Overcoming the Common Pool Problem through Voluntary Cooperation: The Rise and Fall of a Fishery Cooperative

Transcription

1 Overcomng te Common Pool Problem troug Voluntary Cooperaton: Te Rse and Fall of a Fsery Cooperatve Robert T. Deacon ±, Domnc P. Parker, and Crstoper Costello ±# ± Unversty of Calforna, Santa Barbara Montana State Unversty # Natonal Bureau of Economc Researc Resources for te Future Abstract We analyze a seldom used, but gly promsng form of rgts-based management over common pool resources tat nvolves te self-selecton of eterogeneous fsermen nto sectors. Te fsery management regme assgns one porton of an overall catc quota to a voluntary cooperatve, wt te remander exploted as a commons by tose coosng to fs ndependently. Data from an Alaska commercal salmon fsery confrm te model s predctons regardng coordnaton of te arvest, nformaton sarng and provson of sared nfrastructure, and mples rent gans of at least 25%. A lawsut fled by two dsgruntled ndependents led to te coop s demse, an outcome also predcted by our model. Our analyss provdes gudance for desgnng fsery reform tat leads to Pareto mprovements for fsermen of all skll levels, wc suggests a structure tat enables reform wtout losers. EL classfcatons: Q22, D23, L23 1. ntroducton Te common pool problem llustrates a key nsgt from economcs tat ndvdually ratonal beavor can lead to a suboptmal outcome. 1 After decades of study and years of practcal experence, tere s broad consensus tat property rgts-based management can solve te Our respectve emal addresses are deacon@econ.ucsb.edu, domnc.parker@montana..edu and costello@bren.ucsb.edu. We acknowledge valuable comments from Don Leal, Randy Rucker, Tm Ftzgerald, Qunn Wennnger, m Wlen, m Roumasset and semnar partcpants at Montana State Unversty, te Unversty of Hawa, UC San Dego, Carleton Unversty, te Unversty of Maryland and Resources for te Future. We are ndebted to Mark Stcert of te Alaska Department of Fs and Game for elpng us acqure necessary data. Ts researc was supported by a grant from te Paul Allen Famly Foundaton. 1 Two semnal contrbutons are Gordon 1954) and Scott 1955). 1

2 common pool problem n a varety of contexts. 2 roncally, owever, lttle progress as been made mplementng ts soluton n te fsery, te conc example of a common pool resource; despte te collapse of economcally mportant fseres around te world, management scemes based on grantng quanttatve rgts to te catc are used n less tan 2% of te world s fseres. 3 Apparently, mplementaton of rgts-based management of fseres and oter common pool resources as been ndered by te transactons costs nvolved n sftng away from an exstng regulatory regme. 4 n te fsery, ndvduals wo are well-suted to competng under an exstng regme ave ncentves to block te transton. 5 We examne a varant of rgts-based management tat as te potental to overcome ts ncentve to block. n te system we examne te regulator assgns a porton of te aggregate catc to a group of arvesters, formed voluntarly, to manage as te group decdes. Tose coosng not to jon contnue to fs ndependently, under te pror regme. Under condtons we spell out, te transton to a cooperatve-based fsery management regme can be Pareto mprovng, a feature tat enables reform wtout losers n te sense of Lau, et al. 2000). Tese autors pont out tat desgnng polcy reform to be Pareto mprovng by protectng status quo rents under te preexstng regme) mnmzes poltcal opposton; tey cte ts as a feature tat enabled Cna s economc reforms over te last 30 years, as well as less sweepng polcy sfts elsewere. 6 Te polcy of assgnng a porton of te allowed catc to a group tat s empowered by contract to allocate ts members resources n a unfed way gves te co-op te same contractual structure as te frm. As Coase 1937) ponted out, allocatng nputs centrally va contracts wt a manager rater tan across markets allows an enterprse wt ts structure to capture gans from 2 By rgts-based management we mean creaton of rgts to te good, or bad, n queston and relance on market transactons to allocate tese rgts. By solve we mean tat suc systems can allocate a gven quantty effcently. Applcatons nclude markets for rgts to emt sulfur doxde and oter ar pollutants as well as greenouse gasses. See Tetenberg 2002) for a revew. On te use of a rgts-based regme to manage common pool groundwater, see Provencer and Burt 1994). 3 Costello, et al. 2008). Tese systems, often called catc sare or ndvdual transferable quota systems, ave been sown to yeld sgnfcant effcency gans. Te ndvdual transferable quota, or TQ, s a wdely promoted rgts-based system n wc quanttatve rgts to portons of a fxed total catc are assgned to ndvduals and may be traded. Te collapse of fseres s descrbed and documented n several studes see, Halpern et al. 2008; Myers and Worm, 2003; ackson et al. 2001; Worm et al. 2006). Wle polluton, clmate cange, and abtat damage ave all been mplcated, poor governance structures are wdely beleved to be te root cause Beddngton et al. 2007; Hlborn et al 2005, Wlen 2005). 4 Lbecap and Wggns 1984) and Wggns and Lbecap 1985) sow tat rgts-based approaces to managng common ol reservors also suffer from scant mplementaton due to transactons costs. 5 Compoundng te problem, neffcent fsery regulaton can nduce excessve nvestment n vessels and processng plants. Owners of ts captal ave ncentves to resst regulatory cange tat would elmnate or mpar ts value. 6 We focus on Pareto mprovements wtout sde payments. Condtons ensurng a Pareto mprovement clearly depend on te form of te pre-exstng regulaton. 2

