I've been waiting for today for several months. In preparation, I've been working on a set of calculations and maps from the 2008 general election to set the context for analysis that will start tomorrow. I rarely show work in progress, but here's a sneak peak of what I'm thinking about.
This year, my election thinking is centered around the US House of Representatives and 127 (129 minus 2, see dataset note in the next paragraph) districts with Tea Party candidates on the ballot. (After a hell of a time trying to decide on a "good enough" way of deciding who is and is not a "Tea Party candidate" given the different organizations giving endorsements, theNew York Times was considerate enough to compile those endorsements, local press, and candidate positions in order to produce a list.)
Before we get started, a note about the datasets: In an effort to acknowledge that political districts are often nothing more than gerrymandered political constructions, I've decided against using the district boundaries themselves when looking at voting records. Instead, election results are mapped by county. This comes with pros and cons. Pro: It offers a higher resolution for the data (counties are usually, though not always, smaller than House districts) and reminds us that districts are not comprised of monolithic voters. Con: District boundaries do not always follow county lines, occasionally running right through a county. For my purposes right now, I've decided I can live with this. Additionally, despite my family history in Hawai'i, this study is only concerning itself with the 48 contiguous states. While Hawai'i has Tea Party candidates on the ballot in two House districts, this study includes spatial statistics that axiomatically require geographic proximity. Regardless of where the numeric data lies in the national distribution, Hawai'i will remain a geographic outlier, and thus its two districts are not included.
These maps are nothing new. We've seen these before. We saw the map above on CNN's telestrator two years ago. It shows the basics of how the American counties voted for their congressional Representatives. The ends of the spectrum indicate those counties where people generally (almost entirely) agree on the candidate best suited for the job. The middle colors show where there is a little more diversity of opinion.
This one isn't much different: same basic information with the addition of population (voter) density. Each dot represents 1,000 votes either Republican or Democrat within a given county. Just as with the always-delightful, psychedelic election cartograms, these dot density maps are useful to remind us that often large swaths of one color might only represent a handful of people.
[OK. Because this is a blog post about work in progress and not a paper about a finished study, I feel like I can interject a side note here. Cartograms are fun, but I can't take them seriously because of two things they obscure: (1) The boundaries are important. I mean, really important. Administrative and political boundaries are among the most explicit spatial constructions within our sociocultural landscape. They are weird and mean almost nothing in our daily lives, but in that insignificance and oddity lies their relevance. The importance of where they came from and the impact they have cannot be overstated. (2) Places with more people are not bigger. They are usually smaller. It's actually how density works. I realize these are purely representational decisions, but cartograms based on population numbers confuse the meaning of urbanity. And maybe it's just me, but I map to find relationships between phenomena and the spaces we inhabit as such. Next time you feel the need to make a cartogram, maybe do this instead:]
Now that we've refreshed our memories about what went down two years ago and allowed a self-indulgent minute to rant about political cartograms, What is the Tea Party trying to do today? As I understand it, these are clearly not moderate Republican candidates. Yet, check this out:
If I turn off the lights on all the counties NOT confronted with a 2010 Tea Party House candidate, these are the 2008 election results we're talking about. In many ways, this is not entirely surprising. Why would the Tea Party focus on areas that are already solidly Republican? That wouldn't make much sense. Another thought is that the potentially polarizing nature of their platforms could pull voters out of the woodwork and into the booth, but die-hards are not generally the ones who don't vote. So, maybe they are planning on capitalizing on the hardcore conservative independents (read "Libertarians") whose votes never get mapped? I'm willing to entertain that option, with skeptical reservations, for the areas falling closer to the middle of the red-to-blue spectrum, but more of these counties are solidly blue than that theory might account for.
We can take this another step and look at the dot density map, turning off the lights for the non-Tea Party counties. (I warn you, this map is more about what you can't really see in it.)
I did warn you. There is very little color on this map...and that's basically the take away. We are really not talking about our most urban areas or districts that represent a ton of people. And that being said, the minimal color we can see is mostly blue. (You might have to just trust me on that.)
Honestly, saying that a county seems solidly Democrat because I crunched some numbers then drew the county really blue shouldn't be taken super seriously. Before tonight's results, we need a map that means something a little more concrete. For that purpose, I offer this:
Explanation: This map is the result of a statistical test for spatial Clusters and Outliers (go here and here). Statistically significant clusters are those regions wherein we can say with some confidence that the collection of similar voting patterns is not due to chance. Statistically significant outliers are those counties that are considerably different in their voting outcomes than the surrounding counties, specifically different enough that this difference is also probably not due to random chance. (We know what these places are like: They are the isolated pockets of difference, enclaves or college towns.) In this map, I'm representing Clusters and Outliers that satisfy a 95% confidence interval, meaning that the chance that the results are random is less than 5%.
If you ask me, these are the only maps that should be considered when we talk about a Red-and-Blue America. You'll notice that most of the country is grey, counties wherein I can't with any certainty claim a spatial pattern.
Alright. Check it:
Here's the breakdown: Of the counties whose residents will see a Tea Party name on their ballot today, two years ago,
- 29.8% were part of a Democrat Cluster.
- 1.72% were part of a Republican Cluster.
- 2.39% were Democrat Outliers.
- 0.13% were Republican Outliers.
- 61.9% are grey on the map above.
Yes, almost thirty percent of the counties in question are in areas of the country that voted Democrat in 2008 at a statistically significant level. And more than two percent were so Democrat that they achieve statistical significance despite their regional trends. Compare that to the Republican numbers from that map, and I have to ask again What is the Tea Party trying to do today? My guess is that they are trying to redden the grey, but I can't see the political strategy involved at this level. Regardless of whether they are successful today, the answer is going to say something about our political landscape.
In the coming weeks, expect more maps (including maps at local scales), analysis of today's results, and comparative statistics of changes between 2008 and 2010 in an attempt to quantify the effect of the Tea Party. Additionally, expect a more formal presentation of this study (a paper), complete with the specifics of my methodologies.
If you have questions in the meantime, as always, you can email me.