Theoretical model of US walking duration and validation using 2003 American time use survey data

Transdisciplinary research across the fields of health, transportation planning, and community design share common interests in walking; a behavior promoted by current public health agencies. Purpose: To use the 2003 American Time Use Survey (ATUS) to test theories of walking duration (2,444 observations) and purpose (e.g., for exercise, transportation, etc.) for Americans. Methods: The ATUS time-defined activities on a single day in a nationally representative sample (n=1,700 interviews) are used to develop empirical models of walking propensity (i.e., the likelihood of walking) for exercise and other purposes. Trade-offs between walking purposes are explicitly modeled, as are socio-demographic covariates. Zero-altered negative binomial (ZNB) regression models were fit to the data , and account for both the time allocation trade-offs between walking purposes and the presence of ‘true-zero' walkers (i.e., non-walkers) in the sample. Results: The probability of engaging in zero walking is influenced by age, marital status, geographic region, access to a motor vehicle, and household income. These factors have varying effects depending on walking purpose. Various factors influence walking duration (in minutes), including geographic region, the presence of children in the household, ethnicity, marital status, and income. These factors influence walking duration in different magnitudes depending on walking purpose. Conclusions: Complex time-allocation trade-offs between walking purposes occur across demographic ‘markets'; non-walkers vary across walking purposes; and a range of regional and demographic attributes influence average walking duration. Additional survey work is needed to capture some of the theoretical influences on walking propensity, despite the good fit of the empirical model.