History of Prediction Markets
How prediction markets developed from election trading into online exchanges, and why earlier markets closed or never opened.
People were trading on elections centuries before prediction markets appeared online. In sixteenth-century Rome, merchants used information from cardinals' attendants to trade on who would become pope. By the nineteenth century, New York newspapers were publishing presidential-election market prices alongside campaign news. Readers could follow those prices as forecasts, while traders stood to gain or lose money when the result was settled.
The history of prediction markets also shows how disputes over the result shaped the rules of trading. After the contested 1876 US election, a leading New York organizer canceled his pools and returned participants' money minus commission. Later organizers made payment depend on who took office and sometimes waited for a concession before settling.
When did prediction markets start?
People traded on public events by 1503, when accounts of papal succession already described the practice as old. Economists Paul Rhode and Koleman Strumpf traced these early prediction markets in their historical survey. The first modern academic market began in 1988 at the University of Iowa, according to the US Commodity Futures Trading Commission's historical timeline.
Trading on papal elections was recorded in Rome by 1503 and prohibited in 1591. New York's organized election markets ran from 1868 through 1940, and Iowa's electronic market opened 48 years later, in 1988.
During the 1549 papal conclave, Roman merchants obtained information about the voting through partnerships with cardinals' attendants. They traded on both the next pope and how long the selection would take. In 1591, Pope Gregory XIV prohibited trading on papal elections and related events, with excommunication as the penalty. Rhode and Strumpf also followed political markets through Italian city-states and later Britain, where people traded on who would hold public office.
Who invented prediction markets?
The early records describe an existing practice without identifying an inventor. For the modern research tradition, the University of Iowa credits John Wright, Robert Forsythe, Forrest Nelson, and George Neumann with devising its 1988 market. It let researchers compare prices formed during a campaign with the eventual election result.
In his own account, Robin Hanson dates his proposal for idea futures, also known as information markets, to autumn 1988. He proposed trading on scientific claims and other disputed questions and acknowledged earlier related ideas. Polymarket and Kalshi arrived decades after these academic projects, so neither was the first prediction market.
How did Wall Street election betting work before polls?
From 1868 through 1940, organized New York markets let people trade contracts on presidential elections, and newspapers published their prices as election news. Organizers known as betting commissioners matched the two sides, held their money, and charged a commission. Rhode and Strumpf reconstructed the markets from thousands of newspaper reports.
The activity moved through poolrooms, the Curb Exchange, and Broadway hotels before specialist Wall Street firms took over. Rhode and Strumpf estimated New York trading on the 1916 election at $165 million in 2002 dollars, more than twice that year's presidential campaign spending. Across 15 elections from 1884 through 1940, the mid-October favorite won 11 times, an underdog won once, and three races were approximately even. Calibration tests the probabilities themselves, asking whether candidates priced at 60%, for example, win about 60% of the time.
New York's mid-October market favorite won 11 of the 15 US presidential elections from 1884 through 1940. Of the remaining four races, three were priced roughly even and one was won by an underdog.
Disputed results could leave traders waiting. After the contested 1876 election, organizer John Morrissey returned customers' stakes minus commission. Later contracts made taking office the condition for market resolution, and commissioners sometimes waited for a concession before paying. In 1916, settlement came on November 23, more than two weeks after the vote.
New York's election-market organizers held money from both sides and paid the winning side once the result was settled, deducting a commission. In the disputed 1876 election, John Morrissey's customers instead received their own stakes back minus commission, regardless of which candidate they had backed.
The decline took decades. In May 1924, the New York Stock Exchange and Curb Market barred members from election trading. New York's legalization of pari-mutuel horse-race betting in 1939 gave customers another place to participate, while newspapers increasingly used scientific polls for election coverage. By 1940, Gallup articles outnumbered market-price reports. Rhode and Strumpf's longer historical survey places the disappearance of the formal US markets largely by 1944, with enforcement and the deaths of leading commissioners also contributing.
