Nate Silver Weighs In
Three days after the Orosco Execution, Nate Silver posted the following:
The whole head guillotine (WHG) is a lot like poker. The great thing about poker is that you have skin in the game, and the outcome is clear and legible. Texas has just begun a natural experiment. Whether the WHG deters murder is ultimately an empirical question. We’re going to get much more data than we have now. When we get enough, the answer will probably be pretty clear.
Historically, Texas has executed only about 2 percent of convicted murderers. Up until today, killing someone in Texas was like playing roulette. If the wheel came up 00, you got a lethal barbiturate cocktail. If it came up 0 through 26, you got life in prison. If it came up 27 or higher, you were acquitted or never even caught.
Phil Mason’s experiment isn’t that weird by global standards. It’s very similar to what Saudi Arabia does: they behead murderers. Their murder rate is roughly one-tenth of Texas’s.
I don’t want to be too long-winded, but this is why I’m launching the Texas Body Count Tracker.
The red line shows the number of Texas murders since the Orosco execution.
The blue line shows the average number of Texas murders on the same calendar days over the last three years.
Two methodological notes:
1. Murder is highly seasonal, so we’re comparing November to November, March to March, etc.
2. I only went back three years for the baseline because there was a pandemic homicide spike, and I don’t want simple mean reversion to be mistaken for the deterrent effect of the Cylinder of Death.
The light-blue shaded area represents a 95% confidence interval. We’ve modeled the variance in weekly and monthly body counts using a proprietary process with fat-tailed t-distributions. This is a nerdy way of saying: if the actual murder rate falls below the lower bound of the shaded region, the guillotine is probably working.
Right now is like the day after a presidential debate. Even if there’s a strong signal, we won’t have enough data to confirm it. But Texas has averaged about 40 murders a week, so a really strong signal might show up by Thanksgiving.
————
Coach Chief Judge Dodson was encouraged. He bought an Apple Watch so he could check the Body Count Tracker during practice. But he didn’t trust Nate’s proprietary modeling — he thought it might be a trick by the Indigo Blob. In his hierarchy of threats to Texas justice, the Indigo Blob ranked somewhere between the Houston Chronicle editorial board and stochastic terrorism.
So Dodson doubled down on studying Cholesky decompositions. To him, they were the best way to extract a clean correlation signal from messy, heteroskedastic crime data. His assistant coaches began finding eight- and ten-dimensional Gaussian eliminations scribbled across playbooks. Sometimes he’d shout “Pivot! Pivot!” and a sophomore cornerback would think it meant to shift coverage, not realizing Dodson was talking to an upper-triangular matrix in his head.
As time went on, Dodson came to see that Nate’s models were a little better than his Cholesky decompositions. But he kept doing them anyway, just in case Nate’s boyfriend — or his Manhattan cronies — ever got him to cook the books.
Don’t sue me Nate, this is satire.