“Americans Greatly Overestimate Percent Gay, Lesbian in U.S.”

This sort of thing is not new but it’s still amusing. From a Gallup report by Frank Newport:

The American public estimates on average that 23% of Americans are gay or lesbian, little changed from Americans’ 25% estimate in 2011, and only slightly higher than separate 2002 estimates of the gay and lesbian population. These estimates are many times higher than the 3.8% of the adult population who identified themselves as lesbian, gay, bisexual or transgender in Gallup Daily tracking in the first four months of this year.

Trend: Just your best guess, what percent of Americans today would you say are gay or lesbian?

Newport provides some context:

Part of the explanation for the inaccurate estimates of the gay and lesbian population rests with Americans’ general unfamiliarity with numbers and demography. Previous research has shown that Americans estimate that a third of the U.S. population is black, and believe almost three in 10 are Hispanic, more than twice what the actual percentages were as measured by the census at the time of the research. Americans with the highest levels of education make the lowest estimates of the gay and lesbian population, underscoring the assumption that part of the reason for the overestimate is a lack of exposure to demographic data.

But there’s a lot of innumeracy even among educated Americans:

Still, the average estimate among those with postgraduate education is 15%, four times the actual rate.

Newport summarizes:

The estimates of gay and lesbian percentages have been relatively stable compared with those measured in 2011 and 2002, even though attitudes about gays and lesbians have changed dramatically over that time.

Math class is tough.

35 thoughts on ““Americans Greatly Overestimate Percent Gay, Lesbian in U.S.”

    • And a third of respondents said OVER 25%. Without even considering bisexuals or other queer identities!

      I wonder what the gay and lesbian respondents said. I know the “1 in 10” estimate was drilled into our heads pretty hard back in the 90s, and I also feel pretty confident that nobody who was ever a gay or lesbian high school student would believe that the right answer was 1 in 4 or higher…

        • I think Andrew should have said, “most people”. And I assume that since you are reading this blog, you are not in the “most” category for the purpose of the question at hand.

        • I guess I read Andrew as saying people don’t have a sense for small but not tiny percentages like 15%, and that didn’t square with my experience of people having pretty clear ideas about what fractions mean in everyday life. Maybe its easier for people to think about fractions than those same fractions expressed as percentages.

  1. This really knocks the legs out from under the studies on perception of income mobility discussed a few days ago.

    These surveys are asking for estimates of fairly well defined variables for which we have pretty good estimates, and respondents have wildly biased estimates. The notion that the opinions of respondents to an online survey of income mobility (who were paid $0.50 for their participation!) are worthy of serious consideration is, therefore, pretty funny.

    But of course, the income-mobility-type papers aren’t interested in such things. Their purpose is to produce a soundbite that the New York Times can write up as “a troubling indictment of inequality in America”.

    The income-mobility-type papers are much easier than p-hacking noise. When you p-hack noise, you have to explore enough forking paths to find that magical 0.05. That can be a lot of work. Income-mobility-type research is easier because you are working with strongly biased data. It is, therefore, much easier to get a 0.05 as long as you ignore the bias. (Don’t even think about checking whether your subjects understand what they are doing – just assume it.)

    The best of all research projects is when you get publishable results regardless of which way the bias goes. Ideally, you would like to be able to interpret significance in either direction as “troubling”. Almost as good, though, is if you can interpret one result as “troubling” and the opposite result as “surprising”. It’s a win-win situation.

    • Pretty sure the point of the income mobility paper was to measure how wrong the public was about income mobility (and thus how much our political conversation might be suffering from misinformation). It wasn’t actually meant to be an independent estimate of income mobility.

      • Thanks for the correction.

        I had in mind the income-inequality papers that got written up in the NYT and conflated them with the immobility papers. The mobility papers are much less tendentious.

        • Hey Terry, could you provide a link to these income inequality papers (or the NYT link)? My lab is investigating these kinds of perceptions, and income inequality is one we’re interested in…

    • The way to measure inequality is to get the total amount of income (including interest and capital income and non-monetary benefits like health income) for a wide range of people and track it over time. No one would dispute the numbers if this is what they did. Most of the studies trumpeting the large increases in inequality don’t correct for the change in the sample. For instance, we might measure wage income, rather than even total labor income. Alternately, we might solely be looking at aggregate statistics, such as those from the IRS. As a result, it doesn’t take into account changing demographics and household formation. An increase in the percent of people who are divorced or immigrants might bias the results. Further, younger people make less money than older people, and now that we have more older people in the economy, it might look like we have more inequality.

