Reassessing the Prevalence of Party Identification:

The Implications of Forced-Choice Question-Wording Effects

Francis Neely

Department of Political Science, San Francisco State University

San Francisco, CA, USA

fneely@sfsu.edu

Abstract

Although recent scholarship suggests a revival in partisanship, the replication of a question-wording experiment (Niemi, Reed, & Weisberg, 1991) confirms that about one in five people remain non-identifiers—they reject the labels of Republican, Democrat, and Independent. Non-identifiers in 2005 look very much like those measured in 1987, suggesting both the quantity and quality of this group may be stable. What are the implications of mis-categorizing the PID of a significant portion of the population? Descriptively, the overall partisan nature of the population is better described as comprising four, not three, similarly-sized groups. Among those, non-identifiers stand out as younger and a bit less informed and engaged than others; meanwhile, the differences between Independents and partisans on those factors become negligible once non-identifiers are specified. Filtering non-identifiers also affects the estimates of how PID relates to political attitudes and behavior, with the filtered measure mildly outperforming the standard version in explaining approval of and voting for political leaders. In addition, filtering PID alters the estimated influences of other factors. The findings refine competing arguments about the status of party identification. While the prevalence of PID is probably overstated, its influence is likely underestimated.