How much expense will anchoring vignettes add?

There are several sources of additional costs:

1. Survey administration time: Obviously, vignettes will take up more time on the survey. This cost can be ameliorated to a degree because our model allows vignettes to be asked of randomly selected subsets of respondents, if you have variables capable of predicting variation in respondents' thresholds. If not, then you can use the nonparametric model, which only applies for respondents who are asked both vignettes and self-assessment questions.
2. Translation costs: Three vignettes asked of one third of the respondents each will add the equivalent of only one additional item in expense, but it will also add three questions to translate. Some of these costs can be ameliorated by using vignettes and self-assessment questions asked in other surveys, such as those in the World Health Survey.
3. CATI costs: Asking questions of random subsets makes computer assisted interviewing techniques helpful.
4. Costs of anchoring vignettes can be held down by choosing them appropriately. For example, choosing two vignettes that are very close to each other will provide repetitive information and thus will be wasteful. Similarly, vignettes that are too extreme can provide little or no information.
5. The costs of adding anchoring vignettes to surveys should be weighed in the context of the potential benefits of the approach and the costs of not adopting the procedure. Anchoring vignettes provide the only currently feasible method of testing for DIF, and a good way to correct for it. If you are reasonably sure you have no DIF, then anchoring vignettes will at least provide you the opportunity to verify this hypothesis empirically.