Valuing Water Quality Monitoring: A Contingent Valuation...

Valuing Water Quality Monitoring:A Contingent Valuation ExperimentInvolving Hypothetical andReal Payments

Michael A. Spencer, Stephen K. Swallow, and Christopher J. Miller

This paper studies the preferences and willingness-to-pay of individuals for volunteer water

quality monitoring programs. The study involves supporting water quality monitoring at twoponds in the state of Rhode Island. The paper uses both a hypothetical and a real-payment

contingent valuation survey to directfy measure individual preferences and willingness-to-pay

(WTP) for volunteer water quality monitoring at the two ponds. The overall results of the

study suggest that hypothetical WTP is not statistically greater than real WTP, and that the

average survey respondent is willing to support water quality monitoring on one of the two

ponds. The study also finds that the specified purpose of water quality monitoring and certain

socioeconomic characteristics of a respondent significantly affect the respondent’s decision to

support volunteer water quality monitoring.

Since passage of the Federal Water Pollution Con-trol Act (FWPCA) of 1972, water quality has im-proved in some U.S. water bodies (Freeman 1990).Clean water yields positive dividends in terms ofecosystem health, quality of life for humans, andeconomic prosperity (U.S. Environmental Protec-tion Agency 1995). Federal, state, and local gov-ernments have enacted environmental policies andprograms to meet the regulatory requirements ofthe FWPCA, as well as the regulatory requirementsof the Safe Drinking Water Act and the OceanDumping Act, Economic analysis of the potentialbenefits associated with these policies and pro-grams has focused on valuing the benefits of waterquality improvements, such as the associated rec-reational benefits (Bockstael, McConnell, andStrand 1989; Freeman 1995; Needelman and Kealy

The authors are, respectively, graduate research assistrmt, associate pro-fessor, and graduate research assistmrt, Department of Environmentaland Natural Resource Economics, University of Rhode Island.

The authors gratefully thank Linda Green, program director of Uni-versity of Rhode Island Watershed Watch, for helping them establisbreal, potential, “adopt-a-pond” water qrratity monitoring programs forttrk economics experiment. Susan Jaacart, M]ke McGonagle, and EdWatusiak provided valuable assistance during the experiment, Fbrarrcialsupport from the USDA Higher Education Challenge Grant W6-3841 1-2798, USDA NRI/CGP Award 96-35403-3896, U.R.L Partnership forthe Coastal Environment, and National Science Foundation/U.S. Envi-rnnmenkd Protection Agency Partnership for Environmental ResearchGrant # R825307-01-O is gratefully acknowledged. This is Rhode IslandAgricultural Experiment Station contribution no. 3567. Any errors re-main tbe sole responsibilh y of the authors.

1995). Achieving and maintaining “good” waterquality, however, entails various costs, includingadministrative, capital, labor, enforcement, andmonitoring costs, Though these costs can be quitesignificant, certain costs, particularly monitoringcosts, can be reduced with volunteer help and do-nations.

The purpose of this paper is to consider the pref-erences and willingness-to-pay of individuals forvolunteer water quality monitoring programs. Ourstudy involves supporting water quality monitoringat two ponds in the state of Rhode Island. As ex-plained below, the two ponds differ in several re-spects, including the purpose of water qualitymonitoring. We use the contingent valuationmethod (CVM) to directly measure individualpreferences and willingness-to-pay for volunteerwater quality monitoring at the two ponds.

Moreover, because previous research suggestsdifferences between individual hypothetical andreal willingness-to-pay (WTP) for goods and ser-vices (Brown et al. 1996; Cummings, Harrison,and Rutstrom 1995; Cummings et al. 1997; Loomiset al. 1996; Neill et al. 1994; Seip and Strand1992),1 we consider both a hypothetical and a realpayment format for our CVM study. In an experi-

1See Hanemann (1994) for a review of studies which report no sta-tistical differences between hypothetical and real WTP.

Spencer, Swallow, and Miller Valuing Water Quality Monitoring 29

mental economic setting, we randomly assign re-spondents to one of two groups. One group re-ceives aCVM survey involving hypothetical pay-ments, while the other group receives a CVMsurvey involving real money, In valuing publicgoods, such as water quality monitoring, incentive-compatibility remains an issue for both paymentformats. In the hypothetical payment format, re-spondents may overvalue their WTP if they do notfully appreciate the financial obligations (i.e.,costs) behind their decisions. Conversely, in thereal payment format, respondents may undervaluetheir WTP if incentives to “free-ride” exist.

Efforts to address the so-called hypothetical bias(i.e., the difference between hypothetical and realWTP) are currently following three main direc-tions: (1) adjusting or calibrating hypotheticalWTP (Blackburn, Harrison, and Rutstrom 1994;Champ et al. 1997; Fox et al. 1995; Harrison et al.1998; Shogren 1993), (2) designing hypotheticalcontingent valuation surveys that yield resultssimilar to real payments (Cummings and Taylor1997), and (3) developing incentive-compatiblecontributions mechanisms for obtaining real-money values, which provide more accurate refer-ence points for assessing hypothetical bias (Ron-deau, Poe, and Schulze 1996). The third directionsuggests that hypothetical bias will be overstated ifreal-money payments are solicited with a contribu-tions mechanism that provides little economic in-centive to contribute to the public good. That is,calibration of hypothetical WTP may not beneeded if real WTP is truthfully revealed,

Although all three directions remain importantto the study of hypothetical bias, we approach thecurrent study more in the spirit of the third direc-tion.

Consequently, our study will interest not onlyenvironmental agencies and citizen groups in-volved with water quality but also researchers andagencies interested in experimental-economic ap-proaches to CVM (Fisher, Wheeler, and Zwick1993; Shogren 1993). Our overall results suggestthat hypothetical WTP exceeded real WTP; how-ever, given that our hypothetical WTP estimateshave high standard errors, we cannot conclude thathypothetical WTP is statistically greater than realWTP. In terms of respondent preferences for waterquality monitoring, we find that the average re-spondent is willing to support water quality moni-toring on one of the two ponds. We also find thatthe specified purpose of water quality monitoringand certain socioeconomic characteristics of a re-spondent significantly affect the respondent’s de-cision to support volunteer water quality monitor-ing. Although our respondent sample does not per-

mit us to make inferences about the general public,our results nevertheless may be of interest to vol-unteer monitoring groups and environmental agen-cies in their efforts to educate the public, recruitnew volunteers, and seek financial support.

Background on Volunteer WaterQuality Monitoring

Water quality monitoring is a key component inefforts to protect water resources. Information pro-vided by water quality monitoring can help iden-tify the actual environmental impacts resultingfrom pollution, detect trends in water quality, warnof potential problems, and, at times, locate sourcesof pollution and stimulate corrective action inproblem areas (Keeney 1996; U.S. EnvironmentalProtection Agency 1990).

Though water quality monitoring is important,public budgets, staff size, and/or geographic scopehave represented limitations to governmental ef-forts at state, federal, and local levels. However,with an increasing awareness that volunteer moni-toring programs can offer a cost-effective way toobtain credible information on water quality, manystates are turning to volunteer programs (U.S. En-vironmental Protection Agency 1990). Moreover,volunteer water quality monitoring programs helpdevelop an educated and involved citizenry dedi-cated to protecting water resources (Simpson 1991;U.S. Environmental Protection Agency 1990),

In Rhode Island, volunteer water quality moni-toring is coordinated by the University of RhodeIsland’s Watershed Watch Program. WatershedWatch provides volunteers with both classroomand field training, and it supplies the volunteerswith all necessary equipment. The annual cost forwater quality monitoring is $500 per water body(Green and Gold 1993). Funding for WatershedWatch has come from a variety of sources, includ-ing donations by individual citizens, local citizenassociations, local and state governments, privatecorporations, and nonprofit organizations (Herronand Green 1996). The stated goals of WatershedWatch are (1) to promote active citizen participa-tion in water quality protection, (2) to educate thepublic about water quality issues, (3) to obtainmultiyear surface water quality information both toascertain current conditions and to detect trends,and (4) to encourage management programs basedupon water quality information (Green and Gold1993, p. 1).

Currently, over 90% of Rhode Island’s data con-cerning the conditions of its lakes and ponds issupplied by volunteer monitors. Following struc-

30 April 1998

tured weekly, biweekly, and seasonal schedules,volunteers monitor several water quality param-eters, including water clarity, algal density, dis-solved oxygen, water temperature, alkalinity andpH, nutrient levels, salts, and bacteria (Green andGold 1993).

