WTP design to avoid lost load- sign's confusion

[izakro]

Hi all,

We conducted a pilot study intended to guide a WTP design to avoid interruptions in electricity supply (in terms of the number of shutdowns, avg. during of each, whether an alert will be given etc.) In the pilot, respondents were noted that:

Each alternative will include a payment that will be added to your electricity bill to prevent the level of non-supply described in the alternative.

The pilot data were analyzed in MNL and the results were as expected. All coefficients were significant.

Estimated parameters
Estimate
b_cost -0.017
b_duration -0.018
b_shutdowns -0.298
b_alert 0.321

When using these priors to prepare the final design, I found myself confused.
It feels that the signs of the coefficients should appear oppositely, that is:

Code: Select all

U(optiona) = cost[(n,-0.017, 0.002)] * cost[8, 32 ,56, 80, 104, 128] 
         + duration[(n,[b]0.018[/b], 0.001)]  *  duration[15, 30,60,120,180,240]
         + shutdowns[(n, [b]0.298[/b], 0.052)] * shutdowns[1,2,3,4]
         + alert.dummy[(n, [b]-0.321[/b], 0.076)]*alert[1,0]   


since the participant will be willing to pay more to avoid more frequent and longer shutdowns without alert, what am I missing here? I can't sort that out.
Thanks for any help!
RR.

[Michiel Bliemer]

Just looking at your utility function, note that coefficients/parameters in the utility functions are not WTP values. For example, a negative coefficient for duration simply means that an alternative is less preferred if duration is longer. The WTP for a shorter duration is (-0.018)/(-0.017) = 1.058 (in whatever units you chose for cost and duration), which is positive.

Michiel

[izakro]

Thanks, Michiel
I understand what you are saying
I think that the WTP responses have been somehow influenced by their immediate appearance after the WTA set, and given the fact that this is WTP to avoid a (bad) service.
The results of the WTA (compensation for unreliable energy supply) are as follows:

Estimate
b_cost 0.0016
b_duration -0.0043
b_shutdowns -0.019
b_alert 0.4027

Do they make sense?
Can I use these priors for the WTP (to avoid supply interruptions) with opposite signs? (i.e. cost and alert will be positive and duration and shutdowns positive)?
Thanks!
R.

[Michiel Bliemer]

It is difficult for me to advise here since I was not involved in the study and do not really understand the choice experiment. If you believe other signs make more sense to you then you can change the signs of the priors. If you worry about using wrong priors, you can always use zero priors as a safe choice.

Michiel

[izakro]

Thanks, Michiel
last question.
Our design looks like this (a scenario example):
https://https://1drv.ms/i/c/317ea84f1036d885/EUSHUpkcRfVKufdaLN0D8OUBtjAx1Rmua1fE1TNMSqzElA?e=WmJUh9

To integrate a (hypothetical) status quo option, we thought of describing it by implementing the 'worse' attributes among the two alternatives in each menu (llongest duration, the maximal number of shutdowns, not alert and zero payment).

Is this a reasonable approach?
Thanks!
R

[Michiel Bliemer]

That link gives an error for me.

Why would you be interested in a status quo option that describes the worst option? A status quo option is usually included to look at the demand of a new product versus an existing product (whereby the status quo is the existing product). It is not clear to me what the usefulness of a choice probability for the worst option would be? Of course there is nothing stopping you from including such an option, and it may be useful to capture more trade-offs within a single choice task, but it would not be useful for computing market shares or demand forecasting.

Michiel

[izakro]

Sorry for the link, I don't know why it gives an error on your side. I tried to attach an image but the Img button doesn't allow it.

The rationale for the status quo is as follows: Since Israel is considering adopting a demand flexibility system, I thought that by adding a status quo (describing the potential service interruptions due to increased demand) it would be easier to capture less energy-sensitive households.

Looking at several papers including a 2018 paper (see reference below), Fig. 2 provides a menu for illustration. If a respondent chooses option 1, does it mean that they avoid option 1's supply interruption, but in return are exposed to option 2's supply interruption? This appears in many papers, and I feel that respondents may not fully understand the meaning of their choices.

Reference: Morrissey, K., Plater, A., and Dean, M. (2018). The cost of electric power outages in the residential sector: A willingness to pay approach. Applied Energy, 212, 141–150.

To avoid this issue and to better capture residents with inelastic demand, I considered adding a fixed status quo option (perhaps with a worst-case scenario).

[Michiel Bliemer]

If the worst case is the status quo (what would happen in a do-nothing scenario) then it could be fine. This is the same as in health, where the status quo option of 'no treatment' is often the worst health outcome.

[izakro]

Thanks a million!