Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Valuing Environmental Quality Changes Using Averting Expenditures: An Applica...Abdalla, Charles W.; Roach, Brian A.; Epp, Donald J.Land Economics; May 1992; 68, 2; ABI/INFORM Globalpg. 163

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

164 Land Economics May 1992

the level of personal environmental qualityconstant, are equal to the benefits of a mar­ginal pollution reduction. In reality, indi­viduals will alter their personal environ­mental quality level as pollution is reduced.Ifpersonal environmental quality decreaseswith increases in pollution and pollutiondoes not directly enter into the utility func­tion, averting expenditures are a lowerbound to willingness to pay. When pollu­tion enters into the utility function directly,Courant and Porter stated that it is moredifficult to determine if averting expendi­tures bound willingness to pay.

Watson and Jaksch (1982) and Harford(1984), in studying the effect of air pollutionon personal or household cleanliness, de­veloped theoretical models considering the"price" of a cleaning episode as a functionof pollution and cleaning frequency. Theirmodels indicated the need for empirical re­sults to test assumptions of cleanliness andaverting expenditure behavior. Harringtonand Portney (1987) reported that the sum ofchanges in averting expenditures and costsof illnesses is likely to be a lower boundestimate of willingness to pay (WTP) , as­suming individuals do not increase avertingexpenditures in response to a decrease inpollution. Also, they conclude that the pos­sibility exists in principle to observe exactbenefits but obtaining the necessary infor­mation regarding market and household be­havior normally prohibits the analysis.

Bartik (1988) also stated that theoreti­cally correct measures of WTP could be es­timated using averting expenditures if infor­mation was obtainable concerning thehousehold's production technology. In theabsence of such information, upper andlower bounds to WTP could be obtainedfrom averting expenditure information formarginal and nonmarginal changes in pollu­tion. Rather than defining the cost of acleaning episode, Bartik's model utilizesthe defensive expenditure function basedon the household's choice of personal envi­ronmental quality. Bartik noted that the ca­pability of averting expenditure estimatesto bound WTP is dependent upon the valid­ity of several assumptions. In particular,averting inputs should not exhibit jointnessin the production of household outputs and

averting expenditures should not involvesunk costs in the purchase of durablegoods. If an averting good violates thejointness assumption, expenditures on thegood are theoretically divided among theproduction of each commodity involved.Violation of the sunk cost assumption alsoimplies that the full purchase price of thedurable good cannot be attributed to thechange in water quality.

Roach (1990) demonstrated that includ­ing household time as an input into the pro­duction function for personal environmen­tal quality did not change Bartik's (1988)conclusions about the ability of averting ex­penditure estimates to bound willingness topay. Also, graphical analysis indicated thataverting expenditures will present a lowerbound to compensating and equivalent vari­ation (CV and EV, respectively) based onreasonable assumptions. ShortIe and Roach(1989) extended the graphical analysis todemonstrate that averting expenditures arelikely to be a lower bound to CV and EVeven if pollution enters into the utility func­tion directly. However, the bounds in thiscase will be less exact than when pollutiondoes not enter into the utility function di­rectly.

Averting expenditure estimates appearto provide a conceptually valid conserva­tive estimate of actual costs or benefits ofchanges in drinking water quality. How­ever, averting measures may be difficult todefine for different types of pollution. Forexample, Watson and Jaksch (1982) in­cluded house painting and window washingas averting measures related to air pollu­tion. These activities are likely to violatethe nonjointness assumption since theymay also be inputs in the production of"household appearance." Averting mea­sures related to drinking water are easilydefined, such as buying bottled water or ahome water filtration system.

Some researchers including avoidancecosts in their studies have simulated avert­ing behaviors and expenditures based onthe assumption that households did in factengage in such behaviors in response topollution. For example, Spofford, Krup­nick, and Wood (1989) used prespecifiedprobability distributions to model house-

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

68(2) Abdalla et al.: Groundwater Contamination 165

holds' averting expenditures in a study ofgroundwater remediation efforts at a fed­eral Superfund site in Massachusetts. Oth­ers, including Smith and DesVousges(1986), Harrington, Krupnick, and Spofford(1989), and Abdalla (1989), have providedempirical evidence of household avertingexpenditures in response to water qualitydegradation.

In this study, averting expenditures wereused to approximate the economic costs tohouseholds in a southeastern Pennsylvaniacommunity affected by groundwater con­tamination. The estimates obtained are in­terpreted given the underlying theory andimplications for policy are discussed.