3 coordnaton wtout ncurrng excessve transactons costs. 7 Alcan and Demsetz 1972) extended ts nsgt, argung tat te gans to coordnaton are greatest wen producton requres te use of a sared nput. Use of a sared nput s a allmark of commercal fsng and oter actvtes based on common pool extracton, as te pool tself s jontly used by all partcpants. 8 We contrbute to te lterature on common pool management by developng a model of ts alternatve regme and testng ts mplcatons wt avalable data. Our analyss s motvated by te formaton of a salmon arvestng cooperatve tat operated n Cgnk, Alaska durng Data from ts fsery and oter nearby fseres form te bass for emprcal tests. Orgnally, te Cgnk sockeye salmon fsery was regulated by a lmt on lcenses and by closng te season wen a target level of catc was reaced, a common form of management known to nduce a race to fs. 9 Allowng a voluntary co-op to form, wt non-joners fsng as before, led to a dramatc ncrease n te proftablty of te fsery by causng canges n te lengt of te fsng season, te number of permt olders wo actually fsed, te spatal and temporal deployment of effort, and te coordnaton of effort across fsermen. Te co-op s demse n 2005, te result of a lawsut, s an mportant part of te story tat we examne n some detal. Our model and emprcal analyss focus on tese beavoral sfts and on weter te regme cange was or could ave been) Pareto mprovng. Two aspects of te regme we examne, te voluntary jonng decson and assgnment of some catc rgts to a group rater tan entrely to separate ndvduals, are noteworty and lnk our analyss to broader lteratures. Because jonng te co-op s voluntary, te dsgruntled can opt out and ts makes Pareto mprovements easer to attan. Ts feature also mples, owever, tat te fnal outcome generally wll not be Pareto optmal. Te reason s tat, wt varatons n fsng skll, some fsermen wll opt to stay wt te orgnal regme wc s known to be neffcent. Te co-op we examne forms voluntarly and provdes excludable, sared nputs for members,.e., club goods. As a consequence our teory and emprcal work contrbute to te broad lterature on te teory of clubs Cornes and Sandler, 1996). n addton to formng voluntarly, te co-op coordnates ts members decsons, wle non-joners form an uncoordnated frnge. Models of cartel formaton d Aspremont, et al., 1983) and nternatonal envronmental agreements Barrett, 1994; Kolstad and Ulp, 2008) ave te same structure, so our results may make useful contrbutons to emprcal knowledge of ow suc structures 7 n fact, Coase 1937) refers to te frm s manager as an entrepreneur-coordnator. 8 Scott 2000) argues persuasvely for assgnng management responsbltes to fsery-wde frms or arvester assocatons for smlar reasons. Scott also noted gans from coordnatng arvestng actvty n s orgnal Scott, 1955) contrbuton, but ts pont was overlooked n te lterature tat followed. 9 Te race and ts consequences ave been extensvely documented n te lterature; see Wlen 2005). 3

