Abstract
Correlational techniques exclude Don't Know respondents by necessity. However, doing so makes the implicit assumption that excluded respondents are identical to non-DK respondents. As this paper and others have shown, the propensity to give DK responses is strongly related to sociological and psychological variables. By randomly assigning DK respondents, we can better compare correlations between subgroups with differing percentages of DK respondents. Doing so, we find that, contrary to Nie, et al., education is as strongly related to attitudinal constraint in 1968 as in 1956. Furthermore, expected relationships between knowledge and attitude stability, among adolescents, are substantially strengthened after randomly assigning DK respondents.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
