choice-metrics.com

Dear all,

in my CBC I am using percentages as levels for my price attribute.
As these percentages are quite low I was thinking about showing reference amounts to help respondents decide what a percentage means in absolute amounts.
Besides showing general reference amounts like: "1,8% of an average monthly income is 100€" I might add respondent specific amounts as reference. I would ask the respondent for his monthly income at the beginning of the survey and would show as an information below each CBC task:
"1,8% of your monthly income is 250€"

Now my question is:
I´ve read that the attributelevels should not be respondent specific as they need to be independent from respondents.
Do you see that my idea harms this assumption or can I add this respondent specific reference amounts without any doubt? (The attribute levels are always 1,6%, a,8%, 3,3% for each repsondent)

Any help is highly appreciated.
Thanks and best regards,
J.

It is not a problem that attribute levels are respondent specific. Clearly this is reality, different people experience different attribute levels as you would in revealed preference. Therefore, when trying to make choice tasks more realistic for each respondent one would often create respondent-specific attribute levels.

For example, Rose et al. (2008) described respondent-specific reference alternatives with respondent-specific attribute levels.

Rose, J.M., M.C.J. Bliemer, D.A. Hensher, and A. Collins (2008) Designing Efficient Stated Choice Experiments in the Presence of Reference Alternatives. Transportation Research Part B, Vol. 42, pp. 395-406.

You can still use a single underlying design if you like, for example an orthogonal design. But you can also optimise for difference income classes by creating a separate design for low, medium, and high income classes where you use specific attribute levels and priors for each design. There is a lot of flexibility in what you can do.

Michiel

Dear Michiel,

thank you very much for your answer.
Just to get you right: You do not see any issue in generating a single design for all respondents and adding kind of a "further information for reference" section below each choice set showing:
1,6% means 48€ based on your stated income
1,8% means 54€ based on your stated income
3,3% means 99€ based on yours stated income

It would not be inherent part of each choice set in the levels though but a general, respondent specific information below each choice set.

Thank you for re-confirming.
Best regards,
J.

You can express a signle experimental using design coding, e.g. using attribute levels 0, 1, 2, ... etc.

Then depending on the type of respondent, level X means something differrent for a respondent with low income than a respondent with high income, but all that changes is assigning the exact text shown in the survey to this level. In other words, the design remains the same for all respondents, but the labels you attach to each attribute level (i.e., the text or image you display) varies by respondent.

Michiel

Now everything is perfectly clear, thank you very much Michiel!
Your support is highly appreciated!