4 functon. Te fsery cooperatve also sares mportant features wt worker-formed teams voluntarly formed groups of workers tat coordnate members actons and are pad on te bass of overall team productvty Hamlton, et al. 2003). Our analyss seds lgt on two central questons from te worker-team lterature: te productvty gans attanable by groups of workers wo self-coordnate, and te dstngusng attrbutes of tose wo coose to jon. Several features of te Cgnk experment make t deal for emprcal analyss of tese aspects of regme cange. Frst, eac sector fsed at separate tmes tat were determned by fsery regulators; te co-op as a coordnated fleet and te ndependents as compettve ndvduals. Ts feature allows us to cleanly compare ow wtn-season beavor dffered under cooperatve and compettve fsng. 10 Second, tere were two regme sfts at Cgnk, eac movng n an opposte drecton. n 2002 te co-op polcy supplanted lmted entry fsng wt a sngle fsery-wde season, and n early 2005 an Alaska Supreme court decson sut down te co-op, abruptly renstatng te orgnal system. Tese polcy treatments elp us to cleanly separate te co-op s fsery-wde effects from fsery specfc tme trends. 11 Trd, te skll of te 100 permt olders at Cgnk vared substantally. Ts fact enables us to assess ow fsng skll affected te decson to jon or reman ndependent and also ow relatve skll affected wc co-op fsermen actvely fsed as co-op employees. Our fndngs generally confrm teoretcal predctons by us and oters tat coordnated use of a common pool resource can end te race to extract and enance rent capture see, e.g., Wlson, 1990; Scott, 2000). Some of te co-op s gans were realzed by consoldatng effort among ts ger sklled fsermen and by ncreasng te value of te output, te effcency cannels often empaszed n dscussons of fsery reform. 12 However, te co-op aceved addtonal gans troug cannels seldom mentoned n te reform lterature: by coordnatng on te locaton and tmng of fsng, by sarng nformaton on stock locatons and by provdng sared nfrastructure. 13 We fnd no evdence of suc beavor by ndependents. Despte clear 10 Te mgratory nature of te speces, sockeye salmon, made t possble for te regulator to separate te stock between te two groups of fsermen by openng and closng separate seasons for eac. By ts sort of separaton, one group s arvestng opportuntes were unaffected by te actons of te oter group. Clearly, suc separaton s not possble for all common pool resources. 11 Te fact tat te treatment swtc was turned on, and ten off, strengtens te case tat te cange n beavor of te treated fsery between treated and untreated perods s causal Meyer, 1995, p. 158). 12 Tese are key benefts of TQ reforms, wc s te most common type of rgts-based system used n te ndustralzed world see Grafton et. al. 2000, Hannesson 2004, Leal 2002, Lnn 2008, Newell et. al. 2005). 13 Several autors ave dentfed condtons under wc a smple assgnment of rgts, e.g. sares of an TQ fsery, wll not fully ncentvze proft maxmzng beavor Boyce, 1992; Costello and Deacon, 2007). An emergng body of evdence examnes te effcency effects of cooperatve management, ncludng te contractng problems teren onson and Lbecap, 1982; Ostrom, 1990; Wlson, 1990; 4

5 evdence of an aggregate rent gan, te lawsut fled by two ndependents ndcates tat some felt dsadvantaged by te regme cange. Ts motvates us to derve condtons under wc allowng a voluntary co-op to form wll be Pareto mprovng and to see f tese condtons were met at Cgnk. n te followng secton we develop a model of te cooperatve-based management regme and derve a number of predctons regardng membersp, fsng ntensty, effcency, publc nput provson, and proftablty n eac sector. Te model also provdes condtons descrbng ow te catc must be dvded n order to yeld a Pareto mprovement. Secton 3 presents emprcal tests of te model s predctons and estmates of te rent gans from allowng te co-op to form. Secton 4 descrbes te court decson tat eventually ended te co-op and draws lessons for te mplementaton of ts approac to common pool resource management more generally. t also comments on te practcal mportance of our results for te ecologcal and economc state of fseres. 2. Model Te polcy we examne grants permt olders te opton to eter jon a proft-sarng cooperatve or to reman ndependent. A fsery-wde total allowable catc TAC), determned by te regulator, s dvded nto two portons tat are separable across tme or space, one for te coop and te oter for ndependents. Under te orgnal polcy all lcense olders fsed compettvely untl te TAC was reaced, at wc pont te entre fsery was closed. Ts regme, wc naturally encourages a race to fs, stll apples to te group coosng to reman ndependent, so ter ndvdual success depends on skll and pror nvestments n racng captal. 14 Tose wo jon te co-op act under te drecton of a central manager wo seeks to maxmze te co-op s proft. We examne ow beavor dffers between te two sectors as well as te gans and losses ndvdual fsermen experence wen te co-op s allowed to form. Our model takes as gven te sze of te stock and te regulator s TAC and focuses on te wtn-season actvtes of a fxed number of lcense olders. t ncorporates tree features tat are present n many actual fseres: 1) eterogenety n te skll levels and alternatve employment opportuntes of ndvdual fsermen; 2) eterogenety over tme and/or space n Knapp and Hll, 2003; Costello and Deacon, 2007). Oters argue tat dvdng a total catc between two sectors, as wen a sare s allocated to a cooperatve, s lkely to be a muc smpler negotatng task tan allocatng unque sares to all ndvdual arvesters Matulc et al., 2001). 14 Ts form of management s common n U.S., Canadan, and European waters. 5

