1161/JAHA.119.013146. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Joassart-Marcelli P, 2010. Leveling the playing field? Urban Disparities in funding for local parks and recreation in the Los Angeles region. Environ plan A, 42 (5), 1174–1192. doi: 10.1068/a42198. [DOI] [Google Scholar]
  30. Katz MH and Mitchell H, 2006. Health programs in Bayview Hunters Point and recommendations for improving the health of Bayview Hunters Point residents. San Francisco Department of Public Health, 19. [Google Scholar]
  31. Lawrence E, Hummer RA, and Harris KM, 2017. The cardiovascular health of young adults: disparities along the urban-rural continuum. The annals of the American academy of political and social science, 672 (1), 257–281. doi: 10.1177/0002716217711426. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lee SM, et al. , 2017. The relation of perceived and objective environment attributes to neighborhood satisfaction. Environment & behavior, 49 (2), 136–160. doi: 10.1177/0013916515623823. [DOI] [Google Scholar]
  33. Leslie E, et al. , 2005. Residents’ perceptions of walkability attributes in objectively different neighbourhoods: a pilot study. Health & place, 11 (3), 227–236. doi: 10.1016/j.healthplace.2004.05.005. [DOI] [PubMed] [Google Scholar]
  34. Lysberg F, et al. , 2021. Age groups changes in self-rated health: a prospective longitudinal study over a 20-year period using health survey of North Trøndelag data. Scandinavian journal of public health, 49 (8), 845–850. doi: 10.1177/1403494820904626. [DOI] [PubMed] [Google Scholar]
  35. Malambo P, et al. , 2016. Built environment, selected risk factors and major cardiovascular disease outcomes: a systematic review. PLOS ONE, 11 (11), e0166846. doi: 10.1371/journal.pone.0166846. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Manson S, et al. , 2021. IPUMS National Historical Geographic Information System: version 16.0. Minneapolis, MN: Minnesota Population Center. https://pop.umn.edu/projects/ipums-usa. [Google Scholar]
  37. McCormack GR, et al. , 2010. Characteristics of urban parks associated with park use and physical activity: a review of qualitative research. Health & place, 16 (4), 712–726. doi: 10.1016/j.healthplace.2010.03.003. [DOI] [PubMed] [Google Scholar]
  38. Meng S, Yang Y, and Lewis R, 2024. Subjective versus objective: divergency between subjective walkability and walk score during the COVID-19 pandemic. Transportation research record: Journal of the transportation research board, 2678 (12), 527–537. doi: 10.1177/03611981231165023. [DOI] [Google Scholar]
  39. Mirabal NR, 2009. Geographies of displacement: Latina/os, oral history, and the politics of gentrification in San Francisco’s Mission District. The public historian, 31 (2), 7–31. doi: 10.1525/tph.2009.31.2.7. [DOI] [PubMed] [Google Scholar]
  40. Nguyen N, et al. , 2023. Changes in perceived harms of tobacco and cannabis and their correlations with use: a panel study of young adults 2014–2020. Addictive behaviors, 144, 107758. doi: 10.1016/j.addbeh.2023.107758. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. O’Hern PJ, 1973. Reclaiming the urban environment: the San Francisco urban design plan. Ecology law quarterly, 3, 535–595. [Google Scholar]
  42. Oreskovic NM, et al. , 2014. Attributes of form in the built environment that influence perceived walkability. Journal of architectural and planning research, 31 (3), 218–232. [PMC free article] [PubMed] [Google Scholar]
  43. Osei Owusu R and Rigolon A, 2025. What has contributed to green space inequities in U.S. Cities? A narrative review. Journal of planning literature, 40 (1), 59–76. doi: 10.1177/08854122241269981. [DOI] [Google Scholar]
  44. Reuben A, et al. , 2020. Association of neighborhood disadvantage in childhood with DNA methylation in young adulthood. JAMA network open, 3 (6), e206095. doi: 10.1001/jamanetworkopen.2020.6095. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Riley T, Schleimer JP, and Jahn JL, 2024. Organized abandonment under racial capitalism: measuring accountable actors of structural racism for public health research and action. Social science and medicine, 343, 116576. doi: 10.1016/j.socscimed.2024.116576. [DOI] [PubMed] [Google Scholar]
