I’m a political science professor who posts regularly on Bolivian politics. I also occasionally discuss interesting books, pop culture, and daily life with a toddler. I’ve recently moved to Oxford, Mississippi.
MABB & I are presenting a paper at this weekend’s MPSA conference in Chicago. Our paper is titled “The Political Economy of Bolivia’s New Regionalism: A Look at Electoral Patterns in Santa Cruz, Tarija, and Chuquisaca” and uses municipal-level election, socioeconomic, and demographic (ethnicity) data to test the conventional wisdom explanation for regional autonomy movements in the Media Luna.
We found that differences in degrees of “ruralness,” poverty, and “indigeneity” across municipalities had no significant relationship (except in Chuquisaca) with support for Evo Morales, MAS, pro-autonomy prefect candidates, or regional autonomy. You can read our paper here; we appreciate any comments or feedback.
Did you guys use a simple linear regression? My gut tells me that a probit model might be a more correct approach for predicting the probability of being a MAS supporter (or autonomy supporter) givin a list of characteristics....but don't take my word for it, I work in finance not statistics, after all the stuff after least square is really a blurr.
Posted by
chasqui
April 2, 2009 10:46 AM
Originally we were going to use panel-estimated cross-sectional time series, but we decided to take a simpler more direct approach. Probit (and logit, too, I think) would work well for binary data, which we don't have. We're not looking to predict whether *individuals* support Evo/MAS/autonomy. We're looking to predict whether municipal-level, aggrgated vote differences are statistically related to municipal-level socioeconomic/demographic differences. But I could be wrong about probit.
With due respect, the premise of this paper is fairly silly.
Your "systematic test" is an analysis of public perseption and support of the "Autonomy" movement, not the actual political motives behind the movement for which you would also have to analysis key documents (such as the New Constitution, and Autonomy Statutes), the pro-Autonomy leadership group, historical context, and key resources-interests backing the "Autonomy" campaign. As a political scientist you should be well aware that distance always exists between actual politics and public perceptions for a variety of very complex reasons. An economistic or purely statistical analysis will never fully illuminate these dynamics.
Posted by
D.Beaulieu
April 7, 2009 12:43 PM
You're absolutely right. Our paper is NOT testing the motivations for *why* people vote as they do. We are ONLY looking to see if there are any correlations between socioeconomic/ethnic variables and vote shares across municipalities. As such, we found no correlations. The conventional wisdom (that Evo is supported by indigenous, poor, rural voters and that autonomy is supported by white, middle-class, urban voters) isn't supported by our statistical analysis. As you point out, it is clearly more complex. We only meant to test an untested conventional explanation. We don't pretend to know *why* voters vote as they do.
Ideally, we would have a large survey to look at. Sadly, there are no such surveys available. We also originally intended to look at differences in fiscal spending by the central & department governments, but that data was unavailable to us. If you have any suggestions for other variables to include in our models, or better ways to specify our models and/or operationalize our independent variables, we would welcome them.
Ok Miguel, then you need to rewrite your abstract and parts of the paper. Opening line:
"While most conventional accounts assume an economic motivation behind the new regional autonomy movement, such claims have not been systematically tested."
You need to be very specific by what you mean by "motivation" and whose "motivations" you are specifically discussing in refence to the larger "regional autonomy movement". Your paper does not for instance discredit the claims made by Alvaro or other Bolivian intellectuals that the "elite" or "oligarchic" leadership of the Autonomy movement has economic motivations for/behind their political position.
Posted by
D.Beaulieu
April 7, 2009 5:40 PM
Fair enough. But we weren't really trying to get inside the black box. We were just trying to predict votes to see if the observable implications of claims that socioeconomic and/or ethnic cleavages drive support for Evo/MAS or autonomy at the polls. Perhaps we weren't very clear about that.
If you have any *specific* advice for other variables to include in our models, or better ways to specify our models and/or operationalize our independent variables, we would welcome them.
fyi: Bloomberg is quoting a Ipsos/Apoyo poll reporting that Evo's support is down to 40%
Posted by
chasqui
April 9, 2009 4:01 PM
@mcentellas: So the biggest determinants to how you vote are: ability to read and department in which you live. Not very important are: rural vs urban, poverty, and ethnicity. Given these results, what would constitute a more effective campaign strategy for the opposition?
@chasqui: I thought that poll indicated his approval was 49%.
@dv: Frankly, we're a bit puzzled by the impact of literacy. Clearly, literacy could serve as a proxy for class (but one that might cut across the ethnic or urban/rural divides). We also suspect that it has something to do w/ access to media (the newspapers tend to be more sympathetic to the opposition, while the MAS government dominates the radio). We also can't say definitively here that department "matters" (per se), but rather that ethnicity (in our models) only matters in Chuquisaca. We want to look at the data more carefully again, and then perhaps also toss in some "Andean" departments (particularly La Paz & Oruro) to see if there are region-specific effects.
Also, it does look like Bloomberg incorrectly cited the poll numbers. Evo has 49% approval ratings (a 6% drop from previous Ipsos poll see Ipsos Apoyo poll).
Edit: Bloomberg got the numbers right. Evo is at 49% approval, but only 40% of respondents would vote for him in next election in a separate electoral poll. Around the world, approval ratings are typically higher than vote intention numbers.
