Next Article in Journal
Determinants of High-Speed Train Demand: Insights from the Jakarta—Bandung Corridor in Indonesia
Previous Article in Journal
Changing Lifestyles in Highly Urbanized Regions of Russia: Short- and Longer-Term Effects of COVID Restrictions
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Geographic Scale Matters in Analyzing the Effects of the Built Environment on Choice of Travel Modes: A Case Study of Grocery Shopping Trips in Salt Lake County, USA

by
Ensheng Dong
1,
Felix Haifeng Liao
2,* and
Hejun Kang
3
1
RAND Corporation, Arlington, VA 22202, USA
2
Department of Earth and Spatial Sciences, University of Idaho, Moscow, ID 83844, USA
3
Transit Services Division, Fairfax County Department of Transportation, Fairfax, VA 22033, USA
*
Author to whom correspondence should be addressed.
Urban Sci. 2025, 9(8), 307; https://doi.org/10.3390/urbansci9080307
Submission received: 13 June 2025 / Revised: 16 July 2025 / Accepted: 25 July 2025 / Published: 5 August 2025

Abstract

Compared to commuting, grocery shopping trips, despite their profound implications for mixed land use and transportation planning, have received limited attention in travel behavior research. Drawing upon a travel diary survey conducted in a fast-growing metropolitan region of the United States, i.e., Salt Lake County, UT, this research investigated a variety of influential factors affecting mode choices associated with grocery shopping. We analyze how built environment (BE) characteristics, measured at seven spatial scales or different ways of aggregating spatial data—including straight-line buffers, network buffers, and census units—affect travel mode decisions. Key predictors of choosing walking, biking, or transit over driving include age, household size, vehicle ownership, income, land use mix, street density, and distance to the central business district (CBD). Notably, the influence of BE factors on mode choice is sensitive to different spatial aggregation methods and locations of origins and destinations. The straight-line buffer was a good indicator for the influence of store sales amount on mode choices; the network buffer was more suitable for the household built environment factors, whereas the measurement at the census block and block group levels was more effective for store-area characteristics. These findings underscore the importance of considering both the spatial analysis method and the location (home vs. store) when modeling non-work travel. A multi-scalar approach can enhance the accuracy of travel demand models and inform more effective land use and transportation planning strategies.

1. Introduction

Understanding trip mode choices is essential for transportation planning, particularly in forecasting travel demand. In the context of sustainable and low-carbon transportation, it is vital to understand mode choices better, such as riding transit, walking, or biking, as these modes can offer significant benefits for community well-being and the environment [1]. To date, trip mode choice studies have primarily focused on the work trip [2,3], using a variety of different modeling techniques [4,5]. Non-work trips, such as customers’ shopping trips [6,7,8] and children’s school trips [9,10], have recently become another major concern in this research domain. Specifically, grocery shopping trips have attracted interdisciplinary attention in the fields of geography, transportation, urban planning, economics, and marketing. Most studies, however, are more concerned about shopping trip travel frequency [11,12,13,14]; the number of studies about grocery shopping trip travel mode choices and their research methods remains limited [15]. Additionally, previous studies also identified a variety of key factors that are associated with shopping travel mode choices, which could be summarized into (1) individual socioeconomic characteristics, (2) households’ shopping preferences, and (3) built environment and land-use configurations near both households and stores.
Individual backgrounds, including age, gender, race, education, personal income, employment status, etc., were found to have an impact on people’s shopping trip travel behavior. Su & Bell mentioned that older people would like to choose a cheaper travel mode [16]. Based on 1000 surveys in Austin, Texas, Handy found that younger women were more likely to walk, which showed that age as well as gender could be important correlated factors [17]. Guo et al. reported that people older than 65 years were more likely to drive for shopping [18]. A more recent record from the National Household Travel Survey (NHTS) highlighted that women, overall, took more trips than men for family errands, including shopping. Other researchers also found that being a female played an important role in deciding whether to use auto or non-auto travel modes [3,17,19]. Also, some black families may have a lower household income than the non-black families, which in turn may influence the number of vehicles per household, shopping frequency, and travel mode [11,20].
In addition to individual sociodemographic characteristics, the household-level socioeconomic profiles would also provide a more comprehensive way to explain the travel behavior. The number of household members (family size), house ownership, household income, car ownership, employment status for adult members, etc., are main factors found to be important in the literature on travel behavior [21]. In addition, the household head’s age and the family’s participation in food-related programs, the Food Stamp Program, for instance, are some other detailed considerations of household characteristics. Larger families need to buy larger quantities of products due to their higher level of consumption, which, given fixed storage capacity in their homes, necessitates a larger number of driving trips. Clifton [22] found that households with a lower income preferred driving, either alone or carpooling, in Austin, Texas. Handy’s research showed people who owned a car preferred to go shopping by driving, because of the convenience and the time duration [17]. In Europe, people prefer to walk for shopping when cars are not available [23]. A renter is more likely to spend less on food storage, so household ownership can also influence the frequency of travel, as well as the travel mode [24]. The number of vehicles for each adult in one household is also strongly related to car usage. If there are more cars per adult in one household, the family members will be more likely to drive to the store. Scheiner & Holz-Rau found that driving was the primary mode in Berlin when the distance was more than 670 m (about 0.4 mi) from the household to the reported store [25]. However, most studies aimed to discuss shopping frequency, departure time choice, or store choice instead of shopping travel mode choices.
Built environment, or the physical neighborhood structure, can significantly shape the mode choice [26,27,28]. In general, suburban residents tend to use private vehicles, while urban residents tend to favor public transit, biking, and walking [27]. Frank & Pivo [29] found that people living in neighborhoods with a higher land use mix were less likely to choose driving alone for shopping. Scheiner & Holz-Rau [30] concluded that a higher land use mix usually meant a good transit system, which may lower the possibility that people will drive their cars. Several studies also proved that the higher intensity of land use meant the surrounding neighborhood was more friendly to pedestrians, and the public transit service was also better [26,29,31,32,33,34]. Guo et al. mentioned that the higher the density of bikeways, the more likely people are to choose to go shopping by bicycle or on foot [18]. Other factors, such as the employment rate, the population density rate, the number of convenience stores, quick-service restaurants, coffee shops, and bus stations, and the number of traffic signals within a certain distance, are all detailed characteristics regarding the built environment [35].
Geographical scale or different spatial aggregation methods could affect the quantitative measure of the land use mix or configuration around a neighborhood, and help in understanding the relationship between neighborhood context and people’s travel behavior [36,37]. However, there is no fixed definition of the size of a neighborhood [38,39]. Different spatial partitions were adopted in previous studies, varying from census block, census block group, census tract, and ZIP code to buffers with different sizes [23,31,39]. For instance, Frank et al. used a 1 km distance from each respondent’s household, while Yamada et al. focused on three different spatial scales: census tract, block group, and 1 km street-network buffer [40]. No matter whether the geographical scale is defined by the census administration boundaries or the buffer distance, they were used to measure the walkability or driving possibility. In one of these works on walkability and obesity in Salt Lake County, Utah, Fan et al. suggest that different focuses should adopt different scales [41]. For instance, when analyzing the large grocery store, the ZIP code zone or a buffer with more than a 1000 m radius is a better choice; when studying limited services, the census tract level or larger would be more effective; for full-service restaurants, the census tract or block group level is considered better; and if several factors are investigated together, the census tract would be a better choice than census block group or ZIP code.
Our research aimed to delve more into understanding these factors with a particular focus on how built environment factors based on measures at different geographic scales affect travel mode choices. More specifically, we attempted to answer the following questions: Why do people prefer a certain travel mode, driving alone, carpooling, or other mode, in their shopping trips? At which spatial scale, like block, block group, or census tract, do built environment and store characteristics play a role in shaping mode choices?

