Thursday, April 6, 2017

Pick a number, but not just any number: valuation uncertainty and maximum willingness to pay

Abstract:
Empirical results suggest that contingent valuation method (CVM) respondents are uncertain of their valuations, which has led critics of the method to raise issues about its validity. Alternative approaches to resolve the problem have been proposed, involving different willingness to pay (WTP) response formats allowing respondents to explicitly express uncertainty. This paper compares differences between certain and uncertain responses for four different response formats. The results suggested that mean and median (WTP) were not significantly different for respondents who were certain about their valuations. This was generally not the case for respondents who were uncertain about their valuations. However, the median WTP was not found to be significantly different for uncertain and certain respondents. A conclusion for a standard CVM application is that the sample median WTP value could serve as a proxy measure of population maximum mean WTP when uncertainty has been removed.
...
The analysis is based on data from a mail survey on outdoor recreation in southern Sweden.... Some previous studies targeted visitors to a specific recreation area with corresponding high familiarity with the good being valued . By contrast, this study targeted ‘an average visit’ to an unspecified forest area the respondent uses based on the Right of Public Access. This customary right allows people to freely roam the countryside and put up camp for one night regardless of land ownership [which] ... is not traded in any market, implying that it has the characteristics of a public good ...
Beechwood Forest Skane Sweden (Nature)
The resulting sample would ... be expected to involve a wide range of familiarity with the good being valued....The hypothetical valuation scenario was the same for all questionnaire versions.... The ... valuation question [was] worded as follows:
Imagine that you would continue to visit and experience the forest as you normally do. You would travel the same distance to the forest and do similar activities. Consider that your expenses to visit the forest would increase, i.e. became larger than what you stated in the previous question. [which referred to the total cost related to an average forest visit, authors’ comment] How large would your total maximum cost for an average forest visit be before you would decide not to visit the forest?
The respondents were thus faced with the task of expressing their valuation as an equivalent variation (ev), i.e. the maximum WTP to avoid losing their current forest recreation.

The four response formats were open-ended (OE), interval open-ended (IOE), multiple-bounded dichotomous choice (MBDC) and what we have here chosen to call line scale (LS).... Three (OE, IOE and MBDC) of the four samples included a follow-up question on valuation certainty. Information was given that 100% probability would mean 100% acceptance of the stated amount and 0% probability would mean 0% acceptance of the stated amount, but the respondent was allowed to state any probability within those values....
...
[Responses are presented in the table below]
The OE survival curve is generally bounded by the IOE lower and upper bounds. The ranges differ substantially, as the highest reported value in the OE format was SEK 5000, whereas it was SEK 10,000 for the upper bound of the IOE format. Consequently, respondents gave considerably higher values than in the OE format when explicitly asked for their upper bound valuations in the IOE format. By weighting the responses with the reported uncertainties, the response distributions became more clustered as seen in Figure 4.

Figure 4. Survival curves for open-ended responses adjusted for valuation uncertainty.
...
In the unadjusted calculations (UNADJ), the respondent is assumed to be completely certain about her/his WTP response which is the usual assumption in CVM. For the OE and IOE unadjusted calculations, the OE_UNADJ mean (and median) value was contained within the lower and upper bound mean (and median) values of IOE_UNADJ (similar to the survival curves.... All unadjusted responses produced a mode value of 100 SEK....
...
The adjusted calculations took into account the respondent's stated probability of acceptance of their corresponding stated WTP values.... The median values are 100 SEK for all adjusted question formats, with the exception of the upper bound IOE question (80 SEK).
...
The subsample of respondents who stated 100% probability of accepting their stated WTP value were selected to calculate the central tendency measures for OE_CERT and IOE_CERT.... The median WTP values were consistently close to 100 SEK in the certainty-adjusted cases, as well as for those respondents who claimed a 100% certainty in their WTP.
...
It was concluded that socioeconomic differences did not have a significant effect on the distribution of WTP for the different question formats. Mean estimates fluctuate between 212 and 591, whereas median stays between 112 and 150. The estimation results for the certain and unadjusted OE and IOE responses are displayed in Table 5.

Conclusion
Past research has shown that respondents are uncertain of their valuation, which has motivated insight into alternative methods to resolve the problem. One such proposed approach involves alternative question response formats allowing respondents to explicitly express their uncertainty with regard to their stated WTP value of a non-market good or service.  Keske and Mayer (2014 Keske, C.M.H., and A. Mayer. 2014. Visitor Willingness to Pay U.S. Forest Service Recreation Fees in New West Rural Mountain Economies. Economic Development Quarterly 28 (1): 87–100) used a 10-point Likert scale to determine the certainty of reported WTP for recreationists asked about paying a fee at Quandary Peak, a fourteener in Colorado. Ninety percent of the respondents reported a high degree of certainty in their WTP values. Our research produced expectedly different results as we asked a random sample of the whole population in a region with varying degrees of recreation participation about their WTP for an average forest visit, which could be expected to lead to a lower valuation certainty. This paper also investigated the differences of certain and uncertain responses....

The different WTP question formats resulted in different levels of certainty....

The median was not found to be significantly different from the mean in the case of ‘100% certain’ responses. In all other cases except MBDC, the mean was found to be significantly larger than the median. These findings suggest, as indicated by theory, that the ‘100% certain’ responses conform most closely to the representative individual model, as the mean and the median coincide.... This is then a new argument in favour of using the median rather than the mean in applied economic welfare analysis. To further test this suggestion, future research should look at the population distribution of WTP and also investigate actual payments and the associated certainty, as was done by Whitehead, Weddell, and Groothuis (2016 Whitehead, J.C., M.S. Weddell, and P.A. Groothuis. 2016. “Mitigating Hypothetical Bias in Stated Preference Data: Evidence from Sports Tourism.” Economic Inquiry 54: 605–611.

by Mattias Boman and Lindsey (Ellingson) Doctorman
Journal of Environmental Economics and Policy http://www.tandfonline.com/toc/teep20/current via Taylor and Francis http://www.tandfonline.com/
Published online: 02 Mar 2017 Pages 1-22
Keywords: Contingent valuation, willingness to pay, valuation uncertainty, response format, outdoor recreation

No comments:

Post a Comment