Empirical analysis of walkability using data from the Netherlands

•A data-driven approach present to measure walkability based on regression analysis.•We use a large and comprehensive nationwide database of neighborhoods.•The study compares the resulting regression model to existing walkability indices.•The results indicate which factors are key for creating walkable neighborhoods. In urban planning and transportation research, indices have been developed to measure the walkability of neighborhoods based on land-use and transportation characteristics. Despite the extensive work on the development of walkability measures, the empirical validation of the measures has received only limited attention. The aim of this study is to test assumptions made in existing walkability indices. Based on a national neighborhood data set combined with three years of national travel survey data from the Netherlands, the relationships between physical neighborhood characteristics and walking frequency are identified through regression analysis. In the regression model, distance to supermarket, number of daily goods stores within 1 km, number of cafeterias within 1 km, total inland water, land use for residential buildings, and high urban density were found to be significant. Our analysis indicates that existing indices only partly capture variation in walkability. We find that mismatches emerge on the level of both the selection and weighting of variables. Based on the results we identify ways to improve existing indices to measure walkability.