Selected Economic Aspects of Water Quality Trading: A Primer and Interpretive Literature Review

Transcription

1 January 2007 E.B Selected Economc Aspects of Water Qualty Tradng: A Prmer and Interpretve Lterature Revew Rchard N. Bosvert, Gregory L. Poe, and Yukako Sado Department of Appled Economcs and Management College of Agrculture and Lfe Scences Cornell Unversty Ithaca, NY

2 Selected Economc Aspects of Water Qualty Tradng: A Prmer and Interpretve Lterature Revew Rchard N. Bosvert, Gregory L. Poe, 1 and Yukako Sado Department of Appled Economcs and Management Cornell Unversty I. Introducton and Motvaton The economc objectve of envronmental regulatory desgn s to acheve a targeted reducton of polluton at the lowest possble cost. 2 Conceptually, ths cost effectveness objectve can be acheved through voluntary adopton of new technologes, strct regulatory approaches, and/or market-based ncentve programs such as tax/subsdy mechansms or marketable permt systems. Encouraged by the sngular success of the hghly vsble Unted States (U.S.) acd ran program n acceleratng reductons n sulfur doxde polluton, whle also provdng substantal savngs relatve to alternatve command-and-control measures, envronmental polcymakers are ncreasngly drectng ther attenton to market-based approaches that allow flexblty n meetng caps set on aggregate polluton levels. That the longstandng academc nterest n polluton tradng s beng translated nto actual water polcy s evdent n the rapd growth n the number of water qualty tradng ntatves over the past ffteen years: n 1990 three such programs exsted n the Unted States, by 1999 there were 25 programs n varous stages of mplementaton and development, and as of 2004 ths number had ncreased to 70 (Woodward, Kaser and Wcks; Envronomcs; Breetz et al.). Water qualty tradng programs have been establshed nternatonally n such dverse locatons as Australa, Chle, Chna and Slovaka (NCEE). 1 Durng a porton of the tme that ths report was beng prepared, Poe also served as a Vstng Fellow at the Crawford School of Economcs and Government, Australan Natonal Unversty. The authors also benefted, wthout mplcaton, from helpful dscussons wth Rchard Woodward and Josef Kardos. 2 Here we assume, as s approprate for bophyscally determned total maxmum daly loads (TMDLs) appled to water qualty at a watershed level, that the target level of ambent water qualty, and hence total emssons and total abatement of emssons, are exogenously determned through poltcal or other processes. Gven ths predetermnaton, economc effcency s equated wth cost-effectveness. Ths condtonal cost-effectveness benchmark, however, devates from a great body of envronmental economcs lterature drected toward dentfyng the socally optmal level of polluton (or polluton abatement) n whch optmalty s dentfed to be the level of polluton n whch the margnal benefts of polluton equal the margnal costs (see Baumol and Oates, 1971, 1988). Snce explct consderaton of relatve benefts and costs s precluded n the Clean Water Act decson-makng processes, we nstead make the cost-effectveness crteron paramount n our dscusson. It s ths cost-effectveness crteron that underpns the arguments made by Dales n hs orgnal proposal for tradng polluton rghts: economc analyss, whch s all but useless n helpng us to decde on a polcy, s all but ndspensable n helpng us decde on the best way of mplementng a polcy once t has been chosen. The crteron s smply that the best way of mplementng a polcy s the least costly way (1968a, p. 99). As dscussed n footnote 9 below, ths cost-effectve orentaton s further bolstered by Montgomery s semnal theoretcal paper and subsequent work n ths area.

3 2 In the Unted States, the growth of tradng ntatves has been fostered by the U.S. Envronmental Protecton Agency s (EPA s) explct endorsement of water qualty tradng. In 1996 the U.S. EPA ssued a Draft Framework for Watershed-Based Tradng to encourage tradng and to assst n evaluatng and desgnng tradng programs (U.S. EPA 1996). A fnal Water Qualty Tradng Polcy was promulgated n 2003 based on the ratonale that market-based approaches such as water qualty tradng provde greater flexblty and have potental to acheve water qualty and envronmental benefts greater than would otherwse be acheved under more tradtonal regulatory approaches (U.S. EPA 2003, p. 1). The US EPA presently provdes substantal funds to support the development of watershed-based water qualty tradng programs through ts Targeted Watersheds Grant Program (US EPA 2006). Despte ths keen nterest n such programs, the recent ncrease n the number of ntatves, and expanded avalablty of fundng, the promse of the sgnfcant cost savngs for nutrent tradng programs has never materalzed. Wth the excepton of an extremely lmted number of oft-mentoned successes (e.g., Long Island Sound), very few trades have actually occurred (Kng and Kuch). Tetenberg and Johnstone (2003) observed that, []n a general sense, programs targetng water polluton control have generally not been successful (p. 33). Speakng more specfcally about nutrent tradng, Faeth remarked: There are some brght spots, but overall I thnk real progress has been spotty. There s some good news n a few places, but t s not clear that tradng has actually resulted n sgnfcantly mproved water qualty yet. There have been a few trades [But,] [a]re there any functonng markets up and runnng? The answer s no (p. 1-2). 3 Thus, rather than provdng the bass for a how to template from a stream of successfully mplemented programs, the exstng water qualty ntatves are perhaps best vewed as plot programs provdng an mportant range of lessons learned (Keser and Feng). The paramount lesson, as Hoag and Hughes-Popp argued some tme ago, s that translatng theory nto practce may necesstate a reexamnaton of the man prncples assocated wth water polluton credt tradng theory to dentfy factors that nfluence program feasblty (p. 253). If such efforts are undertaken, Kng mantans that, the potental for [water qualty] tradng mght be realzed. If not [water qualty] tradng wll probably end up n the overflowng dustbn of well-ntentoned economc polces that attracted attenton for a whle but never delvered. In hs smlarly guarded conclusons regardng the future of polluton tradng, Tetenberg (2004) argued that the evdence seems to suggest that tradable permts are no panacea, but they do have ther nche (p. 416). Drawng on these lessons learned, specfcally the admonton by Kng and Kuch that the basc economc obstacles pose a greater problem for tradng programs than nsttutonal 3 To avod quotng out of context, Faeth s comments were drected prmarly at pont-non-pont tradng. Nonetheless, Faeth s dsmal observatons apply equally to pont-pont source water qualty tradng (Kng and Kuch).

