How to compute willingness to pay based on different types of surveysGood evening,
I am trying to design a survey using JMP, and my objective is then to use JMP to compute the customers' Willingness to pay based on the answers.
The aim is to evaluate the price customers would be willing to pay for different services. For example, let's say that the customers want to rent a house for a few days, the final objective is to know how much more they would be willing to pay to add Wi-Fi (with different speeds), laundry service (with different options), ...
After reading a few papers, I came up with different types of questions. However, I am not sure of how to compute Willingness to pay with the answers.
Examples of types of questions : (the prices would change with the respondents, being "attributes" for the services)
1) How likely would you be to purchase high-speed Wi-Fi at $7/day ?
- Very likely - Likely - Neutral - Not very likely - Not likely at all
Then I could attribute a score (1-5) to each answer, and try to compute the utility of the service based on that ? Do you have an idea of how I could do that ?
2) Would you rather purchase...
- Low-speed Wi-Fi at $3/day - High-speed Wi-Fi at $7/day - None ($0)
I think it is the closest to the "Choice surveys" offered by JMP. It allows to compute WTP using the Choice analysis, and choosing the "None" answer as a baseline.
3) You just rented a house for a day. You have now a budget of $25. Which items would you spend it on ? (multi-choice)
- Laundry service ($15) - High speed Wi-Fi ($8) - ... - Nothing ($0)
This should also allow to compute Willingness to pay, right ? How could I do that with JMP ?
My other ideas feel less relevant.
I feel that most of my questions do not really fit in the cases provided by JMP. What do you think ?
Thank you for your time,
Have a nice day !
AB
Re: How to compute willingness to pay based on different types of surveysI don't see this as a JMP question (although someone else might have something to suggest). To get price elasticity estimates, you will need a variety of prices and the number of respondents that are, or are not, willing to pay those prices. As we have discussed, the issue really surround eliminating bias from the responses. That is difficult enough, but I am more concerned about the difficulty of eliciting accurate responses - what incentive do people have to respond accurately (how do they even know what is accurate for themselves)? The only suggestion I have is to make sure you pretest whatever survey design you come up with. Perhaps even use a focus group after an initial survey with a small group of respondents.
Re: How to compute willingness to pay based on different types of surveysHi dale_lehman,
Thank you for taking the time to answer my question, and sorry for the late reply. I am mostly concerned about the treatment of data afterwards, using JMP or not (since does not seem to be really adapted to the question). I have read a lot of articles about the shape of questions, and how to avoid anchoring and that kind of things. But before starting to ask the questions, I would like to have a plan to analyze data, and be sure that with the answers to my survey I can come up with WTP and price elasticity, basically. Do you have an idea of a process in JMP that would allow me to do it for the sample of questions I posted ?
Thank you,
ABRe: How to compute willingness to pay based on different types of surveysI have not kept up with the literature, so I really recommend doing a thorough literature search. Environmental economists have probably done the most sophisticated survey modeling and testing in this area. From what I do recall, if you are going to ask questions along the lines of "would you be willing to pay $7 to get....?" you need t randomize the amount of money - and then you need to obtain a sufficient sample size to make reasonable inferences from. Another approach is to pose a series of binary choices; "would you prefer bundle X (e.g., $x + wifi vs. $y without wifi) or bundle Y?" I've been critical of those surveys in the past - they remind me of going to the optometrist and constantly being asked which view is clearer - and often not being able to tell. Notably, in that case, the choices are presented repeatedly to protect against mistakes - something that owuld be hard to do in a survey. One other approach I am aware of is to present a WTP scale with a wide range and ask people to place a mark that best represents their max WTP. The scale may produce anchoring, so you should at least vary the end points of the scale a bit to see if it influences the responses.
Re: How to compute willingness to pay based on different types of surveysGood morning dale_lehman,
Thank you for your answer !
I know that survey answers can be influenced by several factors, and some people in my team are working on that too. This is one of the reasons why I wanted to know how to analyze the answers to different types of questions, because we are trying to find a shape of question that would not influence people too much.
Unfortunately, I do not really have a choice about making a survey or not - there is no data available about the subject I am dealing with, and no way to get any, because most of it involves new kinds of services. It is not a really good way to get information, but it is the best way we found to do so. But in order to limit the bias/randomness, I would like to have more freedom when choosing the type of questions I will ask them.
We also studied incentives and survey distribution and we discussed that with the sponsor.
Do you have an idea of how I could get a theoretical willingness to pay from customers from the answers to the questions I proposed ? Or do you think of another shape of question that would be more appropriate ?
Thank you,
AB
Re: How to compute willingness to pay based on different types of surveysThere is a large environmental economics literature on this question (see https://www.epa.gov/environmental-economics/working-paper-robustness-vsl-values-contingent-valuation-surveys for an initial foray). The bottom line is that you must be very careful with surveys like this. Among the more pronounced issues is one of anchoring - the choices you present can easily influence the size of the willingness to pay estimates. It is easy to generate numbers from these surveys, but not so easy to generate numbers that are consistent and somewhat reliable. I'd prefer to base such estimates on observational data rather than surveys for many of these reasons. For example, if you have data on home sales that includes whether or not houses have broadband available, you can use a hedonic pricing approach to estimate the value of broadband. For me, asking people what they would be willing to pay is the last method I would use (an often overlooked problem is that, even if you avoid the many biases that haunt (it is Halloween after all) surveys, it is almost impossible to give people an incentive to think seriously about the question and give a truthful response. So, what you get from the survey may be unbiased, but almost random responses.