2. Materials and Methods

2.1. Conceptual Framework

Based on individual and household level data, this research grouped factors affecting travel mode choices in shopping trips into four different domains (Figure 1): socioeconomic background; accessibility (e.g., distances to CBD, bus, and transit stations); built environment centered on a person’s home; and store. Multi-scalar analysis was applied to characterize built environment (BE) factors at different geographic scales or ways in which spatial or BE data are aggregated, including a half-mile buffer circle, one-mile buffer circle, half-mile network buffer, one-mile network buffer, census block level, census block group level, and census tract level.

2.2. Data and Variable Selection

The analysis was based on 1294 shopping trips from 1091 households’ responses filtered from the 2009 Utah Travel Survey. We focused on the household information and the adult travel diary in Salt Lake County, Utah, whose public transportation system is more developed as compared to other areas in Utah. Specifically, survey respondents answered various questions regarding their travel behavior on a typical day. The survey recorded trips from approximately 2800 households in three days, which covered different trip purposes and modes, including commuting to work locations and traveling for grocery shopping, walking, or other travel behavior [42]. Notably, people in some households did not report their shopping trips, and all these trips were home-based, internal (origin)–internal (destination) shopping trips in Salt Lake County, Utah, which means both houses and stores were within Salt Lake County boundaries. As demonstrated in Table 1, among the 1294 shopping trips, 61% were carried out by females and 39% were carried out by males. With respect to the education level, 56.18% of the respondents attained bachelor’s degrees or above, and Utah ranked 4th in the nation in terms of persons 25 years and older who obtained bachelor’s degrees (30.8%) [20]. The percentage of white citizens was also slightly higher than that of the state average (89.2%). The average household annual income fell within the USD 50,000 to USD 99,999 category (Table 1). Hence, based on the individual information, most respondents in this study were employed white young females with a good educational background and high household income levels.
Table 1 also shows that the average number of adults in sample households was 2.10; the average number of children per household was 0.83. Compared with the data from the Census Bureau, respondents had a smaller size and more children per family. Almost 80% of families had already owned or bought the house, and more than half of the participating families had lived in the current residence for at least 10 years; 22.64% had lived there more than 20 years (Table 1).
The dependent variable was a respondent’s travel mode choice in shopping trips, including driving alone, carpooling, and other modes (bicycling, walking, taking public transit, etc.), rather than auto/non-auto classification as in other studies [15]. They were coded as 0, 1, and 2 in the following analyses. Additionally, both home locations and the locations of stores were geocoded from 2009 addresses. Of the 1294 shopping trips, more than 90% of shopping trips were completed by either driving alone (799 trips) or carpooling (420 trips), and 5.8% of trips (75 trips) used other transportation modes: public transit (16 trips), bicycle (15 trips), walking (41 trips), and other modes (3 trips). Specifically, Kernel density estimation, or KDE, was used to smooth point events (i.e., trip-level origin locations and their mode choices) across Salt Lake County. The resulting density surface was normalized and can be interpreted as the relative distribution of events rather than absolute counts. As shown in Figure 2, driving alone and carpooling remain the major travel mode choices for shopping; in contrast, taking public transit or other travel modes were mostly found among trips near downtown areas or public transit (Figure 2).
Based on the aforementioned framework, variables characterizing the respondents’ sociodemographic status included individuals’ gender, age, employment status, education, race, household income, house ownership, the number of vehicles, bicycles, children, and adults in one household, and the number of years lived at the current residence. As for the distance factors, employing the OD Cost Matrix, the Network Analyst Toolbar estimated the distance to the downtown area (CBD distance) and the home–store network distance (OD distance). The distance from the nearest public transit station to each respondent’s household (transit distance) was also calculated by taking advantage of the OD Closest Facility function in ArcMap 10.x.
The land use mix index (LUM) is an entropy index that measures the balance and basic characteristics of land zones [33], such as residential, agricultural, commercial, educational, etc. In our study, Frank’s index was adopted to calculate the land use mix index (LUM), reflecting the diversity in different geographic scales [43]. If the value of LUM is close to 1, then there are more different types of land, or maximally heterogeneous land use; when the LUM value is close to 0, there are not too many land use types, or homogeneous land use. Based on the property-type codes provided by the Salt Lake County assessor’s office, this study adopted six categories: single-family residential, multi-family residential, commercial, office, educational/institutional, and recreational.
Density factors were also introduced. We studied the street density, residential density, job density, and population density, as well as densities of points of interest, such as the number of traditional restaurants, bus stations, cafés, quick-service restaurants, convenience stores, liquor stores, and road intersections within different geographic scales. These data were also collected from the InfoUSA website (https://www.dataaxleusa.com/, accessed on 25 May 2025), the American Factfinder website, and the Utah Automated Geographic Reference Center (AGRC).
The store characteristics, such as the sales amount, the food pricing, the number of employees, the size, and the number of parking lots, tend to determine people’s shopping involvement, like the shopping frequency and expenditure, as well as the travel mode choice [12,44,45]. The InfoUSA website provided the location of stores and their annual sales amount, and the number of employees.
Thus, to further investigate the influence of these factors at different geographic scales, seven spatial scales surrounding the household and the store were adopted: a half-mile buffer circle (0.5 mi circle), one-mile buffer circle (1 mi circle), half-mile network buffer (0.5 mi network) one-mile network buffer (1 mi network), census block (Block), census block group (BG), and census tract (CT). In general, the census block is the smallest unit, followed by a half-mile buffer, census block group, one-mile buffer, and the census tract. The size of network buffers varied according to the patterns of roads and streets.

