Literature

PDF storage. All file: <citekey>.pdf entries below resolve to bi-dropbox:poll-sponsor-bias/literature/<citekey>.pdf — the project's canonical Dropbox folder going forward. Citekeys, BibTeX keys in paper/references.bib, the [cite:<key>] tokens, and the PDF filenames all share the same identifier. As of 2026-06-16 the folder has 40 PDFs and covers all 12 papers currently cited in paper/paper.tex; the broader curated set in this document is ~63 entries and the gap is paywalled classics in queue for the next campus-network fetch round.

Sponsor bias and house effects

• Leeper and Thorson (2019) "Should We Worry About Sponsorship-Induced Bias in Online Political Science Surveys?"
  • topic: Political scientists rely heavily on survey research to gain insights into public attitudes and behaviors.
  • relevance: Closest direct test of 'sponsor bias' in surveys: experimentally varies sponsor (university vs marketing firm) and measures total survey error. Sets up the construct but in online opinion surveys, not pre-election polls and not by candidate sponsor.
  • file: leeper2019sponsorship.pdf
• Panagopoulos (2016) "Exit Poll Sponsorship and Response Intentions"
  • topic: Scholars and practitioners alike are increasingly concerned about growing and differential rates of nonresponse in exit polls.
  • relevance: Exit-poll sponsorship (television networks vs generic) shifts response intentions — direct evidence that disclosing the sponsor changes who answers. Mechanism cousin to our story.
  • file: panagopoulos2016exit.pdf
• De Stefano et al. (2022) "Preelectoral polls variability: A hierarchical Bayesian model to assess the role of house effects with application to Italian elections"
  • topic: It is widely known that pre-electoral polls often suffer from nonsampling errors that pollsters try to compensate for in final estimates by means of diverse ad hoc adjustments, thus leading to well-known house effects.
  • relevance: Workhorse hierarchical-Bayesian house-effects model on Italian pre-election polls; decomposes variance into firm-specific bias. Methodological template to extend with a sponsor-level random effect.
  • file: destefano2022preelectoral.pdf
• Pauli et al. (2013) "A hierarchical Bayesian model for house effects in pre-electoral polls"
  • topic: A hierarchical Bayesian model for house effects in pre-electoral polls
  • relevance: Earlier hierarchical Bayesian house-effects model from the Trieste group; conceptual precursor to De Stefano et al. 2022.
  • file: pauli2013hierarchical.pdf
• Torelli et al. (2015) "A heteroskedastic model for estimating house effect from Italian pre-electoral poll data"
  • topic: A heteroskedastic model for estimating house effect from Italian pre-electoral poll data
  • relevance: Heteroskedastic extension of the Italian house-effects framework. Useful if firm-by-sponsor variance components are needed.
  • file: torelli2015heteroskedastic.pdf
• Selb et al. (2023) "Bias and Variance in Multiparty Election Polls"
  • topic: Recent polling failures highlight that election polls are prone to biases that the margin of error customarily reported with polls does not capture.
  • relevance: Adapts Shirani-Mehr-style hierarchical Bayesian decomposition to multiparty German polls 1994–2021. Off-the-shelf code path for splitting bias from variance — extend with sponsor FE.
  • file: selb2023bias.pdf
• Chen et al. (2025) "Electoral predictors of polling errors" (Journal of Politics; preprint 2023)
  • topic: To understand when polls are accurate and when they fail, we adopt a Bayesian hierarchical modeling approach that separates poll bias and variance at the election level, and links error components to a broad range of election features inclu…
  • relevance: Hierarchical Bayesian model on 9,298 US Senate polls; separates election-level poll bias from variance and links to election features. Closest large-N analogue to what we'd run on 14k Brazilian mayoral polls. Now published in JoP (DOI 10.1086/736694).
  • file: chen2023electoral.pdf
• Arzheimer and Evans (2013) "A New Multinomial Accuracy Measure for Polling Bias"
  • topic: In this article, we propose a polling accuracy measure for multi-party elections based on a generalization of Martin, Traugott, and Kennedy's two-party predictive accuracy index.
  • relevance: Multinomial accuracy measure for multi-party polls. Candidate outcome metric when comparing across many mayoral races with >2 viable parties.
  • file: arzheimer2013multinomial.pdf
• Jiang et al. (2023) "Postelection analysis of presidential election/poll data"
  • topic: This paper concerns analyses of the 2016 and 2020 U. S. presidential election data, including the data of pre-election polls and the actual elections.
  • relevance: Decomposes the 2016/2020 US pre-election vs election discrepancy by state and firm — multi-level method useful for the within-race extension.
  • file: jiang2023postelection.pdf
• Crabtree, Kern and Pietryka (2020) "Sponsorship Effects in Online Surveys"
  • topic: Sponsorship Effects in Online Surveys
  • relevance: Political Behavior follow-up to Leeper & Thorson testing how disclosed survey sponsor moves response behavior in online surveys. Second leg of the experimental sponsor-bias literature; pairs naturally with our observational design.
  • file: crabtree2020sponsorship.pdf
• Burgess (2010) "Credibility Crux: How Public Opinion Polls are Affected by News Organization Branding"
  • topic: Credibility Crux: How Public Opinion Polls are Affected by News Organization Branding
  • relevance: Direct precursor on news-outlet branding effects on poll perception/results. Sister mechanism to candidate-sponsor branding; cite in motivation.
  • file: burgess2010credibility.pdf

