The effects of ownership and benchmark competition: An ...

Sujets

• Droits de l'homme
• Sport aérien
• Organique
• Indice de Herfindahl-Hirschmann
• Interactionnisme symbolique
• Série de Fourier

Informations

• Publié par: Shous
• Nombre de lectures: 44
• Langue: English

Extrait

Available online at www.sciencedirect.com InternationalJournalofIndustrialOrganization 26(2008)186 – 205 www.elsevier.com/locate/econbase Theeffectsofownershipandbenchmarkcompetition: ☆ AnempiricalanalysisofU.S.watersystems ScottWallsten ⁎ ,KatrinaKosec AmericanEnterpriseInstituteandAEI – BrookingsJointCenter,115017thStreet,NWWashington,DC20036,USA Received16March2006;receivedinrevisedform30October2006;accepted3November2006 Availableonline26December2006 Abstract Shouldgovernmentsorprivatefirmsownwatersystems?Canyardstick,orbenchmark,competition effectivelydisciplineutilities?Littleempiricalresearchshedslightontheseissues.Weuseapaneldataset thatincludeseverycommunitywatersystemintheUnitedStatesfrom1997to2003totesttheeffectsof ownershipandbenchmarkcompetitiononregulatorycompliance.Wefindthatownershipdoesnot generallyaffectcompliancewiththeSafeDrinkingWaterAct(SDWA).Greaterbenchmarkcompetition, however,isassociatedwithfewerSDWAviolations.Theresultssuggestthatownership,perse,maynot mattermuch,butthatcompetition – eveninitsweakerformofbenchmarkingbyconsumersorregulators canimproveperformance. ©2006ElsevierB.V.Allrightsreserved. JELclassification:L00;L33;L50;L95 Keywords:Industrialorganization;Privatization;Regulation;Competition;Waterutilities – 1.Introduction Privatizingutilitiesremainscontroversial,thoughmoststudiessuggestthatityieldspositive outcomes.Thosepositiveoutcomes,however,aremuchmorerobustinindustriesthatsupport rigorouscompetition,liketelecommunications.Whataboutindustriesinwhichcompetitionis moredifficulttointroduceorsustain?Isaprivatemonopolylikelytooutperformastate-owned ☆ TheauthorsgratefullyacknowledgetheassistanceofCarolineCecot,LeeKyle,AbeSiegel,MollyWells,andthe commentsofRobertHahn,GeorgeClarke,andtwoanonymousreferees.Theauthors,however,aresolelyresponsiblefor anyerrors. ⁎ Correspondingauthor.Tel.:[phone]. E-mailaddresses:[e-mail](S.Wallsten),[e-mail](K.Kosec). 0167-7187/$-seefrontmatter©2006ElsevierB.V.Allrightsreserved. doi:10.1016/j.ijindorg.2006.11.001

S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205781 monopoly?Canyardstick,orbenchmark,competitiondisciplineprivateorpublicprovidersinthe absenceofhead-to-headcompetition?Thispaperattemptstoanswerthesequestionsinthe contextofdrinkingwatersystemsintheUnitedStates. Theidealformofownershipinnetworkindustriesremainsheavilydebated.Water privatizationintheUnitedKingdom,forexample,isstillahotissuenearlytwentyyearsafter MargaretThatchersoldtenregionalwaterauthoritiesandestablishedanewregulator.Inthe UnitedStates,wherenearlyallutilitiesareownedoroperatedbyprivatefirms,around80%ofthe populationreceiveswaterfrompublicly-ownedsystems. Proponentsofprivateownershiparguethatfirmscandeliverservicesmoreefficientlythancan governments.Governmentshavediverseandoftencontradictoryobjectivesthatcanleadtoa misallocationofresources,andmayfaceincompatibleincentiveswhentheyactasbothowner andregulatorofautility.Inaddition,politiciansmayfeelpressuretoprovidejobsintheutilityto constituents,forexample,ratherthaninvestinginthewaternetwork. Opponentsofprivateownershipcounterthatprivatizationisdifficultandimpracticalgiven waterandsewerageexternalitiesandthelimitedscopeforcompetition.Someobjecttothevery ideaofconsideringwateraneconomicgoodorallowingaprivatecompanytorunawatersystem. PublicCitizen(2004),anoutspokencriticofprivatization,highlightstheintensefeelings regardingwaterprivatization,claimingthat “ perhapsthegreatesttheftofcommonresources facinghumanityandtheplanetisthecorporatetakeoveroftheworld'swater. ” Whiledirectcompetitionamongpipedwatersystemsisessentiallynonexistent,benchmark,or yardstick,competitionmaybepossible.Shleifer(1985)formalizedamodelofregulationvia yardstickcompetitioninwhicharegulatedfirm'spricesdependonotherfirms'costs. Benchmarkingmayalsoapplytoobservableinformationotherthanprices.Evenifconsumers cannotchoosewhosuppliestheirdrinkingwater,utilitiesmayfaceperformancepressureif consumersorregulatorscancomparesystems.Consumersdissatisfiedwiththeirutilitiescould expresstheirdispleasurethroughtheirelectedrepresentatives,whileregulatorscouldusethe informationfrombenchmarkcomparisonstodemandpriceorqualitychangeswithpotentially dramaticeffectsonfirmsinthefuture. Despitethepassionstheissueraises,thereisscantempiricalevidenceonwhetherpublicor privatewatersystemstendtoperformbetterandwhetherbenchmarkcompetitionactually improvesoutcomes.Toourknowledge,nosystematic,empiricalanalysiscomparespublicand privatewatersystemsintheUnitedStatesandwearenotawareofanyempiricalliterature exploringthepotentialimpactofbenchmarkcompetitioninthewatersector.Weaddressthese gapsbyexploringU.S.waterqualityandhouseholdexpendituresonwaterbysystemownership typeandthedegreeofbenchmarkcompetitionfaced.Inparticular,weuseapaneldatasetthat includeseverycommunitywatersysteminthecountryfrom1997to2003totesttheeffectsof ownershipandbenchmarkcompetitiononviolationsoftheSafeDrinkingWaterAct(SDWA). Controllingforlocationfixedeffects,income,theshareofthepopulationthatisurban, ownershipconcentrationinthelocalwatermarket,systemsize,watersource,andyearfixed effects,wefindlittledifferencebetweenpublicandprivateownershipwithregardtoSDWA violations.Amongsmallerwatersystems,privateownershipisassociatedwithfewerviolations ofmaximumcontaminantlevelsbutmoremonitoringandreportingviolationsthanpublic systems.Amonglargersystems – especiallythoseservingmorethan100,000people – private ownershipisassociatedwithmorecontaminantbutfewermonitoringandreportingviolations. Theresultsneithersupportcriticsofprivateownershipnorsuggestthatprivateownershipis superior.Overall,compliancewithSDWAregulationsdoesnotappeartodependmuchonsystem ownership.Benchmarkcompetition,however,appearstohaveamorerobustimpact.Greater

