Variable conclusions

Your article “‘Biased’ students give BME academics lower NSS scores, says study” (News, 21 January) cited a University of Reading analysis of National Student Survey data that purported to show that “students are happiest when taught by staff with the following characteristics: white, full professors, holding doctorates, and on fixed-term contracts”.

The original article links to a preprint manuscript, which is an analysis of the 2014 NSS data in relation to a wide range of variables on the characteristics of universities and their staff. The authors have trawled the data for relationships with as many variables as were available. These sorts of fishing expeditions in large datasets are notoriously unreliable at identifying genuine trends in the data: if you try to match enough things to enough data, by chance some will fit. The manuscript deals with multiple variables, many of which are probably closely correlated and potentially confounding, which is a good recipe for selective storytelling.

The effects appear to be small (although the manuscript is unclear on methods): 0.0015 NSS “teaching satisfaction” points and 0.06 overall satisfaction points drop for every percentage decrease in white staff. So, simplistically, the difference in teaching satisfaction between a totally white and totally non-white staff base would be less than 1 per cent; and a 10 per cent difference in non-white make up results in a 0.6 per cent change in NSS score. Although this may have influence on an institution’s placing in the overall ranking, this can hardly be seen as a major factor in determining the NSS score. It is equivalent to arguing that the thickness of my socks (because it is consistent and measurable) is an important determinant of my height.

Philip Wheeler
Senior lecturer in ecology
The Open University

Send to

Letters should be sent to: THE.Letters@tesglobal.com

Letters for publication in Times Higher Education should arrive by 9am Monday. View terms and conditions.