Saturday, November 12
Pretesting Methods
Data Quality and Measurement Error
Sat, Nov 12, 9:00 AM - 10:25 AM
Regency Ballroom-Monroe
New Approaches to Questionnaire Design and Evaluation

Modeling Motivated Misreports to Sensitive Survey Questions (303158)

*Ulf Bockenholt, Northwestern 

Keywords: response set, socially desirable responding, self-deceptive enhancement, item response models

Although self-reports have provided important information in many applications, the limitations of this approach become apparent when studying intentions and private behaviors. In this case, researchers cannot rely on the assumption that self-reports are unbiased, candid, and accurate. The talk present a modeling framework to obtain unbiased self-reports using an item-response theoretic approach. The approach is based on the established assumption that respondents answer personal questions following a three-stage process. First, they arrive at an initial response based on retrieval processes and, subsequently, they decide whether to edit this response and to report a more positive or less revealing answer instead. Several experimental studies on socially desirable responding provide direct support for this conceptualization by showing that social desirability affects the editing process during which participants evaluate retrieved information before responding. The proposed model also assumes that the response-formation and the decision-to-edit stages may be correlated in the sense that individuals whose retrieved response indicates noncompliant behaviors, say, may also be more likely to edit their self-report. Moreover, response-set effects are allowed to be question- and person-specific because questions may differ in the degree to which they elicit editing and participants may differ in the degree to which they edit their self-reports. As a result, it is possible to relate covariates to the hypothesized stages which may prove critical for understanding possible determinants of the response-set effects. By controlling for the effects of editing and response selection, more valid measurements of the behavior in question may be obtained. The proposed approach is applied to a large-scale experimental study and shows that respondents who feel powerful tend to overclaim their knowledge and feel more certain about their knowledge.