Poll Sponsor Bias Home Paper Talk

Reference

Research Summary Working Hypothesis Source of the Bias Institutional Background Data Sources Methods Theory Literature

Working notes

Hypotheses Findings Open Questions & Ideas Key Decisions Design Levers

Tasks

Active Tasks Completed Tasks

Communication

Meeting Notes External Feedback

Briefs

All-Brazil Analysis SP Slice Analysis Heterogeneity Pollster Customer Mix Bulk Extraction Audit

Notes

News Anecdotes Coauthor Candidates Educloud Next Steps

Analyses

AN-001: Do 2024 mayoral polls overstate the sponsoring candidate? AN-002: Does β survive race × week FE with an independent-only comparator? AN-003: Pre-poll trajectory placebo AN-004: Does peak β shift at the 200k runoff-eligibility cutoff? AN-005: Does β shrink as the election approaches? AN-006: Does β depend on which sponsor route maps to the candidate? AN-007: Does pollster β depend on customer mix? AN-008: SP-slice prototype design validation AN-009: Is the by-party heterogeneity in β statistically real? AN-010: Does the headline survive the red-team's five highest-leverage attacks? AN-011: Permutation null distribution + leave-one-pollster / leave-one-UF jackknives AN-012: Wild-cluster bootstrap + WLS + week-window sensitivity AN-013: Channel-A vs Channel-B via digit-frequency tests on raw percent AN-013 v2: forensics targeting AN-013's blind spots AN-014: Where does the K2 "+3 pp renormalization inflation" come from? AN-015: Could LLM extraction-quality differences drive the headline? AN-016: Within-firm β — does the headline survive when PDF style is held fixed? AN-017: Refresh AN-007's customer-mix sorting on the AN-016 sample AN-018: Firm size as the missing axis behind AN-016's β dispersion AN-019: Does coverage_class cluster differently across sponsor types? AN-020: Does coverage_class cluster by specific sponsor route? AN-021: Does audit_pct distribution differ by sponsor type? AN-022: Does methodology completeness differ by sponsor type? AN-023: Pollster methodology fingerprint by candidate share AN-024: Is deferral itself a sponsor-specific tell? AN-025: Within-race FE test of media filtering by firm volume AN-026: Coordination-incentive prediction — selection and bias gradient by rank AN-027: Rank-at-commissioning follow-up to AN-026 AN-028: Registration-date alignment for the rank-at-commissioning test AN-029: Money-controlled selection test for the two-mechanism interpretation AN-030: Joint rank-at-commission + money-controlled selection AN-031: Wild-cluster bootstrap on the sponsor-route split AN-032: Paired sponsored × independent oversample-index test AN-033: Does coverage deferral amplify sponsor bias? AN-040: Does coverage deferral amplify sponsor bias *only* for leaders? AN-041: Does mode of data collection differ between sponsored and matched independent polls? AN-042: Does the sponsored side document interviewer training and supervision less than the matched independent? AN-043: Does nonresponse / undecided handling vary between sponsored and matched independent polls? AN-045: Where does the rank-2 sponsor over-statement live? Race-margin decomposition AN-050: AN-045 sharpened with ex-ante rank (rank-at-commission) AN-051: Does questionnaire-side name rotation and scenario filing differ between sponsored and matched independent polls? AN-052: Does the AN-051 rotation contrast survive within firm? AN-053: Does the sponsor's candidate appear earlier in the headline roster on the sponsored side? AN-054: Does AN-051's rotation contrast map onto AN-018's firm-size discipline gradient? AN-055: Coverage class × candidate-base interaction (cheap Tier 2) AN-056: Within-pair quota distribution distance × bias contrast AN-057: Within-pair ponderação description × bias contrast AN-059: Variance decomposition of +7 pp into within-firm and between-firm AN-060: Race-level visibility shrinks the sponsor effect AN-061: Same-firm × same-race customer-discrimination test AN-062: Sponsor effect shrinks with independent-poll density AN-063: Spec 3c with media-only comparator (dropping pollster-self) AN-064: Inference sensitivities for spec 2 and spec 3c AN-065: Publication-selection proxy via firm-race coverage AN-066: Benjamini-Hochberg FDR correction on the design-inventory table AN-067: Scenario-pick robustness AN-068: Sponsor-label permutation null for Spec 2 AN-069: Same-poll sponsor-vs-opponent DiD AN-070: Event-study trajectory figure AN-071: Per-firm accuracy versus sponsor bias AN-072: Candidate-level fraud-suit rate by sponsor exposure (2024) AN-072v2: Poll-level fraud-suit rate by candidate sponsorship AN-073: Firm specialization by party in candidate-sponsored polls AN-074: CPF cell repeat-dyad test — M1-individual vs M4 single-shot AN-075: CPF cell β decomposition — repeat vs singleton dyad AN-076: CPF β decomposition without institute FE AN-077: All-routes repeat-vs-singleton β decomposition AN-078: Disclosure rate × sponsor route × dyad kind — M1 vs M5 on the committee gap AN-079: Cost × sponsorship × bias — is slant priced? AN-080: Slant unit — firm vs. statistician decomposition AN-081: Audit-request outcomes + follow-on linkage AN-082: Poll accuracy by four-bucket sponsor class AN-083: IPOP makes no observable predictions AN-085: Trusted sources decompose the media bucket AN-087: Trusted-firm robustness across four definitions AN-089: Within-firm sponsor effect on accuracy AN-090: Sponsor-effect spec ladder (paper-ready) AN-092: Sample-size robustness of the AN-090 spec ladder AN-093: Final paper appendix table AN-094: Top-25 other_firm sponsor audit AN-095: VS Publicidade deep dive AN-096: Bipartite structure of the shell–pollster relationship AN-097: |error| degradation by firm β tercile AN-097: Type-2 pollster portfolio audit AN-098: Noise floor — why accuracy can't discipline pollsters AN-099: Consensus deviation as an alternative bias signal AN-100: Sponsor-blind detection — a policy mechanism AN-101: Predicted bias as outcome — shell detection via ML AN-102: Headline tables with shell bucket AN-103: Modern ML for bias detection AN-104: Strict-blind detection AN-105: Theoretical slant signals + within-race demeaning AN-106: Double-ML race residualization AN-107: Universe-scale coverage_class × sponsor interaction AN-108: Sampling-error envelope on biased polls — single-poll detectability AN-109: Per-poll z vs comparator — blind multi-candidate audit-ability AN-110: Empirical DEFF from race × week × candidate cells AN-111: Headline-regression robustness under empirical noise floor AN-119: Negative-β firm diagnostic (CENSUS, EVA FRANCIELI) AN-120: Sponsor-type heterogeneity by funding source (`DS_ORIGEM_RECURSO`) AN-121: Universe-level sponsor-type breakdown with iceberg quantification AN-122: AN-102 headline tables with expanded shell bucket An 125 Stronghold Cluster And Cand Prior

Raw (upstream)

poll_2024 poll_response_2024 poll_sponsor_2024

Assembled

poll cand_poll

cand_poll — regression-ready candidate × poll table

The candidate × poll panel that backs every regression in this project. One row per matched candidate per poll, carrying the candidate's poll percent, ground-truth share, error, sponsor flags, and pollster / race context.

Grain: One row per (protocol, politico_id). · Source: source/assemble/cand_poll.py

Column information

Sample data (50 rows)