First (Still Sleepy) Thoughts

11/05/2008 05:02 pm ET | Updated May 25, 2011
  • Mark Blumenthal Mark Blumenthal is the Head of Election Polling at SurveyMonkey.

The day after an election for a site like ours is a little like the last day of summer camp. We know many of you are checking in one last time to see how those polls did before you turn your attention elsewhere until the next big election comes around. And while all of us would love a chance to kick back, sleep in and mull the results over for a few days before pontificating about how the polls did and what lessons were learned, the reality is that some of you will have moved on by then. So I'd like to share a few quick reactions and also try to give you a sense of what we are planning for once the dust of the elections settles.

First, a quick reaction to how the polls did yesterday. Earlier today, Charles Franklin reviewed the performance of our trend estimates with his usual graphic flair, but I thought some would appreciate seeing my final morning status table updated with the current vote returns. Although there are a few examples of the estimates missing the mark -- mostly in states where polling was relatively sparse -- most produced margins that came very close to the final result.

081105 trends-vote

In fact, by a fluke of luck, the bottom line count in the column labeled "Cum. EV" may end up being a perfect prediction of the final electoral vote count. In all but two states, the nominal leader on our final trend estimate also led in the actual vote (and that includes the 569 lead that John McCain currently holds in Nebraska's Second Congressional District). The fluke of luck is that the two exceptions, Missouri and Indiana, each have 11 electoral votes each (Charles Franklin also posted a chart today that makes a similar point about how the trend estimates predicted the electoral college vote).

Again, votes are still being counted in some states, so the numbers in the table may still change, but one thing seems unambiguous: There was no "Bradley effect" yesterday -- no hidden McCain vote lurking among the undecided. In the states that were polled most heavily, the trend estimates came remarkably close to the actual result. The undecided vote did not appear to "break" decisively toward either candidate. If anything, the undecided may have gone in Obama's direction in Pennsylvania, a state that the McCain campaign suggested was "functionally tied" on the supposition that Obama would get "what he gets" in the polls with the rest going to McCain. Our final Pennsylvania trend estimate showed Obama leading by 7.1 point (with 50.8% of the vote). Obama won Pennsylvania by a 10.3% margin, getting 54.6% of the vote.

The bottom line is that cumulatively, despite all the challenges from cell-phone only households, declining response rates and worries about likely voter models, the polls of late October provided a remarkably accurate picture of voter preferences and the outcome of the election. So our continuing obsession with public opinion polling was not misplaced.

For the next few days, we will continue to look at how the polls, pollsters and our own estimates performed yesterday. And is not going away after that. We have plans to use our Flash charts to display a wider variety of poll data, including Barack Obama's favorable rating and, of course, his job rating as President once he takes office. We are also looking forward to tracking what both the "basic trends" that Charles Franklin charted earlier this week and the reactions that Pollsters will gather to the initiatives of the new Obama administration. Look for new charts and new data coming soon.

Meanwhile: Let me offer a blanket "thank you" to everyone who has emailed with kind words, suggestions for improvements to the site, and especially for those who have made contributions to our "tip jar." I do intend to answer your emails personally, but it may take me a few weeks.