During the past few years, three-dimensional chemical transport models have matured to a point at which their results are often in good agreement with observations, lending confidence to model estimates of ozone loss. In this paper a Lagrangian photochemical model is used to assess the error in model predictions that results from errors in the kinetic and photochemical data used by the model. The results indicate that Arctic ozone destruction can be modeled with an uncertainty of ¿25% (1 sigma) and that midlatitude ozone destruction can be modeled with an uncertainty of ¿50%. It is also shown that two reactions, from a total of more than 100, are responsible for more than a third of the uncertainty in model calculations of Arctic ozone loss. ¿ 1997 American Geophysical Union