Files
Abstract
This paper proposes the use of a generalisation of the Logistic distribution, the Burr Type II distribution, as the error structure in a Type I (Standard) Tobit model. The use of the Burr II is motivated through both an heterogeneity argument and by noticing the need for potentially non-symmetric distributions in Tobit models. Tests for symmetry (the Logistic distribution) and its implied heterogeneity are also proposed. Finally the model is estimated using Australian data on alcohol expenditure.