2.3. Model Selection

The most frequently used travel mode choice models are the binary or binomial model (BL), the multinomial logit model (MNL), and the nested logit model (NL), since the travel mode is usually regarded as categorical data [46,47]. Since the dependent variable had three or more unordered levels (including driving alone, carpooling, and other modes), the multinomial logit model (MNL) was a better choice [48].

3. Results and Discussion

3.1. Descriptive Analysis

As shown in Figure 3, male respondents preferred driving alone and other transportation modes as compared to mode choices of female respondents: 65.6% and 7.3% of male respondents chose to drive alone and used other modes, respectively, while the percentages of those for female respondents were 59.2% and 4.8%, respectively. Young people (18–24 and 25–34 years old) were more likely to choose to carpool, and people aged 25–34 and 65–74 were more likely to choose other modes, compared to other age categories (Figure 3). The disadvantaged financial status of the young and the health pursuits of senior citizens may lead to their preference for choosing to drive alone or to carpool. We also examined the correlations between household income, family size, family composition, and several travel habits. The lower the household annual income was, the higher the possibility that the shopper would walk, bike, or take public transit to the store, while the higher the income was, the higher the possibility that the shopper would drive alone for shopping. These findings agreed with a case study of shopping trips in King County, WA, showing that income per household was positively related to driving for grocery shopping. People in households whose number of vehicles was more than one preferred to drive alone for shopping, and people who had no vehicle in the household were more likely to choose other transportation modes. The number of bicycles in one household was strongly related to its members’ travel mode choices: households with more than two bicycles were more likely to ride a bike for shopping than other groups.

3.2. Results of MNL Regression Analysis

Results for a set of MNL models are listed in Table 2, with different BE factors at a variety of spatial scales. Overall, the MNL models were estimated to investigate whether demographic, proximity, and built environment factors influence shopping trip travel mode choices, by using driving alone as the reference group. The model’s goodness of fit is assessed using the pseudo-R2 statistic and the likelihood ratio test. Most pseudo-R2 values were slightly above or below 0.2, which is relatively low but consistent with typical values reported in previous shopping travel behavior studies, sufficiently capturing the effect of these factors on the likelihood of choosing carpooling or alternative travel modes [27,35]. Results in general showed that individual and household socioeconomic factors, such as gender, age, status as a student, annual income, household ownership, vehicle ownership, and household composition, were the most significant factors in deciding the travel mode, either carpooling or other mode (riding public transit, walking, or biking, Table 2). The transit distance and CBD distance were shown to be significant when not combined with socioeconomic factors. The built environment factors around households and stores, such as LUM, population density, job density, residential density, street density, the number of points of interest, and sales amount, were positively associated with the shopping trip travel mode choices, although their significance could not be applied to all geographic scales (Table 2).
Household annual income under USD 35,000 and the number of adults and children were consistently positive predictors for carpooling, while the number of bicycles and the families who rent were consistently better predictors for riding public transit, biking, and walking (or other). So, consistent with the results of descriptive analysis presented in Figure 3, the household economic condition did influence people’s travel mode choices, especially for those whose annual income was under USD 35,000 and who preferred not driving alone for shopping, and these people would share a vehicle or take other travel modes. In addition, the larger the family size (the number of adults and children), the higher the possibility for this family to share vehicles for shopping. In addition, if there were more vehicles in the household, they were less likely to choose to carpool or take other modes; driving alone would be the best option for them. On the contrary, families having more bicycles, which may suggest they have a healthier life attitude, chose travel modes other than carpooling, and this was consistent with the findings in the descriptive results. If the household was renting a house or an apartment, its members were more likely to choose other travel modes; however, if living in their current locations for a longer period, people preferred not to choose other travel modes for shopping. Hence, we found economic and lifestyle considerations were potential factors behind preferences regarding shopping trip travel modes.
The MNL results also implied that a longer distance to the CBD could lead people to carpool more and take other travel modes less (Table 2). In addition, the long distance to the closest public transit stations (both bus stops and light rail stations) prevented people from choosing other travel modes. Our results also showed that the distances to the aggregated shopping area (CBD) and the nearest transit stations were the most important considerations for shoppers. More detailed discussion of modeling results is as follows.
Individual and household background had a strong impact on mode choices: females aged between 25 and 34, the number of adults, and the number of children in one household were positively related to the likelihood of carpooling. These three factors were highly significant (p < 0.01), with coefficients of 1.051, 0.494, and 0.462 (Table 2), and they were based on the model using 0.5-mile buffer circles. Status as a student, the number of bicycles, and being a renter enabled people to choose other transportation modes such as biking and walking. The number of vehicles in one household was negatively related to selecting both carpool and other modes. The respondents’ employment status, education level, race, household income, and the number of years living in the current residence were not significant in both the model for carpooling and the model for other modes. The number of vehicles and bicycles in the household and families who rent were top predictors for choosing biking, walking, and riding public transit, whose coefficients were −1.202, 0.404, and 1.431. Compared with carpooling, respondents having more vehicles in the household were almost six times less likely to choose other modes to go shopping.
We further assessed three network distance factors: the distance between the household and the reported shopping destination (origin–destination, OD dist.), the distance between the household and the closest public transit (transit dist.), and the distance between the household and the central business district (CBD dist.). The CBD distance is a strong predictor for carpooling (p < 0.01), and being closer to CBD and bus stations also enhances the possibility of riding public transit, biking, and walking.

