Posted by Jay Livingston
Max Weber wrote an essay about “Scholarship as a Vocation” (the more commonly used translation is “Science as a Vocation”). It’s a classic, but Weber, in focusing on the professionals, forgot about the hobbyists tinkering in their garages. Sometimes, they do it better.
When I was an undergraduate, there was a lab technician who worked for the biochemistry department – I don’t think he had a college degree, he may even have been a high school droupout – who knew more than most of the doctoral students and probably some of the faculty. He was also the quarterback on the biochem intramural football team, which is why they usually won.
I was reminded of this by two things this week: my sister-in-law’s birthday dinner and Andrew Gelman’s Social Science Statistic Blog.
Gelman prints a rant (his word, not mine) that someone sent him about a wrongheaded statistical analysis done by some consultant for a local government project.
Gelman agrees and adds:
This certainly doesn't surprise me: I've seen worse from paid statistical consultants on court cases, including one from a consultant . . . who reportedly was paid hundreds of thousands of dollars for his services.The people with credentials can be wrong. And sometimes the uncredentialed guys – like the lab technician – surprise you.
The key problems seem to be:
1. Statistics is hard, and not many people know how to do it.
2. The people who need statistical analysis don't always know where to look.
Take Howie, a guy you would never mistake as an academic. He’s an old friend of my brother- and sister-in-law, and Sunday we were celebrating her birthday somewhere out in Queens at an unpretentious Italian restaurant (decent food, reasonable prices, no tablecloths, Yankees on a couple of TVs in the bar). I was seated down at the end of the table with my brother-in-law and Howie. Talk turned to politics – Hillary, Obama, Rudy – and what about Gore? That got us to Gore 2000 and the electoral college, then to Kerry and the Ohio vote in 2004. Were the elections stolen?
It became clear that Howie knew a lot about voting and irregularities and how you might audit results to detect vote-count tampering. He knew about sample sizes and statistical errors. He knew that no single sample size or percentage was perfect and that you optimize sample size by taking account of of electoral shifts and winning margins. He knew all the ways that a House bill on election audits was flawed. He also knew about hacking electronic voting machines, but he thought the computer scientists were focused on the wrong part of the problem. If you wanted to ensure voting integrity, you had to go for statistical audits. (“The computer science guys, they’re not interested in this stuff.”)
But as far as I knew, this was purely a hobby for him. It certainly wasn’t his job. (The last job he had was for an airline.) He had learned the statistics on his own – books, the Internet – after he’d gotten interested in the problem of election integrity. Now he’s publishing papers with academic co-authors and offering expert testimony on proposed federal legislation.
3 comments:
This is an intriguing post, since I'm both obsessed with elections and learning stats (I tend to focus on other aspects, though). How would you do a statistical audit of an election? The only thing I can envision is looking for large blocks of identical ballots (a la the Freakonomics test for teacher test tampering), but that seems like it would be hard to prove, since people vote on the party line so much. What was Howie's plan?
To verify the electronic vote, you need to have an additional means of recording votes -- a "paper trail." The question is how many of those paper ballots do you need sample in order to make sure that the electronic total wasn't hacked. Howie's point is that there is no fixed number or fixed percentage of ballots for each precinct.
Ahhh... got it.
Sheesh. I always miss the obvious stuff.
Post a Comment