Sunday, May 17, 2015

Crime doesn't pay (procreative dividends)

Pat Buchanan famously wrote that "historians may one day call 'the pill' the suicide tablet of the West". Looking at the sheer numbers, he seems to be on the mark.

I'm not much of an optimist, but I do have a detached sunny disposition, so perhaps we can find a silver lining. The GSS has queried respondents several times, most recently in 2012, on whether or not they have ever been arrested. Male privilege helps get a person cuffed, with 20.4% of adult men having ever been arrested compared to only 5.7% of women. Consequently, we'll be looking at men here, and to allow ample time for procreating to have occurred, we'll be looking at those aged 40-65. The trends subsequently discussed hold across racial lines.

As Heartiste has argued in one hundred different ways, chicks love badboys, and there is hardly a better indicator of badboy status than having had a run-in (or multiple run-ins) with the law. The GSS affirms as much. Among the one-in-five men who've been arrested, 22.1% report having sex no more than once a month and 57.3% report having sex at least once a week. Among the four-in-five men who've never been arrested, those figures are 29.7% and 43.5%, respectively. This sex frequency advantage among those who have been arrested exists despite the fact that arrestees are less likely to be married than men who have never been arrested are (68.6% to 84.7%). Jokes about marriage being a vow of celibacy aside, married people tend to have significantly* more sex than unmarried people do.

Prior to the onset of easily accessible modern contraceptives, this would've translated into badboys outbreeding the rest of the male population. Fortunately, the edge in sexual access doesn't translate into a realized procreative advantage. Men who have been arrested average 2.71 children. Men who haven't average a nearly identical 2.63.

GSS variables used: SEXFREQ(0-2)(4-6), AGE(40-65), SEX(1), MARITAL(1,3-5), CHILDS, ARREST(1-2)

* Controlling for age, mean sex frequency among married adults is about one standard deviation higher than it is among unmarried adults


Anonymous said...

Male fertility is notoriously hard to track down. The only way in which we could do it is to forcibly draw blood from every man in this world and keep a vast paternity database in which every infant and every man has their dna records kept.

That being said, there are some groups of men who do give reliable estimates of how many children they have. I trust high IQ men more than low IQ men. Asian men more than Black men (and White and Hispanic men in between). Men with no arrest records over men with arrest records.

I believe that the men who have been arrested have more kids than they are aware of. I believe that the men who have never been arrested are largely accurate when they say they have x number of children. I believe that low IQ men have more kids than they're aware of. I believe that high IQ men have exactly how many kids they say they have.

gwern said...

You could always look at population registry studies. Here's one very on point recent study:

"Criminal offending as part of an alternative reproductive strategy: Investigating evolutionary hypotheses using Swedish total population data" ; Shuyang Yao, Niklas Langstrom, Hans Temrin, Hasse

Audacious Epigone said...


Could be, although I'm not sure why to necessarily presume that NAM and low IQ self-reports, while less generally accurate, are low ball figures rather than exaggerations. Yes, there is the gag about the confusion that father's day brings with it in the ghetto, but I bet the lower classes tend to trump up number of sexual partners, number of fights, quantity of drugs and alcohol consumed, etc.


The study didn't control for race or ethnicity. Though Muslims make up roughly 10% of Sweden's total population, they comprise more than 50% of its prison population and, crucially different from the case of blacks in the US, they have a much higher TFR than native Swedes do.

I responded at length here.

gwern said...

> The study didn't control for race or ethnicity.

It did include immigration status as a covariate. And since this was a individual-level dataset rather than aggregate, I'm not sure excluding race changes the result. Is there any reason to expect something complicated like an interaction term here? (Why would the correlation of crime and reproduction be different in the ethnic Swedish population than the immigrant-descended population above and beyond the obvious point of the immigrant population having higher average levels of both variables?)

> I responded at length here.

You should probably read studies before criticizing them.

Dan said...

I think one difference is that Sweden hasn't reached 'underclass-saturation' in the way that the US has.

In the US, it is literally impossible for the state to give a good life to most of the poor underclass, because there are so many of them.

In government programs such as housing, the poor displace other poor and wait lists are years long. Programs such as disability insurance in the US can pay very little per person because the denominator is so large.

Brazil and India for instance were socialist for many years but the volume of the poor make this socialism meaningless.

Sweden still has some capacity, at least for a while, room for more in the niche of dependency.

Audacious Epigone said...


Most offences are committed between the ages of 15-25. A lot of those perpetrators are not going to be native Swedes or first-generation immigrants. That's clearly a potential confounding factor. But I'm not doing this for remuneration like the paygaters are, so take this as another data point for your meta analysis.

gwern said...

> That's clearly a potential confounding factor.

Of course it's possibly confounding, but remember what a confounding factor is. If it's independent, it's just noise and the sample size is huge enough that we don't need to care.

> But I'm not doing this for remuneration like the paygaters are

I'm not either. I still know how to use Libgen to get fulltext when it's not available.