Look closer. Think harder. Choose the sound argument over the clever one.

Thursday, March 15, 2007

The "Political Profiling" Study

This political profiling study was quoted by NYT's Paul Krugman:

Data* indicate that the offices of the U.S. Attorneys across the nation investigate seven (7) times as many Democratic officials as they investigate Republican officials, a number that exceeds even the racial profiling of African Americans in traffic stops. ...The current Bush Republican Administration appears to be the first to have engaged in political profiling.

To which Tom Maguire replies:

Well, if you don't even look at data from earlier Administrations you aren't likely to find anything, now are you?

Good point. The study's authors didn't study other administrations, yet claim "The current Bush Republican Administration appears to be the first to have engaged in political profiling."

Pat at Stubborn Facts calls it fatally flawed:

Just a few Google searches found several instances of Republican elected officials who were investigated by the Department of Justice but which were not included in the authors' data set. ...

These omissions are a result of one or more types of statistical bias. The theoretical biases were obvious simply from reading the paper. The concrete examples prove that the study is worthless. Just that quick sample increased the number of Republican officials and candidates investigated by the Bush Administration by about 10%. ...

When an hour of Google searching can increase a crucial data point in a 6-year-long study by 10%, the authors should be deeply embarrassed.

Ouch. Read the whole thing if you're interested.

There's an interesting comment or two, too:

[J]ust from a glance at their data and only their data, the "bias" they claim exists is simply not present in federal investigations of federal and state-wide elected officials, where the "count" is what you would expect from random selection by party numbers. So the "overweighting" of Dem officials "investigated" is limited ENTIRELY to the data on municipal/county officials. No such bias is shown in the federal/statewide data offered by the authors. NONE.

I'd also be interested to look closer at what kicks off an investigation.

Hat Tip: Maguire.

3/31: Still being used.


Comments:

(Please keep in mind that each commenter's opinions are only his/her own.)


Post a Comment



<< Home

This page is powered by Blogger. Isn't yours?