The relationship between urban road design, perceived walkability, and walking behavior

The measurement of perceived walkability (PW) and its relationship with walking has received substantial attention over the past years, especially in the Global North. However, research has largely focused on perceived walkability at the neighborhood level, often overlooking perceived walkability of individual streets. Additionally, while many studies have examined the association from perceived walkability to walking behavior, little is known on whether walking behavior also affects perceived walkability. This paper presents and applies a new measure, the Street-level Perceived Walkability Rating Scale (SPaRS), to assess perceived walkability at the street level across five streets with distinct designs in Addis Ababa: three upgraded and two non-upgraded streets. Walking behavior was measured using self-reported walking frequency (WF). The association between PW and WF was examined using three structural equation models (SEMs) that compared alternative specifications of PW and assessed whether the observed associations were consistent with reciprocal relationship. Descriptive analyses of the SPaRS measure showed consistently higher PW scores on upgraded streets than on non-upgraded ones. The SEM analyses showed that the specification of PW influence its observed association with walking behavior. In the first model in which PW is specified as seven first-order dimensions, none of the dimensions was significantly associated with WF. In the second model in which PW is specified as a second-order latent construct reflected by the seven dimensions, PW was significantly associated with WF (β = 0.31, p <0.001). In the third model with a reversed association, WF was also significantly associated with PW (β = 0.07, p = 0.013). These findings suggest that PW is more appropriately represented as an overarching multidimensional construct than as a set of separate dimensions when examining its association with walking behavior. The results also suggest the possibility of a reciprocal relationship between PW and WF. However, the cross-sectional design of the study limits any definitive conclusion on this reciprocal relationship. Longitudinal studies across diverse contexts are needed to determine the existence of this relationship.