On the approximation bias to benefit measures in discrete choice models

Abstract

We estimate the bias associated to using different approximations for calculating the benefits of transport projects. For this we generated a simulated databank that reproduces the real behaviour of a revealed preference sample on the choice of mode for work trips. Exact measures were computed for three transport policies differing in their degree of impact on mode choices. These values were then compared against the results of applying simpler but less precise measures, showing the existence of significant biases when classical explanatory models were considered. Significant biases were also detected when the correct methodology for marginal changes was applied in the presence of non-marginal policies. In the first case, the proportional error does not depend on the size of the policies, but in the second it grows with their impact. We also simulated another databank with non-linear utilities in income and calculated again some measures of welfare. We found that there are no significant biases if income effects are not considered. Neither did we find a systematic relation between these biases and the size (in terms of change) of the policy.