https://orcid.org/0000-0002-3394-138X
![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
The Cook Inlet beluga whale (CIBW) is an endangered whale found in waters off the state of Alaska that is protected under the U.S. Endangered Species Act. The federal recovery plan estimates recovery costs will total $73 million (in 2013 dollars). In this study, we use data from a stated preference discrete choice experiment (CE) study to estimate the aggregate benefits of recovering the CIBW and generally for improving its conservation status and reducing extinction risk. Estimated CE models account and test for utility scale heterogeneity, attribute non-attendance, self-selection bias, and demographic effects. Aggregation methods that differ in adjustments made to the mean household welfare and/or to the number of population units are compared by assessing their impact on the resulting aggregate (population-level) welfare estimates. Alaska households are willing to pay between $34 million (95% CI of [$25 million, $44 million]) and $103 million ([$74 million, $131 million]) for full recovery of the CIBW, depending upon the model and aggregation assumptions. While some of the state-level recovery value estimates are below the total cost of combined federal and state recovery actions, accounting for welfare benefits beyond the state of Alaska justifies recovery actions by the benefit–cost criterion.
JEL Classification:
Acknowledgement
This article and its findings are those of the author and do not necessarily reflect the views of the National Marine Fisheries Service or the U.S. Department of Commerce. The author thanks Kristy Wallmo and an anonymous reviewer for useful comments that improved the paper. All remaining errors are the author’s alone.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Details not provided here about the survey design, pretesting activities, and survey implementation are available in two other studies (Lew Citation2018; Citation2019), which each used a different subset of the data and explored other issues.
2 Note that the rank choice (best-worst) format was employed in numerous CE studies in economics (e.g., Scarpa et al. Citation2011; Lancsar et al. Citation2013) around the time of this survey due in part to advantages the format has for estimating efficient model parameters. However, along with other SP question formats, it has since been used less after concerns were raised about incentive compatibility properties of SP question formats (Carson and Groves Citation2007; Vossler, Doyon, and Rondeau Citation2012). In particular, Vossler, Doyon, and Rondeau (Citation2012) showed that a CE design exhibiting a particular set of properties ensure it is incentive compatible. Among those properties are that the choice design uses binary referenda between one alternative and the status quo, has independence of choice sets, has at most one alternative from the universe presented being implemented, and is viewed as consequential by respondents. The rank choice format used here does not have all of these features. Thus, it is one that cannot be assured to be fully incentive compatible, despite having features that support its consequentiality. As a result, there is the potential for strategic bias in the responses.
3 Focus group and cognitive interview testing suggested people generally assumed the federal tax was certain and binding and would make up most, if not all, of the ‘added cost’ to their household, which reduces but does not eliminate concerns about this payment vehicle not being fully ‘fixed and unmalleable’ (Johnston et al. Citation2017).
4 ‘Recovery’ is defined as reducing the extinction risk to a near-zero level and removing the species from the ESA list of threatened and endangered species.
5 A two-step process was used to select the experimental design used in the study. First, a programme developed in GAUSS selected the 20 most D-efficient experimental designs based on a main effects utility specification from among 100,000 randomly drawn candidate experimental design combinations that excluded fully dominating choices. Then a series of Monte Carlo simulations were done to evaluate how well each of the 20 designs performed when analysed with a main-effects rank-ordered multinomial logit model using pseudo-data constructed from several sets of assumed true utility parameters. Designs were evaluated in terms of their ability to estimate the true utility parameters, and the best performing one was selected for use.
6 Rural households were oversampled (50% of the mail-out sample), so the estimation models presented here are sample-weighted. The sample was drawn from the Marketing Systems Group’s address-based database of U.S. households, which has nearly 100% coverage of households in Alaska since it is based on the U.S. Postal Services’ Computerised Delivery Sequence File. The survey implementation was carried out by Ipsos.
7 While the survey was initially intended to be conducted on a random sample of U.S. households, administrative approvals limited the scope of the data collection to Alaska households.
8 Response rates varied from 33% to 55% over the 48 survey versions. The response rate excluding undeliverables also excludes those who were deceased or moved out of Alaska (273 total undeliverables).
9 719 from urban areas and 597 from rural areas.
10 The specific question wording is provided in Appendix Figure A2.
11 Allowing the cost parameter (γ in equation 3) to be normally distributed led to model results that suggest a significant percentage of parameter values in the distribution are in the negative orthant, which is counter to theory. We imposed consistency with demand theory by restricting the cost parameter to be positive, which meant estimating models that specify the cost parameter to be fixed, lognormally distributed, or triangularly distributed. Since the lognormally-distributed cost parameter models failed to converge, we restrict ourselves to the triangularly distributed cost parameter models (and fixed cost parameter model results are available upon request).
12 The models were programmed in GAUSS version 22. Halton and Sobol quasi-random draws of varying sizes (500 to 2,000) were evaluated and model parameter stability supported use of 2,000 Sobol draws.
13 Other model specifications that are quadratic in RR had similar model fit and statistical significance but allowed for negative marginal utility for large extinction risk reductions. Here we impose that the utility associated with increases in RR are non-negative.
14 Models that also included cost attribute non-attendance were estimated but not used since likelihood ratio tests rejected the cost attribute being ignored.
15 RPDEV interactions with other utility variables were consistently not statistically significant.
17 Fixed parameter conditional logit specifications of these models indicated statistical significance of ANA parameters for both ESA status and extinction risk reductions, but the lack of significance in the random parameters specifications highlights the potential for ANA to be confounded with preference heterogeneity.
18 The 95% and 90% CIs for these aggregate WTP value estimates, respectively, do not contain the aggregate recovery cost amount, suggesting the costs are statistically lower than benefits in these cases.
19 This assumption was made to illustrate the point that aggregating welfare beyond the state-level leads to aggregate net benefits exceeding costs even when only a small percentage of households have similar levels of WTP as those estimated for Alaska.