La Profesora Abstraída

Weblog of Michelle Dion, Assistant Professor in the Department of Political Science, at McMaster University. My blog has moved to Visit my other website.

Monday, July 03, 2006

Ok. Last post for the night just because I'm annoyed

Univision has the most clueless anchors. They keep reporting the aggregate vote percentages, but provide no analysis beyond "the gap keeps getting smaller, and AMLO is gaining."

A quick look at the disaggregated rapid count votes shows that different percentages of precincts have reported from different regions which have different tendencies.

How hard would it be to do a sub-national analysis using the available data?

I sure hope Mexicans in Mexico have access to better anchors than those that are stranded here with Univision.

Buenas noches. Espero que cuando me despierto manana, haya un presidente nuevo sin violencia o derrota del mercado financiero (pq ya ha bajado el peso....).

posted by Michelle @ 1:12 AM,


At 7/03/2006 2:27 AM, Blogger Chris Lawrence said...

It wouldn't be that hard, but I don't think it's regularly reported anywhere, although it is done internally as part of the projections the networks make in the US for the electoral college and to some extent in UK elections coverage (using the national swing to project uncounted constituencies).

My guess is that a pretty rough regression model based on readily-available census data would work to predict precinct vote shares during a "normal vote" period, particularly once you account for voting in other precincts that have reported.

At 7/04/2006 8:40 PM, Blogger Michelle said...

EXACTLY! Which is apparently what the Electoral Institute did election night when it said it could not conclusively determine a winner. That was based on the analysis of "5 experts" they have on staff running analyses of the rapid count. (The full count starts Wednesday.)

In any case, are there only 5 experts in all of Mexico capable of such analysis? Could the media not have hired any of the several economists or econometrically trained political scientists to do the analysis?

Maybe the networks are still learning...maybe next time they import some expertise if they can't find it locally.

At 7/05/2006 1:36 PM, Blogger MSS said...

The problem with doing these projections is that this election is very different from the previous one, and the short track record of competitive elections in Mexico. So, any regression model and sampling techniques should be treated with caution.

At 7/05/2006 5:39 PM, Blogger Michelle said...

Granted. But the news agencies could have done better than they did Sunday night. The anchors on Univision seemed 'surprised' that Lopez Obrador's lead began to shrink as more votes came in, when a look a the the distribution of missing votes might have implied as much.

Instead, they willingly let it seem like Calderon had a healthy lead most of the night, without enough emphasis on how few of the precincts had reported (especially early, when less than 40% had reported).

They could have done a better job educating the public and explaining that since it was so close and so few precincts reporting, that the result was uncertain. Whoever establishes an early lead will get some sort psychological edge among the people....(I'm sure Kurt Weyland could identify the cognitive process that I'm thinking of).

Univision did a lame job covering the election, IMHO, given the importance of the outcome.


Post a Comment

<< Home

Mexico City slideshow

Go to main page:

La Profesora Abstraída

About me:

Name: Michelle Dion
Location: Toronto, ON
View my complete profile
View my website

traducir este pagina

Previous Posts

Global Voices Online - The world is talking. Are you listening?