Monday, March 7, 2011

Less is More (Sometimes)

 Less is More - Ludwig Mies van der Rohe

Some of the most influential academic journals have very short (one-word) titles. To wit, Science, Nature, Physica, Biometrika, Geology, Circulation, Polyhedron, Endoscopy, Neuron, and Econometrica. Is there anything in particular that we should conclude from this observation? Well, recently, Schreuder and Oosterveld (2008) took a close look at the relationship between the rankings of 6,033 journals in a wide range of scientific disciplines, and the length of those journals’ titles. For their sample as a whole, and for journals in only five of the individual disciplinary groupings that they considered, they found that there is a significant negative correlation between the journals’ so-called “impact factors” and the number of characters in their title. The opposite result was obtained for the “Pediatrics” and “Urology and Nephrology” fields. It’s actually quite important to analyze data from different disciplines separately from one another, because we’re looking at figs and bananas here. By way of an example, in 2006 the average (highest) impact factor for economics journals was 0.8 (4.7), compared with 4.8 (47.4) in molecular and cell biology (Althouse et al., 2009).

So, what do we find when we look at economics journals in this way?

Kalaitzidakis et al. (2003) ranked the top 159 economics journals on the basis of citations in 1998 of articles that were published between 1994 and 1998. They were careful - self-citations were excluded and adjustments were made for different page sizes. Unfortunately, certain well-known journals (e.g., Journal of Finance, and Econometric Reviews) were excluded from the rankings for reasons beyond their control. In addition, given the time-frame for their sample, many recent journals such as the various Berkeley Electronic Press publications, and Econometrics Journal, are also omitted from their analysis. The data that I am going to be using can be found on the Data page for this blog, and the EViews code that I've used is on the Code page.

In the comments that follow I’m going to measure a journal’s title length by the number of characters (including embedded blanks). However, none of the results that I present below change qualitatively when title length is measured in terms of the number of words. Title lengths range from 6 characters (Kyklos: ranking = 81) to 68 characters (Journal of Economics-Zeitschrift für Volkwirtshaft und Socialpolitik: ranking = 66), with mean, median and modal values of 29, 28 and 26 respectively.  Figure 1 shows a scatter-plot of the data, with a Nadaraya-Watson non-parametric fit using the Epanechnikov kernel, and a least squares linear fit. The first impression is that there is really no relationship between title length and ranking – the slope of the linear regression line is -0.2331 (p = 0.454). However, the kernel fit hints of a possible dependency whose sign may depend on the range of the data that we look at.

Now, make sure that you keep in mind that journals with the lowest ranking values (e.g., 1, 2, 3…) are those with the highest perceived importance (impact factor). Table 1 provides a rank correlation analysis using both Kendall’s tau and Spearman’s rho statistics, with proper account being taken of the many “ties” in the data for title length. As well as working with the full sample of data, I’ve also considered several sub-samples, based on both the rankings and the characters data. As you might have anticipated from Figure 1, the signs of the correlations vary by sub-sample, and in general they are not significantly different from zero. However, for the “top 40” journals there is a significant positive rank correlation between journal ranking and title length. For this group, the better journals tend to have shorter titles, consistent with the general finding of Schreuder and Oosterveld (2008). The same result emerges for the 36 journals with the shortest titles (i.e., 6 to 20 characters). In the case of the lowest quality journals – those ranked below the “top 100” - there is a significant negative rank correlation between journal impact and title length. For this group, the better journals tend to have significantly longer titles. You reach the same conclusion if you consider the group of 24 journals with the longest titles.


Table 1: Rank Correlation Analysis
              Ranking vs. Characters                                 Characters vs. Ranking
Rankings
Kendall’s
Spearman’s
Characters
Kendall’s
Spearman’s

τb
ρ
[n]
τb
ρ
1 – 159
-0.0202
-0.0348
6 – 68
-0.0202
-0.0348

(0.71)
(0.66)
[159]
(0.71)
(0.66)
1 - 20
0.1233
0.1712
6 – 20
0.2410
0.3603

(0.47)
(0.47)
[35]
(0.05)
(0.03)
1 - 40
0.2032
0.2829
15 – 30
-0.0882
-0.1274

(0.07)
(0.08)
[88]
(0.24)
(0.24)
21 – 50
-0.0951
-0.1450
21 – 30
-0.0389
-0.0572

(0.48)
(0.45)
[61]
(0.68)
(0.66)
51 – 100
0.0486
0.0620
31 - 40
0.0948
0.1254

(0.63)
(0.67)
[39]
(0.43)
(0.45)
101 - 140
-0.2489
-0.3839
31 – 68
-0.0026
-0.0106

(0.03)
(0.02)
[63]
(0.98)
(0.93)
101 - 159
-0.1482
-0.2318
41 - 68
-0.2132
-0.3305

(0.10)
(0.08)
[24]
(0.16)
(0.12)
Note: 2-sided p-values appear in parentheses. Sub-sample sizes appear in square brackets.

Although we all know that we have to be very careful not to confuse correlation with causality (don't we?), I still think that we can safely say that there's something interesting going on here. Specifically, these results suggest a “U-curve”, or more correctly, a flat-bottomed “bathtub” line-segment relationship between economics journals' title length (on the X-axis, so to speak) and their importance (high impact factor = low rank number, on the Y-axis). While it's not altogether apparent when we look at the aggregate data, this relationship emerges when sub-samples of the data are subjected to my grueling rank correlation analysis. This finding probably won't attract as much attention  as the Kuznets’ curve or its environmental counterpart, but it's sure to be of great interest to authors and journal editors seeking to maximize their citations, and to academic economics departments that want to raise their profile in the profession and their funding base.

This “bathtub” relationship will undoubtedly sound alarm bells in the corridors of publishing houses as they assess proposals for new economics journals. The highly desirable title, Economics, is no longer available, having been shrewdly snapped up not that long ago by an open-access, open-assessment e-journal which thereby managed to “cover all of the bases” in one fell swoop. Even more recently the American Economics Association laid claim to the titles Macroeconomics and Microeconomics, albeit with an “AEA” prefix that they may wish to re-consider in view of my results. The publishers of the journal, SERIEs: Journal of the Spanish Economic Association, which was launched in 2010, will also no doubt ponder the merits of dropping the last six words of its title. However, there is hope. The title, Econometrica, has been spoken for since 1933, but to the best of my knowledge the more worldly journal title, Econometrics, is still available. Obviously, publishers should register their interest forthwith!

Finally, and in the great tradition of concluding academic papers with some conjectures regarding future research, one obvious extension of this posting is to undertake a similar analysis of individual articles' title lengths, numbers of co-authors and citation rates. I’m not claiming any precedence regarding this suggestion, so go for it!


References

Althouse, B. M., J. D. West, C. T. Bergstrom and T. Bergstrom (2009). Differences in impact factors across fields and over time. Journal of the American Society for Information Science and Technology, 60, 27-34.

Kalaitzidakis, P., T. Stengos and T. P. Mamuneas (2003). Academic journals and institutions in economics. Journal of the European Economic Association, 1, 1346-1366.

Schreuder, M. F. and M. J. S. Oosterveld (2008). Who ever said size doesn’t matter? The association between journal title length and impact factor. NDT Plus, 2, 126-127.

Quantitative Micro Software (2010). EViews 7.1, Irvine, CA: Quantitative Micro Software.







© 2011, David E. Giles

No comments:

Post a Comment