I have been teaching my elections class this semester, and we are at that point when the students have to calculate the “effective number of parties” in a number of different countries.  The effective number of parties is defined in a number of places; the assignment has them update tables from Taagpera and Shugart’s book Seats and Votes, Ch. 8.  The ENP is an attempt to estimate, in a way comparable across countries, what Sartori called the number of “relevant” parties.

The calculation is pretty simple: you take the percentage of votes a party received, square it, and sum across parties. This number has the elegant name of the “Herfindahl-Hirschman concentration index”.  Take the reciprocal, and you have the effective number of parties.

For the assignment, I ask students to update Taagpera and Shugart for a set of countries.  Since the book was published in 1989, eastern Europe is one of the most interesting areas.  I generated some figures myself for Latvia using the very helpful archive of constituency level results (CLE) provided by Dawn Brancati of Washington University.

Here’s my question for Thad, though: I am pretty sure I found some errors and anomalies for Estonia.  The errors are in the district names (for example: “Voru maakond, Valga maakond, Polva maakond” is listed in 1992 and “Voru-,Valga, Polva maakond” in 1995–I am pretty sure these are the same).

But is it the case that the magnitude changed in many Estonian electoral districts from 1992-2003?  The aforementioned “Voru-,Valga, Polva maakond” is listed as M=6 (1992), 5(1995), 5(1999), and 10 (2003).