3.3. Home and Store Built Environment Factors at Different Scales

Thirteen household built environment factors were included in the MNL models, including land use mix index (LUM), population density, job density, residential density, street density, number of traditional restaurants, bus stops, cafés, quick-service restaurants, convenience stores, liquor stores, traffic nodes, and store sales amount. Results showed that LUM, job density, and the number of restaurants around the household were negatively related to riding public transit, walking, and biking across different geographic scales, while population density, the number of liquor stores, and sales amount around the household were positively related to the probability of choosing other travel modes (Table 2). Furthermore, geographic scales did affect the model results. Some BE variables remained significant, but not for all geographic scales. For instance, status as a student was positively related to the use of other modes in models using the half-mile buffer circle, one-mile network buffer area, the census block, and the block group (Table 2).
Table 3 further summarizes the correlation between the built environment factors at different geographic scales and the possibility of choosing either carpooling or walking, biking, and riding public transit (others). In general, the store built environment had a greater effect on the travel mode choice than the household built environment, based on the number of significant factors at the 0.05 level. Most significant household built environment factors were within the census level spatial aggregates, which means the community boundaries (census level) had a larger impact than the distance consideration (buffer level) around the household, whereas both census administrative boundaries and the buffer zones were important for factors around the store. The census block group was the level whose numbers of significant factors were the largest in both the model for carpooling and the model for others (e.g., riding public transit, walking, and biking): residential density, street density, sales amount, and the number of restaurants, quick services, traffic nodes, and liquor stores were all significant factors. The significant factors of the census block level included the street density around both the household and the store, as well as the job density, the number of convenience stores, and the sales amount around the store. All of these factors were significant in the model for carpooling, either positively or negatively. The census tract, however, was the only geographical scale that demonstrated the significance of the population density.
In comparison with BE factors based on network buffers, straight-line buffers had a better performance in measuring the significant factor in the model for others (e.g., riding public transit, walking, and biking), which was not expected. Around the household, both the half-mile straight-line buffer and the half-mile network buffer showed that LUM was negatively related to other mode choices. Around the store, job density, the number of cafés, and the sales amount were shown to have significance in the model for others in both half-mile and one-mile straight-line buffers; meanwhile, only the one-mile network buffer had a similar effect. Nevertheless, the network buffers still showed that the residential density, the street density, and the number of convenience stores were significant in the model.
Several factors exerted their effects on the travel mode choice at multiple geographic scales. These factors included LUM, the number of restaurants, and the number of convenience stores around the household for either the model for carpooling or the model for others (Table 2 and Table 3). The job density around the store was a continuously significant factor positively related to the probability of choosing other travel modes. The street density around the store was a continuously significant factor that was positively related to choosing to carpool. The number of cafés and sales amount were strong factors in the model for others, and the number of quick-service and liquor stores was a strong predictor in the model for carpooling.

4. Discussion

This study reveals important insights into how geographic scale and built environment (BE) context influence travel mode choices for grocery shopping trips, a non-work travel behavior that has received less scholarly attention. By testing seven different ways of aggregating spatial or built environment data—two Euclidean buffers (0.5- and 1-mile), two network buffers (0.5- and 1-mile), and three administrative units (census block, block group, and tract)—we show that the impact of BE variables varies significantly by scale and location (home vs. store). First, census units would be more effective as compared to distance buffers in detecting the effect of BE on travel mode choices. For example, modeling results based on census block and block group level measures captured the highest number of significant built environment predictors, especially for stores’ surroundings. Street density near stores, store sales volume, and the number of quick-service or liquor stores were consistently strong predictors when they were aggregated into census block group levels [35]. Among these census units, in general, finer-scale measures would improve the predictive power of all models. For instance, the model using a set of BE factors at the census block level had a larger number of significant factors than that using the BE factors at the block group or the census tract levels (Table 3). Although other studies suggested that the diversity factor or the land use mix index is a significant built environment factor in the travel behavior analysis [32], this factor in our research only demonstrated its significance at the census block level, and it was insignificant at any other geographic scale.
Second, different from recent work on commuting travel [49], particularly for active transportation modes (e.g., walking or biking), our results show that straight-line buffers performed equally well, or in some cases even better, than network buffers in predicting walking, biking, or riding transit to nearby stores. This result indicates that in a relatively high-density urban environment, grocery shopping trips are often short and irregular, and network buffers based on a single pair of origin and destination may not be able to capture the trip chain or multiple errands included in one shopping trip [50,51].
Third, our results highlight that built environment factors centered on stores had greater predictive power, which indicates that destinations’ characteristics matter significantly for non-work trips. More work should be done to investigate the effect of sales volume and the density of café and retail stores, which have been found in our analysis to be strong and consistent predictors for choosing carpooling or other travel modes.
Finally, the impact of socioeconomic factors remained evident across all models. Lower-income households and those with more bicycles were more likely to walk, bike, and ride public transit, while car ownership strongly discouraged the use of these modes.

