Intelligence and education: causal perceptions drive analytic processes and therefore conclusions | International Journal of Epidemiology | Oxford Academic
Article Navigation
Intelligence and education: causal perceptions drive analytic processes and therefore conclusions
Ian J Deary Wendy Johnson
International Journal of Epidemiology, Volume 39, Issue 5, 1 October 2010, Pages 1362–1369,
Intelligence and education: causal perceptions drive analytic processes and therefore conclusions | International Journal of Epidemiology | Oxford Academic
Published: 26 May 2010 Article history
Views PDF Cite
Permissions
Share
Abstract
Background Educational attainment is associated with many life outcomes, including income, occupation and many health and lifestyle variables. Many researchers use it as a control variable in epidemiological and other social scientific studies, often without specifying exactly what environmental effects or set of personal characteristics is being controlled. Other researchers assume that genetically influenced intelligence drives educational attainment, and think that intelligence is the appropriate control variable. Researchers’ different and often unstated causal assumptions can lead to very different analytical approaches and thus to very different results and interpretations.
Methods, results and conclusions We document several examples of this important variation in the treatment of education and intelligence and their association. We recommend greater clarity in stating underlying assumptions and developing analytical approaches and greater objectivity in interpreting results. We discuss implications for study designs.
Education, intelligence, intelligence quotient, health, epidemiology
Topic: educational status intelligence
Issue Section: Theory and Methods
Introduction
Brighter people tend to get more schooling, and the longer-schooled tend to be brighter. These simple facts elicit surprisingly different interpretations among the many epidemiologists and social scientists who measure education and intelligence for research use. Their different interpretations contribute to differences in methodological and analytical treatments that can have profound impacts on study design, methodological choice, results and interpretation of results. Implicit interpretation of the association between these two variables is common throughout epidemiological and other social science research. With regard to health and other outcomes, this observationally ambiguous association involves the statistical issues of mediation, moderation, confounding and direct and indirect effects. These issues are always troublesome because their treatment depends not only on timing of available measurements, but also on understanding of causal pathways. The issues involved in the association between these particular two variables, however, are especially important to the newly emerging field of cognitive epidemiology. One or the other—especially education, due to its greater availability in datasets—is very commonly used as a control variable; intelligence and education are closely inter-related, and they may be measured with varying degrees of precision. Moreover, there is probably some form of longitudinal cascade between them, quite possibly with reciprocal causal and selection effects;1 yet, the optimal longitudinal data sequence to understand the processes involved in these reciprocal and selection effects is often unavailable. At the same time, because they are not perfectly correlated, neither education nor intelligence is a perfect proxy for the other. It is thus often important to understand objectively which (if either) exerts a causative effect on an outcome.
Intelligence and education: clearly correlated, but what is the direction of causation?
Intelligence and education have been studied together since the earliest empirical research on these topics. Spearman2 found teachers’ estimates of intelligence to be correlated with school exam results. Binet3 developed what we now know as intelligence quotient (IQ) tests to identify those children who would not benefit from normal education. When intelligence and educational outcomes—often assessed as years of full-time education or as highest achieved qualification, and also by school grades or educational achievement test scores—are measured at about the same time, a typical correlation is ∼0.5.4 Like any other correlation, a cross-sectional correlation between intelligence and education demands an open mind with regard to causal interpretation. Perhaps more intelligent people gain access to more and higher-level education. Perhaps exposure to more education causes higher intelligence test scores. The problem is one that is basic to epidemiology: what is person and what is situation, what is genetic and what is environmental and what is cause and what is effect? Influences may flow in both directions, and longitudinal studies can help to quantify their relative magnitudes.
Does higher intelligence beget better educational outcomes?
In longitudinal studies that measure psychometric intelligence first and educational attainments later (thus assessing that causal chain), there is a moderate to strong correlation between the two, as assessed by years spent in full-time education, the highest qualification obtained by a person or the scores obtained on educational assessments.5 For example, in a study of approximately 70 000 children in the UK, the general factor from the Cognitive Abilities Test (CAT) battery taken at age 11 years correlated about 0.8 with the general factor of grades on the General Certificate of Secondary Education (GCSE) examinations taken at age 16 years.6 The general factor of the CAT test had very similar loadings from the three domains of verbal, non-verbal (abstract) and quantitative reasoning. Older studies have reported correlations ranging from 0.60 to 0.96.7–9 The conclusion from such studies might be that intelligence has stronger causal effects on educational results than vice versa.