What parents and students should know about college outcomes

Which matters more: the college that you go to, or the field that you major in? The median starting salary of a graduate from Harvard is much higher than that of a graduate from Ohio State. And yet, the starting salary of an Ohio State engineering major easily beats that of a Harvard psychology major.

This post explores a fairly new data on the earnings of undergraduates by school and by major. The data comes from the College Scorecard, compiled by the Department of Education. It has information on earnings for students who received federal financial aid. Since about 70% of full-time students receive federal financial aid, this data is pretty representative overall. However, it is important to keep in mind that the percentage of students receiving federal aid varies across schools, and thus the data may be more representative for some schools than for others. Specifically, wealthy schools tend to replace loans with institutional grants lowering the percentage of students receiving federal aid. For the students that did receive federal aid the earnings data is based on linked individual IRS records and thus likely to be quite accurate.

The latest data combines students who graduated in 2017 and 2018. For each school and the field of study, the Scorecard reports the number of graduates working, those not working (and not enrolled in school), median earnings and various kinds of educational debt. While the Scorecard includes information on people with graduate degrees, I focus only on people who completed their bachelors degree.

The Scorecard currently reports median earnings for periods of the first two years after graduation. I focus on earnings one year after. Thus, the data must be interpreted as starting salaries of fresh college graduates. The Scorecard plans to expand the earnings information beyond the two years after completion, but it will take several years until we have Scorecard data on mid-career earnings.

Earnings by Field of Study

The Scorecard reports fields of study at the four-digit CIP Codes, which include over 300 different fields. To facilitate comparisons across a broader set of disciplines, I categorize fields into the ten broad disciplines used by the NSF in their Survey of Earned Doctorates.

The chart below shows the earnings in the 300 detailed fields of study grouped by the ten broad disciplines. For each field I averaged the median earnings reported by each school, weighting each school by the number of working graduates in that field. The size of each mark in the chart is proportional to the number of working graduates in each field. Students in the engineering and computer science fields have the highest median starting salaries, while those studying physical sciences and the humanities have the lowest. However, there is a great deal of variation within each discipline. For example, among health professions, the registered nursing majors earn almost double that of public health majors. Similarly, within the discipline of social sciences, economics majors do much better than anthropology majors.

It is worth emphasizing that earnings differences one year after college may not accurately reflect long-term outcomes. For example, disciplines such as the physical sciences require a graduate degree to reach the full earnings potential in those disciplines. The graduates we are observing are those who are not pursuing those graduate degrees straight out of college. They may pursue those degrees later, or if they don’t, they may be less skilled than those who do. In contrast, in disciplines such as engineering or registered nursing, the bachelor’s degree is generally considered a terminal degree. As a result, the earnings we observe in these disciplines are likely to be representative of the graduates’ outcomes.

Unemployment by the Field of Study

The Scorecard data includes information on the number of graduates that are working vs those that are not working and not enrolled in school. While not working and not enrolled may be a choice, it may also be an indicator of opportunities available to graduates. The higher the share of those not working could be a proxy for the unemployment rate faced by the fresh graduates. The graph below shows that share plotted against the median salary in each field of study that has at least five thousand working people. It is clear that there is a variation in the share of not working graduates. It ranges from about eight percent for international relations majors to one percent for registered nursing majors. It is also clear that a higher percentage of not working graduates is associated with lower earnings. Supply and demand are probably at work here: the excess supply of graduates in certain fields leads lower wages than in fields where the supply, relative to demand, is lower.

Earnings by the Field of Study and Institution

While the Scorecard data includes information on associate and other two-year institutions, I focus on four-year colleges and universities based in the U.S. This includes nearly 1,800 institutions. I use the Carnegie Classification to split them into four broad categories of doctoral universities, master’s colleges, baccalaureate colleges and special focus institutions. The chart below shows the median earnings by institution category and by field of study. It is clear that the variation between fields is much greater than the variation between different types of institutions. Psychology majors earn pretty much the same whether they attend a doctoral university or a liberal arts college.

To further explore variation between colleges, the chart below show the differences in earning by US News rank range for national universities and nationally ranked liberal arts colleges. Again, we see that the variation between fields, particularly among top ranked institutions, is much greater than any variation between institutions.

The chart below further summarizes that the vast majority of variation in starting salaries is driven by field of study rather than school. The chart shows all 29,773 combinations of school and field of study. Each row of circles represents one of 1,740 schools. The size of each circle and the vertical space given to each row is proportional to the total number of graduates. Thus, the size and distribution of circles corresponds to the actual distribution of earnings. We see distinct clouds of humanities and physical sciences on the low end of earnings, business fields in the middle, and engineering and computer science on the high end of earnings. The bottom part of the chart summarizes variation between schools versus variation between fields. The first box plot shows the distribution of earnings if students from the same school earned the same salary, regardless of field. The interquartile range is about eight thousand. The second box plot shows the distribution of earnings if students in the same field earned the same salary, regardless of school. The interquartile range is about twenty thousand. Using regression analysis I find that the 316 fields explain a 75 percent of variation in earnings, while 1,740 schools explain only 28 percent.

Listing of Schools

For those interested in finding specific institutions and how they compare to similar institutions, I created an interactive visualization that lists schools by average earnings, showing median earnings in each field. Numerous filters allow users to focus on schools of a certain type, rank or geographic location. The tool-tip in the visualization provides additional information about the field and the institution. Below is a static screenshot of the visualization.

Conclusion

Starting salary is certainly not the only metric when evaluating the college experience. However, by this metric, college has a much smaller effect than major. Perhaps students should worry less about where they study, and more about what they study.