5. Conclusions

This study conducted a multiscale analysis of travel mode choices for urban shopping trips, through the case of shopping trips in Salt Lake County, Utah. Results confirmed that individual sociodemographic characteristics and built environment were key factors behind respondents’ mode choices. This study further investigated how geographic scale matters when examining the association between people’s shopping trip mode choice and built environment or BE factors. The result showed that, in general, the smaller the census level the study chose, the more BE factors were significant. This finding reflected the controversial debate over whether the land use mix affects travel mode choices, and their associations might have been inconclusive, partly due to the scale of BE factors. For instance, some past studies did find that the land use mix was not a significant factor in the mode choices of walking and biking [18]. These results echo the recent literature: there is no universal “best” geographic scale, and mode choices for shopping trips, like commuting, are deeply scale-dependent. Researchers and planners should adopt a multi-scale approach to reveal the nuanced and often context-specific influences of the built environment on travel behavior [49].
This study further contributes to the broader field of travel demand estimation and transportation planning, and findings suggest that multiple measures of the built environment should be considered when estimating potential changes in mode choices for shopping travel. Practitioners in the areas of land use and transportation planning should take advantage of such nuanced evidence to design policies that could support multimodal access to shopping destinations, which could further assist in reducing car dependence and creating more walkable and transit-friendly commercial areas.
Lastly, for future work, some other interesting topics deserve more attention, such as the influence of customers’ subjective preferences and the impact of the current and future city development plan. Other related factors regarding store characteristics, such as store size, number of employees, price level, and mixture of stores in the neighborhood, might be worth researching in future studies.

Author Contributions

Conceptualization, E.D., F.H.L. and H.K.; methodology, E.D. and F.H.L.; software, E.D.; validation, E.D., F.H.L. and H.K.; formal analysis, E.D., F.H.L. and H.K.; investigation, E.D., F.H.L. and H.K.; resources, F.H.L. and H.K.; data curation, E.D. and F.H.L.; writing—original draft preparation, E.D., F.H.L. and H.K.; writing—review and editing, E.D., F.H.L. and H.K.; visualization, E.D.; supervision, F.H.L. and H.K.; project administration, F.H.L. and H.K.; funding acquisition, F.H.L. and H.K. All authors have read and agreed to the published version of the manuscript.

Funding

The authors acknowledge the partial funding support from the US Department of Transportation’s University Transportation Center Program grant #69A3551747110 through the Pacific Northwest Regional University Transportation Center (PacTrans).

Institutional Review Board Statement

This study was determined to be non-human subject research or to be exempt from Institutional Review Board (IRB) review at the University of Idaho, as it involved the secondary analysis of de-identified data from the Utah Household Travel Survey, which was collected by the Utah Department of Transportation and other partner agencies for transportation planning purposes. No personally identifiable information was accessed or used in this research, and the study posed no more than minimal risk to participants.

Informed Consent Statement

Participation in the study was anonymous, and no identifying information was collected. Completion of the questionnaire implied informed consent, which was explicitly stated in the introductory section of the form. The study did not involve the collection of sensitive personal data or images, and participants were not required to sign in with a Google account or provide their email addresses.

Data Availability Statement

Restrictions apply to the availability of these data.

Conflicts of Interest

Author Ensheng Dong was employed by the company RAND Corporation. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of the data; in the writing of the manuscript; or in the decision to publish the results.

