Paul Krugman has an interesting blog post today about income distributions, and why the rich tend to feel poorer than they used to. He noted in passing that he had always been taught that income distributions were log-normal for most incomes, switching over to Pareto for large incomes.
I dug in to this dataset and found an interesting follow-up tidbit. His statement about distributions seems to be correct. What’s interesting is that the cross-over point between the two seems to be happening at roughly the 75th percentile, at an income of $87.5k. Apparently incomes above this threshold behave like “large incomes,” with a power-law scaling that suggests a self-similarity of the economies of the rich to the very rich to the really very rich; whereas incomes below this threshold are following a completely different model, bunching up more towards middle values.
Here’s a graph of this income data on log-log axes, to make it clear. The Y-axis is the (decimal) log of household income, in dollars; the X-axis is the (decimal) log of the percent of society, so +2 corresponds to the poorest members of society, 1 to the top 10%, 0 to the top 1%, etc. Note that the curve is linear on its left-hand side (i.e., a power-law curve, a Pareto distribution with α = 0.56) and curves down sharply at a log-percentile of 1.39, i.e. at the 75th percentile of the distribution. The original data is available here.
This suggests that incomes in these two ranges are driven by fundamentally different dynamics, which shouldn’t be surprising. What I think is surprising is that, once upon a time, the “two dynamics” were those of earned income versus investment income, but this cutoff point seems far too low for that to be the case. Instead I suspect that we’re seeing a switch between a “professional class” — which includes much of our modern super-rich [cf. Christina Freeland’s recent article in The Atlantic, and particularly her note that today’s rich tend to be rich from earned rather than inherited or purely passive income] — and a “working class.”
The boundary isn’t the old blue/white-collar boundary; many white-collar jobs are clearly on the bottom side of the hook. (And NB that jobs can pay more than $87.5k and still be logically on the bottom end of the hook; first, actual salaries include things like regional variation, which this chart averages over nationally; and second, a job could represent e.g. the upper end of a job ladder which has a log-normal distribution, and thus be a bit above the threshold. The distribution we’re seeing here is a sum of two distributions, Pareto for jobs which pay higher on the average and log-normal for ones which pay lower)
It may be very interesting to characterize the growing income gap in our society by trying to characterize individual job ladders (i.e., sets of positions which are roughly equivalent, which individuals are expected to move across over the course of their career — although not necessarily at a single company) by their geographically-normalized income distribution. I would bet that if we compared the income distributions for a few hundred job ladders, we would find that they tended to fall into two pretty clear buckets, and that looking at the qualitative characteristics of those two buckets would tell us a lot about who is actually winning and losing in this new world. I’d bet that some of the results would be surprising, especially for jobs close to the edge — e.g., traditionally white-collar jobs now in the low bucket, or blue-collar ones in the high bucket.