6 te unt value of te stock; and 3) te potental for fsermen to gan, collectvely, by sarng nformaton, nfrastructure or oter publc nputs. Tese features allow us to examne tree mportant aspects of te allocaton of fsng effort: 1) avodng redundant capacty and assgnng fsng actvty to te most effcent fsng unts, sometmes referred to as ratonalzaton of effort; 2) coordnatng effort over tme and space to avod races for te most valuable portons of te stock; 3) provdng effcent levels of publc nputs. Fsng skll s parameterzed by te term γ, wc we nterpret as te rate at wt a fserman can apply effort. Fserman s total effort s te product of γ and te tme spends fsng, T. We parameterze te opportunty cost of tme wt φ. f as an attractve opportunty n anoter fsery tat operates at te same tme, or n an entrely dfferent occupaton, φ wll be large. Te unt value of te stock vares over space because t mgrates toward a port were fsng vessels and processng facltes are based and te cost per unt effort falls as te stock comes nearer. 15 We parameterze ts by dvdng te fsng grounds nto two zones and regardng te dstance to eac as a sngle value, 0 or d. Tese zones are called nsde and outsde, respectvely, and dstance s normalzed so tat fsng at an addtonal unt of dstance ncreases te cost per unt effort by 1 unt. 16 We assume te stock spends tme T n eac zone. Te cost per unt effort can be reduced by te avalablty of a publc good nput, G. Examples of G nclude sared nformaton on stock locatons and sared nfrastructure. ndvdual contrbutons to te publc good, denoted x, are costly to contrbutors and total provson of G s determned by te aggregate amount contrbuted. Later, we allow for separate provson by two groups of arvesters and specfy tat te benefts of a publc nput are restrcted to te group tat provdes t,.e., t s a club good. Combnng tese components and ncludng a common cost per unt effort parameter α, fserman s total cost s c = { α + d G Σ x )}γ T + φ T + x. 1) 15 Because te stock mgrates along a route, ts varablty could be caracterzed as occurrng over tme. 16 Te decson to fs nsde versus outsde can be taken lterally or as a metapor for a broader class of gans from coordnaton. Coordnaton gans from tmng arvest actvtes, mprovng catc qualty and sarng nformaton are descrbed n te next secton. 6

7 Te expresson n brackets ncludes all cost components tat are proportonal to s effort. We assumeg 0) = 0, G > 0, G < 0 and α + G ) > 0. d Total catc, Q, s assumed to be a lnearly omogeneous functon of aggregate effort, E, and te stock, Z. Te fsng tecnology s represented by Q = ZF E Z) 2) were F > 0, F < 0, F 0) = 0 and F E Z) < 1. Fg. 1 llustrates te catc tecnology. Gven an avalable stock, Z, applyng E unts of effort wll yeld a catc ofq. f tat effort s appled T sequentally, wt E 0 unts appled frst and E T E0 unts subsequently, te frst batc of effort yelds a catc of ZF E0 Z) and te second yelds a resdual catc of Q T Q = Z F E Z ) F E )). Concavty of F ) mples tat catc per unt effort for te 0 T 0 Z frst applcaton of effort slope of lne ab) s greater tan for te second slope of lne bc). T Z Q T c Z F E Z ) Q 0 b Catc a E 0 E T Effort, E Fg. 1. Catc effort functon n te case we examne te stock mgrates toward port and te batces of effort correspond to effort appled n te two dstance zones. ndvduals wo fs outsde are te frst to apply effort and terefore aceve a ger catc per unt effort tan tose wo fs nsde. All tose fsng n a gven dstance zone are assumed to experence te same catc per unt effort Costello and Deacon 2007) apply smlar reasonng to arvestng of a non-mgratory stock tat nabts patces at varyng dstances from port. 7

8 Te regulator s goal s an escapement target of 1 β ) Z, mplyng tat te TAC equals β Z. Gven 2) ts mples tat total effort must satsfy E ZF 1 β ). Total effort for a set of fsermen s Σγ T and te regulator meets te TAC by constranng fsng tmes to satsfy Σγ T ZF 1 β ) κ. 3) Before fsng starts arvesters are allowed to jon a co-op tat wll coordnate ts members effort, wt te goal of maxmzng total co-op proft. Tose not jonng coose dstance, tme spent fsng and publc nput contrbutons ndvdually, takng as gven te decsons of te regulator and oter ndependents. Te sets of ndependent arvesters and co-op joners are denoted ndependents) and joners) and ter respectve numbers are n) and n ). Te regulator assgns portons of te stock to eac group n proporton to tese numbers and n suc a way tat one group s arvest opportuntes are unaffected by te actons of te oter group. 18 Te separate stock assgnments are denoted Z and Z, te separate TAC assgnments are βz and β Z, and te regulator constrans eac group s fsng tmes accordng to 3) to meet tese TAC assgnments. We assume cost parameters are suc tat eac frm s capable of earnng postve proft by fsng ndependently, regardless of te composton of te ndependent and co-op fleets. 19 Te ntal jonng decson and subsequent decsons on effort deployment are modeled as a two-stage entry game. Subgame perfect Nas equlbra are dentfed by backward nducton. Stage 2 coces by te co-op Te co-op s manager s motvated to maxmze total co-op proft and te co-op s total catc s fxed by te regulator. Consequently, te co-op s optmal polcy solves te followng cost mnmzaton problem: 18 Te possblty of parttonng a stock n ts fason depends on te beavor of te target speces. We examne a mgratng speces, sockeye salmon. n practce te regulator parttoned te stock between te two arvester groups by specfyng dfferent season openngs and closures for eac. As a consequence te two groups could operate on separate portons of te overall run wtout nterferng wt one anoter. A spatal dvson may aceve ts result for a sedentary speces. 19 Ts s consstent wt te presence of postve lcense values under purely ndependent fsng n te fsery we examne emprcally. 8