Brazilian polling accuracy and methodology

• Batista Pereira and Nunes (2024) "Pesquisas eleitorais e mudanças tardias na decisão do voto"
  • topic: Argues that the 2022 Brazilian presidential first-round divergence between Quaest-Genial polls and the actual result is explained by late preference changes — strategic voting + undecided-voter alignment — rather than polling-methodology failure. Combines a pre-election experiment with eve-of-election Quaest data (registered as TSE BR-07190/2022).
  • relevance: Sibling paper using TSE-registered Brazilian poll data; written by Felipe Nunes (Quaest co-founder). Same data, different question — they explain poll-vote divergence with voter-side dynamics, we ask whether sponsors tilt the polls themselves. Important to cite as the in-house alternative explanation that our sponsor-bias result has to be distinguished from. Industry-insider provenance.
  • file: batistapereira2024pesquisas.pdf
• Cantú, Hoyo and Morales (2016) "The Utility of Unpacking Survey Bias in Multiparty Elections: Mexican Polling Firms in the 2006 and 2012 Presidential Elections"
  • topic: Kalman-filter method that uses all polls in a campaign to estimate systematic, firm-specific bias for each candidate in multiparty pre-election polls; applied to Mexican 2006 and 2012 presidential elections.
  • relevance: Closest Latin American methodological predecessor — firm-level bias decomposition in multiparty pre-election polls. Combines multiple polls into a campaign-level latent path; their Kalman-filter pipeline is an alternative to the hierarchical Bayesian house-effects pipeline we plan to extend.
  • file: cantu2016utility.pdf
• Gramacho (2013) "À margem das margens? A precisão das pesquisas pré-eleitorais brasileiras em 2010"
  • topic: This article analyzes 156 pre-election surveys conducted in 2010 on candidates for President and Governor in the 27 Brazilian Federal Units.
  • relevance: Most-cited Portuguese-language audit of Brazilian pre-election polls (2010 president + governors). Documents firm-level error but no sponsor identifier and no within-candidate design — direct predecessor we extend.
  • file: gramacho2013margem.pdf
• Gramacho (2015) "Pre-election polls in the 2014 Brazilian elections: error, accuracy and controversy"
  • topic: Este artículo se ocupa del análisis de encuestas preelectorales de las elecciones brasileñas de 2014, cuando estuvieron en juego los puestos de Presidente de la República y de los 27 Gobernadores de Estado.
  • relevance: English/Spanish synthesis on 2014 Brazil polls (president + governors). Discusses methodological and regulatory debate; cites 'encomendada' problem but does not test it.
  • file: gramacho2015preelection.pdf
• Gramacho (2015) "Surveys pré-eleitorais nas eleições brasileiras de 2014: erros, acertos e polêmicas"
  • topic: Surveys pré-eleitorais nas eleições brasileiras de 2014: erros, acertos e polêmicas
  • relevance: Portuguese version of the 2014 paper; cite alongside the English version for the lusophone audience.
  • file: gramacho2015surveys.pdf
• Bertholini et al. (2022) "Against all Odds: Forecasting Brazilian Presidential Elections in times of political disruption"
  • topic: When the number of observed elections is low, subnational data can be used to perform electoral forecasts.
  • relevance: Brazilian presidential election forecasts 2010–2018; finds polls underperform incumbent-popularity models, hinting at non-random poll error. Sets the Brazilian forecast-accuracy baseline.
  • file: bertholini2022against.pdf
• Kamakura (2016) "Using Voter-choice Modeling to Plan Final Campaigns in Runoff Elections"
  • topic: Even though runoff elections are the most common form of presidential elections in the world, voter-choice behavior in these two-round elections have not received the attention they deserve in the literature.
  • relevance: Brazilian application of voter-choice modeling to runoff elections. Useful for second-round restriction and within-race logic.
  • file: kamakura2016voterchoice.pdf
• Calvão et al. (2015) "Stylized Facts in Brazilian Vote Distributions"
  • topic: Elections, specially in countries such as Brazil, with an electorate of the order of 100 million people, yield large-scale data-sets embodying valuable information on the dynamics through which individuals influence each other and make choi…
  • relevance: Stylized facts on Brazilian vote distributions 1970–2014. Background on the structure of mayoral/vereador electorates.
  • file: calvao2015stylized.pdf
• Meireles and Russo (2022) "Pesquisas eleitorais no Brasil: tendências e desempenho"
  • topic: Analyzes estimates from 2,000+ pre-election polls against the results of five municipal and national Brazilian elections between 2012 and 2020.
  • relevance: Closest existing audit on the same data we use — 2k+ TSE-registered polls, 2012–2020, municipal and national. Examines sample-design predictors of error but does not link polls to sponsors or use within-candidate FE. Direct benchmark and the paper our sponsor-bias result must beat.
  • file: meireles2022pesquisas.pdf
• Lloyd, Turgeon and Gramacho (2016) "Vote buying, undecided voters, and their effects on polling error in Brazil"
  • topic: Vote buying, undecided voters, and their effects on polling error in Brazil
  • relevance: Brazilian-context working paper linking vote-buying and undecided-voter behavior to poll error. Confounder catalog: rules out a non-sponsor explanation we'd otherwise have to discuss.
  • file: lloyd2016vote.pdf

