Gallup conducted a thorough review of its 2012 pre-election presidential polling to determine the factors that

affected its 2012 results. Gallup Editor-in-Chief Frank Newport led the review process, along with Dr. Michael

Traugott of the University of Michigan and a team of Gallup’s most experienced statisticians, methodologists,

and research analysts. Dr. Frauke Kreuter of the University of Maryland, Dr. James Wagner of the University of

Michigan, and Dr. Christopher Wlezien of the University of Texas provided advice and consultation.

The extensive review process involved a significant amount of new research, including the fielding of

experiments and simulations focused on three areas of pre-election polling — survey and sample design, survey

field management, and data handling. Gallup has identified specific factors within these three broad areas that

it believes were core contributors to Gallup’s 2012 presidential election polling issues. Gallup’s election polling

review and experimentation process is ongoing, and Gallup will continue to add to this body of research in the

months ahead.

INTRODUCTION

In the 2012 U.S. presidential election, most pre-election polls underestimated Barack Obama’s popular vote

strength. President Obama won the popular vote by 3.85 percentage points over Mitt Romney, while the

average of the major polls using a landline and cellphone methodology estimated President Obama would win

by about one percentage point. There was a distribution of individual poll estimates around this average.

Gallup’s final pre-election estimate of the 2012 popular vote, based on likely voters, was Romney 49% and

Obama 48%, with a margin of error of ±2 percentage points for each candidate’s estimate. Gallup’s goal in

presidential election environments is to reflect underlying voter opinions and sentiment accurately, and in its

final estimate, to closely approximate the final popular vote outcome. To determine why its final estimate was

different from the actual result, Gallup explored a variety of factors related to its pre-election polling.