Overlooked Benefits of Nutrient Reductions in the Mississippi River Basin S Bryan Parthum National Center for Environmental Economics, U.S

Overlooked Benefits of Nutrient Reductions in the Mississippi River Basin S Bryan Parthum National Center for Environmental Economics, U.S. Environmental Protection Agency, Washington D.C.; [email protected] Amy W. Ando Professor, Department of Agricultural and Consumer Economics, University of Illinois Urbana-Champaign; and university fellow, Resources for the Future, Washington, DC; [email protected]

ABSTRACT Improvements in local surface wa- runoff into local waters, but a major policy ter quality in the Mississippi River Basin (MRB) change such as state subsidies will be needed can contribute to the regional environmental to accomplish the 2040 goals (Coppess 2016). goals of reducing hypoxia in the Gulf of Mexico. State agencies and lawmakers are, therefore, To inform estimates of the benefits of water qual- interested in how much their own residents ity policy, we use a choice experiment survey in would support efforts to meet the INLRS a typical subwatershed of the MRB to estimate targets. How much value do residents of the willingness to pay for local environmental im- MRB gain from changes to water quality in provements and helping to reduce hypoxia far their local watersheds, and to what extent do downstream. We find that residents place large people in a state like Illinois value their local values on reduced local algal blooms, improved watershed’s contribution to nonlocal improve- local fish populations and diversity, and meeting ments such as reducing the scale of the hy- local commitments to help with the regional en- poxic dead zone in the Gulf of Mexico? vironmental problem. (JEL Q52, Q53) Integrated assessment of surface water quality policies and management actions can benefit from information about the total val- 1. Introduction ues of changes in water quality and the distribution of those values among different groups Nutrient pollution and hydrological disrup- of people. A host of previous studies have tion cause water quality impairments through- shed light on the values people place on some out the Mississippi River Basin (MRB) and dimensions of pollution reduction within the serious problems with widespread oxygen United States. That work is surveyed by Berg- depletion called hypoxia in the Gulf of Mex- strom and Loomis (2017); meta-analyses of ico (U.S. Environmental Protection Agency those studies have informed benefit transfer 2008). The U.S. Environmental Protection efforts to estimate aggregate benefits of water Agency’s 2008 Gulf Hypoxia Action Plan quality changes at the national level (John- (GHP) tasked the 12 upstream states with ston, Besedin, and Stapler 2017; Moeltner the responsibility of reducing their transmis- 2019). There is also a long line of research ex- sion of nutrients such as nitrate-nitrogen and ploring the differences between use and non- phosphorus by 45% by the year 2040. In an use values from local and nonlocal improve- approach similar to the other states in the ments in surface water quality in the United MRB, agencies in Illinois created the Illinois States (Greenley, Walsh, and Young 1981; Nutrient Loss Reduction Strategy (INLRS) to Lant and Roberts 1990; Carson and Mitchell coordinate efforts in that state to meet the nu- 1993; Johnston, Besedin, and Wardwell 2003; trient reduction targets. The INLRS promotes van Houtven, Powers, and Pattanayak 2007), voluntary efforts by farmers to reduce nutrient and recent work has emphasized the need to examine these relationships when consid- Land Economics • November 2020 • 96 (4): 589–607 ering the benefits from policies that reduce ISSN 0023-7639; E-ISSN 1543-8325 hypoxia in the Gulf of Mexico (Babcock and © 2020 by the Board of Regents of the Kling 2015; Keiser, Kling, and Shapiro 2019). University of Wisconsin System This paper advances research on water qual-

589 S Appendix materials are freely available at http://le.uwpress.org and via the links in the electronic version of this article. doi:10.3368/wple.96.4.589 This open access article is distributed under the terms of the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0) and is freely available online at: http://le.uwpress.org 590 Land Economics November 2020 ity valuation and integrated assessment with a Previous research in stated preference val- choice experiment survey that estimates three uation shows that spatial dimensions matter conventional benefits of water quality im- in other important ways. First, willingness provements (improvements in local fish pop- to pay (WTP) for an environmental improve- ulations, fish diversity, and reductions in local ment can vary widely across space (Johnston algal blooms) and previously overlooked ben- and Duke 2007; Brouwer, Martin-Ortega, efits (local contributions to reaching a regional and Berbel 2010). In particular, people often nutrient reduction target) that arise from poli- have higher WTP when they are closer to the cies targeting hypoxia in the Gulf of Mexico.1 improvement (Sutherland and Walsh 1985; We then illustrate how to use those values in a Hanley, Schläpfer, and Spurgeon 2003; Cza- spatially disaggregated integrated assessment jkowski, Budziński, et al. 2017; Glenk et al. of a land use management plan and explore 2019). Second, researchers have found that two dimensions of value heterogeneity. when estimating WTP for an environmen- The bottom line of a benefit-cost analysis tal change that has a specific location within often aggregates benefits of environmental the landscape, the quality of responses from improvements to all people affected by the stated preference surveys depends on how policy. However, many policy makers and in- clearly the survey describes the location of the terest groups are particularly concerned about change relative to the respondent (Schaafsma the net impact of agricultural-environmen- and Brouwer 2013; Johnston, Holland, and tal policies on rural residents (Gibbs 2016; Yao 2016). Our survey shows respondents Farber 2018), although evidence regarding exactly where they live relative to the pro- preference heterogeneity between rural and posed improvements. We also vary distance urban areas is mixed. Some research shows from the improvement experimentally across that urban residents give more support for alternatives to identify how WTP varies with environmental policies than people in rural exogenously determined distance from the en- areas of the United States (Salka 2001), and vironmental good. other research finds little difference between We find that people place positive and rural and urban residents in their interests for significant values on local water quality im- environmental quality (Arcury and Christian- provements and on helping to achieve basin- son 1993; Mobley 2016). Racevski and Lupi wide success in reducing hypoxia in the Gulf (2006) find rural residents in Michigan are of Mexico. We do not find evidence of joint less likely to support forest management ef- differences in preferences between rural and forts involving conservation, but conclude this urban residents in the same watershed. We do, is likely because those rural communities rely however, find that rural residents and people on forests products for production or exports. who are familiar with nutrient pollution prob- Melstrom et al. (2015) find that urban rivers lems place more value on moving away from and streams are less valued than rural rivers the status quo conditions in the watershed re- for recreational fishing, but do not estimate gardless of the improvements a program pro- the differences in preferences between rural duces. Finally, we demonstrate how these es- and urban recreationists themselves. We make timates can be used in spatially disaggregated a contribution to this discussion in the context integrated assessments, where benefit totals of surface water quality by testing whether and distributions depend on spatial details the values that people place on water quality of the improvements and the population that improvements vary between people in rural stands to gain. areas and people in urban areas of the same watershed, located in the heart of the MRB. 2. Application 1 Phaneuf (2002) estimates use values within a watershed for achieving total maximum daily load targets (nutrient Freshwater systems throughout the U.S. Mid- reductions). We extend this analysis to estimate the local (within the watershed) benefits of contributing to regional, west have been severely altered due to decades downstream (outside the watershed) nutrient reduction tar- of intensive agriculture production (Manifold gets. 1998; Alexander et al. 2008). Tributaries lo- 96(4) Parthum and Ando: Benefits of Nutrient Reductions 591 cated within the upper MRB carry excess Figure 1 nutrients, byproducts of intensive agriculture Study Area in Upper Sangamon River Basin, production, to the Mississippi River, where Central Illinois: The Four Sections of River (A, B, C, D) Are Highlighted and Included they are eventually released into the northern as Attributes on the Choice Card Gulf of Mexico. An overabundance of these nutrients contributes to the large seasonal hypoxic dead zone off the cost of Louisiana and Texas (Diaz and Rosenberg 2008; Rabalais et al. 2010; Rabotyagov et al. 2014). This paper presents survey results from people in the Upper Sangamon River Water- shed (USRW) in central Illinois (Figure 1). This watershed is listed as a priority watershed due to its high levels of nitrate-nitrogen and phosphorus transmission within the MRB (U.S. Environmental Protection Agency 2008, 2013). The population in the study area is diverse and includes large swaths of rural landscape with several urban clusters. The characteristics of the USRW are representative of many watersheds in the MRB, providing an excellent setting for examining value differ- entials and policy-induced distributional effects across rural and urban populations in the MRB. State agencies, university extension per- sonnel, researchers at the University of Illi- nois, and people from groups like the Illinois Farm Bureau have been active in communicat- and streams. Using this platform allows us ing about the INLRS in the state, explaining to model preferences in the random utility the goals of the INLRS and how agricultural maximization (RUM) framework (McFadden practices such as cover crops, reduced tillage, 1973). Preferences are characterized by esti- and riparian buffers can reduce nutrient load- mating the probability a respondent chooses a ings. It has been shown that stated preferences scenario from a set of alternatives with vary- for environmental goods, and the underlying ing levels of environmental quality (Hanley, latent consequentiality of a survey, are more Wright, and Adamowicz 1998). reliable when respondents have prior knowl- Each respondent began the survey by read- edge of the issues and policies that are being ing a consent form describing the purpose discussed (Whitehead et al. 1995; LaRiviere and nature of the survey and gave consent to et al. 2014; Czajkowski, Hanley, and LaRiv- continue with the survey. Respondents were iere 2015; Czajkowski, Vossler, et al. 2017; then presented a background section that pro- Needham and Hanley 2020)—as is the case vided basic information about nutrient pol- in our survey. lution problems in the MRB and the general nature of the improvements to be evaluated in the survey. After the respondent read the 3. Choice Experiment background section, they answered six choice Methodology questions and supplemental questions about personal characteristics. Choice experiment surveys are widely used to We held a series of focus groups through- elicit preference for nonmarket environmen- out the watershed with attendees from the tal amenities such as water quality in rivers general population. They were asked to take 592 Land Economics November 2020