Poll regulation, disclosure regimes, and pre-registration

• Daoust and Mongrain (2023) "The Regulation of Pre-election Polls: A Citizen’s Perspective"
  • topic: The number of legislations around the world restricting the use and publication of pre-election polls during election campaigns is on the rise.
  • relevance: Survey-experimental evidence on citizen support for banning pre-election polls in Canada. Most directly engages the regulatory framing — useful for the policy intro.
  • file: daoust2023regulation.pdf
• Donovan and Bowler (2016) "Experiments on the Effects of Opinion Polls and Implications for Laws Banning Pre-election Polling"
  • topic: Experiments on the Effects of Opinion Polls and Implications for Laws Banning Pre-election Polling
  • relevance: Experiments on the effects of opinion polls under bans on pre-election polling. Cites international landscape of poll-regulation regimes — background for the TSE pre-registration argument.
  • file: donovan2016experiments.pdf
• Sułek (2008) "The Struggle for the Freedom to Publish Pre-Election Poll Results: The Case of Poland"
  • topic: The Struggle for the Freedom to Publish Pre-Election Poll Results: The Case of Poland
  • relevance: Polish case study on regulation of releasing pre-election poll results; comparative reference for the TSE mandatory-registration regime.
  • file: sulek2008struggle.pdf
• Chatterjee and Kamal (2020) "Voting for the underdog or jumping on the bandwagon? Evidence from India’s exit poll ban"
  • topic: Voting for the underdog or jumping on the bandwagon? Evidence from India’s exit poll ban
  • relevance: Quasi-experimental evidence using India's exit-poll ban. Same regulatory-variation logic but for exit (not pre-) polls.
  • file: chatterjee2020voting.pdf
• Daoust et al. (2020) "Are Pre-Election Polls More Helpful than Harmful? Evidence from the Canadian Case"
  • topic: L’importance grandissante des sondages dans la couverture médiatique soulèvent des questions politiques par rapport à leur régulation lors des campagnes électorales.
  • relevance: Canadian case on whether pre-election polls help or harm — counterpoint to the 'biased polls distort information' motivation.
  • file: daoust2020arepre.pdf