4 3 desgn, we focus much of ths revew on the basc economc prerequstes necessary to mplement a tradng program. Brefly, these can be stated as: 1) there must be an economc demand for permts; 4 2) after accountng for the bophyscal fate and transport relatons, there must be clear gans from trade dervng from dfferences n cost structures between frms n the tradng scheme; 3) there must be a tradng system establshed that accounts for the bophyscal and economc characterstcs of the pollutant and ecosystem under consderaton; and 4) there needs to be a clear authorty to trade, legal foundatons to enforce property rghts exchanges related to polluton permt or credt tradng, and a market structure to facltate trades. Wth an eye toward the phosphorus tradng program beng nvestgated for the upper Passac Rver Basn, we drect the dscusson toward selected economc aspects of pontpont source water qualty tradng programs. In Secton II, we revew the necessary market condtons for trades to occur, wth an emphass on dentfyng the basc demand and supply condtons. In Secton III, we outlne a tradng-rato system desgned recently to provde a least-cost soluton to achevng water qualty targets for a non-unformly mxed assmlatve pollutant. One dstnct advantage of ths tradng-rato system for water qualty tradng s that t explots the property that water flows downhll. Assumng that these market prerequstes are met and that t s possble to mplement a tradng system amenable to the characterstcs of the watershed, we turn our attenton to the range of applcable market structures n Sectons IV and V. We brng together our observatons, summarze our arguments, and draw some polcy conclusons n the fnal secton. II. Market Foundatons: Demand, Supply, and Gans from Trade To desgn a market for polluton tradng, t s crtcal at the outset to dstngush between the famlar concept of a well-functonng commodty market wth that of a nascent polluton tradng program. From an economc perspectve, a market s a predctable socal and legal arrangement that facltates the nteractons among buyers and sellers to communcate and share nformaton effcently and to carry out voluntary exchange. Markets are embedded n, and arse from, a nexus of legal and socal foundatons (Bromley, 1989, 1997). Yet, beyond establshng these prerequste condtons and related enforcement, the trade regulator plays no actve role n conventonal commodty markets; all that s needed are wllng buyers and wllng sellers. In such markets, wllng buyers (ndvduals and frms) demand goods at varous prces based on ther own preferences or proft motves. The supply of goods at varous prces derves from the cost of producton actvtes. 4 Throughout ths report, we refer to emssons permts (or allowances), wth each permt provdng the rght to emt a specfc amount of pollutants. An alternatve tradable commodty s polluton credts. Under a credt system, each dscharger s allowed a certan amount of emssons. Should the actual level of pollutants emtted by the dscharger fall below the amount allowed, the dscharger earns a credt whch can be traded to other dschargers seekng to exceed ther specfed allocaton of polluton. Whle the specfcs of allowance and credt programs dffer, the general economc prncples dscussed here wth respect to permts apply to both settngs.

5 4 By stark contrast, the role of the trade regulator n polluton markets extends beyond the purvew of establshng and enforcng the rules of the game. In such markets t becomes mportant to vew trades as three-party transactons nvolvng actve partcpaton among buyers, sellers, and trade regulators (Kng and Kuch, p.10353). The regulators actons are central to creatng scarcty, and wthout scarcty, there s no demand. Thus, absent the trade regulator, there would be no commodty called polluton abatement, because such markets do not arse naturally (Woodward and Kaser). Also, the trade regulator must balance the objectve of cost-effectveness wth that of assurng that envronmental targets are reached (Kng and Kuch; Woodward, Kaser and Wcks). Greater socal demands engender more strngent regulatons, rase the costs to regulated frms, and, as a result, are often met wth ncreased resstance by those managng the sources of pollutants n ther attempts to mollfy regulatory efforts (Hahn and Hester; La). For pont sources of water polluton, the demand for tradable polluton permts s supported by lnkage between two regulatory mechansms: the Total Maxmum Daly Loads (TMDL s) and the Natonal Polluton Dscharge Elmnaton System (NPDES). A TMDL defnes ambent water qualty n the aggregate, specfyng a quanttatve expresson of the amount of pollutant a waterbody can receve (loadng capacty) wthout causng mparment of the applcable water qualty standard for any porton of that water body. By establshng ths baselne relatve to the status quo wth correspondng wasteload allocatons to pont sources (and load allocatons f expanded to nclude nonpont sources), the TMDL defnes the aggregate, locaton-specfc amount of polluton reducton that must take place. The NPDES permts are then assgned to ndvdual pont sources, defnng the amount and the condtons for dscharge of pollutants nto waterbodes. Once a TMDL s n place, ndvdual NPDES permts should reflect the aggregate waste load allocaton. The relatonshp between the TMDLs and the NPDES permts can stmulate demand for permts n the followng way. Under Prncple 3 of the Draft Plan (US EPA 1996), trades are to be developed wthn a TMDL or equvalent analytcal or management framework. Further, the US EPA supports the ncorporaton of specfc trades nto the ndvdual NPDES permts or the use of flexble approaches for the ncorporaton of provsons for tradng n NPDES permts. Together wth the assumpton that reducng waste s costly, the TMDL and the correspondng dscharger-specfc NPDES wasteload allocatons engender a wllngness to pay on the part of some pont sources to exceed allocated levels. The specfcatons wthn the NPDES permts provde the legal foundatons to ensure that pont-pont source trades are legally enforceable. In essence, under a TMDL program, the NPDES permt has the potental to evolve from ts tradtonal technology-based crtera of How much should each source be allowed to emt? to How should the total polluton load be allocated among varous sources? (Woodward and Kaser). These condtons consttute the frst corner stone of the market demand from wllng buyers s assured. Ths stands n stark contrast to pont-nonpont source tradng n whch

