Responding to a Twitter request, I have compiled competitiveness ratings for all the Super Tuesday Democratic primaries (except American Samoa). The competitiveness rating answers the question of how competitive is this contest, adjusting for how many candidates are vying for the nomination.
To see why one can't compare the vote shares across elections with different numbers of candidates, just think about someone getting 50% of the votes. Just comparing vote shares leads to the conclusion that equal vote share = equal performance. And yet, the person with 50% share almost surely wins a 3-person contest but only ties a 2-person contest.
The following chart shows that most primaries on Super Tuesday were similarly competitive, with ratings ranging from 63% to 68%.
The competitiveness rating is a number between 0% and 100%. Zero is an even race, in which every candidate gets the same proportion of votes. One hundred is a lop-sided race, in which one of the N candidates wins all the votes. The value of the competitiveness rating is that it turns a set of N vote shares into a single dimension, which can be rank-ordered. The methodology takes into account the entire vote share distribution, not just the winner's vote share.
Competitiveness rating is not an attempt to predict how voters would shift their votes if the contest were narrowed to two candidates. I like to talk about the 2-person race just because interpreting competitiveness in a 2-person race is intuitive. If candidate A won 60% of the votes, we know B got 40% so just knowing 60% is enough to describe the competitivness. If A won 100% of the votes, the race is the least competitive, and most lop-sided. If A won 50%, then the race is even, or the most competitive.
Thus, the right way to interpret competitiveness rating is to say that the 6-person race with the vote shares (34.5%, 30%, 14.4%, 11.4%, 6%, 3.7%) has a competitive rating of 65%, which means it's about as competitive as a 2-person race with a 65%-35% vote split. [Those are real values coming from the Texas primary in 2020.]
***
Here is the table that contains the key numbers that went into the above chart:
While doing this work, I realized a couple things that I didn't hear from pundits last week. First, one should never ever, never ever, under any circumstances, compare vote shares from 2016 to 2020 for Minnesota, Maine, Colorado and Utah because those four states switched from caucuses to primaries. I just can't believe how many times I heard Bernie Sanders got a lot lower vote share in Maine.
Second, the 2020 results mirror the 2016 results to a large extent if we consider Joe Biden as representing the centrist lane that Hillary Clinton used to take. Virginia, North Carolina, even Massachusetts were states won by Clinton. The only true "flips" were Oklahoma going from Sanders to Biden but California went from Clinton to Sanders.
The other two states that flipped to Biden - Minnesota and Maine - were caucuses-to-primaries. The former was assisted by Amy Klobuchar, the home state Senator, dropping out and endorsing Biden while Maine was neck-and-neck. In fact, based on this evidence, it seems that the caucuses are not as undemocratic as they are made out to be.
***
Given the pundits' insistence on incorrectly comparing 2016 and 2020 vote shares, I put together the following list of 10 things they're hiding from us. [Sarcasm ahead]
Pundits keep telling you Bernie Sanders did worse this year compared to 2020. Here are 10 things they’re hiding from you:
- With the exception of Vermont, Biden got less vote share in every state compared to Clinton did in 2016. This shows Biden has a ceiling lower than Clinton’s. (And she lost the general election.)
- Biden’s win in Texas is a fluke. His vote share is 31% less than Clinton got in 2016 when she won. He got less than half the vote share she earned. It’s shocking he won Texas on that kind of performance.
- Biden’s wins in Tennessee and Arkansas were actually disappointing because his vote shares were 24% and 26% below Clinton’s in 2016!
- Biden’s wins in Alabama and Virginia were also unimpressive because his vote shares were 15% and 11% below Clinton’s in 2016.
- Biden lost California to Sanders by 8% but don’t be deceived, Biden embarrassed himself because his vote share was 28% lower than Clinton’s when she beat Sanders in 2016. Biden’s share in 2020 was even 22% below Sanders’s share when Sanders lost in 2016!
- Biden’s vote share in Minnesota was the same as Clinton’s in 2016 when she lost to Sanders. This win just proves Biden has a ceiling.
- In Vermont, Biden under-performed Clinton in 2016 because his vote share was 10% below hers. Therefore, Sanders’s win in 2020 is even more remarkable than in 2016.
- In Utah, Sanders won in both years but his vote share in 2020 was less than half of his 2016 share. This is yet another bad omen for him.
- In Maine, while Biden edged out Sanders in 2020, there is grave concern because when Sanders won in 2016, he had 30% higher vote share!
- Sanders was the real winner in Oklahoma because his winning vote share in 2016 was 13% higher than Biden’s vote share in 2020, proving that Sanders has a higher ceiling.
***
Of course, I don't believe in any of those interpretations. But they are the types of statements that were thrown around on the air day after day by the same pundits. I'm just perversely showing how the same logic can be used to denigrate the Biden campaign. Neither campaign should be comparing vote shares between 2016 and 2020. If one side does this, the other can fight back using the same logic.
Finally, here are comments on each item from above using the competitiveness ratings:
What Competitiveness Rating Says about the Above 10 Items
- Competitiveness rating addresses the fallacy of comparing vote shares across elections with different numbers of candidates.
- Texas is rated similar to 2016 despite the 2020 winner taking less than half of the vote share of the 2016 winner.
- Tennessee and Arkansas are both rated similarly in 2016 and 2020.
- Biden’s Alabama win is rated similar to Clinton’s. The Virginia contest is rated less competitive than in 2016.
- Sanders flipped California in 2020, and the contest was less competitive than in 2016. This is one of the big stories of the night, in my view.
-
The same vote share that lost the 2-person contest was enough to win in a 6-person contest. This shows why one should not compare these vote shares without adjustment.
-
Vermont was more competitive in 2020 than in 2016.
-
Did the pundit realize Utah switched from caucus to primary?
-
See 8.
-
Winning margin does not convey the competitiveness of a contest when comparing two elections with different numbers of candidates.
***
I now have a series of follow-up posts looking at various aspects of the Super Tuesday data.
Post #2: debunks talking points about ceilings, winning margins (link)
Post #3: California, caucuses vs primaries (link)
Post #4: overall turnout, youth turnout (link)
Comments