Table 1 Survey Attributes and Levels

Attribute Levelsa Description Fish species (1), 2, 3, 5 Number of different recreational game fish species per 100 yards of river Fish population (15), 30, 45, 150 Number of all fish per 100 yards of river Algal blooms (%) (0), 25, 50, 75 Percent reduction in the frequency of local algal blooms Nutrient target (%) (0), 50, 75, 100 Likelihood that nutrient run-off from this watershed is reduced by the target of “45% by 2040” Location A, B, C, D The section of river where the improvements will be received Distance (Varies) The distance in miles from the respondent to the nearest point on the location attribute; this depends on where the respondent lives and which location is represented in the scenario Annual cost (0), 5, 15, 30, 60 Payment vehicle: annual county fee (e.g., property tax) Note: All attributes listed except for distance were included in the experiment design. a Status quo levels for each attribute are presented in parentheses. the survey and participate in a 30-minute fol- in local agriculture, such as expanded cover low-up discussion. In response to focus group crops, reduced tillage, and riparian buffers; feedback, we revised the survey to incorpo- these voluntary and subsidized practices can rate their suggestions regarding ambiguities reduce sediment and nutrient runoff from the in management mechanisms and wording of surrounding area and are currently well ac- the attribute changes. We deployed the survey cepted and widely used by farmers throughout in a pretest with 79 completed surveys (474 the region. The survey scenarios with water observations) and adjusted the levels of the quality improvements from such changes in cost attribute so that all levels were chosen agricultural practices are within the range of with some frequency. Finally, we distributed future actions actually being discussed in the the survey to a randomly selected group of re- state, and thus not entirely hypothetical. In the spondents living within the watershed. survey background we explicitly state that an environmental change “will NOT result in a Consent and Background change in agricultural acreage or profits,” to further prevent concern about the profitability Several features of the survey were designed of local agriculture from being confounded to increase respondent belief in consequential- with the value people would gain from envi- ity and prevent concern about agricultural reg- ronmental improvements.2 ulation that might trigger protest responses. The consent form explained that “information from this survey will help policy makers, Choice Questions economists, and watershed managers choose A choice question is posed in a “card” that how and how much to improve water quality shows a set of scenarios and asks respondents in your area.” The University of Illinois is re- to choose the scenario they like most. In our gionally known to be connected to state pol- survey, each scenario in a choice card has icy makers and agricultural decision-makers, seven experimentally varied attributes. Four supporting the claim that the survey will be of those attributes relate to biophysical char- consequential. acteristics of water quality, two capture spatial The background section of the survey heterogeneity, and one is the payment nec- tells respondents about the regional goal for essary to implement the proposed improve- nutrient loss reduction to reduce the size of ments. Table 1 summarizes each attribute, the hypoxic zone and the nutrient pollution specifying the status quo and improved levels reduction target for the Upper Sangamon of each attribute. Our choice experiment sur- River watershed’s contribution to that goal. vey is tightly coupled to biophysical models This section explains that the proposed environmental changes would come from changes 2 The full survey text can be found in Appendix A. 96(4) Parthum and Ando: Benefits of Nutrient Reductions 593 of watershed improvements; the levels of the Figure 2 biophysical attributes were informed by the Frequency of Chosen Attribute Levels work of hydrological and ecological modelers in the USRW. Botero-Acosta, Chu, and Huang (2019) model predicted changes in nutrient levels throughout the USRW resulting from hypothetical changes in local agricultural practices. Andres, Chien, and Knouft (2019) use these predicted changes in nutrient levels, climate, and data from 110 monitoring sites across the USRW to model changes in aquatic biodiversity. Three of the four biophysical attributes related to water quality are local, and one is nonlocal. Number of fish species and population of fish (two independent attributes) are local quantitative measures summarizing the current average number of distinct species of fish (diversity) and populations of individual fish per 100 linear yards of river (density). Dissanayake and Ando (2014) find that Il- linois residents have positive value for both species diversity and faunal density in grassland birds; we test whether people value two such attributes of fish in inland streams. Im- provements in local surface water quality are also captured as percent reductions in the frequency of algal blooms in the local watershed vehicle, verifying with focus groups that this including streams and ponds; that reduction is a salient and credibly binding mechanism ranges from 0% to 75%. The fourth nutri- for payment. The survey states that the fee ent-pollution attribute describes the likelihood will be passed on to renters through an an- that this watershed succeeds in meeting its nual increase in rent charged by the landlord. targets for reductions in the level of nutrient Figure 2 shows that all attribute levels were transmission to the Gulf of Mexico and ranges chosen with some frequency by respondents. from 0% (definitely will not succeed) to 100% We designed the survey to increase estima- (certain to succeed). tion efficiency while maintaining reliability in Local water quality–related changes from WTP estimates. In theory, choice experiments a nutrient-loss reduction strategy are not uni- are demand revealing only if they are incentive form throughout a watershed, but rather de- compatible (Carson and Groves 2007), and pend on local details such as depth, flow rate, while a dichotomous choice design (one sta- and shade. We partition the watershed into four tus quo and one alternative) is often argued to equally sized sections. Each scenario specifies be incentive compatible, trichotomous choice the section of the watershed in which water (one status quo and two alternatives) is not. quality attributes improved. The location at- However, trichotomous choice increases the tribute varies as part of the experiment design; amount of information recovered from each as a result, distance (measured as the distance survey response, and some research shows from each respondent to the improved section that values are similar between the two mech- of the watershed) also varies experimentally. anisms (Collins and Vossler 2009). Thus, we The final attribute in each scenario is the include two alternatives along with the status household payment necessary to achieve the quo on every choice card. proposed improvements, cost. We use an in- In stated preference research, hypothet- crease in annual county fees as the payment ical bias can influence estimates of WTP 594 Land Economics November 2020