9 mn d, T ; x α d G x )) γ T + φt + x +, 4) subject to d { 0, d }, T [ 0, T ] for all, and a regulatory constrant on members fsng tmes set accordng to 3). x s te co-op s expendture on te publc nput. Te co-op s optmal polcy s stragtforward. 20 Frst, t sets d = 0 for eac member. Ts s obvous because 4) s non-decreasng n te d for eac member and strctly ncreasng n d for any member wo spends postve tme fsng. Second, te co-op s polcy assgns postve arvest tmes to a subset of members wo ave low values of te ratoφ γ and lmts te number wo actually fs so tat te co-op s season lasts te entre tme fs are avalable, T. 21 Oter co-op members do not fs at all. Te ntuton for ts result s as follows. Snce φ andγ are s cost per unt tme and effort per unt tme, respectvely, te ratoφ sensble to concentrate effort among lowφ γ s 's cost per unt effort, so t s γ members. By slowng te rate of fsng and makng ts season last as long as possble, te co-op concentrates effort among tese effcent members to te greatest extent possble. Trd, te co-op s publc nput provson satsfes Z 1 G x ) F β ) 1, 5) were x 0 and 5) olds wt equalty f x > 0.Ts s a Samuelson condton for optmal provson of a publc good. Tese results are summarzed as Proposton 1 Te co-op s polcy requres tat: ) All actve co-op members fs as close to port as possble; ) Only members wt low cost per unt effort φ γ ) apply effort, tese effcent ) members fs te entre tme te season s open, and te season s open for T perods, te entre tme te stock s avalable; Provson of te publc nput equates te co-op s aggregate margnal beneft from provson to margnal cost, satsfyng a Samuelson effcency condton. 20 Te assumpton tat any co-op member could ave earned postve proft from fsng as an ndependent mples tat te co-op s maxmal proft s necessarly postve. By jonng te co-op, eac member avods any cost assocated wt fsng outsde and benefts from te publc nput. 21 See te Appendx. 9

10 Proof: See text and te Appendx. Stage 2 coces by ndependents Te ndependent fleet s catc per unt effort at any locaton d depends on te effort levels and locatons of all ndependents. We denote catc per unt effort by H d; d, γ, T,, Z ) proft wen te set fses ndependently s and assume eac ndependent takes t as gven. 22 ndependent s π = H d ; d, γ, T,, Z ) γ T α + d G Σ x )) γ T φ T x. 6) ndependent s proft s lnear n T and, by assumpton, maxmal proft s postve. Frm s maxmal proft s terefore ncreasng n T. Ts mplest = T for all,.e., all ndependents fs te entre tme ter season s open. ndependent s optmal publc nput contrbuton satsfes te frst-order condton G Σ x ) γ T 1, 7) were x 0 and 7) olds wt strct equalty f x > 0. Te left-and and rgt-and sdes of 7) are s prvate margnal beneft and margnal cost for contrbutng. Let * be te ndependent wt te gest γ among all ndependents; te prvate margnal beneft of contrbutng s greatest for ts ndependent. Assumng ndvdual fsermens γ parameters are dstnct, f G 0) γ > 1 T ten te unque Nas equlbrum requres ts arvester and only ts arvester to make a contrbuton; * s contrbuton n ts case satsfes 7) wt equalty. 23 Alternatvely, T fg 0) γ 1ten eac ndependent fserman s optmal contrbuton s zero. n eter case, t s clear and unsurprsng) tat ndependents under-provde te publc nput. 22 Te number of ndependents s assumed suffcently large tat eac ndvdual gnores te effect of s/er effort level on te group s catc per unt effort. 23 Gven tat 7) s satsfed wt equalty for ndependent *, te nequalty must be strct for all oter ndependents, mplyng tat ter optmal contrbuton s zero. Ts s a standard free-rder equlbrum. 10

