Abstract: | Recent research has focused attention on the importance of accountingfor measurement error in party identification when modelingthe stability of partisanship and the determinants of the vote.Measurement error estimates have in the past been based on asingle measure of partisanship observed at multiple points intime, a test-retest methodology that requires fairly strongassumptions about the character of change over time. This articleassesses the reliability of the Michigan party identificationscale using multiple measures of partisanship at a single pointin time. Our data not only corroborate previous test-retestresults but also suggest that the accuracy with which partisanshippredicts candidate preferences can be enhanced using multiplemeasures. One measure in particular, a labeled 7-point self-placementcontinuum, is found to hold significant potential to supplementand illuminate the Michigan scale. |