Without wading into the modeling wars - a debate I see as mostly fruitless - guess I'll try to summarize why I think the presidential race is more like "Lean Biden" (60%-75% range) than "Likely/Solid Biden" (90%+ range) as we head into the final night of the RNC...
1) Voters have pretty short memories compared to analysts' priors. As more Americans get used to living in a pandemic, there is potential - not necessarily a likelihood, but potential - for the dominance of COVID-19 to ebb as voters' top issue and other issues to rise.
2) As 2016 showed us, as Trump amplifies/intensifies/just plain makes up his attacks on his opponents, the media tends to react in a way that #bothsides-es each nominee's liabilities - real or perceived - *potentially* bringing their favs/unfavs into closer alignment.
3) At the moment, Trump is losing non-whites by a bit less than he lost them in '16. Whether this holds, who knows. But the voters Trump needs to improve with to get back in contention - non-college whites - supported him before and tend to live in Trump-friendly info ecosystems.
4) Polling suggests Trump's bonus in the Electoral College (relative to the popular vote) is at least as large as it was in '16 (2.9%) or potentially even larger. He could plausibly lose nationally by up to 5 points (!) and still win reelection. nbcnews.com/politics/2020-…
5) The asymmetry in modes of voting - Dems preferring mail and Trump voters set on in-person - creates additional and unique uncertainty, including potential for postal chicanery, voter error & rejected absentees that disproportionately hurt Dems. nbcnews.com/politics/2020-…
That said!! Biden's lead to date is clearly more *stable* than Clinton's was throughout '16, with fewer undecided/third party voters. I still view him as the favorite. And I'd welcome those tempted to label me as a "concern troll" to judge me by @CookPolitical's work in '18.
FWIW, despite what the state polling averages say, I still view Biden's path of least resistance to 270 Electoral Votes as MI, #NE02, PA and *AZ,* rather than FL or WI - based on demographic patterns and what we saw from hard votes in '18. Admittedly, it's a really close call.
And, just as there was a uniquely high chance of Trump winning w/ fewer votes '16, today I believe there's a uniquely high chance (maybe 3-4%) of a 269-269 tie, w/ Biden carrying PA/MI/#NE02 and no other Trump '16 turf - a scenario the House would likely decide in Trump's favor.
One more thing: a lot of the factors I’ve listed are really hard to model/quantify! Especially using historical data that might not apply to recent trends.
Which is why I’m cautious of people who are really confident their quant models are super well-calibrated.
Which is why I’m cautious of people who are really confident their quant models are super well-calibrated.
