Field Experiment in Politics

The generic field-experiment concept spans all disciplines — development economics, public health, education — and emphasizes the broad principle of randomizing real interventions in natural settings. This entry focuses on the political-science tradition, where the canonical application is get-out-the-vote mobilization and the central analytic concerns are validated behavioral outcomes from voter files, one-sided noncompliance in campaign contact, clustered assignment, and randomization inference. The statistical logic is shared, but the substantive questions, outcome measures, and practical conventions are specific to the study of political behavior.

The intent-to-treat effect compares outcomes between everyone assigned to treatment and everyone assigned to control, regardless of whether the treatment was actually delivered; it captures the effect of the assignment or attempt, which is what a campaign can directly manipulate. The complier average causal effect is the effect of actually receiving the treatment among compliers, recovered by dividing the intent-to-treat effect by the share of the assigned group that was successfully treated. The latter answers what the contact itself does, under instrumental-variable assumptions including exclusion and monotonicity.

Randomization inference derives uncertainty directly from the experiment's known assignment mechanism rather than from parametric assumptions about the population. By repeatedly reassigning treatment under the sharp null hypothesis of no effect and recomputing the test statistic, it builds the exact reference distribution against which the observed estimate is compared. This is especially valuable in field experiments with small samples, clustered assignment, or skewed outcomes, where the assumptions behind conventional standard errors may not hold and design-based inference is more trustworthy.