This is a follow-up post to my previous post on Super Tuesday competitiveness ratings. I cover a pair of other insights that came from that line of work.
First, here is the chart the pundits have been waiting for:
Previously, I heard the pundits say that Bernie Sanders was winning but the winning vote shares were too low, compared to 2016. The above chart shows everyone's winning vote shares in 2020 Super Tuesday states for both 2016 and 2020 primaries. Almost all the vote shares earned by Biden in the states won by Biden are in the 30 to 45 percent range, which means he too must have a "ceiling".
This is an example of how not to do data analysis.
***
Next, I looked at why just looking at the margin between the winner and the first-runner-up is not sufficient to describe the competitiveness of an N-person contest (N>2).
When the margins are very high, say 30% or above, they seem to be enough to tell us those contests are lop-sided. But when the margins are 20% or below, the picture is far less clear.
Those margins are a driver of the competitiveness rating but the rating is more than just the margins.
Take Maine as an example. Because the gap between Biden and Sanders was less than one percent, one might expect this contest to have a competitiveness rating higher on the chart, i.e. closer to an even contest.
Maine is rated as less competitive than Massachusetts. In Massachusetts, the gap between the top two was 7 percent. That sounds like a less competitive race than in Maine (1%).
The key is in the vote shares down the ballot.
In Massachusetts, the top 3 got 82% of the votes while in Maine, the top 3 got 83%, basically the same. This means that the 82-3% of the votes are more concentrated in Maine on the top 2 while in Massachusetts, they are more spread out among the top 3. So Massachusetts is rated as more competitive, i.e. more resembling an even contest.
The lesson here is don't throw out too much of the data.
Comments