The myth of centrist electability - why Bernie Sanders is the candidate most likely to defeat Trump

"I support a progressive policy agenda, I just think Sanders and Warren are too far to the left to win against Trump. We need someone more moderate to attract swing voters from the political center." How many times have you heard this statement, or even echoed it yourself when considering which Democrat to vote for in the primaries? In these polarized times, electability is certainly a valid, if not central concern for voters, especially when the political opponent is someone as widely loathed as Trump. If you care about defeating Donald Trump, the argument goes, then centrist candidates are the only safe bet, media keeps telling us. But does this idea hold under empirical scrutiny or is it merely a myth designed to keep voters in check?

The way we reason about political alignment is greatly shaped by a simplistic narrative that is being fed to us relentlessly by mainstream pundits. According to this narrative, voters are aligned along a one-dimensional political spectrum, with most concentrated in the "moderate center", and the distribution trails off as we approach the "far right" and "far left" ends of the axis. It is easy to manufacture empirical support for this model. If you limit people's choices of identification in this way, most of them will indeed identify as moderate left or moderate right of center. The figure below shows the distribution of voters based on one such self-identification survey, with the widely-accepted ideological placements of the four candidate archetypes: the liberal centrist Clinton (also implies Obama, Biden and Buttigieg), the "far left" progressive Sanders (including Warren), the "far right" Trump, and the more traditional Republican exemplified here by Mitt Romney. The large swath of voters in the middle are said to be independents or swing voters that change allegiance based on which candidate inches closer to the center where they reside (in addition to other personal attributes of the candidates, as well as outside factors, such as the state of the economy etc.) 

The problem with this picture is that it is not supported by some very basic observations from the course of recent American presidential elections.
If Trump is really that far to the right, how did he defeat a moderate candidate such as Clinton? Yes, the favorable geographic distribution of his supporters in swing states certainly helped, but he still won about 2 million more votes than Romney did in 2012. If the Republican candidate moved decidedly to the right where a fringe minority of "deplorables" resides, why did voters flock to him as a result?
If Sanders is really that far to the left, why is it that over the last year in numerous national surveys he polls almost as well, and in some instances, even higher than the much more moderate Biden in head-to-head match-ups against Trump? Shouldn't his so-called radical socialism alienate the majority of moderate voters, so that the head-to-head match-up margins are much higher for Biden?   
Some analysts invoke the role of voter enthusiasm on turnout to explain Trump's win - by moving to the right, he motivated formerly disaffected voters on that end of the spectrum to get out and vote. However, the shape of the distribution above would indicate that for every new re-enthused voter gained on the right, he conceded a couple more voters from the center by moving away from it - hardly a winning strategy.

Clearly the one-dimensional political spectrum doesn't explain electoral dynamics very well. Voters have opinions on a variety of issues, everything from gun laws to Medicare for All. On each issue their positions can range from very conservative to very liberal. To assume that everyone falls neatly on the conservative-liberal axis is to assume that people's positions across all topics are highly correlated with one another, which is not the case. A person can be conservative on abortion rights, but liberal on minimum wage laws, for example. Instead of this binary left-right axis, at the very least we should think about the issues as grouped into two categories: issues which relate to the economy and social/identity issues - whereby voters' positions are much more likely to be correlated within each group. If a person supports government intervention in markets when it comes to the minimum wage, they are very likely to also support it in the case of Wall Street regulation. This is exactly what is reflected in the two-dimensional political compass in the chart below. The horizontal economic axis represents how much a person is in favor of government intervention in the economy (left being pro-government intervention, while right being laissez-faire), while the vertical social/identity axis measures where a person stands on social justice and identity issues (ranging from very conservative to ultra-"woke").


The above figure displays the ideological positions on the political compass for a sample of 8,000 voters (source), along with color codes that signify how they voted in the 2016 Presidential election. The first thing to note is that the so called ideological "center" typically thought of as consisting of fiscally conservative, but socially liberal voters in the lower right quadrant is virtually empty, even though they are disproportionately represented among the punditry and the donor class. These pundits maintain that bipartisanship essentially means to govern from this corner of the ideological spectrum, where "moderate" Republican and "moderate" Democrat politicians come together to compromise in the name of the majority of the electorate.
The second thing to note is how well a voter's position on the compass predicts who they voted for in 2016. The figure below shows the likely ideological positions of the Presidential contenders in 2016 and how ideological distance from each might have determined voting outcome. The dividing line is rotated at an angle, which graphically captures the relative importance that economic issues might have had over social justice issues when voters decided which candidate to support. The model predicts that all voters above the dividing line would have voted for Trump, and everyone below it would have voted for Clinton. Even though this forecast didn't realize for every single voter, overall it appears to be a pretty good predictor of voting behavior. The purpose of this model is not to formalize how voters actually decide who to vote for, but as with all modeling, it is a simplified representation of reality meant to show how our conclusions about electability may change entirely once we change our framework of assumptions to better correspond with reality.

