Is education causal? Yes

Abstract

In a provocative analysis piece Deary and Johnson critique the lack of theory behind the use of education as an explanatory variable in epidemiology.1 There are of course several well-defined sociological theories of education in relation to health, although it is a fair comment that these are not always sufficiently heeded in population-based analyses. More specifically, they also challenge the extent to which education has a causal effect on subsequent intelligence (and by implication other health-related outcomes). One of several studies singled out for criticism in this context is that of our path model linking circumstances of origin, childhood intelligence assessed at 8 years, educational attainment by 26 years, occupational attainment (social class) by 43 years and cognitive function a decade later in the British 1946 cohort.2 Briefly, we found strong direct paths between childhood and midlife cognition, and between the former and education; but also a modest independent path from education to midlife cognition, with only a small additional contribution from adult occupation. We conclude that education has a causal influence on subsequent cognition, which is not fully explained by social class of origin, or childhood intelligence, or subsequent occupational attainment. Deary and Johnson direct three criticisms at this conclusion, which we list in ascending order of challenge. First, they concentrate on only one of our outcomes, the NART, which they suggest is, as a mea…