How did the Iowa Electronic Markets begin?
The Iowa Electronic Markets began with a real-money university experiment during the 1988 US presidential election. Political scientist John Wright and economists Robert Forsythe, Forrest Nelson, and George Neumann created the Iowa Political Stock Market to study forecasting through trade, according to the University of Iowa's archives.
The first election market had 155 active traders from the university community and $8,123 in trading volume. Its contracts paid 2.5 cents for each percentage point of the popular vote won by a candidate or the group of other candidates. A 40% vote share would therefore produce a one-dollar settlement. Wolfers and Zitzewitz's 2004 survey also recorded a $500 position limit for Iowa's markets.
That contract design made a price a forecast of vote share. A winner-take-all contract instead pays a fixed amount if a candidate wins, so its price concerns the chance of victory.
With Iowa's rate of 2.5 cents per percentage point, the vote-share contract pays $1.30 for a win with 52% of the vote and $1 for a loss with 40% in these hypothetical results. The winner-take-all contract on the same candidate pays $1 for the win and nothing for the loss.
In 1992, staff at the Commodity Futures Trading Commission issued no-action relief for the university's market, followed by a further letter in June 1993. Under those letters, staff would refrain from recommending enforcement for operating the research and teaching market without exchange registration, provided it met the stated conditions.
How accurate were the Iowa Electronic Markets?
Iowa's vote-share forecasts were closer to the eventual two-party result than individual national polls in 74% of 964 comparisons across the five presidential elections from 1988 through 2004. Berg, Nelson, and Rietz's study compared each poll with the market price on its final fieldwork day. The benchmark was an individual poll adjusted to a two-party split, rather than the polling averages or election models used in other market comparisons.
In the hypothetical comparison, a market forecast of 51% is one percentage point from the 52% two-party result, while the poll's 49% is three points away. Across 964 comparisons in the 1988–2004 presidential elections, Iowa's market price on each poll's final fieldwork day was closer to the result 74% of the time.
When did online prediction markets start?
Web-based prediction markets were running by 1994, when the play-money project later known as the Foresight Exchange began under the name Idea Futures. The Iowa Electronic Markets had already used electronic trading for its real-money university experiment in 1988. Robin Hanson dates the web project to 1994 and credits Sean Morgan with starting it. In his account, it was the first online prediction market to let users introduce their own claims. The project won a Prix Ars Electronica award in 1995.
Early web markets brought event listings and trading controls into a desktop browser.
By 2004, Wolfers and Zitzewitz's survey listed real-money TradeSports alongside play-money venues such as NewsFutures and the Hollywood Stock Exchange. The latter focused on film revenue and awards. Goldman Sachs and Deutsche Bank ran Economic Derivatives, which offered markets on forthcoming economic releases, including employment and retail sales.
How have companies used prediction markets?
Companies used internal prediction markets to forecast sales and project deadlines by combining information held by employees. By the time of Wolfers and Zitzewitz's 2004 survey, Hewlett-Packard had tested a printer-sales market that beat its internal forecasts, and Siemens had run a software-project market that predicted a missed deadline while conventional planning tools still indicated completion on time.
By September 2005, more than a thousand Google employees had traded on 146 events, including product launch dates and office openings, according to project manager Bo Cowgill. Participation was free, and Cowgill reported that prices closely matched how often the forecast events occurred.
A later study of corporate prediction markets by Cowgill and Eric Zitzewitz examined Google's markets from 2005 to 2007, Ford's in 2010, and a private materials company's from 2008 to 2013. The forecasts improved on the benchmarks the researchers tested despite limited trading and weak incentives. Prices in Google's and Ford's markets tended to overstate the chance of favorable company outcomes. These pricing errors generally diminished as experienced participants traded against overly optimistic forecasts.
Why did DARPA cancel the Policy Analysis Market?
