Greg's Opinion Greg's big blog of whatnot


Election Season Math Primer 2.0

The biennial obsession over Early Vote numbers has already begun. What the county makes freely available to the public is a count by EV location. That's always been a problematic way to analyze voting behavior, though.

If you're willing to fork over about $120, you can order the daily roster of voters. That offers some useful information if you know how to use it. But for the most part, it's just a list of names, addresses, and voting locations. If you plug it into a voter database, which both parties have, there are a number of scores that we can derive to tell us what it means for R vs D numbers. And on a good day, those numbers might be accurate. Or they might just be very sophisticated tea leaves.

Right now, I'm in the daily habit of tracking three different scores, each with their own failings to account for. But each worth keeping an eye on. Without divulging a) something proprietary, or b) something I'm not capable of describing accurately, my summary of two of the scores is that we have one that's recognized as too optimistic and another that's somewhat pessimistic. The differing degrees to which each differs from historical patterns and the range between the two is way too broad to consider averaging them out. These two scores are derived from individual voting records. That's one of the nice features of them since a precinct may not be homogenous when it comes to voting blocs. This is a particular issue in Southwest Houston since many precincts have apartment complexes that turnout out decently in Presidential years, not-as-decently in midterm years, and craptacular in odd-year elections. That's partly how we ended up with a City Council District F that voted 55% Dem in even-numbered years, but sent conservatives to City Hall.

The third number is the one I create, using the method outlined here. This early into Early Voting, I'm not comfortable enough leaning on the reading offered by my DPI-by-Precinct calculation for House Districts. The sample sizes for precincts is still small, mail ballots still account for roughly 1/3rd of votes cast through the two days I have data for, and the number of precincts in a State Rep district is small enough to expect some mathematical anomalies that I wouldn't want to be making a money decision on. All this number tells you is whether you're sending voters from the right precincts to the polls. It doesn't account for the possibility that all of the Republicans and none of the Democrats from a precinct may have voted on any given day.

What I've done below is translate all of that into a stock chart. The pessimistic/optimistic scores basically show up an open/close vertical line. And the DPI-by-Precinct is the horizontal bar for each district. In order to make things simple, my definition of the DPI in this case is simply Obama's vote. I've gone through and translated the 2012 precincts into 2008 precincts in order to get a close enough read on what all the split precincts do to the result.

These are four State House districts with little doubt about the outcome:

                                 DPI-pct ... Optimistic ... Pessimistic
HD132: Bill Callegari (R)  ....    41.0% ...   40.8%  .....    32.0%
HD133: Jim Murphy (R)  ........    34.4% ...   31.1%  .....    23.0%
HD139: Sylvester Turner (D)  ..    74.9% ...   75.1%  .....    71.4%
HD143: Ana Luna (D)  ..........    64.1% ...   74.9%  .....    70.6%

Adding a dose of context, the DPI-pct scores compared to Obama-08 actual results show the following:

HD132 ... +1.41%
HD133 ... -0.05% 
HD139 ... -0.65%
HD143 ... +0.04%

There are a lot of qualifiers to keep in mind. Typically, Early Voting behavior suggests that GOP voters turn out more heavily in the first week and Democrats either catch up or move ahead of them in the second week as the hours are longer and the focus on an election is clearer. That said, we have no idea what to conclude about whether the same pattern holds here.

My hunch is that a lot more voters are banking their vote early. In 2008, Democrats did that exceedingly well. So much so that they ended up losing Election Day in Harris County. That may or may not hold this time, since 2010 seemed to have taught many GOP voters the significance of the straight party ballot. It's only a guess for now. But I suspect that the early vote lead we saw in 2008 will be balanced out by Early Voting Republicans. But this may also mean that Election Day is evened out as many of those voters would have already voted. There's no way of knowing until after the fact. But it's a strong enough theory to me to offer it here for discussion.

What stands out to me from the sample of election scores above is that, in the case of the Obama DPI-prec, that grade may be overstated in the GOP suburban areas where he did exceptionally well in 2008, but have been ripe for a return to GOP voting behavior in 2012. In the case of the two GOP districts above, the range offered by the individualized scores seem fairly realistic.

HD139 is a fairly sizable African-American district. In almost all such districts in Harris County, the Obama DPI score is at the top boundary of the range of the individualized scores. It's also worth noting that this district has the shortest of ranges among the four districts. That's another commonality for the Afr-Am districts. My sense is that these districts are easier to score in terms of how they vote. The interesting thing is that the turnout isn't quite as homogenous as many might think. That may be a topic of discussion later on during Early Vote.

HD143 is a heavily Hispanic district, but one with enough working class Anglos, also. I think that the fact that Obama is below the pessimistic score could be read as a feature seen in many Hispanic districts, where Obama under-performed in 2008 or that the working class Anglos skew the district results on a per-precinct level. I haven't gotten enough in the weeds on the methodology here to discern which matters more. And when I say Obama underperforms in Hispanic districts, I offer that as a comparison to district officeholders who usually beat the baseline by a considerable margin, usually by gaining more independent Hispanic voters that aren't quite reliable in their Dem voting behavior.

There exists a deep desire among political professionals to find some magical, singular number that tells us everything we need to know about a precinct, an election contest, a poll, or whatever. I started working on the DPI-by-Precinct calculations as an effort to replace several bad data elements from conversation and replace them with something better. But I think it's better to look at a broader range of data to arrive at a range of expectation rather than try to shoehorn all guesstimation into one number offered with absolute authoriteh. There's nothing gospel-like about DPI-by-Precinct. It has some assumptions that need to be kept in mind for when someone doesn't find themselves in an environment that it applies to very well.

It's that reason that I'm happy to see the individualized scores to compare them against. But even those are an inexact science. And it's an open guesstimate as to which one is the most accurate this time around.

For now, this is just offered as food for thought. Not all of the data is such that I can talk openly about it until the voting is done. But, with this post, I want to set the foundation with this information so that I don't have to write too long of a long backstory about it on Election Night. If the math fascinates you, great. If it bores you, then believe me when I say that the return of Krokus videos is less than a few week away. Sit tight till then.

Related Posts:

Comments (0) Trackbacks (0)

No comments yet.

Leave a comment

Trackbacks are disabled.

Fatal error: Call to undefined function xwp_footerx() in /nfs/c01/h12/mnt/4748/domains/ on line 18