Wednesday, November 14, 2007

You want more turnout?

I know I said I'd expand back out from the Bexley focus post-election, but this is excellent practice for the coming year. I had a commenter ask for precinct-level data, and if the commenter is a long-time reader, they probably knew I was already planning on it. Today's post is race-neutral background, I will add in analyses using actual vote data from the mayoral and council races later.

1) Precincts. I apologize for the quality of the map below. What started out as a map of Bexley was jigsaw-cut before I received it as 12 separate files. You can look at the individual pieces here, but for general reference, the map below should suffice:
If anyone has access to a better digital precinct map, I'd love a copy.

2) Turnout. When looking at turnout, you are always comparing one set of numbers to another set of numbers. The most common comparison is '# of votes' to '# of registered voters'. There are two main problems with the basic comparison, one for each set. The problem with determining '# of votes' is that even in a county like Franklin, which does a relatively excellent job of making the information available, you can almost never determine how many votes came from voters in a given precinct. The easy part is the number of votes cast at machines at the polling place on regular ballots. Then there are the votes that were cast at the polling place but on provisional ballots. These will not show up in the initial unofficial canvass. The ones that are counted will show up in the official canvass, the ones that are rejected, for whatever reason, may or may not be reported elsewhere. Then there are absentee ballots, which can be cast on machines at the Board of Elections HQ, mailed domestically, or sent from overseas (there may be additional, rare methods, such as accomodations for people hospitalized on election day, but we'll let those go). The domestic votes are tallied by election night, and the overseas ballots have extra time to arrive (as long as they were mailed prior to the election). These votes are often combined, but accounted for in their own category, separate from votes tallied at polling places. If you look at Franklin County data from past years, for instance, you will find that after all of the precincts are tallied, an extra line labeled 'absentees' will have a ton of votes which are added to the total. You can narrow down where those votes came from by looking at individual races - for instance in 2003 if there were 10000 absentee votes cast in Franklin County, you could look at how many votes were cast by absentee ballot in the Bexley mayoral race, and that would give you a decent (but most likely low) estimate of how many absentee ballots came from Bexley. You cannot, however, get precinct-level information from this except in rare cases where very local issues (some liquor licenses, for instance) come up. I suppose that the BOE might give you this info if you went down there and asked, but if you're willing to do that I've bored you to tears already... Anyway, this year the BOE has allocated the absentee ballots back to the precincts where the voter was registered, which is great, but it means that the numbers are not entirely comparable to past years' numbers. I will do so anyway (compare them, that is), but be aware. Oh, and this year's numbers do not yet (as of this writing), include accepted provisionals or all overseas absentee votes.


Then, there is the '# of registered voters.' This number is highly suspect, and is almost always higher than it should be. Some of those disgustingly low turnout figures you hear, especially by demographic, are due in part to this distortion. See, when a person moves to a new state, or dies, or commits a felony and goes to jail, they are no longer eligible to vote in their former precinct. It will take a while however, often years, before they are removed from the rolls. If you register to vote at a new address in Ohio, word will hopefully get back via the Sec. of State that you are no longer claiming to be eligible at your old address. But if you move to California, well, they've got better things to do. If you die, well, your loved ones have more important things on their mind, etc. And if you're just a young adult living from lease-to-lease and voting only in presidential elections, you probably won't get around to re-registering until next fall. So, in areas of high residential turnover, especially rental neighborhoods and college campuses, and in areas of high concentrations of senior citizens, there tend to be a lot of 'registered voters' who don't actually exist. In these areas, even if every adult resident came out to vote, turnout would look like 75%.

Having said that, I've pulled turnout numbers from fall of '06 (the most recent general election) and fall of '03 (the most recent Bexley mayoral election) and used them as baselines to test the assumption that "turnout was down" or that "turnout was light in South Bexley."

What we're doing here, first, is comparing the basic comparison ('# of votes' divided by '# of registered voters') across years. Most of the problems I've listed tend to stay constant from year to year, so comparing precincts to their past data is better than directly comparing precincts. Second, we're comparing the raw '# of votes' (remember why this may or may not be a great idea) from year to year, in part to compensate for the flurry of registration in Franklin County prior to th '04 election (which can make the same actual turnout in '03 look larger).

As you can see, turnout is lower across the board in an odd-year election. Furthermore, the majority of precincts are showing lower turnout numbers in '07 than in '03(remember, this includes bonus absentees in '07, but bonus provisionals/recounts etc. in '03 and '06). So, yes, it would appear that turnout was down. Was it particularly light in South Bexley? Well, it certainly wasn't confined to South Bexley, but to better make a determination, I've charted the drop from '06 (the peak value, treated here as the maximum expected turnout) to '07, both in terms of % and raw # of votes:

So, yes, the biggest drop by percent and number of voters was in my home precinct, 3-B. Turnout also dropped by quite a bit in 4-C, also in South Bexley, but not as much as in North Bexley Precinct 1-A. Even in Central Bexley 1-C, turnout dropped more than in South Bexley 3-A. The big story, then, is not where turnout was down, because that happened in all three areas. The story is where turnout was high: 2-A,2-B,4-A,4-B. Three precincts across the heart of central Bexley, and one sharing central tendencies along a strip of Broad St.

What does this mean? Well, if one were so inclined, one could take the vote percentages by precinct, assume that they would also apply to the undervote, multiply them by the 06-07 differential, add the result into each precincts total, sum, and voila - make a claim about what the outcome would have been with 2006 turnout levels. My guess is not much different, but if I have time I'll take a stab at it.

Or you could. I'd be just as happy.


B Kessler said...

Good job, Bonobo. Very interesting... thanks for the hard work.

Verdad said...

Given at least the council results, no big surprise to me that South Bexley (1 candidate running from there, 0 wins) didn't get out to vote relative to Central Bexley (4 candidates running from there, 4 wins).

Either way, the South's representation in city government (council/mayor/school board) dropped by 50% on the 6th. Is that ok with you? Maybe so.

Bexley Voter said...

These are wonderful stats. Nothing gets me on my soapbox quicker than people not showing up to vote, especially when we have so much at stake in our city leadership.

PoliColl said...

If the historical context helps, my research showed highest voter turnout in terms of sheer numbers since 1999 has consistently been: Ward 3, followed by Ward 2, then Ward 1 and finally Ward 4.