(Cummings, Harrison, and Rutström 1995; were not met. The first condition is a no-free- Cummings and Taylor 1999). We include lunch restriction (improvement in any attri- a modified cheap talk script in the informa- bute will come at a nonzero cost) and a wel- tion section of the survey and an opt-out re- fare improving restriction (no improvement minder on each choice card (Ladenburg and across all attributes cannot come at a cost). Olsen 2014) to mitigate such bias.3 After each The second condition checks if any of the 18 choice card, we also include certainty ques- resulting choice cards has an alternative that tions asking how sure the respondents were of was strictly dominated by another alternative the selection they just made (Ready, Champ, on the same card (e.g., a higher level of im- and Lawton 2010; Penn and Hu 2020).4 provements at a lower cost). After seven iterations of the two-step procedure—each iter- ation consisting of many design iterations in Experimental Design the first step—all conditions were met.5 We develop an optimal orthogonal choice ma- With the exception of location and distance, trix resulting in a D-efficient experiment de- we allow the status quo level of each attribute sign (Adamowicz, Louviere, and Swait 1998; to be randomly included in the improved (non– Hensher, Rose, and Greene 2005; Street and status quo) scenarios. We include an alterna- Burgess 2007; Ferrini and Scarpa 2007). As tive specific constant (ASC) to represent the recommended by Ferrini and Scarpa (2007), status quo alternative on the choice card. The the design is optimized for main effects with ASC captures preferences that the respondent zero priors (β = 0) to produce a reliable de- may have for maintaining the status quo that sign when the true underlying data generating are unobservable and not otherwise contained process is unknown and prior information on in our experimental design. parameter values is not available. We produce 18 unique choice cards from the full factorial Individualized Maps and Choice Card design, divided into three blocks of six choice Generation cards. Respondents are randomly assigned one of the three blocks of six choice cards. Following recommendations highlighted by The number of cards and alternatives are cho- Johnston, Holland, and Yao (2016), each al- sen to limit cognitive burden for the respon- ternative on a choice card includes an individ- dents while maintaining statistical power to ually geocoded map highlighting the section estimate WTP (Swait and Adamowicz 2001; of river that would experience the improve- Caussade et al. 2005). ments and a marker locating the respondent within the watershed relative to the proposed After we created an initial design, we im- improvements. Each map was created for the posed two additional conditions for the final individual respondent and geocoded using design, and reran the design if the conditions ArcPy integration in ArcGIS. Eight towns and 3 The cheap talk script included in the information sec- city centers distributed throughout the wa- tion reads: “Experience from previous similar surveys is that tershed are geolocated to provide a “you are people often say they would be willing to pay more money here” marker in each map. The total number for something than they actually would. For example, in of combinations of choice cards, alternatives, one study, 80% of people said they would buy a product, and geolocations resulted in 432 different in- but when a store actually stocked the product, only 43% of people actually bought the new product. It is important that dividualized maps and 432 different levels for you make each of your upcoming selections like you would the distance attribute listed as an attribute on if you were actually facing these exact choices in reality. the choice card. Note that paying for environmental improvement means you In order to accommodate the individualiza- would have less money available for other purchases.” 4 The question asks: “How confident are you in your an- tion of alternatives and choice cards, we cre- swer?” With the range: “0 - not at all confident”; “1 - some- what confident”; and “2 - very confident.” We use these 5 We generated the design using the dcreate package im- responses to recode uncertain responses to the status quo plemented in Stata (Hole 2015). We created a wrapper for alternative. Results and discussion of these regressions are the dcreate package that allows us to impose the additional available in Appendix C. restrictions on the design. 96(4) Parthum and Ando: Benefits of Nutrient Reductions 595 ated images of the choice cards by integrat- Survey Administration ing the mail-merge capabilities of Microsoft The survey was administered online using a Publisher, referencing an underlying matrix of Qualtrics panel of respondents through their all individualized combinations of the experi- survey interface, paired with additional JavaS- ment design. The resulting pages of the docu- cript and HTML to incorporate the individu- ment were then extracted by the survey proto- alized choice cards.6 Respondents were re- col using Python to create an image for each cruited from the 42 zip codes contained within page representing a choice card in the experi- the watershed. Once respondents received an ment. The 432 choice cards images were then invitation to take the survey, they would arrive stored online using Amazon Web Services and at the online interface where they were asked referenced in real time while the respondent to enter their zip code. If the zip code was not was taking the survey. one of the 42 qualifying, they were screened and exited from the survey. The next step in- Other Survey Questions dividualizing the choice experiment was to ask respondents which of the eight locations We designed the survey instrument to test for (towns or city centers) they lived closest to. potential preference heterogeneity between Their response would then cue the system to residents who identify as rural, and those who load a randomly ordered set of choice cards. identify as urban. That characteristic was ex- Our final sample has complete responses from amined in two dimensions: (1) geographical 343 individuals. affiliation and (2) cultural affiliation. The first, geographical affiliation, is simply determined using the U.S. Census Bureau’s classification 4. Econometric Framework of rural—a census block group area with less than 1,000 residents per square mile (Ratcliffe Following choice experiment methodology et al. 2016). Respondents who fit this designa- (Hanley, Wright, and Adamowicz 1998; John- tion are classified as living in a geographically ston et al. 2017), we assume that a respondent rural area, all others are classified as living in derives utility based on the observable char- an urban area. The second, cultural affiliation, acteristics contained within the choice card, is determined by the respondent’s stated af- and some characteristics unobservable to the filiation in the postsurvey questionnaire. The researcher. Specifically, U is the utility re- question was phrased as: “Do you consider spondent i derives by choosing alternative j where you live to be rural?” Respondents in on choice card t: our sample overwhelmingly responded with a cultural affiliation that aligned with their geo- Uijt=−++αβ i p jt i′ xe jt ijt. [1] graphical affiliation. Our design allows us to test the hypothesis that preferences for surface where x is a vector of attributes, p is the price water quality are the same between those who (cost) of the scenario, and e is the stochastic live in a geographically and culturally rural component capturing unobservable character- area and those who live in a geographically istics influencing the respondent’s choice and and culturally urban area. is an independent and identically distributed In order to understand other character- (IID) extreme value. Included in x is an ASC istics of our survey sample, we ask two sets that is equal to 1 for the status quo alternative of personal questions. Three questions come in each choice set, and 0 otherwise. β is the before the choice questions and ask about the vector of preference coefficients, andα is the frequency with which people had seen algal coefficient on cost. Both β and α are indexed blooms, how often respondents visited the to be respondent specific when estimated us- river to go fishing, and how often they recre- 6 The first wave of survey responses was collected from ated nearby the river. A section after the choice January 2019 through February 2019. A second collection questions contains common demographic and period was administered January 2020 through February socioeconomic questions. 2020. 596 Land Economics November 2020 ing a random parameter logit model (Train a full covariance matrix and corresponding 1998). correlation coefficients for the random pa- The variance of the error term also rameters in the model. We follow Thiene and varies with each respondent such that: Scarpa (2009) and estimate the model using 22 Var( eijt )= ki ()π /6 , where ki is the scale pa- maximum simulated likelihood. Halton draws rameter for respondent i. Variation in the error were used in the maximum-likelihood simu- term can be attributable to scale heterogene- lation. The first N prime numbers were used ity or other forms of correlation between the to generate the draws, where N is equal to the model attributes, particularly so in panel (re- number of random parameters in our model.8 peated choice occasion) settings such as ours To develop estimates of total WTP and its (Swait and Louviere 1993; Train and Weeks distribution throughout the watershed for hy- 2005; Hess and Train 2017). Dividing the pothetical improvements in water quality, we preference parameters by the scale parameter allow for location-specific and individual-spe- where λαi= ( ii /)k and cki= (β ii /) results in cific heterogeneity in estimates of marginal a specification that has the same variance for WTP (MWTP) by recovering the conditional all respondents: individual specific means of the parameters in equation [3] (Greene, Hensher, and Rose Uijt=−++ λ i p jt cx i′ jt ijt. [2] 2005; Meyerhoff, Boeri, and Hartje 2014). This is discussed in more detail in Section 6, where ijt is an IID type-one extreme value, where we discuss the integrated assessment now with a constant variance: π 2 / 6. With model exercise. ki in the denominator of each coefficient, al- lowing the coefficients to be independent (not correlated) would constrain the scale param- 5. Results eter to be constant for the sample while al- lowing the preference parameters to vary, or Our sample is evenly divided between people vice versa (Louviere et al. 2002). Equation in rural (53%) and urban (47%) areas, and [2] is the model in preference space (Train 56% of the sample’s respondents own homes and Weeks 2005). To avoid the postestima- instead of renting.9 Respondents are predom- tion difficulties in deriving empirical distri- inantly white (78%) and female (68%); the butions of WTP (Train 1998; Daly, Hess, and former is consistent with the actual demo- Train 2012; Carson and Czajkowski 2019), graphics of the area. Our sample has broad we choose to estimate our model in willing- representation of age, income, and education ness to pay space (WTP-space) directly (Train categories, and the distributions in our sam- and Weeks 2005; Scarpa, Thiene, and Train ple are similar to those of the U.S. Census 2008).7 This is a standard reparameterization demographics for this area. This is an area of equation [2] such that wtpi= c ii/ λ ; utility with little in-migration; half the people in our is then represented by sample have lived in the area for more than 30 years, and only 10% have lived there for Uijt=−+ λλ i p jt() i wtp i′ x jt + ijt . [3] 10 years or fewer. The two subsamples are mostly similar, except that urban respondents Equation [3] is the specification in WTP-space are more likely to hold a graduate degree and (Train and Weeks 2005). We specify the wtp less likely to participate in recreational fish- parameters to be distributed normally, and the ing and hiking.10 coefficient on cost, λi, is distributed log-normally as outlined by Train and Weeks (2005). 8 All specifications and analyses are modeled using the We specify the distributions of the random gmnl package in R (Sarrias and Daziano 2017). parameters to be fully correlated, estimating 9 A full summary of responses to demographic questions and comparison to the American Community Survey can be found in Appendix B. 7 We also estimate our models using conventional prefer- 10 The results of testing for observable differences be- ence-space specifications. These specifications, along with tween rural and urban respondents can be found in Appen- their discussion, can be found in Appendix C. dix B. 96(4) Parthum and Ando: Benefits of Nutrient Reductions 597

