FRR Nov 2010 | A Tale of Two Studies

The Kinsey Institute is at it again! Of course they’ve changed their name (Center for Sexual Health Promotion) and allied with other organizations and sexual policy ‘big-wigs,’ including the Centers for Disease Control (CDC) and former Surgeon General Jocelyn Elders. But to hear them tell it, they have just released the biggest, most important sex survey ever. Well, not quite.

The new study — called the National Survey of Sexual Health and Behavior (NSSHB) — captured headlines all across the world on October 4 with one of its big findings that 8% of U.S. men and 7% of U.S. women said they were homosexual or bisexual! True? Probably not.

Though the survey methodology was touted as highly representative and authoritative, it still left a number of questions. Survey respondents were drawn by Knowledge Networks, an outfit that draws random samples from a larger probability-based panel of potential households. Individuals in the larger panel are queried a few times a month via email to see if they will take part in various surveys, one of which in 2009 was the new sex study. All data collection is done online.

What’s not very clear is how agreeing to participate in such a panel (and possibly getting paid to do it) might or might not be related to one’s sexual proclivities. Pilot tests described on the Knowledge Networks website seem to indicate that their recruitment process over-samples more educated individuals and under-samples the less educated (as might be expected given the need to use a computer and internet connection in order to respond to their survey requests).

Both Knowledge Networks and authors of the NSSHB results claimed that the sample weights for the study were adjusted to match demographic profile of the U.S. Census Bureau’s Current Population Survey (CPS). Yet it is somewhat confusing as to what precisely was done.

One author noted that “Some participant characteristics were previously collected by Knowledge Networks for purposes of sample stratification and sample adjustments using post-stratification data weights. These measures included gender, age, race … and sexual orientation (heterosexual/straight, homosexual/gay/lesbian, bisexual, asexual, other)” (P. 257). However, a few paragraphs later is the statement: “During analyses, post-stratification data weights were applied to reduce variance and minimize bias caused by non-sampling error. Distributions for age, race, gender, Hispanic ethnicity, education, and U.S. Census region were used in post-stratification adjustments.”

So was the sexual orientation measure used to post-stratify the sample or not? Since post-stratification involves adjusting one’s sample — after the fact — so that the weighted proportion of respondents matches a pre-established target, knowing whether or not sexual orientation was used for these adjustments matters a great deal. If so, on what basis was the pre-established target calculated? And how could it have been when there are no regular national statistics or registries on sexual orientation — unlike characteristics such as race or gender — to provide a benchmark, and in fact, when estimating sexual preference was one of the goals of the study to begin with?

Based on the professional write-up, our best guess is that data on sexual orientation was collected separately from the main survey, but not necessarily used to post-stratify the survey sample. But if that is so, exactly how were the estimates of sexual preference compiled and computed? And was the same sample used both to estimate sexual orientation as well as to answer the ‘main’ survey?

Despite all the fanfare, the sample that was reported had a fairly high non-response/noncompletion-rate (as was true of our 1983-84 FRI survey). Consider that “[a] total of 2,172 parents (or legal guardians) reviewed the study, including the survey, and 62% (N= 1,347) consented for their child to be invited to participate. Of 1,347 adolescents [e.g., aged 14-17] contacted electronically, 831 responded, with 99% (820) consenting to participate. An electronic recruitment message was sent to 9,600 potential adult respondents, of whom 6,182 (64%) responded, with 82% (N= 5,045) consenting to participate” (P. 256-7).

Do the math and you find a 50% (5,865/11,772) overall response rate from the target sample. About 70% is common for large national surveys, but of course, sex surveys are inherently harder and fewer people tend to be willing to participate, especially those individuals of a conservative bent.

Estimates of sexual orientation reported by the NSSHB were as follows:

Table 1. Sexual Orientation by Subgroup, NSSHB (2010)

Subgroup Heterosexual Gay/Lesbian Bisexual Other
Boys (14-17) 96.1% 1.8% 1.5% 0.1%
Girls (14-17) 90.5% 0.2% 8.4% 0.9%
Men 92.2% 4.2% 2.6% 1%
Women 93.1% 0.9% 3.6% 2.3%

The estimates in Table 1 are weighted and post-stratified, as discussed above, in order to match the U.S. national demographic profile. However, when only half the target sample participates in the survey, it is quite possible that the respondents were different enough from the non-respondents to create a significant bias in the results.

What we found in our earlier survey (which had a similar number of respondents) was that liberals and the libertine were more likely to respond to a sex survey than the more conservative or sexually conventional. Homosexuals are ‘showy’ about their sexuality (e.g., ‘gay pride parades’), so just as we concluded in the FRI study that our relatively high level of non-response probably led to overestimating homosexual prevalence, so too the same may be true of the NSSHB.

