Transit costs and cost efficiency: Bootstrapping non-parametric frontiers

This paper explores a selection of recently proposed bootstrapping techniques to estimate non-parametric convex (DEA) cost frontiers and efficiency scores for transit firms. Using a sample of Norwegian bus operators, the key results can be summarised as follows: (i) the bias implied by uncorrected cost efficiency measures is numerically important (close to 25%), (ii) the bootstrapped-based test rejects the constant returns to scale hypothesis (iii) explaining patterns of efficiency scores using a two-stage bootstrapping approach detects only one significant covariate, in contrast to earlier results highlighting, e.g., the positive impact of high-powered contract types. Finally, comparing the average inefficiency obtained for the Norwegian data set with an analogous estimate for a smaller French sample illustrates how the estimated differences in average efficiency almost disappear once sample size differences are accounted for.