II. PROCEDURES

Criteria for selecting a community expe­riencing groundwater contamination in­cluded a minimum of 500 households con­nected to a public community watersystem, an expectation that the contamina­tion incident would continue during the en­tire study period, and public notification ofcontamination. The borough of Perkasie insoutheastern Pennsylvania, which has anestimated 2,760 households (population7,877), was selected as an example of acommunity affected by groundwater con­tamination. In late 1987, Trichloroethylene(TCE), a volatile synthetic organic chemi­cal, was detected in one of the borough'swells. TCE levels were as high as 35 partsper billion (ppb) , exceeding the Environ­mental Protection Agency's maximum con­taminant level (MCL) of 5 ppb. Since notemporary solution was available to reduceTCE levels below the MCL, the countyhealth department required the borough tonotify customers of the contamination inJune 1988. As of December 1989, no solu­tion had been implemented.

Mail questionnaires were used to elicitinformation about increases in householdaverting expenditures taken in response tocontamination in Perkasie. Householdswere asked about specific actions they tookto avoid exposure to TCE. This was an at­tempt to address the empirical estimationproblem of including expenditures for unre­lated reasons, such as changes in tastes and

preferences. Since many of the behaviorsfor avoiding a water contaminant require aperson's time as well as purchased inputs,the amount of time required for averting ac­tions and cash expenditures on averting in­puts were sought. The time input was eval­uated alternately at the minimum wage andat the estimated wage of the respondent.Information was collected about possiblefactors influencing averting expenditurechanges, such as health perceptions, atti­tudes, and demographic factors. Also, re­spondents were asked their quantitativeand qualitative cancer risk perceptions as­sociated with the contamination incident.The survey instrument was developed andadministrated following procedures sug­gested by Dillman (1978).

Based on Kalton (1983), a random sam­ple of 1,733 was determined to be neededfor representative results. A mailing list ofresidential customers was obtained and thesurvey was administered in September1989. After three follow-up mailings, 761usable questionnaires were received. Ad­justing for nondeliverable surveys, the ef­fective response rate was 46.9 percent. Theresponse rate was lower than anticipatedand it was felt necessary to investigate non­response bias. A telephone interview witha random sample of 50 nonrespondentswas conducted to determine if respondentsand nonrespondents were similar in keyattributes. Two variables from the mailquestionnaire sample and the telephoneinterview sample, awareness of TCE con­tamination and whether household avertingactions were taken in response to TCE,were statistically compared following amethod suggested by Snedecor and Coch­ran (1980). The hypothesis that the twosamples were different on these attributeswas rejected at the .05 level. Consequently,the mail survey results were concluded tobe representative of the Perkasie popu­lation.

III. AVERTING EXPENDITURES

Only 43.2 percent of Perkasie respon­dents were aware of TCE in their water de­spite mandatory notification of the contami­nation. Of these, 133, or 43.75 percent,

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

166 Land Economics

TABLE 1ESTIMATED COSTS DUE TO TCE CONTAMINATION INCURRED BY PERKASIE

HOUSEHOLDS FROM DECEMBER 1987 TO SEPTEMBER 1989

May 1992

Category of Cost

I. Increased purchases of bottled water"2. New purchases of bottled water3. Home water treatment systemsd

4. Hauling water5. Boiling wate~

Total

Low Estimate ($)a

11,134.5417,341.954,691.46

12,512.7615,632.58

61,313.29

High Estimate ($)b

11,134.5417,341.954,691.46

34,031.4864,134.63

131,334.06

'Low estimate values lost leisure time at minimum wage ($3.35 per hour).bHigh estimate values lost leisure time at the estimated hourly wage. To estimate hourly wage.

the median value of the income category checked by each respondent was used as an estimate ofyearly income which was divided by 2.080 hours.

'An average bottled water cost of $0.83 per gallon was obtained by surveying retail grocers in thePerkasie area.

dThe useful life of a home water treatment system was assumed to be ten years (520 weeks). Sincethe 88-week study period represented 0.17 of a ten-year period. only 0.17 of the purchase price ofeach water treatment system was included in the estimate.

'Information was asked in the survey regarding the number of trips per week to haul water.round-trip distance. and how often the trip was for the sole purpose of hauling water. Averagetravel speed was assumed to be 35 miles per hour and the American Motor Vehicle ManufacturersAssociation indicates an average vehicle operation cost of 32.6 cents per mile.

