Even a dramatic improvement in the economy and the Middle East may come too late to produce a major midterm swing, Verasight CEO Ben Leff told MarketDash.

Asked whether President Donald Trump ending tariffs, reaching an Iran deal and helping send fuel prices sharply lower could turn things around, Leff said voters would need to actually experience the improvement.

"I don't think people necessarily put much weight on what the president says, but if they truly feel gas prices drop, that would drive them more than just being told they would," Leff said.

Voters May Already Be Locked In

Leff said economic dissatisfaction takes time to change, while voters' views of Trump are already "pretty solidified."

"We're unlikely to see a shift where suddenly Republicans are likely to have a great November," he said, adding that any late movement would likely come "at the margins."

Trump helped bring GOP voters to the polls in presidential elections, but his absence this November "might depress turnout" among Republicans, Leff said.

A Reuters/Ipsos poll published Thursday found 30% of people who voted for Trump in 2024 were highly enthusiastic about voting in the midterms, compared with 48% of Kamala Harris voters. Trump's approval rating stood at 32%.

Eurasia Group President Ian Bremmer offered an even starker assessment this week, telling Bloomberg that Trump has become "an anchor at this point. He's dragging them down."

Prediction Markets Point the Same Way

Polymarket traders currently give Democrats a 91% chance of controlling the House, on about $15.1 million in volume, and a 64% chance of taking the Senate, on roughly $6 million.

Leff said Verasight's polling shows Democrats "very likely" to take the House and "likely" to take the Senate.

But he remains cautious about treating prediction market prices as an independent forecasting signal. Verasight does not incorporate them into its election models, in part because, Leff said, it can be unclear whether the market signal comes from ordinary traders or large "whales."

Leff also agreed that combining polls and prediction markets can create a double-counting problem. If a poll moves a market and a forecasting model then uses both, the same information can effectively influence the forecast twice.

Verasight can also exclude candidate-sponsored or poorly vetted polls from its models, Leff said, while prediction-market traders may still react to them.

If polling and markets sharply diverged, Leff said he would trust a high-quality poll because its methodology can be examined.

Asked whether he might then wager against the market, Leff said Verasight has a policy against wagering on prediction markets.

"There needs to be a real wall between pollsters and prediction markets," he said, citing the risk of pollsters developing financial incentives tied to information they produce.