Wednesday, May 28, 2008

Mind-Numbingly Bad Statistics

Since I fit in with neither side, I like to think that I try really hard to criticize the Left as much as the Right. But this week, this has been almost impossible.

A new poll shows that a majority of Californians now support gay marriage and oppose the proposed constitutional amendment banning it. As you may know, I have argued that judicial "activism," to the extent that it protects insular minorities against the tyranny of the majority, is actually a good thing. As such, I don't buy into the arguments made by conservatives these days that the decision validating gay marriage was "legislating from the bench" in a way that usurps the "will of the people." But nonetheless, having a poll that shows that the "will of the people" in fact supports gay marriage pretty much destroys any argument that the "will of the people" has in fact been usurped.

And so, of course, the only response for conservatives who oppose gay marriage and claim that their position is the "will of the people" must either back down from their arguments, or try to discredit the poll. And, given the poll from a week or so ago showing substantially different results from this one, it is entirely possible that this poll is flawed. BUT, if you're going to discredit a poll, it's probably a good idea to be intellectually honest about it, rather than doing what JammieWearingFool does here:

Of course, if you refer to the poll, you see under ideology a full 70 percent of those sampled consider themselves middle of the road (53-35%), moderately liberal (72-23) or strongly liberal (85-11))and they overwhelming favor gay marriage. If you look at the 30% considered moderately conservative (61-32) or strongly conservative (85-11). When broken down by party, the sample is 43% Democrat, 33% Republican, 24% non-partisan. In other words, Republicans are outnumbered 67-33%. Los Angeles County and the San Francisco Bay Area comprise 46% of those polled.In other words, this poll doesn't even come close to a fair representation of the
population. Granted, California is a blue state, but you cannot possibly claim a poll where left outnumbers right 70-30% as valid.

The trouble of course with this is that JammieWearingFool makes several totally absurd assumptions here: 1. That people who consider themselves "middle of the road" are part of the "left"; 2. That people who consider themselves "non-partisan" are part of the "left"; and 3. That even under his unbelievably broad definition of "left," only slightly more than half the voters should be part of the "left."

So when JammieWearingFool says that "left outnumbers right 70-30%" in the poll, he is being more than a little obtuse. In reality, the poll numbers reflect that left and center outnumber right 70-30%, but that self-identified liberals are outnumbered by self-identified conservatives, 30-28%. If California were a pure "purple" state, then you would essentially expect left, right, and center to be about 33% each. Since, as JammieWearingFool acknowledges, California is a "blue" state, a decrease of only 3% from this norm is not exactly earth-shattering. Moreover, this poll from SurveyUSA (and I'm sure there are plenty others from other firms that make similar findings) shows that only about 26% of Californians self-identify as "conservative," with 42% self-identifying as "moderate" and the remaining 22% as "liberal." So, if anything, the poll slightly oversampled conservatives and liberals by about an equal amount.

But the flaws in JammieWearingFool's analysis go further. He decries that only 33% of the respondents to the poll self-identified as Republicans, leaving 2/3 of the people as "not Republicans." In the process, he seems to ignore that 24% of the respondents self-identified as independents, and that Dems only outnumber Republicans 43%-33%. And looking at the numbers from the 2004 elections, we find that in 2004, 43% of voters identified themselves as Democrats, 34.7% as Republicans, and 23% as independents. So, at worst, the poll undersampled Republicans by a whopping 1.7%; of course, given the hard times the GOP has gone through since 2004, it's more likely that self-identified Republicans have decreased by about 2% in California.

Finally, JammieWearingFool criticizes the fact that 46% of poll respondents were from Los Angeles County and the Bay Area. Well, according to the 2006 US Census data, just over a quarter of California residents lived in LA County; and roughly another 20% lived in the Bay Area....which adds up to, you guessed it, 46%!

I look forward to the retraction.

More at memeorandum.

**UPDATE**Lest you think that I'm just picking on one lonely lesser-known Righty blogger, I note that at least one other Righty blogger has picked up the meme from JammieWearingFool.