818S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205 benchmarkcompetition,asmeasuredbytheconcentrationoftheshareofconnectionsheldby watersystemsineachcounty,isassociatedwithbothfewercontaminantandfewermonitoring andreportingviolations.Thissuggeststhatevenaweakformofcompetitioncanhavepositive effectsonregulatorycompliance.Overall,ourresultsareconsistentwiththehypothesisthat publicorprivateownershipmaynotmakemuchofadifferencewhentheoperatorislargely monopolistic,butthateventhen,benchmarkcompetitioncanyieldtangiblebenefits. 2.Theeconomicsofwatersystems Waterandsewerageserviceprovisionentailsuniquechallenges.Fixedcostsarehigh,barriersto entryarelarge,thepotentialforcompetitionislimited,andthesectorhasmanyandlarge externalities.Thelong-livedfixedassetscreateincentivesforbothfirmsandgovernmentstodelay investmentanduserevenuesforotherpurposes.Governmentsoftenencounterpoliticalandlegal obstaclestoprivatization,andprivateinvestorsworrythatgovernmentswillexpropriatetheirassets. Whiledirectcompetitioninwaterprovisionisdifficult,indirectcompetitionismorefeasible. Suchcompetitioncouldtaketheformofprivatefirmscompetingforoperationandmanagement contracts(competitionforthemarket)orbenchmarkcompetition,wheretheexistenceofalarge numberofwatersystemscouldallowconsumersandregulatorstocomparetheirsystemtoothers andaccordinglypressuretheirownutilitytoimproveperformance. Waterservicesmaybesubjecttoavarietyofmarketfailuresthatrequiregovernment involvementorregulation.Externalities,especiallyhealth-related,areespeciallypronouncedin waterandsewerage,complicatingprivatesectorinvolvement. 1 Externalitiesnecessitate mechanismsofensuringthatprices,investment,andotheroperationaldecisionsarenotdivorced fromexternalitiesthataffectactorsotherthanfirms. Somefeaturesofwaterandsewerageinfrastructurecreateincentivesforbothgovernmentsand firmstodelayinvestment.Muchofthesector'sassets,suchasthepipes,aresunk,haveno alternativeuses,andareverylong-lived.Becausethefixedcostscomprisealargeshareoftotal costsandbecausethesystemcanoperateforalongtimewithoutmuchintervention,aself- financingutilitywillearnquasi-rents(Noll,2002).Bothpoliticiansandinvestorsmayfeel pressuretousethequasi-rentsforshort-termprioritiesratherthaninvestment.Consumersmay demandlowertariffs,andgovernmentsthatownandoperatetheirwatersystemsmayface pressuretouserevenuesforotherimmediateneeds.Privateoperators,meanwhile,mayworry aboutexpropriationandfaceinvestorpressuretoavoidreinvestingthequasi-rents. 2 2.1.Theevidenceonprivateversuspublicownership Anumberofstudieshaveevaluatedprivatizationinmanyindustriesaroundtheworld. MegginsonandNetter(2001)andShirleyandWalsh(2000)findincomprehensiveliteraturereviews thatprivatizationhasgenerallyimprovedefficiency.Littleofthatresearch,however,hasfocusedon 1 Theextenttowhichhealthisactuallyanexternalitydependsonwhetherconsumersandtheutilitybearthefullcosts ofanyhealthimpacts.Nonetheless,absentsometypeofoversightconsumershavelittlemeansofmonitoringthehealth qualityoftheirwater. 2 U.S.watersystemsinthe19thcenturywerelargelyprivatelyrun,thoughcitiesbegantomunicipalizetheminthe early20thcentury.TroeskenandGeddes(2001)arguethatmunicipalizationoccurredbecauseprivatefirmswereunder- investingintheirnetworks,andthatthisunderinvestmentoccurredbecausethemunicipalitiescouldnotcrediblycommit tonotexpropriatingtheirresources.Theyconcludethatunderinvestmentwasarationalresponsetothefearof expropriation.

