Pre-analysis plan: Are voters difference-in-differences estimators? Evidence from experimentally benchmarking recent and longer-term municipal homicide rates in Mexico
How to cite
Marshall, John. 2018. "Pre-analysis plan: Are voters difference-in-differences estimators? Evidence from experimentally benchmarking recent and longer-term municipal homicide rates in Mexico." AEA RCT Registry. AEARCTR-0003113. https://doi.org/10.1257/rct.3113-2.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
Drawing inference about incumbent quality or effort from incumbent performance underpins electoral accountability. However, interpreting a performance signal is often complicated by the myriad of unit-specific and period-specific factors beyond the incumbent's control that also influence observable outcomes. Providing voters with benchmarks has the potential to help them to filter the signal from the noise. This project experimentally compares and combines spatial and temporal benchmarks in the context of providing information about recent or longer-term municipal homicide rates across Mexico. Specifically, I assess spatial benchmarks, temporal benchmarks, and a difference-in-differences combination of the two. Leveraging a panel of voters from large municipalities holding municipal elections on July 1, 2018, I examine which benchmarks are most relevant to voters by comparing the effects of different types and combinations of benchmarks on voters' beliefs and self-reported municipal and presidential vote choices, as well as the extent to which voters are able to process benchmarked information---a likely necessary condition for observing any effect.
Intervention
Political accountability is generally regarded as being a central pillar ensuring good governance and its socioeconomic benefits. However, and particularly in developing contexts where governance is weakest, holding politicians to account is challenging for many voters due to their lack of useful information about the actions of their representatives in office. Building on a variety of recent studies examining the effects of information dissemination on electoral, this study focuses on the types of incumbent performance information that influence voters, differentiating recent and longer-term performance and different types of performance benchmark.
I seek to estimate the effects of providing recent v. longer-term information about municipal homicide outcomes on incumbent appraisals and electoral support. To examine the relevance of different types of benchmark, I will compare providing information about each homicide metric with (i) a spatial benchmark (average neighboring municipal administration), (ii) a temporal benchmark (the previous municipal administration), or (iii) a difference-in-differences benchmark. A variety of questions are designed to establish whether the information was understood and illuminate the mechanisms underlying any treatment effects.
Experimental design
Field experiment randomizing the provision of text-based and infographic municipal homicide data as part of a panel survey conducted around the 2018 Mexican elections. There are 10 treatment conditions, including a pure control, placebo control, 4 factorial benchmarks relating to recent homicide information, and 4 factorial benchmarks relating to longer-term homicide rates under the current municipal administration.
Design details
Cited by
No citation data available for this registration yet.