Are Betting Odds Becoming More Trusted Than Polls?
It depends on what you mean by “trusted” and which contest you’re asking about. In the 2024 US presidential election, Polymarket priced Donald Trump at roughly 57% to win on election eve while polling averages showed a toss-up — markets called it correctly and polls did not, which is the case most often cited when people say betting markets have surpassed traditional polling. In 2016, prediction markets failed badly: they gave Brexit roughly a 30% chance and gave Trump’s general-election win about 25% a month out. Both predictions were wrong.
Academic research going back to Wolfers and Zitzewitz finds that markets generally outperform polls on average, but with documented failure modes — thin liquidity in primaries, single-bettor manipulation in 2024, and the 2016 misses on contests that polling also got wrong. The honest answer isn’t “yes” or “no.” It’s that markets and polls measure different things, both have failed in important contests, and the question of which is “more trusted” is partly empirical and partly sociological.
This guide walks through what each system actually measures, the empirical accuracy record across major recent contests, the broader trust-in-institutions context that frames why the question keeps getting asked, and a practical framework for when to weight each more heavily.
Markets are faster (real-time price updates as news breaks) and have outperformed polls in some recent contests; polls are more methodologically rigorous and capture a broader, more representative sample. Use both. Trust the one whose failure mode you’re least exposed to in the specific contest you care about.
The 2024 Election: When Markets Beat Polls
The case for prediction markets in 2024 is concrete and widely cited. On the eve of the November 2024 election, Polymarket priced Donald Trump at roughly 57% to win, while major polling averages showed the race as essentially a coin flip — some models gave Kamala Harris a slight edge. Trump won decisively in the Electoral College with margins in swing states that exceeded most polling estimates. Later analyses noted that prediction markets reached 95% certainty for a Trump win roughly 36 hours before major news outlets called the race.
Markets also reacted faster to events. After the first assassination attempt on Trump in Pennsylvania on July 13, 2024, Polymarket prices moved within hours to reflect a higher Trump probability. Polling averages took days to register the same effect, and in some cases barely moved. When Kamala Harris entered the race on July 21, 2024, prediction markets again adjusted within hours; polling moved more slowly. The speed advantage is real and is part of why markets call major shifts more quickly than polls do.
Polymarket’s 2024 Trump price was potentially affected by a single well-documented French trader who placed tens of millions of dollars in Trump contracts in the weeks before the election. Single-bettor flow can move thin markets meaningfully. The 57% price was right, but it isn’t a clean test of crowd wisdom — it’s a price that reflected, in part, one large directional position alongside the broader market.
The 2016 Counter-Examples: When Markets Failed
The “markets are smarter than polls” framing leans heavily on 2024 because 2016 cuts the other way. In June 2016, prediction markets gave the UK Brexit referendum approximately a 30% chance of passing. Brexit passed. About a month before the November 2016 US election, prediction markets gave Donald Trump roughly a 25% chance of winning. Trump won. In both cases, markets and polling averages were directionally aligned (both expected the status quo outcome), and both were wrong. Markets did not outperform polls in 2016; they failed in tandem.
The honest accuracy record across recent major political contests:
| Contest | Prediction-market call | Polling-average call | Outcome |
|---|---|---|---|
| 2016 UK Brexit referendum | ~30% Brexit (favored Remain) | Favored Remain | Brexit passed (markets wrong) |
| 2016 US Presidential | ~25% Trump a month out | Favored Clinton | Trump won (both wrong) |
| 2020 US Presidential | Favored Biden, narrow margin | Favored Biden, wider margin | Biden won (markets closer) |
| 2024 US Presidential | Polymarket ~57% Trump on eve | Toss-up; some models leaned Harris | Trump won decisively (markets right) |
The record across these four contests is mixed. Markets won 2024 cleanly, were directionally closer in 2020, and failed alongside polls in 2016 (twice). The “markets always beat polls” narrative is overconfident; the “polls are still better” reflex doesn’t survive the 2024 case. Both methods have been wrong on important contests, and neither has a perfect track record.
What Markets and Polls Actually Measure (and Why That Matters)
The accuracy comparison is muddled by the fact that markets and polls measure different things. A poll asks “if the election were held today, who would you vote for” of a representative sample of likely voters; the result is an aggregated stated preference. A prediction market price reflects the equilibrium clearing price between bettors willing to take each side at that probability; the result is an aggregated willingness-to-bet. Those are not the same data point.
Polls have well-known failure modes: declining response rates (from roughly 36% in the 1990s to single digits today), demographic-coverage problems as easy-to-reach voters become an unrepresentative slice, and “shy voter” effects where some respondents don’t truthfully state preferences they expect to be socially sanctioned. The 2016 polling miss is widely understood as a combination of these effects in states where Trump support was understated by stated-preference instruments. The 2024 polling miss is similar in pattern, less so in magnitude.
Markets have their own failure modes. Thin liquidity in primaries and down-ballot races means small bets can move prices meaningfully — primary-election markets are notoriously easy to push. Large directional bettors can distort prices; the 2024 French-trader case is the most-cited example, but the structural concern (a single large position influencing the apparent crowd consensus) is generic. Markets also reflect the demographic of bettors, not voters — and prediction-market participants skew younger, more crypto-fluent, and more politically engaged than the median US voter, which can produce systematic priors that don’t match the electorate.
