Tuesday, June 24, 2008

Nice but irrelevant study: The effect of polling locations on election results.

...while the following post is incredibly dry, it does have the redeeming feature of being non-partisan.

That may be its *only* redeeming feature. :)

The Arizona Republic has an article on a study conducted in Arizona on the impact of polling location on the likelihood that a voter will support an education funding initiative that's on the ballot.

The upshot of the study was that 56% of voters casting their ballots in school supported education measures, while only 54% of those voting in other public buildings did so.

Ignoring for the moment the statistical similarity (56% to 54%? 2% is within the margin of errors of most statistical analyses of election trends, but I digress :) ), what the Rep article notes, but fails to note the significance of, is the fact that the study analyzed data from Arizona's 2000 elections.

In short, while the study may end up serving as a good snapshot of voting trends at the time (or, given the 2% difference, a good hint of a suggestion of a snapshot), changes in voting activity is just 8 short years render the study nearly useless.

In 2000, a negligible number of voters cast their ballots early; so neglibile, in fact, that the Maricopa County Recorder doesn't show the number in the results listed on its website (no direct link; use the drop down menus on this page.)

In 2006, more than 49% off all votes cast in Maricopa County were cast early or by mail.

In 2008, the trend has continued. During the various municipal elections held so far, early voting has accounted for approximately 80% of ballots cast (i.e. - Tempe's early voting percentage for the general election last month was over 82%).

While the early voting numbers for this November's election probably won't be quite so high, given the trend and the parties' efforts to persuade voters to use vote-by-mail, the percentage of early votes should be well over 60% or even 70%.

In short, the effect of location of polling place, while minimal to begin with (2%!) has been rendered all but moot -

There are an average of 1370 voters per precinct in Maricopa County (a smidge more than 1.5 million voters total). In 2006, voter turnout was approximately 60%; assuming a similar percentage turnout this year, that would mean an average of 822 voters casting ballots in each precinct, and assuming 60% EV, that would leave 329 voters physically casting their ballots at each polling place.

2% of 329 is less than 7 (6.579).

Even assuming that all 1142 precincts in the county cast their ballots in a school (they don't), that means the impact of voting in a school is approximately 7500 votes county-wide.

Most school districts (hence, most school-related ballot questions) don't have nearly that many precincts (Scottsdale Unified #48 has 97).

In other words, while the effect of polling place location does exist, it was small to begin with (2%), is minimized by the fact that most education-related ballot questions cover relatively small areas, and is already shrinking due to the growth of 'vote-by-mail' behavior.

Perhaps the scientists who performed this study, Jonah Berger, Marc Meredith, and S. Christian Wheeler, should do a follow up, studying the effects of vote-by-mail (aka "Vote-in-the-comfort-of-home) on voting patterns.

An abstract of the study is here; the article is here, courtesy the website of Dr. Jonah Berger, one of the authors.

What can I say - I'm a numbers geek. Boring is part of the package. :))

Addendum - I emailed Dr. Berger the question that I asked, regarding whether he and his colleagues planned a follow-up regarding the effects of VBM. His reply was surprisingly prompt (I emailed him at his school email addy, after 7 p.m. local time, during the summer. I wasn't sure that he would reply within a month, much less an hour. :)) ). As of right now, they don't have any plans to do a follow-up on VBM effects.

No comments: