STATED PREFERENCE AND THE ECOLOGICAL FALLACY

Stated Preference (SP) data are normally analysed using techniques of Discrete Choice analysis. However, they differ from classical Revealed Preference data in that a single respondent contributes more than one observation. This problem of "repeated measurements" has been widely recognised, but generally ignored in practice. It has usually been assumed that the consequences of ignoring the problem are not serious, at least as far as the coefficient estimates are concerned, though clearly the standard errors are likely to be under-estimated. Recent research appears to confirm the above and recommends straightforward techniques for correcting the standard errors. Although such correction is probably sufficient in many cases, this paper demonstrates that in other circumstances, quite commonly encountered, the coefficients themselves as obtained by standard estimation approaches may be seriously biased. It is shown that for mode choice SP experiments (also called "between-mode"), the error structure assumed in standard software packages fails to account for individual differences in the modal constant. The results may be dramatic: for example, obtaining a significant coefficient of the wrong sign. Moreover, the outcome is entirely incidental and depends on the chance correlation between the explanatory variables presented in the experiment and the individual's modal propensity. The paper suggests a straightforward way of removing the worst kind of bias. Although this may not be feasible with large samples, it can certainly be used with a sub-sample to detect whether the bias is likely to be present. For the covering abstract, see IRRD E101013.