Validation of theoretical assumptions with real and falsified survey data

摘要

Falsification of survey data in face-to-face surveys has been intensively discussed in the literature. The results about the impact of falsifications on survey data are equivocal. While some authors report a strong impact, others find only little differences between real and falsified data. We argue that the impact of falsifications cannot be neglected, particularly when theory-driven analyses are conducted and not ad hoc analyses. The latter reproduce stereotypes used by both, researchers and falsifiers. To test this assumption we compare the results of multivariate regression analyses with real and falsified data by using a) theory-driven predictors and b) ad hoc predictors. As an example of theory-driven analyses we used the theory of planned behavior (TPB) for predicting self-reported healthy eating behavior. As ad hoc predictors we included socio-demographic information about the respondents known to the falsifiers as well as variables, which are indicated by everyday theories. The results show that theory-driven relationships were more strongly pronounced in the real data. In contrast, stereotypical and non-theory-driven relationships were more strongly pronounced in the falsified data. The results provide insights in the area of social cognition when predicting the behavior of others.