This paper presents a robust estimation of two income distribution models using Spanish data for the period 1990-91 under three different concepts of income. The effect on the estimates of the Theil index due to the choice of the definition of income and of the estimation method is also analysed.