Estimating willingness to pay for greenhouse gas emission reductions provided by hydropower using the contingent valuation method

Energy Policy 111 (2017) 362–370

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Estimating willingness to pay for greenhouse gas emission reductions provided by hydropower using the contingent valuation method

MARK

⁎

Benjamin A. Jonesa, , Joseph Ripbergerb, Hank Jenkins-Smithb, Carol Silvab a b

University of New Mexico, Albuquerque, NM, USA University of Oklahoma, Norman, OK, USA

A R T I C L E I N F O

A B S T R A C T

Keywords: Hydropower Externalities Climate change Willingness to pay Contingent valuation Greenhouse gases Glen Canyon Dam

Altering existing operations of large hydroelectric dams in the US, for such reasons as improving downstream environmental habitats and recreation, often constrain the production of hydropower. This results in increased use of electricity from fossil fuel based power plants, which emit greenhouse gases (GHG) that promote global climate change. However, the economic value of hydropower GHG reductions remains unmeasured in the US context. Using a recent proposal to re-purpose operations of Glen Canyon Dam, the largest producer of hydropower on the Colorado River, this study estimates US households’ willingness to pay using the contingent valuation method to preserve GHG reductions provided by current Glen Canyon Dam operations. Results indicate that US households are willing to pay an additional $3.66 per year in increased taxes to prevent increases in GHG emissions due to proposed re-purposing. This study has important policy implications for the role of hydropower in the renewable energy portfolio.

1. Introduction

research on the economic value provided by hydropower externalities, especially in the US context where the extant literature is thin. In particular, the relationship between hydropower and climate change is an area of increasing public interest (Mattmann et al., 2016). Hydroelectric dams are seen as an important renewable energy source that help reduce GHG emissions. Former President Obama publically called for increased federal investments in hydropower in order to reduce US carbon emissions. The 2015 COP21 Paris Climate Conference included discussions on how hydropower can be used within a renewables policy framework to reduce global GHG emissions (International Hydropower Association, 2016). Many countries including China, Brazil, and the US are using hydropower to help meet their COP21 GHG reduction commitments. While discussions of GHG emissions are increasingly being incorporated into conversations around dam re-purposing (e.g., see the Glen Canyon Dam Draft Environmental Impact Statement, US DOI, 2015), little is known, especially in the US, about the value that the public-at-large places on GHG-specific reductions provided by hydropower. Existing studies, with the exception of Longo et al. (2008) in the UK context, have focused on general air pollution (e.g., non-GHG and GHG emissions) and air pollution-related human health impacts of hydropower (Mattmann et al., 2016). By contrast, this study seeks to isolate the specific value that the public places on reduced GHG emissions and associated climate change impacts of hydropower. That is, we

Due to government mandates, advocacy coalitions, and changing public opinion, policymakers are often tasked with deciding between alternative operational arrangements of existing hydroelectric dams. In the US, much of the focus of dam re-purposing has historically been on improving downstream environmental and recreation conditions (e.g., Doremus and Tarlock, 2003; Welsh et al., 1995). However, in an era of increased understanding of the couplings between human and natural systems, it is becoming increasingly difficult for policymakers to ignore the extended social impacts of dams produced through the sale and distribution of hydropower. Hydroelectric power is often associated with both positive and negative social externalities. On one hand, low-cost hydropower is associated with rural community viability, the economic livelihoods of farmers and ranchers, and provides fossil fuel offsets which reduce harmful air pollutants and greenhouse gas (GHG) emissions – see Mattmann et al. (2016). On the other hand, the hydropower production process is not totally free GHG emissions and is associated with downstream species habitat loss and lost recreational opportunities–see Jones et al. (2016). Owing to increased public discourse and mounting political pressures to include the social impacts of hydropower in official records of decision and environmental impact statements on dam operations (Jenkins-Smith et al., 2016), there is a need for additional

⁎

Correspondence to: University of New Mexico, 1915 Roma Ave. NE 1019, MSC 05 3060, Albuquerque, NM 87131, USA. E-mail address: [email protected] (B.A. Jones).

http://dx.doi.org/10.1016/j.enpol.2017.09.004 Received 15 December 2016; Received in revised form 9 August 2017; Accepted 3 September 2017 0301-4215/ © 2017 Published by Elsevier Ltd.

Energy Policy 111 (2017) 362–370

B.A. Jones et al.

purposing would have. Evidence from nationwide surveys in 2008 and 2014 found that large majorities of US residents consistently characterized hydropower as clean, safe, and renewable (Jenkins-Smith et al., 2015). In 2015, hydropower accounted for 6% of total US electricity generation (US EIA, 2016), displacing an average of 225 million MT of carbon per year (US DOE, 2016). The US Department of Energy estimates that over 2017–2050, existing hydropower capacity will reduce cumulative GHG emissions by 4.9 billion MT (US DOE, 2016). Members of the US public may hold non-market (as opposed to market) values for the GHG offsets provided by hydroelectric dams, such as GCD. Put differently, the public-at-large may be willing to pay to maintain existing GCD operations as a way to avoid anticipated increases in GHG emissions that re-purposing would create. With one exception (i.e., Longo et al., 2008), existing research on hydropower externalities has measured the value of “general air pollution impacts” that dams have, which tend to combine GHG emissions, non-GHG emissions (e.g., particulate matter), visibility, and human health outcomes tied to air pollution (e.g., Klinglmair et al., 2012; Ku and Yoo, 2010; Bergmann et al., 2008). Since there is a cause and effect relationship between air pollution, GHG emissions, and climate change, the extant literature has therefore implicitly estimated non-market values for the GHG-climate changes benefits that hydropower provides. However, and more importantly, existing estimates will also contain values associated with other air pollution-related externalities (e.g., non-GHG emissions, human health, visibility), and are therefore likely to overestimate the specific GHG-climate change externality. By contrast, this study seeks to isolate the GHG-climate change externality from other air pollution externalities. To the best of our knowledge, there is only one other study that has estimated WTP for climate change specific GHG reductions provided by a renewable energy portfolio that included hydropower, however, it was in a non-US context (Longo et al., 2008). Longo et al. (2008) find that UK residents are willing to pay £29.65 for a renewable energy policy that includes hydropower that decreases national GHG emissions by 1% a year. We might expect different values in the US because of the greater contribution that hydropower has to the overall energy portfolio and because of the contested history of US dams in the public discourse.2 The contribution of this study is to capture for the first time in the US context the non-market value of avoided GHG emissions and climate change impacts associated with proposed changes to hydropower production at a large US dam. Given that hydropower continues to play an important role in energy production in the US and elsewhere, there is a need for such information in order to help policymakers juxtapose the downstream environmental impacts of dams with the hydropower benefits held by members of the public (i.e., a green-vs-green tradeoff).

