Michael Bastedo, associate professor at the Centre for the Study of Higher and Postsecondary Education at the University of Michigan, thinks it is "best to avoid them".

He has sent me his research on the reputation surveys used by US News & World Report to compile its rankings of US institutions. He concludes that the scores (worth 25 per cent of the overall score) correlate closely with a university's ranking position the previous year but bear little relation to changes to other performance measures.

That is to say, a good ranking produces a good reputation score, which perpetuates the good ranking.

The reputation survey used in the now-defunct Times ÖйúAƬ-QS World University Rankings was one of its most controversial elements: a survey of a tiny number of academics should not determine 40 per cent of a university's score.

ÖйúAƬ

Yet THE has decided to keep a reputational survey in its new world rankings for 2010 and beyond.

With new data partner Thomson Reuters, we have commissioned the polling company Ipsos MORI to carry out the survey. We are committed to achieving a higher response rate, a better targeted and representative sample, and to giving the results less weight in the overall score.

ÖйúAƬ

This will be a vast improvement. But how can we guard against the pitfalls identified by Professor Bastedo?

Although he is unconvinced of the value of reputation polls, he told me that the problems "could possibly be ameliorated by being careful about who is surveyed and how".

We welcome your views on all rankings issues. Post your comments at . University heads may email me to join our reputation survey "platform group".