Figure 3 Responses to Questions about How Familiar Respondents Are with Water Quality Issues in the Watershed and How Frequently They Experience Algal Blooms in Surface Waters in the Watershed

Figure 3 shows the distributions of answers water. Fewer than 20% of respondents re- to qualitative questions about familiarity with ported having fished in the USRW at all. local algal blooms and water quality concerns However, nearly 50% reported having visited described in the survey. Nearly 80% of the the river or walked trails near the river.11 sample reported having at least some famil- Table 2 presents the main regression re- iarity with the water quality issues discussed sults, estimating equation [3] (WTP-space) in the survey, and about the same number of respondents reported experience with algal 11 A full summary of responses to visitation questions can blooms in the rivers or connected bodies of be found in Appendix D. 598 Land Economics November 2020

Table 2 Marginal Willingness to Pay (MWTP) to Reduce Nutrient Transmission to the Gulf of Mexico

(1) (2) Full Sample ASC Heterogeneity Mean MWTP Std. Dev. Mean MWTP Std. Dev. Distance (miles) –0.67*** (0.15) 92.57*** (18.69) –0.68*** (0.15) 1.22*** (0.26) Fish species 0.73** (1.48) 1.06*** (0.26) 4.72** (1.55) 12.32*** (2.14) Fish population 0.17** (0.06) 6.58*** (2.12) 0.16** (0.06) 0.38*** (0.08) Algal blooms (%) 0.77*** (0.11) 0.35** (0.09) 0.88*** (0.10) 0.96*** (0.13) Nutrient target 0.95*** (0.13) 0.85*** (0.16) 1.14*** (0.13) 0.89*** (0.12) Status quo (no program) –69.49*** (14.78) 1.42*** (0.23) –20.25 (13.48) 77.02*** (21.19) Status quo × Rural –48.79*** (14.33) 171.45*** (26.29) Status quo × Aware of water issues –65.82*** (16.34) 106.84*** (21.01) λ (cost coefficient) –3.17*** (0.32) 0.85*** (0.13) –2.71*** (0.42) 0.77*** (0.12) Observations (respondents) 2,058 (343) 2,058 (343) Log-likelihood –1,717.19 –1,717.77 AIC 3,506.38 3,527.54 McFadden ρ2 0.15 0.15 Note: Standard errors in parentheses. Column 1 provides the results of the WTP-space model for the pooled (full) sample. Column 2 introduces an interaction between the status quo dummy and respondent characteristics. Correlation matrices of the random parameters can be found in Appendix C. ASC, alternative specific constant. * p < 0.05; ** p < 0.01; *** p < 0.001. for the full sample. The regression in column likelihood of achieving the watershed’s nutri- 1 includes just the core model parameters. ent loss target. The regression in column 2 introduces an in- The large MWTP to move away from the teraction term between the status quo dummy status quo suggests that respondents have (ASC) and several respondent characteristics. strong preferences for having a new program All mean WTP coefficients in column 1 are instead of the status quo regardless of the statistically significant at the 1% level. The variable attributes in our choice scenarios. coefficient on the status quo (no program) op- Column 2 explores two factors contributing tion is large and negative and suggests respon- to these preferences. Respondents who live dents strongly prefer having a water-quality in more rural areas of the watershed and re- improvement program than not. The coeffi- spondents who are familiar with surface water cient on distance is also negative—people issues in the area are willing to pay signifi- prefer a program focused on the river close to cantly more for moving away from the status quo. Rural residents are estimated to value where they live. The coefficients on fish spe- this move from status quo $49 more than cies and fish population are positive; people urban residents. Those who reported being would be willing to pay nearly $5 per year familiar, very familiar, or very familiar and to have an additional species of game fish in involved with watershed quality issues value the river, and they separately place a positive this move from the status quo $66 more than value on the total number of individual fish those who are less aware. While preferences in the river. The coefficients onalgal blooms for the status quo may vary, the holistic set and nutrient target are positive. People would of preferences is consistent between rural and gain utility from reducing the frequency of urban respondents in this watershed.12 these local problems in their watershed, with an average annual MWTP of $0.77 for a 1% 12 Regressions for the separate rural and urban subsamples reduction in the frequency of algal blooms. can be found in Appendix C. A likelihood ratio test of joint Respondents also place a large value on nutri- preference stability tests the fit of separate regressions for the two subsamples against the constrained pooled sample. ent target, with an average annual MWTP of We fail to reject the null hypothesis that MWTP values are $0.95 for a 1 percentage point increase in the jointly similar across the two subsamples. 96(4) Parthum and Ando: Benefits of Nutrient Reductions 599

