A wicked early look at the 2020 Democratic presidential nomination

On March 11, 2019, the Democratic Party announced that its 2020 national convention will be held at the Fiserv Forum in Milwaukee, WI. This is a reasonable choice, given that Wisconsin—which 2016 Democratic presidential nominee Hillary Clinton lost by 0.8 percentage points—is one of the true swing states in presidential elections.

It also means the formal selection of the 2020 Democratic presidential and vice-presidential nominees will not take place for another 15 months (July 13-16, 2020). Nonetheless, nearly two dozen Democrats have either declared their intention to run for president in 2020—or are strongly leaning toward doing so.

Put another way, just since January 1, 2019, there have been…

37 national polls (17 by Morning Consult alone)
6 Iowa Caucuses polls
5 New Hampshire Primary polls
1 Nevada Caucuses poll
3 South Carolina Primary polls
1 poll each from the primaries in Alabama, California, Michigan, Florida, Wisconsin, Pennsylvania and Oregon

…conducted of the race for the 2020 Democratic presidential nomination, for a total of 59 different polls in three months.

The 37 national polls themselves queried voters about 39 potential nominees; when you include candidates only included in state polls (Massachusetts Representative[1] Joseph P. Kennedy III and billionaire Tom Steyer in New Hampshire; Miramar, FL Mayor Wayne Messam in Alabama and Nevada), the number increases to 42. And not one of these 59 polls listed Ohio Representative Tim Ryan, who declared his presidential candidacy on April 4, 2019—bringing the total to 43.

**********

The point is, despite the Iowa Caucuses—the first contest to choose delegates to the 2020 Democratic National Convention—being nearly 10 months away (tentative date February 3, 2020)—it is not too early to begin to examine public polls assessing voters’ 2020 Democratic presidential nominee preference.

Thus, just as I did with public polls for governor and United States Senate (“Senate”) in 2018, I am collecting and aggregating 2020 Democratic nomination polls released publicly since January 1, 2019. The difference from 2018, though, is that rather than “project” who the 2020 Democratic nominee will be, I am simply taking snapshots of the relative ordering of candidates for that nomination.

Given that convention delegate selection occurs at the state level, I calculated a version of my “weighted-adjusted polling average” (WAPA) for each candidate, both nationally and within each state contest for which I have polling. I then calculated a single WAPA across all of these levels, which I explain below.

I calculated each candidate’s national/state-level WAPA this way:

First, I weighted the raw percentages for each candidate by

The FiveThirtyEight.com pollster rating, converting their letter ratings to a numeric value (A+ = 4.3, A = 4.0, etc.). Thus, each weekly Morning Consult (B-) tracking poll is weighted 2.7/4.3 = 0.628.
A ratio of two values: a) the number of days a poll’s midpoint[2] is since January 1, 2019 and b) the number of days between December 31, 2018 and the contest being assessed (state contest or July 13, 2020, the first day of the 2020 Democratic National Convention). Thus, the University of New Hampshire poll of 240 likely 2020 New Hampshire Primary voters conducted February 18-26, 2019 had a midpoint of February 22, 2019, which is 54 days before the primary. There are 407 days between December 31, 2018 and February 11, 2020, the likely date of the 2020 New Hampshire presidential primary. Dividing 54/407 gives this poll a weight of 0.133.

In essence, recent, higher quality polls count more in WAPA than older, lower quality polls. I also considered weighting each candidate poll percentage by the number of respondents. However, given that the 12 national Morning Consult tracking polls average 13,173 respondents, while the other 25 national polls average 558 respondents, that would give way too much weight to the former.

Second, to account for pollsters releasing multiple polls of the same contest (e.g., the 37 national polls noted earlier were conducted by just 14 unique pollsters[3]), I averaged two versions of the candidate polling average detailed in the first step:

Treating each poll as a statistically-independent event.
Calculating aggregate averages for each pollster then aggregating those.

To complete these steps, I needed to make two decisions regarding each poll.

I assigned a value of 0% to any candidate not listed in a poll who was included in at least one other poll. I realize this may introduce some slight mathematical bias[4], but as FiveThirtyEight.com’s Nathaniel Rakich wrote, “The good news is that, in most polls, the candidates who are rotated in or out are polling poorly, so it doesn’t make a huge difference whether they’re included or excluded.”
If a pollster released two versions of a poll on the same day—with and without one candidate (usually former Vice President Joseph R. Biden, Jr. or Hillary Clinton)—I used the more inclusive poll. While this could potentially introduce significant mathematical bias, this is (hopefully) mitigated by the infrequency of the practice; the bias will also drop considerably over time as the field of 2020 Democratic presidential nomination candidates becomes more set.

