This NYT op-ed by Ellen Ruppel Shell cites my work w/ Brad Hershbein on how college returns vary for persons who grew up in families of different incomes. IMO, it draws the wrong conclusions from our work, and omits some important findings: nytimes.com/2018/05/16/opi…
First, here is the link to our working paper on this topic, which the op-ed did not include: research.upjohn.org/up_workingpape…
The op-ed says that we found "that for Americans born into middle-class families, a college degree does appear to be a wise investment. Those in this group who earned 162% more....than those who didn't. But for those born into poverty, the results were far less impressive."
First our main analysis compared individuals who were from families below and above 185% of the poverty line, not below and above poverty. (Why 185%? That is cutoff for free or reduced price lunch in K-12, and such eligibility is used for many educational purposes)
We found career earnings return to college was 136% for those whose family background was above 185% of poverty line. But return to college was 71% for those from families below 185% of poverty line. 71% is nothing to sneeze at!
Snell probably got the 162% from preliminary results that were released 2 years ago: brookings.edu/blog/social-mo… There, we had found 91% for the lower income group, which is certainly a high return!
Our later numbers are better. The later numbers make technical improvements, including estimating career earnings using actual earnings for different ages rather than projecting earnings by an age quadratic. But both estimates show high returns from college for low-income group.
Actually, in our later working paper, the return to college is HIGHER for persons from families below poverty line: 179% (see Table 7). The big contrast is between individuals from families between 100-200% of poverty line (48%) and those above 400% (129%).
The op-ed says "The authors don't speculate as to why this is the case". But we do speculate, and we did so both in the current working paper, and in the past: brookings.edu/blog/social-mo…
In the new working paper, we find very strong signs that the differential returns is largely due to certain groups, particularly white males from upper-income backgrounds, having differential access to very high earnings jobs.
The differential returns to college with family income background occur only for men, not women, and for whites, not blacks. For blacks, for example, the return to college is 173% for those from lower-income backgrounds, vs 179% for those from non-low-inc backgrounds (Table 9)
And the differential MEAN returns to college are not evident for the median college graduate, but only occur in the upper tails of the earnings distribution, beyond the 75th percentile, & particularly beyond 95th percentile.
This picture is one of the most interesting in our paper, showing how the relative returns to college with family income background vary across percentiles of the earnings distribution. 
We say in paper intro: "Overall, we infer that the differential % return to education across family income background is largely driven by differential access to the right tail of the earnings distribution. College grads.. from high-income families..., particularly men & whites..
have a much higher chance of accessing the top of the earnings distribution than otherwise similar individuals who come from poorer families."
It is true than when we break down the results even further, which does strain the data, that returns to college for near-poor (100-200% of poverty) are not as high, particularly for men and whites. Probably due to better options for this group w/o 4-yr college
The other factor is the VERY high earnings in right tail -- jobs in finance? -- for white men from families above 400% of poverty. So the story is at least as much about increased earnings inequality in the U.S., and how we currently ration access to the earnings right tail.
Furthermore, we are looking at returns to bachelor's degree, not at AA degrees, which is another topic.
In sum, we are certainly not saying that low-income folks shouldn't go to college because it has low returns for them. That is certainly not true for many low-income groups, and should not be inferred from our work.
