Three effects I can see that haven't been mentioned yet:
Prostitutes have a significantly lower life expectancy than non-prostitutes - a mortality rate 40 times as high. Overall mortality rate for, say, a 25 year old woman, is about 1 in 2000. For prostitutes, it's thus about 1 in 50. Compound that for, say, ten years, and you get only 82% of prostitutes surviving past that point.
The lowest-class prostitutes would service the most men and have the highest mortality rate. If they have a 10% mortality rate per year (only five times worse than the average across all prostitutes) over a career of 10 years, then the whole cohort of women older than that point will have about 65% of them dead. After observing the horror of this, we note that death also renders one unable to respond to surveys.
Second effect: prostitutes especially but anyone highly promiscuous in general may have a hard time remembering just how many partners they've had, and may be poor at estimating it, even if they're not trying to minimize it. Especially when they need to discount for repeat customers by a somewhat difficult to estimate margin... it's a genuinely tough problem when you get up into the professional levels, at least without bookkeeping, which will frequently not be kept.
Third effect: prostitutes and ex-prostitutes, being especially disadvantaged, may be harder to reach.
Something I've always wondered about when hearing this 'statistic', but never actually checked - does the gap disappear if you are careful in defining sexual partner?
eg, if the question asked "with how many people have you had consensual penetrative heterosexual intercourse?" rather than "how many sexual partners have you had?"
Sounds about right. I would imagine about half the women I have had sex with would not admit to having had sex with me.
One of taw's old posts is relevant here; his list of possible skewing factors even overlaps with yours quite a bit! He also found a study of students suggesting that the women's reported number of sexual partners changed when they were hooked up to a fake polygraph or thought an experimenter might see their answers. (See also the comments for other explanations.)
Today I came across an article in the Telegraph that states that an average man has 9 sexual partners in his lifetime, whilst a woman has only 4.
Did they include female prostitutes in their samples, when they were calculating said "averages"?
Here's a similar situation: I have read two papers that claimed to explain this situation, in incompatible ways.* So they can't be both true.** It demonstrates that at least one of them is measuring something wrong. This is an extremely common problem in the social sciences. Once there is an established mystery, many people propose mechanisms to explain the discrepancy. They go out and measure and many find that the effect is just about the right size. It never happens that five people each explain 10% of the gap.
* One paper made a point of surveying prostitutes. The other used lie detectors. But the details are not important.
** Conceivably both effect sizes could be true, if there is yet another effect that cancels one out. Age difference is an effect in the opposite direction. But it is extremely unlikely that the three effects are all the same size.
Of course, if you see a large gap like this, and you come up with a theory that explains only 10% of the gap, that's not news, so you forget about it. The only people that have something to report are those whose theory could potentially explain the whole gap.
Honestly, from the things I've heard, setting gays at ten percent of the population, the idea that gay men have 20 times as many partners as heterosexuals or lesbians isn't that hard to imagine.
When it comes to self-identification most studies return results ranging from 1 to 4% depending on country (and most of these where done in Western countries). Even though obviously many people will remain closeted even on a anonymous poll even in a society like Norway, an estimate of 10% is not really the most likley one.
Numbers like 5% or lower seem more plausible.
Lots of rounding. 9 is roughly twice 4. 20 times 10% would roughly double an average. It would actually be slightly less, but I don't really care, especially seeing as the data is by it's nature pretty fuzzy.
http://wiki.answers.com/Q/What_is_the_average_number_of_sex_partners_a_gay_man_has_in_a_lifetime
Ok, gotcha. Thanks. I was looking at the comment by itself and couldn't figure out how to get from "gays are 10% of the population" to "gay men have 20 times as many partners".
See also previous research into this phenomenon like http://www.physorg.com/news10824.html and http://www.mv.helsinki.fi/home/jproos/jonasdottir.htm
Interesting. Do we therefore see a parallel discrepancy in something less socially gender loaded?