6 5 the vablty of such trades s serously undermned because there are no mechansms to guarantee the credts purchased from non-pont sources (Faeth; Keser and Feng). Gven a demand for permts, the potental for market success of pont-pont source trades rests on two addtonal condtons: an effectve supply of permts and the establshment of effectve tradng nsttutons. Assumng proft, or at least cost-mnmzaton motves, frms would be wllng to trade away ther polluton rghts only f they stand to gan monetarly from such a trade. For a transacton to take place, there must be a suffcent dfference n abatement costs across frms for there to be sgnfcant gans from trade from exchangng rghts to pollute. a. Gans from Trade: Margnal Abatement Costs for Unformly-Mxed Pollutants. As underscored n the openng secton of ths revew, the economc objectve of envronmental regulatory desgn s to acheve a targeted level (A*) of polluton (or equvalently polluton abatement R*) at the lowest possble cost. Usng economc optmzaton methods (see Appendx A), and assumng contnuty n abatement costs, the soluton of ths problem nvolves the applcaton of what envronmental economsts refer to as the equmargnal condton that s, the desred level of polluton abatement s acheved at the pont where the margnal abatement costs are equated across all frms. Margnal abatement cost refers to the cost of mplementng one more unt of emssons reducton, where the unt s small, such as a pound of nutrent. For example, an extra pound of reducton by a pont source could be acheved by ncreasng the amount of chemcals used n ts treatment process (U.S. EPA, 1996). The margnal abatement cost n ths smple example would be the cost of the addtonal chemcals. In market tradng, the nature of dfferental margnal abatement costs across regulated enttes plays a fundamental role n determnng the drecton of trade, the number of permts traded, and the magntude of gans from tradng polluton permts. Two features are partcularly mportant. The frst s that for a fxed level of captal nvestment, margnal abatement costs tend to rse wth successve levels of abatement. Wth chemcal treatment processes, for example, the effectveness of each small addton of chemcals beyond a certan pont s expected to dmnsh, rasng the margnal abatement cost of each successve pound of pollutant abated. The second crtcal feature s that cost savngs from trade ncrease wth the dvergence of the margnal abatement cost functons across dschargers. Put dfferently, f all abatement cost functons were the same across dschargers, we would expect no gans from trade. Alternatvely, f there s substantal dversty n margnal abatement cost functons, gans from trade are expected. For nstance, there mght be some dschargers for whom t s costly to reduce pollutant levels, whle for others, ther present technology may be such that addtonal reductons n pollutant levels could be acheved at relatvely low cost. Through a smple example, we can effectvely llustrate how ths heterogenety n cost functons, along wth the applcaton of the equmargnal prncple, can lead to gans from trade relatve to conventonal command-and-control approaches. To do so, consder the

7 6 par of watershed settngs presented n Fgure 1. 5 In each watershed, we assume that the pollutants dscharged have equal effects at the pont of measurement, whch we shall assume s mmedately downstream from Dscharger D. Ths may be due to adjacences of dschargers, or the pollutant could be characterzed by unform mxng, such as s the case of greenhouse gas emssons. 6 The frst column n each of the sub-tables ndcates the total number of pounds of polluton abated by each dscharger. The second and thrd columns provde the correspondng margnal costs of abatng each addtonal unt for Dscharger D and Dscharger U, respectvely. For example, the margnal abatement cost of the fourth pound abated n the homogeneous (heterogeneous) settng for Dscharger D s $8 ($12). Pror to any water qualty control polcy, each dscharger emts sx pounds of pollutants. Now, suppose that the regulatory agency desres to reduce aggregate levels of polluton at the pont of measurement mmedately below dscharger D to 6 pounds of pollutants. In a watershed where the pollutants are perfectly mxed, ths s equvalent to cuttng the total pollutants n half (from 12 to sx pounds). To provde a bass for assessng the gans from trade, we frst assume that no trades are possble; and that abatement responsbltes are to be allocated equally across dschargers. Under these condtons, the total abatement costs n each watershed would be $32: n the watershed where costs are homogeneous each dscharger would ncur total abatement costs of $16 [= $3 (lb. 1) + $5 (lb. 2) + $8 (lb. 3)]; n the watershed where costs are heterogeneous, Dscharger D would ncur total abatement costs of $19 [= $2 (lb. 1) + $5 (lb. 2) + $12 (lb. 3)], whle the frm wth a lower abatement cost schedule, Dscharger U, would ncur costs of $13[= $4 (lb. 1) + $4 (lb. 2) + $5 (lb. 3)]. In both watershed settngs, the aggregate abatement costs would be equal $32, and the amount of abatement by each dscharger would be equal three pounds (resultng n the desred level of sx pounds of pollutants). These abatement costs would be shared equally ($16) by the two dschargers n the homogenous case. If the costs are heterogeneous and wthout trade, Dscharger D would dscharge the same number of unts as Dscharger U, but would ncur about 60 percent ($19) of the total $32 watershed abatement costs. 5 To facltate the presentaton, we examne a stuaton where both abatement amounts and the cost of each ncrement of abatement are reported n dscrete amounts. In most standard elementary or ntermedate economcs texts, smlar examples are presented graphcally utlzng contnuous margnal abatement cost functons. Contnuous cost functons are also utlzed n the mathematcal appendces n ths revew. See Tetenberg (2006) and Sado for more detals. 6 Theoretcal presentatons of permt tradng programs further dstngush between assmlatve pollutants and accumulatve pollutants. Wth our charge of addressng phosphorous polluton n the upper Passac Water Basn, our focus s on assmlatve pollutants. See Tetenberg (2006) for a treatment of polluton tradng permts for accumulatve pollutants.

