Sample Size Requirements For Measuring a Change in Behaviour

Abstract

Before and after surveys are designed to detect a change in travel-behaviour following an intervention policy, such as a travel-modification program. Longitudinal panel surveys are the preferred method for detecting such changes, because the variance of the difference between the before and after surveys is substantially reduced, enabling changes to be detected with smaller sample sizes than if a repeated crosssectional survey is used. A key issue concerns the size of sample required to be able to generalise the Panel results to the population; that is to state, with 95% confidence that if there is a ?% change in behaviour for the sample, there is a ?%± e% change in the behaviour of the population, where e is the sampling error. In this paper we present the rationale for an alternative formulation and demonstrate its applicability both hypothetically and then empirically using data from the Puget Sound Transportation Panel. The results have important ramifications both for those implementing future behaviour change programs and those interpreting the results reported in previous studies.