seek to separate the GHG-climate change externalities of hydropower from the general air pollution and human health externalities.1 Particularly, we use results from a nationally-representative nonmarket valuation survey of the US public on management of the Glen Canyon Dam, the largest producer of hydropower on the Colorado River, to estimate willingness to pay (WTP) for a small reduction in GHG emissions and climate change impacts brought about by maintaining existing dam operations compared to changing them as recently proposed by US government agencies. That is, we identify the intensity of support for hydropower as a reducer of GHG emissions using an actual proposal that was recently considered by the US Department of the Interior (DOI). WTP is estimated using the contingent valuation method (CVM), a widely used approach for estimating non-market, non-use values for public goods. To the best of our knowledge, this is the first such WTP estimate in the US context where hydropower has a long and contested history. This information can be used in benefit-cost analyses of re-purposing hydroelectric dam operations where, as is so often the case, there exist external social impacts that members of the public value. The remainder of the paper is structured as follows. In Section 2, background information on Glen Canyon Dam, the GHG impacts of proposed operational changes in its management, and prior literature on this topic are provided. Section 3 presents the CVM survey and design and the econometric models used to estimate WTP. Section 4 presents results on how responses to the valuation question vary across observable respondent characteristics and also provides estimates of WTP across several different specifications. Finally, Section 5 provides conclusions and policy applications. 2. Background on Glen Canyon Dam and its impacts on GHG emissions Constructed between 1956 and 1966, Glen Canyon Dam (GCD) is located on the Colorado River in northern Arizona and is a significant source of hydropower in the Western US, providing 11,599 MWh of electricity per day to the US states of Arizona, Colorado, Nebraska, Nevada, New Mexico, Utah, and Wyoming (US DOI, 2015). Beginning with the 1996 US DOI Record of Decision, GCD has increasingly been managed to improve downstream environmental conditions and recreational opportunities in the Grand Canyon, which is located less than 20 miles from the dam. This has been achieved by moderating daily fluctuations in water releases and using high intensity, short duration releases to rebuild downstream environmental habitats (US DOI, 1996). Further flow moderations that would change GCD operations were recently considered by the US DOI as part of a 2015 federally-mandated long-term adaptive management plan (US DOI, 2015). Under the 2015 DOI's preferred alternative, GCD hydropower generation is expected to decrease by 1.1%/day and marketable capacity will decrease by 6.7%, requiring an estimated 4.8% increase in system-level generating capacity additions over the next 20 years, which will largely come from gas and coal power plants (US DOI, 2015). This policy proposal is expected to increase regional GHG emissions by 22,908 metric tons (MT) per year (0.042% of total US emissions), equivalent to the annual emissions of 4874 automobiles (US DOI, 2015). The Colorado River Energy Distributors Association estimates that in 2010 hydropower from GCD offset 3 million MT of carbon (CREDA, 2010). Hence, the DOI preferred alternative would reduce by approximately 0.76% the annual GHG offset provided by GCD hydropower. Missing from discussions on operational changes to GCD is economic evidence on the GHG and climate change externalities that re-

3. Methods 3.1. Survey data and design A nationally-representative internet-based survey on management of GCD was fielded by the University of Oklahoma's Center for Energy, Security, and Society. The survey was developed in coordination with scientists and subject-matter experts from multiple organizations and agencies. Following development, the survey underwent rigorous pretesting that included an open-ended survey of stakeholders from a farm association, a species conservation group, and an electric power distribution association on the Colorado River, and cognitive interviews of participants who were unfamiliar with the subject-matter. Following

1 Naturally, air pollution and climate change externalities are related; the latter is likely a subset of the former. We do not dispute that. Rather, we try to isolate the economic value that the public places on the GHG and climate change externalities of hydropower from other externalities of air pollution (e.g., non-GHG emissions, human health, visibility).

2 As a recent example of the on-going discourse over dams, former commissioner of the US Bureau of Reclamation, Dan Beard, has published a book calling for the removal of many US dams and the abolishment of the Bureau of Reclamation, which is the Federal agency tasked with overseeing dam building and operations (Beard, 2015).

363

Energy Policy 111 (2017) 362–370

B.A. Jones et al.

Survey respondents were presented with two policy options for how GCD could be managed in the future. In the first, called Option A here, GCD operations would change according to the preferred DOI alternative in the December 2015 GCD Draft Environmental Impact Statement (DEIS). This was the most current proposed GCD management policy document available to us when the survey was designed. The second, which we call Option B here, would maintain existing dam operations (i.e., no change from how the dam is currently managed). For each policy option – change (A) or continue (B) – the impacts to the resources in the region around GCD were described as if that policy option were implemented. Resource impacts included in the survey were closely based on those described in the DEIS, as described elsewhere (Jenkins-Smith et al., 2016). To evaluate the economic benefits of GHG reductions provided by GCD, two split-sample, experimental treatments were used. The first presented information on GCD impacts that closely mirrored the DEIS: the “no GHG treatment”, hereafter. The second also presented this information and in addition included descriptions of the effects of GCD on GHG emissions and climate change: the “GHG treatment”, hereafter. For respondents randomly selected to receive the GHG treatment, they were provided with the following additional piece of information:

Table 1 Demographic characteristics of the target population and the sample. Demographic

Target population (US Adults, 18+)

Sample

Male Female Age: 18–34 Age: 35–54 Age: 55 or Older Hispanic Non-Hispanic White African American Asian Other HH Income: Less than $49,999 HH Income: $50,000 - $99,999 HH Income: $100,000−149,999 HH Income: $150,000 or more Northeast region Midwest region South region West region

49% 51% 30% 34% 36% 16% 84% 78% 13% 6% 3% 45% 30% 14% 11% 18% 21% 38% 23%

49% 51% 27% 35% 38% 14% 86% 76% 13% 5% 6% 44% 37% 12% 7% 19% 20% 38% 24%

Notes: This table reports demographic characteristics of the target population and the sample. Estimates for the target population come from the US Census (July 1, 2015), Tables PEPASR6H and S1901.

