Studies of the rates of return to research have usually been based on the implicit assumption that if there were no research, then there would be neither growth nor decline in output or productivity. In the case of livestock, particularly in southern Africa, which has a sub-tropical disease ecology and a long history of disastrous losses due to disease, the assumption is especially unreasonable. It ignores the losses that would have occurred in the absence of livestock health research, resulting in underestimation of rates of return. This study draws on data from South Africa to illustrate the magnitude of the error, by separating the maintenance effects of animal health research from output increases due to animal improvement research. This is possible because health and improvements research are conducted at separate research institutes and there are data on cattle deaths due to disease, which allows the effects of health expenditures to be calculated. Explicitly, taking the negative effect of diseases into account considerably increases the returns to the livestock research of the South African Agricultural Research Council (SAARC). Instead of a ROR of 18% for animal research in total, the result is a ROR of at least 35% for animal health research and 27% for improvements research, suggesting a minimum underestimation of about 50%. These results suggest that livestock research is productive, once it is properly decomposed. The implication is that all ROR estimates that implicitly assume that with no research, there would be no change in output, or productivity, must be severely biased downwards.© 2001 Elsevier Science B.V. All rights reserved.