Should we be using social security benefits data as proxies for income poverty?

Administrative data on means-tested benefits have come to be widely used as proxy measures of local poverty rates in the UK. Two such uses are described: allocating funding to local government, and constructing neighbourhood deprivation indices. The paper argues that such uses risk errors of both measurement and interpretation. Income poverty is conventionally defined against percentages of national median income. The paper shows that out-of-work means-tested benefit claims are valid, but incomplete, proxies for income poverty so conceived. Further, there is considerable regional variation in the statistical relationship between rates of income poverty and rates of benefit claims. Benefits data should be used only with caution to describe spatial differences in poverty.