Random Utility Model

Public contributions to support volunteer monitor-ing under the University of Rhode Island Water-shed Watch Program come in two forms: (1) do-nations to support the general operation of Water-shed Watch and (2) contributions to support waterquality monitoring on specific water bodies, Forthe present study we focus on the second type ofcontribution; that is, we investigate public prefer-ences and WTP for volunteer monitoring on spe-cific water bodies.

Numerous water bodies exist throughout RhodeIsland, and each water body has its own character-istics in terms of size, water quality, location (e.g.,near the coast versus inland), surroundings (e.g.,rural versus urban setting), and type (e.g., lake,pond, river, stream, or other). In modeling an in-dividual’s preferences for water quality monitoringon specific water bodies, we assume an individu-al’s utility (i.e., satisfaction or happiness), Uin, as-sociated with monitoring any particular water bodydepends on the characteristics of the water body,the socioeconomic characteristics of the individual,and the individual’s net income after contributingto a volunteer monitoring program on the waterbody:

(1) Uin = U(X~, Sj, Mi - CJ= V(X~, S~, fl’f~– CJ + ‘in,

where Uin = individual i’s utility associated withmonitoring water body n; Xn = a vector of theattribute variables (i.e., characteristics) associatedwith water body n; Si = a vector of characteristicsdescribing individual i; kfi = individual i’s in-come; Ci. = individual i’s required monetary con-tribution to help support monitoring on water bodyn; V(”) = the deterministic component of utilitythat is econometrically measurable by the re-searcher; and &in = the random or unobservablecomponent of individual i’s utility associated withmonitoring water body n.2 Since Uin contains arandom component, sin, model (1) is called a “ran-dom utility model.”

2 Although unobservable to the researcher, &i. is known to individu-al i.

Agricultural and Resource Economics Review

The establishment of water quality monitoringon a water body depends on the collective contri-butions of individuals. That is, Watershed Watchcan add a water body to its list of monitoringsites if

5 Cin~ TC,i=1

where G = group size and TC = total costs ofmonitoring. Our discrete-choice survey design, ex-plained later, relies on variation in Cin among in-dividuals, generating a range of prices from whichto estimate average WTP.

Next, the probability that individual i willchoose to contribute to volunteer monitoring on aparticular water body is modeled as follows. First,we define the set of available alternatives (includ-ing the set of water bodies plus an option not tocontribute to any water body) as W.3 Individual iwill (by assumption) maximize his or her utility bychoosing alternative n if

(2) Ui~> Ui~ for all m + n in W.

Under decision rule (2), the probability that indi-vidual i will choose alternative n is

(3) Pi(n) = Pr(Ui. > Uim,for all m # n in W),

where Pr(”) represents the probability operator.Following the definition of utility in model (l), (3)can

(4)

or

(5)

be written as

Pi(n) = Pr(Vi~+ si, > Vim+ Ei~, for allm+ninw)

Pi(n) = Pr(&in- ~i~ > Vim– Vin, for allm+nin w),

where Vin = V(Xn, Sij AZi– Cin). Next, it can beshown that if we assume the e’s are independentand identically Type I Extreme Value distributedin standard form, then Pi (n) can be defined as thefollowing multinominal logit model (McFadden1974):

exP(VJ(6)

‘i(n)= ~ exp(vim)’

me w

3 In the experiment explained below, individual i faces a choice setconsisting of three alternatives: contribute to volunteer monitoring onPond A, contribute to volunteer monitoring on Pond B, or contribute toneither pond, The “neither pond” alternative Wows for the possibilitythat individual i either (1) receives no utility from having a pond moni-tnred or (2) simply receives greater utility from not having a pnnd moni-tored.


Estimation of model (6) requires specification ofthe functional form for Vin, for all iz=W We as-sume Vi. is linear in its parameters so that

(7a)

and

exp(~’Zin – ~CCiJ(7b)

‘i(n) = ~ eXp(f3’Zim - @Ccim)’

me W

where Zin = z(Xn, S’i),~‘ is a vector of parameters,and (3Cis the marginal utility of income, Note thatthe term 13cMidrops out in estimation because itappears in the utility function of every choice al-ternative.5 Finally, estimates of ~‘ and (3Ccan beobtained through maximum likelihood estimation.

Experimental Design

Our sample consisted of 140 students from threeclasses at the University of Rhode Island: twoseparate classes of a course entitled “Introductionto Resource Economics” and one class from amixed graduate-undergraduate course entitled “In-terdisciplinary Topics in the Coastal Environ-merit.” The experimental sessions took place dur-ing regular class hours, and none of the studentswere forced or pressured to participate in the ex-periment,6 Those who participated were given acash payment of $10.

Preliminary Procedures and Survey Materials

At the beginning of each experimental session, stu-dent-respondents sat in rows and were verballytold that the experiment was not a class assign-ment, but that it would involve them in making“real-life” decisions concerning environmentalgoods. The experimenter also instructed the re-spondents to remain quiet and not to communicatewith other respondents during the experiment.Next, the respondents were randomly split into twogroups. Respondents were asked, row-by-row, tocount-off 1, 2,3, 4, . . . . Then, by the flip of a coin,either even or odd numbered respondents were

4 A linear form is computationally convenient turd may be interpretedas a first-order approximation to a general utility function. Nonlinearfunctional forms were not found superior in the data nnalysis discussedbelow.

s Interaction terms involving M,, however, will not drop out. In ourdata analysis, we investigated several terms involving income interactedwith alternative specific dummies (i.e., constants) and the individualrequired contribution, but none of the terms added explanatory power toour models.

6 Only one student opted to leave, citing physical illness.

escorted to another classroom; these respondentswere assigned to the hypothetical-money survey,while the respondents remaining in the originalclassroom participated in the real-money survey. Inorder to control for interrespondent communica-tion and to give each respondent some privacywhile making decisions, respondents in both class-rooms were separated by at least one desk space.

After the respondents were separated into differ-ent classrooms, each respondent, in both groups,was given a survey and a ten-dollar cash paymentin an unsealed envelope. The survey contained foursections. The introduction summarized various as-pects of volunteer water quality monitoring inRhode Island, including its role in environmentalprotection, its alternative funding sources, whovolunteers in Rhode Island, and what the volun-teers do, Part A asked respondents about their rec-reational use of water bodies and contained severalattitudinal questions related to water quality andwater quality monitoring. The questions in Part Awere primarily used to motivate respondents tothink about what water quality monitoring meansto them, personally, Part B presented two “adopt-a-pond” scenarios in Rhode Island, followed by atrichotomous choice question that elicited a re-spondent’s preference and willingness-to-pay forwater quality monitoring at either pond. Part Ccontained standard demographic and socioeco-nomic questions.

The Adopt-a-Pond Scenarios

Through Watershed Watch, we were able to offer,for sale, a water quality monitoring program at twoponds in Rhode Island. At the time of this experi-ment, neither pond was being regularly monitored;thus, our respondents (in the real-money survey)had a real opportunity to add one or both ponds toWatershed Watch’s monitoring program. Specifi-cally, each respondent was asked to evaluatewhether helshe believed it was worthwhile for himlher to pay a specified amount so that WatershedWatch could add one of the ponds to its monitoringprogram for one year, If a respondent found itworthwhile to “adopt a pond,” then the respon-dent actually paid the specified contribution in thereal-money survey, whereas in the hypothetical-money survey, a respondent simply stated his/herpreference to adopt a pond,7

7 Since we made no promises to continue monitoring beyond one year,our focus was on obtaining the value for adding a pond to WatershedWatch’s list of monitoring sites for the first yew only. If a respondent didnot think one year’s worth of data was helpful, tbe respondent may havedecided not to contribute to any pond.

32 April 1998 Agricultural and Resource Economics Review

The ponds chosen for this study were WakefieldPond, located inland in northwestern Rhode Island,and Melville Pond, located near Narragansett Bayin southeastern Rhode Island, However, in order todevelop a more general model for water qualitymonitoring preferences, we arbitrarily labeledWakefield Pond as Pond A and Melville Pond asPond B. Table 1 lists the general characteristics ofPonds A and B. Pond A is relatively large (72acres), surrounded by wooded area, with no obvi-ous water clarity problem. In contrast, Pond B isrelatively small (5.5 acres), surrounded by housingdevelopment, with an apparent water clarity prob-lem.

