Has anyone informed Politico or Morning Consult that voters are already beginning to cast votes in the midterms? In some places in Pennsylvania, voting started on September 19. Four states will start tomorrow: Minnesota, South Dakota, Virginia, and Wyoming. Voters in Illinois and Michigan can begin filing their general-election ballots starting tomorrow.
So why do pollsters like Morning Consult and sponsors like Politico still report registered-voter outcomes? Perhaps it’s because the news is bad enough for Democrats on that basis:
The POLITICO/Morning Consult poll was conducted Sept. 23-25, surveying 2,005 registered voters. The margin of error is plus-or-minus 2 percentage points.
In other findings, the poll also found Biden’s approval rating dropping to just 41 percent, with 56 percent disapproving. That’s down from 46 percent in last week’s poll — which had represented Biden’s 2022 high-water mark. …
Forty-five percent of voters said they would vote for a Democrat for Congress if the election was today compared to 43 percent who preferred a Republican candidate, the poll found. In last week’s poll, Democrats had a 5-point lead, 47 percent to 42 percent.
Even before we get to those specific results, we have to wonder when Morning Consult plans to apply likely-voter models to its results. November 15th? This criticism doesn’t just apply here but also to a number of pollsters using registered-voter results rather than the far more predictive likely-voter modeling this close to an election.
The apotheosis of this trend belongs to Reuters, which scored Joe Biden’s approval rating this week on results for “adults,” not even registered voters. Even then, Biden only got a 41/53, about the same as Morning Consult found with registered voters.
The biggest problem with the decision to avoid likely-voter modeling comes in the generic-ballot question. A look at RCP’s aggregation shows clearly what happens when using RV results rather than LVs — Democrats look a lot more competitive. In every poll using LV results since Labor Day (except for an outlier from a poll from Selzer, which is unfamiliar to me), Republicans either tie or lead. In every poll using RV results, Democrats lead.
Which leads back to the first question: why use RVs at this stage at all? The point of these polls, and especially the generic ballot surveys, is supposed to be predictive. Without any effort to position these polls to measure the outcomes among the people likeliest to vote, the results cannot be predictive — even though the media consistently treats them as such.
The omission is even more curious in Morning Consult’s poll considering that it includes questions about voting enthusiasm — and their breakdown suggests that Democrats may have an edge on enthusiasm, once you read through their crosstabs.
When you get to the issue responses, on the other hand …
A plurality of voters also did not approve of Biden’s handling of the economy, jobs, healthcare, immigration, climate change and a host of other policy areas surveyed in the poll.
The economy was the top issue among 44 percent of voters (the highest of any issue in the poll), and 61 percent disapproved of Biden’s handling of the economy.
It’s not just the highest issue of any for respondents; it laps anything else, and nearly eclipses anything else. “Women’s issues” gets less than a third of that support as a midterm priority while still coming in second, but a close third are “security issues” — the border crisis and presumably crime being among them:
Under those circumstances, are we really to believe that Democrats have the edge on enthusiasm and that an RV model will be predictive for an election that takes place in less than seven weeks? Not even Morning Consult apparently believes that. According to change in the topline results for Biden and generic Democrats, neither do their respondents.
For now, then, consider any RV poll result for entertainment value alone. Can’t wait to see their likely-voter model roll out in time for Thanksgiving dinner.
Discussion about this post