In willingness-to-pay (WTP) studies, stated preference (SP) data are analyzed by developing econometric models. However, the variation of WTP across different model specifications has not been studied adequately. This paper reports a comparison of various econometric models and estimated WTP values with reference to a case study of rural feeder service to bus stop. Econometric models included Multinomial Logit (MNL), Heteroskedastic Extreme Value (HEV), Nested Logit (NL), Covariance Heterogeneity Nested Logit (CHNL), and Random Parameter Logit (RPL). Various econometric models are found to be consistent and acceptable in terms of signs of coefficient estimates, statistical significance of coefficient estimates and overall goodness of fit. The WTP values are found to vary across different model specifications, although the variations are not consistent and predominant for all attributes. In the present case study, the effect of model specification is found predominant on WTP for a reduction in access walking distance. Also, RPL models accounting for deterministic heterogeneity are found superior to other econometric models such as MNL, HEV, NL, CHNL and RPL models accounting for only random heterogeneity. The work highlights the importance of model specification in WTP studies.
willingness to pay; RPL model; rural feeder service; econometric models