Pages

21 October 2012

How not to count trans people


This week Gallop reported that using the following method:

Results are based on telephone interviews conducted as part of the Gallup Daily tracking survey June 1-Sept. 30, 2012, with a random sample of 121,290 adults, aged 18 and older, living in all 50 U.S. states and the District of Columbia, selected using random-digit-dial sampling
it had put the following question, and received responses as indicated:

The report acknowledges: 
As a group still subject to social stigma, many of those who identify as lesbian, gay, bisexual, or transgender may not be forthcoming about this identity when asked about it in a survey. Therefore, it's likely that some Americans in what is commonly referred to as "the closet" would not be included in the estimates derived from the Gallup interviews. Thus, the 3.4% estimate can best be represented as adult Americans who publicly identify themselves as part of the LGBT community when asked in a survey context.
And furthermore:
Transgender status can be an identity but can also include consideration of behaviors regarding gender nonconformity and an individual's internal sense of gender.
They also asked the respondents to choose between a mishmash of race and language, and reported: *


Let me consider how I would have answered the question if some stranger rang my phone and put it to me.  Well I probably wouldn't, as I usually have no time for intrusive phone calls.  But suppose that I was in a particularly good mood.  I am a married woman, so do I go in the 'no' column?   On the other hand, I am an alumna from Gay Lib, so does that put me in the 'yes' column?  Perhaps I was gay and transgender until I transitioned, but not after?  How does Gallop record that?

The pioneer sexologists, Kinsey, Benjamin, Hirschfeld etc, asked what people did.  Not what they identified as.  Secondly they broke questions down.  Are you gay; are you transgender?  You can be either without the other, neither or both.  This is not allowed for by Gallop. 

How did Gallop expect a heterosexually-married transvestite to answer?  What about a heterosexually married man who goes to the gay baths now and then?  What about a heterosexually married man who pays to be bottomed by a trans prostitute?  The last two types are sort of bi, but rarely speak up at bi-pride events, and almost certainly would opt for the 'no' column.

Then there is that very small word: 'or'.  Easily missed when spoken. How many gays and lesbians said 'no' because they are not transgender?  How many trans said 'no' because they do not identity as gay or lesbian?

I would love to know how many men said that they are lesbian, and how many women that are gay men.

Note how in the race-language mishmash the 'yes' column and the 'DK' column almost add to 10% (Kinsey's estimate).


It is interesting that the "non-Hispanic whites"  gave fewer 'yes' answers.  This is counter-intuitive based on what one sees at gay events.  However given the bad design of the question, the result  doesn't mean much.

Given the size of the sample one would expect over 100 of the respondents to be, or to have been some flavour of trans.  However because of the bad design of the question, Gallop are not able to confirm this.

I suspect that the question was proposed by a Gallop manager, and that his underlings don't feel comfortable enough to point out that the question was a dumb question.



* This type of race-language mishmash is commonly used in the US.  However one must ask a) where are the native Americans?** b) what do Chinese, Indian and Arab have in common such that they are one group?  c) are Spanish-speaking Blacks, Black or Hispanic?

** Native Americans are particularly relevant for the question asked because the Two-Spirit tradition.

3 comments:

  1. This reminds me of a survey about exhibitions in a festival, I wrote the first year to point out the absurdity of the question, it was repeated the very year!

    Was the exhibition as good as you expected, rate from worse than to better than. I saw two exhibitions, one which I thought would be wonderful and was and one which I expected to be terrible but needed to confirm for myself and it was...

    ReplyDelete
  2. I saw this study in a blurb on CNN the other day. I thought the figures looked a little on the low side. On a more positive side of things, I took a survey (not from Gallup) a couple of weeks ago that asked a lot of detailed, well thought out questions about gender and sexuality. It seemed like they were trying to assemble a well-rounded look at the LGBT community.

    ReplyDelete
  3. Closetta_Tv writes;

    (So sorry - wanted to reposnd to your blog but haven't got a blog-reply profile)

    Regarding the GLT surver by Gallop, I totally agree your comments.

    However let us be clear all surveys - whether GLT or otherwise - are equally badly thought through. And always very, very dangerous.

    Further to your always meticulous logic, you might have added that the very phrasing of the question, Leading or otherwise. And especially its juxtaposition in relationship to the other questions in the survey. And of course, a GLT question embedded in a life-style, or sex survey will get a very different response to that embedded in a religious survey. (Hence confusion about the number of GLT Christians, Moslems etc). And who - except a liberal - is likely to participate in any kind of sex survey with a stranger with a clip-board? (Possibly of the opposite sex - gulp!) Survey respondents are frequently shown to be both self-chosen, self-serving, and polite enough to try to give the expected answer ("Do you like these new shit-flavored potato chips?". Oh yes, sweet blond questioner, I love them - and everything else about you. You see, I'm a conformist and would never disagree with someone older/youger/pretiier/kind enough to ask my opinion". )

    There's generally also a very dangerous bias when the sample is chosen. Gallup obviously didn't ask all 250m Americans, and took this "mathematically significant" sample, eg 122,000 (That's about 1in 2,500). For a precise and unambiguous question (eg "Are you alive?") its mostly OK to scale by 2,500. But it always builds in a bias. If you ask this "Are you Alive' question, exclusively of a group of non-english speaking americans, or tiny children, or deaf people, you'd get a DK from them at least, and a huge error.

    Similarly if this GLT question is asked of those in down-town Burbank, you get a very different response than when asked in Salt-Lake City or Vermillion. Gallup almost never gets to those last two places, they like to sample where its convenient. (Even if they did sample in Hicksville USA, would the surveyor do a good job, or be lied to?)

    Lastly, suppose 10% of those responding, lie, or misunderstand the question, or their answer is just recorded wrongly. That means Gallup records 5% GLT when it should have been 4.5% - not a big problem. But it also means that 10% of the 95% who aren't GLT are misreported also. Adding this possible error ie 10% of 95%. swampes the actual GLT rate recorded. ie there are 10%x122,000 wrong answers ie 12,000, Against only 5,000 whigh might be GLT.

    And finally (finally), as my old professor used to say, (and here the nub) "who's asking, and what are they going to do with the information? "

    Keep up the good work!

    Closetta

    ReplyDelete

Comments that constitute non-relevant advertisements will be declined, as will those attempting to be rude. Comments from 'unknown' and anonymous will also be declined. Repeat: Comments from "unknown" will be declined, as will anonymous comments. If you don't have a Google id, I suggest that you type in a name or a pseudonym.