More on Mayer from Manifold…
Jay Manifold sent e-mail about yesterday’s post on the Mayer predictive model of primary campaigns. Here’s what he says:
I think the flavor of your suggestion about the Central Limit Theorem is good, but it doesn’t explain why the Mayer model has worked for 6 elections in a row. Applying the CLT suggests that Mayer should be right most of the time, say for 4 out of the 6; then the other 2 elections would have had a major-party nominee who didn’t lead in the last national poll before the caucuses and primaries started. This would, in turn, be caused by many small, unrelated random effects (see a related comment here).
But he’s nailed it every time. Some things to watch for, then, are:
1- Applicability of other kinds of distributions, or of bounded distributions; unless caucus/primary processes are flawlessly proportional in the allocation of delegates based on their returns, the distribution of delegates is likely to be bounded by some lower cutoff. An interesting cautionary tale is here.
2- I note that election returns where many candidates are on the ballot are likely to exhibit log-normal (power law) distributions. [Ed. Note: The recent California recall.]
3- A complete miss next year. An effective tie among several candidates could produce a situation where not only would no one have a majority, no two could form a majority. Now that would be some fun!
I was struck by Jay’s reminder that Mayer has “nailed it every time.” This is significant. A miss this year would (should) not raise many questions about the model itself. Rather, statistically, it would properly be seen as a normal fluctuation of random events.
As I said in my original essay (link above), if journalists accept Mayer then they have to accept that the primary process is not the thing that we think it is–a dramatic race with an uncertain outcome. If we take away the drama, if we assert–accurately!–the nominee months before the convention, what then is their to report?
Hint: the P-word.
