The evidence index: prediction-market accuracy studies, 1988 to now
Arguments about prediction markets are usually conducted without the evidence in the room. One side cites a famous election result; the other cites a famous failure; nobody produces the list.
This page is the list.
This index covers the published research on how accurate prediction-market prices have been, from the academic election markets of 1988 through the venue-scale studies of 2026. Each entry records what was studied, on what sample, what the study found, and whether it has been through peer review. Every entry was verified against the paper or its publisher record before inclusion, and nothing appears here on the strength of a summary elsewhere.
The index answers a question in the aggregate. Across roughly four decades of study, prediction-market prices have been accurate enough to beat polls and expert forecasts more often than not — and inaccurate in specific, repeatable ways: at the extremes of probability, at long horizons, and on thin books. The research covers elections, corporate forecasting, sports, films, macroeconomic data and the replicability of science, and it now includes venue-scale studies running to hundreds of millions of trades. It also has a shape worth naming: elections and U.S. venues are heavily overrepresented, and much of the modern large-sample work is still in working-paper form.
The interpretation of this evidence — where it supports the category’s claims and where it undercuts them — is the subject of Are prediction markets accurate?. This page is the underlying record, not the argument.
How to read this index
Sample states what the study actually measured — the number of markets, contracts, races or trades, the setting, and the period. Result is the study’s headline finding in its own terms, kept to what the paper’s own abstract or reported figures support. Status is the publication state as of this page’s last update: a peer-reviewed journal article, a paper in conference proceedings, a working paper, or a preprint. Working papers and preprints are labeled every time they appear and are never treated as equivalent to refereed results.
Verbatim titles, authors and links are in the full citations at the foot of the page. Three naming conventions apply throughout. Academic exchanges run as research instruments are named where they are the object of study, because the venue is the dataset and the reader cannot check the result without knowing which one it was. Commercial prediction-market venues are not named in the entries at all — they are described by type, and their names appear only inside verbatim citation titles, where a title is a quotation and is not reworded. Ordinary firms that ran internal markets on themselves are named, because there the firm is the study’s setting, not a venue being graded. Nothing on this page endorses, ranks or grades any operator.
One disambiguation, because the author-year label collides: two 2006 Wolfers and Zitzewitz working papers are cited across this shelf — “Five Open Questions About Prediction Markets” (NBER WP 12060, 2006a) and “Interpreting Prediction Market Prices as Probabilities” (NBER WP 12200, 2006b). Titles distinguish them in the citations below.
Inclusion is deliberately narrow. An entry qualifies by measuring the accuracy or calibration of prediction-market prices against realized outcomes, or by establishing a condition on which that accuracy demonstrably depends — the depth that produces a price, the wording and settlement rules that define what was graded, or the pressure a price can be put under. Pure trading-performance studies, adjacent markets outside this category, and commentary are not indexed here, however widely cited.
Foundations, 1988–2004
Elections came first. Then play-money exchanges, then the corporate experiments that took the mechanism inside firms.
| Study | Sample | Result | Status |
|---|---|---|---|
| Forsythe, Nelson, Neumann & Wright (1992) | The Iowa Political Stock Market, 1988 U.S. presidential election | The market “worked extremely well, dominating opinion polls in forecasting the outcome,” despite traders individually exhibiting substantial judgment biases; the authors attribute this to marginal, not average, traders setting prices | Journal (1992) |
| Forsythe, Rietz & Ross (1999) | Survey of election-market evidence: the 1988 and 1992 Iowa markets, a 1993 Canadian election market, plus laboratory sessions | Individual traders are biased and error-prone in consistent ways — a documented “wishful thinking” effect — yet the markets predicted their elections well; relatively few marginal traders suffice | Journal (1999) |
| Servan-Schreiber, Wolfers, Pennock & Galebach (2004) | 2003 NFL season; a real-money exchange run against a play-money exchange, same questions | The two sets of forecasts were approximately equally well calibrated — money was not the active ingredient | Journal (2004) |
| Pennock, Lawrence, Giles & Nielsen (2001) | 161 expired binary claims on a play-money science-and-technology exchange, priced 30 days before expiry, grouped into six price ranges | Prices “strongly correlate with observed outcome frequencies” — a calibration result on a market with no money at risk | Journal, letter (2001) |