      That’s not to say that these statistics have no value. I believe they do. The issue is that they don’t imply the conclusions that we suggest. For instance, if we look at wage income and find that the top 5% make X times the median whereas 20 years ago they made X*k times the median, where 0<k<1, then calculating that statistics and disseminating it has value. However, it does not imply that the people who are in the top 5% are now richer. They may not be the same group of people. Rather, it just means that to be in the top 5% requires X amount of money now and previously required Xk. It's really a matter of statistical interpretation and using statistics to make an argument.

  2. I dunno Andrew, perhaps this is simply a result of the general belief that percentages of LBGTQ are under-reported on surveys. My prior is that the ‘true’ rate is certainly higher than 3.8%. It would possibly be ignorance/innumeracy if the question asked “What percentage do you think *self-report* as LGBTQ?”. I don’t think you can infer innumeracy from the measures above.

    Careful measurement is important, as we say :)

    • I agree, 3.8% seems quite low. In a lot of states, I guess its still a big deal, especially among religious people. I would give 10% as an upper bound estimate and 3.8% as lower bound.

    • There is actually evidence of this from Coffman, Coffman, and Marzilli Ericson, which suggests that self-reported rates are probably underestimating the level by about 65%. Interestingly, negative attitudes about LGBTQ people are also found to be under-reported.

    • Yes on the measurement, particularly if you think about whether they might be answering what percent has had at least one homosexual experience. Also I think the Kinsey percentages still frame some perceptions.

  3. Huh, every year the mode is at “more than 25%”. Well, since it’s the widest group, density wise it’s not the densest, but still I find it rather surprising that 24 to 35 percent of respondents would sincerely think that at least one fourth of the population is homosexual. I wonder what is the highest guess. In any case that’s pretty amazing.

    Also, people consistently don’t seem to like the 15-20% option. People are strange!

    Also number 2: I find it kind of cute how there’s a constant confusion between LGBT and LG. The small fact brochure is titled “Perceptions_of_Percentage_LGBT_150521%20.pdf”, notice how it mentions perceptions about LGBT, but as we can see from the questions, they asked people only to guess the proportion of LG. The 3.8% estimate that’s throw around is, again, about LGBT, and this is compared against the guesstimates of the proportion of L’s and G’s. To be fair since L and G are a subgroup of that, their proportion is somewhat smaller, so that doesn’t change the conclusions that people seem to overestimate these things, but anyway. Not a big deal, I’m not accusing anyone of anything, just kind of pricked me in the eye in a fun way and amused be slightly.

  4. I think this is Andrew’s subtle way of reminding us of the perils of unrecognized sampling bias. Maybe the “random sample of 1,024 adults” was drawn from an urn that itself is not representative of America as a whole. Maybe the people called on “landlines and cellular phones” via “random digit dial methods” who are willing to opine to strangers about fraught subjects aren’t representative of all Americans with landlines or cell phones. As to how those sampled came to their opinion I think Andrew is correct. If you dumped a bucket of sand on the floor and asked 1,024 to make two piles of it – one containing 85% and the other 15% – I’d bet there’d be half a teaspoon’s worth (on average) in the little pile and the rest in the big pile.

  5. If you ask Americans what is the most common ancestry nationality that Americans report, few would answer “German” either in a fixed list of choices or as an open-ended answer. But according to the Census Bureau American Community Survey about 1/7 of Americans identify as having German ancestry. However most German immigrants arrived before 1900. Also there was a conscious effort starting with World War I to suppress German culture in the U.S.

    My point is that except for the just passed Octoberfest menus and retro biergartens, it is not a group that gets a lot of media attention. And the people in this 1/7 tend to be in the center of the country, away from the media glare.

