Not to pick on Jonathan Capehart because this number has been going around and he's not the only pundit using this exit poll question to argue that LGB's turned Republican in this past election, but here he is:
Gay men, lesbians and bisexuals who self-identified to exit pollsters made up 3 percent of those casting ballots in House races on Tuesday, and 31 percent of them voted Republican. By itself, that number is amazing, especially when you consider that way too many people think being gay and voting Democratic are one in the same. But that percentage is ominous news for a White House viewed with suspicion by many gay men and lesbians, because that's four percentage points higher than the change election of 2008.
Oh my, 4% increase! In fact, it's even worse than that. For some reason Capehart won't share with us he compared 2010 House exit poll results to 2008 Presidential exit poll results, when 2008 House results were available. 80% of LGB people voted for the Democratic House candidate in 2008, making the difference 11%. If you gathered nine LGB people, one of them turned Republican in the last two years!
Except, no, there's no way to know what happened since the poll is (and has always been) pretty much meaningless. Capehart even cites one of the big reasons why at the bottom of his post:
But Post polling director Jon Cohen urged caution when interpreting the exit polls. "[Be] *careful* of extremely low sample sizes," he wrote to me in an e-mail. "Given the way the exit poll was divided into smaller parts, only 126 voters interviewed nationwide described themselves as gay. Sampling error margin is about + or - at least nine points for this group of voters." That should put a damper on LaSalvia's end-zone dancing. But the overall trend and the message it sends should not be ignored by Democrats.
Capehart was warned to be careful with numbers from an LGB poll, but that doesn't stop him from going crazy with the exact analyses he was told to avoid, even saying "More gays voted Republican than in 2008" in the title of his post (and I'm just going to ignore Capehart's conflation of "lesbian, gay, or bisexual" and "gay"). There seems to be something going around the Washington Post's offices, as another Post reporter did the exact same thing a month ago.
Remember, it's the bloggers who are irresponsible and make over-the-top conclusions without being sure of their facts. Without staid, reasonable mainstream media thinkers like Capehart and his colleagues, the world would have already descended into madness.
Anyway, that's a fairly big deal there, that only 126 people were asked a polling question and we're supposed to take it as representative of the whole LGB population that's probably in the tens of millions. A plus or minus 9% margin of error isn't a trifle, and it's not even the entire margin of error. Exit polls actually have a higher intrinsic margin of error than a telephone poll, and 9% is the margin of error a telephone poll that size would have had. One polling firm estimates exit polls have 50%-80% more error than standard polling techniques, meaning the margin of error on that 31% number could have been plus or minus 16%.
Add to that the normal problems with exit polling: not counting early, absentee, or mail-in voters; not counting late voters as the pollsters usually leave an hour before the poll close; lots of people choose not to take the time to respond; non-confrontational people who avoid electioneers, among whom pollsters are forced to stand in some states; and people volunteering to be polled, who should technically be denied by the pollster, but.... All of this is why exit polls are usually adjusted after the official results come in to give a more accurate picture of what happened.
Then there are the specifically LGB problems with the poll. The CNN exit poll asked 126 LGB people and 17,504 people, total. The CNN poll claims 3% of the respondents were LGB, even though 126 is much less than 3% of 17,504. So the polling analysts somehow assumed the size of the LGB population at 3% and then used the results from those 126 people.
Which raises a whole bunch of questions. Where did CNN get that 3% number? Is 3% accurate? Why didn't more than 126 people identify as LGB?
Most importantly: were those 126 LGB people representative of the larger LGB population in the US?
What we do know about those 126 LGB people is that they were willing to complete an exit poll with a pollster in public and felt completely free to tell her that they identify as lesbian, gay, or bisexual. And that's not everybody. It's not just Larry Craig types who would avoid identifying as LGB, but also people worried that others will hear, people who don't want to tell the pollster in front of them that they're LGB, or people who are in too much of a hurry or too anti-social to actually do the poll. CNN also doesn't show an "other" category, which, according to a recent Indiana University study, makes up 1% of the population. Did they assume that everyone who was "other" or refused to answer was LGB or straight?
And that's not even getting into the question I posted on earlier this week: what makes someone lesbian, gay, or bisexual? It's the fundamental problem with all studies of LGB people generally since if we can't even define a population in the abstract, how are we going to define it for the purposes of dissecting it?
If there's something more to gay-ness and bisexuality than the way someone identifies publicly, then exit polling won't get at it. Someone who is only attracted to the opposite sex might identify as "gay" or "bisexual" as a statement of solidarity or as a result of (mis)reading too much Foucault (I knew four people like that in college), some people might not be out yet and say "straight," some people might say "other" because they just had a strange experience but will eventually go back to identifying as what they identified before, some people might say "straight" because they don't think that their same-sex activities/thoughts count as gay, etc.
That same Indiana University study above found that 8% of men and 7% of women didn't identify as straight, so I'm really wondering where this study got its 3% number. Whether people who didn't out themselves to the pollster were more likely to vote Democratic or Republican is open to speculation (Democratic voters, who tend to be poorer, would have more to lose materially; Republican voters, who tend to be more religious, would have more to lose spiritually).
None of what I mentioned above gets into the problems with comparing these two exit polls from 2008 and 2010 and using them to say something about how LGB people have changed. Democratic turn-out was down from what it was in 2008, down more than Republican turn-out was, so maybe that explains why there were fewer LGB people who voted Democratic. Two years have passed, and more conservatives might be out than were back in 2008. The 126 people asked in 2010 and the similar number asked in 2008 surely weren't the exact same people, so the margins of error might have added up to produce a large change over time.
Anyway, all that's to say that there isn't any reason to put much stock in that exit poll. More importantly, many of the problems with that exit poll are found in other polls of the LGB population, and the first question we should ask when an article says "X% of LGB people believe Y" is "Where did they get their queers?" Usually the source won't stand up to scrutiny.
But that won't stop queer or mainstream media from reporting the figure and pontificating over it. Who cares if it's true; it's provocative, and that's more important.