New research by Raj Chetty et al on the "fading American dream"

Raj Chetty et al have just posted important new research (also summarized here) concerning what they call the "fading American dream."  Here is what I wrote about this research,  but was asked to keep offline until its public dissemination date, when I saw it discussed informally a few months back:

The research examines historical data pertaining to the following issue.  Suppose we defined the "American Dream" as positing that in each generation the kids should do better than their parents.  One might imagine the parents wanting this, and also the kids measuring how well they are doing in life by this metric, since it would capture the difference between where they started out and where they've arrived.

Chetty and his coauthors take a look at this, using a wealth of "big data," for U.S. cohorts over the twentieth century and through the present.  They look at absolute, not relative, material wellbeing.  And they define that in terms of income at age thirty.  So the last cohort they look at compares people born in 1980 to their parents, based on how the former were doing in 2010.

An interesting thing about this set-up - because it focuses purely on absolute, not relative, attainment, it could come out at 100% in a society that featured no mobility whatsoever, if we think of mobility as meaning that some multi-generational households rise while others fall.  More on that in a moment.

They find that there has been a great reduction, in recent decades, in achievement of the American Dream as thus defined.  Given how people may tend to benchmark themselves versus their parents, it surely helps to explain the sour mood (to put it mildly) in U.S. politics these days.

Chetty et al also examine the question: What sorts of changes in economic performance would cause the "American Dream" measure of upward movement in absolute terms to start looking better? Suppose we had a policy tradeoff between (a) greater absolute growth and (b) less upwardly-skewed wealth distribution? Using reasonable parameters, would focusing more on growth, or more on distribution, have a more favorable effect with regard to this measure?

If ever the set-up to a research question seemed almost pre-selected to weigh in favor of maximizing growth, rather than taking distribution into account, it is this one.  After all, with zero growth one couldn't possibly have net upward movement from mere relative shifts.  And, with high overall growth, one would think it possible to get very high levels of universal gain even if the rank order were as rigidly fixed as in a feudal society.

But they come up with a surprising answer.  If you start with the level and pattern of GDP growth over the last three or four decades, and have a choice between either (a) ratcheting up the growth a bit, but with the same distributional pattern, and (b) evening out the distribution of gains, it turns out that (b) has significantly more favorable effects than (a) on the percentage of kids who end up out-stripping their parents.  This results from the fact, that in recent decades, nearly all of the real growth has been concentrated at the very top, with everyone else stagnating.