Amidst all of the hubbub over Nate Silver’s departure from the NY Times, Kevin Drum notices something disturbing (boldface mine):
At the same time, this tidbit from Politico about Silver’s decisionmaking process while he was weighing competing offers from the Times and ESPN worries me just a little:
Silver had told The Times that he wanted to expand to weather, economics and anyplace else at The Times that had statistics and numbers he could bring to life….Nate will appear on the air on ESPN and ABC, and will get “verticals,” or web hubs, devoted to a variety of new topics. He’s very interested in education, so there’s been a lot of conversation about that.
Silver has done a great job with sports and campaign polling. But these are fairly unique areas. In the case of sports, Silver has a lot of subject matter expertise to go along with his number crunching. In the case of campaign polling, you don’t really need that much. You can get by with a pretty pure data-driven approach to the whole thing.
But weather, economics, and education? I’m skeptical that you can just parachute into those fields and add a lot of value. They’re far more complex, are already heavily populated with sophisticated statistical modeling, and generally require some serious subject matter expertise in addition to raw number-crunching skill.
When Silver entered the political prediction field, it was relatively wide open. Most popular prognosticators did no statistical analysis whatsoever–recognizing that polls have confidence intervals was a mark of sophistication (such as it was). But political polling is relatively straightforward compared to educational data. ‘Political’ data are relatively simple and uniform, and there are many opportunities for internal validation (e.g., daily tracking polls). But educational data are very different. States administer distinct tests and score them differently (cutoffs, raw scores, adjusted scores, value-added, MGP, etc.). On top of that, the stupid fucking natural history facts matter. Charter schools, for instance, do a couple of things to fudge the numbers, including transferring out students who perform poorly (and are then counted as non-charter students, even if they were in the charter for most of the year) and having a much less challenged student body (fewer English-as-a-second-language and special education students). It’s just not that simple; you have to understand the limitations of the data. Unfortunately, given Silver’s fame, he could potentially do a lot of damage here–and we’re not dealing with sports or political horse races, but childhood development. You have to get it right.
I’m also confused as to what Silver thinks he’ll offer in the weather department. It’s not like there aren’t entire departments of meterology, not to mention multiple government agencies and professional private forecasting firms trying to figure this stuff out. Maybe he’s got some kind of revolutionary special sauce that can be run on a laptop, but, as I understand it, he’s going to need a lot of computing power. Like in the many millions of dollars variety. As Drum puts it, the field is a little crowded.
Anyway, it will be interesting.