The clean takeaway: markets and polls are complementary forecasting tools rather than substitutes. Markets price faster and incorporate news quickly; polls measure stated preferences with rigorous methodology and broader demographic reach. Our breakdown of how prediction markets work and where they’re legally accessible covers the platform mechanics that determine market depth and reliability for any given contest.
The Trust Question Beneath the Accuracy Question
“Are betting odds more trusted than polls” is partly an accuracy question (which has been more right) and partly a sociological question (whose trust, in what context, for what purpose). The sociological half is the part that’s been moving fastest.
Three trends explain why polls are losing trust independent of their actual accuracy:
- High-profile misses get airtime; quiet successes don’t. Polling averages have been broadly accurate across hundreds of state and national races over the past decade, but the 2016, 2020, and 2024 high-stakes presidential misses received massive coverage. Prediction-market 2024 wins received similar massive coverage. Public memory of forecasting accuracy is shaped much more by salient cases than by aggregate track records.
- Trust in institutions has cratered across the board. Pew Research Center’s December 2025 data shows public trust in government at roughly 17% — near the lowest level in nearly 70 years of measurement. Polling, as an institution housed in academia and traditional media, sits inside that broader institutional-trust decline. Markets are read as outside that institutional structure, which makes them feel more trustworthy to readers who distrust traditional institutions, regardless of whether the markets are actually more accurate.
- Real-time price movement looks more transparent than methodological adjustments. When a poll’s results change, the change is mediated through methodology choices most readers can’t evaluate. When a prediction-market price moves, the movement is visible second-by-second on a chart. Transparency of process is conflated with reliability of result, even though the two are distinct.
None of these reasons say markets are actually more accurate. They explain why markets feel more trustworthy in the current information environment, even when the underlying accuracy comparison is mixed. Disentangling the felt-trustworthiness from the actual accuracy is the harder analytical task — and one that the broader gambling-regulation discourse heading into 2026 is starting to grapple with as prediction markets move from financial novelty to mainstream forecasting tool.
When to Trust Each (a Practical Framework)
For a reader trying to make sense of a current contest, the practical framework isn’t “pick one and trust it.” It’s “weight each based on the failure modes you’re least exposed to in this specific contest.”
Weight prediction markets more heavily when: (1) the contest is a major event with deep market liquidity (multi-million-dollar volume on the contract); (2) the news cycle is moving fast and stated-preference polling will be days behind; (3) you trust that bettors include enough informed positions to price information that polls miss. The 2024 US presidential general election checked all three boxes; markets calling it correctly was not a coincidence.
Weight polling more heavily when: (1) the contest is a primary, down-ballot race, or international election where market liquidity is thin and easy to move; (2) the question is one of voter preference rather than event probability (state-level subpopulations of voters are something polls measure that markets don’t directly); (3) the polling methodology is recent and uses adjustments for known 2016/2024 misses. Most state-level legislative races and most international contests fall in this bucket.
And weight neither very heavily when: the contest is an outlier (single-issue referendum, unusual structural conditions, an event polls and markets have never covered before), or when independent verification points (expert forecasts, on-the-ground reporting, base rates from analogous historical contests) disagree with both. The right move in those cases is humility — neither the markets nor the polls have a strong track record on contests like this one.
Play Safe: Gambling should be fun, not stressful. Set limits, stick to your budget, and never chase losses. If you or someone you know has a gambling problem, call 1-800-MY-RESET or visit ncpgambling.org. For more resources, see our Responsible Gambling page.
For the underlying source documentation: the 2025 academic comparative paper is on arxiv; CNN’s November 2024 retrospective covers the markets-vs-polls 2024 election story; and Pew Research Center’s trust-in-government data documents the broader institutional-trust decline framing why this question keeps getting asked.
Frequently Asked Questions
Are betting markets more accurate than polls?
Sometimes — and the record is mixed. Prediction markets correctly called the 2024 US presidential election when polling averages had it as a toss-up. They were directionally closer than polls in 2020. They failed alongside polls in 2016 (Brexit and US presidential). On average across academic studies going back to Wolfers and Zitzewitz, markets generally outperform polls modestly, but with documented failure modes (thin liquidity in primaries, single-bettor manipulation, demographic skew of bettors vs voters).
Why did Polymarket call the 2024 election when polls did not?
Polymarket priced Donald Trump at roughly 57% to win on election eve while polling averages showed the race as a toss-up. Markets reacted faster to events (the July 13 assassination attempt and Harris’s July 21 entry both moved Polymarket within hours; polling took days). Markets also priced in information about voter enthusiasm and early-vote patterns that stated-preference polling missed. The 57% price was potentially also affected by a single French trader’s tens-of-millions-of-dollars Trump position — the price was right, but not purely a clean test of crowd wisdom.
What’s wrong with traditional polls?
Polls suffer from declining response rates (~36% in the 1990s to single digits today), demographic-coverage problems as easy-to-reach voters become an unrepresentative slice, and ‘shy voter’ effects where some respondents don’t truthfully state preferences they expect to be socially sanctioned. These are real methodological challenges that affected the 2016 and 2024 misses. They don’t make polls useless — polling averages are still broadly accurate across hundreds of state and national races — but they do mean a single-poll snapshot is less reliable than the historical baseline.
Can I trust prediction markets for political forecasting?
For major contests with deep market liquidity (multi-million-dollar volume) and a fast-moving news cycle, yes — markets are usually a useful signal. For primaries, down-ballot races, and international contests where market depth is thin, markets become much less reliable and small bets can move prices meaningfully. The honest practice is to use both markets and polls as complementary forecasting tools rather than substituting one for the other.