8 7 Fgure 1. Margnal Cost Functons and Gans From Trade Watershed wth Homogeneous Costs Watershed wth Heterogeneous Costs # of lbs. Margnal Abatement Costs ($/lb.) # of lbs. Margnal Abatement Costs ($/lb.) Abated Dscharger D Dscharger U Abated Dscharger D Dscharger U The mportant queston to ask at ths pont s: Would t be possble to mprove upon ths regulatory approach n terms of cost-effectveness? In the homogeneous case, the answer s no (try any other combnaton of abatement levels by the two dschargers that sum to sx pounds abated and compare the aggregate costs to $32). Snce the abatement cost functons are dentcal for both frms, the equmargnal condton s automatcally satsfed at R* equals sx, wth each frm abatng three unts, and the margnal cost of the last unt beng $8. Hence there s no way to allocate abatement responsbltes n a manner that lowers the aggregate abatement costs for the watershed. In contrast, because the abatement schedules dffer across the two frms n the heterogeneous-cost watershed, there are potental gans from trade relatve to the equal abatement regulatory status quo. In ths case, t s the structure of abatement costs that creates the opportunty for the dscharger (Dscharger D) wth relatvely hgh abatement costs to pay the relatvely low-abatement-cost dscharger (Dscharger U) to undertake some of the hgh dscharger s abatement dutes. In dong so, trades can be undertaken such that each dscharger gans relatve to the regulatory status quo, and the aggregate costs of abatement are reduced. To see how ths works, we need to compare the wllngness to pay to not have to abate the last unt (.e., lb. 3) for Dscharger D wth Dscharger U s wllngness-to-accept compensaton to abate an addtonal unt (.e., lb. 4). Dscharger D s thrd unt of abatement comes at a cost of $12, and hence that dscharger would be wllng to pay up to $12 not to abate ths last unt. Ths s the demand sde of the permt market. Wth respect to the supply of permts, Dscharger U would be wllng to abate an extra unt (e.g. a fourth unt) f compensaton to do so was equal to or exceeded $5. Hence there are clear opportuntes for gans from trade.

9 8 If such trade s permtted, Dscharger D wll pay Dscharger U to abate an extra unt of polluton at a prce lyng between $12/lb. (D s wllngness to pay or the demand prce) and $5/lb. (U s wllngness to accept or the supply prce). Now, f such a trade were arranged at $6/lb., both partes would be better off. Dscharger D would have been wllng to pay $12 but only has to pay $6, resultng n a gan of $5. Lkewse, Dscharger U gans $1 because the amount compensated, $6, exceeds that frm s costs of abatement, $5. In terms or resources devoted to polluton abatement n the watershed, total abatement costs are reduced by $7 wth Dscharger D abatng two unts at a cost of $7 [= $2 (lb. 1) + $5 (lb. 2)] and Dscharger U abatng four unts at a cost of $18 [= $4 (lb. 1) + $4 (lb. 2) + $5 (lb. 3)+ $5 (lb. 4)]. Total watershed abatement cost of $25 for the sx unts abated. Whle the transfer of abatement responsbltes entals a correspondng payment of $6 by Dscharger D to Dscharge U and thus affects the returns of each dscharger, these mones are not regarded as true resource costs to socety from a socal perspectve. They merely represent a transfer of funds from one frm to another. 7 It s mportant to recognze that n ths barganng stuaton the prce of a permt s not a pro determnate, as t depends upon the relatve barganng power of the two dschargers. Note further that after ths ntal one-unt trade there are no further gans to trade n ths watershed. Indeed, the equmargnal prncple s satsfed. Dscharger D s maxmum wllngness to pay ($5/lb.) to not have to abate another unt exactly equals the mnmum compensaton that Dscharger U ($5/lb.) would requre to abate another unt. In relatve terms, the cost-savngs n ths example represent a 22% reducton relatve to the regulatory scenaro, but the savngs of ths magntude are strctly an artfact of the hypothetcal numbers chosen for llustratve purposes. The reader s cautoned that t s mpossble to access whether the 22% reducton demonstrated heren s large or small relatve to the gans that may be realzed n actual tradng programs. All that we know s that ths fgure s n the range of the cost savngs based on ex ante comparsons of potental least-cost emsson reductons wth the costs assocated wth command andcontrol allocatons. These ex ante analyses typcally utlze hypothetcal smulaton technques based on engneerng models and assumed optmzng behavor. In a revew of such studes for ar polluton regulaton, Tetenberg (2006) fnds that the potental cost savngs from adoptng least-cost approaches range from 6 to 95 percent of the commandand-control benchmark (p ). Wth respect to pont-pont source water qualty tradng, separate efforts by Rowles and Sado suggest that cost-savngs wll le at the lower end of the range of the potental savngs dentfed by Tetenberg. We conjecture 7 Bear n mnd that n adoptng a cost-effectve approach, one must recognze that there may be other socal crtera that are also mportant n evaluatng trades, tradng systems, and tradng structures as these elements affect the dstrbuton of polluton and the burden on consumers and taxpayers. Smlarly, potental envronmental justce ssues have been dscussed n general n Drury et al. and Solomon and Lee, and, wth specfc reference to water qualty tradng n Goldfarb.

