Based on a sample of ten of the country's 21 public universities, this article examines the relative efficiency of South African universities between 1994 and 1997 using Data Envelopment Analysis. After outlining the nature and limitations of the technique, a series of seven models were tested. Each used a consolidation of the annual output of graduates and research as the output variable and tested this against various input variables. A high degree of consistency and stability was found. Differences in efficiency during the four year period were studied and the article concludes with a discussion of four factors which appear central in explaining differences in efficiency between universities.