Using linked data to estimate the relationship between physical activity and health

Abstract: There is considerable debate about the relationship between physical activity and health, with policy makers asserting that an increase in participation in exercise will lead to higher aggregate levels of health. However, empirical evidence has been limited by the lack of data available to control for the impact of reverse causation between health and physical activity. In this paper, we create a unique dataset using UK data from Understanding Society and Sport England that allows us to use local authority level information as instruments for physical activity, similar to Huang and Humphreys (2012). This local authority information facilitates the estimation of instrumental variable regression to provide more precise estimates of the effect of physical activity on health. Measuring physical health by the SF12 PCS, we find that an increase in participation in sports activities by one day a week (equivalent to 50 days a year) increases physical health by around 10%. For mental health, measured by SF12 MCS, we find that a similar increase in sports is associated with a 1% increase in health. We observe greater health benefits for lower active people, and people who rate themselves as highly active achieve the lowest percentage increases in their health, both physical and mental.

We wish to acknowledge very helpful comments from seminar participants at the Department of Health, London, NHS England Analysts, Leeds, the ESRI, Dublin, iHEA/ECHE 2014 Joint Conference, Dublin, and InstEAD, Department of Economics, University of Sheffield. We thank Yiu-Shing Lau and Cheryl Jones for research assistance in compiling the database. Funding from University of Manchester, ESRC Transformative Research Seed Fund is gratefully acknowledged.