At Tsinghua I am teaching a class called Frontiers of Psychology. The students are reading *The Man Who Would Be Queen* by Michael Bailey. At one point Bailey mentions what is sometimes called *the older brother effect*: If a man has one older brother, he is more likely to be gay than if he has no older brothers, controlling for several things. This has been seen many times. In 1962 it was reported that gay men have more older siblings than other men but not until 1996 was it determined that this was due to more older brothers.

Bailey doesn’t mention the strength of the effect. *The Wisdom of Crowds* by James Surowiecki is about research that found that non-experts can do an excellent job of estimating this or that number (such as the weight of a particular cow) even when they know little about it. Their answers are excellent in the sense that the average of their answers is very accurate. Perhaps my students, who had read two-thirds of Bailey’s book, could accurately estimate the strength of the effect.

I posed the question like this. Suppose that when a man has no older brothers, his chance of being gay is 2.0%. What is his chance of being gay if he has one older brother? I gathered an estimate from every student. The median of their estimates was 8%. The correct answer is 2.7%.

Did the studies control for the mother’s age at time of conception?

I suspect there is a small social specialization effect, but if they didn’t control for the mother’s age, they might just be picking up that by definition of having an older brother, a younger brother is conceived later. The older an egg is the more degraded its DNA, leading to a greater chance of a tuning error in sexual targeting.

Or if they aren’t stupid, you’ve managed to quantify a crowd response to narrative, which is kinda cool. Maybe PR companies beat us you to that research though…

It’s not clear to me that this is a bad estimate, given the knowledge of the individual students.

What was the mean value of their estimates?

In the land of the One Child policy, questions about brothers must be hard to cope with.

So countries with larger families should have more male homosexuals per capita?

I didn’t compute the mean, but it was probably more than 8.

yes, countries with larger families will have more male homosexuals per capita.

yes, at least one of the analyses controlled for both mom’s age and dad’s age.

Fascinating. That said, i bet their answers would have been far closer if the question had been framed differently. Imagine asking: “How much more likely, in percentage terms (i.e. 100% equals twice as likely), is it that a boy will become a gay man if he has older brothers?

It isn’t clear that this is a bad estimate if you asked them in this manner – giving them 2% instead of X%.

Given the number of 2%, the number of gay men in the other population needs to be at least 8% in order to meet the intuition that at least 1/20 people are gay (I’ve heard numbers ranging from 1/20 to 1/10).

They may have accidentally used this heuristic of trying to get the overall number correct while trying to answer your question.

This does not seem to be a very telling result. Surowiecki does not claim that crowds are good at baseless, or nearly baseless guesses, does he? On what basis would these students have been expected to have any idea that the actual number would turn out to be 2.7%? That example seems less like Surowiecki’s go-to cases where you estimate the number of something like gumballs in a big glass jar, and more like asking a crowd for the exact sequence of cards in a deck of playing cards. This isn’t “The ESP of Crowds.”

Wyman, there are examples in Surowiecki’s book where you’d think it would be impossible for the crowd to do well. But it does. I haven’t seen anything where someone correctly predicts where this sort of crowd-sourcing will do well and where it won’t do well. It was more informative to find out what the answer was — how well the students would do — than be sure I knew what would happen.

I wonder whether the measurement was validated. But it is interesting that someone can write an entire book about a phenomenon without an adequate measurement. Its like measuring the speed of a high-tech train with a thermometer.