choice-metrics.com

Dear all,
Our DCE has two generic alts. (two plans of conservation of marine habitat, each characterized by the level of bio-diversity and size that will be retained for each plan and the cost vehicle) and a an optout option. The attributes attached to this option were fixed (zero biodiversity, zero size etc). A fix prior was assigned in Ngene to the optout in the design.
Analyzing the data in Stata's mixed logit, we found that the ASC for the optout is negative.
What will be the interpretation of the division of the optout param. by the (minus) price coefficient? since it is a negative value, is this a sort of WTA for the lost of marine habitat?
Any advice will be helpful!
Rotem

The interpretation of the ASC depends on the utility specification. In particular, whether some of the attributes were dummy or effects coded. If so, the value of the ASC depends on which attribute level was chosen as the reference level, and can switch from positive to negative depending on the reference level. If no dummy/effects coded attributes exist, then the value of the ASC is directly related to the label of the alternative. So if the ASC for the optout would be negative, then putting the ASC in the two generic alternatives and setting the optout to zero would yield the same but positive ASC, and would mean the positive utility attached to a conservation plan above no conservation plan. Using the price coefficient, this utility can be converted in money. Whether this can be interpreted as WTA I am not sure, I have not done such analysis previously myself, but maybe it can be.

Maybe other choice modellers have done this. Or otherwise you can ask in another forum on choice analysis, e.g. the forum for Apollo.

Michiel

Thank you Michiel, indeed no dummy/effects coded attributes exist, and, ASC was attached only to the optout alt. Assigning WTA interpretation was based on intuition, but it really sounds odd.. I'll try to ask in Apollo forum.