Exact Computation of Censored Least Absolute Deviations Estimator

We show that exact computation of the censored least absolute deviations (CLAD) estimator proposed by Powell (1984) may be achieved by formulating the estimator as a linear Mixed Integer Programming (MIP) problem with disjunctive constraints. We apply our approach to three previously studied datasets and find that widely used approximate optimization algorithms can lead to erroneous conclusions. Extensive simulations confirm that MIP- based computation using available solvers is effective for datasets typically encountered in econometric applications and that, despite the proliferation of competitors, CLAD remains a useful estimator.

Keywords: CLAD estimator, censored regression models, Mixed Integer Programming, disjunctive constraints

JEL Classification: C13, C14, C24, C44, C61

Suggested Citation: Suggested Citation

Bilias, Yannis and Florios, Kostas and Skouras, Spyros, Exact Computation of Censored Least Absolute Deviations Estimator (December 4, 2018). Available at SSRN: https://ssrn.com/abstract=2372588 or http://dx.doi.org/10.2139/ssrn.2372588