Republicans now lead Democrats 51 to 49 among likely voters, according to the latest Harvard-Harris poll.
That two-point edge may sound modest, but the trajectory behind it tells a bigger story. In March, Democrats held a four-point lead in the same poll’s likely-voter measure. By July, that margin had shrunk to two. In August, it flipped. Republicans went from trailing by four to leading by two, a six-point swing in five months using consistent methodology from a single pollster.
According to the original report, this marks the first time any pollster has shown Republicans leading among likely voters since October of last year.
That distinction matters. Likely-voter samples are the number serious analysts and campaigns track most closely as an election approaches. Registered-voter samples tend to overstate the side less likely to actually show up on election day. The same Harvard-Harris poll still shows Democrats ahead by two among registered voters, and partisan spinners will point to that figure. But it’s the likely-voter number that historically predicts outcomes.
The enthusiasm gap inside the poll cuts in the same direction. Roughly seven in ten respondents said they plan to vote, with Republicans reporting slightly higher motivation than Democrats. That is precisely the turnout profile a party wants heading into a competitive fall election.
Context on polling history adds another layer. An analysis of more than 3,000 surveys from the past four election cycles found that polls have overestimated Democratic performance by an average of 3.7 points. The error during summer polling ran closer to six points. If that pattern repeats, the Harvard-Harris result may not be an outlier at all. It may be the first visible sign of the annual autumn correction, as summer samples that skew Democratic drift back toward what the electorate actually looks like.
There is a conventional wisdom in American politics that functions almost like a law of nature. The party out of the White House gains seats in the midterms. Political observers have leaned on it heavily throughout this cycle. A single poll does not overturn that assumption, and two points among likely voters is not a mandate. But a six-point swing over five months, measured consistently by the same pollster, is not noise. It is a direction. And right now, that direction is moving toward Republicans at a moment when the political class expected the opposite.
No single survey should be treated as a final verdict. But if the trendline continues, the comfortable predictions about midterm gravity may need some revision before November arrives.