A cross-country comparison of the middle class as measured by income polarization indices is commonplace in welfare economics. Using the 2001–2007 housing cycle and data for Australia, the United States, Germany, and Switzerland (and an array of methods, including triple-difference design), I show that polarization indices based on disposable income are unreliable. The cycle changes the relative importance of non-monetary income from housing (imputed rent), particularly for middle-income households. Therefore, to ensure that convenient income measures do not misrepresent the size of the middle class, researchers should verify the absence of swings in housing prices during their study period.