Anyone propose a (different) solution?

04.07.2004

I'm in the midst of configuring data on political parties for a paper. And I've concluded that I need a better way of handling pre-electoral coalitions. For example, a common measure is the "effective number of parties" (N) developed by Laakso & Taagepera. Normally calculated as a sum of squares (where si is the % seats for the i-th party):

The formula provides a measure for how many political parties exist in a given political system at any given time (per election). What this means is that if you have a party system w/ eight parties, but four of them win 20% of the vote each (80%) & the other four win 5% each (20%), you don't have eight, but rather 5.9 "effective" parties. The US usually hovers between 1.8 to 2.1 "effective" political parties.

The dilemma's that Bolivian parties sometimes form pre-electoral coalitions. I've yet to find a solution that fully convinces me. Any solution must involve estimating the % of votes for each part of the coalition (e.g. 1997's ADN-NFR-PDC coalition). But. Though I've a few ideas on how to do this — the one that most convinces me is to estimate the vote share for each part of the coalition & disaggregate vote shares into the coalition's component parties — I wonder if there's a "preferred" solution. I've yet to find one. Specifically, I'm looking for less "subjective" solutions that can be transported across cases. Any ideas?

And. Should I carry that over to variables like the "fractionalization index"? Which, though similar to N, measures how fractionalized voter choices are in any given election. Suggestions?

I'm confident I need not adjust measures like the Gallagher "index of disproportionality" (used to measure the proportionality of votes to seats) or the Pedersen "electoral volatility index" (used to measure shifts in electoral choices). Although. If a party campaigned one election & merged w/ another party the next, there's an exaggerated measured volatility.

Essentially, I'm trying to measure trends in party system stability (or instability). It seems clear that Bolivia's party system was in a process of consolidation through 1985-97. A process that reversed between 1997-02. But I want to measure this precisely, using standard variables.

-----
ADDENDUM: A potential problem w/ calculating N by breaking up coalitions into component parts is that it suggests dis-consolidation, by increasing the effective number of parties. One could argue that coalition formation is itself a way of consolidating a party system into fewer choices (the literature on party system consolidation tends to favor fewer parties).

Posted by Miguel at 02:58 PM

Comments

Note to self:

http://www.ics.ul.pt/publicacoes/workingpapers/wp2003/WP3-2003.pdf

http://www.cbrss.harvard.edu/events/ppbw/papers/Ferrara.pdf

http://homepages.nyu.edu/~sln202/rules.pdf

Posted by: miguel at April 7, 2004 06:58 PM

Ok. This entire post/thread is mostly just me thinking out loud, but feel free to jump in.

One of the problems w/ calculating pre-electoral coalitions is that some of them are between important parties, some not. Example, ADN often campaigned w/ PDC. But PDC (Partido Democratico Cristiano) probably at best wins 1% of votes. On the other hand, the ADN-NFR-PDC caolition included NFR, which wasa major regional party and came in third in the next election. So. Estimate maybe 1/3 of total coalition votes? Less? More?

Then you have MNR-MRTKL (1993) & MNR-PS-MBL (2002) where the MRTKL & MBL components probably rated 5% of total national vote? Less? More?

Or should I just forget it, and count each pre-electoral coalition as a single party? That'd be easier, but it washes out lots of information.

Posted by: miguel at April 7, 2004 07:11 PM

Why don't you use the crtiteria for party system consolidation proposed by Mainwaring Scott?
See Mainwaring, Scott, “Party Systems in the Third Wave”, Journal of Democracy, vol.9, no.3, july, 1998,

Posted by: Bianca Buligescu at July 24, 2004 09:24 AM