Do Likert scale scores over-estimate behavior changes?

This is the question I examine, along with my coauthor Jim Thrasher, in our newest publication at Health Education Research.  The title is Examining the relationship between psychosocial and behavioral proxies for future consumption behavior: self-reported impact and bidding behavior in an experimental auction study on cigarette labeling.

Here's the abstract:

Experimental and observational research often involves asking consumers to self-report the impact of some proposed option. Because self-reported responses involve no consequence to the respondent for falsely revealing how he or she feels about an issue, self-reports may be subject to social desirability and other influences that bias responses in important ways. In this article, we analyzed data from an experiment on the impact of cigarette packaging and pack warnings, comparing smokers’ self-reported impact (four-item scale) and the bids they placed in experimental auctions to estimate differences in demand. The results were consistent across methods; however, the estimated effect size associated with different warning labels was
two times greater for the four-item self-reported response scale when compared to the change in demand as indicated by auction bids.  Our study provides evidence that self-reported psychosocial responses provide a valid proxy for behavioral change as reflected by experimental auction bidding behavior. More research is needed to better understand the advantages and disadvantages of behavioral economic methods and traditional self-report approaches to evaluating health behavior change interventions.