Universities and science minister David Willetts has persuaded the coalition to spend millions of pounds on a new national birth cohort study, and Guardian columnist Polly Toynbee – although “unaccustomed to praising this government” – has welcomed the project, saying that such studies are “the crown jewels…in the long and distinguished history of British social science” (“Evidence first, judgement later”, News, 20 June).
But contrary to Toynbee’s endorsement, national birth cohort studies are a poor way to collect social science data. They identify all babies born in the UK on a particular few days, record certain data, then follow them up at intervals and record information such as IQ, educational attainment, earnings, health and so on.
This sounds fine, but the problem is that it has never been possible to find all of the original cohort during the follow-ups. For instance, the second national birth cohort study of those born during the week of 3-9 March 1958 began with 17,419 babies. At the latest follow-up in 2008-09, when the sample was aged 50-51, only 9,790 could be found.
Those who are identified at later stages of the studies are no longer representative samples of the population but tend to be white, middle class and better educated. These unrepresentative samples provide some valid rough-and-ready information, but are not much use for accurate social science research.
Professor emeritus of psychology
University of Ulster