Peter Orszag, drawing on a new study by economists Martha Bailey and Susan Dynarski of the University of Michigan, observes that college completion rates for people raised in low-income and high-income households have diverged in recent decades:
Looking at children born in the early 1960s, the researchers found that only 5 percent of children from families in the lowest-income quartile completed college, while 36 percent of those from families in the highest-income quartile did.
For children born around 1980, the college completion rate among low-income students rose to 9 percent, but among high- income students it jumped to more than half (54 percent). In other words, over two decades, the college income gap widened to 45 percentage points from 31 percentage points. This widening was observed even after the researchers accounted for differences in students’ cognitive skills.
Yet as Orszag goes on to write, Bailey and Dynarski find that gender differences appear to play an important role:
The gap between rich and poor in both college entry and college completion widened by almost twice as much for women as it did for men. (An astonishing 85 percent of girls born in well-off families around 1980 entered college.)
And so it seems unlikely that affluent parents are simply “buying” higher college completion rates for their children, as that would fail to account for this yawning gender gap. Orszag highlights sharp differences across income groups in high school graduation rates and in completion rates once students are enrolled in college:
Not surprisingly, but somewhat depressingly, those who don’t finish are disproportionately poor. Among those born around 1980, only about a third of college students from low-income families got their degrees, compared with about two-thirds of those from affluent families.
Recently, economist Nathaniel Hilger released a job market paper that explores the impact of household income on college attendance:
It is well-known that parental income strongly predicts children’s college attainment. However, there remains debate over whether this relationship is driven by parental income or by other factors, and how impacts of parental income vary across stages of childhood. I develop a new research design to estimate the causal effects of parental income during late childhood on children’s college outcomes using administrative data on the U.S. population. The design compares outcomes of children whose fathers lose jobs before college decisions with outcomes of children whose fathers lose jobs after college decisions. I find that an unanticipated $1,000 decrease in permanent income due to a father’s layoffs reduces children’s enrollment by 0.18%. This impact is precisely estimated and smaller than estimates in prior work that rely on variation in firm closures rather than timing of layoffs. I replicate these larger estimates and show they are driven by selection of workers into closing firms. Causal effects of income during late childhood account for 10-15% of the cross-sectional correlation between income and enrollment. Income losses have even smaller impacts on the lowest-income children, consistent with the fact that these children rely less heavily on parental income to finance college. These findings suggest that relaxing parental liquidity constraints during late childhood will do little to increase enrollment compared to improvements in financial aid, especially for low-income children.
(1) Addressing high male dropout rates, among boys raised in low-income but also in middle-income households, ought to be a high priority. Orszag recommends gender-neutral policy levers, like requiring that young people remain in school until age 18, yet schools might also pursue non-traditional instructional models designed to make persistence more attractive to boys.
(2) Hilger focuses on the relative value of financial aid vs. cash grants to low-income households, but the underlying issue is the cost of higher education. The higher the cost, the more challenging it will be for the public sector to finance subsidies designed to make higher education affordable for low-income households. Even as public subsidies for tuition have declined at the state level, costs have increased and the burden has increasingly been shifted from taxpayers to students and their families. Leveraging federal higher education resources to encourage productivity-enhancing business model innovation, along the lines of the new “Rebalancing Resources and Incentives in Federal Student Aid” report from the New America Foundation, is a strategy that will both alleviate the burden on taxpayers while broadening access to high-quality instruction.
UPDATE: A friend writes in with a potent objection to my argument regarding the cost of higher education:
I think this aspect of your argument is understandable but completely mistaken. As we’ve discussed before, generally speaking, the most expensive schools have the lowest attrition rates. Top private schools have 6-year completion rates of about 95% whereas pretty good state schools (UCLA, Wisconsin, etc.) are at about 65%. You go down to community colleges and the rates are abysmal. Some of this is selection but a lot of it is treatment. Compare the omnipresent wait lists for required courses at a school like UCLA to the mollycoddling house masters at a school like Princeton and you can see that it’s something about the expensive labor-intensive nature of the latter that causes the better completion rate, and not just only that kids from wealthy families get in. The long term trend in undergraduate education has been from an r selection strategy of a residential liberal arts college with hands-on faculty teaching seminars that is extremely expensive and demands the full-time attention of students to various forms of cheaper and cheaper provisioning of education. “Business model innovation” (read: MOOCs) in higher education would be the logical extension in the shift from r-selection to K-selection. Most MOOC courses have extremely low completion rates, something like 20%, and that’s per course, not for an entire degree curriculum.
We might very well decide that for all sorts of reasons it’s well worth it to reduce our use of a high cost and inflexible but relatively high yield model in order to open up space for a low cost and flexible but extremely low yield model. However I am extremely skeptical that increasing the completion rate is a valid reason to do so.
It’s a very good point, yet the implications aren’t entirely clear — mollycoddling might improve college completion rates at great expense, but what will happen to graduates once the mollycoddling is gone? My correspondent recognizes that New America (and I) aren’t calling for MOOCification but rather for gradual reforms of student loans, etc. While this could improve matters at the margin, it won’t change the fundamental fact that the combination of Baumol’s disease (labor-intensive mollycoddling gets more and more expensive over time) and the difficulties involved in educating the marginal matriculant (many of whom have had inadequate K-12 preparation and don’t have the supportive social networks that greatly aid persistence in college).