It’s a common reaction during election campaigns. If your candidate is ahead in the polls, the polls are accurate. If he or she is behind, they’re inaccurate. Most polls tell us Republicans will win big on November 2, gaining as many as 50 to 60 House seats, 7 or 8 Senate seats, and 7 governorships. Some Republican political operatives predict that their party will do even better, picking up 70-plus seats in the House and ten or more seats in the Senate. A few pollsters agree that’s within the realm of possibility, but most estimates are more conservative. We’ll soon know who got it right, but if the higher estimates prove correct, here’s the most likely reason why.
Political polling has come a long way since it failed to predict President Harry Truman’s victory in the 1948 election, and Americans woke up to the infamous November 3, 1948, headline, “Dewey Defeats Truman” in the Chicago Tribune. Back then, election polling was still in its infancy and polls were conducted much less frequently than they are today. The last polls before the 1948 election were more than a week old and failed to capture the Truman surge just prior to Election Day.
Now, instead of relying on a few polls taken weeks apart, the news media rely on many polling organizations, including their own, that conduct polls every day. There are so many polls that RealClearPolitics.com averages them. We’re flooded with polls. Still, polls can vary widely. Some pollsters have a better track record than others, and polls that focus on likely voters rather than registered voters tend to more accurately reflect election outcomes.
The least accurate polls have poorly-worded questions, insufficient samples, or are intentionally skewed. Polls conducted by organizations that work for or favor a particular party or candidate often deliberately skew their publicly-released polls to bolster the party or candidate they would like to win. The New York Times and CBS polling organizations are notorious for polls that are weighted to favor Democrats.
Conservative columnist and political satirist Ann Coulter, in her October 15, 2008, column, “Eighty-Four Percent Say They Never Lied to a Pollster,” documents numerous instances between 1984 and 2000 when polls conducted on the eve of presidential elections by left-leaning news organizations, predominantly the New York Times and CBS, deviated widely from other polls and actual election results. That’s why it’s important to look at numerous polls and the Real Clear Politics average and know who conducted the poll and what biases they might have.
It’s not uncommon, however, for the more objective, more accurate polls to predict the outcome of a specific election race by plus or minus one percentage point. Rasmussen Reports predicted President Barack Obama’s victory in the 2008 election by just such a margin. Accurately predicting the outcomes of numerous individual races in past nation-wide elections, Rasmussen also has accurately predicting aggregate gains and losses. Understanding why polls like Rasmussen have been so accurate in predicting the past elections is the key to understanding why Rasmussen and other polls might not accurately reflect the results this year.
The short explanation is that even with a sufficiently large random sample, the respondents in a poll might not accurately reflect the distribution of actual voters — party affiliation, sex, race, etc. — that will turn out and vote on Election Day in any given district or state. Pollsters, therefore, use models of past elections and census statistics to determine weighting factors for those groups, and they use them to adjust the poll results. Weighting enables pollsters to more accurately predict likely voter turnout for those groups and, therefore, the outcome of the election.