Elisa Jácome, Ilyana Kuziemko, Suresh Naidu write:
We estimate long-run trends in intergenerational relative mobility for representative samples of the U.S.-born population. Harmonizing all surveys that include father’s occupation and own family income, we develop a mobility measure that allows for the inclusion of non-whites and women for the 1910s–1970s birth cohorts. We show that mobility increases between the 1910s and 1940s cohorts and that the decline of Black-white income gaps explains about half of this rise. We also find that excluding Black Americans, particularly women, considerably overstates the level of mobility for twentieth-century birth cohorts while simultaneously understating its increase between the 1910s and 1940s.
This is an interesting paper both for its substantive content and for its use of data. The authors also prepared a teaching supplement that walks through the analysis in detail. Good job!
P.S. For both of the displays above, I think a grid of small plots would work better. There’s this problem where researchers seem to think they need to cram as much as possible into one graph, but they you end up with all these symbols and colors, and the reader needs to go back and forth between the graph and the legend . . . it’s kind of a mess. On the plus side, it’s good to see any scatterplots at all in an empirical paper!
A typo:
“but they you end up”
should be “but then you end up”
I may be imagining this, but the above typo
“they you” instead of “then you”
seems to happen often in this blog. It is as if the writer is racing ahead, thinking of the next word, “you,” and thus, the substitution of the letter “y” for the letter “n” in the preceding word. If so, is there a name for this sort of typo? If not, should there be one? Or, could it be the software built in and is guessing at the next letter?
I like the second graph all together because you can do the four pairwise comparisons of interest in your head.
The usual method to estimate a conditional slope with intercept shift to show the interaction (i.e., interaction plus lower order terms) is great, but because of a recent post on this blog, I wonder how a hinge function with intercept shift would affect things?
Interesting. I skimmed the study. Applying complex modelling to study social mobility and the fluidity of social structures used to be a popular topic in sociology. I don’t know what the applicability would be. It’s social history to me.
These economists are more rigorous. The study also validates that sociologists draw on economics but not vice versa. Bill Bielby’s, (an electrical engineer with a Ph.D in sociology) paper on response error is cited.
Social mobility was the topic of a well known example from the classic text “Finite Markov chains” by Kemeny and Snell (and several later revisions by Kemeny et al).