I am trying to calculate the effective number of parties, based on seats (not votes), for several countries. Currently, I am stuck on the question of how to deal with the residual "other" category often found in seat data. Usually, this is not a big problem, since it is usually only a handful of legislators, and so I have regularly treated it as a party and input it directly into the Taagepera and/or Golosov measures. However, I have found some cases (e.g. Brazil 1962) where the "other" seats account for nearly half of all seats. Clearly, not making an adjustment would artificially deflate the ENP value. But is there a preferred solution? I haven't been able to find one. Thank you.
Good question! I would suggest to disaggregate the category into parties from other sources and to treat unaffiliated/independent MPs as "individual" parties as well... Other opinions?
But isn't the whole point is that in other cases these parties are indeed minor--and in the Brazilian case there are many?. Here they are a sign of the proliferation you are actually trying to measure. So what's wrong with counting all of them? If that does create a problem, the other solution is to lower the cut off # of seats a party has to consider it significant.
Exactly, Peter. I've been using the Taagepera solution (which is to calculate the function for the "remainder" or R using a formula that calculate the logical extremes for N and taking the average). The problem is that Taagepera uses R as a residual for all others. But what if we know that there are two (or more) categories of "others". That gives us more information, but not enough to identify the parties.
I should add that this is for a project where I'm trying to calculate ENPS (using seats, not votes) for countries in Latin America for 1900-1974 (prior to the third wave). Often, there is no seat data available for many countries, and often there is a lot of "other" and/or "independents" listed.
I'm attaching a link to the Taagepera article that offers a partial solution.
http://www.sciencedirect.com/science/article/pii/S0261379497000036
Ideología, orientación económica, afiliación a organizaciones políticas internacionales. Una característica que es interesante en el desarrollo de las políticas públicas tanto en actos de gobierno o en políticas públicas propuestas - Proyectos de Ley, Propuestas. La posibilidad de cuantificación e identificación de las características propuestas podrían agrupar a los partidos políticos de maneras alternas.
Ideology, economic orientation, membership in international political organizations. One feature that is interesting in the development of public policies in both acts of government or public policy proposals - Bills, Motions. The possibility of quantification and identification of proposed features could bring together political parties in alternate ways.
You probably have already read a paper by Golosov introducing an alternative to the Lasso and Taagepera method. I developed an R function on his formulas that can compute the index if you are familiar with R package. I did so, after hearing a college saying he did spend quite few time to figure out how that works. I wrote a post in my blog about the function: https://danielmarcelino.github.io/r/2014/10/06/A-bit-more-fragmented/
Thanks, Marcelino. I think you mean the basic Golosov adjustment to ENP, right? That works well (and I'm familiar with how to use it). But it doesn't give an explanation for how to deal with a large "other" category.
Too often, posted election data treats votes/seats for minor parties as simply a lump residual "other," so it's impossible to know how many parties are in that category (it could be anywhere from 1 to N, if each vote/seat for "other" is for a specific individual party/candidate).
Laakso & Taagepera introduced a method for dealing with "other" by simply estimating the extremes ("other" = 1 and "other" = N) and using that function to estimate a midpoint between the logical extremes for ENP.
My question is whether this is now the accepted method, or whether there is a newer/better way. Golosov's formula for ENP would not work well for estimating ENP with large "other" residual seats/votes.
I see. In the worst case scenario that you cannot distinguish between others+ from others-, I still think Golosov's approach can handle slightly better than others at estimating the effective number of parties as both larger and smaller blocks of parties are not attributed unrealistic scores--or at least in a lesser extent.
I agree, overall, and prefer Golosov. Ideally, however, there should be public reporting of the seats won by each individual party in the legislature. I'm surprised how difficult that is to find, especially for pre 1940s data for most countries in Latin America.
Daniel Marcelino: thanks for pointing my attention to this. You made me realize that there is a problem here, so I wrote this little piece: http://golosovpubs.ucoz.com/index/indices_of_the_effective_number_of_parties_and_party_system_nationalization_in_excel/0-13
Grigorii Golosov: I'll add a link to your thorough piece to the references of the R function 'politicalDiversity' the next time I submit a revision to the repository. Thank you.
The Simplified Version of the Effective Number of Parties Formula looks like this:
N=1/[V12+V22+V32+V42]
You replace the Vs with the Seat Total for each of the first 4 parties, then you divide the seats by the total number of seats in the elected house of the legislature or parliament. You should get a decimal for each party, which you then you square or double the decimal for each party. After that, you add those four decimals together and divide the 1 by the final decimal. Then, you will have your N or Effective Number of Parties.
I know it has been a while since this question was made but CSES has a "Corrected Effective Number of Parties" that deals with the "other" category. It works to take a look.