In designing our contingent valuation survey,8we maintained a particular interest in how an in-dividual’s preference for water quality monitoringmight be affected by a pond’s surroundings and thepurpose of monitoring (table 1). Since water qual-ity is directly affected by what happens on theland’s surface, one might suspect that a pond sur-rounded by housing development remains moresusceptible to pollution problems (e.g., urban rtm-off) than a pond surrounded by wooded area. Achoice between monitoring a pond surrounded bywooded area versus housing development repre-sents a tradeoff between protecting a water re-source in a relatively pristine/wooded area versus aproblem-prone/suburban area, This choice mayalso represent a tradeoff between protecting wild-life health versus (human) public health. As for thepurpose of monitoring, we found no economicstudies that investigated individual preferences forwater quality monitoring, let alone individual pref-erences for specific water quality monitoring ob-jectives.

In order to measure the effect of pond surround-ings and the purpose of monitoring on an individu-al’s preference for water quality monitoring, weimplemented these characteristics as variable in-formation. That is, some respondents were notgiven information on pond surroundings and thepurpose of monitoring, while other respondentswere given this information. The other pond char-acteristics (table 1)—size, water clarity, and loca-tion—were used to give a basic description of the

8 The ovemtl development of the survey involved several steps. Fk’st,we talked with representatives from Watershed Watch and Trout Unlim-ited (another volunteer monitoring organization in Rhode Ishmd) to leanrabout water quality monitoring issues in general and at specific waterbodies. Next, we pretested a hypothetical-money version of the survey on166 students from a course entitled “Introduction to Philosophy” at theUniversity of Rhode Island. The last page of the pretest surveys asked forcomments regarding the survey’s difficulty. Based on the pretest resultsnnd comments, we shortened the survey’s introduction, and we madesome modifications to better clarify the adopt-a-pond scenarios.

Table 1. General Characteristics of PondsA and B

Characteristics Pond A Pond B

Size 72 acres 5.5 acres

Water clarity Good Poor

Location Inland Neor coast

Surrounded by Wooded area Housing development

Purpose of To worn of any To help find source ofmonitoring problems current problem

ponds. Every respondent was given these basic de-scriptions, which represented fixed, base informat-ion.

In addition to the pond characteristics in table 1,each monitoring fchoice alternative required an in-dividual monetary contribution. The required con-tribution for Ponds A and B never equaled eachother, and not every respondent faced the samerequired contributions as other respondents.

We followed the Addelman and Kempthome(1961) fractional factorial design method to createnine different combinations of the variable pondattributes: reauired contributions for Ponds A andB, pond surr&ndings, and purposes of monitoring.This resulted in nine versions of our trichotomouschoice question of choosing to monitor Pond A,Pond B, or neither pond. Each respondent, how-ever, faced only one choice question. Figure 1shows the format of a typical trichotomous choicequestion used in the surveys.

Prior to the trichotomous choice question, re-spondents were informed that a minimum totalcontribution from each of the three experimentalsessions was needed for either pond to be added toWatershed Watch’s monitoring program. Specifi-cally, the experimental instructions for the real-money surveys read:

In order to add either pond to Watershed Watch’8

monitoring program, we need a certain minimum total

contribution from each of the three courses.For thepeople in this room, the minimumtotal contributionneededis $Y for Pond A and $Y for Pond B. If wedo@ collect$Y for Pond A andlor$Y for Pond B,then anyonewho contributedto Pond A andlor PondB will have theirmoneycontributionrefundedtoday.However, if we collect $Y for Pond A and/or$Y forPond B, then Pond A mxflor Pond B will become acandidatefor the WatershedWatch monitoringpro-gram, conditionalon the contributionscollectedfromthe other courses.

That is, if we collect the minimum for Pond Aand/orPondB in all threecourses,thenPondA and/orPond B will be added to WatershedWatch’s moni-toringprogram;otherwise,PondA and/orPondB willQ be added to WatershedWatch’smonitoringpro-

Spencer, Swallow, and Miller Valuing Water Quality Monitoring.-

Given the following choices, I would prefer to (please check only w box)

2 Pay the required 0 Pay nothing and have 0 Pay the requiredcontribution and neither Pond A nor B contribution andhave Pond A monitored, have Pond Bmonitored, monitored.

MONITORING FORPOND A

Descrit)tion of Pond A:Q 72 acre Pond,● Good water clarity,. Located inland. Surround=’

wooded area, and. Monitoring will warn

of ~ problems.

Reauired Contribution:You would have to pay$15 to help findmonitoring for the year1997.

NEITHERPOND A NOR B

I am ~ willing to pay forwater quality monitoringfor either pond A or B.


MONITORING FORPOND B

Descrirltion of Pond B:. 5.5 acre Pond,. Poor water clarity,. Located near coast,. Surrounded by

HQi@’U&Development, and

. Monitoring will helpfind source of currentproblem.


gram and all who contributedto Pond A andiorPondB will have their money contribution refunded onTuesdav November 12.

Based on the choices made by individuals in allthree groups, it is possible for one, both, or neitherpond to be addedto the WatershedWatchprogram.9

The instructions for the hypothetical surveys par-alleled the above instructions, except we added thewording “if payments were collected today,” andwe used the words “did” and “would” instead of“do” and “will.” More details on the experimen-tal instructions are available from the authors uponrequest.

The minimum total contributions for each ex-

Figure 1. Example of Trichotomous Choice Question

perimental session were set based on pretest infor-mation and class size. A pretest of the surveyyielded an approximate $7.00 per person estimateof anticipated average WTP, estimated as the sumof required (individual) contributions times theproportion of pretest respondents who chose apond at that contribution level. However, given theNational Oceanic and Atmospheric Administra-tion’s (1993) suggestion that hypothetical CVMdollar values be divided by a factor of 2, we cali-brated our estimate of anticipated WTP to $3.50.10Next, we multiplied the (calibrated) anticipatedWTP for either pond by the anticipated number ofparticipants in each experimental session to derive

g The likelihood of cross-contamination between the experimental ses-sions was low. Session 2 began ten minutes ufter session 1. Session 3 wasrun several weeks later. Most of the subjects in sessions 1 and 2 werefreshmen und sophomores, while the subjects in session 3 were juniors,seniors, and graduate students. None of the students in sessions 1 rmd 2were enrolled in tbe class that represented session 3. Moreover, tbesubjects in session 3 were first asked if they had heard of the experiment.None said yes.

10Because of different class schedules, we had tbe Opportunity to

adjust our estimate of andcipated WTP ufter our initial day of experi-mentation. Based on the survey results from sessions 1 and 2, we ob-tained an approximate $2.94 and $2.96 per person estimate of (culi-bmted) arrticipated WTP for the hypothetical-money and real-moneysurveys, respectively. We used these latter estimates to determine theminimum total contributions for session 3.


the minimum totrd contributions for that session, 11This approach attempted to create a situationwhere the public provision of funds for water qual-ity monitoring depended on the marginal person.That is, if any one person actually valued monitor-ing at or above the required (individual) contribu-tion for a pond but chose not to contribute (i.e., tofree-ride), then the chance of raising the fundsneeded to establish a monitoring program at thepond would be low.

Incentive-Compatibility of OurContributions Mechanism

Standard economic theory predicts that individualswill not voluntarily contribute to the provision of apublic good because each individual has an incen-tive to free-ride on the contributions of others. De-spite this prediction, one still finds that individualsvoluntarily contribute to public goods, such ascharities, public television, and wildlife conserva-tion. Experimental tests of the ability of the vol-untary contributions mechanism (VCM) to providepublic goods appear mixed (for an overview, seeDavis and Holt [1993, section 6,3] and Ledyard[1995]): free-riding inhibits the ability of the VCMto provide public goods in some scenarios, while itdoes not inhibit the provision of public goods inother scenarios. Such mixed results call into ques-tion the incentive-compatibility of the VCM toeliminate free-riding and truthfully reveal the de-mand for public goods. Recently, research hasturned to the development of incentive structuresthat, when added to the VCM, will significantlyreduce the free-rider problem. The voluntary con-tributions mechanism used in this study incorpo-rates two structures that have been shown theoret-ically and empirically to reduce the incentive tofree-ride.

