SC Look-Back

Good news first. I correctly predicted the order in which the candidates would finish. I also pretty much nailed Senator Clinton's percentage.

Now the bad news. I underestimated Barack Obama's vote share and overestimated Edwards'. Here's the breakdown:

Obama +12%
Clinton -2%
Edwards -10%

My predicted percentages were off for, I think, two reasons. First, I will admit that I placed too much emphasis on the late "surge" in Edwards support. Based on the polling, it does seem that Edwards did gain a bit in the late days of the election. In mid-January, Edwards was polling in the low teens, and ended up at 18%. So he did gain. But not nearly as much as I predicted.

The second reason I was off was the huge turnout. Polls taken in the days and weeks before the election use "screens" to weed out people that the polling firm thinks are unlikely to vote. This is good practice since it does us no good to poll non-voters on their voting preferences. In order to screen out these "unlikely" voters, most pollsters use a combination of the following questions:

1. Do you intend to vote?
2. Did you vote in the last election?
3. How closely have you been following the election?

Some pollsters add in some others, but basically, they use these questions to gauge whether or not that person is "likely" to vote.

In normal circumstances, polling "likely" voters like this leads to more accurate results. But, when you get turnout like we had in South Carolina, which was almost double the turnout from 2004, these likely voter screens become significantly less useful.

It's interesting to note that, putting my predictions aside, the mega-poll combination was a very good predictor of Clinton's and Edward's final vote share. It did, however, underestimate Obama's. I think that is mainly because of the huge turnout of "unlikely" voters who mostly supported Obama. This hypothesis is supported somewhat by the exit polls. Among the 7% of people who never voted before, Obama won 63% to Clinton's 23%. Among the 20% of voters who had never voted in a primary before, Obama won 57% to Clinton's 28%.

Recap: Edwards gained, but not as much as I thought and although Obama was probably going to win anyway, he benefited from big turnout which boosted his vote share from the mid 40's into the 50's.