A Repeated Nested‐Logit Model of Atlantic Salmon Fishing

Abstract

Participation and site choice for Atlantic salmon fishing are modeled in the context of a repeated three-level nested-logit model. Consumer's surplus measures are derived for different levels of species availability in the Penobscot River, the most important salmon river in New England. For comparison, six other travel-cost models are estimated. These include restrictive cases of the nested-logit model, a partial demand model, and two single-site demand models. Comparisons across these models indicate the importance of modeling the participation decision, including income effects, and of adopting a nested-logit structure rather than a single-level logit structure.Key words: Atlantic salmon fishing, nested-logit, travel cost model.Discrete-choice random-utility models are well suited to explain how an individual will choose among a discrete number of alternatives as a function of the costs and characteristics of those alternatives. Such models all share two components. Conditional indirect utility functions are specified for each of the alternatives, and each of these conditional indirect utility functions has a component that is random from the analyst's perspective. If each of the random components is independently drawn from a normal distribution, the model is a probit model; if each of the random components is independently drawn from an extreme value distribution, it is a logit model; and if the vector of random components is drawn from a generalized extreme value dist¡ it is a nested-logit model. In all discrete-choice random-utility models, the deterministic part of the conditional indirect utility function for each alternative is a function of the individual's income, the cost of that alternative and the characteristics of that alternative. We also thank the Atlantic salmon anglers who participated in the study.Review coordinated by Richard Adams.cost for alternative j enter the conditional indirect for alternative j in the separable form of income minus cost of alternative j; i.e., expenditures on the num› given alternative j is chosen. If expenditures on the num› enter each conditional indirect in a linear fashion with the same coefficient for each alternative, the probabilities associated with each alternative are nota function of the individual' s income and the compensating va¡ CV, associated with any change in costs or characte¡ is equal to the corresponding equivalent variation, EV. In this case, the CV is nota function of the individual's income. Alternatively, if expenditures on the num› enter one or more of the conditional indirects in a nonlinear fashion, income effects are present. While logit models of recreational site selection that do not incorporate income effects are quite common, ones that include income effects are rare. ~ Reasons for not including income effects are that it complicates estimation, and the fact that, when income effects are incorporated, the CV and EV do not have closed-form solutions.We develop and estimate a repeated three-level nested-logit model that explains, for e...

Controlling for Observed and Unobserved Site Characteristics in RUM Models of Recreation Demand

“…Unfortunately, variation in the price data is frequently offset by a paucity of information characterizing the attributes of the sites themselves. Researchers are often limited to one or two measures of site quality such as fish catch rates (Chen, Lupi and Hoehn [5] and Morey, Rowe and Watson [20]), fish toxin levels (Phaneuf, Kling, Herriges [23]) or dummy variable indicators capturing different levels of water quality (Parsons, Helm and Bondelid [22]). 3 The risk in this setting is that unobserved site attributes may be correlated with the observed attributes or travel costs (or both), leading to omitted variables bias for the estimated parameters and biasing any subsequent welfare calculations.…”

Valuing the recreational benefits of wetland biodiversity

“…Thus, the repeated discrete choice (hereafter RDC) model, which is one of the random utility models, was used in this study. Morey et al (1993), Needelman and Kealy (1995), Shaw andOzog (1999), andSuwa (2008) are previous studies of the RDC. An advantage is flexibility on benefit calculations.…”

“…Here, ICM equals to Yi in the RDC model; the variable of mi was denoted as TICM. The positive sign of the parameter of income allows us to calculate benefits from quality changes under the concept of compensating variation (Morey et al, 1993). The signs of AREA and TNRS are supposed to be positive following Whitehead et al (2009) and Carson et al (2009).…”

Model Uncertainty in Characterizing Recreation Demand

“…The model is similar to the repeated nested logit model (Morey, Rowe and Watson (1993)) and repeated mixed logit model (Herriges and Phaneuf (2002)). These models integrate individuals' choice among alternatives and the problem of allocating time between multiple recreation sites.…”