11 M ) 1 A κ ) E Z ) A, M κ ) 1 ndependent effort, E Fg. 2. ndependent fserman s catc per unt effort, dependng on were oter ndependents fs κ E Z ) M, Te coce of fsng dstance can be examned usng te margnal and average catceffort functons, E, Z ) Q / E = F E / Z) and A E, Z ) Q / E = F E / Z) / E / Z) M. Te sapes of teses functons, sown n Fg. 2, are determned by te monotoncty and concavty of F ). To meet te catc target, te regulator fxes total ndependent effort accordng to 3), at a level denotedκ. f all ndependents fs at te same dstance, all obtan te same average catc per unt effort, A κ, Z ), regardless of weter all fs nsde or outsde. 24 Suppose ndependent cooses to fs nsde wle all oter ndependents fs outsde. n ts case encounters te stock after oter ndependents ave fsed and obtans te margnal rater tan average) catc per unt effort fromκ unts of effort, M κ ). Alternatvely, f fses outsde wle all oter ndependents fs nsde, s catc per unt effort would be M 1) n Fg. 2, te margnal catc from te frst unt of effort. 25 f all ndependents are fsng outsde, any ndvdual wo devates to te nsde would fnd tat cost per unt effort falls by d, but catc per unt effort falls by A κ ) M κ ). f A κ ) M κ ) > d, wc we refer to as Condton ), ten no ndependent wll fnd t proftable to devate nsde. 26 f Condton ) olds, wc s more lkely wen d s small, te Nas equlbrum strategy profle n ts subgame s unque and requres tat allκ unts of effort fs outsde. Suppose, nstead, tat all ndependents are fsng nsde. n ts case any ndvdual 24 We encefort suppress te second argument n A. ) and M. ), snce t remans uncanged. 25 Fserman s catc equals s catc per unt effort tmes te effort apples, γ T. Catces from te same locaton wll terefore dffer among fsermen n proporton to ter γ parameters. 26 Te common cost term φ T, wc appears n bot proft comparsons, as been gnored. 11

12 wo devates outsde wll fnd tat cost per unt effort ncreases by d, but catc per unt effort ncreases by M 1) A κ ). f M 1) A κ ) < d, wc we refer to as Condton ), ten no ndependent wll fnd t proftable to devate outsde. f Condton ) olds, wc s more lkely wen d s large, a Nas equlbrum n ts subgame s unque and requres tat allκ unts of effort fs nsde. Fnally, suppose A κ ) M κ ) d M 1) A κ ) so neter condton olds. Ts mples tat a Nas equlbrum strategy profle for te second stage subgame cannot ave all effort fsng eter nsde or outsde. We llustrate ts case n Fg. 3. Te orzontal axs now ndcates outsde effort and te dased lne A E) d sows outsder proft per unt effort. To caracterze Nas equlbrum coces of dstance, suppose all ndependent effort was ntally fsng outsde and successve unts were transferred nsde. Te frst unt transferred nsde would earn proft M κ ), sown by pont c, wc exceeds te proft from fsng outsde. Transferrng successve effort unts nsde causes nsder proft per unt effort to ncrease toward pont a, at wc pont all effort s fsng nsde and proft per unt effort equals A κ ). Te dotdas lne labeled nsder proft traces out one possble locus of nsder profts. 27 f Ê unts of effort fs outsde and all oters fs nsde so all earn equal proft, no one as an ncentve to devate. 28 Accordngly, a Nas equlbrum strategy profle n ts case s descrbed by ts dvson of nsde and outsde fsng. M1) nsder proft per unt effort Outsder proft per unt effort a AE) d c ME) b A E) d 1 Ê κ Ι Outsde effort Fg. 3. A NE strategy profle n wc some ndependents fs outsde wle oters fs nsde. 27 t can be sown tat te dot-das lne s monotone and contnuous. 28 Fg. 2 s drawn so tese curves only cross once; we ave not excluded te possblty tat tey cross more tan once. 12

13 We summarze tese results as Proposton 2 n te subgame nvolvng te ndependent sector s coce of tme spent fsng, publc nput contrbutons, and fsng locatons, a Nas equlbrum strategy profle requres tat: ) Eac ndependent arvester fses te entre tme te regulator leaves te ndependents season open; ) Te ndependent sector under-provdes te publc nput relatve to wat s effcent; ) All ndependents fs outsde f A κ ) M κ ) d, fs nsde f > M 1) A κ ) < d and are splt between fsng nsde and fsng outsde f A κ ) M κ ) d M 1) A κ ). Proof: See Fgs. 2 and 3 and te precedng dscusson. We also note tat te TAC constrant 3) and te regulator s stock assgnment, Z = Z n ) n K), mply tat te ndependent sector s season lengt equals T 1 ZF β ) / n K) = 8) Σ γ / n ) and s terefore nversely proportonal to te group s average skll, a result tat wll become useful later. Te Stage 1 decson of weter or not to jon We adopt te conventon tat fsermen are ndexed n ncreasng order of ter γ terms, so low skll fsermen ave low ndex numbers. To obtan a clear dentfcaton on te attrbutes of co-op joners, we assume tat g skll arvesters g γ) ave low cost per unt effort lowφ γ ). Ts wll be true f te φ terms are constant, f φ and γ are nversely ordered, or f φ does not ncrease more tan proportonately as γ ncreases. We start by examnng te second stage proft sares of successve co-ops n wc new members are added n order of ter γ parameters and demonstrate tat larger co-ops formed n ts fason) necessarly ave ger proft per member. Wrtng out te co-op s proft sare equaton and ncorporatng ts optmal polcy coces and te regulator s TAC assgnment yelds Z 1 1 * 1 Zn ) * 1 π ) = β αf β )) + G x ) F β ) x φt 9) n K) n ) n K) n ) mn 13

