Exploiting Rank Ordered Choice Set Data within the Stochastic Utility Model

A Rank‐ordered Logit Model With Unobserved Heterogeneity in Ranking Capabilities

“…However, it has already been noted by Chapman and Staelin (1982) that for the less preferred items, this assumption does not always hold. One of the possible reasons for this is that the respondent perhaps has no experience with some of the items, and hence is not able to indicate the proper ranking order.…”

“…If one assumes k to be the same for the whole sample, this last term can be ignored in the estimation as it becomes a constant in the log likelihood, see Chapman and Staelin (1982) and Hausman and Ruud (1987). These two papers propose estimating different ROL models, each using a different number of ranks, that is a different value for k. Then they have different methods of choosing a model from this set.…”

“…Under the utility assumption in (1) and the assumption of the extreme value distribution, we obtain the rank-ordered logit [ROL] model, see Beggs et al (1981) and Chapman and Staelin (1982).…”

“…Secondly, the respondent may not be able to distinguish between his less-preferred alternatives. In any case, straightforwardly using reported rankings may lead to a substantial bias in the parameter estimates in the ROL model, see Chapman and Staelin (1982). To solve this issue, Chapman and Staelin (1982) suggest to only use the first few ranks in the estimation.…”

“…In any case, straightforwardly using reported rankings may lead to a substantial bias in the parameter estimates in the ROL model, see Chapman and Staelin (1982). To solve this issue, Chapman and Staelin (1982) suggest to only use the first few ranks in the estimation. They consider several rules to determine the appropriate number of ranks to use, in their words "the explosion depth".…”