'Respondents were assumed to use an electric stove for eight minutes to boil one-half gallon at atime. The electricity cost in Perkasie was 5.5 cents per kilowatt-hour.

undertook specific actions to avoid expo­sure after learning of TCE in their water.The actions included: (I) increased bottledwater purchases among households buyingit prior to the contamination, (2) bottled wa­ter purchases by new buyers, (3) installinghome water treatment systems, (4) haulingwater from alternate sources, and (5) boil­ing water. To avoid violation of the joint­ness in production assumption, respon­dents were asked to report only thoseactions taken as a specific response to TCEcontamination. The purchase of home wa­ter treatment systems violates the nondur­ale assumption, since the system providesbenefits lasting beyond the contamination.To overcome this problem, only a portionof the purchase and installation costs wereincluded in the estimates. The useful life offiltration systems was considered to be tenyears. Since the 88-week contamination pe­riod under study represented 0.17 of theuseful life of the system, only 0.17 of thepurchase and installation price was in­cluded in the estimates. Averting goodsmust also not exhibit jointness in produc­tion. A possible display of jointness in pro­duction is the purchase of bottled water.

The water may be purchased for taste aswell as health reasons. However, jointnessis less likely to exist in studying a contami­nation incident of a relative short duration.Consequently, the increase in bottled waterpurchases during the contamination inci­dent is likely to be a direct result of theincreased health risks and not changes intastes and preferences.

The costs of these actions were calcu­lated for the sample and extrapolated to thetotal population of Perkasie residents I (Ta­ble I). The total increase in expendituresfrom December 1987, when TCE was firstdetected, to September 1989 ranged from

I Because costs of the actual groundwater contami­nation episode were of interest, extrapolation to thetotal Perkasie population was based on the existinglevel of households' knowledge ofTCE. Thus, the lossestimate does not address the question of what losseswould have been if all households had informationabout contamination. Since we did not know howhouseholds that were unaware would behave once in­formed of TCE. estimation of losses with full informa­tion was not attempted. Loss estimates under such ascenario can be expected to exceed those obtained dueto the expectation that at least some of the householdsthat were unaware would have increased averting ex­penditures upon learning of TCE.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

68(2) Abdalla et at.: Groundwater Contamination 167

$61,313.29 to $131,334.06, depending onthe wage rate used to reflect the value oflost leisure time. 2 Given the theory re­viewed in Section I, these values are inter­preted as a lower bound estimate of theeconomic losses from TCE contamination.The average weekly increase in averting ex­penditures per household which undertookaverting actions in response to the contami­nation was $0.40.

IV. DETERMINANTS OF AVERTINGEXPENDITURES

Household averting behavior decisionsin Perkasie were considered as a two-stepprocess. The household's first step is to de­cide whether actions should be taken to re­duce exposure to water contaminants.There are two possible outcomes as a resultof this step; either the household decidesthat the contamination is significant enoughthat actions should be taken or the contami­nation is not significant enough to warrantaction. Logit regression models were usedto determine factors influencing decisionsto undertake averting actions. The depen­dent variable was equal to 1 if the house­hold took at least one averting action as aspecific response to TCE contaminationand equal to 0 if no specific actions weretaken (regardless of any previous actionstaken). The logit regression results in Table2 indicate that households were more likelyto take averting actions if they received in­formation about TCE, rated the cancerrisks associated with the levels of TCE intheir water to be relatively high, and chil­dren ages 3 to 17 were present in thehousehold.

Only individuals who made the decisionto avert considered a second decisionstep-selection of the intensity of avertingactions. The estimated increase in house­hold averting expenditures was used as anindicator of averting behavior intensity. Or­dinary least squares regression modelswere constructed to identify factors influ­encing averting expenditure increases. Thedependent variable was the estimatedhousehold averting expenditure directlydue to the TCE contamination over the

TABLE 2

LOGIT REGRESSION RESULTS OF DECISION

WHETHER TO TAKE AVERTING ACTIONS FOR

PERKASIE HOUSEHOLDS AWARE OF TCE

CONTAMINATION

Variable Beta Chi-Square

INTERCEPT -4.47 14.34***TCE RISK 0.47 9.79***TCE INFO 0.35 8.99***CHILD 3-17 0.92 7.21***OTHER PROBLEMS 0.37 1.05CHILD UNDER 3 0.19 0.45GENDER 0.43 2.27FAMILIARITY 0.30 2.11

Number of observations: 263R-Value: 0.296Percent correctly predicted by model: 70.3%

••• Indicates statistical significance at the 0.01 level.

