The conventional wisdom for the Democratic Primary right now is that Senator Bernie Sanders is fading while Senator Elizabeth Warren is rising into a two-way race with former Vice-President Joe Biden. Nate Silver made the case after Debate 3 (September 12) in article form and has continued, via Twitter, to cherry-pick polls to fit this beatified narrative.
I am a bit loathe to engage in the exercise that follows — that is to highlight the evidence that runs counter to this narrative. While it feels a bit like cherry-picking in the opposite direction, it also just is a necessary counter-weight to a herding instinct that does not adequately describe where the race is currently on average or in terms of a view of the race with delegates and chronology in mind, let alone one that analyzes things with an in-depth understanding of demographics like age, race/ethnicity, and education level.
Previous research that I did on the topic of media coverage (high) for polls that Bernie Sanders is doing poorly in versus media coverage (very low) for polls that show him doing well became a touch stone for a war of words between corporate media and the Sanders campaign. Not much has changed in the intervening weeks.
Each of the polling firms in what follows are rated B+ or better. The only exception is the UC Berkeley poll that is headed up by the former director of California’s Field Poll. Field Poll was rated A+ before it closed up shop.
Here are ten polls since Debate 3 that show Sanders doing quite well, but that you would have heard hardly anything about if your primary source of information is major newspapers or television news programming:
SurveyUSA (California) Sept. 13-15 – Joe Biden 27 Bernie Sanders 18 Elizabeth Warren 16. Three of the five California polls in the field in September have shown Elizabeth Warren in the lead, but this poll with a heavy Hispanic respondent pool has Warren third and Sanders in second, comfortably above the 15% threshold for winning delegates by state and district. With 416 delegates available on Super Tuesday, Californians will be responsible for 28% of all delegates available through March 3. The poll shows Sanders in 2nd place with Latinx/Hispanic voters behind Biden and tied for first with him with California’s African American likely Democratic Primary voters. The Likely Voter screen for this poll was set super high, providing a strong handicap against Sanders, many of whose supporters in other polling have said they are not really tuned into the political process at this point.
SurveyUSA (National) Sept. 13-15 – Joe Biden 33 Elizabeth Warren 19 Bernie Sanders 17. While several polls trumpeted in the media since Debate 3 have shown large, even double digit leads for Warren over Sanders for second place, this poll, again with a very high Likely Voter screen, follows Morning Consult and a handful of other high quality or high sample sized polling in showing a continued, tight two-way contest between Warren and Sanders for 2nd place nationally. While Sanders trails Biden by 20% with African American voters in this polling, he holds a slight edge with Latinx/Hispanic voters at 25-23 over Biden (Warren is back of Sanders by double digits in both categories). This national polling sample for Hispanic voters was matched in this time frame by Latino Decisionswhich also shows a two point advantage for Sanders nationally with Warren a significant way back in 3rd. (On a similar note, Sanders leads Biden with Hispanic voters 26 to 23 with Warren at 15% in the latest cross tabs from Morning Consult, which they kindly emailed to me. Biden is up 41 to 20 over Sanders with African American voters, Warren trailing at 13% in the same.)
Monmouth (New Jersey) Sept. 12-16 – Joe Biden 26 Elizabeth Warren 20 Bernie Sanders 18. While Monmouth results for New Hampshire with Sanders trailing badly (based on a sample that is 70% 50-years-old or older) have grabbed a bunch of headlines and been the basis for tons of snark for centrist data-types, this poll by the same outfit in New Jersey, balanced closer to 50-50 with younger and older voters, has received far far far less attention. Clinton won the state by 30%+ in the 2016 primary, so a result with Warren and Biden both within 8% and the poll’s Margin of Error on the gap is … stunning. Warren is tied with Biden among New Jersey’s white likely Democratic voters while she trails Biden (26%) and Sanders (21%) significantly with just 12% of non-white voters. New Jersey is the state that looks most like the Democratic electorate nationally in terms of race and ethnicity. It’s no accident, then, that well-weighted polling from a good polling firm matches what we are seeing nationally on average – a tight three-way race with Warren, Biden, and Sanders all within 10% or less of each other.
Emerson (California) Sept. 13-16 – Joe Biden 25.8 Bernie Sanders 25.5 Elizabeth Warren 20.4. Three of the polls in this list of ten are from California. In this one, Sanders is in a dead-heat with Biden for 1st place. One thing that Emerson is transparent about and does well where other firms either do not disclose or do poorly is in reporting how respondents self-report their vote from the 2016 Democratic primary. While such reporting can be off by a bit here and there, some of the few other firms that have reported such splits in this cycle have had response rates wildly slanted toward Clinton 2016 voters as compared to actual results nationally or by state.
Fox News (National) Sept. 15-17 – Joe Biden 29 Bernie Sanders 18 Elizabeth Warren 16. Not only was Sanders pegged second in this “high quality” phone only poll by a firm rated “A”, but this poll also marked a definitive improvement over Fox’s previous poll in mid-August where he trailed Warren by 10%, a poll result that, predictably, was widely discussed by other legacy media outlets and centrist data pundits in social media. As I have noted, there is now a clear gap between phone and online polling for this cycle with, for instance, Sanders doing 10% better in Iowa in recent online polling as compared to phone only polling. It is not hard to guess why: younger voters are now known quite well for “not answering the phone,” and this would be particularly true for those who have high medical or student loan related indebtedness.