Specifically, our contributions mechanism con-tains a provision point (PP) and a money-backguarantee (MBG). The minimum total contributionrequired for each experimental session represents aconditional PP, whereby the provision of waterquality monitoring on a pond remains conditionalon all three groups reaching their predeterminedPPs. If the sum of the individual contributions fora pond in any group falls short of the group’s con-ditional PP, then anyone who contributed to that

‘1As mentioned earlier, Watershed Watch requires $500 per year tomonitor any water body (Green and Gold 1993). For the real-moneysurvey groups, the minimum totnl contributions required from each ex-perimental session totaled $255 per pond, Since these funds fell short of$500, we set aside $490 ( = $245 x 2) from our budget to cover tbeadditional cost of monitoring at each pond.

pond will receive a full refund. This money-backguarantee reduces any concern that too few contri-butions will be collected to permit provision of thegood at the specified level, in which case respon-dents would have incurred costs for an undeliver-able benefit.

Bagnoli and Lipman (1989) and Davis and Holt(1993, section 6.4) provide game-theoretic argu-ments that show that the use of a PP with an MBGreduces the incentive to free-ride. In fact, forgames of complete but imperfect information, free-riding is not a sensible strategy when a PP with anMBG exists (Bagnoli and McKee 1991). Isaac,Schmidtz, and Walker (1989) report experimentalevidence that adding an MBG to a PP significantlyincreases contribution rates. And Bagnoli and Mc-Kee (199 1) offer experimental evidence that sug-gests a PP with an MBG provides incentives forindividuals to successfully provide a public good.Rose et al. (1997) also offer evidence that suggestsa PP mechanism induces demand-revealing behav-ior for public goods.

Although our contributions mechanism parallelsthe basic nature of the mechanisms used in theabove studies, it is somewhat more complicated.Participants in our study face several provisionpoints: a group-specific PP for each pond and anaggregate PP for each pond, where the latter equalsthe sum of the group-specific provision points, Par-ticipants are fully informed about their group-specific PP and the number of groups participatingin the study, but they do not know the aggregate PPor the group-specific PP for other groups. Addi-tionally, participants are not told the size of theirgroup or other groups, although nothing preventsparticipants from taking a head-count of their owngroup, if so desired. Incomplete information aboutgroup size and the PP may not, however, hinder theeffectiveness of provision point mechanisms inproviding public goods (Rondeau, Schulze, andPoe 1997).12 However, since we do not use in-duced values and run treatments without a PP andan MBG, we cannot test the incentive compatibil-ity of our contributions mechanism. Here, we at-tempt to value a real, deliverable, environmental,public good, which precludes the use of inducedvalues. Rather than formally test our contributionsmechanism in an induced value framework. we

I 2 ~ondeau, fjchup~e, ~d Poe (1997) use a PP in conjuuctiOn with an

MBG and proportional rebate (PR) rale. Under the PR rule, all contri-butions in excess of the PP are divided among, and refunded to, indi-viduals in proportion to tbe amount of their contribution relative to tbetotnl contributions collected in their group. Using induced values in alaboratory experiment, the authors report high levels of aggregate de-marrd revelation regardless of incomplete information ahout the PP orgroup size,

Spencer, Swallow, and Miller Valuing Wafer Quality Monitoring 35

rely more on the findings of previous laboratoryexperiments to defend the incentive compatibilityof our mechanism.

Champ et al. (1997) argue that, without the useof an incentive-compatible contributions mecha-nism, actual voluntary donations to the provisionof a public (environmental) good represent a“lower bound” on the value of the public good.Given this argument in conjunction with the afore-mentioned findings that provision point mecha-nisms reduce free-riding behavior, we argue thatour contributions mechanism should, at least, elicitlower bound estimates of the value for water qual-ity monitoring on each pond in the real-money sur-vey treatment.

Logistics of the Real-Money Survey Treatment

In addition to a survey and envelope filled with tendollars, respondents in the real-money survey treat-ment received an identification card. The identifi-cation card contained an arbitrarily assigned iden-tification number, which also appeared on themoney envelope and the survey. We used the iden-tification number to keep track of monetary ex-changes while ensuring the anonymity of indi-vidual decisions. Respondents were instructed tokeep their identification cards and use the enve-lopes to make monetary exchanges.

For example, if a respondent chose to supportwater quality monitoring on one of the ponds, therespondent was instructed to pay the listed pro-gram cost by placing the exact amount of money inthe envelope. Payment could be made in cash,check (payable to Rhode Island Watershed Watch),or partially cash and IOU. The casMOU paymentoption was available only for program costs ex-ceeding the ten-dollar participation fee, The IOUentailed signing a promissory note that was pay-

13 For convenience, a self-able within one week.addressed stamped envelope was provided for pay-ing an IOU.

When all the respondents finished the survey,the experimental monitors collected all the moneyenvelopes and surveys; respondents who choseneither pond handed in an empty envelope. Next,the experimental monitors checked that each enve-lope contained the correct cash, check, or cashlIOU contribution required by the monitoring pro-gram chosen by each survey respondent. If a moni-toring program was chosen, but the envelope didnot contain the correct money contribution, then

13other “~e~ of tie IOU payment option can be found in HamisOn et

al. (1998), Loomis et al, (1996), and Neill et al. (1994).

the survey was excluded from the results. 14Thenthe monitors summed the valid contributions foreach monitoring program. If the sum of the indi-vidual contributions for any program equaled orexceeded the minimum aggregate contribution re-quired for that experimental session, the respon-dents were instructed to keep their identificationcards until a specified date. In this case, thepond(s) became a candidate for the WatershedWatch program, conditional on the contributionscollected from the other experimental sessions;thus, the respondents needed to retain their identi-fication cards in order to receive a future refund inthe event that the other experimental sessionsfailed to contribute enough money to the monitor-ing program. However, if the sum of the individualcontributions for any program was less than theminimum aggregate contribution required for thatexperimental session, then the respondents whocontributed to the program received a refund im-mediately. Refunds were handled by matchingidentification cards with envelopes. As each re-spondent left the room, he or she presented his orher identification card to an experimental monitor,who in turn gave the respondent an envelope thatmatched the number on the respondent’s identifi-cation card. In this case, respondents who had cho-sen not to contribute received an empty envelopein order to avoid signaling their private decisionsto fellow participants.

Results and Discussion

Contributions Collected for Each Pond

Table 2 summarizes the hypothetical and real con-tributions for each pond, Except for group 3 in thereal-money survey, a relatively small number ofrespondents chose neither pond. A comparison ofthe required group contribution versus the contri-butions collected suggests that, in several in-stances, the marginal person determines whetherthe group reaches its target. For example, in thehypothetical-money survey, group 2 fails to reachthe required fund for Pond B by $2. In the real-money survey, group 3 surpasses the required fundfor Pond B by only $4, but group 1 fails to raise therequired fund for Pond B by $5, and group 2 failsto raise the required fund for Pond A by $12. All of

14@I ~ 5 out of 76 respondents in the real-moneY surveY trea~ent

neglected to make the correct money contribution for their chosen pond,We exumined several approaches for including these respondents in ouranalysis, None of these approaches altered our basic conclusions below,although some numerical estimates changed slightly,


Table 2. Summary of Hypothetical and Real Contributions for Each Pond

Hypothetical-Money Survey Real-Money Survey

Group 1 Group 2 Group 3 Group 1 Group 2 Group 3

Required group contribution for each pond’ $98 $77 $53 $98Total contributions collected fop

Pond A $70 $50 $84 $40Pond B $149 $75 $72 $93

Number of subjects who chose neither pond

(l;%Y (:%) (1:%) (1:%)

Actual sample size 30 16 18 27Number of usable survevsd 30 16 17 26

$77 $80

$65 $42$44 $84

(1:%) (::%)22 2722 22

‘This number represents the minimum total contribution required from each group (i.e., experimental session), for each pond.Groups 1 and 2 represent the two undergraduate classes, and group 3 represents the interdisciplinary graduate-undergraduate class.The required contributions are based on anticipated group sizes of 28, 22, and 18 for groups 1, 2, and 3, respectively, in thehypothetical-money survey, and 28, 22, and 27 for groups 1, 2, and 3, respectively, in the real-money survey.bFor the hypothetical survey, these are staterthypotheticnl contributions.‘Numbers in parentheses represent the number of neither-pond choices as a percentage of usable surveys.‘Some surveys were excluded from the analysis because the respondent either chose both ponds or actually contributed aninsufficient a-mount of money and IOU.

these amounts fall within the range of individualcost variables included in the survey.