DARPA, the Pentagon's research agency, canceled the Policy Analysis Market in July 2003 after senators attacked it as a proposed market in terrorist attacks and assassinations. The Defense Department ended the wider FutureMAP program the day after Senators Ron Wyden and Byron Dorgan publicized their objections on July 28. Public trading never began.
The Policy Analysis Market, or PAM, grew out of research DARPA commissioned in 2001 on forecasting geopolitical developments. The planned questions included economic growth, political instability, and military activity. Researchers also wanted to link questions so traders could express how one event might affect another, such as how a troop withdrawal might change regional stability. Robin Hanson, who worked on the project, described a planned focus on eight countries in the Middle East, with limited testing scheduled for September 2003 and public trading for January 2004.
The Defense Department ended the Policy Analysis Market's parent program, FutureMAP, on July 29, 2003, one day after senators objected publicly. Neither the tests planned for September 2003 nor public trading planned for January 2004 took place.
The website helped provoke the controversy. According to Hanson, a sample screen created by Net Exchange president Charles Polk included the assassination of Yasser Arafat and a North Korean missile attack. Hanson argued that the senators misrepresented the planned research as a market on individual attacks. The senators' account emphasized contracts on assassinations, coups, and military strikes, and the prospect of people profiting from them.
Why were box-office futures banned?
Congress banned box-office futures in July 2010 after film-industry groups opposed them. The Motion Picture Association of America and other industry organizations lobbied against the proposed exchanges. Their objections included the chance that people inside the industry could trade on private knowledge of a film or act to influence its results, along with doubts about the reliability of revenue data. They also disputed whether the contracts would help the industry hedge its risks. The prohibition became part of the Dodd-Frank Act.
Supporters proposed using the contracts to offset financial exposure to a film's performance. A distributor expecting revenue from an opening weekend could hold a position that gained value if receipts disappointed. Someone expecting a stronger opening could take the other side. The CFTC, which regulates US derivatives markets, accepted a risk-management purpose when it approved Media Derivatives' opening-weekend contracts for Takers on June 14, 2010.
In this hypothetical, a film distributor keeps half of opening-weekend receipts and takes a position that gains when receipts fall below the expected $40 million and loses when they exceed it. Across receipts of $0 to $80 million, the position narrows the distributor's revenue range from $0–$40 million to $10–$30 million.
Cantor Fitzgerald's contracts for The Expendables received approval on June 28. The firm already owned the Hollywood Stock Exchange, a virtual-currency game where people traded forecasts of movie revenue and awards. Cantor's proposed exchange would let customers trade film-revenue contracts with real money.
Dodd-Frank became law on July 21, 2010. Section 721 excluded box-office receipts from the definition of a commodity, effective June 1, 2010, so both contract approvals lost their legal basis. The legislation also added box-office futures to the statutory prohibition that already covered onion futures.
Media Derivatives and Cantor Fitzgerald received approval for box-office contracts on June 14 and June 28, 2010. Dodd-Frank became law on July 21 and excluded box-office receipts from the definition of a commodity effective June 1, removing the legal basis for both approvals.
A separate provision, Section 745, authorized the CFTC to prohibit certain event contracts on public-interest grounds, including those involving gaming, war, terrorism, or assassination. That authority later shaped the agency's objections to election contracts.
Why did Intrade shut down?
Intrade withdrew from the US market after the Commodity Futures Trading Commission sued it in November 2012. In March 2013, it suspended trading worldwide and payments to customers while investigating possible financial irregularities.
The Dublin-based exchange had become known for its election forecasts, including its state-by-state prices during the 2008 and 2012 presidential campaigns. It also listed contracts on economic releases, gold prices, entertainment, and other events. Reason's contemporary account reported more than a million transactions during 2012.
The agency's November 26 complaint against Intrade and Trade Exchange Network alleged that they had offered prohibited off-exchange options to US customers and made false statements in regulatory filings. It named contracts on economic figures, gold and currency prices, and acts of war.
That day, Intrade told US residents to close their positions by December 23 and withdraw funds by December 31. Any remaining positions would close at the company's assessment of fair market value.
