I am going to take the opposite view from Lee and say this poll is actually pretty bad news, as it confirms stubbornly high levels of partisanship + leaning. After a decline from 2004 to 2014, the % of Americans choosing something other than true Independent has risen back to 91%
Partisans are more likely than ever to refuse to label themselves as Ds or Rs outright, but (even after the events of the last 2 months) are no more likely to actually disavow their party when pressed. GOP identity is as low as it was all the way back in [checks chart] Dec 2018.
Perhaps more people than ever want third parties, but the parties they want are more extreme than those on offer, so the leaned percentage increases & voting behavior get more predictable as frustration rises. (This is consistent with other Gallup data.)
(Just to be clear about the line connecting the dots: The reason the leaned % would increase as voters drift further from partisan medians is because of negative partisanship.)
The Gallup polling suggests that support for a third party among Republicans is driven by a desire for a more conservative version of the GOP. See: news.gallup.com/poll/329639/su…
That's consistent with GOP leavers being more likely to approve of Trump. See:
Weighting pollsters by their past accuracy is a really good idea that turns out not to improve forecasts that much, if at all. Some suggestive evidence: our vote share projections were better than 538’s in both 2020 and 2016 despite not adjusting for past error
All of this is muddied by assigning a letter grade to a pollster that gives more weight to “gold standard” phone pollsters that have little recent record of outperforming alternatives. It’s kinda easy to see this breaking down if you pay attention to evolutions in methodology.
When America's electoral institutions let you win the House, Senate, and Presidency while losing the popular vote, there are fewer incentives to actually listen to the will of the people
To me, one of the reasons why polling is so crucial to good governance is that there is literally no institution in America's federal government that gives a voice to the raw political majority 100% of the time. Polls work the ways democracies (and republics!) are supposed to.
One of the hardest parts about writing a book about trusting the will of the people to guide our gov right now is that about 25-30% of the public has basically been brainwashed by an anti-fact media ecosystem and radicalized by bad-faith opinion leaders.
So... why trust people?
The answer is at once obvious and a hard pill to swallow: The vast majority of Americans still see Jan 6 for what it (verifiably) was: a violent mob of Trump supporters trying to kill oppo leaders and overthrow the government. So we can still trust the majority. But we also have
to acknowledge that a ~fourth of Americans aren't adding to the wisdom of the public, bc they aren't getting information that would lead them to rationalize or otherwise deliberate about what government could do for them. The hard truth is that we can't always trust the people.
Returning to normal will be a lot harder if 30% of Americans continue to refuse to get vaccinated. The good news: that number is trending down view.e.economist.com/?qs=507097ddb1…
Given that a person's political orientation is the biggest predictor of vaccine hesitancy, and we're geographically polarized, I wouldn't be surprised if we see lots of blue areas approach normalcy before red areas, where refusing vaccines and guidelines could prolong circulation
The good news, for now, is that Republicans don't seem to be getting *more* hesitant to take the vaccine, as we may have expected given typical post-inauguration polarization in attitudes. I consider this one of the biggest sources of optimism in months.
I really wonder how different our conversation about "the future of polling" would be if people just lowered their expectations for horse race precision down to the appropriate level
I am not denying that polling error exists, merely arguing that slightly-larger-than-average errors should be less surprising to people. I think this tweet proves my point. Four points of bias on vote share is not rare, nor does it make polling useless
The easiest way to make progress might just be for pollsters to double whatever traditional margin of sampling error they're reporting. (Would not be opposed to arguments in favor of tripling.)