How close is the MNL model to a Max-utility-model (first choice model)?

Question

Hello everyone,

the HB utility estimation procedure uses a MNL model for the customers´ choice behavior.
In this regard, I need to know how rational (consistent to utility maximizing behavior) the customers behaved. I understood that this can be derived from the Beta values, because the MNL works either similar to a first choice model (high betas) or to a random choice model (low betas).

Thus, where do I find the beta values and is there a benchmark for "rather rational" behavior in contrast to a beta of 0 (random choice)?

Is there a possibility to use a first choice model during the utility estimation procedure?


Thank you so much for your thoughts!

Regards,
Uwe

Answer 1

I'm unaware of a model that would use first choice during estimation, but MNL does assume that respondents are utility maximizers (with some amount of error).  The HB utilities and the fit statistic, root likelihood, should give you the information you seek about where your respondents' choices fall on the spectrum of deterministic to random.  You will likely see a mix of respondents, some acting deterministically while others answer with varying degrees of lesser consistency

Comments

Dear Keith,

thanks very much for the quick reply.

Is there a paper I can quote on the assumption of utility maximization (plus the error term) immanent in the MNL? That would be very helpful.

Do I get it right, that a RLH close to 1 is considered to indicate deterministic behavior and close to 0 random choice behavior?

---

You are right about you interpretation of RLH.  As for a reference, I suggest the Ben-Akiva book on Discrete Choice Analysis, the original 1974 McFadden article in Frontiers in Econometrics or Ken Train's book, which he makes available for free on his website.