All of this writing and data analysis is a lot of work! So after more than five years of posting, I've finally launched a Patreon to help pay the bills.

8/16/19

What's going on with the Sanders / Warren crossover vote?

In short: we don't know. Because to measure this directly, you would need to ask respondents "have you switched from Sanders to Warren, or vice versa?" - and no one is actually asking this. There is, however, an indirect way to look at this: ask voters who their second choice is. If we then suppose that first choice votes are "stolen" from a respondent's second choice, we can get rough idea of what the crossover vote looks like. Here's a graph of how this has played out since Warren began her micro-surge in June:

The general trends here are clear. About 16% of Sanders supporters would otherwise vote for Warren: this fraction has only fluctuated slightly since the beginning of June. Meanwhile, a growing number of Warren voters say that Sanders would be their second choice - from less than 10% a few months ago to 18% today.

One point that we need to bear in mind as we look at these trends is that they represent fractions of larger numbers that have also changed. Sanders has had more supporters than Warren until quite recently, and the blue line is a fraction of that changing total; Warren was far behind Sanders at first but is now neck-in-neck with him, and her red line is just a fraction of that growing base. So when we adjust for how their overall numbers have shifted, the trendlines actually look something like this:














This is probably as close as we can come to actually measuring, in absolute terms, the crossover vote between Sanders and Warren. A few takeaways:
  • Are these candidates really stealing from each other? Again - we don't actually know! Consider this, for example: a Warren voter who wouldn't even consider Sanders in June but warmed up to him by August. This would make Warren's red line go up, but if anything it means progress for Sanders - not that Warren "took" this voter from him.
  • If Sanders "takes" a voter from Warren and Warren "takes" a voter from Sanders, the net advantage to both candidates is zero. For this reason, if you want to see how the crossover vote is impacting their margins, you would have to look at the difference between the red and blue lines, which looks something like this:
  • If the polls tell us anything, what they tell us is that Sanders has more second-choice Warren voters than Warren has second-choice Sanders voters. His advantage narrowed by about 1% in June, and since then it has fluctuated around .5%.
  • Generally, I would say that crossover voting between Sanders and Warren has probably been a wash, at the very most giving Sanders an edge of around half of one percent. And even this is a pretty shaky claim for reasons given above - the polling just doesn't give us the kind of data we actually need to measure what's happening. But this is, on the other hand, a somewhat more rigorous way of looking at what's happening than the standard pundit approach of eyeballing overall polls and monitoring Twitter for discourse trends.