Scientists from a Shanghai technical university have tried to rank all the top universities in the world. A press release is can be found here; links to the various rankings can be found here; and a ranking of only North American universities is here.

First, a good thing about the rankings: they avoid the ridiculous slant towards high-tuition upper class institutions exhibited by the US News & World Report ranking. US News ranks such mediocre universities as Dartmouth, Northwestern, and Washington University at St. Louis ahead of such research powerhouses as Berkeley, University of Michigan-Ann Arbor, and NYU. US News is biased in part due to the ridiculous criteria it uses to measure quality of schools; these criteria include alumni giving rate (which measures the percentage of graduates felt that their 20's were the best time of their life) and graduation rate (which rewards colleges for giving students a pass). Dartmouth College, for example, has no strong departments in any particular area that I'm aware of; I've certanily never seen it appear anywhere especially high in US News' own survey of graduate program reputations (which accurately measure the prestige accorded to different departments). Yet US News ranks it the ninth best school of the United States, one spot below Stanford and Caltech.

Having said this, the methodology of this new Chinese study is flawed as well. First, they measure the number of Nobel Laureates at a given university for 20% of the net score. This is a terrible choice because this number reflects very, very little. Nobel prizes are typically given decades after the research in question was done; they are also typically awarded for a single important discovery and tend to overlook researchers with patters of important discoveries in a variety of different subdisciplines. And not all fields have a Nobel prize associated with the field, and this overemphasises some disciplines at the expense of the others.

Secondly, the Chinese study measures the number of papers published in the journals Science and Nature. Again, this skews the picture towards certain disciplines that tend to publish papers in those journals. Mathematicians never publish papers in Science or Nature. Neither do engineers. Physicists rarely do. Its mostly biologists, chemists, and doctors.

Finally, the numbers used are net numbers, not scaled per number of faculty members (actually, this scaling is done but accounts for only 20% of the net score -- the other 80% depends only on the net values). This tends to reward schools which are big and penalize those that are small -- and I really can't see why a school with a large number of mediocre faculty is better than a school with a small number of exceptional faculty.

But despite these flaws, the ranking is useful. Its fun to look through. My conclusions: Caltech is the best university in the world. On a per-faculty-member basis, it beats the closest competitor (Stanford) by about 25% with a score thats roughly double the score of such universities as Oxford, MIT, and Yale.

Addendum: One other reason why this ranking is not accurate: it measures citations using the ISI citation index. This index overemphasizes English-language journals. Russian scientists, for example, publish papers in Russian journals; French scientists have their own French-language journals, and so on. None of these journals are included in the ISI ranking. The European press relase I linked to above notes with alarm that the best universities in the world tend to be English speaking and wonders why the quality of continental European education is so low; this is why. Its a flaw in the index.

To put in my two cents: in my fields (applied mathematics, electrical engineering) the best European universities are KU-Leuven and Ecole Polytechnique Federale de Lausanne, which have arguably better departments than any American competitors.