An Analytic Test

Back when I was a supervisor of analysts (last millennium, when dinosaurs roamed the National Mall), I had a weathered, paper copy of a Washington Post article in my desk drawer. The article was a summary of data developed about income percentiles in America (who makes how much, from the 1990s). There was little dispute about the data itself, and the article appeared with little fanfare and quickly passed into oblivion except for my faded copy. Sometimes when I wanted to test a potential (or aspiring) analyst, I would pass them the chart and ask them what were the strongest analytic judgments they could draw from it. I made it clear “just use this data, don’t try to add to it or fight it.”

The Brookings Institute has kindly updated that data (here), and the Washington Post covered it again. Here’s the key chart:

If you want to go back to the commentary or data, please do so, but go ahead and take this test (mentally): what judgments can you draw from this data? Again, don’t fight it; the answer is staring at you!

Middling analysts would focus on the point that the middle class is disappearing. Well, the median segment did drop from 47% down to 36%. Yet all three “middle class” segments went from 84% to 85%. Some analysts wanted to argue inequality was growing since the rich segment rose from 0% to 2%; while inequality may be growing (Brookings suggests it is), it is not evident from just this chart. The lowest two segments fell from 47% down to 29%. The bottom three fell from 94% to 65%. The best analysts honed in on the most dramatic change in any single group: the 450% rise in the upper middle class, from 6% up to 33%. So the most telling analytic line is: more Americans moved from the poor, lower, or middle class into the upper middle class.

The data can tell no other story. Why? Simply put, for every person who fell from a higher category, more than one person had to rise, in order for the numbers to hold up.

Three additional points. First, this data is longitudinal, that is, it covers the same people over an extended period of time. This eliminates the possibility the changes in outcome resulted from different people in the study at different times. Second, it does not include government transfers such as welfare, family assistance, etc., so it underreports the actual level of income for the poorest segment. Also, other charts displaying the data in the same study point out that the number of people moving from one income group to another increased (both up and down), meaning these groups are not static. While movement down the income chart grew more, movement up by more than one level also increased. Which is a long-winded way of saying there is still considerable movement between the income groups, and more variability with smaller numbers of big winners.

This does not mean there are not people who have suffered economically over this same time period: by sex, race, ethnicity, undereducation, technology change, and a myriad other reasons. But for every such case, the overall data still improved, which is quite remarkable. Brookings made much of the loss of the middle/middle class, the growth (as they see it) in inequality, and the increased number of people moving “down” the spectrum. But the overall movement into the upper middle class is just as telling. And that’s a little good news.