August 18, 2008

Where are girls most likely to dig older guys?

I don't need to sell this idea with a long-winded set-up, so let's get right to it. The data come from David Buss' (1989) seminal study of mating preferences around the world (free PDF), part of which focused on how older or younger the person's ideal mate was. But we need to adjust this somehow to account for the fact that the samples differed in average age -- a group of 40 year-olds who prefer a guy 10 years older is different from a group of 20 year-olds who prefer a guy 10 years older. (My apologies to the 30 year-old guys who I just gave heart attacks to with that example.)

Luckily the data also show the average age of the sample, so I refer to "preferred age difference" as the sample's average absolute preferred age difference (e.g., "3 years older") divided by the sample's average age. In the previous example, it would be 25% and 50%, respectively. This is not the same as the average relative preferred age difference, since the ratio of two averages is not the average of the ratio. I would need the individual data-points for that. Still, it's a sensible way to correct for the age of the sample.

Also, because I only looked at averages, we expect them to not be "10 years older" no matter what the sample's age, because the average female probably prefers between 3 to 5 years. We are looking for girls in the far-right tail, the ones who are fine with a much older guy. Buss' data do contain standard deviations, but I trust these less than the averages since it's harder to accurately measure a variance than an average, and I don't know how normal each sample's data is. So we'll have to operate on the assumption that where the average is higher, "extreme" values become less extreme, like finding a 6'10 man among 1 million Dutch guys than among 1 million Vietnamese guys. Small differences in the mean have huge effects in the tails of a distribution.

Moreover, how does she know what her ideal guy is? -- girls surely to underestimate rather than overestimate the age of their ideal guy. She says 5 years older, but in reality she'd be fine with 10, at least if the guy made up for being "too much" older in some way, which I assume anyone who undertakes such a project does. The unreliability of the girls' estimates does not matter when we rank the countries -- all girls provide unreliable estimates, but it's doubtful that some countries are systematically more clueless than others.

The samples are not all probability samples -- some are pretty representative, and some are convenience samples (read the PDF). Buss explicitly says that they are not meant to tell us about the individual countries, but only to get a large sample of humanity to study universal preferences. Lacking better preference data, though, we work with what we have. Small samples suggest college students, but that's who you're after anyway, right? So representative or not, they'll do for our purposes, and we'll put more faith in the data from larger samples.

The table below shows the samples arranged from greatest to least preferred age difference (Diff %), together with average age of the sample (Age), average absolute preferred age difference in the sample (Diff Yrs), GDP Purchasing Power Parity per capita, and sample size (N). I included GDP after examining Buss' data and seeing that it was going to be a strong predictor of preferred age differences (more on that later).

country age diff % diff yrs GDP N
Greece 18.71 24.27 4.54 29,172 65
Nigeria 21.13 23.19 4.9 2,035 55
Iran 22.74 22.43 5.1 10,624 27
Indonesia 22.76 20.61 4.69 3,725 55
Colombia 24.34 18.53 4.51 6,724 78
Taiwan 20.54 18.40 3.78 30,126 278
Zambia 22.6 18.32 4.14 1,309 49
Brazil 21.72 18.14 3.94 9,695 355
Bulgaria 23.06 18.13 4.18 11,302 142
South Africa White 19.44 18.00 3.5 9,761 81
Yugoslavia 20.72 17.42 3.61 12,124 74
Israel Palestinian 21.5 17.26 3.71 25,799 55
New Zealand 16.92 17.20 2.91 26,379 76
Israel Jewish 23.29 16.96 3.95 25,799 268
Venezuela 22.52 16.07 3.62 12,166 98
South Africa Zulu 23.52 15.99 3.76 9,761 48
Poland 21.44 15.76 3.38 16,311 118
Japan 19.37 15.75 3.05 33,577 153
Estonia 18.32 15.56 2.85 21,094 150
France 25.83 15.49 4 33,188 91
China 22.46 15.36 3.45 5,292 235
United States Hawaii 22.76 14.50 3.3 45,845 113
Ireland 19.27 14.43 2.78 43,144 67
Norway 22.46 13.89 3.12 53,037 67
India 24.9 13.21 3.29 2,659 144
Germany West 29.14 12.70 3.7 34,181 553
Netherlands 21.65 12.56 2.72 38,486 240
Italy 25.96 12.48 3.24 30,448 55
United States Mainland 20.37 12.47 2.54 45,845 852
Australia 23.12 12.37 2.86 36,258 202
Canada English 23.05 11.80 2.72 38,435 45
Belgium 21.38 11.51 2.46 35,273 90
Finland 24.6 11.50 2.83 35,280 149
Spain 22.75 11.43 2.6 30,120 80
Sweden 26.7 10.90 2.91 36,494 83
Great Britain 21.09 10.72 2.26 35,134 84
Canada French 25.17 7.23 1.82 38,435 71


