The big news this month is that we have now surpassed 700,000 bibliographic items listed in RePEc. The last 100,000 additions took only 7 months, something that seems difficult to beat. The other news is that the ranking methodology has changed following the discussion on this blog. Essentially, authors with multiple affiliations are now treated differently, as they contribute with variable weights to each of their affiliations, and these weights are used to distribute their scores across the countries and regions they are ranked in. Also, a NEP report for the Arab World was created. Finally, we experienced heavy traffic on our sites, with 766,586 file downloads and 2,756,978 abstract views in January 2009.
New participants in RePEc are: Universität Augsburg, Universität Bielefeld, World Agroforestry Centre, University of Craiova, Centre of Financial and Monetary Research “Victor Slavescu”, Hokkaido University, Kobe University, Romanian National Institute of Economic Research, C.D. Howe Institute, Central Bank and Financial Authority of Ireland, University of Virginia (II), Danubius University, INRA-Nancy, Utah State University, University of Applied Sciences Düsseldorf, University of Resita
With regard to major thresholds reached during the month, we can report:
4,000,000 extracted references
700,000 items listed
500,000 abstract views through Socionet
350,000 online articles
125,000 cited articles
100,000 cited working papers
50,000 subscriptions to the NEP email notification service
3,500 series and journals listed
I understand the reasons for changing the ways author’s work are attributed to institutions. But the way this feeds into the individual rankings makes the latter essentially useless. See the example below (mine) in which case I rank 1482 in the “Europan Union” but 287 in the broader category “Europe” (which included the European Union). This makes no sense of course.
Norway (you rank 5 of 160, top 4%)
Sweden (you rank 77 of 313, top 25%)
European Union (you rank 1482 of 9311, top 16%)
Europe (you rank 287 of 10305, top 3%)
The only way I now can contribute to avoid this nonsense is to delete my secondary affilation (in neigbouring Sweden). That would be a pity, because it would lead potentially useful and relevant information not beingn available.
Indivudual rankings should be based on documented scientific achievements. If you want ro regionalize those then the best (and certainly most simple) way would be to make extracts from world rankings (in which case everything would be consistent across different geographical groupings).
If multiple affiliations is a problem I think using the emailaddress to check for the author’s geographical base seems the best option in the short run.
In the longer run authors should be asked to declare their main affilation in which case this problem would disappear.
Congratulations for the work done in repec. The repec rankings with the new method are much better than the previous situation. For example, now the ranking of institutions in Europe looks quite reasonable.
I could imagine that some authors with affiliations in different countries are not happy because they have gone down in within country rankings. One way to correct that would be to locate each author in the country for which he or she has the highest weight according to the method developed in repec, and give him or her full credit for his or her points in that country, and zero elsewhere. This method could be applied for within country rankings only, not for institution rankings.
David de la Croix
A response to these two comments. There has been a lot of reaction through email, a manifestation that it is impossible to improve in a Pareto efficient way. I wish people would rather comment here.
Jan, you seem to be a rather odd case indeed, and there is no way to fix this well apart from Norway joining the EU. But your Swedish affiliation is secondary and treated as so. Indeed, Half of the weight is attributed by default to the institution the closest to your email address (looking at the IP domains). And I agree authors should declare their own percentages across affiliations. But for that I need resources or a volunteer to modify the code in the RePEc Author Service.
David, thanks for the encouragement, I really need it… Your solution is indeed an interesting one, but my principle is that a person cannot be more than 100%. Up to this month, some authors were 500%, that is counting fully in 5 different affiliations, and up to four different countries. That was really bad and reduced the credibility of the regional rankings. I think it is better now. But so far I want to go by the principle that a person is worth 100% in total.
That said, one way or the other, somebody will be unhappy. I try to stick to some rule and introduce as little subjectivity as possible, But I welcome any volunteer willing to modify the RePEc AuthorService code to get these shares trully author driven.
But didn’t David suggest exactly what Christian is looking for :
“One way to correct that would be to locate each author in the country for which he or she has the highest weight according to the method developed in repec, and give him or her full credit for his or her points in that country, and zero elsewhere. This method could be applied for within country rankings only, not for institution rankings.”
This would mean that an author’s work would be attributed to institutions following Christian’s rule (and this would sum to 100% for each author).
Following David’s rule it would also sum to 100 % within and across countries and country aggregates because the work would only be attributed to one country.
Seems like a good compromise. My paradox would dissappear!
Jan, I wish all those sending me comments by email would post them here. Then you would see that many would not be happy with that solution. Indeed, some a genuinely splitting their time between places and are in fact unhappy that I give more weight to one rather than the other.
What I take from this is that it is impossible to please everyone. The new scheme has at least the advantage of fixing the issue that some people counted for more than 100%. The optimal way would be to let authors distribute themselves these 100%. But this requires major work I am unable to perform.
I have noticed that new CEPR Discussion Papers have not been added to RePEc for more than half a year. I know that this is a problem of CEPR and not a problem of RePEc. But nevertheless, since CEPR is one of Europe’s biggest providers of discussion papers to RePEc, maybe you could send them a reminder so that they fix their problem.
We are aware of it and many authors have already complained about this. We have notified the CEPR about this several times, but so far the data has not been updated yet. While decentralizing the data input to publishers allows RePEc to run at no cost, it has the disadvantage that we completley rely on publishers…