S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205981 water.Becauseprivatizationhasbeencombinedwithcompetitioninsectorsliketelecommunica- tions,itislessclearwhethergovernmentorprivateownershipisinherentlysuperiorwhenthe potentialforcompetitionislimited.Withoutappropriateincentives – intheformofregulation,some typeofoversight,oraviablethreatofentryinsomesphereofoperations – itisnotobviousthata privatemonopolywouldbebetterthanapublicmonopoly. Existingresearchcomparingpublicversusprivateownershipofwaterprovisionconsists primarilyofcasestudies,fromwhichitisdifficulttogeneralize.Casestudiesareusuallynot chosenrandomly,butrathertendtofocusonexceptionalcases. 3 Forwaterservices,thosecases oftenfocusonspectacularfailures,suchasthedisastrousprivatizationattemptinCochabamba, 4 Bolivia.However,atleast25,000U.S.watersystemsareprivatelyowned,andothercountries likeFrancerelyheavilyontheprivatesectorwithoutdisastrousconsequences,suggestingthatthe relativelysmallnumberofspectacularfailuresmaynotberepresentative. 5 Casestudiesofwaterownershipinindustrializedcountriesareinconclusive.TheEconomist (2003)arguesthatwaterprivatizationintheUKhasbeensuccessfulwhencomparedtowater systemsinScotland,whichwentfromprivatetopublichands.SaalandParker(2001),contrarily, findhigherpricesbutlittleproductivityimprovementfollowingtheUKprivatization.Likewise, Cowanetal.(2000)estimatethatprivatizationintheUKledtoanetwelfareloss. Mostofthefewempiricalstudiesonwaterprivatizationlackgood,consistentdataandyield generallyinconclusiveresults.OneexceptionisastudybyGalianietal.(2003),whorigorously exploretheeffectofwaterprivatizationonchildmortalityinArgentina.Theyfindthatchild mortalityfelleightpercentinregionsthatprivatizedtheirwatersystems,andthattheeffectswere largestinthepoorestareas. 6 Clarkeetal.(2004)findinahousehold-levelstudyofdeveloping countriesthatconnectionratestopipedwaterincreasedfollowingprivatization,evenamongthe poor,butthatconnectionratesincreasedsimilarlyinareasthatdidnotprivatize. Empiricalresearchontheeffectsofprivatesectorparticipationinindustrializedcountries focuseslargelyonoperatingefficiency,andisinconclusive.Severalstudiesinthe1980susing fairlysmalldatasets(120observationsandfewer)foundnosignificantdifferencesbetweenpublic andprivatewatersystemefficiencyintheU.S.(Byrnesetal.,1986;FeigenbaumandTeeples, 1983;TeeplesandGlyer,1987).Bhattacharyyaetal.(1995),however,foundinasampleof190 publicand31privatewatersystemsthatprivate(public)operatorstendedtobemoreefficient amongsmall(large)systems.RenzettiandDupont(2003)concludedthattheliteraturefoundlittle evidenceof “ unambiguousimprovements ” fromprivatization,butthatpublic – privatepartner- shipsmightbepromising. Whileexistingempiricalstudiesontheeffectsofownershipfocusprimarilyonoperating efficiency,criticsalsoworrythatprivateoperatorsmaygivelessweighttopublicinterestthan 3 Shirley(2002)isanexception.Thiscollectionofsixwaterprivatizationcasestudieswasundertakenwithacoherent frameworkmeanttoensurethatthecases,incountrieswithvaryinginstitutionalandphysicalcharacteristics,were evaluatedinaconsistentmanner. 4 Afocusonfailuresisnotsurprising — theymayinvolvelargeamountsofmoneyandaffectlargepopulations. Atlanta,forexample,signeda20-year,$428millionmanagementcontractwithaprivateoperatorin1998,onlytocancel itafewyearslaterasboththecityandthefirmaccusedtheotherofviolatingthecontract(Brubaker,2003). 5 In2003privateoperatorsowned25,977communitywatersystems(EnvironmentalProtectionAgency,2003).A formerwateranalystforPublicCitizennotedthat “ Contrarytothecritics'contentions,privateoperatorshavea respectablerecordofprovidingqualitywaterandcomplyingwithenvironmentalstandards ” (Tsybine,2002). 6 Theiranalysisisrigorous:theauthorshonetheirfindingsbyseparatingmortalitythatcanbecausedbywater conditionsfrommortalityunrelatedtowater.Privatizationwasuncorrelatedwithmortalityfromnon-watercauses,but stronglycorrelatedwithmortalitycausedbywaterconditions.Moreover,themeasuredeffectmaybeunderestimated sincetheirwateraccessdataunder-sampledthepoorestareas,whichappearedtoreceivethelargestbenefits.

