Causal evidence that the impact of income on pupil outcomes is in fact zero

The correlation between family income and educational achievement is well established worldwide. This correlation is often interpreted in causal terms, leading to calls for income redistribution and other social policy interventions that could improve pupil outcomes. There is a limit, some argue, what education policy can achieve without social programmes that improve the broader life situation of youngsters.

However, the link between family income and educational achievement could be due to several ‘unseen’ variables that affect both pupils’ socio-economic conditions and their achievement – including heritable factors – and it has been difficult to credibly establish that causality runs from the former to the latter.

In a new paper, ‘Wealth, Health, and Child Development: Evidence from Administrative Data on Swedish Lottery Players’, David Cesarini, Erik Lindqvist, Robert Östling, and Björn Wallace shed new light on the issue by studying lottery players in Sweden. Conditional on a couple of factors, which the authors observe and control for, prize winnings are randomly assigned, which gives them a unique opportunity to compare like for like. The lottery prizes are relatively high – 75% of them are in the range of between 1 and 40 Swedish median annual disposable incomes.

The authors analyse a range of outcomes of winning the lottery, including health among both adults and children, but, from an educational standpoint, the most interesting ones are grades and test scores at the age of 15 as well as cognitive and non-cognitive skills among males at the age of 18.

The results are striking. The correlation between family income and pupil outcomes entirely disappears when analysing the effect of winning the lottery. If anything, the effect is in fact marginally negative. This holds true both for children who were under nine years old and those who were older at the time of the win.

In other words, the authors provide causal evidence that the impact of income on pupil outcomes is in fact zero – and that the correlations generally observed are due to ‘unseen’ variables that generate both higher family income and pupil outcomes.

Lottery players are about 10 years older than the average Swede, but once adjusting for age, they are similar to a representative sample of the Swedish population in terms of baseline characteristics. It would therefore appear as if the results could be extrapolated to a larger population than just the lottery players, at least to similar developed countries with publicly-funded education systems.

Certainly, one could always argue that the windfalls were not large enough to improve pupil outcomes, or that they must occur very early in (or before) children’s lives to generate higher achievement. This may of course be true. But this would also imply that social policy interventions and income redistribution, as methods to generate higher achievement, would be wasted as long as children are past a certain cut-off age.

Nevertheless, this is all speculation – anybody can play the God of the Gaps. Overall, the paper is a great example of how important good research design is to unveil causal effects in education research. Indeed, displaying results that fly in the face of much correlational and anecdotal evidence appears the only way to do so.

Gabriel Heller Sahlgren

This comment piece is also the Editor's Pick in the CMRE Monthly Research Digest_03_16. The piece reviews a paper by David Cesarini, Erik Lindqvist, Robert Östling, and Björn Wallace, ‘Wealth, Health, and Child Development: Evidence from Administrative Data on Swedish Lottery Players’ for the Quarterly Journal of Economics (forthcoming), a free copy of which may be downloaded here.

You can download free copies of back issues of the CMRE Monthly Research Digest here.

 to subscribe.