“The impact of education on health behaviours: smoking, drinking and exercise. Evidence from the UK.”

The necessity to refocus health policy from health care services to a broader base of determinants of population health which lie outside health care system is already recognised (Deaton, 2002) and therefore conclusive evidence on determinants of health outside health care system and their magnitude is essential. Although the strong associations between more education and lower mortality had given ground to argue for education policy as a part of health policy (Gwatkin, 1982) it was not enough to justify economic and political decisions. Conclusive evidence on causal effect in the education-health association would provide certainty and help to make better decisions on resource allocation towards health services or education since it is already recognised that determinants of health also lie outside the health sector (Deaton, 2002). Although education is a strong correlate of health regardless of choice of proxy for health (Van Kippersluis et al., 2009, Oreopoulos, 2003, Elo and Preston, 1996), determining causal effect has been and still remains a challenge due to the problems tackling the reverse causality. On the one hand education attainment could determine individual’s health, on the other it could very well be that individual’s health in childhood could determine educational attainment. A third alternative suggests that the relationship between the two is determined by other factors to which both are associated. Identification of a causal pathway or lack of it between education and health has societal importance because knowledge of a determinant, the direction and magnitude of its effect is a powerful tool to improve policy decisions. Along with enhanced capacity to make future policies, an understanding of the causal mechanism allows a more precise and comprehensive evaluation of the effects of current and past education and health policies. Currently the empirical evidence is mixed and therefore inconclusive. There are studies which suggest that health determines educational attainment by reporting that bad poor childhood health and low birtweight has negative effect on schooling later on (Behrman and Rosenzweig, 2004, Case et al., 2005), but findings are challenged by Cutler and Lleras-Muney (2006). There is also a growing body of evidence which supports the causal pathway from education on health (Grossman, 2004, Cutler and Lleras-Muney, 2006). A recent trend in the literature is drawing evidence from analysis of schooling reforms in the U.S. (Lleras-Muney, 2002, Adams, 2002), Denmark (Arendt, 2005, 2008), Sweden (Spasojevic, 2003), the Netherlands (Van Kippersluis et al., 2009), and the U.K. (Silles, 2009, Clark and Royer, 2008). Although the results from schooling reforms are believed to be source of strong evidence (Grossman, 2004), there is variation in the results. While Silles (2009) reports a significant positive effect from the reform to schooling in the UK, Clark and Royer (2008) fail to find a causal effect on health from the same reform. There is also mixed evidence regarding the mechanisms through which education may influence health production. Studies of Grossman (2006) and Kenkel (1990) report evidence that supports the allocative efficiency hypothesis while Cutler and Lleras-Muney, 2006 point towards allocative efficiency hypothesis. The purpose of this study is to investigate whether education has an effect on health behaviours. According to the allocative efficiency in health production hypothesis (Grossman 2006) the more educated alter their behaviour in a way which results in better health. This paper firstly attempts to replicate Silles (2009) findings on the effect of schooling reform on self-reported health using the UK General Household Survey data to infer causal pathway from education to health. Secondly it adds to the current literature in the field by further exploring whether similar patterns exist in health-related behaviours: smoking, drinking, and physical activities. Finding similar disparities in health behaviours would confirm the allocative efficiency hypothesis whereas lack of it would give more plausibility to production efficiency models and therefore would contribute to the existing research in the causal effect mechanisms. While Clark and Royer (2008) explore smoking and body mass index behaviour patterns by education, according to author’s knowledge, drinking and exercise behaviours have not been previously examined in the UK population. Currently most of the studies use instrumental variables (IV) method to solve the endogeneity problem with education and health implied by their reverse causality (Adams, 2002, Arendt, 2005, Spasojevic, 2010, Grossman, 2006). The study also contributes to the existing empirical evidence by applying the method of regression discontinuity design (RDD) which has gained popularity recently and is said to be (Lee and Lemieux, 2009) and is used not only as a complement to IV (Silles, 2009), but lately also as the main method of research (Clark and Royer, 2008) to study health and education relationship. The advantages of RDD include high internal validity and being a very close substitute for a randomised trial and the design is very appropriate for researching natural experiments like education reforms (Lee and Lemieux, 2009). The disadvantage of this method yielding only a very local treatment effect is overcome by the fact that the local treatment effect is the primary interest in this study. The results reported show that no effect of the education reform on the general health of individuals is found although correlations between health and education are consistent with current empirical evidence. An increase in likelihood in reporting a long-standing illness is found, but it is not retain significance in sensitivity analysis. None but two health behaviours are affected by the reform: quitting smoking and on preferring high tar level cigarettes. The likelihood of quitting smoking increased while probability of preferring high level tar cigarettes among smokers decreased substantially. The results of the impact on smoking behaviours do suggest that education could an effective tool for smoking cessation policies and strategies. The magnitude of the effect, however, is sensitive to bandwidth specification and therefore conclusive evidence cannot be drawn. It is likely that having data from more and younger individuals at interview may be a complementary factor to the sensitivity to bandwidth which are driving the differences between this study and those of Silles (2009) and Clark and Royer (2008). Future research using earlier waves of the General Household Survey data could however add more certainty on whether the lack of significant effect is due to no causal pathway from the reform or the effect is mitigated by self-selection as individuals age. This research could also add more robust evidence of the changes in smoking behaviours. This study is organised in five chapters where the in the first one the background, current theoretical frameworks and empirical evidence along with methodology and the background of the 1944 Education Act reform is included. Data used in the study is described in chapter two and the results are reported in chapter three. They are followed by discussion of the results found in health outcome and health behaviours in chapter four and the conclusions of the study are presented in chapter five.