| Pennock, Lawrence, Nielsen & Giles (2001) | Three online games: 50 films (opening weekend) and 109 films (four-week gross), 135 award contracts, 161 binary claims, and 32 of 41 motor races 1999–2001 | Correlation of 0.940 between prices and opening-weekend box office, 0.978 over four weeks; in the 2000 Oscars the highest-priced nominee won all eight categories, against seven of eight for the Wall Street Journal’s published pre-ceremony poll of actual Academy voters | Conference proceedings (2001) |
| Spann & Skiera (2003) | Three business applications: 152 films on a play-money exchange (2000–01), 81 films and 11 music singles on a purpose-built market, and a mobile-services market run with 20 employees of a German mobile-network firm | Mean absolute percentage error of 30.96% on film revenue; against a commercial forecasting service the market was level (69 versus 71 exact hits over 140 films); the internal corporate market beat all four extrapolation models, though not significantly on its small sample. The paper’s 71.1% comparison figure is a different study’s pre-release model on an unrelated 10-film sample, not a same-sample test | Journal (2003) |
| Plott & Chen (2002) | Twelve internal forecasting events over three years inside Hewlett-Packard, roughly 20–30 participants each, forecasting product sales three months out | The market beat the company’s official forecast in six of the eight events where an official forecast existed, and correctly called the direction of the official forecast’s error in all eight; price-derived probability distributions were consistent with outcomes | Working paper (2002) |
Markets against polls, models and panels
The studies that put market prices head-to-head against polls, statistical models and forecaster panels.
| Study | Sample | Result | Status |
|---|---|---|---|
| Berg, Nelson & Rietz (2008) | 964 national polls against market prices across five U.S. presidential elections, 1988–2004 | The market was closer to the eventual result 74% of the time; more than 100 days out it beat the polls in all five elections | Journal (2008) |
| Erikson & Wlezien (2008) | The same election-market record, re-analyzed | A poll records preferences on the day it is taken, not a forecast of election day; corrected for that, properly projected polls beat raw market prices | Journal (2008) |
| Rothschild (2009) | 74 races over the final 130 days of the 2008 U.S. cycle — 50 Electoral College and 24 Senate contests — market prices against poll-based forecasts | Raw market prices were slightly less accurate than a debiased poll-based model; once the market prices were themselves debiased for the longshot effect, they were more accurate and more informative than the debiased polls, most clearly early in the cycle and in uncertain races | Journal (2009) |
| Atanasov et al. (2017) | A multi-year geopolitical forecasting tournament: more than 2,400 forecasters, 261 questions, randomized assignment | Market prices beat the simple average of individual judgments; statistically aggregated forecaster teams beat the market, with the largest advantage early in long-duration questions | Journal (2017) |
| Sethi, Seager, Cai, Benjamin & Morstatter (2021) | 13 U.S. battleground states, daily forecasts from a statistical model and a market, 1 April – 2 November 2020 | Accuracy differed systematically over time: the market was better months out, the model better as the election approached; a simple average of the two beat either alone overall | Preprint (2021) |
| Gruca & Rietz (2025) | The Iowa Electronic Markets in the 2024 U.S. presidential election, set against the venue’s published election-eve accuracy record | On 29 September 2024 the market read 85.7% for a popular-vote winner that lost, and its thin vote-share market showed a 9-point margin the wrong way; the same venue’s presidential election-eve forecasts (10 markets, 22 contracts, through 2020) average an absolute error of 1.34 percentage points, a figure the paper reproduces from the venue researchers’ own earlier accuracy compilation | Journal (2025) |
| Clinton & Huang (2025) | More than 2,500 markets across four venues in the final five weeks of the 2024 U.S. cycle, covering some $2.4 billion in transactions | Accuracy varied widely between venues; identical contracts traded at different prices, with the gaps widest in the final two weeks | Working paper (2025) |
Where an incumbent forecast already existed
These are the studies with a benchmark already in place. A firm that already forecasts its own sales is the cleanest test the literature has.
| Study | Sample | Result | Status |
|---|---|---|---|
| Gürkaynak & Wolfers (2006) | The first two and a half years of a market in macroeconomic data releases — 153 releases — against consensus survey forecasts | Market-based central tendencies were similar to, but more accurate than, the surveys; market-implied probability distributions were well calibrated, and forecastable forecast errors present in surveys were absent from the market | Conference volume (2006) |
| Cowgill & Zitzewitz (2015) | Years of internal prediction markets at Google, Ford and a third large firm, forecasting demand, deadlines and product outcomes | Despite thin participation and weak incentives, the markets improved on the firms’ own expert forecasts by as much as a 25% reduction in mean squared error; an optimism bias was present and shrank as traders gained experience and less-skilled traders exited | Journal (2015) |
Forecasting the replicability of science
The strand with the cleanest control: every one of these markets was run against a survey of the same experts, on the same questions, at the same time.