10 9 that ths relatvely low level of savngs s due to the fact that abatement technologes of the sources consdered n these studes are relatvely homogenous. Unfortunately, as s evdent n both Tetenberg s revew of actual ar emsson programs and Kng and Kuch s assessment of water qualty tradng efforts, the projected cost savngs are typcally nether realzed nor approxmated n practce because of the behavoral responses of partcpants, market mperfectons, and other constrants nhbtng trade. Indeed, Kng and Kuch s ex-post assessment suggests that the cost savngs from most efforts to establsh water qualty programs approach zero. As Woodward notes wth few trades t seems lkely that the cost reductons generated through tradng are mnmal. Ths dsparty between ex-ante projectons and ex-post assessment s evdent n Smth s reflecton on the earlest applcaton of water qualty tradng, the Fox Rver tradng scheme ntroduced by the Wsconsn Department of Natural Resources n 1981: Ths scheme, coverng a number of paper mlls and muncpal wastewater treatment plants dschargng effluent to a heavly polluted stretch of the Fox Rver, had the potental to acheve consderable cost savngs compared to a fxed central drectve requrng unform emssons. A smulaton study by O Nel, Davd, Moore and Joeres (1983) showed that there would be consderable scope for reducng total abatement cost through the allocaton of emssons between sources. Snce abatement costs dffered between sources by a factor of four, there would be consderable cost savngs from such a reallocaton. The fact that, n practce, the ntroducton of the scheme was followed only by a sngle trade, stood n stark contrast to the predcted theoretcal gans, and provded a clear warnng that there would be consderable gap between advance forecasts and practcal outcomes. (p. 29) Furthermore, even for successful tradng programs, ex-post studes of actual cost savngs are lmted (Tetenberg, 2006) and, for those that have been undertaken, there appears to be a lack of consensus regardng the magntude of actual cost-savngs. For example, due to the varyng crtera used, cost savngs examned, counterfactual baselne cases, and other methodologcal dfferences, estmates of the cost-savngs from the much-touted U.S. acd ran tradng program under the 1990 Clean Ar Act Amendments range wdely, from 16%-25% (Keohane), to 25%-34% (Schmalensee et al.), to 43% (Carlson et al.). In the dscusson thus far, we have characterzed gans from trade n the most drect terms the case where hgh-margnal-abatement-cost dschargers pay low-margnalabatement-cost dschargers to undertake addtonal abatement. Applyng these concepts to tradng permts s easly done, smply by recognzng that nstead of specfyng polluton reducton n terms of levels of abatement, one can also express ths objectve n terms of polluton allowed. Hence, f we assume n our example that there are 12 unts of polluton at present,.e. sx from each frm, R* could be expressed not as sx unts abated, but as sx unts of allowable emsson E*. Hence sx unts of emssons permts wll be dstrbuted wthn the watershed, whch for drect comparsons wth our prevous

11 10 dscusson, we assume are equally dstrbuted across dschargers. 8 In the heterogeneous settng, the two dschargers wll trade n a manner consstent wth that prevously dscussed so that Dscharger D abates two (emts three) pounds n the least-cost equlbrum, whle Dscharger U abates four (emts one) pounds. Thus, Dscharger U would sell ts polluton permt to Dscharger D. b. Gans from Trade: Non-Unformly Mxed Pollutants, Accountng for Spatal Dstrbuton The assumpton of unformly mxed pollutants s useful for presentng basc concepts. Ths assumpton too s relevant to some real world cases such as greenhouse gas emssons, and has also been used as an approxmaton n a number of ar polluton tradng programs (Tetenberg, 1985, 2006). For many pollutants and meda, however, such an assumpton s smply napproprate (Tetenberg, 1978, 1980, 1995). For nstance, for nutrent management at the watershed level, the spatal dstrbuton of dschargers relatve to receptor stes s crtcal to program desgn because the fate and transport of the pollutants must be consdered explctly. Due to dluton, dsperson, and other bophyscal nteractons, the mpact on ambent levels of a pollutant at a gven receptor are expected to declne as the dstance between the dscharger and the receptor ncreases. At the extreme, receptor stes upstream wll be unaffected by the downstream dscharger s emssons. Fortunately, the ssue of non-unform mxng and spatal dstrbuton of pollutants s easly accommodated nto the prevous framework (for a mathematcal treatment, see Appendx B). We can do so by defnng a dffuson (or transfer) coeffcent, d j, that measures the contrbuton of one unt of emssons from the th dscharger to the total load of effluent at the jth receptor (Montgomery; Tetenberg, 1978; Hung and Shaw). Formally, let e ndcate an amount of emssons from source, and let e j ndcate the correspondng amount measured at the jth receptor after dscharger emts e. Then, e d = j j e Ths coeffcent s neccessarly bounded between zero and one, where zero ndcates that the th dscharger has no effect on the jth receptor (as n the case of beng upstream) and one ndcates that the unt of polluton from the th source does not dmnsh n any way by the tme t reaches the jth receptor. An ntermedate coeffcent of, say, d j = 0.5 would ndcate that one addtonal unt of polluton for dscharger results n one-half a unt of polluton at receptor j. For a regon or watershed wth I statonary sources of pollutants and J receptor ponts, the dsperson of water emssons for the I sources can be specfed by an I by J matrx of (lnear) unt dffuson coeffcents (McGartland and Oates): 8 In our example, any other ntal allocaton of four polluton permts across frms wll, after tradng, result n the same fnal least cost result ndcated n the text. However, snce the ntal allocaton defnes who has to pay whom, there wll be ncome effects.

