Income distribution data from before the Industrial Revolution usually comes in the shape of social tables: inventories of a range of social groups and their mean incomes. These are frequently reported without adjusting for within-group income dispersion, leading to a systematic downward bias in the reporting of pre industrial inequality. This paper suggests a correction method, and applies it to an existing collection of twenty-five social tables, from Rome in AD 1 to India in 1947. The corrections, using a variety of assumptions on within-group dispersion, lead to substantial increases in the Gini coefficients. Combining the inequality levels with data on GDP, a robust positive relationship between income inequality and economic growth is confirmed. This supports earlier proposals, based on fewer data points, of a "super Kuznets curve" of increasing inequality over the entire pre-industrial period.