190S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205 wouldagovernmentoperator,resultinginlowerqualitywater.Toourknowledgenolarge-scale empiricalstudieshaveexploredtheeffectsofownershiponwaterqualityinindustrialized countries.Inapubliclyavailablepresentation,GasteyerandVaswani(2004)usesomeofthe EnvironmentalProtectionAgency(EPA)datathatweuse,butonlywithcross-sectionaldataand countyaverages. 7 Theyconcludethatprivateoperatorshaveslightlymorecontaminantviolations oftheSDWA.Noresearchappearstohaveexploredcompetitioninwater,totheextentthatit mightexist. 3.Regulationandcompetition TheSafeDrinkingWaterAct(SDWA)of1974coversalloftheapproximately160,000public watersystemsintheUnitedStatesandallowstheEPAtosetmaximumacceptablecontaminant levels,outlinewatertreatmentprocedures,andrequirewatersystemstofollowaprescribedwater qualitymonitoringschedule. 8 TheEPAalsoregulatesthestates,localities,andwatersuppliers thatimplementthosestandards. 9 Waterratesappeartoberegulatedinconsistentlyacrossthecountry(Mann,1993). Unfortunately,toourknowledge,therearenorecentsurveysofstateandlocalwaterregulators orregulations.A1989survey,however,foundthatingeneralstatesregulateprivatebutnotstate- ownedutilities(BeecherandLaubach,1989).Mann(1993)notesthatmayors,citycouncils,or otherelectedorappointedrepresentativesgenerallyapproveratesofwatersystemsownedbythe government.Inotherwords,theevidencesuggeststhatwithsomeexceptions,bothpublicand privatewatersystemsaresubjecttosomepricingoversight — directlybyregulatorsinthecaseof privatefirms,orbypoliticianswhomustapprovepricesinthecaseofgovernment-owned utilities.Perhapsinpartbecauseofscantdata,thereislittleresearchonwaterrateregulation. Wolak(1994)modelstheinteractionsbetweenregulatorsandwaterutilitiesunderasymmetric informationinordertouncovertheutility'scostfunction.Timmins(2002)estimatestheobjective functionofregulatorsandestimatesthedeadweightlosscausedbytheconsistentunderpricingof waterrelativetoitsmarginalcost. Inadditiontoregulatoryorpoliticalconstraintsonutilities'behavior,theymayalsoface pressurefrombenchmarkcompetition(alongthelinesofShleifer,1985).Whiletheredoesnot appeartobeanyempiricalanalysisofbenchmarkcompetition,basedlargelyoninterviews 7 Ouranalysisimprovesontheirsinanumberofwaysthatallowsustoexplorethedatamorerigorously.First,weusea paneldatasetcovering1997 – 2003,whiletheyuseonlyacross-sectionfromyear2000.Second,ourcontaminantanalysis isatthelevelofthewatersystem,whiletheyusecountyaverages.Thatis,forthecontaminantanalysisweusethe ownershipstatusofeachwatersystemratherthantheshareofwatersystemsinacountythatarepublicorprivate.Third, weuseaneconometricapproachsuitedtothistypeofdatathatallowsustocontrolformanyfactorsotherthanownership thatmightaffectquality.Contrarily,GasteyerandVaswanishowonlyaregressionlinehighlightingthecorrelation betweenownershipshareandaveragenumberofviolationsinthecounty. 8 Testingmustbecarriedoutbystate-orEPA-certifiedlaboratories.Anymembersofawaterutilityorlabfoundguilty ofdeliberatelyfalsifyingdatacanbeprosecutedandjailed.Certifiedlabsmustalsopassperiodicon-siteaudits.Given thesesafeguards,theEPAbelievesthatitsdataarecollectedinauniformmannerandarereliable. 9 SeetheEPA'sSDWAwebsitefordetails:http://www.epa.gov/safewater/sdwa/index.html(lastaccessed3/3/05).The 1986amendmentsexpandedtheSDWA'sscopetoprotectaquifersfrompollutantsanddevelopdrinkingwaterstandards forpreviouslyunregulatedcontaminants.The1986revisionadded83additionalcontaminantstothelist(Sapatand Teske,2004).In1996,severalnewamendmentsemphasizedprotectingsourcewater,mandatedoperatortrainingand certification,andrequiredwatersystemstoprovideinformationonwaterqualitytoconsumers.(Seehttp://www.epa.gov/ safewater/sdwa/30th/factsheets/understand.html(lastaccessed3/3/05)andhttp://www.epa.gov/safewater/sdwa/theme. html(lastaccessed3/4/05)).

