The current study examined interactions between perceived safety and built environment variables in explaining physical activity in samples of younger adults and older adults selected from the same regions and from neighborhoods that varied by walkability and income. The results did not support the expected interaction effects between neighborhood environment and safety variables in either sample, particularly for older adults. There were only 5 significant interactions out of 36 models, and only two of the patterns of association confirmed hypotheses that physical activity would be highest when built environments were favorable and perceived safety was high. The observed pattern of results is interpreted as not supporting perceived safety related to crime, traffic, or pedestrian infrastructure as moderating the relation of built environment variables to physical activity. Thus, the nature of the few significant interactions is not interpreted.
Present findings regarding main effects of perceived safety variables were inconsistent, which has been the case in prior studies. Six of 18 models with the NQLS sample of younger adults had significant safety main effects: 4 involved pedestrian safety and 2 involved traffic safety. Only 3 of 18 models with the SNQLS sample had significant main effects, two involving pedestrian safety and one involving crime safety. The pedestrian safety scale was the most frequently significant as a main effect, as it was related to 5 of the 6 outcomes in at least one model. Across the two age groups, it was related to total MVPA minutes and walking for leisure in several different models. This scale assessed built environment attributes that could protect pedestrians from traffic, such as design of intersections, presence of crosswalks, and presence and quality of sidewalks. The relatively consistent support for pedestrian safety main effects may be because the scale assessed perceptions of specific elements in the built environment (e.g., qualities of street crossings), rather than more subjective concerns about crime or volume of traffic. These positive main effects also suggest that neighborhood pedestrian safety may be one of the more important factors in people’s leisure walking and overall physical activity. Further, pedestrian safety factors can be modified, and the present findings suggest that improving such features can positively impact physical activity and walking. There were only two main effects of traffic safety, so improving measures and testing more complex models may be needed to advance evidence for this variable. The crime scale that dealt with concerns about personal safety had no significant main effect in any of the models. Thus, direct associations between perceived crime safety and physical activity were not supported, consistent with most previous literature .
The mixed main effects and null interaction effects suggest that one or more of the following may be true: the current measures of perceived safety (in this case, from the NEWS) lack sensitivity to detect these relationships, the current outcomes (accelerometer-measured MVPA, and self-reported walking for leisure and transportation scales from the IPAQ and CHAMPS ) lack specificity, perceived safety variables are not associated with physical activity, or the links between perceived safety and physical activity are even more complex than could be assessed with these interactions.
The measures of crime, traffic, and pedestrian safety used here may not be sufficiently valid or sensitive to perceptions of safety, warranting better measures. The high mean scores and somewhat small standard deviations suggest a lack of variability and potential ceiling effect, suggesting it may be necessary to develop improved measures or design studies to purposefully select participants with wide variation in perceived safety to ensure hypotheses can be adequately evaluated. The limited variation in safety scores is somewhat surprising because the samples were selected to represent diverse socioeconomic status, and safety variables were documented to differ significantly by neighborhood income . It may also be that the cross-sectional design of the current study is a less sensitive way to uncover these relationships, as opposed to prospective designs. An even better approach would be to conduct quasi-experimental evaluations of efforts to reduce crime, enhance traffic safety, or improve the pedestrian environment.
There are some clues in the criminology literature that could lead to improved measurement and models. First, this literature distinguishes between two related concepts: fear of crime and an assessment of one’s own personal risk of victimization. Fear of crime refers to emotional reactions where perceived risk of victimization is a cognitive judgment of risk . Future research may parse apart these distinctions to assess whether one or both are related to outdoor physical activity behaviors. Further, having witnessed or been the victim of crime can heighten perception of crime , so adding such historical variables as a covariate or third variable in the interactions could be informative. Protective strategies to manage perceived crime and traffic safety, such as avoiding “dangerous” places or routes, traveling with a companion, or carrying a cell phone, could also affect associations with physical activity, so future studies could include such variables in analyses.
The physical activity outcomes examined in this study may lack sufficient specificity to illuminate connections to perceived safety. The amount of physical activity that participants reported on the IPAQ (NQLS) or CHAMPS (SNQLS) or that was demonstrated via use of accelerometers did not necessarily occur in the participants’ neighborhoods. There was a mismatch in locations between the non-specific physical activity measures and neighborhood-specific NEWS safety items. Location-specific physical activity outcomes may help to elucidate these relationships.
Research has pointed to the potential negative effects of safety perceptions on physical activity, particularly for older adults who tend to be more fearful and less active overall [12, 35]. However, both the sample of younger adults and the sample of older adults in the present study had similarly high perceptions of safety from crime and related personal dangers, indicating that crime safety may not be as much of a driving factor in older adults’ low levels of physical activity as has been proposed [36, 37].
This was one of the first studies to use an ecological model to examine built environment by safety interactions in explaining physical activity in seniors and younger adults . The present study employed parallel analyses across two large samples using similar methods, which were also strengths. The weaknesses of the present study were that it relied on cross-sectional data, limiting the conclusions that can be drawn, and that the results have limited generalizability due to sampling bias (high percentages in both samples were white and college-educated). Though the age groupings used in the present study differentiating the younger sample from the older sample are fairly standard, the younger adult sample had a very wide age range (20–65 years), so it is possible there are age-related differences within the younger adult sample that were not revealed in present analyses. The IPAQ has been shown to be valid for total MVPA, but validity of component scores, such as walking, has not been demonstrated . The same is true of the CHAMPS. It is not known from these data where the participants’ MVPA, leisure, or transportation walking actually took place. To the extent the activities occurred outside of the neighborhood, any impact of the safety variables would likely be obscured. The analytic approach of separate models for each interaction was based on a desire to be as sensitive as possible to detecting significant interactions, but this method raises the likelihood of type 1 error. Given that the findings were generally null, type 1 error was not a problem.