WTP= MWTP×∆ x . [5] 6. Integrated Assessment z ∑∑nj zx,,j zj Application From equation [5], the total WTP in the watershed is simply the sum of WTPz over all zip To illustrate how benefits from water quality codes in the USRW. improvements are distributed throughout the Table 3 summarizes the results of our policy watershed, we recover the conditional indi- simulations. Panel A identifies the scenarios, vidual-specific means of MWTP for every re- panel B considers benefits from only the envi- spondent in our sample (Greene, Hensher, and ronmental attributes in the model, and panel C Rose 2005). We use the primary specification adds to panel B by also including the benefits in our analysis (Table 2, column 1) to recover from moving away from the status quo—the conditional individual-specific means. For MWTP associated with the ASC in our model. each zip code in our sample, we average the The first scenario models a 50% reduction in MWTP over the respondents who lived in that only the frequency of algal blooms in river zip code. This gives us zip code–level varia- section A. Scenario 2 models this same im- tion in the MWTP for each attribute. provement except for river section C. This Zip codes are considered rural if there are allows us to hold all other attributes constant fewer than 1,000 residents per square mile, to see how benefits might accrue differently and urban otherwise (Ratcliffe et al. 2016; depending on where the improvement takes U.S. Census Bureau 2019). This allows us to place. Scenario 3 models a 75% likelihood tally welfare changes separately for the rural that the watershed reaches its nutrient loss tar- and urban areas in the USRW. The distribu- get of 45% by the year 2040. Scenario 4 intro- tions of MWTP in the rural and urban zip duces a more complete improvement scenario codes are as expected and have significant in which river section A sees a 75% reduction overlap.13 in the frequency of algal blooms, sections A Policy simulations, or “state-of-the-world” and B receive an additional 50 fish (popula- experiments, simulate a change in the levels tion) per 100 yards of river, section A receives of the environmental attributes to recover an an additional two species of game fish per 100 individual’s total WTP for the suite of im- yards of river, and there is a 100% likelihood provements over the status quo level (Holmes, of reaching the watershed’s nutrient target. Adamowicz, and Carlsson 2017). For ex- Reducing the frequency of local algal ample, an individual’s WTP for a change in blooms in just one of the four reaches of the attribute x1 is that individual’s MWTP for watershed yields around $1 million to $1.6 x1 multiplied by the change in x1’s level: million per year depending on the location WTP= MWTP ×∆ x . If more than one at- xx111 of the improvement (Table 3, panel B, col- tribute is changing, then the individual’s WTP umns 1 and 2). A 75% likelihood of reaching for changes in both attributes is the sum of the the nutrient target is worth $4.4 million per WTP for each attribute j that is changing: year (Table 3, panel B, column 3). Finally, the most comprehensive scenario (Table 3, panel WTP= MWTP×∆ x . [4] ∑ j xjj B, column 4) yields benefits of around $7 million per year. Because we have zip code–specific MWTP Table 3 also provides a summary of the for each attribute, we estimate changes in wel- average values per household for each of the fare for each zip code under different states scenarios. Household WTPs are calculated of the world. Moreover, in zip code z over n at the zip code level. We provide the average households, the total WTP for improvements WTP for each scenario throughout the water- in a set of attributes indexed by j is shed as well as the average WTP in the rural and urban areas separately. Reducing algal blooms by 50% has an average value of $9 or 13 A full summary of the recovered conditional individual-specific means of the MWTP for each attribute can be $15 per year depending on where in the water- found in Appendix D. shed it occurs, and the average value of a 75% 600 Land Economics November 2020

Table 3 Sample Integrated Assessment Value Estimates (Total WTP)

(1) (2) (3) (4) Panel A: Scenarios Algal blooms 50% reduced 50% reduced — 75% reduced Area A only Area C only Area A only Nutrient target — — 75% likelihood 100% likelihood Fish species — — — +2 species Area A only Fish population — — — +50 population Areas A and B Panel B: Without ASC Total annual benefits (dollars) 1,057,497 1,697,818 4,406,411 7,126,757 Rural areas 768,279 985,443 2,612,890 4,512,142 Urban areas 289,218 712,376 1,793,521 2,614,615 Mean benefits per household (dollars) 9.30 14.93 38.75 62.67 Rural areas 13.51 17.33 45.95 79.35 Urban areas 5.09 12.53 31.54 45.98 Panel C: With ASC Total annual benefits (dollars) 3,648,648 4,288,969 6,997,562 9,717,908 Rural areas 2,413,881 2,631,044 4,258,491 6,157,744 Urban areas 1,234,768 1,657,925 2,739,071 3,560,165 Mean benefits per household (dollars) 32.08 37.71 61.53 85.45 Rural areas 42.45 46.27 74.89 108.3 Urban areas 21.72 29.16 48.17 62.61 Note: Benefits are estimated using equation [5]. These are estimates of compensating variation for the improvements modeled in the integrated assessment model exercise. In aggregate, rural areas of the watershed stand to benefit nearly twice as much as the urban clusters. Rural areas of the watershed also tend to have a higher per household WTP for each scenario. This is largely because a majority of the improvements will be realized in more rural areas of the watershed. ASC, alternative specific constant; WTP, willingness to pay. change of the watershed doing its part for receive larger per household benefits because hypoxia reduction is $39 per year per house- the river sections—and the corresponding im- hold. The comprehensive scenario in column provements—are primarily in rural areas of 4 produces average benefits of $63 per year the watershed. per household. To see where the benefits from the policy simulations accrue throughout the watershed, 7. Discussion Figure 4 provides maps of both the total WTP (panel A) and the per household WTP (panel We have carried out a choice experiment sur- B) in each zip code. Benefits are most dense vey to estimate how much residents in a sub- where population is most dense (panel A). watershed of the MRB are willing to pay to However, when we map benefits based on per improve local fish diversity and populations household estimates, we see the distribution in their rivers, reduce the prevalence of local is often higher in the rural areas throughout algal blooms, and ensure that their watershed the watershed (panel B).14 Rural areas tend to does its part to reduce hypoxia in the Gulf of Mexico. While current efforts in the MRB to reduce nutrients and sediment are driven by 14 Refinements could be made when modeling WTP concern about water quality far away in the throughout the watershed, or for use in transfer to similar watersheds, using spatial regression methods such as those gulf, we find that people in our study area discussed by Johnston, Besedin, and Holland (2019) or De Valk and Rolfe (2018). However, the focus of this paper is to effects of policies related to water quality that span geo- provide a proof of concept for estimating the distributional graphically and culturally diverse landscapes. 96(4) Parthum and Ando: Benefits of Nutrient Reductions 601

Figure 4 Spatial Distribution of Total Willingness to Pay in Each Zip Code throughout the Watershed (Panel A) and per Household Marginal Willingness to Pay (Panel B)

