Sponsor bias in Brazilian electoral polls.
Do pre-election polls tilt toward the candidates who pay for them?
Across the universe of 2024 Brazilian mayoral polls matched to
their registered sponsors, candidate-commissioned polls overstate
that candidate by β ≈ +7 pp on average — robust
across specifications from no controls to race-by-week fixed
effects.
Reading Guide
Start here
Paper
Sponsor Bias in Brazilian Electoral Polls
Start here
Working Hypothesis
Living synthesis of the full causal account, hyperlinked to every supporting AN page.
Mechanism
Source of the Bias
Canonical mechanism doc — what carries the +7 pp; honest split between concrete design levers and opacity signals.
Evidence
Heterogeneity
Eight blocks distinguishing the voter-side and supply-side theories.
Evidence
Findings
Curated load-bearing conclusions, one page per finding.
Evidence
Analyses
All 80+ AN pages, grouped by type — figures, tables, robustness checks.
Background
Institutional Background
TSE PesqEle registration regime, mayoral race rules, CONRE.
Reference
Hypotheses
Testable predictions linking theory and the empirical battery.
Reference
Literature
Curated references with author-year cite anchors.
Talk
Talk not yet built — no
talk/ directory yet.All Documentation
Reference
Communication
Hypotheses (folder mode)
- H7: β peaks just below the front-runner (bandwagon)
- H3: β shrinks substantially when methodology features are controlled for
- H6: β peaks at the M+1 viability cutoff (coordination)
- H9: Slant is largest in polls financed by partisan funds
- H1: Self-sponsored polls overstate the sponsoring candidate
- H8: β is larger for first-time / lesser-known candidates (information cue)
- H12: High-β polls are over-represented in poll-related Representações
- H4: Channel A contribution is larger where methodology flexibility is greater
- H2: Opponent-sponsored polls understate the opposing candidate
- H10: Reputable national pollsters show smaller β than small regional firms
- H11: Pollster-self-sponsored polls operate as a hidden-sponsor channel
- H13: Shell-contratante polls show larger residual β
- H5: Channel B contribution decays as the election approaches
Findings (folder mode)
- Per-poll detectability and blind-audit auditability of the +7 pp bias
- Channel A vs Channel B (design slant vs fabrication) — pilot landed; documented levers under-explain the +7 pp
- Peak β shifts from rank 2 in small munis to rank 3 in runoff munis
- Own-CPF sponsorship route gives β ≈ +19 vs +6-9 for committee/party
- Slant decays as the election approaches: +0.037 pp per day
- Polls commissioned by a Brazilian 2024 mayoral candidate overstate that candidate by ~7 percentage points
- Opponent-sponsored polls understate the candidate by ~2 pp
- Sponsored polls flex the population frame and the bairro list — Channel A signatures from a 130-pair eyeball
- Per-pollster β ranges from −10 to +30 across 33 firms
- The race × week FE design gives β = +6.95 pp on 60 cells
- The bias is sender-specific, not a generic pollster house effect
- Slant is 2.3× larger in tight races than in landslides
- Within-candidate jump from preceding independent poll is +6.7 pp (t = 5.2, n = 132)
- Voter coordination dominates bandwagon as the consequential channel