OBJECTIVES To estimate the impact of HIV/AIDS on individual labour productivity during disease progression. METHODS We used a retrospective cohort design to study the productivity and attendance of tea estate workers who died or were medically retired because of AIDS-related causes between 1997 and 2002 in western Kenya. We compared daily output in kilograms of tea leaves plucked, use of paid and unpaid leave and assignment to less strenuous tasks by 54 workers who died or were medically retired because of AIDS to those of comparison workers, matched on time and tea field using longitudinal regression. RESULTS HIV-positive workers plucked less tea in the 18 months preceding AIDS-related termination and used more leave in the 3 years before termination. After adjusting for age and environmental factors, cases plucked between 4.11 and 7.93 kg/day less in the last year and a half before termination. Cases used between 9.2 and 11.0 more sick leave days, between 6.4 and 8.3 more annual leave days, between 19.9 and 11.8 more casual leave days, and spent between 19.2 and 21.8 more days doing less strenuous tasks in the 2 years before termination than did comparison pluckers. Tea pluckers who terminated because of AIDS-related causes earned 16.0% less in their second year before termination and 17.7% less in the year before termination. CONCLUSION These results provide empirical estimates of the impact of HIV/AIDS on labour productivity. As workers often bring unrecorded 'helpers', actual differences may be greater. Decreased attendance and output may put sick workers in jeopardy of losing their jobs and impose financial burdens on employers.