In a November 2013 report for Reason, Katherine Mangu-Ward linked the financial inquiry to nearly $1.5 million in suspect payments to former chief executive John Delaney's personal accounts. Delaney had died on Everest in May 2011. The report said the money should have been held for traders, while also stating that wrongdoing had not been demonstrated.
For someone holding a contract months from settlement, forced closure replaced the eventual outcome with a valuation made while the result was still uncertain. A trader could correctly forecast the event and still receive less than the contract would have paid if held through settlement.
In this hypothetical example of Intrade's forced closure, 100 Yes contracts bought at 40 cents earn $20 when closed at an assessed 60 cents. Holding them until the event happens would produce a $1 settlement per contract and a $60 profit.
Why did the CFTC try to shut down PredictIt?
In August 2022, the Commodity Futures Trading Commission withdrew staff assurances against enforcement for PredictIt, alleging noncompliance with its academic arrangement. The original notice did not identify the breached terms. In March 2023, the agency alleged that Victoria University of Wellington had allowed a for-profit company, Aristotle, to operate the market contrary to those terms, according to the appeals-court opinion.
PredictIt's history began with an academic research arrangement in 2014. CFTC staff issued a no-action letter to Victoria University, promising not to recommend enforcement for operating the proposed small, not-for-profit market without exchange registration. The original terms limited investment to $850 per trader in a contract and participation to 5,000 traders per contract.
The withdrawal set a February 15, 2023 deadline to close positions. Before that deadline, the Fifth Circuit intervened on January 26, blocking the shutdown while it considered the appeal. Its July ruling found the original withdrawal likely arbitrary and capricious because the agency had supplied no explanation. The court also rejected the March letter's attempt to justify the withdrawal after the fact, finding that it violated the earlier injunction. It directed the district court to enter a preliminary injunction while the challenge continued.
The Fifth Circuit blocked PredictIt's shutdown on January 26, 2023, before the CFTC's February 15 deadline to close positions. In July 2023, the court found the agency's withdrawal of its no-action letter likely arbitrary and capricious.
In July 2025, CFTC Letter 25-20 permitted transfer of the academic market to the Prediction Market Research Consortium. It removed the 5,000-trader limit and tied the investment cap to the federal limit on individual campaign contributions, which was $3,500 at that time. The letter clarified the scope of political contracts, excluding war, terrorism, and assassination. The consortium could hire Aristotle International, the market's existing service provider, to run day-to-day operations.
PredictIt's investment cap per trader in a contract rose from $850 under the 2014 CFTC letter to $3,500 under the 2025 letter, a little over four times as much.
Aristotle's exchange entity, Aristotle Exchange DCM, received exchange designation in September 2025. That registration covered the exchange business, while the July letter addressed the terms for servicing PredictIt's academic market.
What happened to Augur?
Users could create and trade event contracts on Ethereum through the Augur prediction market, launched in July 2018. Traders encountered ambiguous contract terms and difficult access. Veil, an interface built on Augur, closed in 2019, while the underlying protocol still hosted trading on the 2020 US election.
Its 2015 white paper described a system of decentralized prediction markets in which REP-token holders reported and disputed outcomes. Those participants put tokens at risk when backing a result and had to interpret the contract and evidence about the event.
A Binance Research case study from 2019 examined a market on Ethereum's price at the end of March. Its deadline came before the time at which the rules said to measure the price. If declared invalid, all three price-range outcomes would settle at one-third of the full contract value. Someone who bought the apparently likely range at a higher price could lose money even if Ethereum finished inside that range. The same analysis described how trading between controlled accounts could inflate apparent activity and attract buyers to a flawed contract.
In this hypothetical Augur position, 300 contracts bought at 70 cents cost $210 and pay $300 if Ethereum finishes in the chosen price range and the market resolves validly. An invalid ruling instead pays one-third of a dollar per contract, returning $100 and turning a potential $90 profit into a $110 loss.