| Study | Sample | Result | Status |
|---|---|---|---|
| Dreber et al. (2015) | Markets over 44 psychology studies undergoing replication, 41 completed; run in two rounds (47 active traders over 23 studies, then 45 over 21) | The markets correctly predicted 29 of 41 replication outcomes (71%, p = 0.012); a survey of the same participants got 23 of 40 (58%), not distinguishable from chance | Journal (2015) |
| Camerer et al. (2016) | 18 laboratory experiments from two leading economics journals, 2011–2014, replicated at high power | 11 of 18 (61%) replicated. The survey’s correlation with outcomes was moderately stronger than the markets’ on this sample (0.52 against 0.30). Both methods “correctly predicted” 61% — but the paper’s own supplement notes every belief on both sides sat above 50%, so neither ever forecast a failure and that hit-rate is simply the base rate | Journal (2016) |
| Camerer et al. (2018) | 21 social-science experiments published in two general-science journals, 2010–2015, replicated at roughly five times the original sample sizes | 13 of 21 (62%) replicated, with the market’s correlation to outcomes slightly stronger than the survey’s (0.84 against 0.76). Both methods called 86% of outcomes correctly — a figure computed from the paper’s supplementary tables rather than stated in its text, and reported in those terms by the same group in Forsell et al. (2019) | Journal (2018) |
| Forsell et al. (2019) | 24 effects in a large multi-laboratory replication project, forecast by markets and by survey | The markets correctly predicted 18 of 24 outcomes (75%, correlation 0.755) against 16 of 24 (67%) for the survey; markets in relative effect size attracted little trading and did not work | Journal (2019) |
Where calibration fails, and where it holds
The results that bound the claim — where prices are miscalibrated, and under what conditions the mechanism underperforms. Note that these results do not agree with each other: the longshot bias is the field’s most replicated finding and also has a published counterexample below, which is why this section is a set of conditions rather than a set of laws.
| Study | Sample | Result | Status |
|---|---|---|---|
| Manski (2006) | Theory, for all-or-nothing contracts | Under heterogeneous beliefs, a price reveals nothing about the dispersion of traders’ beliefs and only partially identifies their central tendency — the mean belief lies in an interval whose midpoint is the price. The interval is widest near 0.5 and narrowest at the extremes, so prices near zero or one are the most informative about mean belief | Journal (2006) |
| Wolfers & Zitzewitz (2006a) | Survey of open questions, with calibration data from the Iowa markets and documented manipulation episodes | Names low-probability events as the exception to the generally good record; documents 2004 manipulation attempts whose price impact reversed within 24 hours | Working paper (2006) |
| Wolfers & Zitzewitz (2006b) | Theory, on when a market price can be read as a probability | Under plausible conditions a market price closely tracks the central tendency of the beliefs of the traders in it — the result that licenses reading a price as a probability at all; the same paper finds the structural pricing biases Manski’s bound admits to be generally small | Working paper (2006) |
| Goel, Reeves, Watts & Pennock (2010) | More than 7,000 NFL games, nearly 20,000 baseball games and about 100 film releases | In football, the market was roughly 3% more accurate than a simple three-parameter statistical model and about 1% better than a poll of enthusiasts; in baseball the market and the simple model effectively tied | Conference proceedings (2010) |
| Healy, Linardi, Lowery & Ledyard (2010) | Laboratory markets with only three traders each, across four mechanisms, in a simple environment (one binary event) and a complex one (three correlated events, eight securities) | The continuous double auction performed well in the simple environment and worst in the complex one — traders concentrated on a few securities and the rest were badly mispriced; an iterated poll performed best when information was complex | Journal (2010) |
| Page & Clemen (2013) | 1,787 markets on a public event-market platform, calibration measured against realized frequencies and split by time to resolution | Close to resolution, prices are reasonably well calibrated; far from resolution they are biased — low probabilities too high, near-certainties too low. Pooling all durations, a price of 0.20 was on average associated with a realized frequency of 15.3% and a price of 0.80 with 87.4%; the S-shaped bias was measurably stronger for markets more than 100 days from expiry | Journal (2013) |
| Tetlock (2008) | Three years of intraday data on one-day binary contracts on sports and financial events, on an online exchange of that era | Liquidity did not reduce — and sometimes increased — deviations of prices from outcomes; the more liquid securities were not better calibrated and showed poorer resolution | Working paper (2008) |
| Berg & Rietz (2019) | Iowa Electronic Markets contracts on monthly financial outcomes — share-price levels and industry performance — run and reinitialized monthly over more than five years | No longshot bias appears in these markets, and no overconfidence effect at short horizons; some intermediate-horizon inefficiency is present | Journal (2019) |
Manipulation and price integrity
Whether a price can be pushed, and what it costs to push it — the condition the accuracy results quietly assume.
