From 2016 underestimating Trump to 2020's swing state misses and Brexit's big surprise, the polling industry has faced intense scrutiny for its election errors, but the numbers reveal a surprisingly complex global story of evolving methods, high-stakes investments, and a persistent struggle for public trust.
Key Takeaways
- The average absolute polling error for the 2020 U.S. presidential popular vote was 3.9 percentage points according to the American Association for Public Opinion Research (AAPOR)
- In the 2016 U.S. election, 72% of polls underestimated Donald Trump's support by an average of 5.1 points per RealClearPolitics analysis
- Gallup's final 2020 election poll had a 2-point error in the national popular vote margin, better than the industry average of 4.0 points
- 70% of U.S. polls in 2020 used online panels, up from 40% in 2016, per AAPOR
- Response rates for telephone polls dropped to 6% in 2019 from 36% in 1997, per Pew Research
- 85% of major U.S. pollsters now weight by education since 2018, per Gallup
- Global polling revenue grew 8% YoY to $2.5B in 2022, per ESOMAR
- U.S. polling market size reached $1.2B in 2023, with 12% CAGR since 2018
- Gallup annual revenue estimated at $90M in 2022 from polling contracts
- Number of U.S. polling firm employees totaled 15,000 in 2022
- 45% of pollsters hold advanced degrees in statistics or social science, per AAPOR survey
- Average salary for U.S. pollster director is $140K as of 2023
- Public trust in polls fell to 29% in U.S. post-2020 per Gallup
- 42% of Americans say polls influence their vote, down from 55% in 2016
- 61% believe polls are biased toward Democrats, per 2023 Rasmussen
Polling accuracy varies widely across different countries and elections.
Industry Economics
- Global polling revenue grew 8% YoY to $2.5B in 2022, per ESOMAR
- U.S. polling market size reached $1.2B in 2023, with 12% CAGR since 2018
- Gallup annual revenue estimated at $90M in 2022 from polling contracts
- Pew Research Center's polling budget exceeded $50M in 2021 for surveys
- YouGov market cap hit $1.5B in 2023, driven by U.S. expansion
- Ipsos global revenue from opinion polling was €450M in 2022
- U.S. political polling spend by campaigns totaled $350M in 2020 cycle
- Number of active U.S. pollsters dropped 15% to 250 firms since 2016
- Corporate polling contracts grew 25% to $800M globally in 2022
- Kantar polling division revenue $300M in 2022, up 10%
- Online polling platforms like SurveyMonkey generated $400M revenue in 2022
- AAPOR membership dues contribute $2M annually to industry standards
- Election polling fees averaged $50K per national poll in 2022 U.S. cycle
- Global market research industry (incl. polling) valued at $82B in 2023
- U.S. academic polling centers funded $100M+ yearly by grants
- Media orgs spent $200M on polls for 2022 midterms, per AdImpact
- Freelance pollster rates rose to $200/hr in 2023 from $150 in 2019
- International polling consortiums like WIN/GIA bill $10M+ per global study
- Tech integration costs 20% of polling budgets, averaging $5M/firm
- Nonprofit pollsters like Monmouth get 60% funding from donations
- Asia-Pacific polling market grew 15% to $500M in 2022
- Europe polling revenues flat at $1B amid regulations
- Latin America polling up 12% to $200M post-elections
- Africa polling market nascent at $50M, 20% growth
- Australia/NZ polling $100M stable
- U.S. pollster profit margins average 18% in 2023
Industry Economics Interpretation
It seems we’re willing to spend billions globally to ask everyone what they think, while quietly acknowledging that far fewer of us are even listening anymore.