S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205911 Sawkins(1995)concludesthatyardstickcompetitionimprovedwaterregulationinEnglandand Wales. Benchmarkcompetitioncouldworkthroughanumberofmechanisms.IntheUnitedStates, theEPApostsdetailedsystem-levelwaterqualitydataonitswebsiteandrequireswatersystems tosubmitannualconsumerconfidencereportstoconsumersandregulators.Thesereportsinclude alistingofallviolationsofEPAhealthregulations,thepotentialsourcesofcontamination,andthe likelyhealtheffects.Thepurposeofthesereports,accordingtotheEPA,isto “ enableAmericans 01 tomakepractical,knowledgeabledecisionsabouttheirhealthandtheirenvironment. ” Consumersandregulatorscancomparetheirwaterqualitytoindustrybenchmarksandpressure systemsthatfallshort. Ourcomprehensivewatersystem-levelpaneldataset,derivedfromdatacollectedbytheEPA undertheSDWAanddatacollectedbytheU.S.Census,allowsustoshedlightontheeffectsof ownershipandbenchmarkcompetitiononwaterquality(asmeasuredbycompliancewithwater qualityregulations).Wediscussourdata,empiricalapproach,andresultsbelow. 4.Dataandanalysis The1996SDWAamendmentrequirestheEPAtocollectdataandmakemuchofitpublic. 11 WeobtaineddataoneachcommunitywatersystemintheUnitedStatesfromtheEPA'sSafe DrinkingWaterInformationSystem,FederalVersion(SDWIS/FED). 12,13 TheSDWIS/FED containsdataoneverypublicwatersystemintheUnitedStates,includingtribalareasandU.S. territories.Asdiscussedabove,watersystemsmustperiodicallytesttheirwaterandreportresults tothestate. 14 Thestate,inturn,determineswhetherwatersystemsarecomplyingwiththe regulationsandreportingviolationstotheEPAperiodicallyasprescribed. 15 TheEPArecordsthis informationintheSDWIS/FED,whichthereforecontainsdataonviolationsofmaximum allowedcontaminantlevelsaswellastreatmentandreportingviolations. 16 TheEPAreports violationsasdiscreteeventsannually,sothedatasetreveals,forexample,howfrequentlywater suppliedbyeachwatersystemexceededthemaximumallowedlevelsofcertaincontaminantsina 10 http://www.epa.gov/safewater/ccr/frequentquestions.html(lastaccessed6/11/06). 11 Recentdataareavailableonlineandolderrecordsareavailableviafreedomofinformationactrequest. 12 WearegratefultoLeeKyleoftheEPAOfficeofWaterforextractingthesedataforusandpatientlyansweringour manyquestionsandrequestsforadditionaldata.Theauthors,however,aresolelyresponsibleforanymistakesmadein usingorinterpretingthedata. 13 Acommunitywatersystemisapublicwatersystemthatservesatleast15serviceconnectionsusedbyyear-round residentsorregularlyservesatleast25year-roundresidents. 14 TheEPAestablisheshowoftensystemsmustmonitorwatersuppliesforchemicalsbasedonsystemcharacteristics. Forinstance,ifasystemhashadnoviolationsofacertaincontaminantandislocatedinanareawherethecontamination riskislow,thesystemcanapplyforanexemptionfromtestingforthatcontaminantasfrequentlyaswouldotherwisebe required.See:http://www.epa.gov/safewater/pws/pdfs/qrg_smonitoringframework.pdf(lastaccessed:12/14/04). 15 Accordingto40CFRpart142,Section15: “ EachState … shallsubmitquarterlyreportstotheAdministratorona scheduleandinaformatprescribedbythe[EPA]Administrator,consistingofthefollowinginformation:(1)New violationsbypublicwatersystemsintheStateduringthepreviousquarterofStateregulationsadoptedtoincorporatethe requirementsofnationalprimarydrinkingwaterregulations,includingviolationsofthepublicnotification requirements … . ” Seehttp://www.epa.gov/safewater/regs.html#cfr(lastaccessed12/15/04). 16 Atreatmentviolationisafailuretoproperlytreatadrinkingwatersourceinordertoreducethelevelofaspecified contaminant.Areportingviolationisafailuretocollecttherequirednumberofsamples(includingconfirmationsamples) inthespecifiedtimeframe,afailuretoensurethesamplesareanalyzedproperly,orafailuretosubmitrequired monitoringinformation.

291S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205 Table1 Numberofactivesystemsin2003 OwnershipSystemsize VerysmallSmallMediumLargeVerylarge Private21,774375163045850 Localgovt.6469935937162879309 F St e a d t e e ra g l o g vt o . vt.1 2 3 0 7 5 1 2 0 4 8 1 8 7 7 5 6 3 7 0 0 1 P N u at b i l v ic e /p A ri m v e a r t i e can4 7 5 1 2 1 1 6 6 1 5 9 2 1 6 356 3 66 0 Unknown54713224130 Total30,29514,37546933516366 3 N 3 o 0 t 0 e: , S M iz e e di c u at m e : g 3 or 3 i 0 e 1 s – es 1 t 0 a , b 0 li 0 s 0 h , e L d a b rg y e: US 10 E ,0 P 0 A 1 b – a 1 s 0 e 0 d ,0 o 0 n 0 a , v V e e r r a y ge la d r a g i e l : y 1 p 0 o 0 p , u 0 l 0 at 1 i + o . nserved:Verysmall:25 – 500,Small:501 – givenyear,butnotbyhowmuch. 17 Thedataalsoincludeself-reportedinformationonsystem ownership,numberofserviceconnections,populationserved,primarywatersource,andthe geographiclocationofeachwatersystem. U.S.watersystemsvaryacrossmanydimensions,makingthemidealfortestingtheceteris paribuseffectsofsuchcharacteristicsasownershiporwatersystemsize.Systemsrangefrom smallpublicutilitiesservingasfewas25peopleandrelyingongroundwatertolargeprivate utilitiesprovidingtreatedsurfacewatertomillionsofcustomers. Ownershipcategoriesincludeprivate,federal,state,local(municipal),tribal,public/private, andunknown.Privatesectorinvolvementcanrangefromcompleteownershipandoperationto smallcontractoperationswork.Wehadhopedthatthepublic/privatecategorywouldcapture thesecontractoperations.Unfortunately,accordingtotheEPA, “ Public/Private ” hasnoexplicit 81 definitionandthecategoryisnotusedconsistently.Fewerthanfourpercentofwatersystems classifiedthemselvesasPublic/Private.Primarywatersourcecategoriesincludegroundwater, surfacewater,purchasedgroundwater,purchasedsurfacewater,andgroundwaterinfluencedby surfacewater.Differenttypesofwatermaybemorepronetocontaminationortohigherfinal waterprices,anditisthusimportanttocontrolfortheprimarywatertype. Theresultingpaneldatasetcovers1997 – 2003andincludes377,629activewatersystem- years. 19 Whiletheprecisenumberofwatersystemschangeseachyear,for2003thedataset 02 contains90,421communitywatersystems,ofwhich53,245areactive.Table1showsthe numberofwatersystemsbyownershipstatusandsizecategory,andTable2showsthetotal numberofserviceconnectionsbyownershipandsize. Thetablesrevealthatmostwatersystemsareprivate,butthatthosesystemstendtobevery small.Acasualinspectionofthenamesofverysmallsystemssuggeststhattheyarelargely mobilehomeparks,apartmentandcondominiumcomplexes,andresortsandhotels.Nonetheless, asubstantialnumberofmediumtolargesystemsareprivate,althoughtheverylargest(thosethat 71 Seehttp://www.epa.gov/safewater/mcl.htmlforalistofcontaminants;http://www.epa.gov/OGWDW/mdbp/ieswtr. htmlfortheFederalRegisternoticeonwatertreatmenttechniques,andhttp://www.epa.gov/safewater/mdbp/qrg_st1.pdf forabrochureonwatertreatment. 81 ConversationswithLeeKyle,EPAOfficeofWater,6/17/04and12/3/04. 19 Theviolationsdataextendbackto1976,butarecordofownershipbeginsonlyin1997. 20 InactivewatersystemsarethosereportedtotheEPAasbeinginactiveaswellasthosethatreportedinoneyearand notatallinthefollowingyear.

