Midterm elections are traditionally brutal for the party in power. What if this November, however, is as
atypical as November 2016?
Trump is a singularly energetic campaigner. His efforts this year will likely move more Republicans and Republican-leaning independents to the polls. Seeing large arenas overflowing with animated Trump-supporters is probably a drag on Democrats. His simple message at these rallies is perfect: things are going well, but the Democrats will take it all away and stop any further progress, so we need more Republicans in Congress.
Re widely publicized negative job-approval ratings for President Trump:
Factors such as the Bradley/Wilder Effect and the "job performance" vs. "handling" question might each count for 5% or more by themselves. If the above factors average just 1% each, the approval picture for Trump changes dramatically, and the GOP's election prospects change dramatically.
[Real Clear Politics's] polling average for the direction of the country is at a five-year high and has been steadily rising for the last year. The peak under Obama was in June of 2009, with 45.8% of those polled saying the country was moving in the right direction. Two years later, Obama bottomed at 17%. On election day 2016, it was 31%. Today it stands at 41%. Because of Trump's unprecedented commitment to keeping his campaign promises, for the first time in a generation, Republicans have a chance to vote with great enthusiasm. They should ignore the polls and do so.
It is my fervent hope that the American electorate is disgusted with the dumpster-fire known as The Resistance and will show up at the polls to do something about it. And that goes double for the next primaries affecting any Senators who wobble on this nomination.
I think another factor impacting poll accuracy is that a fraction of pollees just troll the pollsters. How big a fraction, I don't know, but the fact that I'm seeing increasing numbers of anecdotes of such trolling is illustrative if not indicative. Of course, the anecdotes began after the polls were wrong in 2016, and that may be a fatal contaminant.
ReplyDeleteEric Hines
The conservative movement will be helped by two new voters in our family, both very excited to vote for the first time this November. So there's a local net of +2....
ReplyDeletePolls always depend on turnout assumptions.
ReplyDeleteThe problem with using averages of all midterms is that it includes Democrat debacles like 1994 (54 House, 8 Senate seats) and 2010 (63 House, 6 Senate). You have to go back to Nixon's losses in 1974 to get similar numbers for a Republican President. Even with that loss included Republican Presidents lose few seats on average during their mid-term elections than Democrats.
ReplyDelete"Polls always depend on turnout assumptions."
ReplyDeleteThe revelations that Google deliberately attempted to swing the 2016 election to Democrats by turning out Hispanics, (and, in so far as increasing turnout, succeeded!) is evidence that the assumption is in wide use. But as the actual voting results showed, a larger-than-usual (or expected) fraction of the newly turned out Hispanic voters supported the Republican. Assuming that groups targeted in "get out your base" efforts will behave as desired SHOULD be a dangerous assumption. Neither faction should neglect attempts to persuade the undecided, or uncommitted, or those not bound to the most-obvious-identity-group.
(What does demographic polling data reveal about, say, Hispanic Home-schoolers, or Black Veterans, or Retired White Male Intellectuals? )
That's true too. We may well see that black turnout does not correlate, this time, with as-high Democratic Party numbers; Trump is unusually well regarded for a Republican, apparently, among black men in particular.
ReplyDeleteAll the same, I stand by the point. Your poll may not depend much on an assumption about how a demographic will vote, because you collect demographic data in the same motion as intent-to-vote data. (I.e., you ask them both "How would you describe your race or ethnicity?" and also "How do you intend to vote?" at the same time.) All the same, when you've taken your 200 or 500 or 1000 readings, you still have to make a projection about how to project the whole electorate from that sample. That's a turnout assumption that may simply be flat wrong.
If that assumption is wrong, then the projection is wrong.