Climate Change

• Using hydropower instead of fossil fuels has resulted in lower overall

development and pre-testing, the survey was implemented by Survey Sampling International of Fairfield, CT who recruited adult US respondents to take the survey online and the final sample included 833 complete responses. Respondents were recruited to achieve national representativeness for key demographic characteristics, including age, race/ethnicity, and geographic region. They survey was fielded from May 18–21, 2016. As indicated in Table 1, the sample is generally representative of the target population. Nevertheless, post-stratification survey weights were used to adjust for slight differences between the sample and 2015 US Census estimates. Past research has demonstrated that applications of internet surveys to non-market valuation exercises produce results consistent with mail, telephone, and face-to-face interviews (Lindhjem and Navrud, 2011; Berrens et al., 2004). Table 2 provides summary statistics for respondent socioeconomic and demographic variables. Reported means and standard deviations are weighted for sample representativeness. Respondents are 45 years of age, on average, and have a college degree. 54% of the weighted sample are male and the average respondent is “middle of road” ideologically. Average household incomes in 2015 were close to $50,000. Respondents who have visited GCD in the past constitute 11% of the sample. Strong majorities of respondents view vegetation and wildlife, hydropower, and climate change (in that order) as important issues that GCD officials should consider. Finally, 56% of the weighted sample thinks that government officials operating GCD will consider the survey results when setting future policy.

CO2 (carbon dioxide) and greenhouse gas emissions that are linked to climate change.

• Scientists and government officials believe that hydropower is an im-

portant source of clean, renewable energy that can help reduce the contribution of the US to climate change.

• Impact of Option A: ○ Small increase in the amount of greenhouse gases, including CO2, emitted into the atmosphere, which could increase the negative effects of climate change.

• Impact of Option B: ○ No change in the amount of greenhouse gases, including CO2, emitted into the atmosphere, resulting in little or no impact on climate change. where Option A corresponds to changing GCD operations to moderate flows and Option B corresponds to maintaining current operations. Observe, consistent with projections by the DOI, that constraining GCD hydropower production by moderating river flows would likely result in a “small increase” in GHG emissions with subsequent effects on

Table 2 Summary statistics of respondent characteristics. Variable

Coding

Mean

Std. Dev.

Sample size

Age Education Gender Ideology Household Income (2015$) Visited GCD Importance of vegetation and wildlife when setting policy Importance of hydropower when setting policy Importance of climate change when setting policy Thinks government officials will consider survey results

continuous 1–7 scale; 1 = elementary or some HS, 7 = doctorate 0 = female, 1 = male 1–7 scale; 1 = strongly liberal, 7=strongly conservative 1 = < 50,000, 2 = 50,000–99,999, 3 = 100,000–149,999, 4 = > 150,000 0=no, 1=yes 0–10 scale; 0 = not at all important, 10 = extremely important 0–10 scale; 0 = not at all important, 10 = extremely important 0–10 scale; 0 = not at all important, 10 = extremely important 0 = no, 1 = yes

45.42 4.53 0.54 4.01 1.81 0.11 8.52 8.32 7.88 0.56

0.81 0.07 0.02 0.07 0.04 0.01 0.07 0.08 0.16 0.02

833 805 833 830 833 832 826 827 392 827

Notes: This table reports summary means and standard deviations for respondent characteristics. Importance of climate change question only asked among split-sample respondents presented with information on GHG emissions. Results are weighted to achieve sample representativeness.

364

Energy Policy 111 (2017) 362–370

B.A. Jones et al.

Table 3 Frequency distribution of willingness to pay responses by GCD operational preference and policy treatment. No GHG treatment

GHG treatment

Change

Yes No Not sure Total

Continue

Change

Continue

Frequency

Percent

Frequency

Percent

Frequency

Percent

Frequency

Percent

63 45 46 154

40.91 29.22 29.87 100.00

109 70 68 247

44.13 28.34 27.53 100.00

59 25 33 117

50.43 21.37 28.21 100.00

119 69 54 242

49.17 28.51 22.31 100.00

Notes: This table reports frequency distributions of willingness to pay responses by GCD operational preference and whether respondent was randomly presented with information on GHG effects of hydropower.

change in GHG and climate change externalities from other positive and negative externalities of hydropower. The specific dollar amount seen by an individual respondent was randomly drawn from a continuous uniform distribution ranging $1$600. A continuous bid distribution is used in light of recent research suggesting that it can lead to improved precision of WTP estimators over discrete bid distributions (Lewbel et al., 2011). Possible responses to the valuation question are “Yes”, “No”, or “Not Sure”. The specific valuation question and lead-in text are presented as follows:

climate change. Conversely, maintaining existing hydropower production would not be expected to impact GHG emissions. Immediately following this text, the following question was posed: When thinking about the information provided on this page, would you say that you…

1. Learned something new

You selected [“Option A” or “Option B”]. Both options are costly to operate and will require continued financing. The following question asks whether you, as a taxpayer, would vote for this option in an advisory referendum. The option with the most support would be recommended to government officials managing the Glen Canyon Dam. As you think about your answer, keep in mind the amount of money you and your household would pay for the policy, how much you would be able to afford to pay, and the other things you could spend the money on instead.

2. Already knew the information 3. Don’t understand the information Among respondents receiving the GHG treatment, 77% answered that they “learned something new” about hydropower and GHG emissions after reading the treatment text, while 20% indicated that they “already knew the information” presented, and 3% said that they “don’t understand the information” provided. Aside from the addition of the description of GHG emissions and the information question, the splitsamples are identical. We elicit respondents’ WTP to reduce GHG emissions using the results from two separate questions. In the first, we ask respondents to choose their preferred GCD management option (A or B) if it costs their household no additional money, but keeping in mind the described impacts of their preferred option on the region around the dam. This question separates those preferring continue from those preferring change. Immediately following this question, respondents are asked their household WTP for their preferred policy option. This WTP is a Hicksian welfare change measure of hydropower-induced GHG emissions benefits. We employ a single-bound dichotomous choice (SBDC) format framed as an advisory referendum, which has been recommended for use in evaluating the economic benefits of policy changes by the NOAA Blue Ribbon Panel on Contingent Valuation (Arrow et al., 1993). This approach to estimating WTP for the respondents’ stated preferred policy option allows us to capture both sides of the relevant policy domain (i.e., change and continue), consistent with best practices for using contingent valuation to inform the policymaking process (Carlson et al., 2016).3 Therefore, the contingent valuation scenario is a hypothetical change or continuation of GCD operations (based on an actual recent management proposal) that would impact, among other things, regional GHG emissions and climate change. The hypothetical change (continuation) would lead to a small increase (decrease) in GHG emissions. As described below, we focus only on the results of the continuation scenario since we wish to capture WTP to avoid increases in GHG. By using the previously described splitsample treatment, we can specifically isolate WTP for a small hypothetical

