Looking for Sister Right

Several people have asked me what I thought of this. For example, one colleague wrote:

I thought you might find data of this quantity and quality as interesting as I do.

In their latest blog entry, okcupid uses its very large dating-site userbase to assess the progression of social and economic views with age, and (taking into account self-specified importance of economic and social views) the resulting political affiliations. The final assessment is that the spectrum of views encompassed under “democrat” ends up being much broader than those under “republican”. I’ve seen this mentioned many times before, but I have never seen such pretty data illustrating the phenomenon. Perhaps this is more graphics to impress, rather than graphics in a typical statistical setting, but I find them captivating nonetheless. In any case, I would be curious to hear your opinions of this.

My reply:

It’s interesting stuff, although I wonder a lot about the representativeness of their sample. For example, they talk about libertarian views, which happen to be held by only a small fraction of the population (but a larger fraction of people who are online).

Along similar lines, I think their two-dimensional scheme is misleading; I’d prefer to just label the axes as left/right on economic ideology and left/right on social ideology. For example, where does government support for health insurance fit into this? I don’t see this as “economically restrictive”? Or what about support for a graduated income tax or an estate tax, vs. support for a flat tax or a sales tax? Conditional on total tax revenues being fixed, I don’t see the former as any more or less economically restrictive than the latter–it’s just restrictive to different people.

Their graphs are pretty but I’m not sure about their interpretation. They interpret a cross-sectional pattern as age effects, but they could very well be cohort effects. People in their 40s came of political age around the early eighties when Reagan was president. Not to mention the fact that people in a dating site are far from typical of the population–I’m assuming the people in their 40s are mostly either divorced or never-married; either way, that’s a different perspective than a married person.

Similarly, the trend in social attitudes has got to be, at least to some extent, a cohort effect. Although I suppose it depends on the issue. Americans have become much more liberal on gay rights but not on abortion.

Also note that, before 2006, the 18-30-year-olds were not such overwhelmingly Democratic voters.

Finally,the pretty graph that they show with red and blue is a bit misleading: it’s not a graph of Democrats and Republicans, it’s a graph of average opinions of people at different ages.

Getting to the specifics, I also don’t know why they’re so sure that nuns won’t have liberal views on gay marriage. Lots of nuns are gay, no? I once met a woman whose reason for joining an order of nuns was that, in her words, “I want to meet Sister Right.”

In any case, I agree that age is a good variable to look at. And, considering that these guys are starting from scratch on their analyses, many of the things they’re saying are reasonable.

P.S. For a deeper look at systematic asymmetries between the Democratic and Republican parties in their political support, see Jonathan Rodden and Chris Warshaw’s article, “Why the Democrats Need Boll Weevils and Blue Dogs: The Distribution of Political Preferences across U.S. House Districts.”

1 thought on “Looking for Sister Right

  1. Okcupid puts out a fair amount of "fun" but seriously flawed analysis based upon non-controlled data.

    In many cases (see the eharmony vs the world) analysis. There's a complete misstatement regarding the "probability" of getting married based upon presence of eharmony or not.

    Its an interesting approach to informing design and decisions for web-interaction, but its masquerading as statistical analysis.

Comments are closed.