| Study | Sample | Result | Status |
|---|---|---|---|
| Hanson & Oprea (2009) | Theory, on a manipulator holding a private preference about where the price should sit | When other traders are uncertain about the manipulator’s target, the average target has no effect on price — and raising the uncertainty about it can raise average accuracy, by increasing the value of informed trading | Journal (2009) |
| Hanson, Oprea & Porter (2006) | Laboratory double-auction market, 96 subjects across eight sessions; in half, some traders were paid in proportion to the closing price, as common knowledge | Manipulation had no significant effect on price accuracy: traders without the incentive adjusted the terms on which they would trade, and the adjustment absorbed the push | Journal (2006) |
| Deck, Lin & Porter (2013) | Laboratory markets where the price feeds a decision-maker, against well-funded single-minded manipulators | The counter-result: such manipulators “can in fact destroy a prediction market’s ability to aggregate informative prices and mislead those who are making forecasts.” The damage concentrated in executed trades; standing bids and asks stayed informative | Journal (2013) |
| Rhode & Strumpf (2008) | A field experiment placing planned, random investments worth around 2% of total volume on an academic exchange in 2000, set beside a century of observational data (1880–1944) and a 2004 episode on an offshore exchange | The attacks moved prices, the moves were quickly undone, and prices returned close to their previous levels; little evidence that such markets can be manipulated systematically beyond short time periods | Working paper (2008) |
| Rasooly & Rozzi (2025) | A field experiment randomly shocking prices across 817 separate markets on a play-money venue, with hourly price tracking; plus a smaller, uncontrolled one-week follow-up on the same venue’s dollar-redeemable markets | The shocks’ effects were still visible 60 days later, fading over time rather than reverting outright; markets with more traders, greater volume and an external probability estimate were harder to manipulate. The dollar-redeemable follow-up found the same pattern — substantial reversion, with the manipulation still clearly detectable | Working paper (2025) |
| Dai, Jia & Yu (2026) | Five-minute and fifteen-minute crypto settlement contracts on a large on-chain venue | Settlement-time order flow in the underlying spikes, and the price reverses immediately afterward, transferring value from ordinary traders to manipulators; the fifteen-minute contract shows almost none of it — lengthening the settlement window removes the attack | Preprint (2026) |
The modern venue-scale record, 2024–2026
The category’s first large-sample era. Note the status column: most of this evidence is not yet through peer review.
| Study | Sample | Result | Status |
|---|---|---|---|
| Bürgi, Deng & Whelan (2025) | More than 300,000 contracts on a large licensed U.S. venue | Prices sharpen as markets approach close; low-price contracts win far less often than their prices imply — the longshot bias, alive at modern scale. Across the sample takers averaged a 31% loss against makers’ 12%, and lose hardest on the cheapest contracts | Working paper (2025) |
| Le (2026) | 353 million trades across 429,000 contracts on two of the category’s largest venues | Calibration is structured rather than uniform: a horizon effect, domain-specific biases, and political prices chronically compressed toward 50 | Preprint (2026) |
| Dubach (2026) | 30 billion order-book events over 52 days on the largest on-chain venue, on a pre-registered panel of 600 markets | Eight stylized facts, including a longshot spread premium, broad maker participation carried by a concentrated tail, and a depth profile explained by a market’s duration, price level and volume — with no residual effect of time-to-close once those are controlled | Preprint (2026, v2) |
| Gebele & Matthes (2026) | More than 100,000 events across ten venues, 2018–2025 | Semantically equivalent markets — the same question, worded, sourced and resolved differently — show persistent execution-aware price deviations of 2–4% that do not close | Preprint (2026) |
| Adegbenro (2026) | Roughly 6,000 contracts covering two under-covered regions | What venues list tracks “settlement legibility” — whether a question can be worded, sourced and credibly resolved — rather than public importance; African inventory concentrated in football, Latin American inventory dominated by one country | Preprint (2026) |
| Bartlett & O’Hara (2026) | 41.6 million trades on a large licensed U.S. venue | Single-name markets carry more informed price impact than broad-based ones, yet effective spreads are only modestly wider; a frequency-magnitude decomposition attributes the gap to traders systematically favouring the Yes side in markets that predominantly settle No | Working paper (2026) |
| Tsang & Yang (2026) | On-chain transaction data for the 2024 U.S. presidential market on the largest on-chain venue | A decomposition of reported volume cuts the headline October figure by roughly 60%; measured price-deviation half-lives fell from hours to under a minute over the period, and large-account activity showed capital flowing into both sides at once — consistent with disagreement rather than manipulation | Preprint (2026) |
Where to start reading
These are surveys and reference works, not individual results. Start here if you want the field rather than a finding.