Would you vote for [“Option A” or “Option B”] if adoption of this option would cost your household [random uniform sample: $1: $600] in increased taxes every year for the next 20 years? Table 3 presents the frequency distribution of responses to the WTP question by GCD operational preference and treatment group. The percentage of Yes responses is higher when information on GHG emissions of hydropower are presented to respondents (50.43% vs. 40.91% for change; 49.17% vs. 44.13% for continue). Additionally, for both treatments the frequency totals indicate that intensity of support is stronger for continuing existing operations than it is for changing them, with a stronger preference for continuation in the GHG treatment (242/359 = 67.4%) compared to the no GHG treatment (247/401 = 61.6%). That is, when respondents are presented with additional information on GHG reductions provided by current management of GCD hydropower through a split-sample treatment, more of them prefer continue to change compared to when such information is not presented. Furthermore, presenting information on GHG emissions increases the percentage of Yes votes for the respondents’ preferred option, suggestive that intensity of WTP responses is influenced by knowledge of GHG and climate change effects of hydropower. Since the valuation question elicits respondents’ WTP for their preferred dam management option, responses to this question will allow for estimation of two WTP figures: (i) WTP to change GCD operations (WTPchange ), and (ii) WTP to continue existing GCD operations (WTPcontinue ). Through the use of two split-sample treatments, differing only in their presentation of information on GHG emission impacts of dam operations, WTP can be estimated separately for change and continue and separately with and without inclusion of GHG emission effects (i.e., four WTP estimators).4 Subtracting estimated WTP

3 Multiple bounded approaches (e.g., double bound) tend to enhance the statistical efficiency of estimation, but are generally not incentive compatible like the single-bound approach employed here, and hence, are not recommended for this reason (Johnston et al., 2017). We follow “best-practice” recommendations by using the single-bound CVM format (Johnston et al., 2017).

4 Respondents were asked in the survey to only consider the resource impacts of GCD operations that were described to them. Thus, respondents in the no GHG treatment should not have considered GHG emission or climate change impacts.

365

Energy Policy 111 (2017) 362–370

B.A. Jones et al.

respondents in our sample. Responses and bid amounts have been pooled across treatments for simplicity. The predicted probabilities of responding Yes, Pr(Yes), to the WTP question are indeed declining in the bid amount. At a $1 bid amount, the probability that a respondent answers Yes to the WTP question is around 85%. At the highest bid amount of $600, this probability falls to below 30%. The consistency of the downward sloping curve in Fig. 2 with economic theory provides an important validity check on our bid design allowing us to proceed in our analysis of WTP responses. We now turn to analyzing how respondent characteristics explain responses to the WTP question and how that information will be used to estimate WTP using the contingent valuation method.

including GHG emissions impacts from WTP excluding these impacts, allows isolation of the GHG only effect on WTP for change or continue. Again, this is because the only difference between treatments is the presentation of GHG-related information. However, we are not interested in estimating WTP for GHG impacts of changing dam operations (i.e., WTP for increases in GHG emissions), but in estimating WTP for the GHG impacts of continuing existing operations (i.e., WTP for GHG reductions provided by GCD).5 To this end, the penultimate result will be calculated as the difference in WTP for continue across split-sample GHG No GHG − WTPcontinue . The penultimate results can be treatments: WTPcontinue interpreted as the WTP to avoid small increases in regional GHG emissions that hypothetically changing GCD operations would produce, or equivalently, WTP to preserve current GHG reductions provided by GHG. This is because the reference point for voting respondents preferring continue is change (voting respondents had to select one option), and thus the WTP we estimate is a relative one (Carlson et al., 2016). The survey closed with the following set of subsidiary questions to ensure that respondents understood the questions posed and were motivated to provide meaningful answers:

3.2.1. The contingent valuation method The contingent valuation method (CVM) is the most widely applied approach for estimating non-market values of public goods (Carson, 2012). It is a stated preference approach that directly asks people their WTP for changes in the provisions of public goods. Despite some early controversy on the hypothetical nature of the approach, a NOAA Blue Ribbon Panel co-chaired by two Nobel Laureates in economics endorsed CVM and concluded that it provided reliable information for capturing non-market, non-use values (Arrow et al., 1993). Stated preference surveys are the only way to estimate total economic value, which includes use and non-use values. Thousands of CVM studies have been done in over 130 countries and the method is widely used to estimate the impacts of proposed policy changes, including regulatory impact assessments (Carson, 2011). CVM has been applied at least twice before to study proposed changes to GCD operations (Jones et al., 2016; Welsh et al., 1995). Estimation of WTP using CVM is based on estimating the underlying household WTP function. We follow the conventional approach, and directly estimate the WTP function following the censoring threshold model of Cameron and James (1987). WTP is assumed to be an exponential function of a linear combination of covariates and an additive idiosyncratic error term,

To conclude, we would like some feedback on this survey. Using a scale from one to seven, where one means strongly disagree and seven means strongly agree, please rate your agreement with the following statements: 1. The survey was confusing; I did not understand some of the information or questions 2. The survey was too long; it was hard to stay focused the entire time 3. There was not enough information on the survey for me to answer the questions 4. The survey was interesting; I enjoyed the information and questions

WTPi = exp {Xi′ β + εi} 5. The survey was irrelevant to me; I don’t care about the Glen Canyon Dam

(1)

where X is a vector of covariates, β is the coefficient vector, ε is a random error component with mean zero and variance σ 2 for individual i , and exp {∙} is the exponential function.6 Since Eq. (1) is not directly observable to the researcher, we used the results of latent WTP obtained from the valuation elicitation exercise, where the probability of an individual responding Yes to an offered bid (Bidi ) is equivalent to the probability of the WTP function being greater than the offered bid amount. Following Haab and McConnell p.55) (2002), the probability of a Yes response is given by the following:

Fig. 1 shows the distribution of responses to these questions. Most respondents disagreed with the first and fourth statements, indicating a high level of comprehension. Likewise, most respondents disagreed with the second and third statements, while agreeing with the fifth statement, indicating a high level of interest (motivation). This is consistent with the qualitative feedback that respondents provided at the end of the survey. Almost all the comments were positive, using words like “interesting”, “good”, and “excellent” or phrases like “very easy to follow and complete” and “I like this survey and it taught me about the dam. Everything was easy and simple to understand.” These results suggest that respondents understood the questions posed and were motivated to provide thoughtful responses. In the next section, we investigate how WTP responses vary across socioeconomic and demographic characteristics and then use this information to estimate WTP.

Pr(WTPi > Bidi ) = Pr (θi > δ ln (Bidi )−Xi′ β *)

(2)

where θ = ε / σ is the standardized error term, δ = 1/ σ is the standardized estimated coefficient on the offered bid amount, β * = β / σ is the standardized vector of covariate coefficients, and ln (∙) is the natural log function. Based on a logistic distribution of the error term, ε , it is possible to recover estimates of β* and − 1/ σ from a standard discrete choice logit model with log-likelihood function:

3.2. Regression analysis of willingness to pay responses

log L (π ) =

∑ {Wi log

(F (πi ))+(1−Wi )log (1−F (πi ))}

(3)

We begin by investigating how responses to the WTP question vary across randomly assigned bid amounts. Economic theory predicts that the bid amount should be inversely related to the probability of observing a Yes WTP response. Fig. 2 illustrates this tradeoff for

if WTPi ≥ Bidi and = 0 otherwise, and where Wi = 1 exp {δ ln(Bidi ) − Xi′ β *} F (πi ) = 1 + exp {δ ln(Bid . i ) − Xi′ β *} We focus on the typically more conservative median WTP estimator

5 As a validity check on results, we also estimated WTP for GHG impacts of changing dam operations across the split samples and found the results to be close to zero and statistically insignificant. This indicates that the public holds a $0 WTP for increasing GHG emissions, as we would expect.

