We estimate the labor market return to an MBA, a JD, and master's in engineering, nursing, education, psychology and social work, and thirteen other graduate degrees. To control for heterogeneity in preferences and ability, we use fixed effects for combinations of field-specific undergraduate and graduate degrees obtained by the last time we observe an individual. Basically, we compare earnings before the graduate degree to earnings after the degree. We find large differences across graduate fields in earnings effects, and more moderate differences in internal rates of return that account for program length and tuition. The returns often depend on the undergraduate major. The contribution of occupational upgrading to the earnings gain varies across degrees. Finally, simple regression-based estimates of returns to graduate fields are often highly misleading.