True Versus Measured Theil Inequality

This article deals with several problems that arise when the Theil coefficient of income inequality is computed in practice. Aggregation of income data into brackets leads to an underestimation of the true Theil inequality, which is defined as the value of the coefficient as computed from individual income data. The assumption that the individual incomes are distributed according to a linear density function within the income brackets is suggested as a method to estimate this aggregation error. Calculations show that this method approximates the true aggregation error reasonably well. Several methods are discussed concerning the treatment of negative incomes. In particular one can construct an income bracket that contains both negative and positive incomes and which in the summation formula is weighted with zero weight. Of all methods this procedure using the assumption of a linear density function within brackets, yields the highest value of the Theil coefficient and is thus preferred to the other alternatives.