The relationship between the costs of central administration and academic teaching and research production was studied, based on a large sample of British universities and a separate analysis of the Free University of Brussels, Belgium. To determine variables important to central administration costs, a cross-sectional and intertemporal regression analysis was undertaken with data from the British universities. The fixed-variable cost method was used to estimate cost functions for the Free University of Brussels. For the British universities, 1969-79 through 1976-77 data were used, while 1978 figures were used for the Belgian university. It was found that resources needed for the central administration program in a university were related to student numbers in an indirect way. A significant part of total costs was fixed and did not depend on any activity parameter. These fixed costs consist mainly of salary expenditures for the leadership function of the different administration programs. Other variables than student number influenced the central administration requirements. This was true for the granted research activities and for the mix of fields of studies in a university. Granted research activities involved higher administration expenditures than teaching and teaching-related research activities. Experimental academic activities (represented by number of students or academic staff) required a more expensive central administration than did nonexperimental activities. Finally, it was found that marginal costs with respect to internal activity parameters were nonconstant. Results were relevant for universities with a size below 11,000 students. (SW)