6 The exponential specification will restrict WTP to be positive values. As a robustness check, we also considered a linear model that allows for both positive and negative WTP values–see pgs. 52–54 of Haab and McConnell (2002). WTP values estimated from the exponential model were conservative lower bounds on the linear specification.

366

Energy Policy 111 (2017) 362–370

B.A. Jones et al.

Fig. 1. Distribution of responses to subsidiary questions on comprehension and motivation Notes: This figure reports the distribution of responses to a set of questions on survey comprehension and respondent motivation. Response options range from 1−7, where one means strongly disagree and seven means strongly agree.

multicollinearity diagnostics for the variables in X and none of the results were particularly concerning. To the extent that we are concerned about strategic behavior influencing responses to the valuation question, we took several steps consistent with best practice recommendations of prominent CVM researchers for presenting respondents with an incentive-compatible valuation exercise that involved a plausibly consequential decision (Johnston et al., 2017). The valuation scenario was closely based on an actual US DOI policy scenario that was being considered at the time of the CVM survey and respondents were told this and that the findings from the survey would be shared with GCD managers. A single-bound dichotomous choice valuation question framed as an advisory referendum was used and a binding payment mechanism (increases in federal taxes) was employed that would apply to all respondents if the policy change were enacted. Prior research has demonstrated that these approaches, while unable to completely eliminate strategic behavior, can minimize the opportunity for such behavior (Johnston et al., 2017; Carson and Groves, 2007). Protest responses to the valuation question were also carefully analyzed since they can be related to strategic behaviors. Protest responses were identified using a follow-up debriefing question asked of respondents who selected No the reason for their response. Responses that were inconsistent with welfare estimation were labeled as protests (e.g., existing money should be used, only users should pay, government wastes too much money, etc.), following recent recommendations by prominent stated preference researchers (Johnston et al., 2017). There is currently no single set of best practices to address protest responses in estimation of WTP (Johnston et al., 2017), so we follow the most conservative approach and assume that protest responses, along with Not Sure responses, represent a WTP of zero dollars (Carson et al., 1998). This approach has been widely-used in other energy policy CVM applications (e.g., Lee and Heo, 2016; Kotchen et al., 2013). In the next section, we report results that control for potential hypothetical bias – people not treating the WTP question as a real money income tradeoff (Parsons and Myers, 2016). Potential hypothetical bias can be attenuated by using an advisory referendum format and adjusting responses to the WTP question using results from a follow-up certainty question (Little and Berrens, 2004). Immediately following the valuation question in the survey, we asked respondents on a scale of 0 (not at all certain) to 10 (completely certain) how certain they are that they would actually vote for their preferred option if it cost their household the assigned bid amount. Following previous literature, we use the responses to this question to recode uncertain Yes responses (defined as Yes responses with certainty < 10) as No responses using a recoding threshold of Yes=7 or higher (Li et al., 2009; Champ and Bishop, 2001). In what follows, certainty recoded results are reported alongside raw results.

Fig. 2. Distribution of Yes responses to willingness to pay question by bid amount Notes: This figure reports the predicted probability of responding Yes to the willingness to pay question, Pr(Yes), across randomly assigned bid amounts. Responses and bid amounts have been pooled across treatments for illustration simplicity. Bids are drawn from a continuous uniform distribution with range of $1-$600.

because it is more robust to distributional and functional form assumptions compared to mean WTP which can be very sensitive to the assumed form of the tails of the distribution (Li et al., 2009; Haab and McConnell, 2002). More importantly, median WTP is adopted because mean WTP was undefined for all but one of our regression models due to estimated σ > 1. Lastly, median WTP has a simple majority rule interpretation in the referendum context employed here (Harrison and Kristrom, 1996). Using the estimates from Eq. (3), median WTP can be calculated as,

MD (WTP ) = exp {X i′ βˆ}

(4)

where X i′ are the covariates evaluated at their means and βˆ is the estimated vector of covariate coefficients. Standard errors are constructed using the delta method (Cameron, 1991). Specific covariates included in X are age, education, gender, ideology, household income (2015$), whether respondent has visited GCD, views on the importance of considering vegetation and wildlife, hydropower, and climate change when setting GCD policy, and belief that government officials will consider the survey results when setting GCD policy. These covariates capture many socioeconomic and demographic characteristics of respondents, in addition to attitudinal beliefs on the environment, climate change, and hydropower that might influence WTP for GHG emissions reductions provided by GCD. Controlling for whether the results will be considered by GCD operators will provide an important check on the consequentiality of the survey instrument (Carson and Groves, 2007). Past experiences with GCD through physical visitation of the dam are controlled for in order to eliminate familiarity as a potential confounding influence. We additionally computed several

4. Results Table 4 reports the censored logit regression results from estimating versions of Eq. (3) by policy treatment. These results will provide some intuition on the socio-demographic drivers of responding Yes to the 367

Energy Policy 111 (2017) 362–370

B.A. Jones et al.

