While individual vote-switching can stem from unobserved informational changes, randomized intervention assignment ensures non-vote-selling reasons are orthogonal across treatment groups.

Download image · 1200 × 630 PNG

Research steps

Construct a vote-switching proxy

  1. Record pre-election favourability on a 7-point scale
  2. Collect post-election reported votes
  3. Classify switching by race; pool into any-race measure

Illustrative summary of the source-described proxy. Single-choice switching means voting outside the highest-rated candidate; multi-choice switching means at least one vote outside the top four. Switching can have reasons other than vote-selling.

See source →

How

When to use

Researchers and civil society organizations studying vote-buying face secret ballots where direct observation of vote-selling is impossible. This proxy method measures vote-selling indirectly by tracking changes between pre-election preferences and post-election reported choices. It applies to electoral studies in secret-ballot jurisdictions where clientelism and candidate cash offers occur.

Construct a vote-switching outcome metric by comparing voters' pre-election candidate favorability ratings with their post-election reported votes across two survey waves.

  1. Enumerators conduct a baseline survey during the month prior to an election, asking voters to rate all candidates in each race on a 7-point favorability scale.
  2. Enumerators administer a post-election survey within a month following the election to record for whom participants actually voted.
  3. Analysts classify single-choice race voters as having switched if they voted for a candidate other than their highest-rated baseline candidate.
  4. Analysts classify multi-choice race voters as having switched if they voted for at least one candidate outside their top four baseline choices.
  5. Analysts pool switching indicators across all targeted electoral races to form a composite binary variable indicating vote-switching in any race.

Why it works

Consequently, treatment-induced changes in vote-switching isolate changes in actual vote-selling behavior.

Example

In Sorsogon City, Philippines, researchers evaluated 883 voters across mayoral, vice-mayoral, and city council races during the May 2013 municipal elections, finding 56% switched votes in at least one race.

About the study

Participants

883 surveyed voters across three municipal races in Sorsogon City, Philippines in the May 2013 elections.

Study Design

Two-wave panel survey field experiment measuring vote-switching across single-seat and multi-seat elections, validated against candidate favorability gaps and enumerator vote-buying ratings.

Results

Overall 56% of voters switched in at least one race (12% mayor, 22% vice-mayor, 44% city council); a one-category increase in favorability gap reduced switching by 13 percentage points, while a one-point increase in candidate vote-buying gap increased switching likelihood by 23%.

Source

See source →