Surveys and Citations

I have just finished calculating the correlation between the scores for the academic survey and citations per faculty on the 2010 QS world University rankings.

Since the survey asked about research and since citations are supposed to be a robust indicator of research excellence we would expect a high correlation between the two.

It is in fact .391, which is on the low side. There could be valid reasons why it is so low. Citations, by definition, must follow publication which follows research which in turn is preceded by proposals and a variety of bureaucratic procedures. A flurry of citations might be indicative of the quality of research begun a decade ago. The responses to the survey might, on the other hand, be based on the first signs of research  excellence long before the citations start rolling in.

Still, the correlation does not seem high enough. At first glance one would suspect that the survey is faulty but it could be that citations do not mean very much any more as a measure of excellence.

It would be very interesting to calculate the correlation between the score for research reputation on the Times Higher Education WUR and its citation indicator.

We would expect the THE survey to be more valid since the basic qualification for being included in the survey is being the corresponding  author of an article included in the ISI indexes whereas for QS it is signing up for a journal published by World Scientific. But it can no longer be assumed that authorship of any article means very much . Does it always require more initiative and interest to get on the list of co-authors than to sign up for an online subscription?

It should also be noted that there is an overlap between the two surveys as both are supplemented with arts and humanities respondents from the Mardev mailing lists.

I have calculated the correlation between the citations indicator (normalised average citations per paper)  in the THE 2010 rankings and the research indicator -- volume ( 4.5 % of the total score)  income (6%)  and reputation (19.5%)

This is .562, quite a bit better than the QS correlation . However, the research indicator combines a survey with other data.

It would be very interesting if THE and/or Thomson Reuters released the scores of the individual components of the research indicator.