Modeling a Share or Proportion with Logit or Tobit: The Effect of Outcommuting on Retail Sales Leakages

In this study the effects of commuting and demographic variables on the amount and distribution of outshopping were modeled using household-level survey data in which the proportion of expenditures within specific categories of goods were reported across neighboring retail market areas. The effects on the propensity to shop outside the core study area were estimated using the two-limit Tobit and logit models. Influences on the relative distribution of that outshopping were modeled by multinomial logit. The multinomial logit and Tobit models were shown to produce similar estimates, with empirical results indicating that retail sales leakages are increased for outcommuters for certain types of goods.