Measurement and definition of the link between unemployment and health

Some of the statistical difficulties inherent in any measure of the interaction between employment and health are considered. Unemployment and health are seen to be difficult terms to define and both are seen to contain subjective elements. Micro and macro models are discussed as both have strengths and weaknesses. Most micro models trace the experience of small groups of individuals usually selected either because they are unemployed and/or are ill. Thus the results cannot be acceptably generalized to the population at large. Macro time series are concerned with a measure of the interaction of unemployment and health at the aggregate national level. In macro models the precise link between the unemployment experience of an individual and the health experience of the same individual is lost in favour of the association of a measure of aggregate unemployment with a measure of aggregate health status. Both model types are liable to error in describing what is happening at the national level. Panel data macro models are likely to be most accurate as they answer most of the criticisms mentioned above, but such data are scarce and expensive to collect. However, it is argued that macro time series models applied to data for subgroups of individuals may provide a low cost way to obtain a useful measure of possible interaction.