Social policy analysis is only as good as the data on which it relies. In “Colleges that Pledged to Help Poor Families have been Doing the Opposite,” Jon Marcus and Holly K. Hacker draw conclusions based on inadequate federal data.
In their article, they claim that even the 100 higher education institutions that pledged to expand opportunities for low-income students at a White House summit held almost a year ago are actually making it harder for those students to afford college. As I show in a recent Brookings Institution working paper, though, the federal data on which they base those conclusions are seriously flawed.
I was inspired to write that paper after Jon Marcus contacted Wellesley College (one of the 100 institutions), where I am a professor of economics, inquiring why the average “net price” that we charge low income students had tripled between 2008-09 and 2012-13. The net price is the amount that a student would have to pay the college after deducting grant aid. It includes loans, work-study jobs, and out-of-pocket costs. According to data reported by the U.S. Department of Education, the average net price for low-income students at Wellesley rose from $3,533 to $10,495 over that period.
Those of us knowledgeable about financial aid at Wellesley were confused. No policy change had taken place. We knew the correct answer was that the average net price has remained roughly constant. Yet we could document that the data reported to the government was accurate. As an economist faced with a data puzzle, I began to investigate more fully.
One problem with these data is that many institutions like Wellesley use a more comprehensive measure of family income, including more detail regarding self-employment income, for instance. Some of those families defined as low income by federal standards actually have far greater resources available and are expected to pay more accordingly.
Beyond that, the results of my investigation proved an important conclusion that I regularly teach in my introductory statistics class – sometimes an average can be a misleading statistic. Averages can be affected by outliers. Calculating average income from a sample that includes Bill Gates is not going to provide a useful answer.
Our financial aid data show that the vast majority of low income students at Wellesley face a net price of around $4,000. That amount consists of a work-study job of around $2,000 and an out-of-pocket contribution of around $2,000, which we anticipate will come from a summer job. No other payment is expected – the College has not asked low-income students to take out loans since 2008. Yet a handful of students receiving financial aid pay considerably more than that – as much as $50,000. Variability in the number of such students (say, two one year, five the next) can lead to substantial variability in average net prices. We happened to have a few more of these unusual cases in 2012-13, leading to a large jump in our average net price.
One might wonder how students with family incomes under $30,000 can afford to pay upwards of $50,000 to attend Wellesley College. Note that these students are enrolled, so they must be able to afford it. Consider, for instance, a student whose parents are retired, so their income is low, but they have considerable assets. Unusual financial circumstances like this explain these patterns.
The standard “statistics 101” solution to this problem is to report the median rather than the average. In my Brookings paper, I show that the median net price at Wellesley was actually flat over this period, just as we had suspected. I can’t say for sure that this is what is going on at other institutions without inspecting their data, but it seems likely to be a problem for many of them, and certainly selective private colleges and universities with large endowments. Without credible data, it is difficult to say what is happening to the net price that a typical low-income student could expect to pay.
This statistical problem is critical for those of us who are worried about issues of college access. If low-income students are misled to believe they can’t afford college, they won’t go to college. Tracking net prices across institutions and over time can help, but not if the data are misleading. I have argued that the best solution would be to introduce simplified net price calculators that provide personalized estimates based on an applicant’s basic financial characteristics.
In the meantime, the federal government should make the minor correction of requiring higher education institutions to report median net prices by income rather than average net prices by income. Those data would be far more helpful. Anyone who cares about the issue of transparency in college pricing should support this simple step.
Phillip B. Levine is the Katherine Coman and A. Barton Hepburn Professor of Economics At Wellesley College.