ABC News’ Reanalysis of Its Own Poll Begs You to Choose Your Own Reality

Readers, please allow me another brief Corner note to follow up yesterday’s New York Times/Siena poll post. ABC News is out today with an exceedingly strange piece: not a new poll — though they attempt to disguise it as such — but a reanalysis of an old poll. Yes, ABC/Ipsos last released a national poll back on September 1, one in fact taken between August 23, the night Kamala Harris spoke at the Democratic National Convention, and — through the four days of post-convention media afterglow — August 27. (I mention these dates because the calendar matters in this case.) That poll showed Harris up over Trump by an unusually wide amount relative to most other major polls, 50–46. The mere fact that it was an outlier is no reason to dismiss it outright; good and well-conducted polls sometimes return anomalous results through no fault of the pollster.

What in fact raised eyebrows among intelligent observers was that this poll, taken during what one might have expected to be the “bounce” phase of Harris’s candidacy, actually showed no change at all relative to ABC/Ipsos’s previous test of the national race, which had also been at an unusually Harris-favoring 50–46 number. The delta, in other words, from pre-convention to post-convention was exactly zero, which is yet another reason why, last week, serious observers immediately began to signal that the winds had shifted against her. (An unstated fear about Ipsos is that there is something artificially tight about their “Likely Voter” screen — remember the low-propensity nature of many Trump voters — and that this means they are envisioning a different electorate from the one that will actually materialize on November 5.)

So now, in the wake of NYT/Siena’s downbeat news for the campaign, we have ABC’s “reanalysis” of the old data rushed out this morning, and guess what? According to the authors, the news is even better for Kamala if we merely “adjust” these numbers a bit. (In this imagining, Harris is now theoretically leading by a whopping 52–46 margin among likely voters.) Nobody is buying it except Matthew Dowd, who is ideologically required to. This sort of ex post facto jiggery-pokery is suspicious enough on its own terms, but when done with old and completely dead data (especially given the post–Labor Day factor), it is an exercise that reveals more about the anxieties of its practitioners than the state of the race.

It’s going to be a wild debate night tomorrow, folks.