Download data and study materials from OSF
Barry C. Burden
University of Wisconsin-Madison
University of Miami
Sample size: 2312
Field period: 10/21/2004-11/1/2004
The aim of this project is to reconcile a standard conceptualization of party identification with its measurement. Theory assumes that party identification is fundamentally affective, yet the standard questions used to measure this concept explicitly ask respondents to respond cognitively. To better align measurement with theory, we seek to alter the original questions to replace cognitive prompts with affective prompts. Through a survey-based experiment we compare the two measurement strategies directly. Rather than assume, as recent revisionist works have, that the measurement is adequate but the original theory flawed, we attempt to make theory and measurement congruent by altering the latter. Not only does this approach make logical sense but our initial substantive results yield new and interesting insights about party identification in the United States that encourage additional exploration of the concept.
Following an earlier phone survey, we expected the affective party identification questions to produce more Republican responses and slower response times, particularly among women.
Respondents were randomly assigned to one of two condition. In the first condition respondents were given the traditional NES party identification battery (Think). In the other condition respondents were given altered versions of those questions where the wording emphasized affect (Feel). All other items were administered identically to all respondents.
The main dependent variable of interest is the seven-point party identification scale. Response times, ideology, and vote choice were also dependent variables.
The experiment yielded modest effects, causing a slight Republican shift in the "Feel" condition. The effects were somewhat larger among low-sophistication respondents and those living in battleground states during the 2004 presidential election. The "Feel" questions also produced somewhat strong correlations with measures of partisanship and shifted women's positions leftward on the abortion question. Significant model effects unfortunately make comparing these results to previous results challenging.
Additional party identification experiments should be conducted across time and modes to identify the factors that cause the affective and cognitive measures of party identification to diverge and converge.
The study was intended to be conducted by telephone as the pilot study and NES studies had been. It was instead implemented using the Knowledge Networks internet platform, which resulted in unexpected mode effects but a larger sample. The KN sample had just 2% Independents, much lower than in most telephone surveys.
Burden, B. C., and C. A. Klofstad. 2005. "Affect and cognition in party identification." Political Psychology 26:869-886.
Barry C. Burden. 2005. "The Social Roots of the Partisan Gender Gap." Presented at the annual AAPOR meeting, Miami, FL.