Gravity and Welfare in Discrete Choice Models with Incomplete Consideration

In spatial choices such as commuting, migration, sourcing and residential location, agents often choose from a large set of options while only considering a subset of alternatives. 

A fundamental identification problem arises: observed choices may reflect both preferences and which options are considered. We show that in discrete choice models with many options, this distinction has limited welfare implications. As the number of independent consideration draws grows, stochastic consideration becomes equivalent to a deterministic shift in utility. 

We show that the expected size of the consideration set is a sufficient statistic for the convergence of welfare and choice probabilities between models with and without consideration frictions. We apply these results to workplace choices of garment workers in urban Bangladesh, using a weekly panel of over 2,200 workers surveyed between 2018 and 2022 in major garment districts. We measure consideration sets directly and calibrate the relative importance of consideration frictions and preferences in determining the gravity of commuting flows. We show that workers operate in extremely localized labor markets where the distinction between preferences and consideration is most consequential.