and traditional ways of life, which continuation will support, and in the case of the GHG treatment, acting to reduce climate change impacts, which tends to be viewed more favorably among progressives than conservatives (McCright and Dunlap, 2003). The signs (all positive) and significance (in columns 3 and 4) on household income are largely as predicted by economic theory: higher income households are more likely to respond Yes to the WTP question because they tend to have more accommodating budget constraints. Attitudinal beliefs held by respondents also play a significant role in explaining WTP responses. Viewing vegetation and wildlife as important issues to consider when setting GCD policy significantly increases the probability of responding Yes to the valuation question across both treatments in Table 4. This is somewhat of a puzzling result because continuing current operations, as described in the survey, would entail further degradations to downstream vegetation and wildlife habitats compared to the positive impact that changing operations would have on these resources. Thus, we would have expected these coefficients to be insignificant or even negative. Perhaps this can be explained by the fact that respondents viewing the downstream environment as important also tend to view climate change as an important issue that GCD operators should consider when setting policy (columns 3 and 4); the correlation coefficient (r = 0.52 ) indicates a strong positive correlation between the two variables. Though the question on climate change was only asked to those in the GHG treatment, the positive and significant signs on these coefficients suggest that this too is an important issue to those preferring continuation, and unlike vegetation and wildlife, continuation would actually result in improved GHG emissions outcomes. Clearly, respondents supporting continue view both impacts to the environment and impacts on GHG emissions as consequential, but they ultimately supported continuation, trading off fewer GHG emissions for potentially worsening riverine effects (a “green-vs-green” tradeoff), perhaps because respondents have stronger feelings about addressing climate change than they do about addressing the downstream riverine environment, if forced to decide between them. Finally, respondents believing that GCD officials will consider the results from this survey when setting policy have a greater probability of responding Yes in Table 4, except for in the GHG treatment model with recoded responses. This indicates that respondents viewing the survey as consequential are more likely to support paying some amount of money for their preferred management policy. The only null result that we feel should be mentioned is the insignificance of coefficients on the importance of hydropower covariate. Aside from the marginally significant and positive result in column 4, viewing hydropower as an important resource that should be considered when setting GCD policy does not affect WTP support for continuation. This is interesting in light of the fact that continuation largely preserves hydropower production compared to change. It seems that Yes responses are driven more by attitudes on the impacts of hydropower on climate and the environment and less influenced by attitudes on the intrinsic importance of hydropower itself. Using Eq. (4) and the results in Table 4, median WTP and associated standard errors are estimated and presented in Table 5 by policy treatment and with and without controls for potential hypothetical bias. The estimate of annual household median WTP for the next 20 years is $61.06 (no GHG treatment) and $83.06 (GHG treatment). This implies that households are willing to pay an additional $22 per year over the next 20 years for the GHG emissions reductions and climate change benefits provided by maintaining current operations of GCD (column 3). After recoding for potential hypothetical bias, we obtain more conservative WTP estimates of $19.95 (no GHG treatment) and $23.61 (GHG treatment), a difference of $3.66. That is, by our most conservative model, we estimate that US households have an annual median WTP of $3.66 to preserve the small GHG emissions reductions provided by continuing existing GCD operations compared to changing operations as preferred by the US DOI. This result is based on regression

Table 4 Censored logit regression results of willingness to pay responses for continuing existing GCD operations by policy treatment.

Age Education Gender Ideology HH income Visited GCD Import. vegetation and wildlife Import. hydropower Import. climate change Will consider survey results Constant Observations Pseudo R-squared

No GHG treatment

GHG treatment

(1) Raw responses

(2) Certainty recoded

(3) Raw responses

(4) Certainty recoded

0.016 (0.013) −0.128 (0.120) 0.052 (0.397) −0.220** (0.112) 0.133 (0.263) 1.449** (0.639) 0.203*

0.034*** (0.012) −0.079 (0.118) −0.146 (0.393) −0.070 (0.108) 0.062 (0.257) 1.436** (0.683) 0.431***

0.012 (0.012) 0.093 (0.125) 1.480*** (0.413) −0.179 (0.124) 0.564** (0.230) 1.208** (0.492) 0.294**

0.017 (0.012) 0.202* (0.120) 1.031*** (0.400) −0.223* (0.119) 0.526** (0.222) 1.495*** (0.492) 0.135

(0.116) −0.147 (0.116) –

(0.152) −0.141 (0.123) –

(0.131) −0.066 (0.170) 0.176**

(0.132) 0.271* (0.164) 0.202**

1.522***

1.634***

(0.081) 0.829**

(0.087) 0.606

(0.381) 2.754* (1.628) 232 0.2687

(0.393) −2.221 (1.461) 232 0.2804

(0.416) −3.254** (1.642) 231 0.3098

(0.391) −6.711*** (1.792) 231 0.2655

Notes: This table reports censored logit regression results for four separate models among respondents preferring to continue existing GCD operations. The dependent variable is a binary variable for response to the willingness to pay valuation question (Yes = 1; No = 0). Not Sure and protest No responses have been conservatively assumed to represent a WTP of zero. Raw responses have not been certainty recoded and certainty recoded responses have been asymmetrically recoded at the 70% threshold to account for potential hypothetical bias. The importance of climate change question only asked of respondents in the GHG treatment. Results are weighted to achieve sample representativeness. Standard errors are in parentheses. *** p < 0.01. ** p < 0.05. * p < 0.10.

valuation question. Results are reported for the subset of respondents preferring to continue existing GCD operations since we seek to estimate WTP for reductions in GHG emissions provided by GCD, consistent with continuation of current policy. The results in Table 4 indicate substantial heterogeneity in Yes responses across respondent characteristics. We will provide some intuition for each of the significant coefficients in-turn. There is weak evidence in Table 4 that older individuals are more likely to respond Yes to the valuation question than younger respondents, though the coefficient is only significant in column 2. Gender plays an insignificant role in determining WTP response among respondents in the no GHG treatment, but does significantly determine responses for those presented with information on GCD effects on GHG emissions. There, males have a significantly higher probability of responding Yes to the WTP question for continuing operations than females. These results suggest that Yes responses are higher among older males, perhaps because these individuals feel more strongly that hydropower production should not be constrained. There is suggestive evidence for both treatments in Table 4 that more ideologically liberal respondents are more likely to be willing to pay a positive amount of money to continue operations (columns 1 and 4). This may be because in the US being more liberal tends to be associated with positive feelings towards preservation of unique cultures 368

Energy Policy 111 (2017) 362–370

B.A. Jones et al.

energy policy that includes hydropower that decreases national GHG emissions by 1% a year. By contrast, extrapolating from the WTP found in this study of $3.66 (corresponding to a national GHG change of 0.042%), would suggest that US households are willing to pay $87.14/ yr. to avoid a 1% increase in GHG emissions associated with constraining GCD hydropower. The finding of a larger value in the US context may be because hydropower plays a more important role in the overall US energy portfolio than in the UK – 6% vs. 1.9% of total generating capacity – suggestive that perhaps US residents have stronger preferences for hydropower over other energy sources. This finding is also broadly consistent with nationwide surveys finding that US residents would like to see reliance on hydropower rise to 20% of the overall mix of energy sources over the next two decades (Herron et al., 2012). How might this WTP estimate be used in a policy setting? Currently, the DOI is evaluating proposed operational changes to GCD for the next 20 years. Their preferred alternative in the 2015 DEIS would produce small increases in regional GHG emissions of 22,908 MT/yr. or 0.042% of annual US emissions compared to a “no action” alternative (US DOI, 2015). The results of this study suggest that US households have a WTP of $3.66/yr. to specifically avoid this increase in GHG emissions by continuing existing operations. This does not mean that GCD operations should not be changed or that continuation of current operations provides greater economic benefits than the proposed change. Rather, we have provided empirical evidence that members of the US public hold positive non-market values for GHG reductions provided by GCD. At a minimum, this is admissible evidence that GCD operators should consider when setting future policy and it should be weighed along with the other economic benefits and costs that changing operations would have on society. A broader policy implication is that, when considering changes in the operation of existing dams, policy makers should be cognizant of the implications for societal value placed on changes in greenhouse gas emissions. Such changes consist of altering operations with implications for hydropower production (as in the case recounted here), adding to existing hydropower generating capacity (new or upgraded generators), or reducing hydropower generating capacity – including dam removal.7 These kinds of changes are consistent with those promoted in the recent US Department of Energy “vision” for the future of hydropower in the US – which chiefly rely on expanding hydropower production from the existing infrastructure (US DOE, 2016). A cost benefit analysis of these kinds of changes should include an assessment of the public WTP for the resulting net changes in GHG emissions. While ours is the first-ever WTP estimate of GHG reductions provided by an existing hydropower dam in the US, our results suggest that the US public may hold positive values for GHG reductions from hydropower more broadly. We believe that this area is ripe for rigorous empirical research, particularly in other hydropower contexts where coupled human and natural systems entail tradeoffs between impacts to downstream environments and hydropower production.

