Alternative-specific constant equal to nest in nested logit

[Ellen Sterk]

Dear all,

I am using a nested logit model to analyze a discrete choice experiment. I have one nest with three alternatives and one nest with only the baseline alternative. I have included an alternative-specific constant (ASC) to determine respondents' general propensity to choose an alternative over the baseline. That means that my ASC is basically equal to my nests (1 for the three alternatives, 0 for the baseline). Is that ok to do or is that problematic?

Thank you in advance!
Ellen

[Hong Il Yoo]

Ellen Sterk: As usual it depends on what choice situations you're modelling and what modelling assumptions you'd like to maintain. For instance, suppose that your first nest includes three unlabelled policy interventions and degenerate nest describes the status quo. Then it makes sense to use the same intercept for all three alternatives in the first nest since there's no reason (other than experimentally assigned attributes) to expect people to prefer one intervention over others. My work with Jose Oviedo follows this approach: https://doi.org/10.1007/s10640-016-0058-7.

[Ellen Sterk]

Hong Il Yoo, thanks a lot for your quick response! That's exactly the setting I have (with products instead of policies), so that is very helpful. Thanks!