However, I always go a step further and ensure failure by trying to predict vote differentials. Join me below the fold to check out my granular predictions in both the presidential and Senate races.
Averaged out, I missed by just 1.26 points which is pretty good! I overstated President Barack Obama's support in four states and the national popular vote, and under-estimated it in five states—so I wasn't blinded by my partisanship.
My two biggest misses were Colorado, where I didn't trust more favorable polling (I unskewed myself!), and Ohio, where the polling was marginally more optimistic than the final results. But even then, I considered the final polls to be overly pessimistic, so I tacked an extra couple of points to Obama's margin. It seems that every time I ignore the data I get burned.
Though not so fast! The polling composite said Obama would win Iowa by 2.6, but I ignored it and went with a single poll—the Ann Selzer one for the Des Moines Register. Selzer has a near-perfect record polling Iowa, and I bet correctly.
In Nevada, the polling composite said 4.7 percent, but polling in the state has traditionally under-polled Democrats, so I added an extra point for good measure. I came close, just off 0.6, but I could've added another point and done even better.
Still, all in all it was a good night for my presidential predictions. My Senate ones, on the other hand, were more off. While I predicted a 55-45 Senate, I blew two races—North Dakota and Nevada. I should've listened to the polling in Nevada, but if I went by the polling composite alone, I would've still blown North Dakota. Then again, I consistently talked about Heitkamp being the most likely Democratic upset based on her mad retail politics skills. So I'm thrilled I was wrong about that one.
In the vote differentials, I overstated Democratic performance in nine of them, and underestimated it in seven. So I was just slightly more optimistic about our victory margins than the final results:
I expected more undecideds to side with the Democratic candidate in places like New Mexico, Massachusetts, and Ohio, while I believed the tied polling in Montana. Indiana was a polling clusterfuck, so I pulled a number out of my you-know-where and still underestimated the damage Richard "Rape is a gift from God" Mourdock did to himself ... and same with Missouri. Apparently, even heavily evangelical states aren't going to reward rape apologists. Good to know moving forward!
The difference between the two sets of predictions? The amount of data. Lots more battleground state polling than in places like Montana. But all in all, I'm pleased with my results, particularly considering the alternative. How did you guys do with your predictions?