Citation:

Merrill, N., J. Bousquin, M. Mazzotta, AND K. Mulvaney. Regional Demand Models for Water-Based Recreation: Combining Aggregate and Individual-Level Choice Data. The Northeast Agricultural and Resource Economics Association Annual Meeting, Bar Harbor, ME, June 19 - 21, 2016.

Impact/Purpose:

This paper presents the a multi-site recreation participation model, in which the underlying choice behavior assumptions are based on the RUM framework of site choices, which can be combined with both observed visitation data and existing population-level participation data. We apply the model to valuing the effect of water quality changes on saltwater recreation on Cape Cod, MA.

Description:

Estimating the effect of changes in water quality on non-market values for recreation involves estimating a change in aggregate consumer surplus. This aggregate value typically involves estimating both a per-person, per-trip change in willingness to pay, as well as defining the market extent that is affected by this change. The change in aggregate consumer surplus will be affected by variations in both estimates (the per unit change as well as the total number of people or trips affected). It has been shown that the total number of people or trips is often a more important determinant of aggregate costs or benefits than the value per trip (Hanley 2003, Bateman 2006, Kozak et al. 2011). Therefore, when the objective is to estimate changes in aggregate consumer surplus, attention should be directed to estimating both the value per person and the number of affected people. This paper presents a multi-site recreation participation model, in which the underlying choice behavior assumptions are based on the random utility model framework of site choices, which can be combined with both observed visitation data and existing population-level participation data. We apply the model to valuing the effect of water quality changes on saltwater recreation on Cape Cod, MA.Recreation data that might be used for valuation come in a wide range of formats, including data on discrete trips, data on visits over various time periods, and population-level participation rates from national surveys (such as the National Survey of Recreation and the Environment), to name just a few. There have been attempts to combine these types of data in various ways to estimate changes in consumer surplus, including using travel distance distributions and benefit transfer (Mazzotta et al. 2015), RUM-based Bayesian approaches (Phaneuf et al. 2015), or linked count and site choice models (Lupi et al. 2005). This work builds on these concepts, but tailors them to be consistent with aggregate visitation observations by specifying the likelihood function based on the distribution of aggregate participation at sites. This will allow us to explore the additional information provided by incorporating aggregate trip estimates with individual visitation data. Combining existing spatial population, participation, and travel behavior data with a benefit transfer of per-trip values, we will conduct a “low-information” application for valuing recreational effects of changes in Cape Cod estuary water quality related to eutrophication. The framework is built such that site-specific data, such as a future Cape Cod-specific intercept survey, or more site visitation data could be added to the model to update the low-information case. By specifying the model in a general way, it allows for a machine learning process as new information becomes available, while providing information to address immediate policy-relevant decisions using the combined participation and benefit transfer model. Both parameter estimates, as well as assumptions of functional form (choice structure) are flexible, allowing for a dynamic learning algorithm of recreation participation. An application of the low information concept is presented for Cape Cod, with associated pseudo-code for the learning algorithm to take advantage of the availability of future data.

Record Details: