Which Comparison Counts? Information, Benchmarking, and Electoral Accountability

Field experiments have mixed findings on how and to what extent information changes voting behavior. Many factors may explain these mixed results, but one that remains under-explored is how benchmarking, or the presentation of performance information relative to other governments, can change interpretation. In a pre-registered survey fielded in Mexico with both descriptive and experimental components, I examine respondents' preferences for benchmarked information. I find that respondents prefer benchmark municipalities that are geographically proximate, highly populated, and governed by the same party coalition as the respondent's municipality. When randomly assigned to treatment conditions that vary whether partisan benchmarks are shown to respondents, I find that only same-political coalition comparisons change respondents' intentions to vote for an incumbent. This effect is driven in part by respondents updating negatively about incumbent politicians relative to opposition party coalitions.