We're looking for a country high on the list with a large N, although again even moderate sample sizes give us a good feel for what college students are up for. Surprisingly, the first choice is Taiwan. If you like Asian girls, that's where you should go, especially since it's a first-world sample. Brazil is up next, and those girls have enough good PR already that I don't need to add to it. Bulgaria also looks like a good bet -- see my post on Balkan babes if you don't know how great they look, kind of Slavic and Mediterranean (if you know what I mean). Israeli Jews, Poles, and Japanese seem OK. Estonia is misleading because the average girl is barely over 18 and prefers a guy who's less than 3 years older than she is, not good.

While still being cautious about the smaller samples, Greece, Colombia, South African Whites, and the former Yugoslavia also look pretty good. I think we can be fairly confident about Greece and Yugoslavia since Bulgaria and Israeli Jews had large samples and scored highly, so there's probably an "Eastern Mediterranean" pattern here. Ditto for Colombia, based on similarity to Brazil: South Americans who are mostly Mediterranean but with a fair amount of African admixture.

If I had to bet, I'd say Greece is the optimal place. Not only do they score the highest, but their sample is so young -- barely legal, as we say. The average girl in this sample is 18 and claims to want a guy who is 23, and tacking on a couple years for people's lack of intuitive precision, I think you could say 25. And this is just the average girl -- you might have to search 100 or even 1000 Greek girls, but it wouldn't be that hard to find an 18 year-old who wanted a 30 year-old, to sleep with, to marry, whatever. I can't recall where I read it, but I distinctly remember some online forum on this topic where a world traveler said that Greek girls were among the most willing to associate with older guys.

And aside from Greece's climate, history, culture, bla bla bla that you really care about, it's also a first-world country and member of the EU, so the girls can easily afford to wax or laser-remove their body hair (hey, no ethnic group is perfect). I've never been there, but judging from the Greek international students I saw at college (somewhere between 20 to 40 of them), they're pretty nice, and they along with the Turkish girls had a reputation for it. The only other group reknowned for their looks were, of course, the Brazilians. It will not escape the reader's notice that where girls dig older guys, the girls have big butts. Somehow it all works.

Another thing to bear in mind is the country's population size: if you're looking for a 1 in a million girl, she'll be easier to find in Brazil than Bulgaria.

Look at how bad things are at the lower end of the table: sure, it's only a moderate sample size, but the average French Canadian girl is 25 and claims to want a guys who's 27, and Finland's pretty bad too. You'd think that as girls got older, they'd want increasingly older guys, but it looks like just the opposite happens on a group level. Here is a graph of the country's preferred age difference by how old the average female is:



Disgusting. The older the sample is, the closer they want the guy to be relative to themselves. On the other hand, thank god for those adventuresome youngsters! The Pearson correlation between these two variables, each point weighted by sample size, is -0.32 (p two-tailed = 0.0552). I tested two hypotheses, so this is more of a "trend" than a "significant" finding, but it depends on how religious you are about p-value cut-offs. We've already hinted at it, but let's take a look at preferred age difference by GDP:



Sure enough, in richer countries, females prefer males who are closer to themselves in age. This is part of the spread of middle class values, culturally and genetically, that began with the emergence of merchant classes during the Late Middle Ages and really took over during the Industrial Revolution. Some middle class values are fine, but I can't stand how segregated these societies are by age, and seeing lovely nubile girls holding hands with losers their own age is just part of it. (Another aspect is that adults are completely clueless about what's going on with youngsters and rely on rumor as their news source.) The weighted Pearson correlation here is -0.63 (p two-tailed = 0.0001), so it's strong and significant. Still, there are some points that are above the trend: Greece is the strongest outlier above and French Canada the strongest outlier below, confirming what we've already said about their suitability.

Worry-warts may say that this shows that girls in poorer countries "really" want guys their own age, but alter their preferences toward even older guys for increased security. Maybe, and maybe there's another explanation. In any event, you're meeting her preference and therefore making her happy if you're an older guy. Why she likes you doesn't matter, as long as she does. Indeed, that's their response to my disgust at the 18 year-old who's dating a 19 year-old -- it's what she wants. The worriers will counter that a small age difference is what the girl "really" wants. But she wants whatever she wants. It certainly isn't the "deep down preference" in the sense of what's been true during human history and evolution -- the alarmists' wishes characterize the developed world.

So there you have it. Again, the data aren't perfect, but let me know if you've got something better. This is just a guide, so you can't sue me if your efforts based on it don't pan out.

17 comments:

  1. next stop, greece!

    funny, but this data matches anecdotes i've heard about may-december greek romances from greek friends who immigrated to the US and tell me about all the older guys getting it on... and marrying!... 18 year old greek girls. it's the one thing they miss the most about leaving greece, other than the food.

    It will not escape the reader's notice that where girls dig older guys, the girls have big butts. Somehow it all works.

    pathogen influenced?

    ReplyDelete
  2. More confirmation is always good.

    Pathogens and big asses -- actually just had a little discussion about this with the GNXP people. I think so. Lots of pathogens leads to shorter-term mating strategies, and the genes or hormones or whatever that achieve this in females give them bigger asses and somewhat masculinized faces.

    ReplyDelete
  3. regarding the supposed preference of older women (28 - early 30s?) for men closer to their age: this is a great example of a declared preference seeming at odds with the observed reality. most unmarried 30ish women i know who are back on the market are formally dating men considerably older than themselves (though some occasionally screw around with younger guys). now either they are bending to biological reality and settling or they actually prefer the older men despite their statements to the contrary. (women behaving at odds with what they claim they want? whatta surprise!)

    a couple hypotheses to explain this discrepancy spring to mind. given women's alternating taste in men based on their ovulatory cycle, they may be responding to these questionnaires while referencing the feelings they have at the time of ovulation when the image of the lantern-jawed hot stud is foremost on their minds. it's like asking a girl why she likes a certain guy and she replies "oh, he's cute!" which is really girlspeak for "oh, he exhibits that suite of alpha traits that presses my buttons!". girls can't really put into words what turns them on in men because men's sexual attractiveness is more complicated than women's, so similarly when asked what age difference they prefer they may just default to one obvious signal of attractiveness like "youth", when so many more variables are involved.

    the other hypothesis may be that the older women have given up on a long term mating strategy in favor of a short term one, and mating with a younger guy who's health and attractiveness signals are more easily discerned would be the better bet for at least ensuring a genetically top notch baby going into single motherhood.

    but, hey, wadda i know about the sexual desires of older women? i only date young chicks.

    ReplyDelete
  4. Here it's probably the difference between a correlation at the level of individuals and at the level of groups. These two can go in opposite directions. Let's say you looked at height and suicide rate in Europe. Probably you find that Scandinavians have higher suicide rates and are taller than Mediterraneans, at the group level.

    But at the individual level -- looking among Swedes -- it's probably the shorter Swedes who kill themselves more than the tall Swedes, and the same among Italians.