There is also a question of whether they are using "average" to refer to mean or median
That's the first explanation I thought of too, but the article explicitly says "mean average" when referring to women. I can't imagine that they used the mean for women, and the median for men....
Ah, then yes, they must be either
1) mistaken,
2) not restricting to heterosexual pairings, or
3) enlightening us as to a difference in reporting frequency of men and women.
Edited to add:
Age could potentially play a role as well?
3) enlightening us as to a difference in reporting frequency of men and women.
This. The survey says "Around 10% of both men and women did not answer the questions on sexual partners.". If women tended to embarrassed by having too many sexual partners, and men by having too few, this would explain the difference.
For the math to work out, the 10% of women who did not respond would have to have 45 partners, on average. I'm skeptical.
Math: The reported average among men is 9, thus the actual average is at least .9(9) + .1(0) = 8.1. If the missing tenth of women have an average of x partners, the actual overall average is .9(4) + .1(x) = 3.6 + x/10. Setting these equal, we get 8.1 = 3.6 + x/10, which yields x = 45.
For the math to work out, the 10% of women who did not respond would have to have 45 partners, on average. I'm skeptical.
This creates an ethical dilemma for the scientists performing the study!
You are right - age could be very important - and I didn't think of it. For example, if men preferred much, much older women, most women in the sample would then have few sexual partners. Both men and women would, by the time they died, have had the same number of (hetero) partnerships on average, but whilst they were alive and around to fill in surveys, the numbers could be quite different.
However, the effect is probably in the wrong direction to help explain the reported difference, women live longer, and men seem to prefer (somewhat) younger women.
With standards like that, the other statistics in the article are probably also untrustworthy. "I do wonder under what other misapprehensions I continue to labour."
The median vs average comment sums it up. Imagine a society with one female prostitute who services all of the otherwise monogamous and mate-for-life males. In this society, the median for women is one, the median for men is two. Similarly, if we average while discarding the outliers we will see an asymmetric average.
I expect there's a way to get more accurate survey results using a randomized response method (which I learned about from this post). I can't quite figure out how you'd do it, though.
Give them a coin and a d20. They flip the coin, and if it comes up heads, they tell you how many partners they've had. Otherwise, they roll the d20 and report the number that comes up minus one (to protect the identity of virgins). This still leaves people with more than 19 partners in the open. A geometric random number would be better, but more unwieldy to implement.
Alternatively: rig a random number generator such that it can be fixed to a known number without the subject's knowledge. Demonstrate the random number generator in its unfixed state.
Then fix the generator for one trial, leave the room, and have the subject report that number plus their number of sexual partners.
Today I came across an article in the Telegraph that states that an average man has 9 sexual partners in his lifetime, whilst a woman has only 4.
Let's assume for a moment that in fact all these men and women are heterosexuals. In that case, each partnership contains one man and one woman. So, in total, the number of partnerships that women enter into is exactly the same as the number of partnerships that men enter into. Given a few other pretty well-known facts - that men are roughly as numerous as women, and live roughly as long, we deduce that on average men enter into roughly as many sexual partnerships as women.
There is, of course, a potential flaw in this - we know not all partnerships are in fact heterosexual. We know that heterosexual partnerships must have as many female participants as male, but we know no such thing about homosexual ones. Perhaps homosexual men have many, many more partnerships than heterosexual men, or women of any inclination? Whilst there is some evidence in favour of this concept, I don't really think it's going to be a big enough effect to skew the entire ratio this much. In order to have a ration of 9 to 4, most male partnerships would have to be homosexual ones.
A better theory is that the data is not very good. Perhaps men tend to exaggerate the number of sexual partners they have, even in anonymous surveys. Perhaps women tend to emphasise their degree of virginity. Perhaps the distributions are different - perhaps for example, most women have fewer partners than most men, but some women have lots? And perhaps this sort of information is lost in the survey?
Probably the only safe conclusion we can draw from these figures is that journalists and statistics don't mix.