Effects of polls on voters and candidates

• Faas et al. (2008) "Polls that Mattered: Effects of Media Polls on Voters’ Coalition Expectations and Party Preferences in the 2005 German Parliamentary Election"
  • topic: Polls that Mattered: Effects of Media Polls on Voters’ Coalition Expectations and Party Preferences in the 2005 German Parliamentary Election
  • relevance: Schmitt-Beck et al. on the 2005 German election: media polls shift voters' coalition expectations and party preferences. Canonical evidence that biased polls would matter for choice.
  • file: schmittbeck2008polls.pdf
• Granzier et al. (2019) "Coordination and Bandwagon Effects: How Past Rankings Shape the Behavior of Voters and Candidates"
  • topic: Candidates' placements in polls and past elections can be powerful coordination devices for parties and voters.
  • relevance: French two-round RDD: first-round ranking shapes second-round voting and candidate dropout. Direct evidence that small shifts in observed poll rank can change outcomes — strong motivation for our question.
  • file: granzier2019coordination.pdf
• Mutz (1995) "Effects of Horse-Race Coverage on Campaign Coffers: Strategic Contributing in Presidential Primaries"
  • topic: Donors update strategically on horse-race poll coverage; pre-primary poll standings predict campaign contributions, consistent with donors backing perceived viable candidates.
  • relevance: Cited in the paper/paper.tex introduction footnote on why polls are commissioned by candidates — slanted polls plausibly affect not just voters but the donor / fundraising margin. The mechanism complements the voter-side coordination story in [cite:granzier2019coordination] with an upstream financial channel.
  • file: mutz1995horserace.pdf
• Andonie and Kuzmics (2012) "Pre-Election Polls as Strategic Coordination Devices"
  • topic: Pre-Election Polls as Strategic Coordination Devices
  • relevance: Game-theoretic model of pre-election polls as strategic coordination devices. Foundation for why sponsors would bother to slant.
  • file: andonie2012preelection.pdf
• Jo (2022) "Informational roles of pre‐election polls"
  • topic: Informational roles of pre‐election polls
  • relevance: Game-theoretic 'informational roles of pre-election polls'. Modeling reference for the welfare cost of slanted polls.
  • file: jo2022informational.pdf
• Araujo and Gatto (2021) "Casting Ballots When Knowing Results"
  • topic: Access to information about candidates' performance has long stood as a key factor shaping voter behaviour, but establishing how it impacts behaviour in real-world settings has remained challenging.
  • relevance: Brazilian 2018 natural experiment: voters who cast ballots after partial results saw a bandwagon effect. Brazil-specific evidence that poll-like information moves vote choice.
  • file: araujo2021casting.pdf
• McAllister and Studlar (1991) "Bandwagon, Underdog, or Projection? Opinion Polls and Electoral Choice in Britain, 1979-1987"
  • topic: Bandwagon, Underdog, or Projection? Opinion Polls and Electoral Choice in Britain, 1979-1987
  • relevance: Classic on bandwagon/underdog/projection effects in British polls. Origin reference for the mechanism.
  • file: mcallister1991bandwagon.pdf
• Cornejo (2024) "Partisan bias in public perception of elections polls: experimental evidence from Mexico"
  • topic: While most of the literature on public opinion has explored how motivated reasoning makes some individuals discredit survey results in advanced industrial democracies like the U.S., few studies have considered how voters perceive electoral…
  • relevance: Mexican survey experiment: voters perceive polls matching their partisan preference as more credible, and overestimate co-partisan vote share. Direct evidence of perception–slant interaction.
  • file: castrocornejo2024partisan.pdf
• Cantú and Márquez (2021) "The effects of election polls in Mexico's 2018 presidential campaign"
  • topic: The effects of election polls in Mexico's 2018 presidential campaign
  • relevance: Effects of polls in Mexico's 2018 presidential race — Latin American analog of the Schmitt-Beck setup.
  • file: cantu2021effects.pdf
• Farjam (2020) "The Bandwagon Effect in an Online Voting Experiment With Real Political Organizations"
  • topic: Since the 1980s, the number of political polls has increased 10-fold, and starting in the last decade, an increasing number of Internet polls have come to be based on convenience samples, no longer even maintaining the appearance of represe
  • relevance: Field-experimental bandwagon evidence with real political organizations. Cite for the 'effects of polls' mechanism.
  • file: farjam2020bandwagon.pdf
• Dahlgaard et al. (2015) "How do opinion polls affect voters? The effect of opinion polls on the Danes’ voting behavior and sympathy for parties"
  • topic: Information like opinion polls can influence public opinion.
  • relevance: Danish poll effects on vote shares and party sympathy — quantifies how much sponsor bias would have to move to matter.
  • file: dahlgaard2015opinion.pdf
• Stoetzer et al. (2018) "Forecasting Elections in Multiparty Systems: A Bayesian Approach Combining Polls and Fundamentals"
  • topic: We offer a dynamic Bayesian forecasting model for multiparty elections.
  • relevance: Dynamic Bayesian forecasting model for multiparty elections — useful for forecast/error baselines.
  • file: stoetzer2018forecasting.pdf
• Strijbis et al. (2015) "Haben die Umfragen das Wahlergebnis beeinflusst? Strategisches Wählen und Mitläufereffekte bei der Bundestagswahl 2013"
  • topic: Whether polls influenced the election outcome — strategic voting and bandwagon effects in the 2013 German Bundestag election.
  • relevance: Direct follow-up to Faas/Schmitt-Beck (2008) on a different German election cycle. Cite alongside schmittbeck2008polls to show the bandwagon/strategic-coordination mechanism replicates.
  • file: strijbis2015umfragen.pdf

Selective publication of polls and survey selection

• Lee et al. (2024) "Do Korean News Media Selectively Report Election Polls? : A Bayesian Analysis of House Effects, Disparities in Poll Outcomes, and Their Relationship to Reporting Frequency During the 20th Presidential Election"
  • topic: Do Korean News Media Selectively Report Election Polls? : A Bayesian Analysis of House Effects, Disparities in Poll Outcomes, and Their Relationship to Reporting Frequency During the 20th Presidential Election
  • relevance: Closest predecessor on the publication side of sponsor bias: Bayesian decomposition of Korean media polls into house effects and selective reporting by news outlets. Same mechanism we 'kill' by using TSE pre-registration.
  • file: lee2024korean.pdf
• West and Andridge (2023) "Evaluating Pre-election Polling Estimates Using a New Measure of Non-ignorable Selection Bias"
  • topic: Among the numerous explanations that have been offered for recent errors in pre-election polls, selection bias due to non-ignorable partisan nonresponse bias, where the probability of responding to a poll is a function of the candidate pref…
  • relevance: New measure of non-ignorable selection (partisan nonresponse) in pre-election polls. Methodological complement: ours is sponsor selection, theirs is respondent selection.
  • file: west2023evaluating.pdf
• Carlson (2018) "The perils of pre-election polling: Election cycles and the exacerbation of measurement error in illiberal regimes"
  • topic: Pre-election polls are increasingly being used to forecast and study elections in new and weak democracies yet the tension that surrounds elections in less liberal regimes may make data from such surveys particularly prone to measurement er…
  • relevance: Pre-election polls in illiberal regimes — measurement error grows with the election cycle. Cite for the broader 'when do polls misreport?' question.
  • file: carlson2018perils.pdf
• Калинин (2016) "The social desirability bias in autocrat's electoral ratings: evidence from the 2012 Russian presidential elections"
  • topic: In authoritarian regimes, election polls can be vastly polluted by measurement error, namely the social desirability bias, which can contribute to substantial inflation in the publicized estimates of an autocrat's electoral support and vote…
  • relevance: Russian 2012: social-desirability bias inflates autocrat support. Polluted-respondent counterpart to polluted-sponsor.
  • file: kalinin2016social.pdf
• Jennings and Wlezien (2018) "Election polling errors across time and space"
  • topic: Election polling errors across time and space
  • relevance: Cross-country audit of polling error across time and space (Nature Human Behaviour). Benchmark for the average error our sponsor effect must be compared against.
  • file: jennings2018election.pdf
• Kennedy et al. (2017) "An Evaluation of the 2016 Election Polls in the United States"
  • topic: The 2016 presidential election was a jarring event for polling in the United States.
  • relevance: AAPOR analysis of the 2016 US polling 'miss'. Canonical post-mortem of poll error decomposition.
  • file: kennedy2017evaluation.pdf
• Stout and Kline (2010) "I’m Not Voting for Her: Polling Discrepancies and Female Candidates"
  • topic: Although there is a large literature on the predictive accuracy of pre-election polls, there is virtually no systematic research examining the role that a candidate’s gender plays in polling accuracy.
  • relevance: Polls underestimate female candidates ('Richards Effect'). Demographic-selection analog to our sponsor effect — same within-candidate logic.
  • file: stout2010voting.pdf
• Daoust (2021) "Blame it on turnout? Citizens’ participation and polls’ accuracy"
  • topic: When pre-election polls fail, citizens make choices in an environment where the information is inaccurate.
  • relevance: Cross-country test of whether voter turnout explains polling error. Rules out one confounder.
  • file: daoust2021blame.pdf