References

  1. Lak, A.; Moussavi.A, S.M.R. Nurturing Healthy and Resilient Communities: An in-Depth Bibliometric Analysis of Walkability within the Built Environment. J. Transp. Health 2025, 41, 101981. [Google Scholar] [CrossRef]
  2. Horowitz, J.L. Semiparametric Estimation of a Work-Trip Mode Choice Model. J. Econom. 1993, 58, 49–70. [Google Scholar] [CrossRef]
  3. Bhat, C.R. Analysis of Travel Mode and Departure Time Choice for Urban Shopping Trips. Transp. Res. Part B Methodol. 1998, 32, 361–371. [Google Scholar] [CrossRef]
  4. Elmi, A.; Badoe, D.; Miller, E. Transferability Analysis of Work-Trip-Distribution Models. Transp. Res. Rec. J. Transp. Res. Board 2015, 1676, 169–176. [Google Scholar] [CrossRef]
  5. Liu, G. A Behavioral Model of Work-Trip Mode Choice in Shanghai. China Econ. Rev. 2007, 18, 456–476. [Google Scholar] [CrossRef]
  6. McDonald, N.C. Household Interactions and Children’s School Travel: The Effect of Parental Work Patterns on Walking and Biking to School. J. Transp. Geogr. 2008, 16, 324–331. [Google Scholar] [CrossRef]
  7. Müller, S.; Tscharaktschiew, S.; Haase, K. Travel-to-School Mode Choice Modelling and Patterns of School Choice in Urban Areas. J. Transp. Geogr. 2008, 16, 342–357. [Google Scholar] [CrossRef]
  8. Zhu, X.; Lee, C. Walkability and Safety around Elementary Schools Economic and Ethnic Disparities. Am. J. Prev. Med. 2008, 34, 282–290. [Google Scholar] [CrossRef]
  9. McDonald, N.C.; Deakin, E.; Aalborg, A.E. Influence of the Social Environment on Children’s School Travel. Prev. Med. 2010, 50 (Suppl. S1), S65–S68. [Google Scholar] [CrossRef] [PubMed]
  10. Lee, C.; Zhu, X.; Yoon, J.; Varni, J.W. Beyond Distance: Children’s School Travel Mode Choice. Ann. Behav. Med. A Publ. Soc. Behav. Med. 2013, 45 (Suppl. S1), S55–S67. [Google Scholar] [CrossRef] [PubMed]
  11. Blaylock, J.R. An Economic Model of Grocery Shopping Frequency. Appl. Econ. 1989, 21, 843–852. [Google Scholar] [CrossRef]
  12. Smith, M.F.; Carsky, M.L. Grocery Shopping Behavior Consumers. J. Retail. Consum. Serv. 1996, 3, 73–80. [Google Scholar] [CrossRef]
  13. Wilde, P.; Ranney, C. The Monthly Food Stamp Cycle: Shooping Frequency and Food Intake Decisions in an Endogenous Switching Regression Framework. Am. J. Agric. Econ. 2000, 82, 200–213. [Google Scholar] [CrossRef]
  14. Cao, X.; Mokhtarian, P.L. How Do Individuals Adapt Their Personal Travel? Objective and Subjective Influences on the Consideration of Travel-Related Strategies for San Francisco Bay Area Commuters. Transp. Policy 2005, 12, 291–302. [Google Scholar] [CrossRef]
  15. Jiao, J.; Moudon, A.V.; Drewnowski, A. Grocery Shopping How Individuals and Built Environments Influence Choice of Travel Mode. Transp. Res. Rec. 2011, 2230, 85–95. [Google Scholar] [CrossRef]
  16. Su, F.; Bell, M.G.H. Transport for Older People: Characteristics and Solutions. Res. Transp. Econ. 2009, 25, 46–55. [Google Scholar] [CrossRef]
  17. Handy, S. Non-work travel of women: Patterns, perceptions and preferences. In Proceedings of the Women’s Travel Issues: Proceedings from the Second National Conference, Baltimore, MD, USA, 23–26 October 1996. [Google Scholar]
  18. Guo, J.; Bhat, C.; Copperman, R. Effect of the Built Environment on Motorized and Nonmotorized Trip Making: Substitutive, Complementary, or Synergistic? Transp. Res. Rec. 2007, 2010, 1–11. [Google Scholar] [CrossRef]
  19. Matthies, E.; Kuhn, S.; Klockner, C.A. Travel Mode Choice of Women: The Result of Limitation, Ecological Norm, or Weak Habit? Environ. Behav. 2002, 34, 163–177. [Google Scholar] [CrossRef]
  20. Farber, S.; Bartholomew, K.; Li, X.; Páez, A.; Nurul Habib, K.M. Assessing Social Equity in Distance Based Transit Fares Using a Model of Travel Behavior. Transp. Res. Part A Policy Pract. 2014, 67, 291–303. [Google Scholar] [CrossRef]
  21. Carrasco, N. Deciding Where to Shop: Disaggregate Random Utility Destination Choice Modeling of Grocery Shopping in Canton Zurich. Master’s Thesis, IVT, ETH Zürich, Zürich, Switzerland, 2008. [Google Scholar]
  22. Clifton, K.J. Mobility Strategies and Food Shopping for Low-Income Families: A Case Study. J. Plan. Educ. Res. 2004, 23, 402–413. [Google Scholar] [CrossRef]
  23. Coveney, J.; O’Dwyer, L.A. Effects of Mobility and Location on Food Access. Health Place 2009, 15, 45–55. [Google Scholar] [CrossRef]
  24. Blattberg, R.; Buesing, T.; Peacock, P.; Sen, S. Identifying the Deal Prone Segment. J. Mark. Res. 1978, 15, 369–377. [Google Scholar] [CrossRef]
  25. Scheiner, J.; Holz-Rau, C. Gendered Travel Mode Choice: A Focus on Car Deficient Households. J. Transp. Geogr. 2012, 24, 250–261. [Google Scholar] [CrossRef]
  26. Cervero, R. Built Environments and Mode Choice: Toward a Normative Framework. Transp. Res. Part D Transp. Environ. 2002, 7, 265–284. [Google Scholar] [CrossRef]
  27. Schwanen, T.; Mokhtarian, P.L. What Affects Commute Mode Choice: Neighborhood Physical Structure or Preferences toward Neighborhoods? J. Transp. Geogr. 2005, 13, 83–99. [Google Scholar] [CrossRef]
  28. Ewing, R.; Cervero, R. Travel and the Built Environment. J. Am. Plan. Assoc. 2010, 76, 265–294. [Google Scholar] [CrossRef]
  29. Frank, L.; Pivo, G. Impacts of Mixed Use and Density on Utilization of Three Modes of Travel: Single-Occupant Vehicle, Transit, and Walking. Transp. Res. Rec. 1994, 1466, 44–52. [Google Scholar]
  30. Scheiner, J.; Holz-Rau, C. Travel Mode Choice: Affected by Objective or Subjective Determinants? Transportation 2007, 34, 487–511. [Google Scholar] [CrossRef]
  31. Cervero, R.; Radisch, C. Travel Choices in Pedestrian versus Automobile Oriented Neighborhoods. Transp. Policy 1996, 3, 127–141. [Google Scholar] [CrossRef]
