Deriving Long - Run Inequality Series from Tax Data ∗

Prior to the last three decades, regular surveys on household income were rare or non-existent in many developed countries, making it difficult for economists to develop long-run series on income distribution. Using taxation statistics, which tend to be available over a longer time span, I propose a method for imputing the incomes of non-taxpayers, and deriving the underlying distribution of income. Because taxation statistics are typically disaggregated by gender, it is possible to derive separate income distribution series for men and women in countries where individuals file separately. I show that over the past four decades, the distribution of adult male incomes and the distribution of family incomes are highly correlated. Applying this method to Australia, I develop a new annual series for inequality from 1942 to 2001. Inequality fell in the 1950s and the 1970s, and rose during the 1980s and 1990s – a pattern similar to that in the UK.