I’m often asked whether there is enough volume and liquidity on PredictIt to use it as a meaningful source of analysis. 

Volume on PredictIt is determined by four main factors:

• The overall popularity of the site;
• Regulatory limits on the size of bets
• Regulatory limits on the number of traders in each market 
• Trader interest in particular markets

   Popularity of PredictIt

PredictIt has seen rapid growth over the last three years. From 22,000 in 2016, the number of traders on PredictIt has jumped to 80,000 as of March 2018. The increase is due in large part to a surge of interest during the 2016 election cycle, when 17,000 traders were joining per month. In early 2015, PredictIt saw about 2 million total shares traded. By 2016, popular markets such as the Michigan primaries alone generated more than 2 million trades, in part because prominent pundits like James Carville announced on TV that they were betting on PredictIt. In 2017, more than 300 million shares were traded on PredictIt.

As part of its agreement with the CFTC, PredictIt is not permitted to disclose the amount of money traded on the site. Brookings Institution fellow Philip Wallach, however, estimated in The Los Angeles Times that PredictIt had $1 million in bets in its 2016 presidential election market. Public information about the volume in political prediction markets is generally sparse, but the volume on PredictIt can be placed in context through comparisons with comparable sites:

• Iowa Electronic Markets: The IEM, as a matter of policy, does not release information on volume. An IEM spokesperson declined even to comment on how volume on the site compared between the 2012 and 2016 presidential elections. Older studies, however, provide some data. A 2008 paper revealed that the IEM markets had about 1,000 active investors and $167,000 in trading volume during the 2000 presidential election. Trading volume fell to about $92,000 during the 2016 presidential elections. The IEM impose a $500 limit on the amount traders can put in their accounts.

• Intrade: The Ireland-based prediction market was active from 1999 until 2012.  The site included contracts not only on politics, but also on subjects such as commodities, business, and entertainment. Intrade generated press attention comparable to PredictIt, and, in its heydays, “mentioning Intrade was becoming a mark of sophistication.” Intrade saw $230 million wagered on the 2012 presidential campaign, a jump from $50 million in the summer of 2008. In terms of volume, there were more than 100,000 registered users on the site in 2011, but that number fell to 7,000 account holders in 2012. Seventy-five percent of users were U.S. citizens.

• Betfair: About $277 million was traded on the 2016 U.S. presidential elections. This was more than any other market on the site in history despite the fact that Betfair excludes Americans.

Regulatory Limits on the Size of Bets

As part of the deal PredictIt struck with the CFTC, PredictIt must place an $850 limit on each investment by an individual participant in any contract. This requirement impacts the markets in three main ways:

• First, it limits the ability of a small number of investors to dictate market prices with large bets. Whether or not this results in more or less accurate market prices depends on how the“wisdom of the crowds” in that particular market compares to the insights of the most active investors. During the 2016 elections, for example, Paddy Power,which does not place bet limits, reported that there were 97 four-figure plus bets on Trump as opposed to 84 four-figure plus bets on Clinton. An open question is whether small investors­—some of whom have as littleas $10 on PredictIt—have enough at stake to make prudent predictions.

• Second, it provides a check againstmarket manipulation, which PredictIt otherwise doesnot seek to control.

• Third,it may exacerbate favorite-longshot bias by encouraging traders to “underprice in-the-money options, and overprice out-of-the money options.” In a market where one is only permitted to bet $850,those taking the longshot can exert a much greater influence on the price than someone who takes No. For instance, if a longshot is priced at 10 cents then you can buy 8500 Yes, but only 944 Nos.

Regulatory Limits on the Number of Investors

CFTC regulations force PredictIt to limit each contract to 5,000 total traders. Traders, to my knowledge, do not have a way of knowing when a contract will reach the 5,000 cap.

In practice, the 5,000 cap does not significantly impact the market price on PredictIt. Only the most popular markets reach the 5,000 total trader limit, and only near the time when the markets are about to close. For markets popular enough to reach the 5,000 limit, PredictIt typically creates multiple markets that effectively test the same question. For example, during the 2016 presidential election, even though the market on which candidate would win reached its cap, PredictIt created several other markets on the presidential election that allowed latecomers to enter. These included markets asking which party would win the election, and whether or not a woman would win the election.

Trader Interest in Markets

The volume in each market on PredictIt varies considerably. Factors that influence the number of traders includes:

• The familiarity traders have (or think they have) with the question at hand
• The amount of polling and/or media coverage on the markets
• The degree to which the market is ideologically polarizing
• The extent to which the markets provide an opportunity for quick profits

The most predicted markets on PredictIt tend to be those involving presidential elections and high-publicity events such as Supreme Court picks. 

Here is some data to give some perspective on PredictIt’s absolute numbers:

• The Alabama Special Election race between Roy More and Doug Jones, one of the more popular markets of 2017, saw more than 10 million shares traded.

• The Supreme Court market saw trade volumes of about 28k on active days.

How does volume impact the insights available from PredictIt?

Conventional wisdom on prediction markets suggests that higher volume markets are more likely to produce accurate forecasts. Consistent with this expectation,PredictIt CEO John Aristotle assessed in 2015 that PredictIt had seen a correlation between “greater liquidity across the marketplace and increasing forecasting accuracy.”

As an active trader, however, I have found that theory does not necessarily conform to reality. I am reluctant to draw any hard conclusions on the relation of volume and accuracy for several reasons:

• Big Misses in Popular Markets: Political prediction markets generally and PredictIt specifically have produced some of their biggest misses in the most popular markets such as Brexit and the election of Donald Trump. In the 2016 elections, the Iowa Electronic Markets came closest to predicting Trump’s victory even though they had significantly less volume in terms of the number of participants, the number of shares traded, and the amount of money on the site. Whereas Betfair and PredictIt put the odds of a Clinton victory in the low 80s, the IEM markets leveled off at 71 percent.

• Typical Trading Patterns:Market movements are driven by a relatively small number of traders even in high volume markets.

• Sophisticated Traders:Low volume markets may draw more sophisticated investors. This is particularly true in markets where pertinent information may only be available to insiders and/or skilled researchers. PredictIt spokesman Will Jennings predicted in a recent podcast interview, for example, that the midterm markets are likely to draw more “casual” traders with particular knowledge of local elections. Although I generally advise beginners to start in tight, high volume markets, as a more advanced trader with an intuitive feel for the markets, I actively look for deals in low-volume markets where I can generate high returns with frequent trading.

• Irrational Influxes from “Noise Traders”: Increases in volume can come from demographics that irrationally distort markets.