  32. Cervero, R. Mixed Land-Uses and Commuting: Evidence from the American Housing Survey. Transp. Res. Part A Policy Pract. 1996, 30, 361–377. [Google Scholar] [CrossRef]
  33. Cervero, R.; Kockelman, K. Travel Demand and the 3Ds: Density, Diversity, and Design. Transp. Res. Part D Transp. Environ. 1997, 2, 199–219. [Google Scholar] [CrossRef]
  34. Schwanen, T. The Determinants of Shopping Duration on Workdays in The Netherlands. J. Transp. Geogr. 2004, 12, 35–48. [Google Scholar] [CrossRef]
  35. Handy, S.; Boarnet, M.; Ewing, R.; Killingsworth, R. How the Built Environment Affects Physical Activity: Views from Urban Planning. Am. J. Prev. Med. 2002, 23, 64–73. [Google Scholar] [CrossRef]
  36. Duncan, M.J.; Winkler, E.; Sugiyama, T.; Cerin, E.; DuToit, L.; Leslie, E.; Owen, N. Relationships of Land Use Mix with Walking for Transport: Do Land Uses and Geographical Scale Matter? J. Urban Health Bull. N. Y. Acad. Med. 2010, 87, 782–795. [Google Scholar] [CrossRef]
  37. Brownson, R.C.; Hoehner, C.M.; Day, K.; Forsyth, A.; Sallis, J.F. Measuring the Built Environment for Physical Activity: State of the Science. Am. J. Prev. Med. 2009, 36, S99–S123.e12. [Google Scholar] [CrossRef]
  38. Gauvin, L.; Robitaille, E.; Riva, M.; McLaren, L.; Dassa, C.; Potvin, L. Conceptualizing and Operationalizing Neighbourhoods: The Conundrum of Identifying Territorial Units. Can. J. Public Health 2007, 98 (Suppl. S1), S18–S26. [Google Scholar] [CrossRef]
  39. Hess, P.; Moudon, A. Site Design and Pedestrian Travel. Transp. Rec. Rec. 1999, 1674, 9–19. [Google Scholar] [CrossRef]
  40. Yamada, I.; Brown, B.B.; Smith, K.R.; Zick, C.D.; Kowaleski-Jones, L.; Fan, J.X. Mixed Land Use and Obesity: An Empirical Comparison of Alternative Land Use Measures and Geographic Scales. Prof. Geogr. 2012, 64, 157–177. [Google Scholar] [CrossRef]
  41. Fan, J.X.; Hanson, H.A.; Zick, C.D.; Brown, B.B.; Kowaleski-Jones, L.; Smith, K.R. Geographic Scale Matters in Detecting the Relationship between Neighbourhood Food Environments and Obesity Risk: An Analysis of Driver License Records in Salt Lake County, Utah. BMJ Open 2014, 4, e005458. [Google Scholar] [CrossRef] [PubMed]
  42. Wei, Y.D.; Wu, Y.; Xiao, W.; Garcia, I.; Wen, M. Urban Form, Air Pollution, and Walking Behavior: A Study of Salt Lake County, Utah. J. Transp. Health 2023, 33, 101686. [Google Scholar] [CrossRef]
  43. Frank, L.D.; Andresen, M.A.; Schmid, T.L. Obesity Relationships with Community Design, Physical Activity, and Time Spent in Cars. Am. J. Prev. Med. 2004, 27, 87–96. [Google Scholar] [CrossRef]
  44. Putrevu, S.; Ratchford, B.T. A Model of Search Behavior with an Application to Grocery Shopping. J. Retail. 1997, 73, 463–486. [Google Scholar] [CrossRef]
  45. Bawa, K.; Ghosh, A. A Model of Household Grocery Shopping Behavior. Mark. Lett. 1999, 10, 149–160. [Google Scholar] [CrossRef]
  46. Wen, C.-H.; Koppelman, F.S. The Generalized Nested Logit Model. Transp. Res. Part B Methodol. 2001, 35, 627–641. [Google Scholar] [CrossRef]
  47. Recker, W.W.; Kostyniuk, L.P. Factors Influencing Destination Choice for the Urban Grocery Shopping Trip. Transportation 1978, 7, 19–33. [Google Scholar] [CrossRef]
  48. McFadden, D. Quantitative Methods for Analyzing Travel Behavior of Individuals: Some Recent Developments; Routledge: London, UK, 1977. [Google Scholar]
  49. Jian, W.; Liu, X.; Liu, H.; Hu, Y.; Gao, L. The Impacts of the Multiscale Built Environment on Commuting Mode Choice: Spatial Heterogeneity, Moderating Effects, and Implications for Demand Estimation. J. Adv. Transp. 2023, 2023, 9346631. [Google Scholar] [CrossRef]
  50. Wei, Y.D.; Xiao, W.; Wu, Y. Trip Generation, Trip Chains and Polycentric Development in Metropolitan USA: A Case Study of the Wasatch Front Region, Utah. Appl. Geogr. 2021, 133, 102488. [Google Scholar] [CrossRef]
  51. Kim, K.; Kwon, K.; Horner, M.W. Examining the Effects of the Built Environment on Travel Mode Choice across Different Age Groups in Seoul Using a Random Forest Method. Transp. Res. Rec. 2021, 2675, 670–683. [Google Scholar] [CrossRef]
Figure 1. Factors that affect people’s travel mode choices in shopping trips.
Figure 1. Factors that affect people’s travel mode choices in shopping trips.
Urbansci 09 00307 g001
Figure 2. Kernel density of household travel mode distribution.
Figure 2. Kernel density of household travel mode distribution.
Urbansci 09 00307 g002
Figure 3. Shopping trip travel mode choices and respondents’ gender, age, and income.
Figure 3. Shopping trip travel mode choices and respondents’ gender, age, and income.
Urbansci 09 00307 g003
Table 1. Individual and household characteristics.
Table 1. Individual and household characteristics.
VariablesPercentVariablesPercentVariablesPercent
Gender Household Income# of Adults/Household
Male39.33Under USD 35,00016.771 person17.39
Female60.66USD 35,000–USD 49,99912.212 persons65.15
Age USD 50,000–USD 99,99936.553 persons10.82
[18, 24]6.03USD 100,000 or more22.14 persons4.33
[25, 34]21.64Prefer not to answer12.365 persons1.55
[35, 44]18.93Household Life Cycle6+ persons0.77
[45, 54]15.07No children or retirees40.8# of Children/Household
[55, 64]19.78W/children no retirees36.710 child62.21
Above 6418.55With retirees22.491 child13.29
Employment Rent or Own the Home 2 children11.05
Employed full-time37.33Rent18.73 children7.96
Employed part-time11.44Own/Buying79.294 children4.1
Self-employed6.41Other0.935 children1.08
Student5.18Prefer not to answer1.086+ children0.31
Homemaker16.31# of yrs lived at residence # of vehicles/household
Retired18.32Less than 1 year9.660 vehicle1.78
