By Sally Raskoff
Sexual behavior is challenging to measure. Alfred Kinsey famously studied sex in the mid-twentieth century, and although groundbreaking, his study relied on convenience sampling which prevents us from being able to generalize the results to the entire population.
The National Health and Social Life Survey (NHSLS), conducted in 1992, has been considered a more scientifically rigorous study. Two more recent studies, the National Survey of Family Growth (NSFG) and the National Survey of Sexual Health and Behavior (NSSHB) provide us with a more current picture of sexual behavior in America.
The NHSLS, NSSHB, and NSFG are all national probability samples, which means that we can generalize the findings to the larger population even though they didn’t survey everyone in the country.
The older NHSLS is based on 3,432 respondents (1,901 women, 1,531 men), ages 18-59. The NSFG is sampled an astounding 13,495 people (6,139 men, 7,356 women), ages 15 to 44. The NSSHB sampled the largest age range, including 5,865 people (2,929 women, 2,936 men,), ages 14 to 94.
The findings from all these studies are quite interesting--and not just because they have to do with sexual behavior.
Each study asked about sexual orientation identity. The older NHSLS data showed 98.6% of women and 96.9% of men said they were heterosexual, 0.9% women and 2.0% men said homosexual, gay, or lesbian, and 0.5% and 0.8% men said they were bisexual.
The newer studies show slightly different data:
| Sexual Orientation | NSSHB | |
| | 2009 | |
| WOMEN | 14-17 | 18-70+ | NSFG, 2006-2008 18-44 |
| Heterosexual | 90.5 | 93.1 | 93.7 |
| Homosexual, Gay, or Lesbian | 0.2 | 0.9 | 1.1 |
| Bisexual | 8.4 | 3.6 | 3.5 |
| MEN | 14-17 | 18-70+ | 18-44 |
| Heterosexual | 96.1 | 92.2 | 95.7 |
| Homosexual, Gay, or Lesbian | 1.8 | 4.2 | 1.7 |
| Bisexual | 1.5 | 2.6 | 1.1 |
Source: NSFG: Tables 12 & 13; NSSHB: Table 1.
The table above shows the primacy of the heterosexual category, with which most people identify. However, comparing data on identity to those based on behavior, a fascinating pattern emerges: Identity does not always match behavior.
| Sexual Behavior | WOMEN | MEN |
| NSFG (2006-2008) | 15-24 | 15-24 |
| Any Opposite Sex Contact | 70.1 | 71.7 |
| Any Same Sex Contact | 13.4 | 4.0 |
| No Sexual Contact with another person | 28.6 | 27.2 |
Source: NSFG: Table 7.
Notice how the identity data patterns show very few people aligning with the homosexual or bisexual categories. Yet when asked about homosexual or bisexual behavior, much higher percentages appear.
Since both studies utilize probability samples, they are both representative of the larger population. There can be some sampling error, wherein some groups might be systematically excluded in ways that might bias the data. When this happens one sample may not fully represent its population. Is that what’s happening here? Or is there more going on?
Part of the answer might lie with methodology. The NHSLS used face-to-face interviews and focus groups. The NSFG used in-person interviews using “Audio Computer Assisted Self-Interviewing” technology. The NSSHB used “Research Panels accessed through Knowledge Networks” via the internet, although they did provide hardware and internet access when necessary. When dealing with a sensitive subject like sex, how the data are collected will have a big impact on the results.
Another clue would rest with the different ways the questions were asked. Each study asked about the issues in slightly different ways.
For the NSFG, what was considered “sexual behavior” was different if it was same-sex or other-sex contact. For the NSSHB, questions were about specific behaviors based on who were their partners.
While the studies were conducted at different times, that is not necessarily problematic. Cultural patterns such as these do not tend to shift quickly.
Our scientific techniques for high-quality research are based on systematic methodologies. Because such techniques can yield different results we need to replicate or repeat research studies as often as possible. Many studies on the same topic can give us a lot of data patterns which then can be compared and compiled so that we can see more clearly what is going on in our social world. What other factors do you think might create more high-quality data on sensitive issues like this one?