Wednesday, July 31, 2019

Any conversation about electability has got to start with the fact that a white male has not won the popular vote since 2004.

The electoral discussion among pundits and data journalists has been taking some especially silly turns of late and before the bullshit accumulates to the same dangerous level it did in 2016, we need to step back and address the bad definitions, absurd assumptions, and muddled thinking before it gets too deep.

We should probably start with the idea electability. While we can argue about the exact definition, it should not mean likely to be elected and it absolutely cannot mean will be elected.

Any productive definition of electability has got to be based on the notion of having reasonable prospect of winning. With this in mind, it is ridiculous to argue that Hillary Clinton was not electable. Lots of things had to break Trump's way for him to win the election and, while we can never say for certain what repeated runs of the simulation would show, there is no way to claim that we would have gotten the same outcome the vast majority of the time.

This leads us to a related dangerous and embarrassing trend, the unmooring of votes and outcomes. This is part of a larger genre of bad data journalism that tries to argue that relationships which are strongly correlated and even causal are unrelated because they are not deterministic and/or linear. In this world, profit or even potential profit is not relevant when discussing a startup's success. Diet and exercise have no effect on weight loss. With a little digging you can undoubtedly come up with numerous other examples.

The person who wins the popular vote may not win the electoral college, but unless you have a remarkably strong argument to the contrary, that is the way that smart money should bet.

No comments:

Post a Comment