Rates of tuition increases in both private and public higher education that continually exceed inflation, coupled with the fact that the United States no longer leads the world in terms of the fraction of young adults who have college degrees, have focused attention on why costs keep increasing in higher education and what categories of higher education expenditures have been growing the most rapidly. In a series of publications, the Delta Cost Project has shown that during the last two decades median instructional spending per full-time equivalent (FTE) student in both public and private 4-year colleges and universities in the United States grew at a slower rate than median expenditures per FTE student in many other categories of expenditures (research, public service, academic support, student services, and scholarships and fellowships). Similarly, the Center for College Affordability and Productivity reports that during the same time period, managerial and support/service staff at colleges and universities grew relative to faculty. Do such changes reflect increased inefficiency and waste or do some non instructional categories of employees and expenditures contribute directly to the educational mission of American colleges and universities? In this paper, the authors use institutional level panel data and an educational production function approach to estimate whether various non instructional categories of expenditures directly influence graduation and persistence rates of undergraduate students in American colleges and universities. They find, not surprisingly, that the answer is several of these expenditure categories do influence students' educational outcome, but that the extent that they matter varies with the socioeconomic backgrounds and the average test scores of the students attending the institutions. Appended are: (1) Graduation Rate Equations Using Shares of Expenditures (Subsamples); and (2) Allocative Differences in Expenditure Ratios. (Contains 2 figures, 8 tables and 29 footnotes.) [Additional support for this research was provided by the TIAA-CREF Institute. For a related report, see ED506277.]