Veil had launched in January 2019 to make participation easier. When it closed, co-founder Paul Fletcher-Hill cited a small user community, difficult onboarding, and trouble combining decentralization with regulatory compliance. Users were told to withdraw their positions before Veil stopped supporting trading.
Augur's 2020 election market exposed another difficulty. In a 2021 firsthand account, Vitalik Buterin described buying most of his position after the election had been decided. Contracts paying one dollar if Donald Trump lost stayed around 85 cents for more than a month afterward. Buterin considered capital tied up until settlement and the technical work needed to participate among the reasons more traders had not corrected the price. He reported that the market ultimately resolved correctly after two dispute rounds.
For more than a month after the 2020 election result was known, Augur contracts paying $1 if Donald Trump lost traded around 85 cents. The shaded band marks the 15-cent difference from the $1 settlement value in this schematic of Vitalik Buterin's account.
Why was Polymarket banned in the US in 2022?
On January 3, 2022, Polymarket settled charges from the Commodity Futures Trading Commission for offering event contracts without the required registration. The settlement required a $1.4 million penalty and the wind-down of noncompliant markets. By January 24, The Block reported that US users could still view markets but were blocked from trading.
The order identified the operator as Blockratize, doing business as Polymarket, and dated its operations to approximately June 2020. Its contracts asked whether future events would happen, including cryptocurrency-price changes and election results. The CFTC classified them as swaps and found that the operator needed to register as a trading venue. The order applied that requirement to the contracts despite their use of blockchain software.
How did Polymarket become legal in the US?
Polymarket acquired a regulated US exchange and clearinghouse through its purchase of QCEX, completed on July 21, 2025. The company presented the acquisition as its route back to the US market. Polymarket and Polymarket US refer to the international platform and its US exchange business.
How did Kalshi become legal?
The CFTC, the US derivatives regulator, registered Kalshi as an exchange in November 2020. A district-court ruling in September 2024 then let it list contracts on which party would control a chamber of Congress. The ruling overturned the agency's prohibition on those contracts.
Kalshi's registration made it a designated contract market, the regulatory category for an exchange operating under the Commodity Exchange Act. The CFTC had previously registered HedgeStreet in February 2004 for internet trading in binary options. HedgeStreet became Nadex in 2009.
In April 2012, the CFTC prohibited Nadex's proposed federal-election contracts, finding that they involved gaming and were contrary to the public interest. Exchange registration still left individual contracts subject to regulatory review.
When Kalshi proposed contracts on control of Congress, the CFTC rejected them in September 2023, citing gaming, activity unlawful under state law, and the public interest. In Kalshi v. CFTC, the district court focused on the underlying events. It held that elections and control of Congress fell outside the relevant statutory categories of gaming and unlawful activity, and set the prohibition aside in September 2024.
HedgeStreet registered as an exchange in 2004, but the CFTC prohibited its election contracts in 2012, after it had become Nadex. Kalshi registered in 2020, faced rejection of its congressional-control contracts in 2023, and obtained a court ruling overturning that rejection in 2024.
Kalshi listed the contracts on September 12. They traded for about eight hours before the appeals court paused trading while it considered a stay. On October 2, the D.C. Circuit declined to keep the judgment on hold, finding that the CFTC had not shown the necessary irreparable harm. Election trading could proceed during the appeal. The CFTC later requested dismissal, which the court granted on May 7, 2025, leaving the district court's ruling in place.
When did prediction markets become legal in the US?