would gain significant benefit from the local be quite small (Keiser, Kling, and Shapiro environmental improvements that could re- 2019). In contrast, we find that people would sult from reduced nutrient pollution and from gain large value from reducing the frequency helping to reduce environmental problems in of local algal blooms, with respondents will- the gulf. ing to pay nearly $40 per year to reduce the Much traditional research on surface water frequency of nearby algal blooms by 50%. quality values has focused on generic mea- Algal blooms are becoming more prevalent sures of whether waters are boatable, fishable, as climate change expands hot summer con- and swimmable, and the resulting values can ditions; our result implies that economists and 602 Land Economics November 2020 water quality modelers should pay increased alone, total WTP for reaching a 75% likeli- attention to the impact of management and hood of reaching nutrient reduction targets policies on those particularly harmful mani- (scenario 3) is estimated at approximately festations of nutrient pollution. $4.4 million annually. And when modeled Residents of the U.S. Midwest gain no use with improvements that will likely come as value from reducing hypoxia in the Gulf of complements for any policy targeting reduc- Mexico. However, we find that people in our tions in nutrient loss and transmission to the study area would benefit from increasing the gulf (scenario 4), total annual benefits within likelihood that their watershed reaches the tar- this small watershed are estimated to exceed get set for it under the INLRS; the average re- $7 million per year. spondent would be willing to pay $48 to have Debate over nutrient loss reduction strate- even a 50% chance of the watershed’s goal gies continues. To inform that debate, analysts being met. This finding provides further com- should quantify the full range of costs and pelling rationale for investments in programs benefits and how costs and benefits are distrib- that would reduce nutrient loss, a discussion uted among groups of people in the landscape. in which government agencies, NGOs, and in- Our findings can play an important role in that dustry groups are all currently engaged. effort. However, more work needs to be done Our estimates suggest that people in this in order to further uncover and understand the landlocked part of the Midwest would gain overlooked benefits of reductions in nutrient large value from improving local game fish loss and transmission. Future research would diversity and fish populations. This result do well to explore how values vary through- seems to be capturing significant nonuse val- out the MRB for improving local fish habitat, ues for having thriving river ecosystems in the avoiding local algal blooms, and solving re- region, since only a small fraction of respon- gional environmental problems like hypoxia dents reported engaging in local fishing. Most in the Gulf of Mexico. Additional work is previous research on the value of fish species also needed to understand the factors driving and populations comes from travel cost and people in our study to express such strong an- recreational site choice models that can only tipathy for a status quo that does nothing to capture use values (Phaneuf et al. 2013; Mel- address pervasive surface water pollution in strom et al. 2015). The large nonuse values we the United States. estimate in this study support the well-known claim that revealed preference estimates may not capture the full range of benefits from en- Disclaimer and Acknowledgments vironmental improvements (Adamowicz et al. The views expressed in this paper are those of 1998; Hanley and Czajkowski 2019). the authors and do not necessarily reflect the Economists, other social scientists, and views or policies of the U.S. Environmental policy makers have wondered if there is a Protection Agency. The paper has not been rural-urban divide in the values people place subjected to the agency’s required peer and on environmental improvements. In this case, policy review. No official agency endorse- we find that rural and urban preferences are ment should be inferred. We would like to similar. If anything, rural residents may place thank Maria Chu, Jason Knouft, and Alejan- more value on a move away from the status dra Botero-Acosta for their work modeling quo toward environmental improvement. This the biophysical characteristics of the water- finding implies that people in the rural areas, shed, Ketong Xie for valuable research assis- where many of the changes needed to improve tance, participants from the W4133 working water quality will be implemented, may also group (2018) and the Program in Environ- have high WTP for those improvements them- mental and Resource Economics (pERE) at selves. the University of Illinois Urbana-Champaign, Finally, the results from our simple simula- Ben Gramig, Richard Ready, Klaus Moelt- tions suggest that the total values water quality ner, Robert Johnston, participants of the So- improvement could bring to a watershed like cial Cost of Water Pollution Working Group our study area are not trivial. For the USRW at Cornell University, and an anonymous 96(4) Parthum and Ando: Benefits of Nutrient Reductions 603 reviewer for valuable feedback and sugges- Botero-Acosta, A., M. L. Chu, and C. Huang. tions made during the design, analysis, and 2019. “Impacts of Environmental Stressors on submission stages of this study. This paper is Nonpoint Source Pollution in Intensively Man- based in part on work funded by USDA-NIFA aged Hydrologic Systems.” Journal of Hydrol- grant #1008843. All data and replication files ogy 579: 124056. https://doi.org/10.1016/j.jhy- can be found at https://doi.org/http://doi. drol.2019.124056. org/10.5281/zenodo.3692738. Brouwer, Roy, Julia Martin-Ortega, and Julio Ber- bel. 2010. “Spatial Preference Heterogeneity: A Choice Experiment.” Land Economics 86 (3): References 552–68. https://doi.org/10.3368/le.86.3.552. Carson, Richard T., and Mikołaj Czajkowski. 2019. Adamowicz, Wiktor, Peter Boxall, Michael Wil- “A New Baseline Model for Estimating Will- liams, and Jordan Louviere. 1998. “Stated ingness to Pay from Discrete Choice Models.” Preference Approaches for Measuring Passive Journal of Environmental Economics and Man- Use Values: Choice Experiments and Contin- agement 95: 57–61. https://doi.org/10.1016/j. gent Valuation.” American Journal of Agri- jeem.2019.03.003. cultural Economics 80 (1): 64–75. https://doi. Carson, Richard T., and Theodore Groves. 2007. org/10.2307/3180269. “Incentive and Informational Properties of Adamowicz, Wiktor, Jordan Louviere, and Joffre Preference Questions.” Environmental and Swait. 1998. Introduction to Attribute-Based Resource Economics 37: 181–210. https://doi. Stated Choice Methods. Final report to Re- org/10.1007/s10640-007-9124-5. source Evaluation Branch, Damage Assessment Carson, Richard T., and Robert Cameron Mitch- Center, NOAA. Washington, DC: National ell. 1993. “The Value of Clean Water: The Pub- Oceanic and Atmospheric Administration. lic’s Willingness to Pay for Boatable, Fishable, Alexander, Richard B., Richard A. Smith, Gregory and Swimmable Quality Water.” Water Re- E. Schwarz, Elizabeth W. Boyer, Jacqueline V. sources Research 29 (7): 2445–54. https://doi. Nolan, and John W. Brakebill. 2008. “Differ- org/10.1029/93WR00495. ences in Phosphorus and Nitrogen Delivery to Caussade, Sebastián, Juan Dios Ortúzar, Luis I. the Gulf of Mexico from the Mississippi River Rizzi, and David A. Hensher. 2005. “Assessing Basin.” Environmental Science and Technol- the Influence of Design Dimensions on Stated ogy 42 (3): 822–30. https://doi.org/10.1021/ Choice Experiment Estimates.” Transportation es0716103 Research Part B: Methodological 39 (7): 621– Andres, Kara J., Huicheng Chien, and Jason H. 40. https://doi.org/10.1016/j.trb.2004.07.006. Knouft. 2019. “Hydrology Induces Intraspecific Collins, Jill P., and Christian A. Vossler. 2009. Variation in Freshwater Fish Morphology under “Incentive Compatibility Tests of Choice Ex- Contemporary and Future Climate Scenarios.” periment Value Elicitation Questions.” Journal Science of the Total Environment 671: 421–30. of Environmental Economics and Manage- https://doi.org/10.1016/j.scitotenv.2019.03.292. ment 58 (2): 226–35. https://doi.org/10.1016/j. Arcury, Thomas A., and Eric Howard Christian- jeem.2009.04.004. son. 1993. “Rural-Urban Differences in Envi- Coppess, Johnathan. 2016. “Dead Zones and ronmental Knowledge and Actions.” Journal of Drinking Water, Part 3: The Illinois Nutrient Environmental Education 25 (1): 19–25. https:// Loss Reduction Strategy.” Farmdoc Daily 6: doi.org/10.1080/00958964.1993.9941940. 53. Available at https://farmdocdaily.illinois. Babcock, Bruce A., and Catherine L. Kling. 2015. edu/2016/03/dead-zones-drinking-water-part3. “Costs and Benefits of Fixing Gulf Hypoxia.” html. Iowa Ag Review 14 (4): article 4. Available at Cummings, Ronald G., Glenn W. Harrison, and E. https://lib.dr.iastate.edu/iowaagreview/vol14/ Elisabet Rutström. 1995. “Homegrown Values iss4/4. and Hypothetical Surveys: Is the Dichotomous Bergstrom, John C., and John B. Loomis. 2017. Choice Approach Incentive-Compatible?” “Economic Valuation of River Restoration: An American Economic Review 85 (1): 260–66. Analysis of the Valuation Literature and Its Uses https://www.jstor.org/stable/2118008. in Decision-making.” Water Resources and Cummings, Ronald G., and Laura O. Taylor. 1999. Economics 17: 9–19. https://doi.org/10.1016/j. “Unbiased Value Estimates for Environmental wre.2016.12.001. Goods: A Cheap Talk Design for the Contin- 604 Land Economics November 2020

