Monday, December 01, 2008

The chicken or the egg?

Through her exceptional book The Nurture Assumption, Judy Harris has profoundly influenced the way I read and understand the findings of social science research. One gem was her explanation of how data can be sifted, sliced and diced to produce puzzling results, which are sometimes utterly meaningless. Another was an insight about the (dis)connection between causality and statistics.

It is by now a commonplace that children who are praised tend to be sunnier and well-behaved, whereas children who are scolded and beaten tend to be ill-mannered. It's been presented that way as a finding of social science -- there are probably dozens of citations because it seems to me I've been reading this for decades. (I won't look them up for you; I'm offline at the moment.)

But statistics can't tell you whether kids are well-behaved because they're praised, or praised because they're well-behaved. In other words, the correlation between good behavior and good treatment could just as well be explained as:
Nice children get praised; surly children get beaten.
So why did we hear the converse for so long? Well, "nice children get praised, surly children get beaten" just doesn't sound very interesting. I mean, Duh! How much would you pay a researcher to tell you that? An article with that "shocking result" would never get printed! But "praised children are nicer" -- now that's an interesting conclusion -- even if it's misleading and useless!

Here's another one. I read somewhere that happy people tend to be healthier, more successful in their careers (and thus financially) than unhappy people. Or was it that healthier, richer, more successful people tend to be happier? (The comparisons were within the same culture; it wasn't saying that Americans were happier than Ethiopians for example.) Cross-sectional data won't tell us; we need time-series data -- we need to monitor people for decades. We must find happy 25-30 year olds, and see how healthy and rich and successful they are at 45-50. Or we can just assume that the causality works one way, and declare our assumption to be the assured results of a rigorous statistical analysis!

I tend to think that happy people tend to succeed, rather than the other way round. People who tend to be unhappy don't become happy when their plan works, when they get that promotion, or when their bank account grows. And happy people, who smile and encourage others -- realistically happy people I mean, not Pollyannas -- are easier to work with than Eeyores. It's easy to believe they get more promotions and find success more often, too. But that's just a hunch; the statistics so far neither prove nor disprove my guess.

A final example: religious people give more blood -- this according to Jonathan Haidt, the honest liberal atheist and psychology professor. Do these religious people give more blood because they believe in God, or attend church or synagogue? Or do generous people (who give blood as part of a generous lifestyle) tend to believe in God and join churches and synagogues?

I tend to believe the former, because I think God has the power to change people, but honestly I don't know how much of the effect is due to each cause. It could be that generous, religious people get that way because they're more aware of how much they've received. Statistics don't say.

As an engineer, how do I deal with all this? Well, as an engineer I simply note the correlation when it's useful to me. As a human being and a Christian, though, I pray (no statistics on that) and I give. And when I read "A new study shows that..." I think a little more than I used to about whether the study might show something else.

No comments: