Matching to Categories: Learning and Compliance Costs in Administrative Processes

Abstract

A perennial task for the state is the creation and policing of categories. State-created categories have real world impacts on the public. The consequences of racial categorizations, for example, are well-documented. We examine a less studied consequence of state categorization, which are the administrative burdens created when individuals attempt to match themselves to state-created categories. Matching requires time and effort, and failure to match to an advantageous category can mean a loss of material benefits. The matching problem may sometimes result from obscure categories, or an overwhelming number of categories. The matching problem is also amplified when the state uses identity categories—such as self-employed or unemployed, a retiree, parent, spouse or disabled—where individuals hold pre-existing beliefs about such identities that map poorly onto equivalent state categorizations. To study the matching problem and ways to reduce it, we undertook a field experiment in a California welfare program, CalFresh, the state version of the Supplemental Nutrition Assistance Program (SNAP). Claimants often fail to select into the category of “self-employed” even though it would be more favorable for them to do so. We show how a more intuitive presentation of information about the category and its benefits increased the rate of those identifying as self-employed from 8.8% to 12.1%, approximately one-third. We also show that providing a simple self-attestation template to convey information about self-employment status, a means of reducing compliance costs while meeting state documentation requirements, increased the number of claimants providing an acceptable form of documentation to match to the category. The results show that people frequently lack an intuitive understanding of state categories, that the presentation of categories can reduce this matching problem, and that the state can make it easier to document the match.