chapter  16
14 Pages

Gauging the value of short- term site closures in a travel- cost random utility model of recreation demand with a little help from stated preference data


Introduction Random utility models of recreation demand are well suited for valuing closure of sites and changes in the characteristics of sites such as an improvement in water quality or increase in fish catch. In these applications the welfare effects are realized through site substitution or the choice of taking no trip on a given choice occasion. Parameter estimates from the models are used to measure the decline in utility implied by substitution along these lines and the coefficient on a trip cost variable, in turn, is used to monetize the change in utility. The models, however, ignore the possibility of substitution across choice occasions within a given season in response to a site closure or change in site quality. For example, a closure of a beach for a weekend or two weeks may result in people delaying trips to the closed site until later in the season. In effect, these people are substituting across time instead of sites. This is a common occurrence in damage assessment cases where the short-term closure of a site may have little impact on the total visitation to the site over a season implying that people have delayed trips in response to the closure.1 We have designed and estimated a random utility maximization (RUM) model that accounts for substitution over time and use it to gauge the value of short-term closures. The model combines revealed and stated preference data and is applied to beach use on the Texas Gulf coast. The data were gathered by a phone survey in 2001. Respondents were asked to report information on trips to 65 beaches during the year.2 As part of the survey, all respondents visiting the Padre Island National Seashore (14 percent of the sample) were asked if they would have visited another site if Padre had been closed. If they responded yes, they were asked to report which site. If they responded no, they were asked if they would take a trip later in the season to “make up” for their lost trip to Padre. These stated preference (SP) data along with the reported trip revealed preference (RP) data are used to estimate a RUM model where a trip to Padre later in the season is treated as an alternative in the choice set. This allows us to estimate the utility for delaying a trip versus making a trip to another site and, in turn, to estimate the loss of a

beach closure at Padre Island that accounts for substitution of delayed trips. Our approach does not use a dynamic choice model over a season. Instead, it offers a practical alternative; we believe a strong alternative given the limitations of our simple SP follow-up choice question. Part of our motivation for pursuing this topic was the large number of respondents who reported that they would “make up” a lost trip by visiting Padre later, leading us to wonder if conventional RUM analysis might be missing a key behavioral response to a closure and hence measuring welfare loss inaccurately. The SP questions are shown in Figure 16.1. If a person responded “Definitely Will” or “Very Likely” or “Likely” to the follow up question we assume that they will take a make-up trip. There are drawbacks to this question format. First, instead of “closure,” it uses the wording “. . . if you had not been able to visit . . .” which, in principle, could be due to reasons other than a closure. If people took a broad enough interpretation, such as being sick or the traffic being bad, there may be a great deal of latitude in our response data, and that is a reason for caution. Furthermore, even if respondents are thinking in terms of closure, the assumed reason for the closure could vary in the minds of the respondents. Oil spill? Red tide? Other? If behavioral responses are sensitive to the reason for the closure this will lead to even more latitude in the response data. Second, it does not mention how long Padre will be closed. The length of closure will affect the amount of time available during the balance of the season for a make-up trip, when during the balance of the season make-up trips would be possible, and even how many trips may be affected. We, in effect, assume a single day closure in our modeling strategy. Third, the construction of the follow-up question is

such that everyone is asked if they would take a make-up trip to Padre, even if they choose another site in the first question. Only a small fraction choose both and we treated them as choosing another site only. Fourth, with any stated preference response data, what people say they will do and what they actually do can be quite different things. In the recreation demand field there is some evidence that people overstated their expected number of trips for a season when asked at the beginning of a season. We may have a similar optimistic overstatement in our make-up trip data. So, our data clearly come with some caveats, and we urge the reader to keep these in mind when interpreting or using our results. We find that accounting for delayed trips to Padre reduces the estimated welfare loss by about 70 percent. In the next two sections we present our model and study design. Then, we turn to a short presentation of the data and the results.