willingness-to-pay cbc-hb covariates

Question

Hi Sawtooth Community,

I would be very grateful if you could help me with the following special request:

BACKGROUND: I have a dataset from a CBC and added a few covariates to my analysis (all zero-centred as described in Orme & Howell 2009: Application of Covariates in Sawtooth)
to test their effect on the preference for the attributes and their levels according to following logic:

Beta x = Intercept x + Parameter x × Expression Covariate

I got the intercept value and values for each covariate, the results are quite interesting and insightful (being aware of the downside of adding too many covariates to a model).

GOAL: Being aware of the downsides of WTP calculations with CBC data, I still want to try to convert the utilities for the intercept value and covariates into monetary figures (according to the WTP logic in Omre, 2001 Assessing the Monetary Value of Attribute Levels with Conjoint Analysis: Warning and Suggestions).

THE QUESTION: If I calculate the value in EUR for one utility point for the intercept, can I use this EUR-value/utility point also for the calculation for the WTP for covariates or do I need to calculate an EUR-value per utility point for each covariate separately?

I hope I expressed my situation clearly enough to use the power of this awesome forum!

I am very curious about your answers/thoughts?

Many thanks in advance!

Kind regards,
Alfons J.P.

Answer 1

I think it is easier and more straightforward to use the market simulation approach to estimating WTP (given all the caveats and problems with WTP estimation that I will not address in my comments).

Use your HB run wherein you applied covariates within Sawtooth Software's market simulator.  Specify two products in the market simulator: the product with the enhanced feature and the same product without the enhanced feature.  Change the price of the product with the enhanced feature until the simulated shares of preference (under the Share of Preference, logit, simulation method) achieves 50/50.  That price delta is a representation of WTP for the feature for the sample.

You can do this simulation approach for each of the market segments that you employed in your covariates (by using the covariate variable as a filter, to isolate the market simulation among a subset of respondents).  For illustration, if gender was a covariate, then you could perform this market simulation approach to calculate the WTP for males separately from females.

An extension of this approach for calculating WTP is to use more than 2 products in the market simulation.  You'd include the outside good (the None alternative) and all major competitive products.  In that case, you would adjust the price for the enhanced replicate of the product until their two shares were identical (and those identical shares would no longer be 50/50 because other alternatives would take up share).

Comments

Bryan, thanks a lot! That was very helpful!! I will try to implement your suggestion - I am curious about the results :-)

Nevertheless, coming back to my initial idea (even though being less meaningful, but I would like to compare these figures with the simulation WTP): would you use the price per utility point of the intercept value for calculating the WTP for covariate utilities or would you calculate for each covariate a price per utility point? Sorry for bringing it up once again...I not meaningful at all, you can ignore this question. Thanks a lot!