| Work | Scope | Status |
|---|---|---|
| Wolfers & Zitzewitz (2004) | The field’s standard survey: election markets, sports, economic data releases; finds market forecasts fairly accurate and better than most moderately sophisticated benchmarks | Journal (2004) |
| Arrow et al. (2008) | A short statement co-signed by twenty-two economists and legal scholars, citing mounting evidence of lower prediction error than conventional methods and urging regulators to clear a path | Journal, policy forum (2008) |
| Tziralis & Tatsiopoulos (2007) | An extended literature review classifying 155 articles published between 1990 and 2006 | Journal (2007) |
| Snowberg, Wolfers & Zitzewitz (2013) | The handbook chapter on prediction markets for economic forecasting, including the contract-specification cases | Book chapter (2013) |
What this index is not
It is not a scoreboard of venues. Several entries measure a specific exchange’s prices, and where that exchange is an academic research market it is named, because the result is uncheckable otherwise; commercial venues appear only inside verbatim citation titles. Either way nothing here ranks operators, and no operator’s conduct is assessed on this page — a study reporting that a venue’s prices missed is a finding about the mechanism under those conditions, not a judgement about the venue.
It is not complete, and it is not neutral in coverage, because the literature is not. Elections and U.S. venues dominate; sports, films and corporate forecasting are thinly covered by comparison; large categories of question appear to have no published accuracy study at all. An index inherits the bias of the field it indexes, and this one states that rather than smoothing it over.
And it is not final. The venue-scale entries are the newest and the least settled: several are working papers or preprints that may be revised, published, or contradicted. This page is swept for supersession on the same cycle as the rest of this shelf, with statuses re-checked and superseded results corrected in place; the date at the top records the last such review.
Send corrections and additions to research@majjha.com. A study missing from this list is a gap in the record, not a gap in the argument — and it will be added.
Corrections
Material corrections are recorded here, dated, with the wording they replaced.
2026-08-06 — Rhode & Strumpf (2008), Sample. The row read “planned, randomly timed investments.” The paper’s own term is “planned, random investments”: what was randomized was the side taken — Democratic or Republican, fixed by the hundredth digit of the previous day’s Dow — not the timing, which ran to a schedule of eleven trading episodes, roughly ten days apart, beginning 110 days before the election. The 2% share of total volume and the study’s result are unchanged.
2026-08-06 — Dai, Jia & Yu (2026), Result. The row read “Settlement-time order flow in the underlying spikes and reverses immediately afterward.” What reverses is the price, not the order flow. The paper’s own words are “settlement-time spot order flow spikes, causing large price reversals after settlement,” and the distinction is load-bearing: the price snapping back is the paper’s evidence that the settlement-window flow carried no information, which is a different claim from the flow itself unwinding. The direction of the value transfer, the five-minute versus fifteen-minute contrast, and the study’s publication status did not move.