Brazilian electoral context, campaign finance, and parties

• Mancuso et al. (2023) "Money in Electoral Campaigns: The Relationship between Money and Politics as a Cause of the Judicialization of Electoral Processes in Brazil"
  • topic: The electoral court is a specialized branch of the Brazilian judiciary that not only organizes and regulates elections in the country but also resolves conflicts between political actors involved in the electoral process.
  • relevance: Money in Brazilian electoral campaigns and the judicialization of the electoral court — direct background for who 'sponsors' a poll and why TSE pre-registration exists.
  • file: mancuso2023money.pdf
• Hunter and Power (2019) "Bolsonaro and Brazil's Illiberal Backlash"
  • topic: Brazilian voters delivered a sweeping victory to presidential candidate Jair Bolsonaro, putting the far-right populist at the helm of the world's fourth-largest democracy.
  • relevance: Background on the 2018 Bolsonaro election and the polarized environment in which 2024 mayoral polls are run.
  • file: hunter2019bolsonaro.pdf
• Power and Rodrigues-Silveira (2019) "Mapping Ideological Preferences in Brazilian Elections, 1994-2018: A Municipal-Level Study"
  • topic: This article investigates the electorally expressed ideology of Brazilian voters via ecological analysis at the municipal level between 1994 and 2018.
  • relevance: Municipal-level mapping of Brazilian ideology 1994–2018 — useful for thinking about race-level controls in the FE design.
  • file: power2019mapping.pdf
• Trudeau (2022) "How Criminal Governance Undermines Elections"
  • topic: How does criminal governance affect elections? Existing accounts explore various types of politician-criminal group relationships, but have largely overlooked the reasons why criminal groups might be appealing candidate partners.
  • relevance: Criminal governance and elections in Brazilian municipalities — relevant for the 'who sponsors what kind of poll where' selection story.
  • file: trudeau2022criminal.pdf
• Lawson et al. (2010) "Looking Like a Winner: Candidate Appearance and Electoral Success in New Democracies"
  • topic: A flurry of recent studies indicates that candidates who simply look more capable or attractive are more likely to win elections.
  • relevance: Brazilian/Mexican candidate-appearance and electoral success. Brazil-context behavioral reference.
  • file: lawson2010looking.pdf

Forecast accuracy and audits in other settings

• Magalhães (2005) "Pre-Election Polls in Portugal: Accuracy, Bias, and Sources of Error, 1991–2004"
  • topic: Pre-Election Polls in Portugal: Accuracy, Bias, and Sources of Error, 1991–2004
  • relevance: Portuguese pre-election polls 1991–2004 — accuracy and sources-of-error audit. Comparative reference.
  • file: magalhaes2005portugal.pdf
• Andersen and Fox (2001) "Pre-election polls and the dynamics of the 1997 Canadian federal election"
  • topic: Pre-election polls and the dynamics of the 1997 Canadian federal election
  • relevance: Canadian 1997 federal election: pre-election polls and campaign dynamics. Comparative reference.
  • file: andersen2001preelection.pdf
• Callegaro and Gasperoni (2008) "Accuracy of Pre-Election Polls for the 2006 Italian Parliamentary Election: Too Close to Call"
  • topic: Accuracy of Pre-Election Polls for the 2006 Italian Parliamentary Election: Too Close to Call
  • relevance: Accuracy of 2006 Italian polls. Comparative reference for the Italian house-effects literature.
  • file: callegaro2008accuracy.pdf
• Lauderdale et al. (2019) "Model-based pre-election polling for national and sub-national outcomes in the US and UK"
  • topic: Model-based pre-election polling for national and sub-national outcomes in the US and UK
  • relevance: Multilevel model-based pre-election polling for sub-national outcomes in US/UK — closest method for mayoral (sub-national) polls.
  • file: lauderdale2019model.pdf
• Graefe (2014) "Accuracy of Vote Expectation Surveys in Forecasting Elections"
  • topic: simple surveys that ask people who they expect to win are among the most accurate methods for forecasting Us presidential elections.
  • relevance: Vote-expectation surveys vs vote-intention polls. Useful for thinking about which scenario to use as outcome.
  • file: graefe2014accuracy.pdf
• Penadés, Mateos and Bejarano (2024) "Encuestas controladas por cuotas: una aproximación empírica a su error total"
  • topic: Most Spanish and Latin American public-opinion data are collected via quota sampling, a non-probabilistic method whose total survey error is poorly understood.
  • relevance: Empirical total-error decomposition for quota-controlled surveys in Spain and Latin America — the dominant sampling method for Brazilian commercial pre-election polls. Methodological reference for bounding the variance against which our sponsor bias is measured.
  • file: penades2024encuestas.pdf