In total, the respondents in the hypothetical sur-vey raised $204 and $296 for water quality moni-toring on Ponds A and B, respectively. The respon-dents in the real-money survey actually contributeda total of $147 and $221, respectively, for Ponds Aand B. However, since the real-money respondentsfailed to reach the required group contribution inevery group, each respondent received a full refundand neither pond was added to the Rhode IslandWatershed Watch monitoring program.

Analysis of Individual Preferences

Table 3 describes all the variables used in our artaly-sis of individual preferences for water quality morti-toring. The SURROUND.A and SURROUND_B at-tribute variables represent, respectively, a choicebetween monitoring a pond in relatively natural(i.e., wooded) versus nonnatural (i.e., housingarea) surroundings (table 1), The PURPOSE_A andPURPOSE_B variables measure the effect of in-cluding information on the purpose of monitoring,with purposes to “warn of any problems” in PondA or to “help find the source of [a] current prob-lem” in Pond B (table 1). We developed no a priorihypothesis concerning which purpose might be im-portant to an individual. We did, however, hypoth-esize that the inclusion of information on the moni-toring purpose will have a positive effect on anindividual’s probability of choosing a pond overchoosing the neither-pond option.

The required contribution/cost variable is sepa-rated into HYPOTHETICAL COST, which equals

the respondent’s required payment for programschosen in the hypothetical-money surveys andwhich equals zero for programs chosen in the real-money surveys; and REAL COST, which equals therespondent’s required payment for programs cho-sen in the real-money surveys and which equalszero for programs chosen in the hypothetical-money surveys. HYPOTHETICAL COST andREAL COST are expected to have a negative effecton an individual’s probability of choosing a pond.Lastly, we consider several socioeconomic vari-ables, characterizing a respondent’s gender, aca-demic major, residence, and attitude toward socialresponsibility for water quality. We interact thesocioeconomic variables with the NEITHERPOND alternative. These interactions indicate howthe socioeconomic variables affect an individual’sdecision to choose a pond over choosing the nei-ther-pond option. Specifically, a negative coeffi-cient on any of the socioeconomic–NEITHERPOND interactions (table 3) indicates that a re-spondent with a higher value for that socioeco-nomic characteristic is more likely to choose tomonitor one of the ponds, while a positive coeffi-cient indicates a respondent is less likely to chooseto monitor a pond.

Table 4 reports the estimation results for fourspecifications of the multinominal logit model (7aand 7b). For each specification, the Xz-statistic fora likelihood ratio test of the model is significant atP <0,001. Specification 1 includes all the variablesfrom table 3, except the NEITHER POND con-stant. As expected, the coefficients on the costvariables are negative. The REAL COST variable ishighly significant, with P <0.001 for a one-tailed


Table 3. Description of Variables

Variable Description

Alternative specific constantsPOND APOND BNEITHER POND

Pond/program attribute variables’SURROUND_A

SURROUND_B

PURPOSE_A

PURPOSE_B

HYPOTHETICAL COST

REAL COST

Alternative specific socioeconomic variables’WQAITITUDE_N

GENDER_NECOLOGY_NRESIDENT_N

Dttmmy variable = 1 formonitoring on Pond A; = Otherwise.Dtrrrrmy variable = 1 formonitoring on Pond B; = Otherwise,Dtrmmy variable = 1 formonitoring on neither pond; = Otherwise.

Dtunmyvariable = 1 ifinformation onthesurroundings of Pond Misgiven;= Otherwise.

Dmnmyvariable = 1 ifinformation onthesurroundings of Pond B is given;= Otherwise.

Dummy variable = 1 if information onthepurpose ofmonitoring Pond A isgiven; = Otherwise.

Dummy variable = 1 if information on the purpose of monitoring Pond B isgiven; = O otherwise.

The required contribution/cost for an individual in the hypothetical-moneysurvey treatment to support monitoring on chosen pond.b

The required contribution/cost for an individual in the real-money surveytreatment to support monitoring on chosen pond.b

Discrete variable recorded as 2 = strongly agree, 1 = agree, O = neutral,–1 = disagree, and –2 = strongly disagree with the statement:“Everyone is responsible for water quality, even if they do not directlyuse a water body. ”

Dummy variable = 1 if female; = O if mafe.Dummy variable = 1 if ecology/environmental majord; = O otherwise.Dummv variable = 1 if a Rhode Island resident: = O otherwise.

‘In all the choice questions, the surrounding and/or monitoring information either appeared for both ponds or did not appear at all,bFor both the hypothetical-money and real-money survey treatments, groups 1 and 2 faced individual costs of $5, $10, or $15 forPond A, and $2, $7, or $12 for Pond B, Group 3 faced individual costs of $6, $11, or $14 for Pond A, and $4, $12, or $16 for

Pond B.

CThe socioeconomic variables are interacted with the NEITHER POND dummy variable.‘Ecology/environmental majors included animal science, aquiculture and fishery technology, biology, botany, environmentalmanagementiscience, natural resources science, plant science, soil and water resources, wildlife biology/management, and zoology,

test, but the HYPOTHETICAL COST variable isonly significant at P <0.18 for a one-tailed test.15Another expected result from specification 1 is thatthe coefficients on the monitoring variables arepositive, although a one-tailed test reveals that onlyPURPOSE_B is significant (P c 0,06). This sug-gests that respondents are more likely to supportvolunteer water quality monitoring when they aregiven information on the purpose of monitoring,especially in the case of Pond B. We return to thisissue below.

The SURROUND.A and SURROUND.B vari-ables are negative but insignificant (two-tailed P >0,45). The insignificance of these variables sug-

15 we ~t~o ~~dma~ed ~h~ ~odel with a nonlinear/quadraric cOst stmc-

ture (Boyle 1990). This functional form, however, made no significantimprovement (P > 0.70) in our model.

Additionrdly, one might suspect that the ten-dollar pmticipation pay-ment, given to each respondent, might produce a smatl income effect.That is, individuals might he more willing to pay for monitoring whenthe required contribution is less than or equal to ten dollars. However, byconducting a number of attemative nested and nonnested tests, we findno evidence for such an income effect. All tests were insignificant (P >0.20).

gests that the respondents value volunteer waterquality monitoring, but they remain indifferent be-tween monitoring ponds in a relatively natural ver-sus a nonnatural setting. The alternative specificconstants, POND A and POND B, also are negativebut insignificant (two-tailed P > 0.25). These nega-tive constants would (if significant) tend to suggestthat monitoring on either pond is not preferred toleaving both ponds unmonitored; however, a pref-erence for monitoring at either pond depends onthe average effect of other variables that alter theconstant/intercept terms, and we see below thatrespondents prefer monitoring at least one pond(compare specifications 1 and 3, table 4).

As for the socioeconomic–NEITHER POND in-teraction variables, all are negative and, except forRESIDENT_N, all remain influential in the model,although GENDER_N is significant at only the11% level (table 4). The negative coefficient onWQA7TITUDE_N is expected, since one wouldexpect people who agree that everyone is respon-sible for water quality to be more likely to supportvolunteer water quality monitoring. Similarly, the


Table 4. Estimation Results for Various Specifications of the Multinominal Logit Model

Specification 1 Specification 2 Specification 3 Specification 4

Parameter Pr>lZl Parameter Pr>lZl Parameter PD-IZI Parameter PrAZlVariable Estimate (P-vahse~ Estimate (P-value) Estimate (P-value) Estimate (P-value)

POND A

POND B

NEITHER POND

HYPOTHETICAL COST

REAL COST

SURROUND_A

SURROUND_B

PURPOSE_A

PURPOSE_B

WQA7T[TUDE_N

GENDER_N

ECOLOGY_N

RESIDENT_N

-1.223(-1.117)-1.186

(-1.135)

-0.037(-0.936)-0.177

(-4.087)-0.422

(-0.610)-0.432

(-0.678)0.111

(0.1 88)0.869

(1.557)-1,129

(-2.515)-0.833

(-1.608)-1.064

(-2,047)-0.237

(-0.459)

