As student numbers in the UK's higher education sector have expanded substantially during the last 15 years, it has become increasingly important for government to understand the structure of costs in higher education, thus allowing it to evaluate the potential for expansion and associated cost implications. This study applies Data Envelopment Analysis (DEA) to higher education institutions (HEIs) in England in the period 2000/01-2002/03 to assess the cost structure and the performance of various HEI groups. The paper continues and complements an earlier study by Johnes, Johnes and Thanassoulis (forthcoming), who used parametric regression methods to analyse the same panel data. Interestingly, the DEA analysis provides estimates of subject-specific unit costs that are in the same ballpark as those provided by the parametric methods. We then extend the previous analysis by examining potential cost savings and output augmentations in different HEI groups using several different DEA models. The findings include a suggestion that substantial gains of the order of 20-27% are feasible if all potential savings are directed at raising student numbers so that each HEI exploits to the full not only operating and scale efficiency gains but also adjusts its student mix to maximise student numbers. Finally we use a Malmquist index approach to assess productivity change in UK HEIs. The results reveal that for a majority of HEIs productivity has actually decreased during the study period.