  46. Rollings KA, et al. , 2024. Built environment and public health: more than 20 years of progress. American journal of public health, 114 (1), 27–33. doi: 10.2105/AJPH.2023.307451. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Saelens BE, et al. , 2003. Neighborhood-based differences in physical activity: an environment scale evaluation. American journal of public health, 93 (9), 1552–1558. doi: 10.2105/ajph.93.9.1552. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Sallis JF, et al. , 2011. Income disparities in perceived neighborhood built and social environment attributes. Health & place, 17 (6), 1274–1283. doi: 10.1016/j.healthplace.2011.02.006. [DOI] [PubMed] [Google Scholar]
  49. Sallis JF, et al. , 2012. Role of built environments in physical activity, obesity, and cardiovascular disease. Circulation, 125 (5), 729–737. doi: 10.1161/CIRCULATIONAHA.110.969022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Sampson RJ and Raudenbush S, 1999. Systematic social observation of public spaces: a new look at disorder in urban neighborhoods. The American journal of sociology, 105 (3), 603–651. doi: 10.1086/210356. [DOI] [Google Scholar]
  51. Scott J, et al. , 2020. Structural racism in the built environment: segregation and the overconcentration of alcohol outlets. Health & place, 64, 102385. doi: 10.1016/j.healthplace.2020.102385. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Shareck M, et al. , 2023. Measuring walkability and bikeability for health equity and intervention research: a scoping review. Cities & health, 7 (6), 1108–1117. doi: 10.1080/23748834.2023.2260133. [DOI] [Google Scholar]
  53. Shashank A and Schuurman N, 2019. Unpacking walkability indices and their inherent assumptions. Health & place, 55, 145–154. doi: 10.1016/j.healthplace.2018.12.005. [DOI] [PubMed] [Google Scholar]
  54. Sokol R, et al. , 2017. Variability in self-rated health trajectories from adolescence to young adulthood by demographic factors. Preventive medicine, 105, 73–76. doi: 10.1016/j.ypmed.2017.08.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Stronegger WJ, Titze S, and Oja P, 2010. Perceived characteristics of the neighborhood and its association with physical activity behavior and self-rated health. Health & place, 16 (4), 736–743. doi: 10.1016/j.healthplace.2010.03.005. [DOI] [PubMed] [Google Scholar]
  56. Suarez-Balcazar Y, et al. , 2020. Walkability safety and walkability participation: a health concern. Health education and behavior, 47 (3), 430–438. doi: 10.1177/1090198120903256. [DOI] [PubMed] [Google Scholar]
  57. Sugiyama T, et al. , 2008. Associations of neighbourhood greenness with physical and mental health: do walking, social coherence and local social interaction explain the relationships? Journal of epidemiology & community health, 62 (5), e9. doi: 10.1136/jech.2007.064287. [DOI] [PubMed] [Google Scholar]
  58. Tabatabaie S, Litt JS, and Muller BHF, 2023. Sidewalks, trees and shade matter: a visual landscape assessment approach to understanding people’s preferences for walking. Urban forestry and urban greening, 84, 127931. doi: 10.1016/j.ufug.2023.127931. [DOI] [Google Scholar]
  59. Van Dyck D, et al. , 2011. Do adults like living in high-walkable neighborhoods? Associations of walkability parameters with neighborhood satisfaction and possible mediators. Health & place, 17 (4), 971–977. doi: 10.1016/j.healthplace.2011.04.001. [DOI] [PubMed] [Google Scholar]
  60. Venerandi A, et al. , 2024. Walkability indices—the state of the art and future directions: a systematic review. Sustainability, 16 (16), 6730. doi: 10.3390/su16166730. [DOI] [Google Scholar]
  61. Vie TL, et al. , 2019. Self-rated health (SRH) in young people and causes of death and mortality in young adulthood. A prospective registry-based Norwegian HUNT-study. SSM - Population health, 7, 100364. doi: 10.1016/j.ssmph.2019.100364. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Wali B and Frank LD, 2024. Redefining walkability to capture safety: investing in pedestrian, bike, and street level design features to make it safe to walk and bike. Transportation research part A: Policy and practice, 181, 103968. doi: 10.1016/j.tra.2024.103968. [DOI] [Google Scholar]
  63. Wang H and Yang Y, 2019. Neighbourhood walkability: a review and bibliometric analysis. Cities, 93, 43–61. doi: 10.1016/j.cities.2019.04.015. [DOI] [Google Scholar]
  64. Whitfield GP, et al. , 2018. Racial and ethnic differences in perceived safety barriers to walking, United States National health interview survey – 2015. Preventive medicine, 114, 57–63. doi: 10.1016/j.ypmed.2018.06.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Wood L, Frank LD, and Giles-Corti B, 2010. Sense of Community and its relationship with walking and neighborhood design. Social science and medicine, 70 (9), 1381–1390. doi: 10.1016/j.socscimed.2010.01.021. [DOI] [PubMed] [Google Scholar]

RESOURCES