S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205391 Table2 Totalpopulationservedin2003 OwnershipSystemsize VerysmallSmallMediumLargeVerylarge Private3,082,8964,564,5023,608,74713,691,53419,550,550 Localgovt.1,547,28413,854,54921,707,49381,009,849102,691,886 F St e a d t e e ra g l o g vt o . vt.2 4 6 4 , , 5 8 0 5 2 4 1 3 8 7 8 3 , , 2 1 7 7 3 5 5 4 1 2 9 0 , , 0 2 7 4 5 8 1 8 , 0 5 5 0 , 7 7 , 2 3 3 520 103,414 P N u a b ti li v c e /p A ri m va e t r e ican 1 8 4 2, 6 7 ,9 6 5 4 08 2 8 2 0 3 , , 5 2 1 5 0 5 6 1 8 4 4 1 , , 0 4 3 7 0 7 1 4 , 8 5 , 1 8 7 6 , 0 7881 0 ,275,450 Unknown79,223184,846148,678307,0920 Total5,010,47320,269,11027,229,74898,888,198123,621,300 3 N 3 o 0 t 0 e: , S M iz e e di c u at m e : g 3 or 3 i 0 e 1 s – es 1 t 0 a , b 0 li 0 s 0 h , e L d a b r y ge: US 10 E ,0 P 0 A 1 b – a 1 s 0 e 0 d ,0 o 0 n 0 a , v V e e r r a y ge la d r a g i e l : y 1 p 0 o 0 p , u 0 l 0 at 1 i + o . nserved:Verysmall:25 – 500,Small:501 – serveover1,000,000 — notidentifiedseparatelyinthetable)areownedonlybylocal governments. Privateandlocalsystemsaccountformorethan90%ofallcommunitywatersystems.Water systemsownedbystatesappeartoserveprimarilyprisonsanduniversities.Federally-owned systemsservemilitarybases,nationalparks,andsomeprisons.Finally,NativeAmericantribes oftenownandoperatetheirownwatersystems. Asdescribedabove,theSDWIS/FEDcontainsdataonviolationsofthemaximumallowed levelofanumberofhealth-relatedcontaminants,treatmenttechniques,andmonitoringand reporting. 21 Table3showstheaveragenumberoftheseviolationsbysystemownershipandsize from1997to2003.Theaveragenumberofhealth-basedcontaminantviolationsissmall — generallylessthan0.1persystem,peryear.Theaverage,however,masksvariationacrosssystem types.WatersystemsownedbyNativeAmericantribestendtoexperiencethemostfrequent contaminantviolations,followedbyprivatesystemsandthenlocalsystems.Federally-owned systemsreportthefewestviolationsofhealth-basedmaximumcontaminantlevels. Violationsbysizegiveasomewhatdifferentview.Smallersystemstendtohavethelargest numberofviolationsofalltypes,although “ large ” watersystemsappear,onaverage,tohave almostasmanycontaminantviolationsasverysmallsystems.Differencesacrosssizemeanthat simplecomparisonsacrossownershiptypesmaybedeceptive,ascertaintypesofsystemsare morelikelytobeofacertainsize.Typically,largersystemshavemoresamplingpoints,anda violationatanysamplingpointcountstowardsthetotalnumberofviolations.Onemight thereforeexpectmoreviolationsinlargersystemsforreasonsthathavelittletodowithinherent waterquality.Thisobservationalsomeansthatalargeraveragenumberofviolationsinsmall systemsmaybeevenworsethanthefigureimpliessinceallelseequal,systemswithfewer samplingpointsshouldhavefewermeasuredviolations. 21 TheEPAcompilesdataonthenumberofviolationsperwatersystemsperyear.Itisnotobvioushowbesttousethis variable.Onemightnormalizeorweightitbypopulationservedonthegroundsthataviolationinalargesystem potentiallyaffectsmorepeoplethanaviolationinasmallsystem.However,weusedtheunweightednumberof violationspersystembecauselargerwatersystemstypicallysampletheirwateratmorelocationsthansmallersystems. Largesystemscanthusreceivemoreviolationsthansmallersystemssinceaviolationatanysamplingpointcountsasa separateviolation.(conversationwithAbeSiegel,EPAOfficeofWater,2/2/05).Assuch,weconsideredthevariable – violationspersystemperyear – tobealreadynormalized,albeitimperfectly.Thesefacts,however,makeitespecially importanttocontrolforthesizeofthewatersystemsincesizewillbecorrelatedwiththenumberofviolations.