Not currently employed5.021–5 years28.671 vehicle24.5
Education 6–10 years20.022 vehicles49.15
Less than high school1.0811–15 years10.513 vehicles16.62
High school graduate10.1216–20 years8.54 vehicles5.56
Some college20.17More than 20 years22.645 or more2.4
Vocational/Technical3.94Race
Associates degree8.5Hispanic4.25Asian2.86
Bachelor’s degree35.16African American0.62White90.73
Graduate/Post-graduate21.02American Indian or Alaskan Native0.46Other2.4
Table 2. Final results for MNL models based on multi-scalar built environment (BE) factors. (Boldface indicates significance, or p < 0.05, of the test results.)
Table 2. Final results for MNL models based on multi-scalar built environment (BE) factors. (Boldface indicates significance, or p < 0.05, of the test results.)
Straight-LineNetworkCensus Level
0.5 mi Circle1 mi Circle0.5 mi Network1 mi NetworkBlockBGCT
Variables for carpooling
Socioeconomic factorsFemale age [25, 34]1.0511.0021.0781.010.9330.9940.989
Income < USD 35,0000.5610.5790.5780.5440.7110.6260.609
# adults0.4940.4610.5120.4620.5470.550.523
# children0.4620.4720.4610.4610.5040.4720.478
# vehicles−0.236−0.191−0.226−0.219−0.214−0.217−0.228
AccessibilityOD distance0.0160.0090.0160.0150.0270.0210.005
Transit distance−0.0680.000−0.115−0.026−0.155−0.057−0.100
CBD distance0.011−0.0250.0120.0140.0100.0160.000
Household built environmentResidential density0000−0.003−0.0010
Street density0.659−0.6830.0250.0220.1120.2270.139
# convenience stores−0.0680.056−0.0740.114−0.6670.0730
# traffic nodes0.0010.0010.0050.0020.0080.0040.001
Store built environmentJob density−0.0420.065−0.037−0.028−0.066−0.072−0.07
Residential density00000.0020.0010
Street density1.2051.6880.1570.0770.1380.5440.65
# quick services0.0160.033−0.0020.0450.0170.020.053
# convenience stores0.026−0.0630.017−0.0310.6340.0130.062
# liquor stores−0.3010.004−0.2710.055−0.819−0.513−0.319
Sales amount0.027−0.115−0.054−0.2240.7430.1690.206
Variables for the model for others (e.g., riding public transit, biking, and walking)
Socioeconomic factorsFemale age [18, 24]−1.101−1.003−1.058−0.906−0.681−1.632−1.293
Student0.6320.7130.5880.6321.0190.8760.392
Non-white0.9280.9080.8190.7210.9020.6480.374
Income < USD 35,0001.0791.0441.1121.0820.881.0890.959
# adults0.3990.5760.3410.5270.2740.2420.476
# vehicles−1.202−1.336−1.215−1.347−1.23−1.336−1.334
# bicycles0.4040.4840.450.4620.5260.520.506
Rent1.4311.4851.5461.3841.0481.2051.19
Years at current residence−0.295−0.213−0.276−0.209−0.126−0.143−0.141
AccessibilityOD distance0.038−0.0420.0010.007−0.0440.0030.016
Transit distance−0.705−1.044−0.698−0.854−0.304−0.758−0.583
CBD distance0.1050.0450.0990.043−0.0770.0550.051
Household built environmentLUM−3.692−0.839−2.728−1.387−0.086−1.579−0.806
Population density−6.7665.384−0.043−0.0690.6090.0467.068
Job density−0.596−0.388−0.239−0.3−0.0210.01−0.201
# restaurants0.0170.0220.0580.0250.138−0.0850.016
# liquor stores−0.1560.1160.033−0.449−15.2481.1380.17
Sales amount0.4730.0740.8820.0364.090.2660.272
Store built environmentLUM−2.375−0.156−3.036−2.5881.2231.454−2.793
Job density0.7380.9420.2050.566−0.010.1140.339
Residential density0.00100.0030−0.0030.0010
Street density−0.375−3.824−0.855−2.5570.2060.741−0.381
# restaurants0.02−0.0020.028−0.0180.3030.1160.049
# cafés0.7270.3450.0150.450.103−0.4170.058
Sales amount−1.17−0.6730.063−0.587−0.144−1.17−0.655
Table 3. Results of MNL modeling related to the influence of built environments at different scales. (Highlighted cells mean the significance of corresponding factors on a certain scale level. C: significant in the model for carpool; O: significant in the model for others.)
Table 3. Results of MNL modeling related to the influence of built environments at different scales. (Highlighted cells mean the significance of corresponding factors on a certain scale level. C: significant in the model for carpool; O: significant in the model for others.)
VariablesStraight-LineNetwork BufferCensus Level
0.5 mi1 mi0.5 mi1 miBlockBGCT
Household built environmentLUMO O
Population density O
Job density O
Residential density C
Street density C
# restaurant O CO
# bus stop
# cafe
# quick service C
# convenience store C O
# liquor store O
# traffic node CO
Sales amount O
Store built environmentLUM O
Population density
Job densityOO OC
Residential density O
Street densityC OCC
# restaurants OO
# bus stops
# cafeOO O
# quick service C C
# convenience store C
# liquor store CC
# traffic node
Sales amountOO CO
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Dong, E.; Liao, F.H.; Kang, H. Geographic Scale Matters in Analyzing the Effects of the Built Environment on Choice of Travel Modes: A Case Study of Grocery Shopping Trips in Salt Lake County, USA. Urban Sci. 2025, 9, 307. https://doi.org/10.3390/urbansci9080307

AMA Style

Dong E, Liao FH, Kang H. Geographic Scale Matters in Analyzing the Effects of the Built Environment on Choice of Travel Modes: A Case Study of Grocery Shopping Trips in Salt Lake County, USA. Urban Science. 2025; 9(8):307. https://doi.org/10.3390/urbansci9080307

Chicago/Turabian Style

Dong, Ensheng, Felix Haifeng Liao, and Hejun Kang. 2025. "Geographic Scale Matters in Analyzing the Effects of the Built Environment on Choice of Travel Modes: A Case Study of Grocery Shopping Trips in Salt Lake County, USA" Urban Science 9, no. 8: 307. https://doi.org/10.3390/urbansci9080307

APA Style

Dong, E., Liao, F. H., & Kang, H. (2025). Geographic Scale Matters in Analyzing the Effects of the Built Environment on Choice of Travel Modes: A Case Study of Grocery Shopping Trips in Salt Lake County, USA. Urban Science, 9(8), 307. https://doi.org/10.3390/urbansci9080307

Article Metrics

Back to TopTop