One topic that comes up often in conversations with my ultra-conservative father is homosexuals and what rights are they owed. A sticking point we’ve come to is what portion is actually gay? He maintains that it is at most 2%, while I’ve heard figures as high as 10%. I’ve searched in vain for cites, and have come up with nothing. So I thought I’d ask here. Can anyone help me?
As with most statistical claims, there is a lot of play in the joints. Here is a presentation “sponsored” by accuracy in media that pegs the number at 1.1%.
Another one criticizes the 10% figure.
Here is an article That discusses some of the issues involved in attempting to estimate the gay population.
And here is one reporting similar results and difficulties in Scotland.
Here are two links to articles discussing the history of counting gays, and claiming that they are undercounted:
[URL=http://content.gay.com/channels/finance/glmoney/60years_000717.html]http://content.gay.com/channels/finance/glmoney/60years_000717.html http://www.gaydemographics.org/USA/2000_Census_Total.htm
In short, to some extent, it depends who you ask, and how you count.
I hope this was helpful.
Indeed it is. Thank you.
Impossible to say for a couple of reasons.
First people lie to interviewers. I have heard it said there are no reliable statistics for how often people brush their teeth for the same reason.
Second, it is impossible to describe “Homosexual” in a way we all agree upon. One homosexual experience? Count me in. Look for a high percentage.
An exclusively gay life? No heterosexual contact ever? Pretty rare, I bet.
As a buddy once said defending the 10% figure. “It must be true, everyone at the party last weekend was gay.”
Haha. Or, I’m not gay, but my boyfriend is.
Not to get all GD on anyone’s ass but my response to the question “how many homosexuals are there” when asked in the context of protecting their civil rights is generally “how many do there have to be before protection is deserved?”
You absolutely can measure things where there is incentive to lie to interviewer. Pretty neat trick.
You sit down with subject. Give subject a coin and a mini-cubicle. Tell them to flip the coin (privately, where interviewer cannot see the result).
If it’s heads, answer the question honestly, do you in fact cheat on your taxes?
If it’s tails, then flip it again. If heads, answer yes, if tails, answer no.
Note that no matter which answer the subject gives, the interviewers has no way of knowing whether this particular subject cheated on their taxes (or brushed their teeth, or did illegal drugs, etc.). However, when the results are aggregated together, it is simple statistics to derive the true answer.
Judging from the response I get from the females at the clubs, I’d say 100% of them are…
To expand on a certain point: “homosexuality” is a very bland and vague definition. For example, in many places in the world, people still enjoy public baths. There is no implication of homosexuality in going to baths with same-sex friends, or perhaps even scrubbing each other (well, maybe not getting too close, but the point remains). So, basically, I think the question would have to be: have you had sex (oral, anal, or “whatever-it-is-they-call-lesbian-sex” with someone or your own sex? I mean, in some cultures, its hardly unsuall for men to kiss, hold hands, or hug.
But then a gay person who has not had such sex could truthfully answer “no” and would therefore not be counted as “homosexual.” Or on the flip side, what if you ask a male virgin if he’s had penetrative sex with a woman and he truthfully answers “no,” how do you fit him in your survey? And then of course there are those pesky people who, while self-identifying as “homosexual,” have had sex with persons not of their own sex. And then there’s our very own Paul who fessed up in this very thread to being both straight and a participant in some form of “homosexual experience.” Not having clicked on any of the links in this thread, that idea of defining orientation based on behaviour has been attacked pretty much from the time it was formulated.
Don’t know if this helps, but self-identified gay, lesbian or bisexual voters accounted for about 4% of the participants in the California recall election, according to the NGLTF.
When I studied survey design in grad school, this technique was relatively new. While it is certainly valid mathematically, at that time it was open to question that the methodology could be explained to the average person-on-the-street sufficiently well that they could trust its anonymity and therefore be trusted to give truthful responses. I am curious to see if studies since that time has validated the technique in real-world application.
When I was exposed to the technique in 1989, I was taught it had been used in the real world to measure tax cheating, and illegal drug usage. I have no personal experience of it’s usage. On the other hand, I see no reason to doubt it. Perhaps the average person may not understand/believe from the short description I gave, but it certainly seems doable to me. I am also curious to see if any Dopers have first-hand experience with this.
i never studied statistics. could you explain–in really simple terms–how you extract your results with this technique? i understand that with enough coin flips half will come up heads and half tails, but how does that translate into accurate conclusions about the interviewee’s claims? suppose the question is one to which only a very small percentage of respondents would honestly answer yes (say, “do you brush your teeth while hopping on one foot?”) Almost all of your “yes” responses will be the result of the coin toss. how do you sort it all out?
Actually, that’s not quite true. While theoretically if you flipped a coin a squillion times, you’d get roughly half a squillion heads and half a squillion tails, in practice you tend to find that the coin will go on rather striking runs one way or the other. It’s actually highly unlikely that whatever point you stop flipping at will be a point at which the results happen to be close to even.
As far as the OP: I do have it on good authority (from a survey commissioned by the Board of Regents a couple years back) that 8.6% of 4,000 students polled at my university self-identify as gay or lesbian (not bisexual).
It’s a very large (8th largest student body in the nation) commuter campus in Orlando, Florida. I imagine if you corrected first for the body of students who are gay but didn’t want to say so, then corrected for the fact that this is a university population and thus younger (and used to a society which is relatively tolerant of homosexuality) than average, you’d probably end up somewhere between 6 and 9%. Of course, IANA statistician, nor am I gay.
If you get tails on the first coin flip, your second flip determines your answer. You will end up answering “yes” or “no” regardless of whether you really cheat on your taxes. If you get heads on the first flip you can answer the question honestly because the testers have taken away the “need” to lie. If you answer “yes” you can explain it away by saying that you were forced to by the coin. Nobody knows whether you really did cheat on your taxes or whether the coin made you say “yes”. A bit of math lets the statisticians remove the answers that were forced by the coin (it is fairly predictable, as you said).
Are you asking for the actual math? Let’s see if I can rederive it off my head, it’s been a long time since I did this.
Num(Yeses)=p(head coming up yes and person does cheat on taxes) +
Oops…
p(yes response)=p(heads coming up on flip and person does cheat on taxes) + p(tails on first flip and tails on second flip)
p(no response=p(heads on first flip and person does not cheat on taxes) + p(tails on first flip and heads on second flip).
The second term in both those equations is obvious 1/4. So we get to:
p(yes)=p(heads coming up and person cheats) + 1/4
p(no)=p(heads coming up and they don’t) + 1/4
Next take out the 1/2 probability of the heads.
p(yes)=1/2(p(person cheats))+1/4
p(no)=1/2(p(person doesn’t cheat))+1/4
So let’s say we ask 100 people, and get 65 yeses.
.65 = .5(probability of cheating) + .25
.4=.5(probability of cheating)
.8=probability of cheating
we estimate 80% of responders cheat on their taxes.
Another way of looking at that is, 25 of those yeses were just coin flips. We had 40 yeses from the 50 people who flipped heads the first time.
Another example, we ask 200 people and get 60 yeses
60/200 = .3
.3 = .5(probabily of cheating) + .25
.05 = .5(probability of cheating)
.1 = probability of cheating
We estimate 10% of responders cheat on their taxes.
Hope that makes it clear. Of course their is an additional step of calculating the p-value to see how confident you are in the answer, but that step is standard statistics and is identical to situation where responders answer openly and freely. Note that we were able to get good answers even though we have no idea whether any particular individual cheated on their taxes.