Today’s post title is stolen from a paper by the same name “Cohort Increases In Sex With Same-Sex Partners: Do Trends Vary by Gender, Race, and Class?” in Gender & Society, by Mishel et al, which came out in January.
We examine change across U.S. cohorts born between 1920 and 2000 in their probability of having had sex with same-sex partners in the last year and since age 18. Using data from the 1988–2018 General Social Surveys, we explore how trends differ by gender, race, and class background. We find steep increases across birth cohorts in the proportion of women who have had sex with both men and women since age 18, whereas increases for men are less steep. We suggest that the trends reflect an increasingly accepting social climate, and that women’s steeper trend is rooted in a long-term asymmetry in gender change, in which nonconformity to gender norms is more acceptable for women than men. We also find evidence that, among men, the increase in having had sex with both men and women was steeper for black than for white men, and for men of lower socioeconomic status; we speculate that the rise of mass incarceration among less privileged men may have influenced this trend.
The paper is of interest because of the topic itself, but also because of the statistics. It’s a rare paper that uses predictive probabilities; in part, anyway.
The Abstract gives the hint. The model is a logistic regression, unfortunately using p-values, which exaggerate evidence. However, it’s a rather tepid, strictly additive model, with sodomy predicted by age cohort, race, region, and a couple of others. The effect is thus to smooth the observed data. They, as is common, call these measures “controls” or “adjustments”, which regular readers will recognize as unfortunate terminology.
The predictive probabilities aren’t true predictive probabilities, but merely model output probabilities. I mean the uncertainty in the parameters is not accounted for (not integrated out). Thus the signals demonstrated will be too sharp.
That, plus all the other usual and regular caveats that apply to survey data should be kept in mind. Perhaps the exact percentages aren’t right, but we have some information about trends which can be of use.
If you can’t see it, the caption reads:
Percentage of Women and Men in each Cohort Who Have Had Both Male and Female Sexual Partners.
NOTE: Percentages are predicted probabilities (× 100) from regression models that adjust for cohort differences in composition by age, race, and other control variables. “Both” implies at least one male and at least one female sexual partner. Numbers are from Table 4.
Maybe it only feels high, the 12% rate for men in the 1984-2000 birth cohort. This doesn’t mean people who “identify” as homosexuals, but this is men who have practice sodomy at least once however they “identify”. The story is worse for women, but probably for obvious reasons.
This bit about “identify” is not a small point. The authors say “There is no trend in having sex with only same-sex partners since age 18 for either men or women; the predicted probabilities are approximately 0.01 or 0.02 in every cohort for both men and women”. Meaning strict homosexuality, if we can call it that, is fairly constant, and low.
More shocking is the black-white difference.
Top caption is “Percent of White and Black Men in each Cohort Who Have Had Both Male and Female Sexual Partners”. As many as 20% of black men have had sodomy?
Anecdotal culture has long said blacks view homosexuality, and hence presumably sodomy, more negatively than whites. Or is there perhaps a more mundane explanation?
We suggest that the dramatic rise in incarceration might be one factor in these group differences in trends. Being black or hispanic and having lower education are well-known predictors of incarceration, with the highest rates for black men with little education. If, as we hypothesize, incarceration increases a man’s probability of having a same-sex partner, for the majority of men who saw themselves as heterosexual before incarceration, then we would expect this increase to show up in trends in having ever had sex with both sexes, and this is what we find.
That explains the black-white difference, but not the increase for both races per se.
The most obvious objection was given by the authors:
One limitation of our analysis is that, while we have interpreted increases in reports of sex with same-sex partners as indicative of real behavior changes, we cannot be certain that the observed change is not an increase in the willingness to disclose sex with partners of the same sex rather than a change in actual behavior. We believe, however, that the increases are large enough that they are unlikely to be entirely a function of declines in underreporting.
What about those born after 2000? Good guess, using these models, is about 15% for men and maybe 25% for women. After these folks become sexually active, of course.