The Iowa Electronic Markets received academic no-action relief in 1992, followed by HedgeStreet's exchange registration in 2004 and PredictIt's no-action letter in 2014. Kalshi gained exchange registration in 2020 and court decisions allowing its congressional-control contracts to trade in 2024. The CFTC's appeal was dismissed on May 7, 2025, leaving the district court's election-contract ruling in place.
| Date | Decision | What changed |
|---|---|---|
| 1992 | CFTC staff issued no-action relief for the Iowa Electronic Markets. | The academic market could operate within stated conditions without staff recommending enforcement for lack of exchange registration. |
| February 2004 | HedgeStreet became a registered exchange and clearing organization. | It could offer internet trading in binary options under that regulatory framework. |
| October 2014 | CFTC staff issued a no-action letter to Victoria University of Wellington for PredictIt. | Staff would not recommend enforcement for operating the proposed small academic market without exchange registration, subject to the letter's conditions. |
| November 2020 | Kalshi received exchange designation. | It became a regulated contract market, with proposed products still subject to review. |
| September and October 2024 | A district court overturned the congressional-control prohibition, and the appeals court declined a stay. | Kalshi could list those election contracts and continue trading them during the appeal. |
| May 2025 | The appeals court granted the CFTC's request to dismiss its appeal on May 7. | The appeal ended, leaving the district court's ruling on Kalshi's congressional-control contracts in place. |
When did prediction markets become popular?
The 2024 US presidential election brought Polymarket and Kalshi into widespread election coverage. A court victory that autumn let Kalshi list election contracts on a registered US exchange weeks before the vote. By November, Reuters reported that mainstream news sites were citing market odds and that election trading had drawn billions of dollars. Its reporting also captured disagreement over whether prices reflected useful information or the influence of large positions.
Market forecasts appear both in newspaper coverage and in the event contracts readers can follow on a phone.
New York's election markets supplied newspaper forecasts before scientific polling. Intrade later drew attention for its forecasts of the 2008 and 2012 elections. Political enthusiasts and campaign reporters followed its prices closely in 2012, according to The Washington Post.
Joshua Clinton and TzuFeng Huang's 2025 research preprint studied more than 2,500 political markets across the Iowa Electronic Markets, Kalshi, PredictIt, and Polymarket in the final five weeks of the 2024 campaign. The authors reported $2.4 billion in trading activity across that sample. A later study of Polymarket's 2024 presidential market found that adding blockchain transactions without accounting for how shares are created and canceled can overstate turnover, so reported dollar totals depend on the counting method.
What caused past prediction market failures?
Earlier prediction markets closed or failed to launch because of regulatory decisions, weak participation, and problems handling contracts or customer money. The practical effects ranged from forced early settlement to the loss of an interface through which people traded.
| Problem | Historical example | Effect on traders or prospective participants |
|---|---|---|
| Access was withdrawn or threatened. | Intrade's US withdrawal in 2012; PredictIt's 2022 wind-down order. | Intrade customers had to close positions early. PredictIt users faced a similar deadline before a court intervened. |
| A proposed market was stopped. | PAM in 2003 and box-office futures in 2010. | PAM never opened to public trading. The film-revenue contracts lost the legal basis for their approvals. |
| A service attracted too few users. | Veil's 2019 closure. | Users lost Veil's trading service and were told to withdraw positions, while the Augur protocol continued. |
| Financial problems interrupted access to money. | Intrade's worldwide suspension in 2013. | Trading stopped and payments were suspended during the investigation. |
| The contract could not settle as its buyer expected. | The 1876 election dispute and Augur's ambiguous contracts. | Customers could face cancellation or an alternative settlement value despite correctly anticipating the event. |
In a May 2024 essay, Nick Whitaker and J. Zachary Mazlish argued that limited demand mattered more than regulation in explaining why the field had remained small. Writing before that year's election surge, they asked who would take the other side of an informed trader's position. Savers had little reason to use contracts whose trading gains came from other participants' losses. People seeking entertainment might find distant deadlines and specialized subjects unappealing, and demand for hedging was limited. Without enough of those participants, an informed trader had fewer opportunities to earn money from research.
Event Contracts
How event contracts turn a question about the future into a tradeable price.
Prediction Market Platforms
How exchanges, academic markets, and decentralized protocols differ.
Markets vs. Polls
What markets and polls forecast, and how to compare their records.
Market Resolution
How contract rules and outcome sources determine settlement.