Methodological references (fixed effects, hierarchical Bayes, meta-analysis)

• Shirani-Mehr, Rothschild, Goel and Gelman (2018) "Disentangling Bias and Variance in Election Polls"
  • topic: Hierarchical Bayesian decomposition of US state-level pre-election poll error 1998–2014 into bias and excess variance components beyond the reported margin of error.
  • relevance: Parent paper of the modern house-effects literature; the model the Selb et al. 2023 and Chen et al. 2023 papers extend. Direct methodological template we extend with a sponsor random effect on TSE-registered Brazilian polls.
  • file: shiranimehr2018disentangling.pdf
• Gelman and King (1993) "Why Are American Presidential Election Campaign Polls So Variable When Votes Are So Predictable?"
  • topic: Why Are American Presidential Election Campaign Polls So Variable When Votes Are So Predictable?
  • relevance: Classic: campaign polls are variable but votes are predictable. Conceptual foundation for hierarchical modeling of poll error.
  • file: gelman1993why.pdf
• Tomaselli (2013) "MULTILEVEL MODELS IN META-ANALYSIS OF PRE-ELECTION POLL ACCURACY MEASURES"
  • topic: MULTILEVEL MODELS IN META-ANALYSIS OF PRE-ELECTION POLL ACCURACY MEASURES
  • relevance: Multilevel meta-analysis of pre-election poll accuracy. Method reference for pooled estimates across races.
  • file: tomaselli2013multilevel.pdf
• Bouton et al. (2018) "A Theory of Small Campaign Contributions"
  • topic: We propose a theory of small campaign contributions driven by an electoral motive, i.e., the desire to influence election outcomes.
  • relevance: Theory of small campaign contributions: closeness and underdog effects. Theoretical reference for sponsor motivation.
  • file: bouton2018theory.pdf
• Mundim et al. (2019) "Bolsa Família, informação e preconceito: uma análise com o uso de experimentos"
  • topic: Este artigo apresenta e discute os resultados de experimentos de lista e de informaçãorealizados em pesquisas de opinião pública encomendadas pela Secretaria de ComunicaçãoSocial da Presidência da República, em 2014 e 2015.
  • relevance: Brazilian survey experiments commissioned by the federal government's communications secretariat — Brazilian institutional reference for government-as-sponsor polls.
  • file: mundim2019bolsa.pdf
• Dunning et al. (2019) "Voter information campaigns and political accountability: Cumulative findings from a preregistered meta-analysis of coordinated trials"
  • topic: Voters may be unable to hold politicians to account if they lack basic information about their representatives' performance.
  • relevance: Preregistered meta-analysis of voter-information field experiments — methodological reference for the pre-registration angle.
  • file: dunning2019voter.pdf

Notes on positioning

Has the paper been done? No. Six closest predecessors, none doing the same combination: (i) [cite:leeper2019sponsorship] Leeper & Thorson (2019) experimentally vary survey sponsor (university vs marketing firm) but in online opinion surveys, not pre-election polls, and not by candidate sponsor; (ii) [cite:lee2024korean] Lee, Zhang & Pak (2024) Bayesian-decompose Korean news-media selective reporting of polls — same spirit, but for the publication-selection problem the TSE pre-registration regime explicitly removes; (iii) [cite:gramacho2013margem] Gramacho (2013) and [cite:gramacho2015preelection] Gramacho (2015) audit Brazilian poll accuracy at the president/governor level but do not link polls to sponsors and do not exploit mandatory pre-registration; (iv) [cite:meireles2022pesquisas] Meireles & Russo (2022) is the closest existing audit on the same data (2k+ TSE-registered polls, 2012–2020, including municipal races), but analyzes only sample-design predictors of error and does not link polls to sponsors or use within-candidate FE — direct benchmark our sponsor-bias result must beat; (v) [cite:batistapereira2024pesquisas] Batista Pereira & Nunes (2024) explain the 2022 presidential polls-vs-results gap with late voter-side change (strategic voting + undecided alignment) on the same TSE-registered universe — competing alternative explanation we must distinguish sponsor bias from; (vi) [cite:cantu2016utility] Cantú, Hoyo & Morales (2016) build a Kalman-filter pipeline for firm-level pre-election poll bias in multiparty Mexican races — closest Latin American methodological predecessor, treats pollster as the unit (no sponsor split, no mandatory pre-registration). The Italian house-effects pipeline ([cite:destefano2022preelectoral], [cite:pauli2013hierarchical], [cite:selb2023bias]) gives us the estimator template but at the pollster level only. Nothing in the pool combines (a) mandatory pre-registration to kill publication selection with (b) a within-candidate FE design that splits sponsor from pollster house effects.