    So, it could be that individual women do prefer increasingly older guys as they age. But then some force operating at the level of societies causes some to prefer smaller age diffs than in some other society. That would make the older samples prefer closer age diffs.

    ReplyDelete
  5. Do you think big-butt wilder types would be more into older guys in general? Maybe they look for more of a dominating presence?

    ReplyDelete
  6. Patrick Bateman8/19/08, 10:09 AM

    Among American girls, I've noticed that country girls are far more willing to date much older men than city girls are. They also dig the scruff.

    ReplyDelete
  7. I visited Taiwan for a week and was surprised to see so many older men (by 20-30 years) with younger women.

    My Taiwanese friend said that her sister was an airline stewardess and her co-workers told her to grab a businessman about 10 years older than her.

    She also said that older guys have more money to spend on a girl and have more free time, since they are already established in their careers.

    So bring your money if you're going to Taiwan or China...

    ReplyDelete
  8. I read this quote from Roissy's blog and it made me think of you...

    Chic, Nicole, that’s why I’m doubting how real some of these net PUAs are. People put more value into things they have a hard time getting and if you’re really getting hot young girls interested all the time, you’re just not going to be that impressed by mere youth and beauty. If, on the other hand, you’re hanging out with hot people all the time and not getting any…

    The bloggers with the biggest obsession with youth and beauty are also exactly those who most sound like they’re faking success.
    --------------------------------

    Trust me. Young girls are NOT that great. You should have partied in high school and hooked up with girls then so you wouldn't have such a fetish with them now. I just came back from studying abroad in Thailand and I dated a lot of girls and you tire VERY QUICKLY with having to deal with an 18 year old. Are they attractive? Sure, but they get old just like everybody else.

    ReplyDelete
  9. Not me -- I prefer being around young girls, not only for their looks, in case the recurring theme of girliness here hasn't made that clear. My best friend since I moved out to grad school is 18, will be 19 within a month or so.

    Now, guys that age I am repulsed by, as in I couldn't hang out with them. Too desperate and whiny, and I have too much contempt for those with negative dignity. That's why they're a perfect match for abandoned cougars.

    ReplyDelete
  10. Luckily the data also show the average age of the sample, so I refer to "preferred age difference" as the sample's average absolute preferred age difference (e.g., "3 years older") divided by the sample's average age.

    I'm not sure that's helpful. Ceteris paribus, is a woman who at 20 prefered a man three years older than her likely to shift his preferred age advantage when she's 23? If not, the method decreases the preferred age difference percentage 2% (from 3/20=15% to 3/23=13%) for no reason, translating to a four or five spot change on the comparative rankings of the countries. Anyway, the inverse correlation between PPP and desired age difference among females is stronger if the absolute difference in years is used rather than the percentage method (the case is the same for average IQ, which I looked at for an upcoming post). But you've presented them both, so obviously readers can come to their own conclusions.

    ReplyDelete
  11. Oh, friend. You are kinda creepy.

    --Signed a city girl who is currently digging on a dirty guy 12 years older than her

    ReplyDelete
  12. Heh, guys aren't susceptible to peer pressure / shaming like girls are, so don't even try. :) And btw, girls don't approach creepy guys en masse with naughty looks on their faces.

    Ceteris paribus, is a woman who at 20 prefered a man three years older than her likely to shift his preferred age advantage when she's 23?

    I thought about using the absolute number, but the idea is that she ought to be increasing this absolute number as she gets older. If a 20 y.o. girl wants a 25 y.o. guy, that's very do-able. But if a 30 y.o. woman wants a 35 y.o. guy, why would he? He can get 27. So she should aim for 7 or 8 years older instead of 5.

    It's like the income gap she prefers -- it ought to increase as her income increases. Maybe there's some saturation level, but the ages here aren't so old.

    Remember to weight each data-point by sample size when you run the correlation -- sample sizes vary a lot here. That's could be why it came out higher when you used absolute numbers.