Table 5 Median willingness to pay estimates for continuation of existing GCD operations by policy treatment and certainty recoding.

Median WTP

(1) No GHG treatment

(2) GHG treatment

(3) Isolated GHG emissions (1)–(2)

Raw responses Recoded responses Observations

61.06 (14.63) 19.95 (7.15) 232

83.06 (27.88) 23.61 (10.10) 231

22.00 (2.07) 3.66 (0.81) –

Notes: This table reports median willingness to pay estimates for continuing existing GCD operations by policy treatment and certainty recoding level. Median WTP calculated as exp {X i′ βˆ} where X i′ are the covariates evaluated at their means and βˆ is the estimated vector of covariate coefficients from Table 4. Covariates are the same as those in Table 4. Raw responses have not been certainty recoded and certainty recoded responses have been asymmetrically recoded at the 70% threshold to account for potential hypothetical bias. Column (3) is the difference in willingness to pay between the GHG and no GHG treatments, which isolates the WTP for GHG emissions reductions provided by continuing GCD operations. Results are weighted to achieve sample representativeness and are weighted by voting support for continuation. Standard errors constructed using the delta method are presented in parentheses.

models that account for potential hypothetical bias and include controls for socioeconomic and attitudinal characteristics held by respondents. Furthermore, this result has economic significance because it is significantly different than zero. Aggregating the most conservative nationally-representative value ($3.66) to all US households produces an annual estimate of $493 million. Since this result is based on median WTP, it can be interpreted as the price at which a hydropower project proposal that leads to small reductions in GHG emissions and climate change impacts would be accepted by a simple majority vote of US households. Alternatively, this value is the minimum aggregate WTP that would be agreed to by US households using a simple majority voting rule. If we were to aggregate only to those households residing in the states receiving GCD hydropower (i.e., Arizona, Colorado, Nebraska, Nevada, New Mexico, Utah, and Wyoming), then the corresponding annual regional value would be $83.2 million. Both results are suggestive of significant economic value associated with GHG emissions and climate change impacts of hydropower. 5. Conclusions and energy policy implications This paper produces estimates of WTP for GHG emissions reductions provided by maintaining current operations at the largest hydropower producing dam on the Colorado River, the Glen Canyon Dam. A hypothetical GCD re-purposing scenario, based on an actual policy proposal recently considered by the US DOI – to change dam operations (resulting in small increases in regional GHG emissions) or to continue current dam operations (avoiding the GHG increases) – was presented to respondents in a split-sample nationally-representative stated preference CVM survey. Survey respondents presented with information on GHG reductions provided by GCD hydropower were more likely to support continuation over the change preferred by the DOI in the 2015 GCD DEIS – 67.4% support for continue vs. 32.6% for change. That is, a majority of respondents prefer to not change GCD operations when presented with information on both the downstream and extended regional impacts that change would have on the environment, climate change, hydropower production, and ways of life tied to hydropower. Isolating just the economic benefit provided by GHG and climate change reductions, we estimate annual US household median WTP over 20 years of $3.66. Our results are slightly larger than those obtained from a comparable study by Longo et al. (2008) in the UK, where it was found that residents were willing to pay £29.65/yr. ($58.38/yr.) for a renewable

Funding sources The Office of the Vice President for Research at the University of Oklahoma funded the design and implementation of the contingent valuation survey. References Arrow, K., Solow, R., Portney, P.R., Leamer, E.E., Radner, R., Schuman, H., 1993. Report of the NOAA panel on contingent valuation. Fed. Regist. 58 (10), 4601–4614.

7 Note that while this paper dealt with GHG impacts of re-purposing an existing dam, the construction of new dams and reservoirs raises additional difficult issues, including the possibility of environment disamenity, forced migration, and that the new dam and reservoir lead to net increases in GHG emissions (Fearnside, 2006, 2001).

369

Energy Policy 111 (2017) 362–370

B.A. Jones et al.