0,264

0.256

0.350

<0.001

0.542

0.498

0,851

0.119

0.012

0,108

0.041

0.646

1,309(1.458)

-0.039(-1.005)-0,182

(-4,217)

0.085(0.156)0.871

(1.678)-1,097

(-2.490)-0.766

(-1.519)-1.088

(-2.144)

0.145

0.315

<0.001

0.876

0.093

0.013

0,129

0.032

-1.445 0.002 1.266 0.137(-3.050) (1.487)-0.043 0.249 -0.039 0.316

(-1.153) (-1.003)-0,187 <0,001 -0,182 <0.001

(-4,518) (-4.244)

0,097 0.845(0,196)0.877 0.063 0,801 0.002

(1.860) (3.163)-1.106 0.011

(-2.531)

-0.761 0.130(-1.514)–1,079 0.032

(-2,145)

Log-likelihood –1 14.299 -114.618 –121.676 –1 14.630Xz-statistic 63.632 <0.oolb 62.995 <0.001 48.880 <0.001 62.971 <0.001

(d.f = 12) (d$ = 8) (d.f = 5) (d.$ = 7)

NOTE:Numbers in parentheses represent Z-statistics.‘The P-values reported in this table correspond to a two-tailed test of Ho: ~ = O versus H.: ~ # O.‘The level of significance (i ,e., P-value) for the Xz-statistic.

negative coefficient on ECOLOGY_N is expectedbecause student-respondents in ecology/environ-mental majors have expressed a preference to ad-dress environmental problems as professionals,and this attitude may correlate with a positive will-ingness to support pond monitoring. This result isconsistent with Kuitunen and Tormala’s (1994)finding that students with more knowledge aboutnature and conservation issues are more willing tosupport endangered species conservation.

The negative coefficient on GENDER_N sug-gests that the female respondents are more likelythan the male respondents to support volunteer wa-ter quality monitoring. Similar differences in pref-erences between genders is noted in other studies(Kuitunen and Tormala 1994; Swallow et al. 1994;Day and Devlin 1996).

Finally, the insignificance of RESIDENT_N (P >0.60, specification 1, table 4) indicates that RhodeIsland residents were not significantly more likelythan non–Rhode Island residents to support volun-teer monitoring in their own region of residence.

Given the statistical insignificance of the ‘‘sur-round” variables and RESIDENT_N, we reesti-matedthe multinominallogit model (7a and 7b) with-out these variables in specification 2 (table 4). Notethat we use the NEITHER POND constant, insteadof the POND A and POND B constants jointly.This specification more directly measures an indi-vidual’s preference for a pond over the neither op-tion.16 Estimation of specification 2 results in nosubstantive changes in the sign, magnitude, andsignificance of the remaining variables. A likeli-hood ratio test between specifications 1 and 2 re-veals no statistically significant difference (X2 =0.638,4 d.f, P > 0.95). Thus, we conclude that therespondents value volunteer monitoring regardless

16~he ~~eofNEITHER POND, instead of POND A and pOND B

jointfy, imposes the restriction that respondents view these ponds asequivalent, a priori, so that pond acreage, clarity, and location differences(table 1) did not significantly affect respondents’ choices. The restrictionis not significant (P > 0,90).


of either pond’s surroundings and the respondent’sstate of residence.

Next, realizing that the socioeconomic–NEITHER POND interaction terms may alter theNEITHER POND constant term, we reestimatedspecification 2 without the socioeconomic-NEITHER POND interactions, using specification3 (table 4). No major change in the cost and moni-toring variables occurs, but the NEITHER PONDconstant becomes negative and statistically signifi-cant (P < 0.003). This suggests that the respon-dents prefer to monitor one of the ponds over leav-ing both ponds unmonitored. A likelihood ratio testbetween specifications 2 and 3, however, revealsthe restriction is statistically significant (X2 =14.116, 3 d.j, P e 0.005), This implies that differ-ent socioeconomic groups maintain different pref-erences for volunteer pond monitoring.

Subsequently, we include the socioeconomic–NEITHER POND interaction terms in our final,fourth specification. 17Specification 4 differs fromspecification 2 by excluding the PURPOSE_Avariable. Given that PURPOSE_A remains insig-nificant in specifications 1, 2, and 3 (P > 0.80), itdoes not appear influential in the respondents’ de-cisions over volunteer pond monitoring; thus, wedrop PURPOSE_A from our final analysis, A com-parison of specifications 4 and 2 reveals no majordifferences in the sign, magnitude, and signifi-cance of the parameter estimates, except that thesignificance of PURPOSE_B improves from a one-tailed P <0.05 to P <0.001 in specification 4.

The insignificance of PURPOSE_A but signifi-cance of PURPOSE_B suggests a “catch the pol-luter(s)” preference among the respondents. Re-call, the purposes of monitoring Ponds A and Bare, respectively, “to warn of any problems” and‘‘to help find source of [a] current problem”;hence, the respondents appear willing to tradeofftrying to detect any unknown and uncertain prob-lem(s) for finding the source of a known and cer-tain problem, Several explanations for this resultmay exist. First, the respondents may view the ben-efits of monitoring Pond A as less certain than thebenefits of monitoring Pond B, A preference formore certain benefits over less certain benefits re-mains consistent with Kahneman and Tversky’s(1979) “prospect theory” and with Macmillan,Hanley, and Buckland’s (1996) finding that people

IT For our founh specification, we used the Hausman-McFadden test

(Hausman and McFadden 1984; Green 1995, pp. 500-501) to test forviolations of the independence of irrelevant alternatives (11A) as sump.tion. The resulting chi-square of 0.66 (P > 0.85) indicates that the 11Aassumption is satisfied and suggests that the current specification repre-sents the respondents’ decision-making process,

prefer environmental projects with certain gainsover projects with uncertain gains. Second, the re-spondents may deem a current environmentalproblem more important than monitoring to estab-lish a baseline from which to detect possible futureproblems. Indeed, these results suggest that, in theabsence of more information, respondents assumedmonitoring was intended only to detect possiblefuture problems. While respondents value monitor-ing services, in general, their value for monitoringa particular pond increased when monitoring couldidentify a current pollution source.

Estimated WTP for Volunteer WaterQuality Monitoring

Following Hanemann (1984), we can obtain a util-ity-theoretic welfare measure, or willingness-to-psy, for volunteer pond monitoring. A respon-dent’s maximum willingness-to-pay, WTP, is cal-culated as the program cost, C, that will make theindividual indifferent between monitoring the pondchosen and the status quo (i.e., the neither pondalternative), which has zero cost. The indifferencebetween choosing Pond k and the neither pond op-tion can be represented symbolically as follows:

(8) U(XN>Si)Jfi) = U(xk,Si, ~i - w’Tpk)for all k # N,

where N denotes the neither pond option, whichhas zero cost; k denotes the chosen pond; WTPkequals the maximum willingness-to-pay for moni-toring on Pond k; and X, S, and M areas defined inequation (1). Next, using our empirical utility func-tion, as in equation (1) with (7a, 7b)–(8), we cansolve for WTPk as follows:

(9)P’

‘Tpk=‘E ‘Zik-‘iN)’Estimates of WTP for Ponds A and B can be ob-tained by using the parameter estimates in table 4.

Table 5 reports WTP estimates for the averagerespondent based on equation (9) and the param-eter estimates for specification 4 (table 4). ForPonds A and B, hypothetical-money WTP exceedsreal-money WTP by a factor of 4.67.’8 This factorlies well within the range of factors reported acrossthe few other discrete choice studies that comparehypothetical-money and real-money WTP for apublic good (Foster, Bateman, and Harley 1997);

18Beca”se of the linearity in specification 4 (table 4), the ratiO Of

hypothetical-money WTP to reaf-money WTP (for Ponds A and B) re-duces to a ratio of the REAL COST coefficient (–O.182) to the ffYPO-7“ETICAL C(XW coefficient (-0,039).


