David and the rest of the SSP crew have been kind enough to give me a soapbox here, and I think I’ll be starting a series on breaking down large jurisdictions through the lens of some election.
Having gotten my hands on precinct data for the city for both 2008 and the 2010 Special, I thought I’d continue to examine the disparities between Obama’s and Coakley’s respective performances.
As you can see on the map, the geographic central core of the city, Roxbury and Mattapan, remained strong with little dropoff from Coakley to the Obama. Jamaica Plain, Allston/Brighton, and Back Bay – all strong Obama areas as well – showed slightly greater drop-offs. Even greater drop-offs were noticeable in the already swingy areas of the city, such as West Roxbury, Dorchester, Charlestown, and Southie. McCain won only 3 precincts throughout the entire city’s 254; Brown increased that to 33.
Putting this statewide perspective, we get this:
Again looking at the map, South Boston was pretty darn brutal for Coakley, with Brown scoring 60%+ in several precincts. Many people (including one Stephen Lynch) indicated particular hostility for Coakley in the neighborhood. She did get destroyed here, but was it any worse than how badly she got destroyed across the rest of the state?
I think not. Sidenote: I’m defining “South Boston” the same way the Boston City Council does, that is, all nine precincts in Ward 6 and precincts 1-7 in Ward 7.
In 2008, in the 16 precincts constituting “South Boston” (or Southie), Obama beat McCain by a margin of 3,100 votes, or roughly 59-39. In 2010, Coakley lost by a margin of 1,500 votes, or roughly 43-56. Overall, this was a 16.0% swing; this is somewhat worse than that 15.31% swing experienced by Coakley across the state.
But, despite my election-night model assuming so, Coakley didn’t experience a uniform dropoff. Instead, dropoffs are quite correlated with how well Obama performed in the area was to begin with. (This makes sense – Democratic strongholds are likely to remain so, while swingy areas in which Obama did well might have been particularly receptive to Republicans in a close election.)
Throwing this up on a graph (with Coakley’s dropoff on the vertical axis and Obama’s margin on the horizontal), we get:
You’ll see a few outliers here: the point at the origin you can throw out – that’s Boston Precinct 01-15, which last had a voter in 2004. The correlation on that is 0.83 0.816, suggesting quite a strong relationship.
Taking the geekery to the next level, I busted out the extraordinarily helpful Stata (how academic of you, my SPSS-using friends tell me…):
For those who are less of statistics nerds than I am, the regression tells us two main things:
- For every point increase in Obama’s margin in a voting unit (precincts within Boston, towns elsewhere), we can expect Coakley’s performance relative to Obama’s to improve by 0.14%.
- For a hypothetical voting unit that was exactly tied between Obama and McCain, we should expect a 17% swing away from Coakley.
Applying this to South Boston, we see that there isn’t really a pattern: some precincts had drop-offs more than to be expected, others had less.
There really isn’t much a discernible pattern here, again, supporting the conclusion that while Southie didn’t like Martha, they didn’t indicate their dislike for her through their votes more than the rest of the state did.
This can all be represented visually as well:
The last benefit of getting the Boston data was I could finish results of the Senate Race by CD. As we’d already known, they weren’t pretty, but here’s the results table just as a freebie:
In the article, you state that the correlation coefficient (r) was 0.83, but you give the r^2 as 0.6651, which would indicate an r of 0.8155. . . Am I missing something?
I expected to see better proof of my pet theory about Coakley being crushed by low non-white turnout. (I know, reductionist and a dangerous thing to say…hear me out…)
Two things jumped out at me:
1. Coakley lost Lowell, barely won Worcester and Holyoke, and had much smaller than usual margins in Fall River, New Bedford, and Springfield. Other than Lowell (large SE Asian population), the non-white populations of those cities (mostly Hispanic, specifically Puerto Rican, the most uniformly Democratic of all Hispanic subgroups – Springfield also has a sizable African-American community) usually are nearly unanimous for Democrats, when they turn out. The white populations of these cities are more willing to vote Republican.
2. Coakley did just fine in towns that consisted mostly of white liberals. Famed strongholds Newton and Brookline; wealthy Metrowest towns like Lexington, Lincoln, and Concord; Five Colleges towns and Berkshire and Franklin County towns out west; and the Vineyard and Outer Cape.
Both these things suggest that the turnout for this election was far whiter than 2008, 2006, or 2004.
…what the heck is a CD ? Why is Worcester there more than once ?
I still have to write up my Massachusetts Republican gerrmander, which has been sitting on my hard drive for well over a month.
But I hadn’t thought to work out Brown-Coakley numbers for the current districts. I shall have to integrate these numbers into the diary when I finally finish it.
It would be easier to get to other diaries on the front page if you had more “below the fold”.
And as someone getting a Ph.D. in statistics, I’m wondering why you didn’t use R instead of Stata. 🙂