Polling Accuracy
- The average absolute polling error for the 2020 U.S. presidential popular vote was 3.9 percentage points according to the American Association for Public Opinion Research (AAPOR)
- In the 2016 U.S. election, 72% of polls underestimated Donald Trump's support by an average of 5.1 points per RealClearPolitics analysis
- Gallup's final 2020 election poll had a 2-point error in the national popular vote margin, better than the industry average of 4.0 points
- State-level polling errors in 2020 averaged 4.3 points, with swing states like Wisconsin showing 5.5-point errors per FiveThirtyEight
- The 2012 U.S. election saw national polls accurate within 1.2 points on average, per AAPOR historical data
- Brexit referendum polls had an average error of 6.5 points, underestimating Leave by 7 points in final polls
- In the 2019 UK general election, MRP models outperformed traditional polls with errors under 2 points nationally
- Australian federal election 2019 polls erred by 2.8 points on average for Labor's vote share
- French presidential 2017 polls missed Macron's first-round by 1.5 points average across IFOP and others
- Canadian 2019 election polls had 2.1-point average error, per CBC Poll Tracker
- Midterm 2018 U.S. Senate polls averaged 4.0-point errors in competitive races
- Scottish independence 2014 referendum polls erred by 4.5 points underestimating Yes vote
- Brazilian 2018 presidential polls missed Bolsonaro's first round by 3.2 points
- Indian 2019 Lok Sabha polls averaged 5.7-point errors for BJP vote share, per CSDS
- German 2021 federal election polls had 1.8-point average error
- South Korean 2020 election polls accurate within 1.9 points, per Realmeter
- Spanish 2019 election polls erred by 3.4 points for PSOE
- Italian 2018 polls missed 5E Movement by 4.1 points
- Dutch 2021 polls had 2.3-point error for VVD
- Israeli 2021 election polls averaged 3.6-point errors
- New Zealand 2020 polls accurate to 1.4 points
- Swedish 2022 polls erred by 2.7 points for SD
- Argentine 2019 polls missed Fernandez by 4.2 points
- Turkish 2018 polls had 5.1-point average error
- Mexican 2018 polls accurate within 2.0 points for AMLO
- Polish 2019 polls erred by 3.0 points
- Irish 2020 election polls had 1.7-point error
- Belgian 2019 polls missed N-VA by 2.9 points
- Danish 2022 polls accurate to 1.6 points
- Norwegian 2021 polls had 2.4-point error
- U.S. generic ballot polls in 2022 midterms erred by 3.2 points on average
Polling Accuracy Interpretation
Modern polling, while often impressively close, reminds us that its predictions are more like a well-informed dart throw than a psychic reading, with global average errors ranging from reassuringly precise to alarmingly off-target.
Public Trust Perceptions
- Public trust in polls fell to 29% in U.S. post-2020 per Gallup
- 42% of Americans say polls influence their vote, down from 55% in 2016
- 61% believe polls are biased toward Democrats, per 2023 Rasmussen
- UK trust in polls at 35% post-Brexit, per Ipsos Mori
- 52% of Republicans distrust polls vs 20% Democrats, Pew 2022
- 67% say media polls inaccurate after 2016 election, Gallup
- Global trust average 40% in 2023 World Values Survey polling module
- 38% follow election polls closely, up 10% in battleground states
- 45% cite 2020 errors as reason for distrust, Monmouth 2023
- Younger gens (18-29) trust polls at 25%, vs 50% boomers
- 55% of independents view polls as entertainment, not serious
- Post-midterms 2022, trust rebounded 5 points to 34%
- 70% aware of polling house effects/biases, educated voters
- Women trust polls 8 points more than men, per 2022 survey
- 62% say transparency improves trust, demand full methodology
- Conservative media audiences distrust at 75%, liberal 30%
- 48% in EU trust national polls, varying by country 20-70%
- 35% say polls sway undecideds, self-fulfilling prophecy concern
- Trust higher for university polls at 45% vs commercial 25%
- 50% follow aggregator sites like 538 over single polls
Public Trust Perceptions Interpretation
The polling industry now faces a conundrum worthy of a Greek tragedy: nearly everyone is watching the show, most are convinced the script is biased, yet the audience, armed with popcorn and deep distrust, can't quite look away.
