The role of zero-centered Diffs for ACBC utility aggregation

Question

Dear all,

In a nutshell I don´t get how the zero-centered diffs scaling enables the averaging of attribute utilities across the individuals of a sample, since it considers all attributes for one individual instead of one attribute for the entire sample. Can anyone shed some light on this please? I probably have an error in reasoning at this point.

Thank you very much!

Best,
Chris

Answer 1

In any respondent-level (or aggregate) logit model, the magnitude of the utilities is confounded with the reliability (the inverse of the amount of response error) with which a respondent (or respondents) made choices.  Zero-centered diffs is our way to try to ameliorate this but making the utilities of all respondents to have about the same magnitude.  Once we've done as good a job as we can of making the utilities comparable by putting them on the same scale then you can compare groups, examine averages and so on.  If you don't do this, then more consistent respondents will have greater weight in your averages and such than will less consistent respondents.