A simulation-based method for testing an abnormally high incidence of specific choice patterns in discrete choice data

This paper introduces a simulation-based method designed to detect unusually high frequencies of specific choice patterns in discrete choice data. Its empirical size and power are then evaluated through a series of simulation exercises. The results show that the empirical size is consistently lower than the assumed theoretical significance level, and that the test power approaches 100 % when the observed choice pattern appears approximately 2.5 times more frequently than expected. The proposed test is applied to a real data set obtained by means of a discrete choice experiment focused on air pollution in South Delhi, India. • Non-trading, lexicographic or inconsistent behaviour can cause a high frequency of a specific choice pattern. • A simulation based procedure for the testing of an abnormally high frequency of a specific choice pattern is proposed. • The empirical size and power of the test is analysed by simulation exercises. • The proposed test is applied on a data set focused on air pollution in South Delhi.