ApolloChoiceModelling forum

Dear Prof. Hess and Team,

I have conducted a Stated Choice Experiment for the willigness-to-subscribe / willigness-to-pay for a subscription service.
The choice was between 3 plans with different characteristics and prices or the opt-out option.
I have estimated a nested-logit model showing a way too high willigness-to-pay. I would strongly assume this is due to the hypothetical bias.

As I already imagined this might happen, I have included an second experiment in the survey to get (kind of) revealed preference data. In particular, the participants were asked to select a voucher for the subscription service or a lower cash amount as a reward for taking the survey. (With a 10% chance of actually getting this reward.)

The first choice was beween a 30$ voucher and 1$ cash, which was increased in 5$ increments until cash was selected or 30$ in cash was reached.
A substential part of the sample actually selected very low cash amounts despite previously stating their willigness to subscribe for a rather high price.

I am now thinking if there is a sensible way to include this data into my SP model to correct the willingness to pay.
However, in contrast to typical approaches of combining RP and SP data (e.g. Buckell & Hess, 2019) I cannot directly use the cash/voucher price attribute, as it has been a different setting.

Do you have any ideas what might be suitable direction or do you have a paper in mind that could be helpful? My current idea would go into the direction of scaling the price parameter in the SP model, but I am a bit stuck on how to do this.

Hi

first, in relation to your SP model, you could try to refine the specification and see if that helps. Feel free to show us your current results on the forum as we may then have more suggestions.

second, in relation to joint "RP"-SP, your datasets look quite different, but you could of course try to have a joint model. However, there is a bigger issue. Your second dataset is potentially affected by endogeneity as you change the values of the attributes depending on what they choose

Stephane

Dear Stephane

Thanks a lot for your response.

The method for the cash/voucher I described was only used in this way to find the "jumping point". So the result is a value like $10 or $15 at which cash is preferred. I could have just asked to enter this value, but then the decision might have been biased. So while the individual choice situation might be endogenous, the jumping point itself would still be exogenous, right?

May I ask if you could give me a few more hints how a joint model may look like? Do you have any papers in mind that do something related? Unfortunately, I was not able to find anything.

You could use a set of logit models for this, each time with the choice between the values you show them