Climate Change Economics

Willingness-to-Pay Surveys, Forced Choice Experiments and "Revealed-Preference"

Economic analysis needs price data to weight alternatives. Prices are found in markets as suppliers and demanders interact. Some markets are less free than others (think about utility services, for which prices are set by regulatory commissions, as an example). But some things never enter the market at all: a good view, clean air, friendly neighbors, reduced carbon emissions and less impact on climate change, and many, many others.

For all these goods and services, some non-market mechanism has to be developed to determine some substitute "price" for economic analysis such as a cost-benefit analysis or other quantitative measure of which of a number of alternative actions is best. Economists use a number of different tools for approximating prices. The logic involved in three major approaches helps to illustrate the assumptions economists may make in estimating non-market "prices."

Willingness-to-pay surveys are sometimes conducted to get at the value of a good that will not be on the market. For example, a survey looking at provision of recreational services might ask, "How much would you be willing to pay to have a fishing lake within 1 mile of your home?" Then the survey might ask the same question for a 2-mile distance, a 5-male distance and so on, to see how much being close to such a lake is worth. This seems simple enough: in principle, the analyst can put the average price up for each distance and figure out from the line drawn through the points how much being closer is worth.

But there are problems that enter into the process:

1) If the survey offers specific dollar choices, the numbers offered shape the answer (you can't offer a positive amount less than the minimum or more than the maximum). Open-ended questions, where each person writes in an amount, give different results from those with dollar choices - but tend to give results that make little sense (a 2 mile distance might be preferred to both 1 mile and 5 miles, which is strange for a recreational fisherman).

2) The answers are not consistent for "how much per month?" versus "how much per year?" where the latter should be 12 times the former but very rarely is.

3) The question does not offer alternative recreational options: a swimming facility, a ball park, a playground or a hiking trail, for example. If all options are offered, the answers vary depending on the order in which they are asked.

4) The answers can vary depending on where the possible payer sees the money going. So the answer will change if the question reads "... pay to the city ..." as compared to "... pay to the park authority ..." and even more if the payment is tied to a particular purpose group of beneficiaries, as in "...pay to the part authority to introduce young people or the disabled to fishing ..."

This means that two surveys on the same issue can arrive at vastly different results even if they surveyed exactly the same people! You might not be able to tell why the difference arises if you didn't have the full survey in front of you.

But that's not all that enters the picture and makes establishing the value of something not in the market so difficult ...

5) The income of the respondent can influence the answer, and the influence is not always clear or consistent - the people with more money do not always express a willingness to pay more. (In the case of the fishing lake, once someone has an income above some level, he or she might be more likely to be willing to spend the time and money to drive to a far distant lake, so one nearby may not be worth much!)

6) Sometimes the answer is affected by immediate or short term political or economic factors, so answers obtained in one year (or one season of the year - say, when the weather makes fishing more or less attractive as an activity) may not apply to another time.

And, finally, the phrasing of the question as a positive or a negative has an overwhelming effect. Taking the lake example, this question can generate completely different answers from the first one above:

"How much would you be willing to pay to avoid pollution of a lake that is 1 mile from your home so you could continue to fish there?"

In other words, "willingness to pay for a gain" is not the same as "willingness to pay to avoid a loss."

Forced Choice Experiments address the fact that most real choices involve more than one factor, and that you can't ask everyone all the questions about different combinations of the elements of a choice. Think about that fishing lake again and add in some other features of the lake:

Distance: 1, 2, or 5 miles away

Size: 5, 10, or 20 acres

Stocked?: Yes or No to whether or not the lake is regularly stocked with fish for people to catch

Boats?: Yes or No to whether or not boats of any sort are allowed on the lake

Now, consider asking someone about how much he or she would be willing to pay for each combination of lake type. You would have to ask 36 questions and hope each person kept track of the features! (That 36 is 3x3x2x2, for all the different combinations of Distance, Size, Stocked, and Boats possibilities available.)

Instead, economists and statisticians have come up with a way of asked each person surveyed a subset of the choices and combine their answers to get results. Thus, each person might be 3 to 6 questions about discrete pairs of combinations, such as:

"Which of these options do you prefer?

A) A 20 acre fishing lake 5 miles away that is not stocked and does not allow boats

B) A 10 acre dishing lake 2 miles away that is stocked and does allow boats

How much would you be willing to pay for your preferred choice?"

Another way of posing the same forced choice is to add in a price for each of the options - but that would involve picking a series of specific prices to use, rather than leaving it open-ended as in the "How much ...?" question here. And the prices would make the set of choices even bigger: adding 5 price choices would mean 180 different combinations would need to be tested - and that would require a very large sample of people to be surveyed.

These forced choice questions enable analysts to look at multiple features than can change in a way that simple willingness to pay questions cannot. However, the approach suffers from all the problems we described above about willingness to pay with the issues multiplied by having to always pick some specific number of options for each aspect of a choice.

"Revealed Preference" is the economist's favorite method of assessing relative value to one choice over another where there is no market. In the case of the fishing lake, the idea would be to measure the number of people fishing at different lakes, and finding out how far they traveled to fish. The rationale for this approach is that people are "revealing" their true preferences by the actions they take when they have choices. This sounds reasonable enough - but the approach would break down if there were no lakes within a mile of some residents since no one could choose to only travel a mile to fish! (From a sampling point of view, it would also fail if there were not a significant number of people with the option of a fishing lake within a mile of their home, which is a much harder test to meet.)

Observed behavior is most likely an accurate reflection of the choices people prefer from among the options they see as available to them. But this raises a whole new set of questions about the people whose action is observed:

a) Do they really have choices? (Is a stocked, 10 acre fishing pond not allowing boats available within 1 mile?)

b) Do they know all the choices available to them? (Do they know about the 50 acre lake 20 miles away that is stocked and allows boats?)

c) Do they know all the costs and benefits to them from all the choices they perceive themselves as having? (Do they know and that there is a free express bus from their neighborhood running to the lake 20 miles away?)

d) If they know all the negative and positives that are associated with the different choices they have, do they have good "price" data on those aspects of their choice? (Even if they know about the free bus to the lake, do they really know how much it costs to take their own car and how long it might take them to find a place to park?)

Putting a price on non-market goods and services based on the preferences revealed through people's choices involves assuming that the answer to all these questions is "Yes." In reality, the correct answer is much more likely to actually be "No."