In addition, many of the media headlines got it wrong to begin with (what a surprise!). While only 92% of adult men and 93% of women identified themselves as “heterosexual/straight,” a significant minority of the rest (1% of men; 2.3% of women) called themselves “other” (which presumably, from the professional report, grouped together the responses “asexual” and “other”). The FRI study similarly found that 2.1% of men and 3.9% of women said they were “not sexual,” so the “other” subset of the NSSHB should not be assumed to be homosexual. In fact, the reported tallies for combined adult gays/lesbians/bisexuals were 6.8% men and 4.5% women, somewhat less — especially for women — than the media headline of 8% gays and 7% lesbians.

Table 1 is also odd in other respects. Not the high prevalence of homosexuality, per se, but the difference between the sexual orientations of adolescents vs. adults. If we take these figures at face value, then about 3.3% of boys were homosexual, compared to 7.8% of men. For the girls, it’s just the opposite — 8.6% are homosexual in youth, but that shrinks to only 4.5% in adulthood. If these estimates are the ‘best ever,’ what mechanism explains this pattern? For instance, if individuals are ‘born’ homosexual, how could the percentages grow in boys and shrink in girls over time?

FRI suspects that the numbers themselves aren’t all that accurate. Not only do the estimated rates of homosexuality/bisexuality run considerably higher than other studies (such as the CDC’s 2005 report), but other results also don’t seem to jibe with previous research. An example of this is the reported prevalence of anal sex. The Guardian reported “More surprisingly, perhaps, the reported rate of anal sex has also increased dramatically, effectively doubling since the National Health and Social Life Survey was carried out by researchers from the University of Chicago in 1988” (, 10/5/10). So did the prevalence of anal sex really double in the past 20 years, or are the NSSHB results simply overstating the reality of things?

Part of the answer comes from ‘across the pond’ in Great Britain. It’s Office of National Statistics recently reported a much, much larger (~238,000 person) and more traditionally-executed study covering all of Britain. The Integrated Household Survey (IHS) was not primarily a study about sex, but for the first time included a question on sexual identity. Surveying those aged 16 or over (adulthood in Great Britain is reached at age 16), it found the results in Table 2.

Table 2. Self-Perceived Sexual Identity by Gender, IHS (2010)

Gender (age 16+) Heterosexual Gay/Lesbian Bisexual Other Don’t Know/Refusal/Non-response
Men 94.6% 1.3% 0.3% 0.5% 3.3%
Women 94.9% 0.6% 0.7% 0.5% 3.2%

While it is certainly possible that rates of homosexuality differ between the U.S. and Great Britain, if anything, there is even more societal and legal support for homosexual expression in England than in America. Marriage-like civil partnerships, for instance, are legal throughout Britain. So you might guess that rates of homosexuality, were they to differ, would favor the English. And yet the IHS only found 1.6% of men and 1.3% of women claiming a homosexual or bisexual identity, about a third of the rates estimated in the American NSSHB. Why the difference?

At least two factors come to mind. One, the overall response/completion rate for the IHS was upwards of 96%, compared to only about 50% for the NSSHB. Very few individuals, therefore, failed to answer the question on sexual identity in Britain, meaning that the self-selection bias was likely fairly minimal. Indeed, the Office of National Statistics (ONS) did several tests of the survey either with or without the sexual identity question and found very similar response rates each time.

Second, and related to the first factor, because the rest of survey was not about sex, but only had this one question on sexual identity, it seems likely that it was not perceived as a significant invasion of one’s privacy. This in turn should have kept non-responses to a minimum and improved the overall accuracy of the results. The ONS also compared the IHS results to several other recent studies of sexual identity/orientation and found similar estimates of prevalence.
It is also interesting that — like nearly every other study on this topic over the past 50+ years — the IHS found a significant drop-off in the prevalence of homosexuality at older ages. Table 3 documents the reported results. Although the age groupings are fairly wide, note how the proportion claiming a homosexual or bisexual identity drops from 1.8-2.0% up to age 44, to 1.2% in middle-age, and down to 0.6% for those in old age.

These results provide further confirmation of the notion FRI has studied and documented for many years: either homosexuals tend to die at younger ages than heterosexuals, or they drop out of the lifestyle as they get older. The first possibility underscores the risks and unhealthiness that seem to be inherent in homosexual practice. The second puts the lie to the ‘born that way, always that way’ claim of homosexual activists. Either way, being homosexual may be ‘fun’ for awhile, but in the end it’s no picnic.

Table 3. Sexual Identity by Age Group, IHS (2010)

Age Group Heterosexual Gay/Lesbian Bisexual Other Don’t Know/Refusal/Non-response
16-24 92.8% 1.1% 0.7% 0.4% 5.0%
25-44 94.7% 1.4% 0.6% 0.5% 2.9%
45-64 95.6% 0.8% 0.4% 0.5% 2.7%
65+ 95.1% 0.3% 0.3% 0.6% 3.6%