Jacob Felson writes,
In a 2002 article in the Journal of Economic Perspectives, economists Samuel Bowles and Herbert Gintis attempt to decompose the correlation in income between parents and children into genetic and environmental components. In addition, the authors attempt to calculate the extent to which the intergenerational correlation in income is due to inherited IQ. But unlike most data analyses, they do not use any particular data. Instead, the authors collect estimates of effect sizes from a variety of meta-analyses and attempt to estimate ballpark estimates for parameters in a theoretical path analysis. I am wondering — is this “meta-meta-analysis” appropriate? Can one conduct path analyses and evaluate relative magnitudes of correlations taken from a variety of studies?
My reply: I haven’t read the paper, but I’m skeptical of path analysis in general, and the meta-meta-analysis setting doesn’t help any!
I'd be especially wary of correlations of environmental/genetic effects, as they are determined to a large extent by the amount of variance in the environment. When everyone gets enough food, environmental effects on height reduce to zero, when there is variance in the amount of food people get, the variance due to environmental effects increase.
Hence the proportion of variance due to genetics decreases.
I imagine that there might be similar issues in, for example, variance in access to quality education.
That said, I've not read the paper either.
Jeremy