My #10at10 2020 Democratic Primary Model is now live. It includes delegate projections down to the state and Congressional District level (State Senate District in Texas) for every state voting from the Iowa Caucus on February 3rd through to Super Tuesday, when thirteen states vote a month later. The Model uses a variety of factors in an attempt to route around the difficult data-based issues that menace the potential success of such a projection.
There were 177 unique state polls in the three week periods before relevant contests in the 2016 Democratic Primary. Eighty of them (45%) missed outside their margin of error (MoE) in projecting Vermont Senator Bernie Sanders’ share of the vote. Fifty-six (32%) underestimated his share of the vote outside their MoE. Of the twenty-four polls that overestimated his share, twenty-two were in nine southern states including Texas. The scientific standard for accuracy in polling is 95% – for example “accurate to within +3.2% 19 times out of 20.” Across all 177 polls, Sanders’ share of the vote was underestimated by an average of 3.4%.
Three-quarters, or more [pdf], of likely Democratic voters for the 2020 primary have not settled into support for a single candidate. The contest itself will take place over a nineteen week period in which several of the candidates will drop out if they have not already in the next six months. All of this should give some idea of why models projecting delegate totals by candidate for the Primary are exceedingly rare and worthy of suspicion. At this stage, what follows is not a prediction of where things will likely wind up, but of how things would go if this model is reasonably accurate and if the Iowa caucuses were imminent. Nevertheless, the model does include defensible innovations to account for obvious data difficulties.
A big question looming over the Primary is whether any candidate can prevent a brokered convention. Even though Biden has lost ten or more points on a national lead, the former Vice-President maintains a large popular vote advantage and a delegate lead in thirteen of seventeen states that will vote through Super Tuesday. Taken together, Biden could manage, if the model is correct, to spin 44% (642 of 1470) of pledged delegates available through March 3rd out of just 28% of the national vote on average. Still, Gaffe Master Flash would have to recover serious momentum to avoid at least a second ballot at the Democratic Convention scheduled for July 2020 in Milwaukee. After his debacle last month, the debate on Wednesday may be the biggest moment in his decades as a career politician.
In Iowa, a tight race between the top-4 candidates suggests Biden, if the caucuses were tomorrow, would likely capture less than 25% of the vote but punch out a four delegate win (15-11) over Massachusetts Senator Elizabeth Warren, with Sanders (10) not far behind and California Senator Kamala Harris (5) trailing but viable statewide with 14.6% of the popular vote. In New Hampshire, Biden enjoys a sixpoint lead in the popular vote projection, but shares the delegate lead, nine a piece,with Sanders on account of the vagaries of Democratic delegate math.
Sanders is the only candidate besides Biden that the model sees winning delegates in all seventeen states and competing seriously in every one of the four major regions nationally. Sanders’ strength in every region flows out of his racially diverse, working class coalition that continues to hold strong appeal for voters under forty-five years of age. The model has Sanders at 15% (the cut off by delegate contest) or greater with voters in each of the four major groups by race or ethnicity, with White, non-Hispanic voters as his weakest point (14.9%) and Hispanic voters (20.2%) as his strongest.
Elizabeth Warren would be projected currently to win no delegates in Alabama and South Carolina and just fourteen delegates total in four other Southern states with large African American populations. Warren is strong in many other parts of the country, however, including now pushing Sanders (30 delegates) and Biden (34) in her home state of Massachusetts where the model would project her to pick up 27 delegates. Warren also does well on the left coast in California (98 delegates, good enough currently for 2nd there) but also in places like Colorado (17 of 75 delegates, 3rd) and even Texas (47 of 228, 3rd).
Could Warren and Sanders combine their delegates at a contested convention to put one of them forward as the nominee? Perhaps. But one or both would have to substantially improve their current standing. With just enough delegates for the two of them together (657) to push past Biden alone (643), they would, if the contest started today, have to catch fire in the final three months of the primary or broker a much larger deal to prevent Harris, candidates with much smaller delegate totals, or the superdelegates, which return this time for the second ballot, from easily pushing Biden over the top.
Harris is projected to do reasonably well in California and also, potentially, in the South. As the Model was initially being built out, Harris was enjoying the crest of her post-first-debate surge, hitting an average of 15% nationally. She has since lost 3.2% on average nationally while Biden has recovered three points. Dipping below the magic 15% threshold has seen Harris go from a projection of 275 delegates or more through March 3, 2020 to her current state in the initial full launch at just 122 delegates, almost all of them (97) from California. Even at her best, however, Harris polled fairly poorly with Hispanic, Asian, Indigenous, and “other” survey respondents. Outside the South and California, Harris’ best state in the model is Iowa, showing that a well organized, legacy media supported campaign could see her capitalize on a traditional Democratic Primary trajectory for previously not well-known candidates.
South Bend Mayor Pete Buttigieg polls fifth in terms of popular vote in the Model at 5.8% in strict average, 4.9% as weighted together with the demographic projection portion and numbers from my preliminary model, which will continue to play a small part (2-8%) going forward. Buttigieg’s failure to gain any traction with people of color and his bad luck in being a politician from Indiana have him stuck on zero in the #10at10 2020 Democratic Primary Model’s delegate count through Super Tuesday. Former Texas Congressman Beto O’Rourke is at 3% and 11 delegates, all from Texas. Minnesota Senator Amy Klobuchar is at less than 1.5% but 27 delegates, all from Minnesota. Former U.S. Housing and Urban Development Secretary Julián Castro appeal to Hispanic voters sees him at 10 delegates split between California and Texas.
There is a long way to go, with many assumptions to be challenged and at least moderately surprising, cart tipping events to unfold. So long as it can reasonably account for those dips and doodles, I will fully update the #10at10 2020 Model bi-weekly, with some adjustments around more sporadic polling and events ongoing between full updates.
Will the Democratic Convention be brokered or contested – the “Holy Grail” of political reportage? If the #10at10 2020 Democratic Primary Model is accurate and the contest started tomorrow, the answer would be technically yes but practically no. It is extremely unlikely absent a major change in one candidate’s fortunes that any of then will reach 1885 delegates or 50% on the first ballot. But if Biden can reach 44% by delegate total through Super Tuesday and maintain or increase that share as candidates drop out after March 3rd, he would almost certainly ask Harris to be his Vice-President. Superdelegates would rubber stamp the arrangement, and the whole deal would be widely understood months ahead ofMilwaukee.
Brokered ahead of the convention and requiring a second ballot, but not actually contested.
That, however, would not have been the conclusion ten days ago with Harris at 15% nationally and Biden at 25%. A candidate with less than 30% support nationally breezing to a second-ballot win would sit horribly with large segments of the Democratic electorate. The #10at10 2020 Model is built around the likelihood that three or more candidates will regularly poll over 15% nationally. In that case, Biden is unlikely to maintain a 10%+ lead over the long haul.