UPDATE here at the top: For whatever reason, my link to Bitecofer’s article no longer works, and it appears that perhaps the Niskanen Center may have pulled the article. Looking to see if Bitecofer discusses in her twitter feed, but haven’t found anything yet. Regardless, I have a saved copy of the whole thing. If anybody figures out what happened, do share.
Yeah - that’s like about 12-15 points higher than the next most optimistic forecast I’ve seen and while I love reading it, I don’t buy it (yet), given several items I will discuss below. But regardless, here are the maps and charts for president and senate win that too)
You need to read her full piece , (note, link now points elsewhere but to the correct article) but some of the highlights include:
After discussing the highlights of her new model (discussed below), she noted:
“I opined in March that even if the pandemic was mismanaged, even if that mismanagement did cause Americans great death and despair, and even though what had been a strong “economic fundamentals” for Donald Trump’s reelection campaign was inevitable, hyperpartisanship meant that it was unlikely these massive political stimuli would have much impact on the fundamentals of the race. Regrettably, that is exactly where we find ourselves heading into the fall general election. I use the word “regrettable” because the inelasticity of American public opinion is a symptom of a democracy in full-blown crisis. A healthy “body politic” does not remain unresponsive to political stimulus on an epic scale.”
And that seems to have held true, as we are about where we were before the Dem primaries began (during which time there was lots of distraction and upheaval due to the intense early primary season). She specifically references two Quinnipiac surveys, one in December, 2019, showing Biden with a 9 point lead over trump, and one this September, showing a 10 point lead.
As to the “what about the swing states and 2016?” issue, she had this to say:
“And what about the swing states? The refrain of “there’s the popular vote, and then there’s the Electoral College” has become like nails on a chalkboard for this analyst, because it doesn’t mean what people think it does. Yes, it’s true that we don’t have a national election, we have 50 state elections, and that the Electoral College decides the presidency. It’s also true that the Democrat’s voter coalitions are not as evenly, or ideally, distributed as the Republican Party’s coalition. But, as was revealed in the 2018 midterms, Democrats did not need to win the popular vote by 11 points to retake the majority of the House of Representatives at just 23 seats. They won 40 seats at just over an 8 point margin. Likewise, there are some misconceptions as to how favorable the Electoral College currently is to the GOP, largely powered by the misunderstanding of exactly how Donald Trump won in 2016, and conversely, how Hillary Clinton lost.”
She goes on to add:
“In this cycle though, disunity between the popular vote and the Electoral College is unlikely. The Republican Party’s “small tent” strategy of appealing primarily to white voters and using voter suppression strategies to mitigate the damage is likely to keep Trump’s share of the two-party vote share low, well below 50%. To win, the Trump campaign needs to replicate 2016- they need to drag the winning two-party vote margins in the swing states to “plurality” vote shares: something below the 50% mark. Based on Trump’s polling, which is consistently stuck at 46-47%, well below. Then they must subtract from Biden enough votes that his ultimate two-party vote share ends up even lower than Trump’s- as what happened to Clinton. This can only happen with another round of atypically high third-party and write-in balloting. This time, polling data makes no suggestion of the voter sentiments that drove 2016’s high defection rates. Not only are voters much more likely to choose either Biden or Trump, the percent indicating undecided is already low- the complete opposite of the conditions everyone should have seen plainly heading into 2016.”
She then spends a number of paragraphs pointing out any number of other harbingers of change (e.g., if Jaime Harrison is coming close to beating Lindsay Graham, then it is a pretty safe assumption that Cal Cunningham is ahead of Tillis in NC). She also devotes a lot of time to the “known unknowns” of pandemic, vote by mail fiasco, college student voting uncertainty because of location uncertainty, etc.
Finally, after some other discussion, well worth reading, she lays out her weighting methodology:
- 30% to trump approval/disapproval (using CIVQS)
- 30% to state-wide margin of Democratic US house candidates relative to GOP
- 15% to 21st century incumbent disadvantage relative to midterm performance (you’ll have to read the article to understand that one)
- 15% to something called “Realignment: Measuring State Performance Beyond National Trends” (a good example might be how VA and CO have moved leftward faster than the country as a whole)
- 5% to state-wide population density. I think this is a way to capture the urban/suburban/rural divide.
- 5% to statewide white, no college population
My reactions:
- While I don’t believe the odds are that good, it was nice to see another model that showed Biden as a strong favorite
- Given the current estimated national lead of around 7% on the low end to maybe 10% at the high end, that range, per Nate Silver and 538, does in fact comport with a high 90s likelihood of winning the EC.
- The limitations, aside from all of the other issues, e.g. voter suppression in it many forms, the pandemic, Bill Barr’s acknowledged pursuit of something, anything to dump in October, are at least two-fold: One, this is almost totally based on data OTHER than “who are you going to vote for?” polling - the closest that comes is the trump approval/disapproval number. Second, in this and many other models, each state is treated as independent, i.e. if, for example, both GA and NC poll as “toss-up,” each is analyzed independently (coin flip) and you have a 25% chance you win/lose both and a 50% chance you win one. However, we know that isn’t how things actually work (witness WI, MI, and PA circa 2016), and in fact states do correlate with each other, such that if both states were truly equally likely to go to Biden, if you knew on election night that one state was definitely headed that way, then you could bet better than even money that the other was gonna go that way too.
- Finally - this is certainly consistent with what Charlie Cook is quoted as saying, which is that the cake is baked. If so, let’s do hope it’s a dark blueberry cake and not raspberry or strawberry.
Thursday, Sep 17, 2020 · 5:56:30 PM +00:00 · Denver11
UPDATE #2 Here is Bitecofer’s explanation re the dropped link:
“FYI: the forecast update is temp unavailable while we make changes to it so that people can understand how this modeling approach represents a MASSIVE departure from the models they are used to- even though it does offer probability estimates. Hope to be back up in a few hours.”
