During the 2012 Presidential election, Gallup’s tracking poll data significantly overstated support for Mitt Romney and as a result Gallup decided to engage in an in-depth analysis to determine what went wrong. Their conclusion is that they made four important mistakes:
1) Gallup’s likely voter estimates were off. They put too much weight on past voting patterns missing some of President Obama’s voters.
2) Problems with their regional control. Because of time constraints they ended up underrepresented specific time zones with assigned regions.
3) Problems with defining race. The way Gallup asked people to define their race/ethnicity resulted in an unusually high number being classified as multiracial. This was a problem when their numbers weighted against the Census Bureau’s data on race. Gallup has begun addressing this issue.
4) Back in 2011 Gallup tried switching from RDD list-assited landlines samples to a listed landline sample. This listed landline sample proved to be slightly older and more conservative. Gallup has began moving back to the old system.
This analysis is mostly good news for progressives moving forward. Gallup is one of the most prolific and widely cited pollsters. Gallup actually does relatively light election polling, most of their polling is about issues.
It is likely that Gallup has recently been slightly underestimating the support for a range of progressive policies like gay marriage, marijuana legalization, and immigration reform. Now that Gallup has started to address their sampling issues, it is likely that their polling will correctly find a slightly higher level of support for these issues.
Image from Gallup on Facebook