/ About / Archive / Other media
Nate Silver's new candidate ranking method favors some candidates, penalizes others - 4/10/19
Nate Silver has posted his "current thinking on the tiers in the Democratic primary, in terms of likelihood of winning the nomination." As usual, it has come under a lot of criticism from people with different intuitions about how the candidates should be ranked.

Fortunately, however, we don't have to rely on subjective intuitions - because Silver has done this before. Back in December, Silver posted another tiered ranking of the candidates; but then, he explained that his ranking was based on Likert scores. This makes the comparison straightforward: all we have to do is calculate updated Likert scores based on recent polling, and we can figure out how the candidates would rank using his previous methodology. Here's how it shakes out:

Here I have also assigned each candidate a "corrected tier" that - like Silver's post in December - is based on their Likert ranking. This gives us a clear view of how his new ranking differs. Red cells indicate that Silver has underranked the candidate; green cells indicate that he has overranked them.

When Silver posted his December rankings, I noted a serious problem in his method for calculating Likert scores that significantly changed candidate ratings when it was corrected. This time, instead of correcting that error, I have simply adopted his previous method to demonstrate that Silver is not even applying his flawed methodology consistently. Note that there are, between his last ranking and this one, two major points of continuity:
  • In both rankings, he significantly errs in favor of Kamala Harris. The last time, failure to account for "not sure" ratings boosted her score by .3 points; this time, he has ranked her in tier 1a, when she should at best be in tier 1b.
  • In both rankings, he significantly errs at the expense of Bernie Sanders. The last time, failure to account for "not sure" ratings dropped his score by .03 points, placing him in tier 1b rather than 1a; this time, he again places Sanders in tier 1b when he should be in 1a.
I don't think that much further elaboration is necessary. There is no math or method to Nate Silver's madness; he's simply making it up as he goes along, and changing his approach in ways that appear to consistently favor some candidates and penalize others.