UC Berkeley (California) Sept. 13-19 – Elizabeth Warren 29 Joe Biden 20 Bernie Sanders 19. The West has long been Joe Biden’s weakest region, and California has a ton of Hispanic/Latinx voters who, as previously noted, slightly favor Bernie Sanders. White Californians are going big for Warren per this polling and even more so in a Capitol Weekly poll from Sept. 1-13 that matches this topline result nearly perfectly (Warren 29 Sanders 21 Biden 18). The strict average of five California polls in the field in September (Change Research had a Warren 25 Sanders 23 Biden 18 poll to add to the mix of ones already mentioned here) is a very tight three way race: Warren 23.8 Biden 21.7 Sanders 21.3.
High Point University (North Carolina) Sept. 13-19 – Joe Biden 31 Bernie Sanders 20 Elizabeth Warren 15. North Carolina is among the top-ten states that “look like the Democratic” electorate as a whole. This B+ rated poll, fairly weighted by race/ethnicity and age, further shows why Warren must still make up significant ground with people of color while Sanders, once again, does significantly better with non-white voters than he does with white voters. Sanders kept it closer than expected in North Carolina in 2016, losing by 14%, the best of his showings in Southern states. A further slip by Biden, with support of less than a third of the Democratic electorate already, and Sanders is the likely winner here. Significantly, these results are among registered Democratic voters only. Warren, in the same poll, has the support of just 7% of independent voters. While South Carolina gets all the attention, North Carolina votes just three days later, has twice as many delegates available, and, as noted above, is more representative of the Democratic party.
Emerson (National) Sept. 21-23 – Joe Biden 25 Elizabeth Warren 23 Bernie Sanders 22. While slagged by the New York Times’ Nate Cohn and others for using an Amazon Turk panel for the online portion of their polling, Emerson is a B+ rated polling outfit that has called more races correctly in recent years than highly regarded live caller polling firms at NBC/WSJ and CNN. Emerson is one of the few pollsters that learned from mistakes over the course of the 2016 Democratic primary (where 45% of polls in the last three weeks before a state contest missed at least partially outside their Margin of Error) and has not gone back to old ways of bad demographic splits by age, race/ethnicity, education, and more. If this national result is most accurate and nothing changes substantially in the next four months, we will be in for a heated and lengthy contest.
CNN/SSRS (Nevada) Sept. 22-26 – Bernie Sanders 22 Joe Biden 22 Elizabeth Warren 18. Even with absolutely indefensible age splits that hurt Sanders, he still showed up tied in first place for an all phone poll by CNN/SSRS, a firm that has persistently been unfavorable to him in this cycle. Taken together with the CBS/YouGov poll earlier in September showing Sanders 29 Biden 27 Warren 18, and it is not unrealistic to see Sanders taking two of the first four states (New Hampshire included where the RealClearPolitics average shows a tight three-way race) with only minor improvements in his overall numbers.
Ipsos/Reuters (National) Sept. 27-30 – Joe Biden 18 (All Voters) 21 (Registered Voters) Elizabeth Warren 14 (AV) 15 (RV) Bernie Sanders 15 (AV) 16 (RV). It would be bad data not to acknowledge that this result, while a little less so that the Sept. 23-24 result from Ipsos, is a shift away from Sanders in Ipsos/Reuters polling which had previously seen a tighter two-way race between Biden and Sanders with Warren trailing a distant third. That said, either way of sizing things up (one of which accounts for new voters coming into the process, one of which emphasizes those more likely to vote based on previous ways of understanding elections) shows, as we know also on average nationally, that we have a three-way race with a 10% spread or less among the top candidates. In this case, the spread is now just 4-6%.
As acknowledged above, this discussion of ten polls is intended primarily as a counterweight to an opposite way of cherry-picking. It should be noted, however, that using the same way of measuring (B+ or better polling firms in the field since Debate 3) shows just six polls where Sanders is in third place or worse, trailing second by 3% or worse. While those six polls (NBC/WSJ national, Selzer/DMR/CNN Iowa, Suffolk Nevada, Quinnipiac national, CNN/SSRS South Carolina, Monmouth New Hampshire) have sucked up the lion’s share of media coverage and have to be taken seriously, even if many of them have serious demographic or other problems, they are far from the whole picture.
It is a three-way race per the majority of relevant high-quality polls, and even more so in my most recent update forecasting the delegate race through Super Tuesday. We will not know for a few more weeks whether Warren’s recent surge is a temporary bump after a solid performance in debate three, where Biden looked lost at times and Sanders had laryngitis, or whether it is more attributable to a steady march upward toward a high likelihood of securing the Democratic nomination. Until that picture becomes clearer, anyone who insists it is a two-way race is engaging either in poorly supported punditry that likely is based on derivative analysis or is up to axe-grinding rather than solid data work.