It also got me thinking about Michigan. The marriage ban passed in Michigan in 2004 by 59%. By the formula in given in 538.com, a 2% drop per year in the support of the ban means we could go for a repeal right about … next November. After all, we will be 5 years after the ban was enacted and support for it by then should be about 49%
Alas, I've heard no talk of mounting a repeal and even 538.com predicts Michigan won't repeal and enact gay marriage until 2013. I'm well aware that support for a gay marriage ban (which was at 59% in 2004) is quite different than support for the repeal of the ban and different again from enacting actual permission for gay marriage.
My friend and debate partner has a doctorate in Mathematics, used to be a college math professor, and still occasionally gets his hands dirty with the stuff. He offers additional insights. You shouldn't need a math degree to understand his basic ideas.
I agree that the battle has turned and will eventually be lost (or in our view, won). Living through that prolonged loss should discourage the fundies and help to weaken their movement. They have staked a lot of reputation on this issue.
This of course intrigued my mathy side. I read the 538.com article and stored it for future reference.
Two lines of comment:
1) The existence of the model and attention paid to it will in itself change the phenomenon, leading to errors -- perhaps large -- in the predicted outcomes vs. actual. For example, I mentioned the damage these defeats will do to evangelism, whose demise we have recently, separately, discussed.
Bigger: Silver's list of the years in which victory "should" be available state by state will tempt activists to mount campaigns to reap those victories (perhaps too early: people are optimistic). The resulting public attention will alter the trends in many states -- all that attention has a non-linear effect. Nothing in his model accounts for that emotional process. It's hard to say whether this will be more or less success for pro-marriage campaigns; the effect may vary state by state, depending on who gets the media sound bite du jour.
I base this comment on the famous insight from physics: measuring something changes it. The classic example is that we can't simultaneously measure (know) velocity and momentum.
I would not be surprised to learn that the errors that result are a couple of years or larger. This could lead some activist campaigns (on either side) into major errors in judging what is politically possible in a given state and year: a chaotic dimension.
2) Silver's choice of variables is reasonable and he tested his dimension reduction plan. But he appears to use linear regression, giving a linear model. My evidence: 2% per year, not an X% compounded reduction per year, which would give an effect that slows down exponentially, being X% of a shrinking number.
I would not expect a social science phenomena to follow a linear model for very long. Instead, I'd expect a power law and, even then, large errors. The result of that change in methods is to introduce multiplicative effects -- change accelerates by feeding on itself or slows down geometrically, yielding a long tail on the distribution. These could happen together, giving a faster-than-expected bulge of state victories in the middle, transitioning to a long tail.
I do expect a long tail: For deep south states, 2024 seems very optimistic. They're still fighting the Civil War.
And, as some responses to the 538.com article point out, Utah is a special case. Hard to imagine that turning by the predicted 2013.
The long tail means at some point the Supremes will act, ending the need for individual states to do anything.
No comments:
Post a Comment