MPG’s Steve Koczela and Rich Parr wrote the paper below for the 2015 AAPOR conference.
Download the full paper (PDF)
Public polls and poll reporting use a variety of descriptors of party affiliation. Most common is party self-identification: asking respondents with which party they most identify. This is the only available method in most national polls, which cover states with no party registration. In states with party registration, pollsters use a mix of party self-identification and party registration, either self-reported or drawn from a list of registered voters used for sampling. The implications of each method for analysis are not well-understood.
Using data from five waves of the WBUR Tracking Poll on the 2014 Massachusetts General Election, we explore the differences among three methods of identifying voters’ political affiliations. Because the polls were conducted using lists of registered voters including party affiliation, we are able to compare self-reported and actual party registration, as well as party self-identification with actual registration.
We found that a substantial minority of voters misreport their own party registration, and many self-identify with a party with which they are not registered. Further, we find self-reported measures of partisanship create more polarized groups than do actual registration figures. People who say they identify with or are registered with a party are a more concentrated voting bloc than actual registered members of that party. Finally, our research supports those who cast a skeptical eye on the independence of independents. No matter how they are identified, most independents act more like partisans.