On Nesting Location Choice Models Correctly: A Reply
As shown by Guimaraes, Figueiredo and Woodward (2003), a particular class of conditional logit models yield identical parameter estimates to a Poisson count data model. In Schmidheiny and Brülhart (2011), we have pointed out that the conditional logit model and the Poisson model can be seen as polar cases of a continuum of intermediate cases which emerge from a random utility nested logit model. Herger and McCorriston (2013) have proposed an alternative nested logit model. In this note, we show that their paper misrepresents our model, that their alternative specification does not in fact nest the Poisson model, and that their analysis contains a number of formal errors.
• The nested logit model in Schmidheiny and Brülhart (2011) is a random utility model
that nests the conditional logit and Poisson models.
• The nested logit model in Herger and McCorriston (2013) is a restricted version of
Schmidheiny and Brülhart (2011) that does not nest the Poisson model.
• The proposed estimation strategy in Herger and McCorriston (2013) is incorrect.