Customer Taste With Mixed Logit

28 Pages 6903 Words

ily be used instead. For convenience, we call the approach “maximum
likelihood with conditioning of individual tastes,” or ML/COIT.
In the empirical application, we investigate residential customers’ choice of energy
supplier and estimate the value that these customers place on various attributes of
4
suppliers’ offers, such as the contract length and the type of pricing (e.g., time-of-day
rates that apply a different price for electricity consumed at different times of the day and
seasonal rates that charge different prices in different seasons.) The data consist of
surveyed customers’ responses to conjoint-type experiments. Each surveyed customer
was presented with a series of choice situations in which hypothetical offers of several
suppliers were described and respondents were asked to identify which offer the customer
would choose. We estimated a mixed logit on these data (excluding the last choice, which
we retain for use in assessing the predictive ability of the models.) We calculated each
customer’s conditional...