Earnings Premium by Institutional Selectivity
Data from Payscale.com shows that graduates of the most selective private colleges tend to get slightly greater lifetime income than the most selective public colleges. This is partly due to a higher concentration of students in science, technology, engineering and mathematics (STEM), since STEM graduates earn higher salaries, and partly due to the aggregation of students who will be successful no matter where they enroll. But, when one calculates the rate of return on investment, the most selective public colleges rank at the top, because of much lower cost.
Nevertheless, the release of College Scorecard data has contributed to a proliferation of top ten lists of the colleges with the highest average earnings for Bachelor’s degree recipients. The College Scorecard data presents average post-graduation income for the institution as a whole and does not provide major-specific data.
A recent paper by Dirk Witteveen and Paul A. Attewell of the CUNY Graduate Center, The Earnings Payoff from Attending a Selective College, finds that Bachelor’s degree recipients from very selective colleges earn 8 percent more than graduates from moderately selective colleges (the next tier down in selectivity), 11 percent more than minimally selective colleges and 19 percent more than open admission colleges. The study considered earnings 10 years after graduation, and controlled for gender, age, race, admissions test scores, academic major, college GPA, parent income, parent education, institutional control and whether the student pursued a more advanced degree after graduation. Thus, there is an earnings advantage to attending a very selective college, but it is much smaller than most people believe. Also, there is significant differentiation in earnings by academic major, so the choice of major has a greater impact on earnings after graduation than institutional selectivity. This partially explains why some STEM-only colleges have greater average earnings in the College Scorecard data than liberal arts institutions.
Dirk Witteveen and Paul A. Attewell have also published an interesting article about college success, The College Completion Puzzle: A Hidden Markov Model Approach. This paper uses Hidden Markov Models, a statistical method that is often used in computer speech recognition algorithms, to predict whether a student will graduate or not based on only a few semesters of transcript data. They found that students who alternate semesters of challenging and easier courses are more likely to graduate. They also found that students who drop out are more likely to drop classes, especially during semesters with more challenging classes. This research demonstrates the impact of class selection on college completion.