Question Order and Multidimensionality of Partisanship

Abstract

The recognition of question-order problems has prompted a reexamination of theory and data in several areas of political behavior. The possibility of question-order effects is of special concern to the study of party identification, the original conceptualization of which has been criticized for disregarding multi dimensionality. We have yet to discover whether the sequence and/or proximity of items measuring multiple, related dimensions may influence findings. An area of research in which results may be especially susceptible to reactivity is that of multiple partisan identification. Most of the studies in this literature asked respondents a context-differentiated sequence of items about their partisan identification in national and state and (infrequently) local politics, respectively, separated only by the usual probes for intensity and direction. The responses obtained logically risk contamination by either consistency or contrast effects, as well as by salience or frame-of-reference effects generated by intervening or antecedent, nonrecursively-related items. In this study, we report the results of two experiments used to control for potential question-order effects in the measurement of multiple party identification. In the first, national and state partisan identification items were alternated in sequence in order to test whether responses to national partisan identification items structure responses to state (and local) partisan identification items. In the second, party thermometer items, national, state, and local partisan identification items, and national, state, and local retrospective evaluations of party governing performance were rotated sequentially. Generally, responses to these three measures of partisanship did not differ significantly as the order of appearance changed.