  6. I think the issue is that for heterosexuals any observation of homosexual behaviour labels the person as homosexual even if the person themselves doesn’t identify in that way e.g. because they are married and have a family, because the homosexual relationship was experimental or a long time in the past (prison/armed services/boarding school) or the non-observed relationships are heterosexual.

    It’s like sexuality is on a continum and we’re trying to dichotomise it at different points which depends on where we personally stand on the continum – to be like me you have to be exactly like me.

  7. As others have stated, 3.8% seems very low to me. I agree that there’s a lot of bias in self-reported estimates like this.

    I have some data to back this up. My department recently did a climate survey of graduate students, with 100 responses. Of those that answered questions relating to LGBT identities (89), 22 self-identified with LGBQ identities. 4 identified as trans/non-binary/genderqueer.

    This is by no means a representative sample; it’s people who self-selected to be in the survey (out of about 130 graduate students in the department). It’s in academia. It’s in the most trans-friendly city in the US (Seattle). At least in my department, the percent of LGBT graduate students sits in the 15-25% range. Much higher than 3.8%.

    I have personally seen some very real identity policing in the conservative area where I grew up. Many people would be afraid to fill out an “anonymous” survey stating that they were LGBT in any way. BYU has been known to record students’ browsing history and expel them on circumstantial evidence of “homosexual behavior.” [1] They also had a big to-do in 2014 when they asked students if they were “heterosexual, heterosexual, or other.” [2] People were legitimately afraid that if they answered “other” they would be expelled.

    Having seen this kind of policing first-hand, I think that survey data from more liberal areas is definitely more accurate, but I have no idea how to incorporate that into my mental model for the percentage of queer folks in the country. I firmly believe that 3.8% is still far too low an estimate. It also depends a lot on how you would define “LGBT.” Are you asking after people who self-identify as LGBT, or are you asking after people who have same-sex contact. There is a group of people, men who have sex with men (MSM) who do not identify as gay, but still participate in regular sexual contact with other men. [3] One CDC study found that 6.5% of men in the US had had sex with another man at some point in their lives. [4] Or maybe you want to define it as homosexual folks who live together? That will give you an estimate somewhere between 1.5% (Wyoming) to 18.5% (San Francisco County). [5] And how do you want to include trans folks? There are a lot of differing opinions on that, to be sure.

    Anyway, the whole issue is complicated and deeply entwined with local culture. I’d personally put the estimate somewhere between 10-15%, and it honestly wouldn’t surprise me if it were higher.

    [1] https://en.wikipedia.org/wiki/Brigham_Young_University_LGBT_history
    [2] https://www.huffingtonpost.com/2014/04/28/byu-survey-straight-heterosexual_n_5227529.html
    [3] (slightly NSFW) https://melmagazine.com/the-straight-men-of-the-rural-midwest-who-have-sex-with-each-other-9a1eb5fc448e
    [4] https://www.cdc.gov/nchs/data/ad/ad362.pdf
    [5] https://www.ncbi.nlm.nih.gov/pubmed/27227149

  8. I think this survey data and the survey data on what percentage of Americans are black or Hispanic cannot serve as evidence of innumeracy. Andrew is correct that people are innumerate if they don’t realize that a number as high as 25% would mean that I would be interacting with people in that category all the time. However, there is segregation geographical and social with respect to African American and LGBT. The average American could believe that the number is extremely large, but they live elsewhere. For a real example of innumeracy, you need some example of a demographic group that respondents won’t assume are geographically or socially isolated. It is hard to find such a group. Perhaps, asking people about professions would get us a better test as to whether respondents are giving such high numbers because of innumeracy or because of some other set of assumptions. If you asked people what percentage of the population is a medical doctor for instance. If you got a big number, that would be much better evidence for Andrew’s innumeracy hypothesis.

  9. Late reply here but I am looking at this topic 2019. I suspect this is because people are developing their internal estimator based on what they see on television which reflects urban areas and more recently a stress by the media on representing gay Americans at higher rates. Gallup study this year also shows overestimation of gays. To disprove my thesis we would look for rates of overestimation that go back consistently to periods of time when minorities and gays were not as well represented. But I don’t think we have consistent data earlier than 2001.

Leave a Reply to mpledger Cancel reply

Your email address will not be published. Required fields are marked *