Survey Methodology
- 70% of U.S. polls in 2020 used online panels, up from 40% in 2016, per AAPOR
- Response rates for telephone polls dropped to 6% in 2019 from 36% in 1997, per Pew Research
- 85% of major U.S. pollsters now weight by education since 2018, per Gallup
- RDD telephone sampling coverage reached only 70% of U.S. adults by 2020
- Probability-based online panels like AmeriSpeak have MOE of ±1.0% for n=4,000, per NORC
- 62% of 2020 polls used likely voter screens, varying from simple to complex models
- Multimode surveys (phone+online) reduced bias by 1.5 points in 2022 tests, per AAPOR
- Weighting by age, race, education explained 40% of 2016 polling errors, per FiveThirtyEight
- Cell phone-only households in U.S. hit 59% in 2021, necessitating dual-frame sampling
- Nonresponse bias in polls averages 2-3 points for partisanship, per Pew
- 45% of global pollsters use AI for text analysis in open-ended questions as of 2023
- Sample sizes in national U.S. polls averaged 1,200 respondents in 2022
- Opt-in online polls have 4x higher MOE than probability samples, per AAPOR guidelines
- 78% of polls now adjust for turnout models post-2016, per YouGov
- Spanish-language interviewing in U.S. polls increased to 15% of samples in 2020
- MRPs using 100,000+ respondents predicted 2020 swing states within 1 point
- House effects show Rasmussen polls +3R bias, NYT/Siena -1D in 2022
- 92% of AAPOR members disclose methodology details online, up from 65% in 2010
- Global use of mobile surveys rose to 55% in 2022 from 20% in 2015
- U.S. polling firms spent $50M on tech upgrades post-2016, per industry reports
- 67% of polls use raking over iterative weighting for multivariable adjustment
- Voter file matching covers 95% of registered voters in battleground states
- 40% reduction in bias from including refusal questions in surveys, per 2021 study
Survey Methodology Interpretation
As pollsters scramble to address plummeting response rates and the complex demographic shifts of modern America, their trade increasingly resembles a high-stakes archaeological dig, where they must meticulously reassemble a picture of the public from digital fragments while constantly refining their tools against the persistent specter of hidden error.
Workforce Employment
- Number of U.S. polling firm employees totaled 15,000 in 2022
- 45% of pollsters hold advanced degrees in statistics or social science, per AAPOR survey
- Average salary for U.S. pollster director is $140K as of 2023
- 30% growth in data scientists employed by polling firms since 2018
- Women comprise 52% of polling industry workforce in U.S., per 2022 census
- 2,500 certified professional pollsters via AAPOR/MQP in 2023
- Turnover rate in polling analysts at 12% annually due to election cycles
- 65% of senior pollsters have 10+ years experience, per ESOMAR
- Remote work adopted by 75% of polling staff post-COVID
- Entry-level interviewers earn $18/hr, 40% part-time
- 20% of workforce freelance/contract during election seasons
- Training hours average 120/year for pollster staff on ethics
- Diversity: 25% non-white in U.S. polling leadership, up 10% since 2015
- Global pollster workforce estimated 50,000, 40% in U.S./Europe
- Age median 42 for analysts, skewing older post-retirements
- Unionization low at 8% in private polling firms
- Internships in polling grew 50% to 1,000 slots yearly
- Programmer/analyst roles up 35%, salary $110K avg
- Fieldwork staff seasonal peak at 5,000 during U.S. elections
- 55% have certifications like MRIA or Insights Association
- Job postings for pollsters doubled in 2024 cycle vs 2020
- Retention boosted 15% with mental health support post-2020
- Global south pollsters employ 10,000 locals, 20% growth
- UK pollster staff 3,000 total, avg salary £45K
- Field manager roles scarce, 500 U.S. openings yearly
- AI training upskills 40% of workforce, reducing manual coding
Workforce Employment Interpretation
A surprisingly robust and increasingly diverse industry of 15,000 highly educated, well-paid, and often overworked professionals—armed with advanced degrees and AI tools—is trying desperately to read your mind between election cycles while working remotely and hoping you'll pick up the phone.