gent Valuation Method.” American Economic Gibbs, Bob. 2016. “EPA Edicts Hurt Rural Amer- Review 89 (3): 649–65. https://doi.org/10.1257/ ica.” The Hill, February 8. https://thehill.com/ aer.89.3.649. opinion/op-ed/268695-epa-edicts-hurt-rural- Czajkowski, Mikołaj, Wiktor Budziński, Danny america. Campbell, Marek Giergiczny, and Nick Hanley. Glenk, Klaus, Robert J. Johnston, Jürgen Mey- 2017. “Spatial Heterogeneity of Willingness to erhoff, and Julian Sagebiel. 2019. “Spatial Pay for Forest Management.” Environmental Dimensions of Stated Preference Valuation and Resource Economics 68: 705–27. https:// in Environmental and Resource Economics: doi.org/10.1007/s10640-016-0044-0. Methods, Trends and Challenges.” Environ- Czajkowski, Mikołaj, Nick Hanley, and Jacob mental and Resource Economics 75: 215–42. LaRiviere. 2015. “The Effects of Experience https://doi.org/10.1007/s10640-018-00311-w. on Preferences: Theory and Empirics for Envi- Greene, William H., David A. Hensher, and John ronmental Public Goods.” American Journal of M. Rose. 2005. “Using Classical Simula- Agricultural Economics 97 (1): 333–51. https:// tion-Based Estimators to Estimate Individual doi.org/10.1093/ajae/aau087. WTP Values.” In Applications of Simulation Czajkowski, Mikołaj, Christian A. Vossler, Wiktor Methods in Environmental and Resource Eco- Budziński, Aleksandra WiŚniewska, and Ewa nomics, ed. Riccardo Scarpa and Anna Alber- Zawojska. 2017. “Addressing Empirical Chal- ini, 17–33. Dordrecht: Springer Netherlands. lenges Related to the Incentive Compatibility of https://doi.org/10.1007/1-4020-3684-1_2. Stated Preferences Methods.” Journal of Eco- Greenley, Douglas A., Richard G. Walsh, and nomic Behavior and Organization 142: 47–63. Robert A. Young. 1981. “Option Value: Em- https://doi.org/10.1016/j.jebo.2017.07.023. pirical Evidence from a Case Study of Rec- Daly, Andrew, Stephane Hess, and Kenneth Train. reation and Water Quality.” Quarterly Jour- nal of Economics 96 (4): 657–73. https://doi. 2012. “Assuring Finite Moments for Will- org/10.2307/1880746. ingness to Pay in Random Coefficient Mod- Hanley, Nick, and Mikołaj Czajkowski. 2019. “The els.” Transportation 39: 19–31. https://doi. Role of Stated Preference Valuation Methods in org/10.1007/s11116-011-9331-3. Understanding Choices and Informing Policy.” De Valck, Jeremy, and John Rolfe. 2018. “Spa- Review of Environmental Economics and Policy tial Heterogeneity in Stated Preference Valu- 13 (2): 248–66. https://doi.org/10.1093/reep/ ation: Status, Challenges and Road Ahead.” rez005. International Review of Environmental and Re- Hanley, Nick, Felix Schläpfer, and James Spur- source Economics 11 (4): 355–422. https://doi. geon. 2003. “Aggregating the Benefits of org/10.1561/101.00000097. Environmental Improvements: Distance-De- Diaz, Robert J., and Rutger Rosenberg. 2008. cay Functions for Use and Non-use Values.” “Spreading Dead Zones and Consequences for Journal of Environmental Management 68 Marine Ecosystems.” Science 321 (5891): 926– (3): 297–304. https://doi.org/10.1016/S0301- 29. https://doi.org/10.1126/science.1156401. 4797(03)00084-7. Dissanayake, Sahan T. M., and Amy W. Ando. Hanley, Nick, Robert E. Wright, and Vic Adamo- 2014. “Valuing Grassland Restoration: Prox- wicz. 1998. “Using Choice Experiments to imity to Substitutes and Trade-offs among Con- Value the Environment.” Environmental and servation Attributes.” Land Economics 90 (2): Resource Economics 11: 413–28. https://doi. 237–59. https://doi.org/10.3368/le.90.2.237 org/10.1023/A:1008287310583. Farber, Dan. 2018. “Politics, the Environment, Hensher, David A., John M. Rose, and William and the Rural/Urban Divide.” Legal Planet, H. Greene. 2005. Applied Choice Analysis: October 29. Available at http://legal-planet. A Primer. Cambridge: Cambridge University org/2018/10/29/environmental-law-and-the-ru- Press. ral-urban-divide. Hess, Stephane, and Kenneth Train. 2017. “Cor- Ferrini, Silvia, and Riccardo Scarpa. 2007. “De- relation and Scale in Mixed Logit Models.” signs with A Priori Information for Nonmar- Journal of Choice Modelling 23: 1–8. https:// ket Valuation with Choice Experiments: A doi.org/10.1016/j.jocm.2017.03.001. Monte Carlo Study.” Journal of Environmental Hole, Arne Risa. 2015. DCREATE: Stata Module Economics and Management 53 (3): 342–63. to Create Efficient Designs for Discrete Choice https://doi.org/10.1016/j.jeem.2006.10.007. Experiments. Statistical Software Components 96(4) Parthum and Ando: Benefits of Nutrient Reductions 605

S458059. Boston: Boston College, Department Ladenburg, Jacob, and Søren Bøye Olsen. 2014. of Economics. “Augmenting Short Cheap Talk Scripts with a Holmes, Thomas P., Wiktor L. Adamowicz, and Repeated Opt-Out Reminder in Choice Experi- Fredrik Carlsson. 2017. “Choice Experiments.” ment Surveys.” Resource and Energy Econom- In A Primer on Nonmarket Valuation, ed. Patri- ics 37: 39–63. https://doi.org/10.1016/j.rese- cia A Champ, Kevin J. Boyle, and Thomas C. neeco.2014.05.002. Brown, 133–86. Dordrecht: Springer. Lant, Cristopher L., and Rebecca S. Roberts. 1990. Johnston, Robert J., Elena Y. Besedin, and Bene- “Greenbelts in the Cornbelt: Riparian Wetlands, dict M. Holland. 2019. “Modeling Distance De- Intrinsic Values, and Market Failure.” Environ- cay within Valuation Meta-analysis.” Environ- ment and Planning A 22 (10): 1375–88. https:// mental and Resource Economics 72: 657–90. doi.org/10.1068/a221375. https://doi.org/10.1007/s10640-018-0218-z. LaRiviere, Jacob, Mikołaj Czajkowski, Nick Johnston, Robert J., Elena Y. Besedin, and Ryan Hanley, Margrethe Aanesen, Jannike Falk-Pe- Stapler. 2017. “Enhanced Geospatial Validity tersen, and Dugald Tinch. 2014. “The Value of for Meta-analysis and Environmental Bene- Familiarity: Effects of Knowledge and Objec- fit Transfer: An Application to Water Quality tive Signals on Willingness to Pay for a Public Improvements.” Environmental and Resource Good.” Journal of Environmental Economics Economics 68: 343–75. https://doi.org/10.1007/ and Management 68 (2): 376–89. https://doi. s10640-016-0021-7. org/10.1016/j.jeem.2014.07.004. Johnston, Robert J., Elena Y. Besedin, and Ryan Louviere, Jordan, Deborah Street, Richard Carson, F. Wardwell. 2003. “Modeling Relationships Andrew Ainslie, J. R. Deshazo, Trudy Cam- between Use and Nonuse Values for Surface eron, David Hensher, Robert Kohn, and Tony Water Quality: A Meta-analysis.” Water Re- Marley. 2002. “Dissecting the Random Compo- sources Research 39 (12): 1363. https://doi. nent of Utility.” Marketing Letters 13: 177–93. org/10.1029/2003WR002649. https://doi.org/10.1023/A:1020258402210. Johnston, Robert J., Kevin J. Boyle, Wiktor (Vic) Manifold, Beverly. 1998. Executive Summary: The Adamowicz, Jeff Bennett, Roy Brouwer, Trudy Quality of Our Nation’s Water. Report to U.S. Ann Cameron, W. Michael Hanemann, Nick Congress. Washington, DC: U.S. Environmen- Hanley, Mandy Ryan, Riccardo Scarpa, et al. tal Protection Agency. 2017. “Contemporary Guidance for Stated McFadden, Daniel. 1973. “Conditional Logit Preference Studies.” Journal of the Association Analysis of Qualitative Choice Behavior.” In of Environmental and Resource Economists 4 Frontiers in Econometrics, ed. Paul Zarembka, (2): 319–405. https://doi.org/10.1086/691697. 105–42. New York: Wiley. Johnston, Robert J., and Joshua M. Duke. 2007. Melstrom, Richard T., Frank Lupi, Peter C. Essel- “Is Willingness to Pay for Farmland Preserva- man, and R. Jan Stevenson. 2015. “Valuing Rec- tion Transferable across States? Evidence from reational Fishing Quality at Rivers and Streams.” a Choice Experiment.” American Agricultural Water Resources Research 51 (1): 140–50. Economics Association (New Name 2008: Ag- https://doi.org/10.1002/2014WR016152. ricultural and Applied Economics Association). Meyerhoff, Jürgen, Marco Boeri, and Volkmar Milwaukee, WI: Agricultural and Applied Eco- Hartje. 2014. “The Value of Water Quality Im- nomics Association. provements in the Region Berlin–Brandenburg Johnston, Robert J., Benedict M. Holland, and as a Function of Distance and State Residency.” Liuyang Yao. 2016. “Individualized Geocoding Water Resources and Economics 5: 49–66. in Stated Preference Questionnaires: Implica- https://doi.org/10.1016/j.wre.2014.02.001. tions for Survey Design and Welfare Estima- Mobley, Catherine. 2016. “What Matters When tion.” Land Economics 92 (4): 737–759. https:// Explaining Environmentalism at the Watershed doi.org/10.3368/le.92.4.737. Level: Who You Are, Where You Live, What Keiser, David A., Catherine L. Kling, and Joseph You See, or What You Perceive?” Environment S. Shapiro. 2019. “The Low but Uncertain and Behavior 48 (9): 1148–74. https://doi. Measured Benefits of US Water Quality Pol- org/10.1177/0013916515586058. icy.” Proceedings of the National Academy of Moeltner, Klaus. 2019. “Bayesian Nonlinear Sciences U S A 116 (12): 5262–69. https://doi. Meta Regression for Benefit Transfer.”Jour - org/10.1073/pnas.1802870115. nal of Environmental Economics and Man- 606 Land Economics November 2020