194S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205 Table3 SDWAviolationsbysystemownership,size,andprimarywatersource MaximumcontaminantTreatmenttechniqueMonitoringandreporting levelviolationsviolationsviolations Ownership Federalgovt.0.0490.0411.312 Localgovt.0.0950.0650.691 NativeAmerican0.1220.0914.983 Private0.1030.0311.449 Stategovt.0.0750.0251.016 eziS Verysmall0.1060.0371.48 Small0.0880.0540.653 Medium0.0830.0550.619 Large0.1010.0870.738 Verylarge0.0450.1411.061 Primarywatersource Groundwaterundertheinfluence0.0750.3533.029 ofsurfacewater(GU) Purchasedgroundwaterunderthe0.0290.1741.377 influenceofsurfacewater(GUP) Groundwater(GW)0.1050.0181.166 Purchasedgroundwater(GWP)0.0590.0130.299 Surfacewater(SW)0.1610.2921.65 Purchasedsurfacewater(SWP)0.0370.0260.29 Note:verysmallsystemsserve25 – 500people,smallserve501 – 3300,mediumserve3301 – 10,000,largeserve10,001 100,000,andverylargeservemorethan100,000. – Monitoringandreportingviolationsarethemostcommonviolationtype,withallbutlocal systemsinviolationatleastonceperyear,onaverage.Asforcontaminantviolations,tribal systemsexperiencethemostfrequentmonitoringandreportingviolations.Locally-owned systemshave,onaverage,thefewestsuchviolations.Verysmallsystemstendtohavehigher numbersofmonitoringandreportingviolationsthanlargersystems.Theaveragenumber decreasesthroughmediumsystemsandthenincreasesforlargeandverylargesystems. Treatmenttechniqueviolationsaregenerallytheleastcommonviolationtype.Again,Native Americansystemsarethemostfrequentviolators,followedbylocalsystemsandthenfederal. Privateandstate-ownedsystemsreportthefewesttreatmenttechniqueviolations. Thetablealsorevealsthatviolationfrequencyvariesacrosswatersource.Systemsthatrely primarilyonsurfacewaterhavethemostcontaminantviolations,andonaveragehaveabout50% moreviolationsthansystemsthatprimarilyusegroundwater,whichhasthesecond-highest numberofviolations.Groundwaterundertheinfluenceofsurfacewater,meanwhile,generates themostmonitoringandreportingandtreatmentviolations.Thesetabulationsemphasizethe importanceofcontrollingforwatertypeinanyanalysis. 4.1.Econometricanalysis Asdiscussedabove,thedataincludethenumberofviolationseachyear,andwedonothave dataontheseverityofeachindividualviolation.Asaresult,thedependentvariable(violations)is

S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205195 countdata,andshouldbemodeledbyanappropriatediscretedistributionthataccountsforthe largenumberofzeroesandsmallvalues.InthecaseofviolationsofSDWArules,itseems appropriatetoexpectthat,forsmalltimeintervals,theprobabilityofaviolationisproportionalto thelengthofthetimeinterval.ThissuggeststhatthedependentvariableisdrawnfromaPoisson distributionwithparameter λ it : 22 e −k it k y it Pr ð Y it ¼ y it Þ¼ it !y ti Here,y it isthenumberofviolationssystemireportsinyeart, ln k it ¼bVx it ; and x it isavectorofindependentvariables Itthereforefollowsthat ; E ½ y it j x it ¼ Var ½ y it j x it ¼k it ¼ e bxV it : ð ð ð 1 2 3 Þ Þ Þ ThePoissondistribution,however,assumesthatthemeanofthevariableequalsitsvariance. Inourcase,thevarianceofeachdependentvariableofinterestfarexceedsthemean, suggestingoverdispersioninthedataandthatanegativebinomialdistribution(forwhichthe varianceisaquadraticfunctionofthemean)ismoreappropriate(Greene,1993;Longand Freese,2003). 23 ThenegativebinomialregressionissimilartothePoissonregressionexceptthat 42 k it ¼ exp ðbVx it þe it Þ andE ðe it Þ¼ 0 :ð 4 Þ Asnotedabove,y it isthenumberofviolationsinsystemiinyeart.Wefocusoncontaminant levelandmonitoringandreportingviolations,asthenumberoftreatmenttechniqueviolationsis toosmalltogenerateconsistentresults.Thevector x it includessystemsizeanddummyvariables indicatingownershiptype.Becauseofthedifferencesbysysteminsizeandownership,wealso interacttheownershipvariableswithfivesizecategorydummies.Becausewaterqualityandthe economicsofdrinkingwaterprovisioncanbeaffectedbythephysicalandtopographical characteristicsofthewatersystem'slocation(Noll,2002),wealsocontrolforlocationfixed effects-dummyvariablesindicatinginwhich “ core-basedstatisticalarea ” (CBSA)thewater systemislocated. 25 Allthefixedeffectdummiesequalzeroforwatersystemsnotlocatedinany CBSA.Wealsoincludedummyvariablesindicatingthewatersource(e.g.,groundwateror 22 SeeGreene(1993),pp.676 – 679. 23 Themeansandvariances(mean,variance)oftheviolationsare:contaminant(0.10,0.27),treatment(0.05,0.23), monitoring(1.14,54.69). 24 SeeLongandFreese(2003)foradescriptionofthenegativebinomialregressionmodel. 25 County-levelorevensystem-levelfixedeffectswouldbesuperiortoourCBSAfixedeffects.Unfortunately,those provedimpossiblegivenourresources.First,AllisonandWaterman(2002)pointoutthattheSTATAprocedurefor estimatingthefixedeffectsnegativebinomialregressiondoesnotactuallycontrolproperlyforthefixedeffects.Second, giventhatprobleminSTATA,wedidnothaveenoughcomputerresourcestoincludeall3141countyfixedeffectsinthe equation,letalone50,000+watersystemdummies.However,includingfixedeffectsmanuallyratherthanbyusing STATA's “ xt ” commandshastheadvantageofallowingustouseproceduresdevelopedbyLongandFreese(2003)to simulateexpectednumberofviolationsforeachtypeofsystembasedontheregressionresults.

