Stochastic DEA constructs production frontiers that incorporate both inefficiency and
stochastic error. This results in a closer envelopment of the mean performance of the companies
in the sample and diminishes the effect of extreme outliers. We use the Land, Lovell and Thore
(1993) model incorporating information on the covariance structure of inputs and outputs to study
efficiency across a panel of 17 European airlines in the 1990s during the early phase of
liberalisation. After allowing for stochastic error in computing the relative efficiencies we
conclude that the airlines that were efficient in 1995 resembled those that were efficient in 1993
but not those in 1991. The airlines that were efficient in 1995 were the larger companies.