Seasoned political operatives are dismayed at Donald Trump’s staying power in the Republican presidential primary.  

But for those Republicans waiting for Trump’s surge to crest and more establishment candidates to rise to challenge him, there’s reason not to dump their Jeb Bush stock.  

At least, that’s what political prediction markets suggest.  

Going into the second GOP debate, PredictIt , an online political prediction site, shows the former Florida governor leading the GOP pack, followed by Florida Sen. Marco Rubio.  

Contrast that with the double-digit lead Trump holds over Bush in Real Clear Politics’ polling average,  and it looks like prediction markets know something the rest of America doesn’t.  

Michael Burleson is a young software engineer in the D.C. metro area who’s put $750 into bets on PredictIt.  

Traders such as him, self-described “political nerds,” bet on candidates they think will win the nomination and against those who they think won’t.  

There’s evidence to suggest that kind of betting has more predictive value than polling likely voters.  

“Markets are a place to look when trying to see what’s going to happen. Polls tell you what’s happening now,” said David Rothschild, an economist at Microsoft Research, a division of the software company. 
PredictWise , a research site Rothschild helps run, aggregates PredictIt data with that from another betting site, Betfair. As of Wednesday morning, Bush led Rubio by 22 points and Trump by 20.  

In the 2012 election, there were at least 11 lead changes in GOP primary polls, Rothschild said Tuesday. But there was no lead changes in prediction markets. “They had Romney shooting straight up,” he said.  

When former Texas Gov. Rick Perry briefly took the lead in the 2012 primary, “the markets knew something,” Rothschild added. “The markets were anticipating that ‘Oops’ moment.” Rothschild is in touch with major polling companies trying to adopt from prediction markets the kind of expectation questioning that Rothschild says has been more accurate. Traditional polls ask voters who they plan to vote for; expectation polling asks who voters think will win.  

With economist Justin Wolfers, Rothschild published a 2012 paper  which showed that expectation polling has more accurately predicted the outcome of presidential elections over the previous 60 years than traditional polls.  

In last year’s midterm elections, expectation polling better predicted the outcome of every Senate contest except North Carolina.  

Why would that be?  

Rothschild and Wolfers argue that asking voters about their expectations effectively widens the sample size. In order to answer the question, voters have to take into account what their social networks think and how they will vote.  

That’s especially useful in an era of ubiquitous cell phone use when reaching a random sample of voters is becoming increasingly difficult, Rothschild said.  

But prediction markets are hardly a random sample, either.  

“Prediction markets are getting a select group of people who know a lot about what is going on,” Rothschild said.  

Many PredictIt players are concentrated in the Washington, D.C., area. The top five states by concentration of players are California, New York, Texas, Florida and Illinois. Notice that Iowa and New Hampshire are not on that list.  

Prediction market enthusiasts acknowledge there may be some biases in the market data from the players Rothschild called “data geeks,” but they maintain that markets are much more consistent.  

Yes, markets can change in seconds, and yes, bettors do bet some candidates short even if they expect them to win long-term. But PredictIt player Burleson has noticed the markets are relatively sticky.  

That’s because there’s money on the line.  

If you’re placing a bet, Rothschild said, you are incentivized to do your research and focus on the “net effects” of a campaign, rather than the daily ups-and-down.  

“The polls I don’t pay much attention to,” Burleson said. Instead,  he tends to look at political endorsements when he’s betting on who’s going to do well and who isn’t.  

“It’s not that the market is always right,” said John Aristotle Phillips, CEO of PredictIt. “It’s that the profit motive clears your mind… It sobers you up.” In other words, there’s reason to suggest that today’s polls still reflect an ephemeral Trump boom, whereas prediction markets know more about what’s going to happen in November 2016.  

That’s not to say Trump isn’t doing well in the prediction markets. He’s in the top three, below Rubio and above Ohio Gov. John R. Kasich as of Wednesday morning.  

But his stock, said Phillips, has been more volatile than any other Republican’s.

Get breaking news alerts and more from Roll Call in your inbox or on your iPhone.