    ReplyDelete
  13. Why do you want to weight here, if the purpose is to get a comparison between countries? Do you want the US numbers to have 16x the weight of the Greek data? Anyway, the .63 correlation comes in at .60 unweighted--the differences are minor.

    ReplyDelete
  14. The weighted correlation is stronger using absolute year differences, too. It is almost identical to not weighting (.652 versus .654). Again, I'm not trying to be obnoxious--the relationships are similar whether the percentage method or the absolute method is used, and whether the numbers are weighted or not.

    I just don't have a good way to weight in Excel--I have to use a tedious formula, which is why I keep whining about it (I'm coming up with a post related to this one). What do you use? If I had SPSS or knew anything about VBasic I could probably come up with something, but I don't.

    ReplyDelete
  15. I want to weight since there's a wide range of sample sizes, and some of the small ones appear to hold up the correlation -- like Iran. You don't want that to come crashing down if you weight by sample size. Plus someone's going to bring it up, so you might as well.

    I actually don't have a stat software program that I regularly use, since I'm not really a data person. More of a modeling guy. You can do it in Excel, though.

    The formula for the Pearson correlation between X and Y is Cov(X,Y) / SQRT(Var(X)*Var(Y)). Starting with Cov(X,Y), this is E((X-Xbar)(Y-Ybar)), where Xbar and Ybar are the *weighted* means. For Excel, set one column = (N / Nt)*X, where N is the country's sample size, Nt is the total sample size, and X is the country's value for the variable. Drag this down through all countries, then sum up this column. That gives you the weighted mean for X, and ditto for Y.

    Do the same for Cov(X,Y). Set a column = (N / Nt)*(X-Xbar)*(Y-Ybar), where Xbar and Ybar are whatever you found in the weighted mean step. Drag it through all countries, sum up this column, and then you've got it.

    Var(X) = E(X-Xbar)^2, so set a column = (N / Nt)*(X-Xbar)^2, where Xbar you found in the weighted mean step. Drag it down, sum it up, and bingo. Do the same for Var(Y).

    Now you have the weighted covariance, and the weighted variances, so you plug these into the original formula for the Pearson correlation.

    Another thing to remember is calculating the p-value. If find correlations between lots of things, you'll find one that's significant at the 0.05 level 1 in 20 times by chance alone. In my post, the correlation between age and preferred diff is not significant by the usual standard, even though the scatter-plot and moderate correlation make it look convincing.

    Pearson's r has a t-distribution, so find an online calculator for this. The df = N-2, where N = number of data-points, in this case number of countries, not the total sample size. The t-value is:

    r * SQRT((N-2)/(1-r^2))

    Then see if it holds up under a two-tailed test.

    ReplyDelete
  16. Thanks. I'll give it a shot and see if I can keep up. I've been using, as an array formula in Excel:

    {=(SUM(CX:CY*(AX:AY-SUMPRODUCT(CX:CY*AX:AY)/SUM(CX:CY))*(BX:BY-SUMPRODUCT(CX:CY*BX:BY)/SUM(CX:CY))))/(SQRT(SUM(CX:CY*(AX:AY-SUMPRODUCT(CX:CY*AX:AY)/SUM(CX:CY))^2)*SUM(CX:CY*(BX:BY-SUMPRODUCT(CX:CY*BX:BY)/SUM(CX:CY))^2)))}

    where X is the beginning and Y the end of the range of values. The C column is the weight (N from specific samples) and the A and B columns are actual data. That just gives me a correlation though.

    ReplyDelete
  17. Greece is most certainly NOT a first world country. Athens is a dump and feels barely better than Albania. That being said, Greek girls are seriously cute. So are girls in Istanbul, and Istanbul is a much cleaner, nicer city. Don't know if Turkish girls put out though. I get the feeling many do.

    ReplyDelete

You MUST enter a nickname with the "Name/URL" option if you're not signed in. We can't follow who is saying what if everyone is "Anonymous."