Construction of a theoretical measure of educational output is attempted in this paper. The effort begins by specifying the purpose for which this measure is intended: centralized allocation, according to the equimarginal principle, of limited resources to alternative units producing education. It then specifies a simple and narrow goal that education is to meet: satisfaction of an individual student's preferences for market good, on-the-job consumption, and the direct benefits of education. Educational output is defined as a compensating variation, or lump-sum dollar amount. The measure depends upon the particular set of preferences chosen, and as a result there will, in general, be as many distinct and equally correct measures as there are students. Consideration of the preferences of faculty and other relevant individuals will only add to the multiplicity of measures. It is concluded that centralized allocation of resources to alternative education units according to the equimarginal principle cannot be accomplished if the preferences of each individual student are to be taken into account. (Author/MSE)