I've been thinking of changing the formula under which the impact factor is calculated for the most-cited list of legal periodicals. Comments welcome if anyone has an opinion. From the vantage point of November 2004 (when I do the 1997-2004 calculations) I cannot determine the number of articles published in each journal, simply because not all of the 2004 articles have been indexed or included in full-text databases as yet. I can only reasonably accurately count articles through the end of 2003, yet the recorded number of citations to each journal includes cites to 2004 volumes, as cited up through October 31st. 2004. So an impact factor calculation of, (cites to 1997-2004 articles)/(number of 1997-2003 articles published by the journal), doesn't quite mesh. So a few articles are added to the denominator by adding 10/12th of the average number of articles published in the previous 7 years. There's a reasonable argument though, for not bothering with the formula adjustment because the impact factor of the current year's volume is very slight. For example, looking at Columbia Law Review (from the vantage of Nov 1, 2004) the impact factors of each volume are as follows (with the number of 2004 Columb. L. Rev. articles estimated):
Year Cites Articles Impact-factor
1997 1252 62 20.2
1998 1090 38 28.7
1999 983 51 19.3
2000 910 54 16.9
2001 588 54 10.9
2002 461 67 6.9
2003 257 52 4.9
2004 39 43 0.9
Because the 2004 volume has had so little time in which to accumulate citations its impact factor is very small. It would be expected that the oldest volume in the cycle would have the highest impact factor, as the article count accumulates each year. It would be feasible to come up with an average expected increase in impact factor and thus to declare a deviating volume good or bad (in this case plainly either 1997 was a bad year, or 1998 was a good one, citation-wise), not sure who'd be interested in that though, other than the then editorial board. Anyway, the 39 cites to 2004 articles represents less than 1% of the total cites to 1997-2004, and calcuating the impact factor simply by (1997-2004 cites)/(1997-2003 articles) seems reasonable.