correlation o' the day

new from oktrends, 10 charts showing random significant correlations about sex, drawn from their user data base. the last one gives me an excuse to re-blog it!

it compares per capita GDP of a nation to the tendency of okcupid users in that nation to indicate that they are looking for casual sex. the fine folks there found that:

"...money seems to be a more powerful influence on sex drive than culture or even religion. You have, for example, Portugal, Oman, Slovenia, and Taiwan within a few pixels of each other on the right side of the graph, and Syria, Sri Lanka, and Guatemala almost stacked on the left, and all of them sit along the trend line."

you can click through the link above to toggle over the data points to see individual country data, but here's an image of the graph by region:

right. so i think this is an interesting jumping off point, if not a satisfying final conclusion. several things occur to me, and they mostly have to do with how we're defining our independent and dependent variables: prosperity an sexual proclivity.

Prosperity as independent variable measured by per capita GDP

1) Is per capita GDP here nominal or PPP-adjusted? is it what we should be using as a measure of prosperity? in my limited understanding, median income may be preferable, since it eliminates outliers in countries with large wealth gaps, where a few very folks may monopolize a lot of wealth (like the US!)

2) what other, more specific economic variables or variables associated with prosperity did the okT folks try? literacy? income inequality? family size? i wonder if any of these would produce a closer relationship...

3) what about non-economic variables? the okT folks also have access to data about strength of religious conviction which i suspect would correlate cross-nationally. there must be a way to compare this to other data sets about average religiosity in a given country. which brings us to the second set of issues...

"National" sexual proclivity as a dependent variable

...as measured by okCupid users in a given country, a self-selected population

and there we have the crux of the issue. okT doesn't try to obscure this fact, but i think it's something even more interesting to investigate, and something that quantitative analysis will only get you so far on [resists urge to climb atop methodologies soapbox].

basically, the most pressing issue here is that, for many of the countries shown, the dependent and independent variables are describing two significantly different populations. in a countries like the US or australia, with a high number of internet users and a (larger) social acceptance of internet dating, i think it's fair to say that those individuals willing to get on the internet and declare that they're looking for a random shag may be, if still self-selecting, to some degree representative of a larger population.

but think of other countries, especially developing world countries, where both populations, internet users and internet daters, are smaller. it seems to me, the self-selection bias gets much much stronger in those circumstances. now, this is surely still tied to some measure of economic prosperity, but i think it also implies that, in a lot of these places, we might be looking at a lot of different things: changing social trends and cultural mores, how early adopters of internet technologies are using the unique public/private space online to buck these mores, or again, push change. i small a comparative case study coming on!

anway, this is just off the top of m head this morning. what do other people think? 

i've asked the nice people at okT for their data, although i don't know if they can share it, so maybe we can test some of this stuff later.