Donald Trump’s chances of winning the Presidential election have been rising on prediction markets. That’s not surprising. Although no single event stands out, most commentators see October’s news flow as being mildly net positive for the former president.
Here's where we stand currently: Expert forecaster Nate Silver has increased his estimated probability of a Trump victory by 10 percentage points, from 43% to 53%. The PredictIt prediction market has made a slightly larger adjustment, from $0.48 to $0.59 (or a 59% chance of a Trump victory), while Polymarket is up a bit more, from $0.50 to $0.64.
There would be nothing to see here if we were talking about financial prices or sports betting spreads. This election is very close, and even small events or bits of information can move probabilities a lot; the swings we’ve seen in the past few weeks don’t require big explanations.
Also, in a near-even contest, it is not surprising to see differences of opinion of this size. For example, in the upcoming World Series, major sports books are quoting odds on a Los Angeles Dodger win from 52% to 58%, similar to prediction market prices for Trump to win the presidency. Meanwhile, experts who favor the Yankees imply less than a 50% chance for the Dodgers. Others think the Dodgers are more than a 71% favorite based on predicting they win in four or five games. And there’s far more useful data and analytics for predicting baseball than Presidential elections — especially one as unusual as this year’s — and far more liquidity in sports betting than election markets for arbitragers to take advantage of odds differences.
But this is a hyper-partisan election with swirling paranoia feeding conspiracy theories. Some Democrats are suggesting that the prediction market prices, especially Polymarket, are being manipulated by people hoping to encourage Trump voters and dishearten Democrats. On the other side, there are Republicans extrapolating from the gains, figuring that if Trump’s chances increase 10 percentage points every two weeks, by Election Day they’ll be around 70%.
Is there support for either of these views?
Going back to July 22, the day after President Joe Biden withdrew from the race, PredictIt has given Trump an average price of $0.49, compared to 52% for Silver and $0.52 from Polymarket. So there does not seem to be a long-term prediction market bias for Trump. If the charge is about a short-term attempt to boost Trump in recent weeks, we must again exonerate PredictIt traders, since its price for Trump has increased a similar amount to Silver’s prediction. But Polymarket started the period with the highest chance for Trump and has also increased the most. Four large bettors — who could conceivably be one bettor using four accounts — are responsible for much of the Trump-buying volume.
One obvious question is why people don’t buy Trump on PredictIt and sell him on Polymarket, locking in a small profit, in the type of arbitrage trade that is common for financial markets. Polymarket consistently prices Trump bets an average of $0.04 higher than PredictIt.
One reason is the contracts are not identical. PredictIt pays off if Trump wins the majority of Electoral College votes, in the sole opinion of PredictIt, and pays nothing if that doesn’t happen, for example if there is an Electoral College tie or Trump dies or withdraws. Polymarket pays off if a Republican is elected President, and pays $0.50 if neither a Republican nor a Democrat is selected. If something unexpected occurs, at worst you get $0.50 from Polymarket but you can get zero from PredictIt.
Another consideration is credit risk. PredictIt offers the conventional security of traditional institutions, Polychain offers crypto security. Some people have a vast preference for the former, and my guess is they tend to be pro-establishment supporters of Vice President Kamala Harris. Others vastly prefer the latter and are probably more likely to support anti-establishment Trump.
What needs explaining?
If we take the $0.04 as the equilibrium fundamental spread between Polymarket and PredictIt, then the event to explain is why the spread went from $0.02 — below fundamental value — in mid-September to $0.05 — above fundamental value — today. It could be noise, of course, unrelated to bettors’ intentions or electoral prospects, but it could be market manipulators trading to influence posted prices rather than to maximize profits. Logically, it could be Harris supporters manipulating PredictIt or Trump supporters manipulating Polymarket.
To be robust against manipulation markets need large open interest and trading volume, low transaction costs, transparent reporting, liquidity, deep-pocketed market makers with views of fundamental value and rigorous monitoring. No existing prediction market is as secure as major financial markets—and manipulation is not unknown in those markets.
Manipulator wannabes face obstacles on both markets. PredictIt limits exposure to $850 per account, so multiple accounts (and participants) are required to move prices significantly. And they’ll need to be U.S. adults to boot. On the other hand, relatively low liquidity means that prices can be moved without significant cost. There appear to be few opportunistic market makers who would offset any attempt to manipulate prices. The other public election betting sites, other than Polymarket, are similar but smaller.
On Polymarket, one individual can essentially bet unlimited amounts without disclosing the sort of information (name, address, financial institution) required by PredictIt. But the site has many deep-pocketed sophisticated market makers, so it takes far more money to move prices. U.S. citizens are forbidden to bet on Polymarket, but there’s little means to enforce that ban.
Tale of the tape
An important clue is trading patterns. Manipulators trade in ways that maximize price impact. Large orders are placed at times of low liquidity, and bets may be quietly reduced in small, random-looking trades at times of high liquidity. Large bettors seeking profit trade in opposite ways. The available data doesn’t appear to show signs of manipulative trading on either market. That doesn’t disprove the manipulation charge, but it makes it less plausible.
How about the momentum idea? I have previously pointed out (read this) that both Silver’s probabilities and PredictIt prices have daily momentum; if the probability estimate or price went up yesterday, it’s more likely to go up than down today as well. But what about two-week momentum? Do changes over the last two weeks predict prices over the next two weeks?
They do, but perhaps not enough to matter much. My calculations suggest that, based on recent moves, Silver’s estimate will be 56% for Trump by Election Day, PredictIt might be up to $0.61, and Polymarket could be at $0.66. Of course, these are simply extrapolations of historical momentum, not forecasts, and they ignore all the events and news of the next two weeks.
I have made no serious attempt to forecast the election, but I have been skeptical of the wide swings of the summer and early fall in prediction markets. The average of the three estimates today is almost exactly what it was on June 1, before the Trump-Biden debate. Back then, the estimates were 15 points apart, today they have converged somewhat to a 11-point difference. All have shown big ups and downs in between — sometimes in different directions from the others — but I suspect the same party will win on Nov. 5 as would have won if the election had been between Biden and Trump on June 1. And you could have saved yourself the trouble and anxiety of following the political news, without a net loss.