Income, relative income, and self-reported health in Britain 1979-2000

According to the relative income hypothesis, an individual’s health depends on the distribution of income in a reference group, as well as on the income of the individual. We use data on 231,208 individuals in Great Britain from 19 rounds of the General Household Survey between 1979 and 2000 to test alternative specifications of the hypothesis with different measures of relative income, national and regional reference groups, and two measures of self assessed health. All models include individual education, social class, housing tenure, age, gender and income. The estimated effects of relative income measures are usually weaker with regional reference groups and in models with time trends. There is little evidence for an independent effect of the Gini coefficient once time trends are allowed for. Deprivation relative to mean income and the Hey-Lambert-Yitzhaki measures of relative deprivation are generally negatively associated with individual health, though most such models do not perform better on the Bayesian Information Criterion than models without relative income. The only model which performs better than the model without relative income and which has a positive estimated effect of absolute income on health has relative deprivation measured as income proportional to mean income. In this model the increase in the probability of good health from a ceteris paribus reduction in relative deprivation from the upper quartile to zero is 0.010, whereas as an increase in income from the lower to the upper quartile increases the probability by 0.056. Measures of relative deprivation constructed by comparing individual income with incomes within a regional or national reference group will always be highly correlated with individual income, making identification of the separate effects of income and relative deprivation problematic.