Variable DescriptionsTCE RISK: Qualitative rating of the cancer risk associated

with the levels of TCE in the water (I = insignificant risk to5 = very serious risk).

TCE INFO: Addition of two qualitative integer responsesto the amount of information received or obtained concerningTCE (2 = no information to 10 = a lot of information).

CHILD 3-17: 0 if no children ages 3-17 were living in thehousehold, I if at least one child age 3-17 was present.

OTHER PROBLEMS: 0 if individual was not aware of prob­lems in addition to TCE, 1 if aware of at least one additionalproblem.

CHILD UNDER 3: 0 if no children under 3 years of age.were present in the household, I if at least one child under 3years of age was present in the household.

GENDER: 0 if respondent was male, 1 iffemale.FAMILIARITY: Qualitative ranking of respondent's famil­

iarity with chemical substances (I = not familiar to 4 = veryfamiliar).

88-week study period. Table 3 shows thathouseholds with children less than threeyears of age spent more as a result of theTCE contamination than those withoutyoung children.

While the explanatory power of the mod-

2Uncertainty about the cost estimates may be aresult of sampling error. The number of observationsobtained (761) fell short of the number that were calcu­lated (1,136) for a representative sample of the popula­tion. Since this was lower than the desired numberof responses, follow-up telephone interviews with arandom sample of 50 nonrespondents were conductedto test for possible bias. The hypothesis that nonre­spondents' and respondents' were different for twoattributes was rejected at the .05 level of significance.Despite this evidence that nonresponse bias does notappear to be present, sampling error cannot be com­pletely ruled out.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

168 Land Economics May 1992

V. CONCLUSIONS

Note: Variable descriptions are the same as given in Table 2.•• Indicates statistical significance at the 0.05 level.

TABLE 3ORDINARY LEAST SQUARES REGRESSION RESULTS

FOR INTENSITY OF AVERTING ACTIONS FOR

PERKASIE HOUSEHOLDS AWARE OF TCE

CONTAMINATION

els is relatively low, the signs of the statisti­cally significant coefficients were in the ex­pected directions. The results are alsoconsistent with previous work. For exam­ple. the positive influence of informationabout environmental contamination re­ceived by households upon averting expen­ditures was also found by Swartz andStrand (1981).

tures included the presence of young chil­dren in a household and income levels. Re­sults of this study imply that policy analystsshould consider using studies to measureaverting expenditures when examining pol­icy alternatives. We have shown that mea­suring averting expenditures to estimate thecost of environmental pollution is concep­tually valid and empirically feasible. In cer­tain situations, it may be less expensive tomeasure averting expenditures than to de­velop estimates of pollution costs usingother methods. Drinking water contamina­tion appears to be such a case. Where alower bound estimate of costs is sufficientto justify policy action, such as setting anMCL for a drinking water contaminant,there is no need for methods that are moreexpensive in time and money, even thoughthey may be more inclusive and thus pro­vide a complete estimate of total costs.Failure of policy analysts to consider theuse of averting expenditure studies may re­sult in more expensive or less timely deci­sions.

Another implication for policy thatemerges from this study draws on the find­ing that averting expenditures vary withhouseholds' qualitative perception of thehealth risk and knowledge of contamina­tion. Risk communication strategies whichaffect perception of drinking water risksmay change the estimates of benefits andcosts of environmental policies which af­fect health.

A surprising finding of this study wasthat only 43 percent of Perkasie residentswere aware of the TCE contamination de­spite legislation which requires public wa­ter authorities to notify its customers ofcontamination. This significantly differsfrom the results of an earlier study (Abdalla1989) in which 96 percent of residential cus­tomers were aware of a contamination inci­dent in central Pennsylvania. Thus, existingpublic notification rules and procedures ap­pear inadequate to inform all affected cus­tomers.

