Most empirical work on inequality uses measures that are based on household surveys. These aim to provide a comprehensive overview of income inequalities, covering all social strata and comparable both through time and between countries. Gini coefficients are the index mostly commonly computed from these sources, though various quintile ratios are also frequently employed. Deininger and Squire [1996] have compiled an impressive data set of available Gini and quintile measures of inequality. Yet, the limitations of this data for studies of the evolution of inequality through time are evident from Table 1, which shows the number of data points in a 26-year period (1970-1995) for those countries for which more than three data points are available. Only four countries show data for virtually every year, and most do not have data for even half of the years. These gaps are irreparable. There is no way to construct Gini coefficients for countries and years for which adequate household sample surveys were never conducted. Fortunately, the decomposability properties of the Theil measure make it possible in part to repair this gap, albeit in most cases only for the manufacturing economy. In particular, one can compute the between-group measure of inequality (T" hereafter) across industrial sectors, as delineated by national or international industrial classification schemes. Data on industrial wages, earnings and employment are very easily found. The data are also reasonably reliable; there is little reason to suspect that they are faked in any systematic way that would affect a Theil measure. Where gross errors do occasionally enter into the recording, the regularity and hierarchical structure of the data sets often means that these can be detected.