Industry-insider angle. Felipe Nunes — co-author of [cite:batistapereira2024pesquisas] and the main academic voice on Brazilian poll performance — is the co-founder of Quaest, one of the major Brazilian polling firms. He continues as a researcher (UFMG, now FGV-EESP) while running the firm. Implications: (1) the most academically engaged Brazilian pollster has not analyzed sponsor effects; (2) our paper needs an intro framing that does not claim malfeasance, only documents an average bias, and acknowledges that the industry has serious academic voices.

Cleanest positioning. "We use Brazil's mandatory pre-registration regime (TSE PesqEle) to estimate sponsor bias in pre-election polls free of publication-selection contamination, identifying the effect from a within-candidate fixed-effects design that compares the same mayoral candidate across polls with different sponsors." Two distinguishing claims: (1) the regime removes the selection on publication that limits Leeper-Thorson and Lee-Zhang-Pak; (2) the candidate×race fixed effect identifies the client-specific effect separately from the pollster's generic house effect — most prior work does only one or the other.

Citation-graph expansion targets (most central nodes the next pipeline step should follow forward and backward):

1. Leeper & Thorson (2019), DOI 10.1017/xps.2019.25 — direct sponsor-bias predecessor.
2. Lee, Zhang & Pak (2024), DOI 10.20879/kjjcs.2024.68.6.007 — selective-reporting predecessor.
3. De Stefano, Pauli & Torelli (2022), DOI 10.1214/21-aoas1507 — house-effects estimator.
4. Selb, Chen, Körtner & Bosch (2023), DOI 10.1093/poq/nfad046 — multiparty bias-variance decomposition (pulls in Shirani-Mehr et al. 2018, which is not in candidates).
5. Gramacho (2013), DOI 10.1590/S0104-62762013000100004 — the Brazilian benchmark on poll accuracy.
6. Schmitt-Beck, Faas & Mackenrodt (2008), DOI 10.1093/IJPOR/EDN034 — voter-side effects of media polls, motivation anchor.
7. Castro Cornejo (2024), DOI 10.1080/17457289.2024.2409642 — Mexican partisan-bias-in-poll-perception, closest Latin American analog.
8. Chen, Körtner, Selb & Wiederspohn (2023), DOI 10.33774/apsa-2023-t1vh8 — large-N hierarchical model on US Senate polls, methodological scale-up template.

Missing from this pool — flag for next pipeline step / hand-add. Shirani-Mehr, Rothschild, Goel & Gelman (2018) "Disentangling Bias and Variance in Election Polls" (JASA) is the parent paper of half the house-effects work in the pool and should be pulled in explicitly. It did not surface in the citation-graph expansion of Selb 2023 / Chen 2023 / De Stefano 2022 either — needs hand-add via direct DOI lookup. Also worth searching: Holland on poll release and political behavior in Latin America; David Samuels' / Bruno Speck's broader Brazilian campaign-finance work; the AAPOR ad-hoc committee reports on 2020 US polls; Soroka et al. on media-bias-in-polls.

Supply-side theory: persuasion, disclosure, career concerns

Theoretical anchors for the project's supply-side mechanisms (docs/theory.md §"Polls as Bayesian persuasion", §"Polls as verifiable disclosure", §"Polls as career-concerns games"). Added 2026-06-02 in a targeted refresh focused on filling theory.md placeholders. Canonical 1980s-1990s references were added by direct Crossref lookup; modern Bayesian-persuasion follow-ups came from pollster reputation accuracy, Bayesian persuasion information design, and citation-graph expansion on Kamenica-Gentzkow 2011.

