Hi Erwin,
Neat, but speaking of assumptions, how about autocorrelation? Isn't it
simply the case that if a journal's two-year IF is rising sharply (AMJ
recently), then its lagged (five-year) IF will trend lower; conversely, if
an IF is dropping (ASQ recently, to an extent), predicting from older data
will make things look better?
Put another way: The JCR data are historical, so if anything it is more
meaningful to predict the two-year IF from the five-year data (with the
obvious violation due to partly overlapping data). In which case the story,
messy interpretation and all, is set to reverse itself. Of course, in time
we will have ready-made data from JCR to provide a proper lag structure.
Finally, note that the IF is computed from publication date (with the
roughness of yearly aggregation), so the time to publication is not an
issue. Indeed, a "perverse benefit" of long publication lags can be that
forthcoming papers have a chance to gain visibility as such and get built
into future citing papers. Nowadays, early attention is facilitated by
preprint and advance online publication systems, and the commercial
publishers (hence SMJ, JIBS) look to have gained a bit of an edge from this.
Xavier Martin, Tilburg University
-----Original Message-----
From: Business Policy and Strategy List
[mailto:
BPS-NET@AOMLISTS.PACE.EDU]On Behalf Of Danneels, Erwin
Sent: Friday, July 24, 2009 00:08
Subject: Five versus two year journal Impact Factors
Dear Colleagues:
I have long thought the journal impact factors calculated based on two-year
prior citations (e.g., the score for 2008 includes citations to articles
published in 2006 and 2007) were dubious. In our field publication from
first submission often takes two years, and some really good papers are
late-bloomers.
So I was curious when I learned that there are now (since 2009) five-year
impact factors produced by Journal Citation Reports, and I ran some
preliminary analysis comparing the scores and rankings of journals using the
2-year versus 5-year criterion. I combined the lists for the "Business" and
"Management" fields arriving at 123 journals with data for both 2-year and
5-year impact factors.
Their Pearson correlation is .91, the Spearman’s rho is .927, and Kendall’s
tau-b is .774, so they are extremely highly correlated. There some very
interesting exceptions though, as noted from changes in ranks and residuals
from the OLS prediction of the 5-year impact factor from the 2-year impact
factor (see attached Excel sheet).
There is one prestigious journal that has a lower long term impact that what
would be expected from its short term impact: AMJ. There several prestigious
journals that have a higher long term impact that what would be expected
from their short term impact: MISQ, ISR, ASQ, JM, SMJ, and Org. Sc.
I warn the purists that normality and homoscedasticity assumptions are
(somewhat) violated; they may wish to look only at the rankings and their
correspondence.
Erwin.
Erwin Danneels, Ph.D.
Associate Professor
www.mgt.wpi.edu/People/Danneels<http://www.mgt.wpi.edu/People/Danneels>
www.wpi.edu/News/Transformations/2007Summer/investigations.html<http://www.w
pi.edu/News/Transformations/2007Summer/investigations.html>
(if double-clicking on these links does not work, cut and paste into your
browser)