Presentation Title

Socioeconomic Disparities in Sustainable Travel: Examining the Equity of Regional Bikeability

Faculty Mentor

Georgiana Bostean

Start Date

17-11-2018 10:00 AM

End Date

17-11-2018 10:15 AM

Location

C158

Session

Oral 2

Type of Presentation

Oral Talk

Subject Area

behavioral_social_sciences

Abstract

It is increasingly evident that active modes of transportation such as biking have extensive personal and social benefits (see Giles-Corti et al. 2010, for review). The extent to which neighborhoods are bikeable is influenced by the built environment, which varies by area socioeconomic status (Fuller and Winters 2017; Gordon-Larsen et al. 2006), creating socioeconomic disparities in access to active transportation. This study examined whether there are socioeconomic disparities in census tract bikeability in Orange County, California (N = 561). We created an index of bikeability (range 1-10) using Geographic Information System (GIS) methods and data on bike route density, route connectivity, separation of bike facilities from streets, topography, and density of common destinations. Linear regressions predicted census tract bikeability using income and education (American Community Survey 2012-2016, 5-year estimates). There was no evidence of multicollinearity based on VIFs, which were below 3. Preliminary results show that tracts with lower education have lower bikeability (β = -0.54, p < .001), but higher median income is associated with lower bikeability when education is controlled (β = -0.55, p < .001). Moran’s I (0.44, p < .001) provides evidence of spatial autocorrelation in the regression residuals; therefore, we will also explore spatial regression models to predict bikeability. These preliminary findings suggest that investment is needed in areas of lower socioeconomic status to improve the equity in access to bikeable routes.

Summary of research results to be presented

We compared bivariate models predicting bikeability score using education and income, with a fully adjusted model including income and education, and found that the adjusted model has better fit, based on lower AIC value and higher R2. In the adjusted model, both income and education are statistically significant predictors of bikeability. A one standard deviation increase in the percentage of tract with less than a college education is associated with a .53 standard deviation decrease in predicted bikeability score (p < .0001), while a one standard deviation increase in median income is associated with a .55 standard deviation decrease in bikeability. Based on a Moran’s I of 0.44 (p < .001), there is evidence of spatial autocorrelation in the residuals from this regression, which violates the OLS assumption of uncorrelated errors. Therefore, we also explored a Geographically Weighted Regression (GWR) to predict bikeability. The GWR model explains 42% of variability in bikeability and scored a lower AIC value, but residuals may still be slightly spatially autocorrelated.

Subsequent analysis of best model fit will take place over the next two weeks, and preliminary results (indicated as such in the abstract) will be finalized before the presentation date.

This document is currently not available here.

Share

COinS
 
Nov 17th, 10:00 AM Nov 17th, 10:15 AM

Socioeconomic Disparities in Sustainable Travel: Examining the Equity of Regional Bikeability

C158

It is increasingly evident that active modes of transportation such as biking have extensive personal and social benefits (see Giles-Corti et al. 2010, for review). The extent to which neighborhoods are bikeable is influenced by the built environment, which varies by area socioeconomic status (Fuller and Winters 2017; Gordon-Larsen et al. 2006), creating socioeconomic disparities in access to active transportation. This study examined whether there are socioeconomic disparities in census tract bikeability in Orange County, California (N = 561). We created an index of bikeability (range 1-10) using Geographic Information System (GIS) methods and data on bike route density, route connectivity, separation of bike facilities from streets, topography, and density of common destinations. Linear regressions predicted census tract bikeability using income and education (American Community Survey 2012-2016, 5-year estimates). There was no evidence of multicollinearity based on VIFs, which were below 3. Preliminary results show that tracts with lower education have lower bikeability (β = -0.54, p < .001), but higher median income is associated with lower bikeability when education is controlled (β = -0.55, p < .001). Moran’s I (0.44, p < .001) provides evidence of spatial autocorrelation in the regression residuals; therefore, we will also explore spatial regression models to predict bikeability. These preliminary findings suggest that investment is needed in areas of lower socioeconomic status to improve the equity in access to bikeable routes.