• Kamenica and Gentzkow (2011) "Bayesian Persuasion"
  • topic: Sender chooses a signal structure to maximize receivers' posterior. The canonical formal model behind theory.md §"Polls as Bayesian persuasion (supply-side / Channel A)".
  • relevance: Direct theoretical anchor for Channel A — sponsor's choice of poll methodology is the signal structure. The disclosed-methodology regime makes the model especially tractable: the signal structure is publicly committed before realizations.
  • file: kamenica2011bayesian.pdf
• Kamenica (2019) "Bayesian Persuasion and Information Design"
  • topic: Survey of the Bayesian persuasion / information design literature post-2011, including extensions to multiple senders, multiple receivers, persuasion under commitment problems, robust persuasion.
  • relevance: Annu. Rev. Econ. survey — useful for the paper's theory section to situate the Brazilian polling-disclosure setup within the broader framework. Cites most of the relevant extensions.
  • file: kamenica2019bayesian.pdf
• DellaVigna and Gentzkow (2010) "Persuasion: Empirical Evidence"
  • topic: Survey of empirical persuasion estimates across media, voting, finance, charitable giving. Provides effect-size benchmarks across domains.
  • relevance: Reference for the magnitude question — is +7 pp consistent with persuasion effects elsewhere? Useful for the introduction's stakes paragraph.
  • file: dellavigna2010persuasion.pdf
• Crawford and Sobel (1982) "Strategic Information Transmission"
  • topic: Canonical cheap-talk model. Sender's optimal slant decreases in the precision of the receiver's ex-post verification signal.
  • relevance: Theoretical anchor for theory.md §"Polls as verifiable disclosure" prediction 1 (β^B declines with proximity to election day — verification is sharper close to election).
  • file: crawfordsobel1982strategic.pdf
• Milgrom and Roberts (1986) "Price and Advertising Signals of Product Quality"
  • topic: Full-disclosure equilibrium under cheap verification. When receivers can cheaply check the seller's claims, optimal sender behavior collapses to truth-telling.
  • relevance: Companion to crawfordsobel1982strategic for the verifiable-disclosure framework. Predicts the sharp boundary case: as election day approaches and verification becomes nearly costless, β^B → 0.
  • file: milgromroberts1986price.pdf
• Dziuda (2011) "Strategic Argumentation"
  • topic: Sender holds private information and chooses which subset of arguments to disclose; receiver knows total argument count but not content. Partial-verifiability model.
  • relevance: Closest model to the Brazilian polling regime — pollster discloses methodology elements (sex/age/education/income/area weighting) but receivers can't audit the actual sample. Partial verifiability is the empirically relevant case for Channel A.
  • file: dziuda2011strategic.pdf
• Holmström (1999) "Managerial Incentive Problems: A Dynamic Perspective"
  • topic: Canonical career-concerns model. Agent's current-period effort is a signal about future productivity; reputational incentives substitute for explicit incentives.
  • relevance: Theoretical anchor for theory.md §"Polls as career-concerns games". Predicts the heterogeneity gradient by pollster firm size / reputation stock (H10 in hypotheses/). Also used by theory.md §"Pollster reputation and customer-mix sorting" for the supply-side sorting prediction.
  • file: holmstrom1999career.pdf
• Dewatripont, Jewitt and Tirole (1999) "The Economics of Career Concerns, Part I: Comparing Information Structures"
  • topic: Extends Holmström by characterizing how different information structures (e.g., the precision of public output signals) shape career-concerns incentives.
  • relevance: Closer match to the polling-firm setting than Holmström 1999 alone — the polling market has graduated information precision (close-to-election polls have high-precision verification; far-from-election polls have low-precision verification). Companion: Part II (DOI 10.1111/1467-937X.00085) addresses application to public-sector accountability.
  • file: dewatripont1999economics.pdf
• Meirowitz (2005) "Polling Games and Information Revelation in the Downsian Framework"
  • topic: Pre-election polls modeled as games where pollster and respondents have strategic incentives; equilibrium poll responses convey partial information about candidate viability.
  • relevance: The most directly relevant formal model — polls as strategic games. Models the receiver-side of what our paper measures empirically (the slant the sender produces). Useful for the theory section's "why is this hard" framing.
  • file: meirowitz2005polling.pdf
• Gentzkow and Shapiro (2006) "Media Bias and Reputation"
  • topic: Profit-maximizing media outlets in a Bayesian setup with consumers updating about firm accuracy. Outlets slant toward consumer priors to maintain reputation; slant declines as consumers gain information.
  • relevance: Closest cousin to the polling-firm problem in theory.md §"Polls as career-concerns games". Substitute media → polling firms, consumers → media clients + sponsors. The reputation-slant trade-off is structurally identical, with the addition that polling-firm reputation is verified by election results, not by consumer welfare.
  • file: gentzkow2006media.pdf
• Klein and Leffler (1981) "The Role of Market Forces in Assuring Contractual Performance"
  • topic: Canonical reputation-as-self-enforcement model. Firms earn a quality premium that is the discounted value of future honest output; this premium is the rent that disciplines current-period behavior in the absence of contract enforcement.
  • relevance: Cited in paper/paper.tex § Mechanism alongside [cite:shapiro1983] for the reputation-as-premium reading of the firm-volume β-discipline gradient — large firms have a credibility premium that disciplines per-poll slant because the cumulative market signal (across many polls) is what generates the premium.
  • file: kleinleffler1981.pdf
• Shapiro (1983) "Premiums for High Quality Products as Returns to Reputations"
  • topic: Equilibrium price premium for reputable producers in a market with imperfect quality observation. The premium compensates firms for the cost of maintaining reputation.
  • relevance: Companion theoretical anchor to [cite:kleinleffler1981] for the §Mechanism reading. Provides the formal price-premium machinery that maps to "large-firm β ≈ 0 versus small-firm β ≈ +12 pp" in the firm-volume tertile cross-cut.
  • file: shapiro1983.pdf

These 12 entries close the theory-side gaps surfaced in docs/theory.md §"Polls as Bayesian persuasion (supply-side / Channel A)", §"Polls as verifiable disclosure", §"Polls as career-concerns games", and §"Reputation as premium-supported quality". The [ref needed] placeholders in theory.md can now be replaced with the citekeys above. No PDF download attempted yet — many are paywalled (JPE, REStud, GEB, AER, QJE) and require campus access.