Oklahoma. Johnston, R.J., Boyle, K.J., Adamowicz, W., Bennett, J., Brouwer, R., Cameron, T.A., Tourangeau, R., 2017. Contemporary guidance for stated preference studies. J. Assoc. Environ. Resour. Econ. 4 (2), 319–405. Jones, B.A., Berrens, R.P., Jenkins-Smith, H.C., Silva, C.L., Carlson, D.E., Ripberger, J.T., Carlson, N., 2016. Valuation in the Anthropocene: exploring options for alternative operations of the Glen Canyon Dam. Water Resour. Econ. 14, 13–30. Klinglmair, A., Bliem, M., Brouwer, R., 2012. Public preferences for urban and rural hydropower projects in Styria using a choice experiment. IHS Kärnten Working Paper. Accessed from: 〈http://www.carinthia.ihs.ac.at/HydroVal/files/working_ paper.pdf〉. Kotchen, M.J., Boyle, K.J., Leiserowitz, A.A., 2013. Willingness-to-pay and policy-instrument choice for climate-change policy in the United States. Energy Policy 55, 617–625. Ku, S.J., Yoo, S.H., 2010. Willingness to pay for renewable energy investment in Korea: a choice experiment study. Renew. Sustain. Energy Rev. 14 (8), 2196–2201. Lee, C.Y., Heo, H., 2016. Estimating willingness to pay for renewable energy in South Korea using the contingent valuation method. Energy Policy 94, 150–156. Lewbel, A., McFadden, D., Linton, O., 2011. Estimating features of a distribution from binomial data. J. Econ. 162 (2), 170–188. Li, H., Jenkins-Smith, H.C., Silva, C.L., Berrens, R.P., Herron, K.G., 2009. Public support for reducing US reliance on fossil fuels: investigating household willingness-to-pay for energy research and development. Ecol. Econ. 68 (3), 731–742. Lindhjem, H., Navrud, S., 2011. Are Internet surveys an alternative to face-to-face interviews in contingent valuation? Ecol. Econ. 70 (9), 1628–1637. Little, J., Berrens, R., 2004. Explaining disparities between actual and hypothetical stated values: further investigation using meta-analysis. Econ. Bull. 3 (6), 1–13. Longo, A., Markandya, A., Petrucci, M., 2008. The internalization of externalities in the production of electricity: willingness to pay for the attributes of a policy for renewable energy. Ecol. Econ. 67 (1), 140–152. Mattmann, M., Logar, I., Brouwer, R., 2016. Hydropower externalities: a meta-analysis. Energy Econ. 57, 66–77. McCright, A.M., Dunlap, R.E., 2003. Defeating Kyoto: the conservative movement's impact on US climate change policy. Social. Probl. 50 (3), 348–373. Parsons, G.R., Myers, K., 2016. Fat tails and truncated bids in contingent valuation: an application to an endangered shorebird species. Ecol. Econ. 129, 210–219. US DOE, 2016. Hydropower Vision: a New Chapter for America's 1st Renewable Electricity Source. US Department of Energy, Office of Scientific and Technical Information, Oak Ridge, TN (Accessed from). 〈https://energy.gov/sites/prod/files/ 2016/10/f33/Hydropower-Vision-10262016_0.pdf〉. US DOI, 2015. Glen Canyon Dam Draft Environmental Impact Statement. US Department of the Interior Bureau of Reclamation, Upper Colorado Region (Accessed from). 〈http://ltempeis.anl.gov/documents/draft-eis/〉. US DOI, 1996. Record of Decision, Operation of Glen Canyon Dam, Final Environmental Impact Statement, Appendix G. US Department of the Interior Bureau of Reclamation (Access from). 〈https://www.usbr.gov/uc/rm/amp/pdfs/sp_appndxG_ROD.pdf〉. US EIA, 2016. Hydropower Explained. Accessed from: 〈https://www.eia.gov/ energyexplained/index.cfm/data/index.cfm?Page=hydropower_home〉. Welsh, M., Bishop, R., Phillps, M., Baumgartner, R., 1995. GCES Non–Use Value Study: GCES Non–Use Values Final Study Report. Hagler Bailly Consulting, University Park, Madison WI.

Beard, Dan, 2015. Deadbeat Dams: Why We Should Abolish the US Bureau of Reclamation and Tear Down Glen Canyon Dam. Johnson Books, Boulder, Colorado. Bergmann, A., Colombo, S., Hanley, N., 2008. Rural versus urban preferences for renewable energy developments. Ecol. Econ. 65 (3), 616–625. Berrens, R.P., Bohara, A.K., Jenkins-Smith, H.C., Silva, C.L., Weimer, D.L., 2004. Information and effort in contingent valuation surveys: application to global climate change using national internet samples. J. Environ. Econ. Manag. 47 (2), 331–363. Cameron, T.A., James, M.D., 1987. Efficient estimation methods for "closed-ended" contingent valuation surveys. Rev. Econ. Stat. 69 (2), 269–276. Cameron, T.A., 1991. Interval estimates of non-market resource values from referendum contingent valuation surveys. Land Econ. 67 (4), 413–421. Carlson, D.E., Ripberger, J.T., Jenkins-Smith, H.C., Silva, C.L., Gupta, K., Berrens, R.P., Jones, B.A., 2016. Contingent valuation and the policymaking process: an application to used nuclear fuel in the United States. J. Benefit-Cost. Anal. 7 (3), 459–487. Carson, R.T., 2012. Contingent valuation: a practical alternative when prices aren't available. J. Econ. Perspect. 26 (4), 27–42. Carson, R.T., 2011. Contingent Valuation: A Comprehensive Bibliography and History. Edward Elgar, Northampton, MA. Carson, R.T., Groves, T., 2007. Incentive and informational properties of preference questions. Environ. Resour. Econ. 37 (1), 181–210. Carson, R.T., Hanemann, W.M., Kopp, R.J., Krosnick, J.A., Mitchell, R.C., Presser, S., Martin, K., 1998. Referendum design and contingent valuation: the NOAA panel's novote recommendation. Rev. Econ. Stat. 80 (2), 335–338. Champ, P.A., Bishop, R.C., 2001. Donation payment mechanisms and contingent valuation: an empirical study of hypothetical bias. Environ. Resour. Econ. 19 (4), 383–402. Colorado River Energy Distributors Association [CREDA], 2010. Hydropower and Glen Canyon Dam. Accessed from: 〈http://www.creda.org/Documents/Messaging %20Final%20100510.pdf〉. Doremus, H., Tarlock, A.D., 2003. Fish, farms, and the clash of cultures in the Klamath basin. Ecol. LQ 30, 279. Fearnside, P.M., 2001. Environmental impacts of Brazil's Tucuruí Dam: unlearned lessons for hydroelectric development in Amazonia. Environ. Manag. 27 (3), 377–396. Fearnside, P.M., 2006. Dams in the Amazon: Belo Monte and Brazil's hydroelectric development of the Xingu River Basin. Environ. Manag. 38 (1), 16. Haab, T.C., McConnell, K.E., 2002. Valuing Environmental and Natural Resources: The Econometrics of Non-Market Valuation. Edward Elgar, Northampton, MA. Harrison, G.W., Kristrom, B., 1996. In: Johansson, P.O., Kriström, B., Mäier, K.G. (Eds.), On the Interpretation of Responses to Contingent Valuation Surveys in Current Issues in Environmental Economics. Manchester University Press, Manchester. Herron, K., Jenkins-Smith, H., Silva, C., 2012. US Public Perspectives on Security (Sandia Report: SAND2012-0165). Sandia National Laboratories, Albuquerque, NM. International Hydropower Association, 2016. What will the Paris Agreement mean for hydropower development? Accessed from: 〈https://www.hydropower.org/blog/ what-will-the-paris-agreement-mean-for-hydropower-development〉. Jenkins-Smith, H.C., Silva, C.L., Carlson, D., Gupta, K., Jones, B.A., Ripberger, J., Wehde, W., 2016. Estimating Non-Use Values for Alternative Operations of the Glen Canyon Dam: An Inclusive Value Approach, Phase 3B Project Research and Findings. University of Oklahoma, Center for Energy, Security, and Society. Jenkins-Smith, H., Silva, C., Carlson, D., Ripberger, J., Gupta, K., Carlson, N., 2015. NonMarket Values for Alternative Operations of the Glen Canyon Dam: Explorations in Choice and Valuations. Center for Energy, Security, and Society, University of

370