Moreover, in the presence of heavy tails, Mitzenmacher (2004) has shown that a log-normal distribution can exhibit a Pareto tail by expanding its shape parameter, σ2. The convenience argument is justified as follows. In inequality analysis, decomposition is an important argument as it allows the decomposition of inequality into within-subgroup and between-subgroup components. The additive structure of a mixture model can preserve the decomposability of an inequality index. The log-normal distribution provides analytical formulae for the main inequality indices and their decomposability is preserved by the linear structure of the finite mixture. This decomposability would have disappeared if we had considered a finite mixture of normal densities on the log of the income variable.