An inexact science: Accounting for measurement error and downward bias in mode and location choice models

Using commuting data for Brisbane, Australia, we find that accounting for measurement error in travel times causes the magnitude of parameters in mode and location choice models to increase approximately three-fold and 30–40%, respectively. Errors appear to be somewhat systematic, with travel times being underestimated for short journeys and vice versa for long journeys—especially by public transport. We find similar results when we use alternative transport cost measures and independent commuting data from London. Our findings are likely to have important implications for transport and land use policy as well as the many types of economic models in which travel times—and transport costs, more generally—occupy a central role.