Model-Directed Attention and the Persistence of Wrong Mental Models: Evidence from Firm Pricing

Abstract

We document a mechanism through which wrong mental models survive in data-rich environments: by remaining silent about the patterns they deem irrelevant, wrong models direct attention away from the patterns that refute them. We examine this in firm pricing under consumer intertemporal substitution (IS), where shifts in purchase timing add to the sales spike on the discount day and produce dips before and after. A simple model in which lower prices lead to higher sales is silent about the dips and attributes the spike entirely to new demand. In survey and administrative data from 13,000 gas station managers, over 40 percent neglect or underestimate IS, and those holding the simple model perceive demand as more elastic, set lower prices, and earn lower profits. Online experiments provide causal evidence. Holding the data fixed, subjects assigned the simple model are less likely to recall the dips and more likely to neglect IS than those assigned a model that explains the dips. Two interventions further support this mechanism: prompting attention to the dips lowers IS-neglect, and encountering an alternative model that is also wrong, but explains the dips, dislodges the simple model. Joint with Yi Han and David Huffman.