On the validity and reliability of self-reported vote: validity without reliability?

Data from seven Swedish election study panels, 1973-1994, were analyzed to assess simultaneously the validity and reliability of self-reported vote. Eight categories were created, based on self-report at Time 1, an external check, and self-report at Time 2. Overall the validity and reliability of this measure were quite high. However, the cases in which the measure was valid but not reliable outnumbered those in which it was reliable but not valid. Subsequent turnout behavior was most strongly predicted by what people had done previously, but the two self-report measures were also significant predictors in the regression analysis. The eight categories were then compared on a series of demographic and political variables.