Scholiaster
No bio...
User ID: 840
Scholiaster
7mo ago
Not really, because Biden’s disfluencies left the door open for another interpretation (that doesn’t have your problems) and thereby created a scissor statement that won’t get traction. At most, it will extend Puerto-Rico-gate for another news cycle, which probably marginally benefits the Democrats in PA and other swing states.
Scholiaster
8mo ago
Pew research shows consistent voting is a U-shaped curve.
https://www.pewresearch.org/politics/2014/06/12/political-polarization-in-the-american-public/
Scholiaster
8mo ago
Usually it’s with policies with concentrated benefits for a few and diffuse costs for the rest, like rent control, occupational licensing, NIMBY, the list goes on.
Scholiaster
8mo ago
I’m not asserting a link between turnout and quality, though there may be one. Indeed, Public Choice Theory suggests that for certain policies democracy and majority voting is no guarantee of quality.
Scholiaster
8mo ago
The middle drops when the parties rely on negative campaigning to energize their (more extreme) base at the expense of the middle, who are usually already annoyed by politics. Negative campaigning affects partisans differently than non-partisans. When the middle drops out, the party that is marginally more effective at turning out their base win.
Scholiaster
8mo ago
The Median Voter Theorem states that a majority rule electoral system will elect the candidate preferred by the median voter. However, if the middle drops out and doesn’t vote, the candidates can easily be more extreme than preferred by the broader populace.
Scholiaster
8mo ago
Voting is a Prisoner’s Dilemma, where if everyone cooperates (by voting) the politics are sane and geared after the medal voter. But one’s vote is so insignificant, it is more beneficial to not waste one’s time and defect by not voting. However, if too many people defect, the parties polarize and scare the median voters away from the pools, and polity goes bouncing from extreme to the other. Compulsory voting, as in Australia, solves the PD.
Awfully hard to say, because the best evidence is the polls and the polls are maximally uncertain. It’s like guessing the polling error, which is normally a fool’s errand. Still, one can reason through various sources of error:
a) Sample error won’t be much of a factor. We have lots of polls and we can average out their sample errors. We can’t entirely get rid of it, of course, but the real problem is that the polls we’re seeing don’t have enough variance. This suggests …
b) Herding. The polls are not really independent of each other. This means that averaging them won’t diversity away the errors they have in common, and a big one is …
c) Model error. The fundamental problem is that it is believed that the traditional sampling is just missing voters. It has certain missed Trump voters the past two cycles. So a lot of pollsters try to estimate what they can’t sample. They can be sophisticated about it but it’s ultimately a guess, and it seems that a lot of pollsters are reluctant to guess too far away from the NYT/Siena and other leading pollsters. However, this guess can be wrong about the electorate that will really show up and this will produce a bias against the true result. Currently, pollsters seem to be herding around and 50-50 outcome, which minimizes their chances of being seriously wrong unless there’s a blow out.
d) Latency. It is often stated that polls are not a prediction of the future but a snapshot of the present electorate. Not quite. It’s a snapshot of a recently past electorate. Polls take a number of days to complete. Sometimes they are held. For most polls we’re looking a delay of a couple of days. Poll aggregators, like RCP, reaching back weeks have big latency problems, given in composite picture of the race one or two weeks ago. Latency error means that polls may be too slow to respond to late breaking changes or late deciders.
How does this affect the current election? The biggest sources of polling error seem to be model error and latency error, and we’re not completely in the dark about them. We have some data points showing the effects of different models. In particular, the IA Selzer poll arand the IA Emerson poll released on the same day has D+3 and R+10, where the former does not even try to find missing Trump voters and other latter does. That’s a 13-point spread. Some of it could be sampling error, but these polls shows that model error could be a significant chunk of the polling error. There’s also a leaked internal IA poll from the Trump campaign which is R+5, suggesting that Trump may be underperforming the modeling by 3-5 point. That’s huge. If you give 3 points to Harris, she sweeps the swing states and cleans up with about 319 to 219 EVs. There are of course lots of complications with the IA Selzer polls, but it opens the possibility for significant model error in most posts (and they’re herdering around this, so not completely independent) and a major polling miss.
As for latency, the last week of the Trump campaign has been disastrous for non-online Latinos. The “floating island of garbage comment,” which Trump refused to condemn personally, appears to have liquidated the undecided Latino vote in Pennsylvania. Given how close in the polls PA is, it probably hands the commonwealth to Harris and her blue wall holds. If there’s a 3-5 point model error, as there are some signs for, she could end up sweeping the swing states or even start winning Red states.
Of course, model error could favor Trump. Polls are trying to account for low-propensity Trump voters (generally young and non-college educated men) but the problem is with low-propensity voters is that they don’t turn out. Maybe they did more when Trump was fresh and cool. The last week of campaigning with emptier and emptier rallies suggest he’s past his expiration date.
Final call: lean Harris with upside.
More options
Context Copy link