I have supported efforts to get more graphs in the social sciences — rather than tables with regression coefficients. This is not the place to reiterate all the advantages, but graphs make it easier for the reader and are useful for the author to make correct inferences.
There is, however, one reason we should continue making these numbers available as tables (e.g. in an appendix): meta-analysis. Yes, there are tools out there to get numbers from published figures, but that’s additional work that might discourage researchers from undertaking these valuable meta analyses.
Take-away message? Keep using these graphs (– use more of them!), but also put the numbers in an appendix. Especially in the age of online (first) publications, there is no excuse for not doing both.
Brian Gaines and Rein Taagepera have recently clarified how to measure two-partiness (or should that be two-partyness? It isn’t in my dictionary, so I went with the analogy with happiness, although both forms seem to be in use). They highlight that the effective number of parties (Neff) is inadequate as a measure of two-partiness: very different constellations can lead to similar values of Neff.
Interestingly, the suggest two distinct measures for measuring two-partiness (T, D2), and I have implemented them in my R package polrep, available on R-Forge. It’s probably worth reiterating at this point that Gaines and Taagepera do not suggest we stop using the effective number of parties in general, but that we use more appropriate measures if we’re interested in two-partiness.
Gaines, Brian J., and Rein Taagepera. 2013. “How to Operationalize Two-Partyness.” Journal of Elections, Public Opinion & Parties 0 (0): 1–18. doi:10.1080/17457289.2013.770398.
A few years ago Ken Okamoto-Kaminski and I did the exercises in Agresti & Finlay “Statistical Methods for the Social Sciences“. We did the exercises for which there is a key, and the document shows how we solved all these exercises. These are my notes, and everything is handwritten. The edition we used had a misprint in the answer for exercise 3.5.f (I’ve checked this with Agresti, so the misprint should be fixed in newer editions).
Agresti & Finlay Statistics Exercises