3 ms·

> Though that can't be taken as a justification of "20% of the men get 80% of the women", which is either (obviously) false as stated, also true of the distribu

> Though that can't be taken as a justification of "20% of the men get 80% of the women", which is either (obviously) false as stated, also true of the distribution of X among women, or incoherent as a meaningful metric. Not that I agree with the quote, but... why would any of these assertions be true? I suppose that depends on what "get" means in the quote. I think it's immediately evident that, say, 20% of men aren't married to 80% of women, for example -- so if that's what "get" means, then that doesn't add up. But I see nothing inherently wrong with the notion that 20% of men could be taking 80% of women off the market. Not because these men are dating approximately 4 women each at the same time, but because they could be rapidly playing the field. If these men are engaging in short flings with women, it would seem reasonable that the women might not want to settle on anything less than the 80 percentile of men, knowing that they can briefly hold out until another top 20%er comes along. That effectively takes them off the dating market for the other men. Again, the stats could disagree with such a hypothesis, but I don't think there's any inherent logical contradiction in the hypothesis itself. If we're talking about hooking up with people met at the bar, I can maybe see the quote making sense. But if we're talking about meaningful relationships, I actually find it hard to believe that a man is having a hard time finding a partner because "20% of men are stealing all of my opportunities." Though that could be easy for me to say if I'm just that damned attractive, but I don't suspect that's the case :).

scarmig 4y ago
Noting that we agree that the actual stats would disagree about the hypothesis. If I might take the liberty of riffing off your model, let's take it as "looking at the the sexual encounters that occurred over a duration on the scale of months, you can select a certain subset (~80%) of women who will have had sexual relationships exclusively with a certain subset (~20%) of men." I'd put that version down as falling into my first category of obviously false. I.e. sexual relationships don't form a disjoint graph, even excluding the edges from that 20% of men to the nonselected 20% of women. I don't think it's inherently incoherent, just wrong. We can weaken it to "Over some duration on the scale of months, you can select some relatively small subset of men (~20%) who will have been the partners in a disproportionate number (~80%) of sexual relationships." This is likely and plausible, but it's likely and plausible that you could apply that same procedure to women to get a similarly surprising pair of numbers, because of how social graphs work, i.e. the bulk of a measure is driven by the extremes of the distribution. It's interesting, but it doesn't tell us much about the differences in how heterosexual men and women pair. I'd read your proposed model as stronger than the second version and perhaps a bit weaker than the first. Maybe modifying the first model by adding something like "Half of those selected 80% of women will also have formed a single edge with the nonselected 80% of men." I think this is a viable way to recover it into something more plausible, but in doing so we've the changed the actual claim to something like "you can select a certain subset (~80%) of women who will have had sexual relationships exclusively with a certain subset (~40%) of men." That's a much weaker and less shocking claim than the original, and I wouldn't be surprised if you could again apply that same procedure to women and get similar numbers out of it. I think we can make statements about the structure of the sexual relationship graph that point to differences in how men and women form sexual pairs, but those aren't as easily understandable. "The distribution of sexual encounters among men have a Gini coefficient of 0.654, while among women it's 0.562" doesn't really make for a punchy soundbite.

scarmig 4y ago

Noting that we agree that the actual stats would disagree about the hypothesis. If I might take the liberty of riffing off your model, let's take it as "looking at the the sexual encounters that occurred over a duration on the scale of months, you can select a certain subset (~80%) of women who will have had sexual relationships exclusively with a certain subset (~20%) of men." I'd put that version down as falling into my first category of obviously false. I.e. sexual relationships don't form a disjoint graph, even excluding the edges from that 20% of men to the nonselected 20% of women. I don't think it's inherently incoherent, just wrong. We can weaken it to "Over some duration on the scale of months, you can select some relatively small subset of men (~20%) who will have been the partners in a disproportionate number (~80%) of sexual relationships." This is likely and plausible, but it's likely and plausible that you could apply that same procedure to women to get a similarly surprising pair of numbers, because of how social graphs work, i.e. the bulk of a measure is driven by the extremes of the distribution. It's interesting, but it doesn't tell us much about the differences in how heterosexual men and women pair. I'd read your proposed model as stronger than the second version and perhaps a bit weaker than the first. Maybe modifying the first model by adding something like "Half of those selected 80% of women will also have formed a single edge with the nonselected 80% of men." I think this is a viable way to recover it into something more plausible, but in doing so we've the changed the actual claim to something like "you can select a certain subset (~80%) of women who will have had sexual relationships exclusively with a certain subset (~40%) of men." That's a much weaker and less shocking claim than the original, and I wouldn't be surprised if you could again apply that same procedure to women and get similar numbers out of it. I think we can make statements about the structure of the sexual relationship graph that point to differences in how men and women form sexual pairs, but those aren't as easily understandable. "The distribution of sexual encounters among men have a Gini coefficient of 0.654, while among women it's 0.562" doesn't really make for a punchy soundbite.