Data analyst G. Elliot Morris has warned that the Republican Party may be heading into the midterm elections with an inflated sense of confidence, arguing that current polling methodology is quietly tilting survey results toward the GOP.

"The polls are [possibly] understating support For Democrats, not Republicans, as you might have heard," Morris said in a discussion of the data generating process with David Nir, host of The Downballot Podcast.

Morris acknowledged why many political observers would be skeptical of such a claim. "If you pay any attention to politics and the polls over the past couple of cycles, then you will probably be triggered by large apparent misfires in public opinion surveys in the direction of the Democrats over the last four election cycles, basically every cycle since 2016, except for the midterms," he said, referring to President Donald Trump's unexpected victories in 2016 and 2024.

But "once you look under the hood," Morris said, pollsters appear to be inadvertently introducing bias in the opposite direction, setting Republicans up for greater disappointment than they already anticipate. "Pollsters now are increasingly weighting or adjusting their samples by political variables, in particular, the results of the last election." Weighting on past vote is a longstanding survey adjustment in which pollsters rebalance their samples so respondents' recalled choices match the outcome of the most recent election, a technique meant to correct for the possibility that one party's voters answer surveys at higher rates than the other's.

The complication, according to Morris, is that survey respondents do not always admit they voted for Trump when they did. "Maybe you don't want to say you voted for that guy. Maybe you were like a soft Trump voter in 2024 and a pollster is going to ask you, 'hey, who did you vote for last time? And you were under no obligation to the pollster's statistical methodology to tell them the truth. Maybe you say, 'oh, I didn't vote.'"

Follow-up surveys, Morris noted indicate that respondents have been revising their recollections of past support, declining for whatever reason to acknowledge that they backed Trump. "Those sort of soft members of the Trump coalition now won't admit that they voted for him. So that's not necessarily a problem unless you are equalizing your sample by past vote," he said. "... [T]his level of recalled vote error could be increasing support for Republicans in the generic ballot by a little less than one percentage point in terms of vote margin. So, in other words, if there wasn't this level of hypothetical bias we're exploring, maybe Democrats are winning the generic ballot by like eight points instead of seven, which is what our average shows today." The generic ballot, which asks voters which party's congressional candidates they plan to support, is a standard barometer of the national political environment heading into a midterm.

That margin, Morris argued, could prove decisive for Republicans in narrowly contested state and district races — particularly if they are counting on that single point to carry them across the finish line. For poll-watchers, the thread to follow from here is whether survey firms revisit how they weight by past vote — and whether generic ballot averages move as more fieldwork is completed before Election Day.