12 11 D = d j When these spatal condtons must be accommodated, the prevously descrbed equmargnal prncple must be modfed:..t s not the margnal costs of emsson reducton that are equalzed across sources n a cost-effectve allocaton (as was the case for unformly mxed assmlatve pollutant) t s the margnal costs of polluton reducton at each receptor locaton that are equalzed (Tetenberg, 2006 p. 34). Ths relatonshp s most readly understood when there s a sngle receptor ste, such as that used n the Long Island Sound n whch an equalzaton factor s used to relate each faclty s geographc locaton to ts relatve mpact on the oxygen levels n Western Long Island Sound where the mpact of excess ntrogen s most severe (Connectcut Department of Envronmental Protecton). Smlarly the Mnnesota Rver tradng program (Hall) and the proposed Lower Bose rver tradng program (Schary) defne the trades and tradng ratos throughout ther respectve watersheds n terms of the mpact at a sngle pont of egress. To add further to an understandng of ths modfcaton n the equmargnal prncple when there n non-unform mxng, we contnue the prevous example of the heterogeneous watershed wth two dschargers. The left-hand sde of Fgure 2 presents the Heterogeneous Watershed from Fgure 1, but wth two notable, albet mnor changes. The frst change, ndcated n the ttle of the left-most column s that, the quantty of pollutant s measured n the area of Dscharger D The second modfcaton s that dffuson coeffcents are now specfed. In ths case, the dffuson coeffcents are both equal to one (e.g. d DD = d UD = 1), whch s equvalent to unform mxng. As such, the cost-effectve outcome of reducng sx unts of aggregate emssons s dentcal to that dscussed prevously: Dscharger D reduces polluton by two pounds and the lower cost Dscharger U reduces polluton by four pounds. Recall that ths allocaton of abatement reduces aggregate polluton costs relatve to an equal abatement command-and-control approach. Compare ths result wth the watershed condtons descrbed on the rght hand sde of Fgure 2, characterzed by non-unform mxng. Here the dffuson coeffcent for Dscharger D equals one (e.g. d DD =1) as would be expected because water qualty measurements are agan taken n the area of Dscharger D. However, the dffuson coeffcent, d UD = 0.5. Ths mples that for every unt of polluton reduced by Dscharger

13 12 U (where U represents upstream) there s only a 0.5 unt reducton n polluton at the downstream receptor. The effectve margnal abatement cost of the nth unt of polluton reducton at receptor ste D by Dscharger U s that assocated wth the 2*nth unt reducton by Dscharger U. Stated dfferently, to reduce one unt at D, Dscharger U must reduce two unts and ncur the addtonal cost of abatng both unts. Ths effectve margnal cost s presented n the fnal column of Fgure 2. The last three entres n ths column are empty because to acheve these levels of abatement at the pont of measurement would requre U to abate more than t ntally pollutes. We assume that ths s mpossble n ths settng, although such an outcome could exst f offsets from other sources are avalable to Dscharger U. Under these condtons, applcaton of the equmargnal prncple demonstrates that the cost-effectve way to reduce sx unts of polluton at ste D s now accomplshed through equal abatement of three unts by each frm. Hence, n the heterogeneous case there are no longer gans to trade relatve to the equal abatement command-and-control regme once spatal dstrbutons are accounted for n ths nstance. Indeed, t s nterestng to note that total abatement n terms of the sum of abatement levels measured at each ste s larger: Dscharger D abates three unts and Dscharger U abates sx unts (equvalent to three unts measured at D). Fgure 2: Cost-Effectve Solutons Under Unform and Non-Unform Mxng Watershed wth Heterogeneous Costs (Unform Mxng) # of Margnal Abatement Costs lbs. ($/lb.) Abated d DD =d UD =1 Watershed wth Heterogeneous Costs (Non-Unform Mxng) # of Margnal Abatement Costs lbs. ($/lb.) Abated d DD =1; d UD =0.5 (@ D) Dscharger D Dscharger U (@ D) Dscharger D Dscharger U Although ths smple example represents an extreme case where the transfer coeffcents elmnate any gans from trade, ths wll not be true n general, However, by consderng ths extreme case, we do underscore the lesson to be learned from our llustraton: that gans from trade depend not only on the dfferences n abatement costs across frms, but also on the relatve dffuson of transfer coeffcents between pars of upstream and downstream stes. In cases such as ths one presented n Fgure 2, these effects can be at odds wth one another. One effect can act to negate the other f lower cost abaters are also characterzed by relatvely low dffuson coeffcents.

14 13 III. Tradng Systems In the prevous secton, we dentfed the fundamental characterstc of cost-effectve polluton abatement--that of equatng margnal abatement costs across frms after accountng for spatal characterstcs and transport. The orgnal suggeston that tradable polluton rghts could acheve ths least-cost allocaton of resources was provded ndependently by Dales (1968a, b) and Crocker. Drawng from Coase s semnal work on property rghts, Dales proposed hs concept for a market for fully transferable polluton rghts wthn the context of water qualty. Crocker s vson of a market prcng system for emsson rghts was orented to atmospherc polluton. 9 a. Ambent Permt Systems and Emsson Permt Systems Motvated by Dales arguments, as well as a closely related polcy proposal to establsh markets n BOD bonds or lcenses for ndustral sources n the Delaware Estuary, Montgomery sought to prove whether or not such conjectures could be supported theoretcally. In hs semnal paper, Montgomery explored whether compettve market equlbrum n lcenses can acheve externally gven standards of envronmental qualty at least cost to the regulated ndustres (p. 396). In explorng lcenses, Montgomery took care to dstngush between polluton and emsson lcenses. The former s assocated wth the rght to emt pollutants n terms of pollutant concentratons at a set of receptor ponts. In the present lexcon of tradable permts, ths s most frequently referred to as an ambent permt system. Emssons lcenses nstead confer a rght to emt pollutants up to a certan rate. Such lcenses correspond to what s now commonly referred to as emsson permt systems. Under an ambent permt system, permts are receptor specfc, wth the allocaton to each receptor ste correspondng to the targeted ambent qualty at that ste. The Montgomery paper provdes a theoretcal proof for the exstence of a compettve market equlbrum, and that moreover, the equlbrum concdes wth the least-cost soluton. For a result that Tetenberg (2003) calls remarkable, Montgomery further proved that ths least-cost outcome s ndependent of how the ntal permts are allocated across dschargers or sources. That s, theoretcally at least, any ntal allocaton rule across dschargers stll engenders the cost-effectve allocaton after tradng. 9 Asde from the pollutant type, the Crocker and Dales presentatons also dffered n terms of the extent of the market. Crocker clearly envsoned a market that would nclude polluters and pollutees, and hence framed hs arguments for a market system to lnk receptors and emtters n terms of optmal levels of ar qualty over tme. As dscussed prevously n footnote 1, Dales orented hs arguments toward provdng a least-cost soluton to achevng a specfed level of polluton abatement. In hs subsequent theoretcal formulaton, Montgomery (1972, p. 395) argued t appears unlkely that markets n rghts, contanng many sufferers from polluton as partcpants, wll lead to overall Pareto Optmalty. They can only serve the more lmted but stll valuable functon of achevng specfed levels of envronmental qualty. Wth some exceptons (e.g. Farrow et al.) economc dscussons subsequent to these ntal efforts have centered on the cost-effectve aspects of tradable polluton permts, mplctly or explctly adoptng the arguments of Dales and Montgomery.