Having calculated 12 different candidate WAPA, the final step was to combine them into a single “national-and-state-weighted WAPA” (NSW-WAPA).

It is well-established that candidate performance in Iowa affects candidate performance in New Hampshire, in turn affecting candidate performance in every contest that follows; I still think this is the definitive work on the subject:

Bartels Presidential Primaries.JPG

Therefore, I weigh the earliest primaries and caucuses—Iowa, New Hampshire (February 11[5]), Nevada (February 22) and South Carolina (February 29)—higher than subsequent contests, which in turn I weigh higher than national polls:

Iowa, New Hampshire = 5
Nevada, South Carolina = 4
The weighted-adjusted average of subsequent state contests = 2
National polls = 1

The weighting formula to combine 12 WAPA into a single NSW-WAPA is thus:

(IA*5 + NH*5 + NV*4 + SC*4 + OtherStateAverage*2 +National)/21

Yes, the results from Iowa and New Hampshire will change the subsequent national and state polling, but in ways that are currently unknowable. This formula represents a compromise between underweighting the early contests on one hand and overcomplicating the formula on the other hand.

Also, I chose to use an average of the WAPA from the post-South-Carolina contests because, as of this writing, I only have a single poll from each of seven such contests. I may revisit this decision as more polls are conducted of 2020 Democratic presidential nomination preferences in these states.

**********

As of March 31, 2019, here is the relative position of 2020 Democratic presidential nomination candidates, regardless of announcement status.

Table 1: Weighted-adjusted polling averages for selected 2020 Democratic presidential nomination possibilities

Candidate	Running?	National	Iowa	NH	*NSW-WAPA*
Biden	???	29.9	26.6	24.5	*28.5*
Sanders	Yes	20.6	18.5	23.5	*20.8*
Harris	Yes	9.9	9.0	11.9	*10.1*
Warren	Yes	6.5	8.4	8.4	*8.2*
O’Rourke	Yes	7.2	5.1	5.0	*6.2*
Booker	Yes	4.0	4.5	3.8	*4.2*
Klobuchar	Yes	2.0	3.4	4.4	*2.7*
Buttigieg	Yes	1.1	4.2	0.6	*2.4*
Gillibrand	Yes	0.8	0.3	1.4	*1.1*
Gabbard	Yes	0.5	0.4	0.3	*0.9*
Yang	Yes	0.5	0.1	0.0	*0.8*
Castro	Yes	1.0	0.9	0.2	*0.6*
Bloomberg	No	1.3	0.0	1.5	*0.4*
Brown	No	0.6	0.5	0.7	*0.3*
Delaney	Yes	0.4	0.7	0.3	*0.3*
Hickenlooper	Yes	0.7	0.3	0.0	*0.2*
Inslee	Yes	0.4	0.6	0.0	*0.2*
H. Clinton	No	1.6	0.0	0.0	*0.09*
Swalwell	???	0.0	0.1	0.0	*0.09*
Bullock	???	0.2	0.2	0.0	*0.06*
Bennet	???	0.0	0.2	0.0	*0.05*
M. Obama	No	0.7	0.0	0.0	*0.03*
J. Kennedy III	No	0.0	0.0	0.1	*0.03*
McAuliffe	???	0.1	0.0	0.0	*0.02*
Holder	No	0.2	0.0	0.0	*0.02*
Kerry	No	0.3	0.0	0.0	*0.02*
de Blasio	???	0.1	0.0	0.0	*0.01*
Moulton	???	0.0	0.0	0.0	*0.007*
Winfrey	No	0.1	0.0	0.0	*0.005*
Abrams	???	0.1	0.0	0.0	*0.004*
Williamson	Yes	0.1	0.0	0.0	*0.004*
Steyer	No	0.0	0.0	0.01	*0.003*
Cuomo	No	0.02	0.0	0.0	*0.001*
Messam	Yes	0.0	0.0	0.0	*0.0*
Ojeda	No	0.0	0.0	0.0	*0.0*
Ryan	Yes	0.0	0.0	0.0	*0.0*
7 Others	No	0.0	0.0	0.0	*0.0*
DK/Other		9.1	16.0	13.3	*11.7*

There is a glorious absurdity to listing candidates averaging well under 1.0 percentage points, but 32 candidates garnered at least 1% support in one or more of the 59 polls assessed. However, given that one Democratic National Committee criterion for participating in the first 2020 Democratic presidential debate (hosted by NBC, MSNBC and Telemundo June 26-27, 2019 [6]) is to register at 1% in at least three designated national polls released after January 1, 2019, even that low level of support matters.