Full citations
1. Forsythe, R., Nelson, F., Neumann, G. R., Wright, J. (1992). “Anatomy of an Experimental Political Stock Market.” American Economic Review 82(5): 1142–1161. econpapers.repec.org 2. Forsythe, R., Rietz, T. A., Ross, T. W. (1999). “Wishes, expectations and actions: a survey on price formation in election stock markets.” Journal of Economic Behavior & Organization 39(1): 83–110. ideas.repec.org 3. Servan-Schreiber, E., Wolfers, J., Pennock, D. M., Galebach, B. (2004). “Prediction Markets: Does Money Matter?” Electronic Markets 14(3). tandfonline.com 4. Pennock, D. M., Lawrence, S., Giles, C. L., Nielsen, F. Å. (2001). “The Real Power of Artificial Markets.” Science 291(5506): 987–988. science.org 5. Pennock, D. M., Lawrence, S., Nielsen, F. Å., Giles, C. L. (2001). “Extracting Collective Probabilistic Forecasts from Web Games.” Proceedings of the Seventh ACM SIGKDD International Conference on Knowledge Discovery and Data Mining: 174–183. clgiles.ist.psu.edu 6. Spann, M., Skiera, B. (2003). “Internet-Based Virtual Stock Markets for Business Forecasting.” Management Science 49(10): 1310–1326. pubsonline.informs.org 7. Plott, C. R., Chen, K.-Y. (2002). “Information Aggregation Mechanisms: Concept, Design and Implementation for a Sales Forecasting Problem.” Working paper, California Institute of Technology Social Science Working Paper 1131. hss.caltech.edu 8. Berg, J., Nelson, F., Rietz, T. (2008). “Prediction market accuracy in the long run.” International Journal of Forecasting 24(2): 285–300. sciencedirect.com 9. Erikson, R. S., Wlezien, C. (2008). “Are Political Markets Really Superior to Polls as Election Predictors?” Public Opinion Quarterly 72(2): 190–215. academic.oup.com 10. Rothschild, D. (2009). “Forecasting Elections: Comparing Prediction Markets, Polls, and Their Biases.” Public Opinion Quarterly 73(5): 895–916. academic.oup.com 11. Atanasov, P., et al. (2017). “Distilling the Wisdom of Crowds: Prediction Markets vs. Prediction Polls.” Management Science 63(3). pubsonline.informs.org 12. Sethi, R., Seager, J., Cai, E., Benjamin, D. M., Morstatter, F. (2021). “Models, Markets, and the Forecasting of Elections.” Preprint, arXiv:2102.04936. arxiv.org 13. Gruca, T. S., Rietz, T. A. (2025). “Iowa Electronic Markets: Forecasting the 2024 US Presidential Election.” PS: Political Science & Politics 58(2). cambridge.org 14. Clinton, J. D., Huang, T. (2025). “Prediction Markets? The Accuracy and Efficiency of $2.4 Billion in the 2024 Presidential Election.” Working paper, SocArXiv. ideas.repec.org 15. Gürkaynak, R. S., Wolfers, J. (2006). “Macroeconomic Derivatives: An Initial Analysis of Market-Based Macro Forecasts, Uncertainty, and Risk.” NBER International Seminar on Macroeconomics 2005: 11–64. Also NBER Working Paper 11929. nber.org 16. Cowgill, B., Zitzewitz, E. (2015). “Corporate Prediction Markets: Evidence from Google, Ford, and Firm X.” The Review of Economic Studies 82(4): 1309–1341. academic.oup.com 17. Dreber, A., Pfeiffer, T., Almenberg, J., et al. (2015). “Using prediction markets to estimate the reproducibility of scientific research.” Proceedings of the National Academy of Sciences 112(50): 15343–15347. pnas.org 18. Camerer, C. F., Dreber, A., Forsell, E., et al. (2016). “Evaluating replicability of laboratory experiments in economics.” Science 351(6280): 1433–1436. science.org 19. Camerer, C. F., Dreber, A., Holzmeister, F., et al. (2018). “Evaluating the replicability of social science experiments in Nature and Science between 2010 and 2015.” Nature Human Behaviour 2(9): 637–644. nature.com 20. Forsell, E., Viganola, D., Pfeiffer, T., et al. (2019). “Predicting replication outcomes in the Many Labs 2 study.” Journal of Economic Psychology 75: 102117. doi.org 21. Manski, C. F. (2006). “Interpreting the Predictions of Prediction Markets.” Economics Letters 91(3): 425–429. Also NBER Working