Sources & References
- Reference 1AAPORaapor.orgVisit source
- Reference 2REALCLEARPOLITICSrealclearpolitics.comVisit source
- Reference 3NEWSnews.gallup.comVisit source
- Reference 4FIVETHIRTYEIGHTfivethirtyeight.comVisit source
- Reference 5POLITICSpolitics.ox.ac.ukVisit source
- Reference 6YOUGOVyougov.co.ukVisit source
- Reference 7ABCabc.net.auVisit source
- Reference 8IFOPifop.comVisit source
- Reference 9CBCcbc.caVisit source
- Reference 10WHATSCOTLANDTHINKSwhatscotlandthinks.orgVisit source
- Reference 11DATAFOLHAdatafolha.com.brVisit source
- Reference 12LOKNITIlokniti.orgVisit source
- Reference 13WAHLRECHTwahlrecht.deVisit source
- Reference 14REALMETERrealmeter.netVisit source
- Reference 15CIScis.esVisit source
- Reference 16IPSOSipsos.comVisit source
- Reference 17PEILpeil.nlVisit source
- Reference 18ENen.idi.org.ilVisit source
- Reference 19ROYMORGANroyMorgan.comVisit source
- Reference 20SCBscb.seVisit source
- Reference 21LANACIONlanacion.com.arVisit source
- Reference 22KONDAkonda.com.trVisit source
- Reference 23MITOFSKYmitofsky.mxVisit source
- Reference 24CBOScbos.plVisit source
- Reference 25VRTvrt.beVisit source
- Reference 26VOXMETERvoxmeter.dkVisit source
- Reference 27NRKnrk.noVisit source
- Reference 28PEWRESEARCHpewresearch.orgVisit source
- Reference 29AMERISPEAKamerispeak.norc.orgVisit source
- Reference 30CDCcdc.govVisit source
- Reference 31ESOMAResomar.orgVisit source
- Reference 32GALLUPgallup.comVisit source
- Reference 33TODAYtoday.yougov.comVisit source
- Reference 34ADWEEKadweek.comVisit source
- Reference 35CATALISTcatalist.usVisit source
- Reference 36ACADEMICacademic.oup.comVisit source
- Reference 37IBISWORLDibisworld.comVisit source
- Reference 38YOUGOVyougov.comVisit source
- Reference 39OPENSECRETSopensecrets.orgVisit source
- Reference 40KANTARkantar.comVisit source
- Reference 41INVESTORSinvestors.surveymonkey.comVisit source
- Reference 42NAPOLCOMnapolcom.orgVisit source
- Reference 43STATISTAstatista.comVisit source
- Reference 44NORCnorc.orgVisit source
- Reference 45ADIMPACTadimpact.comVisit source
- Reference 46GREENBOOKgreenbook.orgVisit source
- Reference 47WINGIAwingia.orgVisit source
- Reference 48QUIRKSTRATEGIESquirkstrategies.comVisit source
- Reference 49MONMOUTHmonmouth.eduVisit source
- Reference 50ROYMORGANroymorgan.comVisit source
- Reference 51BLSbls.govVisit source
- Reference 52GLASSDOORglassdoor.comVisit source
- Reference 53INDEEDindeed.comVisit source
- Reference 54UPWORKupwork.comVisit source
- Reference 55ZIPRECRUITERziprecruiter.comVisit source
- Reference 56CENSUScensus.govVisit source
- Reference 57INSIGHTSASSOCIATIONinsightsassociation.orgVisit source
- Reference 58LINKEDINlinkedin.comVisit source
- Reference 59SHRMshrm.orgVisit source
- Reference 60ONSons.gov.ukVisit source
- Reference 61MONSTERmonster.comVisit source
- Reference 62MRSmrs.org.ukVisit source
- Reference 63RASMUSSENREPORTSrasmussenreports.comVisit source
- Reference 64WORLDVALUESSURVEYworldvaluessurvey.orgVisit source
- Reference 65MEDIAPOSTmediapost.comVisit source
- Reference 66EUROPAeuropa.euVisit source
- Reference 67HARRISINTERACTIVEharrisinteractive.comVisit source