Since awareness of contamination influ­ences averting behavior, the policies andprocedures for public notification are also

0.642.21**1.161.02

t-Value

54.3364.7117.3037.56

Parameter EstimateVariable

INTERCEPTCHILD UNDER 3TeE RISKOTHER PROBLEMS

Number of Observations: 113F- Value: 3.16 (Entire model is significant at

0.01 level)R-Squared value: 0.12

Data on the averting expenditure in­creases of households in a Pennsylvaniacommunity was collected to estimate thecosts of a groundwater contamination inci­dent. Averting expenditures were esti­mated to range from $61,313.29 to$131,334.06 during an 88-week TCE con­tamination period. Under specific assump­tions, the change in averting expendituresassociated with a change in environmentalquality provides a conservative estimate ofthe true cost, or benefit, of the environmen­tal change (Roach 1990). These assump­tions appear to be reasonably appropriateto situations involving drinking water con­tamination.

Qualitative risk perception and knowl­edge of contamination are important deter­minants of households' decisions to under­take averting behaviors, while the factorsthat influence the level of averting expendi-

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

68(2) Abdalla et al.: Groundwater Contamination 169

important factors affecting public realiza­tion of the costs. The regression resultssuggest that households may not be equallyconcerned with health risks posed by con­taminants in drinking water. Notificationefforts could be intensified towards thosegroups which appear to be more concernedwith water quality. For example, house­holds with young children tend to spendmore on averting activities related to wateruse. Notification programs targeted at par­ents of young children could be deliveredthrough child care centers or pediatricians'offices.

In light of the few studies documentingthe existence and nature of behaviors toavoid environmental contaminants, the re­sults are significant. While the averting ex­penditure method does not encompass allimpacts, this study indicates that themethod is capable of yielding conceptuallyvalid estimates of an important category ofthe costs of environmental pollution.

References

Abdalla, Charles, W. 1989. "Measuring theCosts of Groundwater Contamination: AnEmpirical Application Using Averting Be­haviors." Agricultural Economics and RuralSociology Staff Paper No. 161, PennsylvaniaState University, University Park, April.

Bartik, Timothy J. 1988. "Evaluating the Bene­fits of Non-Marginal Reductions in PollutionUsing Information on Defensive Expendi­tures." Journal ofEnvironmental Economicsand Management 15: 111-27.

Courant, Paul N., and Richard C. Porter. 1981."Averting Expenditures and the Costs of Pol­lution." Journal of Environmental Econom­ics and Management 8(4):321-29.

Dillman, Don A. 1978. Mail and Telephone Sur­veys: The Total Design Method. New York:John Wiley.

Harford, Jon D. 1984. "Averting Behavior andthe Benefits of Reduced Soiling." Journal of

Environmental Economics and ManagementII :296-302.

Harrington, Winston, and Paul R. Portney.1987. "Valuing the Benefits of Health andSafety Regulation." Journal of Urban Eco­nomics 22:101-12.

Harrington, Winston, Alan J. Krupnick, andWalter O. Spofford, Jr. 1989. "The Eco­nomic Losses of a Waterborne Disease Out­break." Journal of Urban Economics 25:116-37.

Kalton, Graham. 1983. Introduction to SurveySampling. Beverly Hills: Sage Publications.

Roach, Brian A. 1990. "Valuing Changes inDrinking Water Quality Using Averting Ex­penditures." Master's thesis. Department ofAgricultural Economics and Rural Sociology,Pennsylvania State University, UniversityPark.

Shortie, James S., and Brian A. Roach. 1989."A Diagrammatic Analysis of Defensive Ex­penditures in a Household ProductionFramework." Agricultural Economics andRural Sociology Staff Paper No. 171, Penn­sylvania State University, University Park.

Smith, V. Kerry, and William H. Desvousges.1986. "Averting Behavior: Does it Exist?"Economics Letters 20:291-96.

Snedecor, George W., and William G. Cochran.1980. Statistical Methods. Ames: The IowaState University Press.

Spofford, Walter A., Alan J. Krupnick, and EricF. Wood. 1989. "Uncertainties in Estimatesof the Costs and Benefits of Groundwater Re­mediation: Results of a Cost-Benefit Analy­sis." Discussion Paper QE89-15. Washing­ton, DC: Resources for the Future.

Swartz, David G., and Ivar E. Strand, Jr. 1981."Avoidance Costs Associated with ImperfectInformation: The Case of Kepone." LandEconomics 57(May): 139-50.

U.S. Environmental Protection Agency. 1990.Progress in Groundwater Protection andRestoration. EPA 440/6-90-001, Office ofWa­ter, Washington, DC, Feb.

Watson, William D., and John A. Jaksch. 1982."Air Pollution: Household Soiling and Con­sumer Welfare Losses." Journal of Envi­ronmental Economics and Management 9:248-62.