agement 93: 44–62. https://doi.org/10.1016/j. tainty to Mitigate Hypothetical Bias in a Stated jeem.2018.10.008. Choice Experiment.” Land Economics 86 (2): Needham, Katherine, and Nick Hanley. 2020. 363–81. https://doi.org/10.3368/le.86.2.363. “Prior Knowledge, Familiarity and Stated Pol- Salka, William M. 2001. “Urban-Rural Conflict icy Consequentiality in Contingent Valuation.” over Environmental Policy in the Western Journal of Environmental Economics and Pol- United States.” American Review of Pub- icy 9 (1): 1–20. https://doi.org/10.1080/216065 lic Administration 31 (1): 33–48. https://doi. 44.2019.1611481. org/10.1177/02750740122064820. Penn, Jerrod, and Wuyang Hu. 2020. “Certainty Sarrias, Mauricio, and Ricardo Daziano. 2017. Follow-up Efficacy under Potential and Actual “Multinomial Logit Models with Continuous Hypothetical Bias: A Meta-analysis.” American and Discrete Individual Heterogeneity in R: Journal of Agricultural Economics forthcom- The gmnl Package.” Journal of Statistical Soft- ing. ware 79 (2): 1–46. https://doi.org/10.18637/jss. Phaneuf, Daniel J. 2002. “A Random Utility Model v079.i02. for Total Maximum Daily Loads: Estimating Scarpa, Riccardo, Mara Thiene, and Kenneth Train. the Benefits of Watershed-Based Ambient Wa- 2008. “Utility in Willingness to Pay Space: A ter Quality Improvements.” Water Resources Tool to Address Confounding Random Scale Research 38 (11): 36-1–36-11. https://doi. Effects in Destination Choice to the Alps.” org/10.1029/2001WR000959. American Journal of Agricultural Economics Phaneuf, Daniel J., Roger H. Haefen, Carol Man- 90 (4): 994–1010. https://doi.org/10.1111/ sfield, and George van Houtven. 2013.Mea - j.1467-8276.2008.01155.x. suring Nutrient Reduction Benefits for Policy Schaafsma, Marije, and Roy Brouwer. 2013. “Test- Analysis Using Linked Non-market Valuation ing Geographical Framing and Substitution and Environmental Assessment Models. Final Effects in Spatial Choice Experiments.” Jour- Report on Stated Preference Surveys. Report to nal of Choice Modelling 8: 32–48. https://doi. the U.S. EPA. Washington, DC: U.S. Environ- org/10.1016/j.jocm.2013.04.007. mental Protection Agency. Street, Deborah J., and Leonie Burgess. 2007. The Rabalais, N. N., R. J. Díaz, L. A. Levin, R. E. Construction of Optimal Stated Choice Exper- Turner, D. Gilbert, and J. Zhang. 2010. “Dy- iments: Theory and Methods. Hoboken, NJ: namics and Distribution of Natural and Hu- man-Caused Hypoxia.” Biogeosciences 7: 585– John Wiley and Sons. 619. https://doi.org/10.5194/bg-7-585-2010. Sutherland, Ronald J., and Richard G. Walsh. Rabotyagov, Sergy S., Catherine L. Kling, Philip 1985. “Effect of Distance on the Preservation W. Gassman, Nancy N. Rabaliais, and R. Eu- Value of Water Quality.” Land Economics 61 gene Turner. 2014. “The Economics of Dead (3): 281–91. https://doi.org/10.2307/3145843. Zones: Causes, Impacts, Policy Challenges, and Swait, Joffre, and Wiktor Adamowicz. 2001. a Model of the Gulf of Mexico Hypoxic Zone.” “Choice Environment, Market Complexity, Review of Environmental Economics and Pol- and Consumer Behavior: A Theoretical and icy 8 (1): 58–79. https://doi.org/10.1093/reep/ Empirical Approach for Incorporating Deci- ret024. sion Complexity into Models of Consumer Racevskis, Laila A., and Frank Lupi. 2006. Choice.” Organizational Behavior and Human “Comparing Urban and Rural Perceptions Decision Processes 86 (2): 141–67. https://doi. of and Familiarity with the Management org/10.1006/obhd.2000.2941. of Forest Ecosystems.” Society and Natu- Swait, Joffre, and Jordan Louviere. 1993. “The ral Resources 19 (6): 479–95. https://doi. Role of the Scale Parameter in the Estimation org/10.1080/08941920600663862. and Comparison of Multinomial Logit Models.” Ratcliffe, Michael, Charlynn Burd, Kelly Holder, Journal of Marketing Research 30 (3): 305–14. and Alison Fields. 2016. Defining Rural at the https://doi.org/10.1177/002224379303000303. U.S. Census Bureau. Washington, DC: U.S. Thiene, Mara, and Riccardo Scarpa. 2009. “De- Census Bureau. Available at https://www2.cen- riving and Testing Efficient Estimates of WTP sus.gov/geo/pdfs/reference/ua/Defining_Rural. Distributions in Destination Choice Models.” pdf. Environmental and Resource Economics 44: ar- Ready, Richard C., Patricia A. Champ, and Jennifer ticle 379. https://doi.org/10.1007/s10640-009- L. Lawton. 2010. “Using Respondent Uncer- 9291-7. 96(4) Parthum and Ando: Benefits of Nutrient Reductions 607

Train, Kenneth. 1998. “Recreation Demand Basin. Washington, DC: U.S. Environmental Models with Taste Differences over Peo- Protection Agency. ple.” Land Economics 74 (2): 230. https://doi. ———. 2013. Hypoxia Task Force Reassessment org/10.2307/3147053. 2013: Assessing Progress Made Since 2008. Train, Kenneth, and Melvyn Weeks. 2005. “Dis- Washington, DC: U.S. Environmental Protec- crete Choice Models in Preference Space and tion Agency. Willingness-to-Pay Space.” In Applications Van Houtven, George, John Powers, and Subhrendu of Simulation Methods in Environmental and K. Pattanayak. 2007. “Valuing Water Qual- Resource Economics, ed. Riccardo Scarpa ity Improvements in the United States Using and Anna Alberini, 1–16. Dordrecht: Springer Meta-analysis: Is the Glass Half-Full or Half- Netherlands. https://doi.org/10.1007/1-4020- Empty for National Policy Analysis?” Resource 3684-1_1. and Energy Economics 29 (3): 206–28. https:// U.S. Census Bureau. 2019. 2013–2017 American doi.org/10.1016/j.reseneeco.2007.01.002. Community Survey (ACS) 5-Year Estimates. Whitehead, John C., Glenn C. Blomquist, Thomas Washington, DC: U.S. Census Bureau. J. Hoban, and William B. Clifford. 1995. “As- U.S. Environmental Protection Agency. 2008. Mis- sessing the Validity and Reliability of Con- sissippi River/Gulf of Mexico Watershed Nutri- tingent Values: A Comparison of On-Site Us- ent Task Force. Gulf Hypoxia Action Plan 2008 ers, Off-Site Users, and Non-users.” Journal for Reducing, Mitigating, and Controlling Hy- of Environmental Economics and Manage- poxia in the Northern Gulf of Mexico and Im- ment 29 (2): 238–51. https://doi.org/10.1006/ proving Water Quality in the Mississippi River jeem.1995.1044.