Table 5. Willingness-to-Pay Estimates for theAverage Respondent

Hvuothetical WTP Real-Money WTP

Pond A $42.69 $9.15*

($38.24) ($1.79)Pond B $63.23 $13.55*

($58.67) ($2.42)Difference(WTP, - WTPJ’ $20.54 $4,40**

($21.78) ($1.76)

NOTE:The WTPestimates reported here are based on the pa-rameter estimates to specification 4 in table 4 and the averagerespondent, identified by the sample mean of the variablesWQATTITUDE, GENDER, and ECOLOGY, which are, respec-tively, 1.6692, 0,4662, and 0.6767. The numbers in parenthesesrepresent standard errors for the WTP estimates.‘Given the definition of WTP in equation (9), the difference(WTP, - WTPJ can be written as (-~’/&-) (Zw - Zti).*Significant at P-value <0.001 for a one-tailed test of Ho: WTP~= O versus HA: WTP~ >0.**Significantatp-value <0.01 for a one-tailed test of Ho: WTf’E

= WTPA versus HA: WTP~ > WTPA.

however, it exceeds the National Oceanic and At-mospheric Administration’s (1993) suggested cor-rection factor of 2. This suggests that the averagerespondent in our sample may have overstated his/her WTP for volunteer monitoring in the hypo-thetical survey. However, because of the high stan-dard errors on the hypothetical-money WTP esti-mates, we cannot conclude that these WTPestimates are statistically greater than the real-money WTP estimates. 19

Next, given specification 4, a comparison of theWTP estimates for Ponds A and B reveals howmuch the average respondent is willing to pay tohelp identify a current pollution source in Pond B,The average respondent appears willing to pay anextra $20,54 hypothetical dollars, or $4,40 real dol-lars, to help find the source of the water clarityproblem in Pond B. The real-money difference isstatistically significant (P t 0,01), while the hypo-thetical-money difference is not statistically sig-nificant (P > 0. 15), These results suggest, espe-cially in the real-money surveys, that respondents

19 ~ch s~and~d error in table 5 is based on an estimate of the vtiance

of WTPk Var(WZ’Pk). Here, we followed Kmenta’s (1971, p. 444) ap-proximation formula based on a Taylor series expansion, as follows:

‘ar(wp’w2”vu(D~’+(ia2”var@c)‘2”(iNi?)c0v(D~c)

where D~ = -~ ‘(Z~k- Zw) and Cov(Dti &) equals the covariarrce be-tween Dk and PC Note that the elements of Var(WTPk) account for the samp-le size, through the vasiance-covariance matrix of estimators; in this case,- is the standard error of WTP,

value monitoring at Ponds A and B equally untilthey are informed that monitoring on Pond B willhelp identify a current source of pollution. Thisinformation motivated respondents to raise theirWTP by about 48%. Without this information, thedifferences between Ponds A and B (table 1), in-cluding the size, water clarity, and location differ-ences, left respondents statistically indifferent be-tween the ponds.

Concluding Remarks

While most economic studies of water quality fo-cus on estimating the benefits of water quality im-provements, this paper initiates study of the pref-erences and willingness-to-pay of individuals forvolunteer water quality monitoring programs. Thepaper uses both hypothetical and real payment con-tingent valuation formats to directly measure indi-vidual preferences and willingness-to-pay for vol-unteer ‘monitoring at two ponds in Rhode Island.An analysis of the contingent valuation survey re-sponses suggests that our sample of respondentsvalues pond monitoring regardless of whether apond is surrounded by wooded area or housingdevelopment, and that a respondent’s state of resi-dence plays no significant role in determining therespondent’s probability of choosing to supportmonitoring on one of the Rhode Island ponds.

Factors that appear to influence a respondent’sdecision to support volunteer water quality moni-toring include (1) the stated purpose of monitoring,(2) the individual cost of monitoring, (3) the re-spondent’s attitude toward social responsibility forwater quality, (4) the respondent’s gender, and (5)the respondent’s academic major. The respondentsappear willing to tradeoff monitoring “to warn ofany problems” (as in Pond A) in exchange formonitoring “to help find [the] source of currentproblem” (as in Pond B). In terms of WTP, theaverage respondent values the two ponds equallywhen information on the purpose of monitoring isnot given; however, the average respondent is ac-tually willing to pay 4890 more in real money tomonitor a pond for the purpose of helping identifythe source of current water quality problems. Thismay reflect a preference for more tangible/certainenvironmental gains, an ethical preference to catchthe polluter(s), or a higher concern for ponds withCtItTHU moblems.

In te~s of the socioeconomic characteristics ofthe student-respondents, those who believe every-one is responsible for water quality, females, andecology/environmental majors are more likely tosupport pond monitoring, The ecology/environ-

Spencer, Swallow, and Miller Valuing Waler Quality Monitoring 41

mental major effect suggests that monitoring pro-grams might obtain more public support if envi-ronmental agencies better educate the public aboutthe importance of water quality monitoring.

Overall, water quality monitoring is a key ele-ment in efforts to protect water resources, andstates are finding that volunteer monitoring pro-grams are a cost-effective way to obtain credibleinformation on water quality, In Rhode Island, vol-unteer water quality monitoring data aids environ-mental managers to set priorities for various man-agement or enforcement actions for watershedsand water bodies. Supplementing monitoring datawith information on public preferences and WTPcould help environmental managers identify thewater body types that both the public and waterresource specialists consider important and worthyof investment.

To accomplish this task, future research mightinvestigate a two-stage process for setting priori-ties for monitoring and management actions forwater bodies. The first stage would involve waterresource specialists using monitoring data andother scientific information to identify environ-mentally important water body types. In the secondstage, environmental economists could use thefirst-stage results to build a model of public pref-erences for water quality monitoring on variouswater body types. Based on information from bothsteps, environmental managers could identify a listof water bodies that better satisfies the preferencesof both water resource specialists and the public.

Lastly, environmental economists may gain in-sights from this case application comparing hypo-thetical-money and real-money contingent valua-tion surveys. For our sample of respondents, hy-pothetical-money costs are less influential thanreal-money costs in determining a respondent’sprobability of choosing to support monitoring onone of the ponds. The present study used a split-sample design wherein respondents in each samplereceived either a hypothetical-money or a real-money survey, but the actual differences betweenthe surveys were quite minimal. This design im-plied that the hypothetical survey contained exten-sive details concerning how money would havebeen collected and, if necessary, refunded if realmoney had actually been solicited from the respon-dents. Further research might involve comparingreal-money surveys with two versions of the hy-pothetical survey, wherein the additional hypo-thetical survey follows a more traditional, andbriefer, format for the payment vehicle, such aspayments of hypothetical new taxes. Without suchadditional research, and especially in light of ourstatistical tests, we believe it is premature to con-

clude that hypothetical and real contingent vahta-tion estimates of WTP necessarily will differ bythe factor of four obtained here.

References

Addelman, S., and O. Kempthorne. 1961. Orthogonal Main-

.l&ect Plans. ARL Technical Report 79, Aeronautical Re-

search Laboratory, Office of Aerospace Research, United

States Air Force, Wright-Patterson Air Force Base, Ohio,

Bagnoli, M., and B.L. Lipman. 1989, “Provision of Public

Goods: Fully Implementing the Core through Private Con-

tributions.” Review of Economic Studies 56(4):583-601,

Bagnoli, M., and M. McKee. 1991. “Voluntary Contribution

Games: Efficient Private Provision of Public Goods.”

Economic Inquiry 29(2):35 1-66.

B1ackbum, M., G.W. Harrison, and E.E, Rutstrom. 1994. ‘‘Sta-

tistical Bias Functions and Informative Hypothetical Sur-veys, ” American Journal of Agricultural Economics

76(5):1084-88.

Bockstael, N. E., K.E, McConnell, and I.E. Strand. 1989. “Mea-

suring the Benefits of Improvements in Water Quality: The

Chesapeake Bay.” Marine Resource Economics 6(1 ): 1-

18.

Boyle, K.J. 1990. “Dichotomous-Choice, Contingent-

Valuation Questions: Functional Form Is Important. ”

Northeastern Journal of Agricultural and Resource Eco-

nomics 19(2): 125–3 1,

Brown, T. C., P.A. Champ, R.C. Bishop, and D.W. McCoUum.

1996. “Which Response Format Reveals the Truth about

Donations to a Public Good?” Land Economics 72(2):

152-66.

Champ, P. A., R.C. Bishop, T.C. Brown, and D.W. McCoUum.

1997. “Using Donation Mechanisms to Value Nonuse

Benefits from Public Goods.” Journal of .Envircmmental

Economics and Management 33(2): 151–62.

Cummings, R. G., S. Elliott, G.W, Harrison, and J, Murphy.