Paper 10359. nber.org 22. Wolfers, J., Zitzewitz, E. (2006a). “Five Open Questions About Prediction Markets.” Working paper, NBER Working Paper 12060. nber.org 23. Wolfers, J., Zitzewitz, E. (2006b). “Interpreting Prediction Market Prices as Probabilities.” Working paper, NBER Working Paper 12200. nber.org 24. Goel, S., Reeves, D. M., Watts, D. J., Pennock, D. M. (2010). “Prediction Without Markets.” Proceedings of the 11th ACM Conference on Electronic Commerce. dl.acm.org 25. Healy, P. J., Linardi, S., Lowery, J. R., Ledyard, J. O. (2010). “Prediction Markets: Alternative Mechanisms for Complex Environments with Few Traders.” Management Science 56(11): 1977–1996. ideas.repec.org 26. Page, L., Clemen, R. T. (2013). “Do Prediction Markets Produce Well-Calibrated Probability Forecasts?” The Economic Journal 123(568): 491–513. academic.oup.com 27. Tetlock, P. C. (2008). “Liquidity and Prediction Market Efficiency.” Working paper. business.columbia.edu 28. Berg, J. E., Rietz, T. A. (2019). “Longshots, Overconfidence and Efficiency on the Iowa Electronic Market.” International Journal of Forecasting 35(1): 271–287. ideas.repec.org 29. Hanson, R., Oprea, R. (2009). “A Manipulator Can Aid Prediction Market Accuracy.” Economica 76(302): 304–314. doi.org 30. Hanson, R., Oprea, R., Porter, D. (2006). “Information aggregation and manipulation in an experimental market.” Journal of Economic Behavior & Organization 60(4): 449–459. ideas.repec.org 31. Deck, C., Lin, S., Porter, D. (2013). “Affecting policy by manipulating prediction markets: Experimental evidence.” Journal of Economic Behavior & Organization 85: 48–62. doi.org 32. Rhode, P. W., Strumpf, K. S. (2008). “Manipulating Political Stock Markets: A Field Experiment and a Century of Observational Data.” Working paper, Natural Field Experiments No. 00325. ideas.repec.org 33. Rasooly, I., Rozzi, R. (2025). “How manipulable are prediction markets?” Working paper, arXiv:2503.03312. arxiv.org 34. Dai, D., Jia, R., Yu, S. (2026). “Settlement Manipulation in Prediction Markets.” Preprint, arXiv:2606.31675. arxiv.org 35. Bürgi, C., Deng, W., Whelan, K. (2025). “Makers and Takers: The Economics of the Kalshi Prediction Market.” Working paper, UCD Centre for Economic Research WP25/19. ucd.ie 36. Le, N. A. (2026). “Decomposing Crowd Wisdom: Domain-Specific Calibration Dynamics in Prediction Markets.” Preprint, arXiv:2602.19520. arxiv.org 37. Dubach, P. D. (2026). “The Anatomy of a Decentralized Prediction Market: Microstructure Evidence from the Polymarket Order Book.” Preprint, arXiv:2604.24366. arxiv.org 38. Gebele, J., Matthes, F. (2026). “Semantic Non-Fungibility and Violations of the Law of One Price in Prediction Markets.” Preprint, arXiv:2601.01706. arxiv.org 39. Adegbenro, A. (2026). “What Prediction Markets Can See: Market Formation, Settlement Legibility, and the Geography of Tradable Uncertainty in Africa and Latin America.” Preprint, arXiv:2606.17503. arxiv.org 40. Bartlett, R., O’Hara, M. (2026). “Adverse Selection in Prediction Markets: Evidence from Kalshi.” Working paper, Stanford Rock Center for Corporate Governance No. 266. law.stanford.edu 41. Tsang, K. P., Yang, Z. (2026). “The Anatomy of a Blockchain Prediction Market: Polymarket in the 2024 U.S. Presidential Election.” Preprint, arXiv:2603.03136. arxiv.org 42. Wolfers, J., Zitzewitz, E. (2004). “Prediction Markets.” Journal of Economic Perspectives 18(2): 107–126. aeaweb.org 43. Arrow, K. J., et al. (2008). “The Promise of Prediction Markets.” Science 320(5878): 877–878. science.org 44. Tziralis, G., Tatsiopoulos, I. (2007). “Prediction Markets: An Extended Literature Review.” The Journal of Prediction Markets 1(1). ubplj.org 45. Snowberg, E., Wolfers, J., Zitzewitz, E. (2013). “Prediction Markets for Economic Forecasting.” Handbook of Economic Forecasting, Vol. 2, Elsevier. Also NBER Working Paper 18222. nber.org