Validation of voter recall: Time of electoral decision making

Accuracy of respondent recall has long been a concern in political science research. Earlier analyses of voter behavior discovered errors which included 10 to 25 percent of the sample. The present work examines an important attitudinal question which appears in each national election study: the time at which the respondent recalls making a final presidential vote choice. Data from the four-wave 1980 National Election Panel Study are used to validate the recall variable for that election campaign. A new variable — candidate preference patterns — is created which reflects consistency in individual voter preference. Only forty percent of the sample reflect consistent responses on the two measures. Furthermore, partisan strength and political involvement are found to be positively related to inconsistent behavior. Cognitive dissonance and the bandwagon effect are offered as explanations for these results.