The data in Table 1 suggest the following as of March 31, 2019:

The leading candidate, Biden (28.5%), has not yet formally announced whether he will run for president in 2020.
On average, more than half (50.7%) of poll respondents choose someone other than Biden or Vermont Senator Bernie Sanders (20.8%) to be the 2020 Democratic presidential nominee.
Competing for 3^rd place is California Senator Kamala Harris (10.1%) and Massachusetts Senator Elizabeth Warren (8.2%).
The three candidates whose support in national polls is understated relative to Iowa and New Hampshire polls are Warren, Minnesota Senator Amy Klobuchar (2.7%) and, especially, South Bend, IN Mayor Pete Buttigieg (2.4%; 4.2% WAPA in Iowa). Hawaii Representative Tulsi Gabbard (0.9%) is also buoyed by WAPA of 2.0% in Nevada and 1.4% in South Carolina.
Despite ranking #5 overall, however, former Texas Representative Beto O’Rourke (6.2%) currently does not poll as well in the early states as he does nationally.
Rounding out the top 10 are New Jersey Senator Cory Booker (4.2%) and New York Senator Kirsten Gillibrand (1.1%).
Half of the current top 10 choices for the 2020 Democratic presidential nomination are women—including four Senators!
Just outside the top 10 are entrepreneur Andrew Yang (0.8%) and former Secretary of Housing and Urban Development Julián Castro (0.6%).
Other than Hillary Clinton (0.09%), the top-listed candidates who have declared they are NOT running for president in 2020 are former New York City Mayor Michael Bloomberg (0.4%) and Ohio Senator Sherrod Brown (0.3%).
Both Bloomberg and Brown continue, on average, to poll just ahead of three other announced candidates for the 2020 Democratic presidential nomination: Maryland Representative John Delaney (0.3%), Colorado Governor John Hickenlooper (0.2%) and Washington Governor Jay Inslee (0.2%)
Seven potential 2020 Democratic presidential nomination candidates—California Representative Eric Swalwell (who may announce on April 8), Montana Governor Steve Bullock, Colorado Senator Michael Bennet, former Virginia Governor Terry McAuliffe, New York City Mayor Bill de Blasio, Massachusetts Representative Seth Moulton, former Georgia House Minority Leader Stacey Abrams—all have NSW-WAPA less than 1.0%.
As for the remaining declared 2020 Democratic presidential nomination candidates: author Marianne Williamson (0.004%) is languishing at 15^th among declared candidates (and #30 overall), while Messam registered 0% in both polls listing him; Ryan has yet to be included in a poll.
One candidate, former West Virginia State Senator Richard Ojeda (0.0%), withdrew his candidacy on January 25, 2019.
The other potential candidates (all of whom have announced they are not running for president in 2020) to register at least 1% in any of the 59 polls assessed here are former First Lady Michelle Obama (0.03%), Kennedy (0.03%), former Attorney General Eric Holder (0.02%), former Secretary of State John Kerry (0.02%), businesswoman Oprah Winfrey (0.005%), Steyer (0.003%) and New York Governor Andrew Cuomo (0.001%).
The other seven names to be included in at least one national poll (all registering 0% support) are attorney Michael Avenatti, Los Angeles Mayor Eric Garcetti, Virginia Senator Tim Kaine, California Governor Gavin Newson, former Massachusetts Governor Deval Patrick, former Starbucks CEO Howard Schultz and Facebook CEO Mark Zuckerberg.
Despite being asked about 42 different candidates, roughly 1 in 8 respondents did not state a preference or wanted an unlisted candidate.

In sum, there is currently a clear pecking order among potential 2020 Democratic presidential nominees: Biden and Sanders are the leaders, splitting just under half of the overall vote between them, followed by Harris, Warren, O’Rourke and Booker (29% total). Just behind these six candidates are Klobuchar and Buttigieg (5% total), meaning just eight candidates currently have 83% of the vote between them.

The only other candidate to top 1.0% is Gillibrand, though Gabbard—who rounds out the top 10—rounds up to 1.0%, as do Yang and Castro. However, with 16 declared candidates for the 2020 Democratic presidential nomination appearing in at least one poll (plus the just-announced Ryan) and another eight people contemplating running as well, the race is very much in flux with more than 15 months until Democrats convene in Milwaukee on July 13, 2020.

Until next time…

[1] That is, member of the United States House of Representatives.

[2] If the midpoint falls between days, I use the later day.

[3] I treated Morning Consult tracking polls and the Morning Consult/Politico polls as coming from distinct pollsters.

[4] The arithmetic difference between an estimate and the “true” (unmeasured) value.

[5] Post-Iowa primary/caucuses dates obtained here.

[6] It will be held over two nights to accommodate up to 20 candidates, with no more than 10 appearing each night.