The controversy over the veracity of polls, which some conservatives argue are skewed in favor of President Obama, led The Hill to examine the surveys released in the final days of the 2008 election to see which had the best track record.
The result: Most major polling companies were within 1 or 2 points of the final tally when Obama took 52.9 percent support to Sen. John McCainJohn McCainPentagon should have a civilian chief to give peace a chance McCain to support waiver for Mattis, Trump team says Senators crafting bill to limit deportations under Trump MORE’s (R-Ariz.) 45.6 — a difference of 7.3 percentage points.
None of the major polls was more than 3.7 percentage points off the final mark.
Four polling outlets in 2008, the liberal-leaning Democracy Corps, Fox News-Opinion Dynamics, CNN-Opinion Research and Ipsos-McClatchy, hit the final result almost exactly, with each predicting a 7-point Obama victory.
In addition, seven polling outlets erred by only 1 point, predicting either an Obama 6- or 8-point victory.
Gallup, Reuters-C-SPAN-Zogby and CBS News-New York Times fared the worst, each predicting an 11-point Obama victory.
This election cycle has seen a debate over whether polls showing Obama in the lead are inaccurate, as The Hill reported in late September.
Conservatives argue that many mainstream polls are skewing in Obama’s favor because of sample sizes that base 2012 turnout projections on 2008, when Democrats — and Hispanics, blacks and young voters in particular — voted in record numbers.
Republicans say it’s foolish to expect a similar proportion of Democratic voters in 2012. They argue a much smaller proportion of those voters will turn out on Election Day, while a larger proportion of Republicans eager to deny Obama a second term can be expected to vote.
But in 2008, Republicans and Democrats alike made similar claims of bias, as polling metrics and assumptions were complicated by the unique and historic circumstances surrounding an election that produced the country’s first black president.
The final polling results, however, ended up being pretty accurate.
There was a 7.3-point spread between Obama and McCain when all the votes were tallied, and the RealClearPolitics average of polls at the time showed Obama with a 7.6-percentage-point lead heading into Election Day.
Of course, there were outliers on both sides, ranging from a 2-point to an 11-point Obama advantage, but overall, the polls closely predicted the results.
Every poll makes assumptions based on the peculiarities of an election. In 2008, pollsters had to decide which way undecideds would lean in a race without an incumbent; how big of a turnout to expect from Hispanics, blacks and young voters; and which way the growing population of independent voters might lean.
In 2012, pollsters are again finessing their models to keep up with the changing times, opening themselves to criticism that they’re making the wrong assumptions.
Undecideds typically lean toward the challenger, but what percentage of these should be allocated to Republican candidate Mitt Romney? Will a smaller proportion of Democrats come to the polls? Are Republicans so eager to deny Obama a second term that party identification should mirror the 2010 midterm elections instead of the 2008 presidential election?
The Hill will revisit the performance of pollsters after the election, but for now, we offer a look at the best and worst from 2008.
|Predictions by firm|
|FOX News/Opinion Dynamics||43||50||-0.3|
|American Research Group||45||53||0.7|
|Predictions by firm|
|NBC News/Wall Street Journal||43||51||0.7|
|ABC News/Washington Post||44||53||1.7|
|Daily Kos.com (D)/Research 2000||46||51||-2.3|
— Poll rankings provided by Dr. Costas Panagopoulos and first published in the essay Polls and Elections: Preelection Poll Accuracy in the 2008 General Elections, which ran in the Presidential Studies Quarterly; Dec. 2009; 39, 4
— A negative spread indicates the poll underestimated Obama’s final margin of victory, a positive spread overestimated Obama’s final margin of victory.
— The rankings are not solely based on the spread, but according to Panagopoulos, a predictive accuracy model “based on the natural logarithm of the odds ratio of the outcome in a poll and the actual election outcome” to account for “the direction of bias.”