14 * were ndcates te set of co-op members selected to fs and x s te co-op s optmal publc mn nput contrbuton. Te rs conssts of tree components. Te frst s catc per member mnus te common cost term nvolvng a. Gven te TAC allocaton formula, ts does not depend on co-op sze. Te second component s te co-op s maxmal net publc good beneft per member. As sown n te Appendx, t necessarly s ncreasng n n ). Te trd component s te opportunty cost of tme spent fsng dvded by te number of co-op members; t decreases wt co-op sze for te followng reason. f a new member s added te TAC allocaton rule causes a proportonate ncrease n te co-op s effort, so effort per member remans uncanged. Consequently, te effect of a new member on te trd component n 9) concdes wt te new member s effect on te co-op s average tme cost per unt effort. Gven te order n wc members are added, te new member s tme cost per unt effort φ γ ) s necessarly less tan tat of exstng members. Terefore, te new member wll be desgnated to fs and te co-op s average tme cost per unt effort falls. Taken togeter tese results mply tat co-op proft per member ncreases wt co-op sze, as llustrated by te upward slopng lne π C γ ) n Fg Te postve co-op proft sown for te lowest skll level follows from te assumpton tat all fsermen could earn postve proft by fsng ndependently, plus te fact tat ) a one member co-op would receve ts own TAC allocaton and tereby avod fsng outsde and ) ts allocaton exceeds wat te least sklled arvester would catc as an ndependent. Ts reasonng also mples tat a 1 member co-op s proft exceeds wat te same fserman could earn by fsng ndependently wt all oter arvesters, a result tat s useful sortly. 29 Ts sold lne s a smoot curve connectng a set of dscrete ponts ndcatng te per member profts for co-ops of dfferent szes. 14

15 Proft π m γ ) πˆc π C γ ) 1 e e+1 Sze of co-op fsermen ordered by γ) nk) Fg. 4. Equlbrum co-op sze Next, we examne te proft of te margnal least proftable) co-op member n a sequence of ndependent fleets formed by successvely addng lower sklled arvesters and demonstrate tat te margnal ndependent s proft necessarly falls as te sze of te ndependent fleet grows. To smplfy, we assume te ndependent fleet s equlbrum publc nput provson s 0, wc s always approxmately true. We also make use of te conventong 0) = 0 and te fact tat catc per unt effort equals β / F 1 β ) due to te TAC constrant. ncorporatng tese smplfcatons n 6), ndependent arvester s proft n te case were all ndependents fs outsde s π β = α d T F γ φ, β ) ) 1 wc we wrte as π ) = 1 β φ α d γ T. 10) F β ) γ Our earler assumpton mples tatφ γ falls as γ ncreases, so ndependents wt ger skll parameters ave ger profts. Te margnal least proftable) ndependent n any group s terefore te one wt te lowest γ and formng a sequence of ndependent fleets by successvely addng lower skll fsermen causes margnal proft to declne. Te same concluson apples n te case were all ndependents fs nsde because s proft n ts nstance s found by replacng te constant d n 10) by zero. Ts result also extends to te case were some ndependents fs nsde and oters fs outsde because equlbrum n te second stage requres 15