We can use this model to explain how Trump was able to flip 7-9 million former Obama voters (depending on estimates) and defeat Hillary Clinton. The green line in the figure below represents the vote split during the 2012 Obama-Romney election. For all practical purposes, we can assume that Obama and Clinton occupy roughly the same ideological placement. Trump, on the other hand, is ideologically quite different from Romney. With his racially inflammatory statements, anti-immigrant diatribes and disdain for political correctness he is significantly more extreme than Romney on social justice issues. But when it comes to economics, he presented himself decidedly to the left of his more traditional Republican rivals during his 2016 campaign: he criticized NAFTA and the TPP, he said he would not cut Social Security and Medicare, vowed to increase infrastructure spending and promised to tax corporate earnings abroad. These typically leftist stances had been unheard of from a Republican contender. Thus Trump represented an ideological shift to the left and up from Romney, which translated into winning many more white working class votes (the Obama-Trump voters caught in the triangle in the upper left corner) than those that he managed to lose with his inflammatory rhetoric (fiscally conservative-socially liberal types who in 2012 had voted for the Wall Street candidate, depicted in the lower right triangle).


What can this model tell us about whether a progressive candidate such as Sanders can win against Donald Trump? Sanders is significantly to the left of Clinton in terms of economics, but he is also somewhat less "woke", because he gives priority to class politics over identity politics. Being less "woke" here should not be interpreted as not being cognizant of social justice issues, but rather as a difference in approach to addressing them - relying on universal programs and class organizing vs. identity-based policies and identity-based organizing. We can also see this in the type of criticism aimed at him by his Democratic rivals, accusations often grounded in radical interpretation of identity politics (the moniker "BernieBros", retweeting the "transphobic" endorsement from Seth Rogan, the alleged misogyny against Warren etc.). The chart below shows the vote share that Sanders would be expected to win against Trump. We see that he is well positioned to win back a lot of the white working class Obama-Trump voters, while losing only a small number of "centrist" fiscally-conservative wealthy suburbanites, who would be turned off by his progressive agenda. In effect, the model predicts that Sanders would restore the balance that existed when Obama won against Romney, only this time between two far more populist candidates.


Thus we see how the argument about centrist electability falls apart once we take into consideration that ideology has two dimensions: economic and social. Analyzed with this additional layer of complexity, we can see how Sanders may in fact be the most electable candidate against Trump. Other empirical data specific to the 2020 election also supports this claim.

As imperfect evidence as the polls are this early in the election, it's worth noting that they unambiguously show that Sanders beats Trump in almost all battleground states. The list below shows the Real Clear Politics polling averages (as of February 7, 2020) for 15 swing states taking into account all poll surveys (no cherry-picking) in which voters were asked who they would vote for in a general election contest between Sanders and Trump. A positive margin indicates that Sanders beats Trump.

  • Florida: Tie
  • Colorado: +10%
  • Michigan: +6.7%
  • Minnesota: +9%
  • Ohio: +2.5%
  • Nevada: +2.7%
  • Iowa: -6%
  • New Hampshire: +5%
  • North Carolina: +1%
  • Pennsylvania: +3.7%
  • Virginia: +3.5%
  • Wisconsin: +2%
  • Arizona: -5%
  • Maine: +8%
  • Georgia: +0.5%

It is true that Biden's polling average margins are higher by 0-3 percentage points in most of these states, however one of the main problem with polls is that they don't take voter enthusiasm into account, which translates into voter turnout on election day. This was the reason that Trump managed to shock the nation and defeat Clinton, despite what many polls had predicted. Voter enthusiasm is especially high for non-traditional populist candidates, such as Trump and Sanders, compared to more traditional establishment candidates, such as Biden and Clinton. The large enthusiasm gap between Sanders and Biden is evident from the number of people that show up to their rallies, the number of individual donors that have donated to their campaigns (Sanders leading the field by far with 1.4 million individual donors, compared to Biden's 450,000), as well as the latest Iowa showdown. While the RCP polling average predicted Sanders to win 23% of the Iowa vote, on election day he won 24.8%. For Biden, the polls predicted he'd win 19.3% of the vote, but in the end only 15% showed up to vote for him. This constitutes an estimated enthusiasm gap of 6.1%, which is well over the margin advantage that Biden has over Sanders in all swing states. 

The geographical distribution of Sanders' supporters is also favorable to his electability argument. The NYT in August published a detailed map of the individual donors for all candidates. A low resolution version of it is reprinted below.

If we examine this map in detail, we'll find that Sanders dominates everywhere except understandably in the home states of certain candidates (such as Klobuchar in Minnesota, O’Rourke in Texas, Buttigieg in Indiana, Bullock in Montana etc) and in the coastal liberal urban bubbles (such as Manhattan & Williamsburg, Palo Alto, LA, Boston, DC, downtown Chicago and other similar pockets where the professional managerial class is heavily concentrated, which support either Buttigieg, Warren, Biden or Harris). It is virtually certain that the coastal liberal bubbles listed above will go to the Democrats in the general no matter who the nominee is. The real battleground areas, where the working class and rural voters live, is precisely where Sanders crushes his competition. 
Sanders' cross-over appeal to working class Republicans and independents is also illustrated by the fact that his Vermont Senate seat had been Republican for 145 years before he won it for the first time in 2006. Since then, he has been re-elected with 67.5% and 71% of the vote (40%+ vote margin), including 25% support from Republicans.  
While all this evidence doesn’t guarantee that Sanders ultimately will win against Trump, what it does show rather convincingly is that Sanders is the most electable candidate and stands the best chance to defeat Donald Trump in November. Most of Sanders' signature progressive policies enjoy majority support not just among Democrats, but among the general electorate as well, and he is building a historically unprecedented multi-racial working class grassroots movement which is poised to expand the electorate by bringing in many young and disaffected voters to the polls. If you're considering voting for Sanders, electability, far from being a concern, is in fact a chief argument why he should be the nominee.

Comments