Man, those singularity people don’t fuck around.
I can’t find the link, but apparently intergenerational income mobility in the US is about the same as in Sweden.
A couple related papers:
It looks like the correlation measure is “how far is the father’s income from the average compared to the son’s”; I think distance is measured in standard deviations from the mean (although technically it shouldn’t matter too much what the units are). 0 is uncorrelated (the son’s income distribution has no relation at all to their father’s), 1 is perfect correlation (same position).
Using the scatterplots there as a reference, imagine the father’s position in the income distribution on the x axis and the son’s on the right. IF they’re the same (1 correlation coefficient), you see a 45% straight line. As the correlation becomes weaker, the cloud of data points “opens up” from the line, until at 0 a scatterplot of father/son income looks like random noise. This paper talks about very high correlations at the upper end of the income spectrum ( . 8 ) and very low ( 0 , for the lowest income) at the other end.
I don’t know, I just wasn’t expecting close to half correlation.
I can’t find anything talking about if the number’s changed, but apparently the old consensus used to be that it was .2; then a bunch of papers came out criticizing various methodological issues and suggested revised numbers in the .4-.6 range, and a different set in the .1-.2 range. That Canadian paper gives a .2 number for Canada.
The non-linear nature of the correlation is the most interesting thing; the Canadian paper talks about the 95th percentile having a very strong likelihood to move even farther up, and the 5th percentile having a likelihood to move even farther down.