15 14 [T]he logc behnd ths result s rather straght forward. Whatever the ntal allocaton the transferablty of permts allows them ultmately to flow to ther hghest-valued uses. Snce those uses do not depend on the ntal allocaton, all ntal allocatons result n the same outcome and that outcome s cost-effectve (Tetenberg 2003, p. 401). The mportant mplcaton of ths nvarance result from the perspectve of economc theory s that such ndependence mples that there need not be a conflct between poltcal feasblty or equty and cost-effectveness. Further, ambent standards are always met. Unfortunately, Montgomery also demonstrates that ths result does not extend to emssons permts. Instead, [a]n extremely restrctve (and sometmes unattanable) condton s requred to ensure that the market equlbrum s also the least-cost soluton. Ths fndng s partcularly dsturbng on two counts. Frst, the envronmental authorty may not be able to fnd an ntal allocaton of permts that ensures an effcent outcome. And second, should such an allocaton exst, a substantal degree of flexblty n the choce of ths ntal allocaton may be lost. Such flexblty can be extremely mportant n desgnng a system that s poltcally feasble (as well as effcent) (Krupnck, Oates, and Van De Verg, p. 234). To summarze, from a theoretcal perspectve ambent permt systems have an advantage over emssons permts. Beyond theoretcal modelng arguments, however, the relatve advantage s less clear. One practcal concern about ambent permt systems s: [f]or every receptor a separate market would have to be establshed. In order to emt one unt, each source would have to keep the approprate number of dsposton permts for each receptor t affects. If a source wants to ncrease emssons t must obtan addtonal [ambent] permts for each of the receptors ts emsson reaches. (Klaassen, p. 48) Ths multplcty of markets s lkely to create large burdens on each market partcpant who must create a portfolo of permts and engage n smultaneous deals across all of the affected receptor ponts (Krupnck, Oates and Van De Verg; McGartland). Ths contrasts wth emsson permt systems n whch trades smply nvolve one-to-one exchanges of rghts to emt. Hahn (1986) suggests that havng to smultaneously address the consequences on many receptors may lead to thn markets where compettve market prce-takng assumptons can no longer be assumed, even as a rough approxmaton. 10 Ths s of partcular concern to water qualty tradng as trades are lmted to ndvdual watersheds or sub-watersheds (Woodward and Kaser). Whle attenton to ths ssue has centered on ambent polluton systems, related concerns about susceptblty to market manpulatons have smlarly been expressed for emssons permt systems (Hahn, 1986). 10 Tetenberg (2006) provdes a comprehensve revew of the related ssue of market power n a chapter devoted to that topc.

16 15 Concerns have also been rased about the relatonshps between each permttng approach and realzed water qualty. Wth respect to ambent standards, usng the Fox Rver Dscharge Program as a case study, Devln and Grafton argue that property rghts assocated wth ambent polluton permts are poorly defned unless ambent qualty relatonshps are clearly specfed between the emssons of a frm, the emssons of other frms, and relevant receptor stes. Because dfferent frms have dfferent mpacts on ambent water qualty due to locatonal or bophyscal ssues, what the seller sells and what the buyer buys s typcally not clearly defned n past applcatons of ambent permt systems to water qualty. On the other hand, to the extent that they rely on one-toone tradng, trades of emssons permts may create unacceptably hgh local concentratons or hot spots n areas near sources that acqure permts nstead of abatng (Tetenberg, 1995). To ensure that no volatons of ambent targets occur, aggregate levels of abatement may thus have to be hgher under emssons permt systems than command-and-control regulatons. In turn, ths suggests that, dependng on local condtons, emssons permt systems may actually ncrease aggregate abatement relatve to command-and-control approaches. Tetenberg (1985) provdes compellng emprcal examples demonstratng ths possblty. b. Tradng Rules: Nether the emsson permt nor ambent permt systems are optmal from all ponts of vew (Atknson and Tetenberg, 1982, p. 103). In an effort to search for more pragmatc alternatve permt system desgns that wll garner the benefts of both the ambent and emsson permt systems whle mnmzng ther respectve shortcomngs, a number of tradng rule systems have been proposed n the lterature. The unfyng feature of these structured rules of trade s that emssons are traded under the constrant that ambent targets are not volated. Klaassen and Tetenberg (2006) categorze these rules as follows: The polluton offset system (Krupnck, Oates and Van de Verg) meets ambent standards by allowng emssons tradng among dschargers as long as they do not volate ambent standards at any receptor. The modfed polluton offset system (McGartland and Oates) nstead only allows trades that meet the followng two condtons: ambent standards and pre-trade water qualty are not exceeded at any receptor. The non-degradaton (or constant emssons) offset system (Atknson and Tetenberg, 1982) only permts trades that meet the followng two condtons: ambent standards are not exceeded at any receptor and total emssons do not ncrease relatve to the baselne. Ths latter approach has, n effect, been appled through regulatory terng n the U.S. acd ran program n whch sulfur emssons are controlled both by the relatons desgned to acheve local ambent ar qualty standards as well as by sulfur allowance tradng rules (Tetenberg, 2006).

