Conditional Logit, IIA, and Alternatives for Estimating Models of Interstate Migration

Many researchers have used the conditional logit model to examine migration. One common objection to this model is that it carries the independence from irrelevant alternatives (IIA) assumption, which may be too restrictive. This study compares the conditional logit with models that partially relax (nested logit) or fully relax (mixed logit) the IIA assumption. We will begin to learn whether assuming IIA holds poses serious estimation problems for migration modeling. Given the substantial computational cost of the more complex models, a finding that a well-specified, but computationally much simpler, conditional logit model may suffice would be useful.