1997. “Are Hypothetical Referenda Incentive Compat-

ible?” Journal of Political Economy 105(3):609–21.

Cummings, R. G,, G,W. Harrison, and E.E. Rutstrom. 1995.

“Homegrown Values and Hypothetical Surveys: Is the Di-

chotomous Choice Approach Incentive-Compatible?”

American Economic Review 85(1):260–66.

Cummings, R.G, and L.O. Taylor. 1997. “Unbiased Value Es-timates for Environmental Goods: A Cheap Talk Design

for the Contingent Valuation Method.” Unpublished ins.,Department of Economics, School of Policy Studies, Geor-

gia State University. August.

Davis, D. D., and C.A, HoIt. 1993. Experimental Economics.

Princeton: Princeton University Press.

Day, K,M,, and R,A. Devlin. 1996. ‘‘Volunteerism and Crowd-

ing Out: Canadian Econometric Evidence. ” Canadian

Journal of Economics 29(1):37–53.

Fisher, A., W.J. Wheeler, and R. Zwick, 1993. “Experimental

Methods in Agricultural and Resource Economics: How

Useful Are They?” Agricultural and Resource Economics

Review 22(2): 103–16.

Foster, V., I.J. Bateman, and D. Harley, 1997. “Real and Hy-

pothetical Willingness to Pay for Environmental Preserva-


tion: A Non-Experimental Comparison. ” Journal ofAgri-

cultural Economics 48(2): 123–38.

Fox, J.A., J,F. Shogren, D.J. Hayes, and J.B. Kliebenstein.

1995, ‘‘CVM-X: Calibrating Contingent Values with Ex-

perimental Auction Markets.” Kansas Agricultural Ex-

periment Station Contribution no. 95-470-J. Manhattan:

Kansas State University. November.

Freeman, A.M. 111.1990. “Water Pollution Policy.” In Public

Policies for Environmental Protection, ed. P.R. Portney,

pp. 97–149. Washington, D. C.: Resources for the Future.

1995. <‘The Benefits of Water Quality Improvements

for Marine Recreation: A Review of the Empirical Evi-

dence.” Marine Resource Economics 1O(4):385AO6.

Green, L.T., and A.J. Gold. 1993. “Rhode Island Watershed

Watch.” Fact Sheet no. 89-2. Cooperative Extension and

Department of Natural Resources Science, University ofRhode Island.

Greene, W. II. 1995. LIMDEP: Version 7.0 User’s Manual.

Bellport, N. Y,: Econometric Software, Inc.

Hanemann, W.M. 1984. “Welfare Evaluations in Contingent

Valuation Experiments with Discrete Responses.” Ameri-

can Journal of Agricultural Economics 66(3):332-41.

. 1994. ‘‘Vafuing the Environment through Contingent

Valuation.” Journal of Economic Perspectives 8(4):19-

43.

Harrison, G.W., R.L, Beekman, L.B. Brown, L.A. Clements,

T.M. McDaniel, S.L. Odom, and MB. Williams. 1998.

“Environmental Damage Assessment with Hypothetical

Surveys: The Calibration Approach.” In Topics in Envi-

ronmental Economics, ed. M. Boman, R, Brannlund, and

B. KristrtJm. Amsterdam: Kluwer Academic Press (forth-

coming).

Hausman, J., and D. McFadden. 1984. “Specification Tests for

the Multinominal Logit Model. ” Ecorsometrica 52(5):

1219-40.

Herron, E.M., and L.T. Green, 1996. “University of Rhode

Island Watershed Watch: 1995.” Natural Resources Sci-

ence Technical Report 96-3. Cooperative Extension andDepartment of Natural Resources Science, University of

Rhode Island.

Isaac, R. M., D. Schmidtz, and J.M. Walker. 1989. “The As-

surance Problem in a Laboratory Market.” Public Choice

62(3):217-36.

Kahneman, D., and A. Tversky. 1979. “Prospect Theory: An

Arrafysis of Decision Under Risk.” Ecorzometrica 47(2):

263-91.

Keeney, T.R.E. 1996. “Volunteers Needed to Monitor Ponds.”

Providence Journal-Bulletin, April 13.

Krnenta, J. 1971. Elements of Econometrics. New York: Mac-

millan Publishhrg Co.

Kuitunen, M., and T. Torm51a. 1994. r‘Willingness of Students

to Favour the Protection of Endangered Species in a Trade-

off Conflict in Finland.” Journal of Environmental Man-

agement 42(2): 11 1–18.

Ledyard, J.O. 1995. “Public Goods: A Survey of ExperimentalResearch. ” In The Handbook of Experimental Economics,

ed. J.H. Kagel and A.E. Roth, pp. 111–94. Princeton:

Princeton University Press.

Loomis, J., T. Brown, B. Lucero, and G. Peterson. 1996. ‘‘Im-

proving Validity Experiments of Contingent Valuation

Methods: Results of Efforts to Reduce the Disparity of

Hypotheticsd and Actual Willingness to Pay.” Land Eco-

nomics 72(4) :45LM1.

Macmillan, D., N. Hanley, and S. Buckland. 1996. “A Contin-

gent Valuation Study of Uncertain Environmental Gains.”

Scottish Journal of Political Economy 43(5):519-33.

McFadden, D. 1974. r‘Conditional Logit Analysis of Qualita-tive Choice Behavior,” In Frontiers in Econometrics, ed.

P. Zarembka, pp. 105-42. New York: Academic Press.

National Oceanic and Atmospheric Administration. 1993. ‘‘Re-

port of the NOAA Panel on Contingent Valuation.” Fed-

eral Register 58(10):4601–14.

Needelman, M. S., and M.J. Kealy. 1995. “Recreational Swim-

ming Benefits of New Hampshire Lake Water Quality

Policies: An Application of a Repeated Discrete Choice

Model.” Agricultural and Resource Economics Review

24(1):78-87.

Neill, H. R., R.G. Cummings, PT. Ganderton, G.W. Harrison,

and T. McGuckirr. 1994. “Hypothetical Surveys and Real

Economic Commitments.” Land Economics 70(2): 145–

54.

Rondeau, D., G.L. Poe, and W.D. Schtdze. 1996. “Developing

a Demand Revealing Market Criterion for ContingentVahration Validity Tests.” Working Paper 9616. Depart-

ment of Agricultural, Resource, and Managerial Econom-

ics, Cornell University. November.

Rondeau, D., W.D. Schnlze, and G.L. Poe. 1997. “Voluntary

Revelation of the Demand for Public Goods Using a Pro-

vision Point Mechanism. ” Working Paper 97–20. Depart-

ment of Agricultural, Resource, and Managerial Econom-

ics, Cornell University. October.

Rose, S. K., J. Clark, G.L. Poe, D. Rondeau, and W.D. Schulze.

1997, “The Private Provision of Public Goods: Tests of a

Provision Point Mechanism for Funding Green Power Pro-

grams.” Workkrg Paper 97-09. Department of Agricul-

tural, Resource, and Managerial Economics, Cornell Uni-

versity. July.

Seip, K., and J. Strand. 1992. “Willingness to Pay for Envi-

ronmental Goods in Norway: A Contingent Valuation

Study with ReaJ Payment.” Environmental and Resource

Economics 2(1):91-106.

Shogren, J.F. 1993. “Experimental Markets and Environmental

Policy.” Agricultural and Resource Economics Review

22(2): 117-29.

Simpson, J.T. 1991. Volunteer Luke Monitoring: A Methods

Manual, EPA 440/4-9 1-002. Office of Water. Washington,

D.C.

Swallow, S.K., T. Weaver, J.J. Opaluch, and T.S. Michelmarr.

1994. “Heterogeneous Preferences and Aggregation in

Environmental Policy Analysis: A Landfill Siting Case.”

American Journal of Agricultural Economics 76(3):43 1-

43.

U.S. Environmental Protection Agency. 1990. Volunteer Water

Monitoring: A Guide for State Managers. EPA 44014-90-010, Office of Water. Washington, DC.

—. 1995. Liquid Assets: A Summertime Perspective on the

Importance of Clean Water to the Nation’s Economy. EPA

800-R-96-002. Office of Water. Washington, D.C.

Valuing Water Quality Monitoring: A Contingent Valuation...

Documents

Transcript of Valuing Water Quality Monitoring: A Contingent Valuation...