17 16 Conceptually, each of these systems s expected to lower transactons costs relatve to the ambent permt system because dschargers have only to be concerned wth receptors that threaten to volate ambent targets, rather than all receptors that are affected by the dscharger s emssons. However, such systems may create prohbtve admnstratve costs f ambent smulaton models have to be conducted before each trade s approved (McGartland, 1988). Ths necessty elevates the costs to dschargers because they do not know the value of the permt beforehand and whether or not, once an applcaton for a trade has been submtted, the trade wll be approved (Hung and Shaw). Further, the rght to ncrease emssons s endogenous or state-dependent:.e., a trade between two sources affects deposton patterns and alters tradng possbltes for other dschargers. Hence prces are endogenous, rasng the costs of prce dscovery. Therefore, n the fnal analyss, these tradng rules may only be useful n cases where, a pror, the regulators have smulated the consequences of trades n areas of concern, f dffuson coeffcents are made avalable and f the number of receptors where volatons are a concern s small and stable, n the sense that the number of potental volatons does not change drastcally wth trades (Klaassen). As demonstrated by the correspondence between the nondegradaton rule and the acd ran program, such polcy condtons may be satsfed n specfc nstances. b. Zonal Tradng Systems An alternatve feasble approach to dealng wth non-unform mxng pollutants s a zonal permt approach: In ths approach, the control regon s dvded nto a specfc number of zones; each zone s allocated a zonal cap. In pure zonal systems, permts can be traded wthn each zone on a one-for-one bass, but tradng among zones s prohbted (Tetenberg, 2006 p. 89). At one extreme, the emssons permt system mentoned above represents a sngle zone system, wth an exchange rate of unty across dschargers. By ncorporatng tradng ratos, defnng tradng zones, or combnng both, we may modfy ths admnstratvely smple system. The sngle market ambent permt system defnes all trades n terms of ther effects on the worst case or most constranng receptor. As ndcated above, such an approach has been utlzed n water qualty tradng n the Long Island Sound and proposed for tradng schemes elsewhere. In these crcumstances, the ambent condtons of the other receptors are presumed to be met, as long as the condtons at ths sngle receptor are met. Emssons permts are traded based of ther ratos of mpacts on water qualty at the crtcal receptor. Such an approach s admnstratvely facle, and provded that the domnance of the crtcal receptor s assured, s a potental least-cost means of meetng ambent standards. However, f such domnance s not actually the case, n ether the short or the long term, then the emergence of hot spots may requre that total allowable permts must be ratcheted down to avod such outcomes. Ths would reduce any cost advantages from adoptng a smpler exchange program. 11 Or n the case of the Mnnesota Tradng Program, the use of a sngle bndng endpont may have to be reconsdered (Hall, personal comments accompanyng a formal presentaton). 11 Tetenberg (2006) also rases long-run equlbrum concerns assocated wth the possblty that frms wll relocate to dstant ponts wheren the effect of emssons on the bndng receptor s low.

18 17 Ths tradng of permts wthn several zones has certan appeal on the surface. It allows the regulatory agency to choose the sze and locaton of zones, takng nto account the spatal dstrbuton of sources and possble dfferences n allowed [ambent] levels. Therefore t offers more protecton for [ambent] targets than emsson tradng n one zone and reduces control costs. However, ths s only the case when the envronmental agency has complete and correct knowledge of control costs. Wth lmted nformaton, the cost wll be above the cost mnmum because emssons cannot be traded among zones. Ths s because wthout knowng the cost-mnmum soluton, the envronmental agency does not know how many permts t should allocate to each zone. In addton, there s no protecton aganst volaton of the standards, even n small zones, because t s not exactly known where emssons wll take place after tradng (Klaassen, p ). Smulatons based on full nformaton about costs of each frm have demonstrated costeffectveness gans relatve to standard emsson permt programs (because n a sngle market wthout tradng ratos dstant sources over-control). However, Tetenberg (2006) concludes that n realstc crcumstances wth only lmted nformaton about dscharger abatement costs, zonal permt systems wth no tradng between zones do not appear promsng [they] do not n general provde much of an opportunty ether to reduce costs or to control the hot spot problem (p. 93). Ths pessmstc concluson has led to proposals to expand beyond no-tradng zones to allow drectonal tradng (as has been done n the RECLAIM program, see Harrson) or rely on exchange rates between zones. We now turn to one such proposal desgned specfcally for water qualty tradng. c. A Tradng-Rato System for Water Qualty Recognzng that [t]he fact that water flows to the lowest level un-drectonally s a very specfc and useful property of water (p. 83), Hung and Shaw desgned a tradng rato system that lnks emsson permt tradng to ambent water qualty. From a theoretcal perspectve, these authors prove that, by assumng cost-mnmzng dschargers and no transactons costs or strategc behavor, ther proposed tradng-rato system s costeffectve for both smultaneous and sequental tradng equlbra. The fact that, theoretcally at least, the cost-effectve outcome can be acheved by sequental trades s partcularly sgnfcant because ths concern has been underscored wth respect to many of the rules and systems (Hahn 1986; Atknson and Tetenberg 1991). The tradng-rato system by Hung and Shaw s based on the assumptons that emssons at a ste can be translated nto ambent concentratons, and that the mpact of these addtonal emssons on affected stes can be expressed fully by the dffuson coeffcents defned above. By defnton, the dffuson coeffcent for an upstream dscharger expressed n terms of a downstream receptor wll be bounded by zero and one.