Ciao!

MPSA preview

03.12.2005

I've run a few test models on my Bolivian election data, and this is a preview of some of my findings. The following table shows cross-panel, time-series regressions of province-level election data. Also, Chuquisaca provinces get dropped from the models since their "media luna" cells are empty (frankly, I can't decide whether to code Chuquisaca as "media luna" or not, ideas?).

Only the numbers followed by (*) or (**) are statistically significant. The constant isn't a variable, it's just a residual generated from most statistical analysis, and is almost always statistically significant (if it's not, it usually indicates a problem).

The between-province panel estimates test for variations between provinces, over time. The dependent variables are features of the party system: effective number of parties (ENPV), the percentage of blank & null votes, electoral volatility index, and votes for systemic parties (MNR, ADN, MIR). Notice that the most predictive variable to predict structural differences in provinces' party systems is the "media luna" dummy variable. The department's effective threshold (the estimated % of votes needed to win at least one seat) was also frequently significant.

Basically, "media luna" provinces (when controlling for all other variables): have fewer political parties (meaning less party system fragmentation) by a large magnitude (reducing ENPV by one whole party); tend to cast fewer blank & null votes (the difference is a pretty big 5%); have much smaller electoral volatility (14.35% of voters fewer change their party vote between elections); tend to support the three systemic parties (they vote for them 23.44% higher than in the rest of the country). Basically, media luna provinces have fewer parties, concentrated on the three large systemics, and voters tend to have stronger party identifications.

In the within-province regressions (which test for variations w/in provinces over time), the regional variables drop out (for obvious reasons). But, here, the effect from the change to multi-member proportional (MMP) electoral system become more interesting.

I used two MMP dummy variables, since the electoral system change's effects might be more pronounced in the second election, rather than the first. In 2002, w/in province differences were: a reduction in number of parties (0.3 parties), an increase in electoral volatility (12.66%), and a dramatic reduction in votes for systemic parties (23.28%). In 1997, w/in province differences were: a small reduction in blank & null votes (1.91%), a slight reduction in electoral volatility (4.61%), and a decrease in votes for systemic parties (9.52%).

So, basically, adopting MMP accentuated already-existing regional cleavages, and strengthened the creation of locally-entrenched party systems. The result (if you've seen Bolivia in the last several months), is tragically obvious. At least two of these parties (MAS & MIP) have little incentive to moderate their politics, since they're almost completely entrenched regionally (especially MIP, which is entrenched only in the rural provinces of La Paz). Why? Because MMP means these parties have to win very few votes (even in individual provinces!) to be guaranteed a relatively large number of seats in the national legislature. And wooing votes outside specific constituencies (to build broader, multi-constituency parties) is discouraged, since the returns at the national level are slim (winning fewer, if any new seats) but risk winning the local seats (which other, more radical parties may take over).

More later. Or, rather, I'll just post a link to the conferences paper once it's finished.

-----
ADDENDUM: This is an extension of some very preliminary data I compiled about a year ago here & here.

Posted by Miguel at 06:46 PM

Comments

Posted by: eduardo at March 12, 2005 07:43 PM

Posted by: Miguel at March 12, 2005 07:54 PM