The freakonomist has struck again. His latest article in the New York Times is on a topic we've hotly debated before: is human sexuality biologically determined?
In the article, Levitt and Dubner discuss a paper by Andy Francis. The paper might be more fulfilling for the hardcore among you.
The paper reacts to some interesting correlations. Apparently, while homosexual males are more likely to have a friend with AIDS, they are far less likely to have a relative with AIDS. For homosexual females, these correlations flip. If the data holds, what conclusions can be drawn?
Francis thinks this provides strong evidence that biology is not the sole determinant of sexual identity. If fear of AIDS is playing a role in determining sexual identity, other seemingly innocuous costs and benefits might be too.
From the abstract:
People who have a relative with AIDS, on average, have more knowledge, awareness, and fear of AIDS than those who do not. ...I find that AIDS causes people to shift from less safe sexual activities to safer ones. I find that AIDS [awareness] causes men to shift from homosexual to heterosexual behavior, desire, and identity, whereas AIDS [awareness] causes women to shift from heterosexual to homosexual desire. Neither genetic nor hormonal theories of sexual orientation can explain these findings. Therefore, biology is not the sole determinant of sexual behavior, desire, and identity.
Isn't their a stronger risk factor associated with men with AIDS than women? Would this explain the turn among both AIDS aware genders to engage in sexual activity with the female gender?
ReplyDeleteThat's his explanation exactly, yes.
ReplyDelete150 people does not causation prove, as noted in the NYT article: "Because the sample size was so small - simple chance suggests that no more than a handful of men in a group that size would be attracted to men - it is hard to reach definitive conclusions from the survey data."
ReplyDeleteIf there was some kind of causal effect, wouldn't we have seen a pronounced drop in gay population--unrelated to deaths--since about 1985? Is there any suggestion in the materials that such a drop has occurred?
150 people does not causation prove,
ReplyDeleteProving causation has little to do with sample size. Causation may be suggested if there are few alternative explanations for the correlation. Here, genetics and hormonal trends cannot explain the trends, as he notes in the paper.
Also, extra statistical safeguards are used with the small sample size, like an emphasis on the correlations which had small p-values, the data that was highly unlikely to result as random chance.
If we can eliminate third factors and chance, and we know which factor was prior to the other, we can start to feel more comfortable with the idea of causation. All these are addressed in the research.
As Steven Landsburg recently noted:
"That's largely what good empirical economics is about—finding thoughtful and creative ways to distinguish between correlation and causation. Whenever I write on an empirical topic, readers send me e-mail "explaining" that correlation and causation are not the same thing. When they read about a medical breakthrough, do these same readers write to science reporters "explaining" that lab results can't be trusted unless the test tubes are clean? Competent economists always address the causation/correlation issue, just as competent biochemists always clean their test tubes."
-Slate
Are you honestly suggesting there's not a "third factor" that is at least as plausible as a bunch of guys deciding to be hetero because they fear AIDS? How about social stigma? Family pressure? Religious inclination? Or unwillingness to self-identify as gay to the pollster?
ReplyDeleteTo say that we should just trust economists' conclusions because we should accept that "their test tubes are clean" is to reject the idea of academic discourse--and to give economists and their interpretations of statistics way too much power.
I find it interesting that (if anonymous is correct and the results are skewed by sample size/third factors/etc.) the prerequisite of having a relative with AIDS causes men to underreport homosexual behavior and women to over/correctly report homosexual behavior.
ReplyDeleteOtherwise, if women underreported too, then anonymous' argument would be moot. Is this correct?
Are you honestly suggesting there's not a "third factor" that is at least as plausible as a bunch of guys deciding to be hetero because they fear AIDS? How about social stigma? Family pressure? Religious inclination? Or unwillingness to self-identify as gay to the pollster?
ReplyDeleteThese factors were discussed and controlled for in the paper.
If you see other relevant factors, you have a positive burden of proof to mention them, no matter how obvious you think they are. Otherwise, causation would be impossible to assert.
To say that we should just trust economists' conclusions because we should accept that "their test tubes are clean" is to reject the idea of academic discourse--and to give economists and their interpretations of statistics way too much power.
I agree that we shouldn't trust people blindly. But he dealt with these concerns in the research, as good economists tend to do.