Commentary: Liberty, fraternity, equality.

relationship between income inequality and health has been independently discovered several times by people who appeared not to know of each other's work. 1,2 When I first published on the subject 3 I too was unaware of Rodgers' paper 4 until George Davey Smith pointed it out to me. Rodgers introduced an income distribution term into his regressions simply to take account of the well-known curvature of the international relation between gross national product (GNP) per capita and life expectancy as countries go through the so-called epidemiological transition. In contrast, I came to the subject through an interest in health inequalities within Britain. While using occupational incomes and death rates to discover whether death rates were responsive to changes in incomes for which people were not self-selected, 3,5 I also wanted to know whether the shape of the individual relation between income and mortality within Britain was linear or curvilinear. My interest in this point was explicitly to find out whether, if income was redistributed from the rich to the poor, the health of the poor would benefit by more than the health of the rich would suffer: I wanted to know if average health would be improved by redistribution. Having been convinced by my results both that mortality was responsive to changes in income and that the health of the poor was more sensitive to changes in income than that of the rich, I thought it might be worth looking to see whether national mortality rates were lower in countries with a narrower income distribution. Mildred Blaxter had mentioned to me that she thought there was such as relationship though she had not looked at it herself. My first three papers on income distribution 3,5,6 were all accompanied by evidence that the individual relation between income and health or mortality was curved: that not only seemed to lend plausibility to a causal interpretation of the cross-sectional relationships between income distribution and mortality in different countries, it was why I thought they existed. It was then a little annoying when, a few years later Gravelle 7 suggested, as if it were a new hypothesis, that the income distribution relationship might be explained by the curvature of the individual relation between income and health. He did so without reference to the evidence in my earlier papers, although I had pointed it out to him when he sent me a draft of …