This study examined the correlation between the level of walkability and noncommuting trips by conducting multilevel logistic regression analysis with the two odds of active transportation (i.e., walking and cycling) for leisure and shopping purposes. It is an expansion of a prior study that revealed a positive correlation between neighborhood walkability and active commuting in Seoul [52
], since it is likely that noncommuting trips are more influenced by the neighborhood environment than commuting trips when people walk or cycle [30
]. As a result, we empirically discovered that the walkability level of Seoul’s neighborhoods was positively correlated with the probability of active transportation for shopping purposes while showing no statistical correlation in leisure purposes.
Discussions based on the findings are presented as follows. First, the correlation between the Walkability Score and the nonmotorized trips varied in the leisure and shopping models. In the shopping models, the Walkability Score was positively correlated with the odds of nonmotorized trips (Model S–1: OR = 1.015, 95% CI = 1.003–1.026; Model S–2: OR = 1.018, 95% CI = 1.005–1.031). However, there was no significant correlation between the Walkability Score and the odds of nonmotorized trips for leisure purposes. This differs from the findings of previous literature that showed that the odds ratios of walking for the Pedestrian Index of the Environment (PIE) were 1.05 for leisure and 1.04 for shopping purposes [21
]. This difference is interpreted based on the indicators included in the walkability index. While the PIE considers the 5Ds of urban form, including density, diversity, design, destination accessibility, and distance to transit [21
], the Walkability Score mainly considers destination accessibility [25
]. As mentioned earlier, the Walkability Score was assessed based on the accessibility of amenities including grocery stores, restaurants, shopping centers [61
]. Therefore, it can be demonstrated that the Walkability Score is an index based on the accessibility of amenities that are more suitable for shopping than for leisure purposes. Second, other neighborhood-level variables had no significant correlation with the odds of nonmotorized trips both in the leisure and shopping models. From Models L–2 and S–2, no variable was significant among land use mix and sidewalk length. This result is different from those of previous studies in which active transport (e.g., walking) and/or physical activity was positively correlated with land use mix [31
]. Meanwhile, some studies showed mixed-use development may not provide a positive effect on walking and cycling in some high-density cities because these cities would reduce the chance for people walking or cycling as they can move easily to another area by public transportation [5
]. This might be applied to Seoul because it is one of the high-density cities [76
] and has good public transportation in its neighborhoods. Regarding sidewalk length, this analysis showed a similar result to the previous study that estimated the association between sidewalk length and walking for different trip purposes [66
]. It found that neighborhood-based walking for transportation had a positive association with sidewalk length but had no association in recreation walking, which was in line with our research on active transportation for leisure and shopping purposes. Third, the types of individual variables that were significantly associated with the odds of nonmotorized trips differ in the leisure and shopping models. Specifically, older age was positively correlated with the odds of nonmotorized trips for leisure purposes, while female gender had a positive correlation with the odds of nonmotorized trips for shopping purposes. These results can be used to develop various urban policies. For example, older people were more likely to have higher odds of nonmotorized trips for leisure purposes in this study. This finding presents an important issue for further research on age-friendly community design. Environmental factors that promote leisure walking (e.g., greener landscape, well-designed street furniture, wide pedestrian roads, safer pathway) in the development of age-friendly communities may be considered. In addition, a higher Walk Score was correlated to a lower risk of obesity for females in urban areas [43
]. As noted from the result, the female gender tends to walk and cycle for shopping purposes; therefore, planning a wider choice of commercial facilities in a walkable and bikeable distance in urban areas or creating a more walkable environment for shopping could induce more physical activity for females. Finally, according to ICC values, the level of walkability of the neighborhood was an important factor in influencing individuals’ odds of nonmotorized trips for shopping purposes. From the results of the multilevel analyses, the proportion accounted for by the Walkability Score in the odds of nonmotorized trips for shopping purposes was about 4.5%. Although the ICC value was high at 9.3% in the leisure model (Model L–1), the Walkability Score was insignificant; therefore, it was not discussed. Additionally, compared to the results of the previous study by Kim et al. [52
] (ICC was 2.1% for active commuting model), active trips for shopping had a higher ICC value (4.5%) in this study. This indicates that the Walkability Score is a walkability index that responds more sensitively and effectively to the odds of nonmotorized trips for shopping purposes when compared to commuting purposes.
This study has several limitations; directions for future studies to address them follow. At the outset, even though this study found a significant correlation between the walkability and odds of active transportation for shopping purposes, it showed no correlation between them for leisure models. This may be due to the Walkability Score’s sensitive nature toward walking and cycling for shopping purposes. For further direction of the study, the Walkability Score can be developed and customized based on trip purposes. For example, it will be possible to develop a walkability index, “Walkability Score for Leisure,” where people can search for environmental conditions for leisure purposes in their daily lives by considering the characteristics and contents of each facility that can attract and affect leisure trips. Second, this study examined the correlation of neighborhood walkability with active transportation in Seoul. The city of Seoul is promoting policies regarding walking and cycling, and the neighborhood environment is being developed toward a more walkable environment [51
]. Future research will be able to experiment with other cities in Korea, which could help identify practical policies in urban design and transportation planning that can be used across the country. Third, since we focused on active transportation for noncommuting trips, this study combined walking and cycling into nonmotorized transportation. However, walkers and cyclists can have different characteristics of shopping and leisure trips. In a further study, an analysis comparing walking and cycling behavior on each trip’s purposes could be discussed. Fourth, the mean ages of the participants for leisure and shopping purposes were relatively high, corresponding to middle-aged and older adults. This can reflect that retired people may have more time for recreational and shopping activities, and thus they are likely to spend their time on the active mode of the trip. Nevertheless, the older age of respondents remains a limitation. Future studies need to properly extract samples and use them for analysis so that respondents are not biased at a specific age group. Finally, this study is a cross-sectional design that cannot identify causal relationship over time. Cross-sectional analysis cannot identify whether individuals walk frequently because of its walkable environmental conditions or whether the individuals who walk a lot more choose to live in walkable neighborhoods. For further study, a longitudinal analysis can be performed to examine the causality between walkable environment and active transportation.
Despite these limitations, this study is significant in several aspects. First, since the correlation of environmental walkability and active trips was analyzed using a multilevel logistic model, factors from both individual and neighborhood levels were taken into account. Although the ICC values were generally low at 4.5% in Model S–1 and 4.7% in Model S–2, it is suggested that multilevel modeling is required when dealing with nested data [73
]. That is, because individual-level variables were nested within the neighborhood, the multilevel analysis allowed an examination of the impact of multilevel factors on the dependent variable (active transportation). Second, there are few studies that comprehensively examined active transportation for both leisure and shopping purposes. In this study, we compared the probability of active transportation (i.e., walking and cycling) for each travel purpose (i.e., leisure and shopping purposes) depending on the Walkability Score. Based on our findings, a tailored policy guideline can be provided based on trip purposes. Finally, we found that the Walkability Score, measured in Seoul, had a significant validity for examining the odds of active transportation for shopping purposes. However, this study demonstrated that more variables should be considered when assessing active transportation, especially for leisure purposes, as they affect not just accessibility to local destinations but also other urban form factors such as density, diversity, and design, and the quality of urban environments. From the findings of this study, policymakers and researchers from the urban and transportation planning and public health fields can obtain a comprehensive understanding of enhancing active travel in leisure and shopping, based on the primary concerns.