The writer is editor in chief of Gallup Inc.
Gallup’s publicly released election polling has been the object of review and scrutiny this year, as has always been the case since Gallup first published pre-election polls in the 1936 election. It is not unusual to hear from critics, including operatives from the presidential campaigns who do not like poll results.
This year, critiques also came from aggregators who do not poll on their own but, rather, borrow poll results from organizations, including Gallup, that are willing to invest very significant time and resources conducting polls.
The reality is that Gallup’s final estimate of the popular vote in the presidential election this year was generally within the same ballpark range as most other polls — one point different, for example, than the final poll estimates of CNN, the New York Times/CBS and other polls — and within the margin of error from the final actual popular vote result. Further, Gallup arguably produced more insights about the U.S. electorate throughout the presidential campaign, and interviewed more Americans, than any other organization.
It is important not to confuse aggregation of state poll results designed to estimate the Electoral College with estimates of the national popular vote. Gallup is not involved in estimating state or Electoral College results, and as we learned in 2000, the national popular vote and the Electoral College are not always in sync.
This year in particular, the national popular vote reflected Mitt Romney’s strong support in the South, even though his popular vote support didn’t for the most part help him in key Midwestern or Western swing states.
Gallup’s tracking from April through Election Day shows that this year’s popular vote was generally quite close all along, with a decided movement toward Romney after the first presidential debate, and then an equally decided movement back to Barack Obama in the final week of the campaign.
One often-quoted aggregator, Sam Wang from Princeton (a very smart neuroscientist but not a pollster), has been quoted as saying that Romney was never in the lead during the entire course of the campaign. This conclusion is highly debatable and belies a number of other poll results in addition to Gallup’s showing Romney ahead at times in October.
For example, the Pew Research poll showed a movement from an 8-point Obama lead from Sept. 12-16 to a 4-point Romney lead for Oct. 4-7. In the week following the final debate, the ABC News/Washington Post tracking poll showed Romney with a 3-point lead in its Oct. 24 release. On that day, the Gallup tracking poll showed the same 3-point lead.
The goal of any final pre-election poll before an election, which Gallup certainly acknowledges, is to come as close as possible to the actual national vote. Gallup’s methodologists constantly review the processes involved in election polling and this year made a number of shifts and changes to accommodate such things as the increase in early voting. All pollsters recognize the increasing challenge of measuring election sentiments in the context of rapidly changing forms of communication and changes in the ways in which people vote.
Gallup provides pre-election polling as a public service, although this is a small part of our overall business. Further, our continuous polling, during election years and non-election years alike, extends well beyond the basic horse race to provide deep analytical insights about the views and priorities of the U.S. electorate.
Gallup is nonpartisan and independent. We do not work for or accept support from any candidate, party or partisan organization for our election polling. We believe this nonpartisan public service is a valuable contribution to our ability to understand the U.S. electorate, and we will continue to evolve how to best measure and analyze voters’ views and intentions in election years to come.