The researchers calculated the correlation between the earnings of parents and sons for a cohort of 2,163 babies born in 1958 and another cohort of 1,976 born in 1970. They showed that the correlation was higher for the 1970 cohort, and also higher than equivalent figures from countries like Canada.
Therefore, apparently, poor social mobility is a UK-specific problem, worsening over time.
But the 1958 NCDS cohort had 16,460 babies, and the 1970 cohort had 16,695 (see, for example, http:www.cls.ioe.ac.ukCohortNcds2000mainncds00.htm). Therefore, 87 per cent of the 1958 cohort and 88 per cent of the 1970 cohort have dropped out of this new study.
This level of attrition means that the reported small differences between the two studies, in terms of their already small inter-generational correlations, must be viewed with extraordinary caution.
The 11-12 per cent still involved in the Sutton Trust study are not, it must be recalled, a random sample of the original cohort - merely those who are relevant, traceable, and responded.
Therefore, quoting statistical significance as evidence of the importance of the changes over time is simply an error. I have not looked at the quality of the figures or appropriateness of the analyses from the other countries mentioned in the story.
Professor Stephen Gorard Department of Educational Studies University of York