Randomization Inference With Natural Experiments
Daniel E. Ho, Kosuke Imai · Journal of the American Statistical Association · 2006
Since the 2000 U.S. Presidential election, social scientists have rediscovered a long tradition of research examining the effects of ballot format on voting. Using a new dataset collected by The New York Times, we investigate the causal effect of being listed on the first ballot page in the 2003 California gubernatorial recall election. California law mandates a unique randomization procedure of ballot order that, when appropriately modeled, can be used to approximate a classical randomized experiment in a real world setting. We apply randomization inference based on Fisher's exact test, which directly incorporates the exact randomization procedure and yields accurate nonparametric confidence intervals. Our results suggest that being listed on the first ballot page causes a statistically significant increase in vote shares for more than 40% of the minor candidates, whereas there is no significant effect for the top two candidates. We also investigate how randomization inference differs from conventional estimators that do not fully incorporate California's complex treatment assignment mechanism. The results indicate appreciable differences between the two approaches.
Cited by 142 works. doi:10.1198/016214505000001258
Authors
References
- Design of Experiments
- Generalized Linear Models
- Individual Comparisons by Ranking Methods
- The central role of the propensity score in observational studies for causal effects
- Design and Analysis of Experiments
- Sequential Treatment Assignment with Balancing for Prognostic Factors in the Controlled Clinical Trial
- Generalized Linear Models (2nd ed.).
- Planning of Experiments
- A Generalization of Sampling Without Replacement From a Finite Universe
- Statistics and Causal Inference