916S.Wallsten,K.Kosec/Int.J.ofInd.Organ.26(2008)186 – 205 surfacewater),asthisaffectsinherentwaterquality,andyearfixedeffectstoaccountforchanges inregulatoryoversightorcompliancethatmightchangeovertime,suchasutilitiesgaining experiencewiththeSDWAovertime. 26 Wealsocontrolforcounty-levelvariables,includingaveragecountyhouseholdincomeand theshareofthepopulationthatlivesinurbanareas.Thesevariablescomefromthe2000U.S. Census,whichmeansthattheyunfortunatelyvaryonlyovercounty,andnotovertime.Income andurbanizationdonotchangeradicallyinashortperiodoftime – anddifferencesacross countiesprobablychangeevenless – sothesearereasonableproxiesforincomeandurbanization eveninyearsotherthan2000. Finally,whileweknowofnopipedwatersystemsdirectlycompetingwithoneanotherfor givenhouseholds,theexistenceofalargenumberofwatersystemsinagivenregionmay facilitatebenchmarkcompetition.Weconjecturethatutilitiesfacemorepressuretoperformwell incountieswithnumerousutilitiesbecauseitiseasierforconsumersandregulatorstocompare utilitiesingeographicallysimilarareasandmoredifficultforutilitiestoblamepoorqualityon factorsbeyondtheircontrol(e.g.,geographicfactors). Tocapturethispossibility,wecreateaHerfindahl – HirschmanIndex(HHI)tomeasure concentrationinthewatersystem “ industry. ” Wedefineacountyasamarketandtheshareofthe totalconnectionsinacountythatawatersystemservesasitsmarketshare.Acompletely monopolisticmarketwouldyieldanHHIof10,000,andtheU.S.JusticeDepartmentdeemsan industrytobeheavilyconcentratediftheHHIexceeds1800.Becausepipedwatersystemsdonot competewithinthesamegeographicspaceandcustomersdonothaveachoiceofwaterprovider shortofmovingtoanewwaterdistrict,amoretypicalHHIcalculationwouldimplyacompletely monopolisticmarket.However,weconsiderourmeasureofHHIappropriateforexploring benchmark,ratherthanhead-to-head,competition. Evenourmeasureofbenchmarkcompetition,inwhichamarketisacounty,suggestsa highlyconcentratedmarket.TheaverageHHIacrossallU.S.countiesis4224 — morethan twicetheJusticeDepartment'sthresholdfora “ heavilyconcentrated ” industry.Still,nearlyone- thirdofthepopulationlivesincountieswithatleastmoderatebenchmarkcompetition (HHI ≤ 1800). 4.2.Results — nointeractions Table4showsboththecoefficientestimatesandtheincidentrateratios(IRRs)resulting fromestimatingthisequationwithouttheinteractionvariablesandwithrobuststandard errors. 27 Thetableshowsthatcontrollingforsize,watersource,location,year,income,and urbanization,privately-ownedfirmshavefewercontaminantviolations,butmoremonitoring andreportingviolations,thandomunicipally-ownedfirms.TheHHIispositivelycorrelated withviolations,suggestingthatthenumberofviolationsdecreaseswhenthereismore competition,consistentwiththehypothesisthatcompetition – eveninitsweakbenchmark form – improvesoutcomes. Thetablealsoshowsthat,ingeneral,thenumberofviolationsofeithertypegenerally decreasesassystemsizeincreases.Thiscorrelationmayreflectthegreaterresourcesavailableto 26 Wedropsystemswhoseownershipstatusis “ unknown ” and,unfortunately,systemswhosestatusis “ public/private ” becausethoseareeffectivelyunknownaswell,asdiscussedabove.EPAofficialshadnopriorsastowhichownership typethesewerelikelytobe. 72 Weadjustedthestandarderrorsforintra-CBSAcorrelation.