I used the map from http://fivethirtyeight.blogs.nytimes.com/ to generate a list of the 9 states that they have listed with <81% probability who will win it: CO, FL, IA, MO, NC, NH, OH, VA, WI. Then I created 2 separate lists from my data set. The first list is all counties who had a swing in the last 3 elections, with just the last election having been a <5% win, and in swing states. The second list was of the counties in the U.S. with "close" elections for all of the last 3 elections (the victor having won with <5%), had "swung" in any of the last 3 elections to either side, and in swing states. Here are my two lists:
|LIST 1||LIST 2||Swing States|
Combining these lists leaves 14 likely swing counties. No county from WI made either list, so can be presumed an Obama win, despite Romney's pick of Paul Ryan for VP (Obama won WI the last election 56%-42%). Four MO counties are swing, and total, they comprise 6% of the state's voter turnout from 2008--all four went to Obama, but not by much.
Unemployment is often a key predictor for elections--the current national average is 7.5% (BLS, June). In this case, MO still goes to Obama, because they are below the national average with 7.1%. NH, IA and VA all have unemployment < 6%, so are likely Obama wins. OH is on the border with 7.2% unemployment, and CO, FL and NC are all above average, ranging from 8.2-9.4%. If we use these as predictors for the electoral votes, (above average unemployment states go to Romney, and below average go to Obama), then Obama wins 286-252. However, most pundit models give MO to Romney, in which case Obama still wins 276-262.
Despite my county-level-swing-state analysis, most pundit models have FL, CO and OH leaning towards Obama, generating an Obama win with >330 electoral votes (I haven't seen any models giving him <300 electoral college votes, as of August 15). Of course, swing counties aren't the only predictors, or even the primary predictors, since many of these counties represent <1% of the state's votes. Another important non-economic, non-demographic factor is whether candidates can generate turnout.
A bigger question for me, is if, as Brian Williams suggested, that strategists (people who do *real* analysis professionally, not my on-the-side hobby) have advised the candidates that there are 12 key counties in the United States that determine the presidency, what kind of democracy is this? This absurdity that we call the electoral college, that on the one hand, twice this century has given the presidency not to the person who won the popular vote, but to he who won the electoral college, and on the other hand, creates a situation where a handful of undecided voters in a handful of counties determine the fate of the entire country, and the candidates win these votes by pouring vast sums of money into those specific counties?
Coincidentally, last year, almost to the day, I posted an analysis of presidential voting patterns, the crux of which is that we almost always re-elect the sitting president, with the predictive variable being whether the economy was worse the 6 months prior to the election than when the president took office. Fortune-Telling the Presidency