Professor Todd Zywicki of George Mason University Law School has a post today
on VC discussing a study by Dan Klein
that found about a 9-1 ratio of Democrats to Republicans among the faculty at UC Berkeley and Stanford (combined). In this post I address (1) why the study (at least as presented by Zywicki) is complete bull shit and (2) how Zywicki manipulates the flawed study even more. In part 2, I will address the issue of intellectual diversity at UC Berkeley and Boalt.WHAT THE NUMBERS SAY
1. This is a survey of voter registration records of faculty members of each school.
2. Therefore they only take into account PARTY registration based on those records. Did I say PARTY? I mention this because blogfather Eugene Volokh heartily endorsed
the concept of Democrats voting across party lines in November.
3. This, in turn, means that the beginning of Zywicki's post and the article ("The popular vote for President went 48 percent Democrat and 51 percent Republican. This nearly one-to-one national diversity is unlike colleges and universities, where a one-party system prevails") are examples of a nice bait and switch. Do you see the sleight of hand? Implicitly Klein is suggesting the breakdown in the country is 48-51 for Democrats to Republicans, NOT, that the breakdown AMONG VOTERS is 48 for KERRY and 51 for BUSH. Now I'm not arguing this is a huge error that explains away all the results of the survey, but it's at the very least intellectual inconsistency and at the most downright dishonesty.
4. Ignoring confound variable #1 -- base rates: Now as I said in 3, the base rate used is the result of the 2004 Presidential election rather than the breakdown by party registration nationally. But of course the measure is of 2 schools, located in Northern California, specifically in the Bay Area. You might wonder what national party breakdown numbers woul have to do with two schools in this area. I wonder the same thing. I'm also wondering why local numbers weren't used. So let's take a look shall we? Hmm, vote for President in 2004
in Alameda County 75% for Kerry, 23% for Bush. And if we now turn to Santa Clara (I think Stanford falls under Santa Clara, not sure because the County line is very close) but anywho, that county had 64% for Kerry and 35% for Bush. So roughly in those two counties combined the ratio is 2.5 to 1. Not exactly a 9:1, but then again not exactly a perfect 1:1 either. Could it be a factor? Well if it was an actual "scholarly" study, not just one claiming to be one, these things would be taken into account.
5. Ignoring confound variable #2 -- time: Important to Klein and Zywicki is the almost complete lack of Republicans among junior faculty. They believe that as full professors retire, assistant and associate professors coming up the tenure ladder are replacing the more conservative full professors. Umm, paraphrasing Churchill, "Anyone under 30 and not a liberal doesn't have a heart, and anyone over 30 and not a conservative doesn't have a brain." Naturally, I'm willing to bet a pint that ALL people get more conservative as they get older, up to and including assistant and asssociate professors. Does this mean that the possible effect that Klein cries wolf about does not exist at all? The answer of course is that I don't know because not all the variables were controlled for. WHAT ZYWICKI THINKS OF THIS (INCLUDING THE 3 UPDATES)
1. The prima facie case for bias: "There seems to be some ambiguity about what I wrote. First, I did not say that there was a bias here. I said that if we saw a ratio of 30 to 1 in a general population that we know to be roughly 1 to 1, this usually will create a prima facie case of bias."
Putting aside with the flaws of the 1 to 1 ratio as discussed above, there is the other problem of "general population." I'm a first year law student, and Zywicki is a tenured professor of law, so I'm not going to even remotely pretend that I know any fraction of the knowledge of the law that he does. HOWEVER, I'm also fairly confident that a comparison to the general population is not always the best comparison. Specifically he's using a legal standard, one used in discrimination cases. I have no background in this area so I don't know the relevant case law, nor do I have the time to look things up, BUT, if it's anything like proving a malpractice for example, then of course it's not so obvious that the comparison should be to the general population. I have a sneaking suspicion that the ratio of women to men (to borrow Zywicki's example) at say Cisco systems is not nearly close to the ratio in the general population. Does this create a PF evidence for discrimination thereby shifting the burden on Cisco? I don't know, but my hunch is that it wouldn't. But even if it did, a quick showing of the ratio in the relevant population (tech employees, Silicon Valley, whatever) would rebut that presumption. As a matter of fact Klein notes that his numbers are not that different than nation-wide studies. Well if that's the case, then we can shut the door can't we? The job just attracts certain people (cf. military discriminates against Democrats based on ratio compared to general population).
2. "Where is the blue-ribbon panel at Stanford on intellectual diversity? "
This, of course, is a legitimate concern. One that I actually share with Zywicki, which I'll turn to in Part 2 (probably tomorrow).
Labels: Law School, Only In Berkeley, Rabid Conservatives, The Red Menace