Pre-treatment surveying verifies that randomization successfully created balanced experimental groups prior to intervention delivery.

Download image · 1200 × 630 PNG

Research procedure

Measure baseline before campaigning

  1. Randomly sample households across study villages
  2. Interview voters before campaign launch
  3. Compare baseline means across assigned groups

Illustrative summary of the source-described baseline survey across 60 treatment and 90 control villages in Benin. Demographic, economic and political measures support balance checks; random assignment does not guarantee identical measured means.

See source →

How

When to use

Evaluating political campaigns or electoral field experiments requires establishing demographic and attitudinal baselines prior to intervention. Conduct pre-treatment voter surveying to verify sample balance across experimental conditions and document baseline political knowledge, assets, and preferences. This allows researchers and planners to identify potential confounders before campaign treatments begin.

Administer a structured pre-treatment survey to a representative sample of registered voters across treatment and control units before initiating campaign activities.

  1. Draw a representative random sample of households within every selected study unit.
  2. Deploy trained enumerators to interview prospective voters in selected households prior to campaign launch.
  3. Collect demographic data including respondent age, gender, spoken languages, and education level.
  4. Measure socioeconomic status by recording steady income source and household landholding size.
  5. Gauge political knowledge and preferences, including familiarity with local mayors, presidential candidates, and expected turnout.
  6. Compare mean baseline values between treatment and control groups to verify statistical balance across covariates.

Why it works

It allows researchers to identify initial imbalances and control explicitly for baseline differences in subsequent empirical estimation.

Example

Prior to the March 2011 presidential election in Benin, research staff surveyed 5,009 representative household respondents across 150 villages in 30 electoral districts, collecting data on demographic, political, and economic variables to verify balance before campaign town halls and rallies began.

About the study

Participants

5,009 representative household respondents across 150 villages in 30 communes across Benin.

Study Design

Pre-treatment baseline survey administered across randomly assigned treatment (60 villages) and control (90 villages) units, evaluated using two-sided t-tests of mean equality across demographic, political, and economic covariates.

Results

Balance tests showed baseline female proportion of 0.61 in treatment vs 0.59 in control (p = 0.08), mean age of 36.97 vs 37.08 (p = 0.40), steady income index of 1.78 vs 1.77 (p = 0.64), education of 0.49 vs 0.46 (p = 0.04), and expected turnout of 1.05 vs 1.08 (p = 0.05).

Source

See source →