"In recent years there appeared numerous articles about best suited methods to estimate income distributions and polarisation tendencies. The Austrian census data 1989-1999 corroborate findings that summary statistics like dispersion or Gini remain rather stable over time when evaluated for the economy as a whole. However across space and time there is considerable movement of incomes and skills in locational differentiation of housing tenures. To make the argument precise, this contribution compares Kullback-Leibler distances with parametric distribution estimates. The ""middle class"" generates a rather flat income pattern around median and mean; a newly developed type 3 extremal value distribution (a log-Weibull type) indeed outperfoms the classic lognormal distribution in subpopulations where that middle class is relatively dominant. This is reflected in the shape of the distributions which hence permit better insights than summary statistics. It is also shown that these movements create a tendency towards polarisation in certain parts of the population but also that pattern is not uniform. The paper argues therefore that chasing one single and ""best"" type of income distribution is misleading because the shape depends on the social composition. A thorough analysis of the movements appears necessary when drawing conclusions about housing policies and about future public support for housing needs."