Food environments are widely believed to be a driver of increasing obesity rates . Within developed countries such as Australia, Canada, UK, Europe and the US, unhealthy food environments in poorer and minority neighborhoods have also been blamed, in part, for higher obesity rates among poor and minority populations compared to richer, non-minority counterparts [2–4]. However, empirical evidence on how food environments actually affect dietary behaviors and health outcomes such as obesity rates has been inconclusive and limited in scope [5–8].
Despite the difficulty in establishing a clear causal link between food environments and obesity, there is still much interest in policy and planning interventions to change the obesogenic nature of urban environments, particularly for vulnerable groups such as the low-income and children [9–11]. For instance, in 2009, South Korea implemented ‘Green Food Zones’ 200 meters around schools, where sales of unhealthy foods were restricted . In December 2017, the London Mayor announced a similar initiative to ban new fast food outlets within 400m of schools .
Such policies that regulate food environments around schools and other locations must designate specific distances in order to operate. Their effectiveness are thus dependent on the accurate identification and definition of the physical extent of one’s food environment and on a proper understanding of how far people are willing to travel for food. However, current literature on food environments does not sufficiently support the accurate identification and definition of a bounded ‘food environment’ because of four key limitations: the inconsistent definitions of ‘food environments’ across studies; a lack of justification for geographic extent and potential spatial misclassification; assumptions of a uniform neighborhood effect across different groups and urban environments; and the exclusion of non-residential environments.
Firstly, food environment boundaries have been inconsistently defined across various research studies. Researchers commonly use administrative units such as census tracts or postal sectors, or buffers of varying radii around address points or centroids of administrative units [6, 14].
Secondly, few food environment studies explicitly justify their choice of buffer distance [6, 15], which may reflect a lack of literature on food-related travel behavior  that food environment researchers can draw upon. Furthermore, the few papers on this topic have found that actual trips to restaurants and grocery-stores tended to be longer than commonly assumed buffers in neighborhood environment studies [16–18], which in turn suggests that much of existing neighborhood effects research into ‘food environments’ might be threatened by spatial misclassification, where chosen analytic boundaries fail to capture the true food-related travel behavior.
Thirdly, most food environment studies currently assume a uniform neighborhood effect across all groups, despite the fact that the ‘relevant contextual unit’ is likely to vary depending on the population group, location, and the type of built environment . Resolving the ‘uncertain geographic context problem’  in food environment research will thus require studies to account for people’s travel patterns, taste preferences, cost considerations or social/cultural norms [5, 6]. Further motivating the need for deeper exploration into the heterogeneity in food environments’ extents are findings from activity-based travel research, which shows that trip frequencies, trip lengths, mode choice are dependent on both built form and individual socioeconomic characteristics [21–23].
To date, however, few studies explicitly investigate whether the association between food environments and obesity risks might vary by demographic characteristics, such as income or race/ethnicity. Of the empirical studies that have explored the intersection between income, race/ethnicity and environment, these have found that associations between BMI, food environments  and neighborhood walkability  differed by SES and race/ethnicity. There is also relatively little research on how built environmental characteristics, such as the availability of public transit, might affect one’s food-related travel patterns, and thus the extent of one’s food environment [16, 25]. The few available studies have found some significant associations. For instance, Thornton et al.(2017) tracked 56 participants’ food purchases over two weeks, in Melbourne, Australia, and found that younger age groups tended to travel further for food purchases, as did participants living in richer neighborhoods with lower access to supermarkets . Kerr et al. (2012) found, from a travel survey of 4800 Atlanta residents, that lowest income, non-White participants, those without a degree, and those travelling from less accessible environments travelled further for food . Zenk et al. (2011) found socioeconomic differences in estimated activity spaces of 131 participants in Detroit, U.S: Participants without cars or were not in the labor force had smaller overall activity spaces than counterparts with comparative higher socioeconomic status, which in turn suggests that they might have a smaller food environment.
Fourthly, people often move, shop and make food purchases outside of their home environments [8, 27].For instance, Thornton et al.(2017) found many food purchases occur outside participants’ residential neighbourhood . Similarly, a 2010 survey of 50 individuals in Philadelphia found many visited stores beyond traditionally defined residential neighborhoods . However, much of food environment research still focuses on the residential environment. Little attention is paid to food environments around workplaces, schools, and other ‘anchor points’ [5, 7, 27, 29].‘Anchor points’ are places with important material and symbolic meaning for the individuals, around which they organize their daily activities . Transport scholars theorize that one’s use of time and space is conditioned by one’s basic anchor points, such as home, work and school, since the time available for visiting other places for other activities is bounded by departure from and return to these bases. Within these spatio-temporal constraints, Within these spatio-temporal constraints, individuals make locational and scheduling choices to balance time spent on an activity, such as eating a meal, with travel time to a sufficiently attractive option, such as a good restaurant [31, 32]. If people’s food-seeking behaviors differ by anchor points types, then definitions of food environments should then be contingent on the types of land uses, such as residential, entertainment, office, or retail, within these areas. This study thus examines different types of respondent-defined ‘anchor points’, and whether these affect how far and long people travel for food.
To date, most food environment studies have been conducted within the US, Australia, and New Zealand [15, 29], though a growing number of studies have also be carried out in East Asian cities (e.g. [33–38]). There are significant differences between Western and Asian cities in terms of food environment patterns [33, 39], population density and public transport provision , as well as social and cultural norms around food. More research focused on Asian cities and how people there interact with their food environments is needed to facilitate context-specific policy formulation within the region. The relative lack of research into food environments in Asia is a particularly pressing concern, given that Asia is home to 54% of the world’s urban population, as of 2018, compared to 7% in North America, 13% in Europe, and 0.7% in Oceania . The region has also seen a rapid increase in Type 2 diabetes and obesity prevalence [41, 42].
This study is based in Singapore, a highly urbanized, densely-populated city-state in South East Asia. While behaviors will necessarily vary by city, there are three reasons why results from Singapore might provide generalizable insights for many other cities in the Asia Pacific region: Singapore has a reputation for being a food paradise where eating outside of the home is a common practice, with about 60% of Singaporeans eating out at least four times a week . The propensity to eat out is shared elsewhere in Asia, in Hong Kong, Taiwan and Malaysia . Secondly, Singapore is a multi-ethnic, multi-religion country with three primary ethnic groups: Chinese, Malay, and Indian. There is thus significant overlap between Singapore’s population composition with its regional neighbors Malaysia, Indonesia, Hong Kong, China, India and others. Thirdly, like many of the major cities in Japan, China, South Korea and South East Asia (e.g. Tokyo, Hongkong, Bangkok, Kuala Lumpur), Singapore has an extensive public transport network, high population density and high built density. Given the similarities in population, food cultures, foodscapes and built form between Singapore and other cities in the region, this study provides a useful case study for researchers, health professionals and planners interested in Asian urban food environments.
This study combines insights from an ecological model of food-related behavior, which asserts that built environment, social, and individual factors interact to affect eating patterns [1, 45] with models of travel behavior which postulate that travel patterns are dependent on built form, locational characteristics, and individual attributes. It examines how individual and built environment characteristics may be associated with how far (distance) and how long (time) people travel to food venues. In doing so, we seek to contribute empirically-informed, theoretically-derived estimates of food environment extents by population group and built environment type, that provides a starting point for further policy-oriented research in Asian cities.