(Rasmussen Reports) – While public polling generally gave a good projection of what to expect around the country in Election 2010, that was not the case in Nevada.
Rasmussen Reports’ final poll on the race showed Senator Majority Leader Harry Reid trailing by four points and the Real Clear Politics average showed the Democratic incumbent down by 2.7 points. When the dust settled, however, Reid won by 5.6 percentage points.
After every election, polling firms review the things that went well and those that didn’t to learn and make modifications. In this process, Nevada seems like an ideal place to start. That’s especially true because it’s the second straight election in which the Democrat has substantially outperformed both Rasmussen Reports polling and the average of all public polls in Nevada.
In 2008, President Obama won the state by 12 points and outperformed the average poll by six points. Rasmussen Reports had shown a four-point edge for Obama. So it seems as if there is something substantive that public pollsters are missing in understanding the Nevada electorate. However, the misunderstanding is fairly recent. Both Rasmussen Reports and the public polling averages did very well in the 2006 Nevada governor’s race and the 2004 Bush-Kerry presidential match-up.
At Rasmussen Reports, we began by comparing our results to the exit polls to see if there were any obvious clues. Our final poll showed Obama’s job approval in the state at 44%, while the exit polls found 45%. Our final sample had 39% Democrats and 35% Republicans, while the exit polls showed 35% Democrats and 33% Republicans. Their final numbers showed 32% not affiliated with either major party, while we showed 26%. None of those variances would explain the difference in projected outcome.
Our final numbers showed that 76% of voters would be white, while the exit polls showed 72% white. This could certainly account for some of the difference between the poll and the actual results, but not all. It also confirms the sense that the Reid campaign did a good job of turning out its base.
One discrepancy that really jumps out, however, are the results among unaffiliated voters. The Rasmussen Reports numbers showed Angle winning these voters by more than 20 percentage points. The exit polls showed her with just a four-point advantage. This strikes us as an especially significant clue because the Angle poll numbers in Nevada were not out of step with the preference of unaffiliated voters for Republicans around the country.
Another important clue can be found in the fact that our final poll showed Angle leading among those who had already voted a week before the election and those who were absolutely certain to vote. Among those who said something might come up to prevent them from voting, however, Reid held an 11-point edge. At first glance, it appears as if that group should have made up a larger segment of our sample.
At Rasmussen Reports, we also asked people to rate how closely they were following news about the elections on a scale from 1 to 9. Those who were following it most closely responded with an 8 or a 9 and preferred Angle over Reid. Those who said 5, 6 or 7 preferred Reid over Angle. The lower the rated level of interest, the higher the advantage was for Reid.
This higher interest among less committed voters seems borne out by the exit poll data showing that six percent (6%) of voters did not vote in the presidential election two years ago. These voters strongly preferred Reid over Angle.
Some credit the Reid campaign for an incredible get-out-the-vote effort to bring these less engaged voters to the polls. They may have done a good job and certainly deserve kudos for the effort, but there is no way such a campaign effort can account for a nine-point swing from our final poll on the race.
Others say the difference might be caused by the cell phone factor. Rasmussen Reports and some of the other public polling firms didn’t call cell phones. While that’s a possible contributing factor, if it was truly significant, we would have expected to see it crop up in more places. Additionally, in terms of moving forward, the question about cell phones is moot. The reality is that every polling firm will need to find ways to reach people who have abandoned land lines as we begin polling for 2012 and beyond.
At the moment, our working hypothesis centers around the measure of enthusiasm and certainty to vote questions. Republicans and Republican-leaning unaffiliated voters were clearly fired up, ready to vote and following the election closely. But a larger-than-anticipated number of less fired up and less committed voters came to the polls for Reid.
As we work our way through this, we are fortunate to have a great pool of data to work from. The key is not just understanding what was off in our mix of potential Nevada voters but what changes we could have made in our sampling approach or Likely Voter model to capture a more accurate sample. As we do this, we will need to make sure that any adjustments we consider also would work in all the other states we polled this year.
The end result for us will be the same as it is after every election. We will learn from both our successes and our failures to modify our polling process for the next cycle.