There are polls, and there are polls. And then there’s the Ipsos “KnowledgePanel” model, which isn’t quite a poll as commonly understood.

Let’s start with the results from today’s ABC/Ipsos survey, which created a stir on social media. A majority of the KnowledgePanel think Donald Trump deliberately broke the law in the Stormy Daniels/Karen McDougal payoffs that resulted in his indictment. Trump’s approval level has also dropped to its lowest level in the series history:

With former President Donald Trump now formally charged on criminal charges, a majority of Americans (53%) believe he intentionally did something illegal, according to a new ABC News/Ipsos poll.

An additional 11% say he acted wrongly but not intentionally. Only 20% believe Trump did not do anything wrong, and 16% say they don’t know, per the ABC News/Ipsos poll conducted using Ipsos’ KnowledgePanel. …

Of note, the former president is enjoying weaker-than-usual support from his own partisans with only a plurality of Republicans (45%) thinking Trump did nothing wrong, and the rest of the party split among the belief he intentionally did something illegal (19%), he was wrong but it was unintentional (18%) or they simply do not know (17%).

Trump’s current approval rating has hit its nadir, 25/61, the worst in a series whose best rating came in October 2020: 35/57. (For what it’s worth, Joe Biden only scores marginally better at 34/48, which is only slightly better than his series-low last week, 32/49.) Among all respondents, a 48/35 plurality thinks Trump should suspend his presidential campaign until the indictment gets settled one way or the other. That’s not surprising, given that almost two-thirds of the electorate think Trump broke the law and that less than half of his own base “supports” him, to use Ipsos’ language.

We’ll get back to the internals, which Ipsos does not supply, in a moment. There’s another topline result that doesn’t figure into the reporting quite as prominently, however:

Clearly, this assessment doesn’t mitigate the respondents’ dislike of Trump. But it’s still quite notable that a majority believes Alvin Bragg’s indictment to be a political act rather than a dispassionate application of the law, and it’s not even that close. ABC mentions it in the thirteenth paragraph, which seems a bit low considering how significant that question is regarding the credibility of the indictment.

But then again, just how predictive is this poll, really? In the first place, the sample is very small for a national poll — only 566 adults, not even registered voters. Both the sample size and the respondent type are not terribly reliable for national polling. But the methodology is also questionable:

Our recruitment process employs a scientifically developed addressed-based sampling methodology using the latest Delivery Sequence File of the USPS – a database with full coverage of all delivery points in the US. Households invited to join the panel are randomly selected from all available households in the U.S. Persons in the sampled households are invited to join and participate in the panel. Those selected who do not already have internet access are provided a tablet and internet connection at no cost to the panel member. Those who join the panel and who are selected to participate in a survey are sent a unique password-protected log-in used to complete surveys online. As a result of our recruitment and sampling methodologies, samples from Knowledge Panel cover all households regardless of their phone or internet status and findings can be reported with a margin of sampling error and projected to the general population. KnowledgePanel members receive a per survey incentive, usually the equivalent of $1 (though for some it is $2) in points, that can be redeemed for cash or prizes.

So … basically this is an Ipsos Grand Jury. Their selection may be relatively random, but they become a kind of professional poll sample for repeated uses. This is an attempt to get around real issues about survey responses in an electorate that refuses to answer the phone any longer. Essentially, Ipsos has incentivized these people to always answer the call, but that introduces risks of other effects. That brings into question just how representative they may now be of the electorate, as opposed to being motivated to impact polling on a regular basis. The “observer effect” seems like a big risk here, especially with Ipsos providing incentives to participate such as tablets, Internet connections, and “cash or prizes.”

The upshot of this: don’t bet the farm on results from this process. This may well reflect the mood of the electorate to a fine detail, but neither the sample nor the methodology provides any confidence on that point.

However, the numbers on this poll likely provide a broad, non-specific sense of how Trump and the indictment are being received, especially with gaps this wide. RCP no longer tracks Trump’s favorability numbers, but those were always negative and likely got worse in the immediate aftermath of the indictment. Nate Silver suggested last week that the polling response after the grand jury unveiled it was that Trump may have gotten what he deserved, not so much in the specifics but as a kind of “lifetime achievement award” for transgressions going back at least to the Access Hollywood tape, if not further back than that.

And that raises questions of its own for Republicans in the 2024 cycle. Assuming other polling substantiates these findings, or even approximates them, the risk of running Trump in the general election is much higher than it was in 2016 or 2020. Trump is too much a known quantity to get people to shift from negative to positive now, especially with one indictment on the table and two others that may drop along the way. How far will Republican primary voters go with a candidate with this much negative drag — especially with legitimate alternatives at the ready? Adam Baldwin and I discuss that in today’s Amiable Skeptics, so be sure to check that out.