Are leading papers of better quality ? Evidence from a natural experiment

European countries in which universities rely on public funding increasingly follow the lead of the United Kingdom and run Research Assessment Exercises. Given the subjective nature of such evaluations, some scientists prefer verifiable measures such as citation counts. This, however, also is prone to problems since the number of cites is correlated, among others, with the order of appearance in an issue. In particular, leading papers are more cited. It is, thus, difficult to assess whether they are of better quality, or whether this happens because they appear first in an issue. We make use of a natural experiment that was carried out by a journal in which papers are randomly ordered in some issues, while this order is at the editors' discretion in other issues. Our estimates suggest that approximately two thirds of the additional cites are due to going first, and one third to higher quality.