16 tat eac ndependent earns te same proft per unt effort at eter locaton. Ts mples tat te nsde vs. outsde dfferental n catc per unt effort exactly matces te dfferental n cost per unt effort, d, so once agan ndependents wt ger skll parameters ave ger profts. 30 m γ Te dased lne π ) n Fg. 4 llustrates te margnal proft n a group of ndependent fsermen wo ave effcency parameters greater tan or equal to a gven level γ ι. Te left m γ C vertcal ntercept of π ) les below te π γ ) ntercept because, as explaned earler, a 1 member co-op s proft exceeds wat te same fserman could earn by fsng ndependently wt all oter arvesters. Te rgt vertcal ntercept of π ) s sown to le above te m γ correspondng ntercept for te co-op, ndcatng tat te gest sklled fserman could earn more by fsng as a lone ndependent tan by jonng an all-nclusve co-op, but ts s not te only possblty. f bot condtons on ntercepts are met ten π ) must cross π γ ) from m γ C below at least once. Suc a crossng pont dentfes a tresold skll level tat separates co-op joners from ndependents. n Fg. 4 te tresold s ndex value e, referrng to a fserman wt skll levelγ e. f all arvesters wt skll less tan or equal to γ e are n te co-op ten: ) all tose n te co-op earn π γ ), wc exceeds tat tey would earn by fsng ndependently, and ) all C e tose wo fs ndependently earn more tan tey would n te co-op snce π m γ ) > π C γ e) > e. 31 Ts allocaton of fsermen to groups, togeter wt Nas equlbrum m γ strategy profles n stage 2, s terefore a subgame perfect Nas equlbrum. f π ) les C entrely below π γ ), te allocaton n wc all arvesters jon te co-op s te only Nas equlbrum. f te two curves cross more tan once, tere s an equlbrum for eac occason were π ) crosses π γ ) from below. Te generc Stage 1 predcton, tat g γ fsermen m γ C coose to fs ndependently, s not surprsng; by defnton, glners compete most successfully n te race to fs and jonng te co-op would necesstate sarng ter arvest profts wt less sklled fsermen Te 1 β F β) α d term s replaced by one of two expressons n ts case, dependng on weter te ndvdual nvolved fses nsde or outsde, but tese two expressons take on te same value. 31 We assume a fserman jons te co-op f profts from te two coces are equal. Te condton stated n te text s equvalent to te nternal and external stablty condtons for cartel formaton developed by d Aspremont, et al. 1983). 32 We ave not demonstrated tat π m γ ) ncreases monotoncally. As te ndependent fleet s average skll level ncreases te season lengt falls, wc works aganst te proft ncrease from greater skll. 16

17 Key results on te jonng decson are summarzed as: Proposton 3 Under our assumpton on te relatonsp between effort rate and tme cost parameters, a subgame perfect Nas equlbrum strategy profle as te followng propertes: ) Te group coosng to fs ndependently conssts of glners; more precsely, all ndependents ave skll levels greater tan any co-op member; ) Te coces of dstance, fsng tme and publc nput contrbutons are descrbed by Propostons 1 and 2. Caracterzng Pareto-mprovng catc allocatons Te above dscusson caracterzes te membersp and economc beavor of eterogeneous fsermen composng te two fleets. Here we dentfy condtons under wc allowng formaton of te self-selected cooperatve s Pareto mprovng. Te answer nges on te allocaton of catc between te two sectors. We ave assumed tus far tat te regulator assgns catc n proporton to membersp: Z = Z n ) n K). To explore ts ssue more completely, we generalze ts noton to allow for dsproportonate assgnments: = Z θn ) n K) were te scalar θ controls te proportonal assgnment to te Z cooperatve sector. For example, f θ=0.9 ten te cooperatve s assgned a stock allocaton tat provdes nne-tents of a per capta sare for eac co-op joner. ntutvely, t would seem tat cooperatve members would be advantaged and ndependents dsadvantaged by larger values of θ, but te endogenety of self-selected membersp may blur ts ntuton. We start by dervng te proft for an arbtrary fserman,, n a completely ndependent fsery, a term we use to refer to te counterfactual stuaton were no co-op s allowed to form. We ten compare ts proft to wat would earn wen te cooperatve s allowed to form. Naturally, we smultaneously solve for weter fserman fses ndependently or as a member of te cooperatve fleet. n a completely ndependent fsery.e. f te co-op were not allowed for form), would earn te followng proft from ndependent fsng: ~ π β φ ~ α d γ T 1 F β ) γ, 11) = ~ 1 were T ~ s te season lengt n te absence of a cooperatve, gven by: T = ZF β ) / Wen te voluntary cooperatve s allowed to form, s proft depends on weter e/se decdes to jon or to fs ndependently. Suppose cooses to fs n te ndependent fleet. Te resultng proft s: γ. 17

18 π β φ α d γ T F β γ, 12) ) = 1 Here, T s te season lengt for te ndependent fleet, gven by T = Z F 1 β ) / γ. Te stock n ) assgnment Z depends on te allocaton rule as follows: Z = 1 θ ) + θ Z. Fserman n K) gans from te co-op s formaton f ~ π < π, and loses f ~ π > π, wc clearly depends on te allocaton parameter θ. Settng te rgt-and sdes of equatons 11) and 12) equal, we can solve for te crtcal parameter value, θ c,tat yelds te same proft for regardless of weter te co-op forms: γ / n ) θ c =. 13) γ / n K) were s te set wo would jon. 33 Te rs of 13) s te rato of average skll for tose wo would jon to te average skll of all fsermen. By Proposton 3, joners ave below-average skll, soθ < 1. Tose wo would coose to fs as ndependents are dsadvantaged by allowng c te co-op to form f θ > θ c and tey are advantaged fθ < θ c. Next, consder te fate of tose wo opt to jon te cooperatve f t s allowed to form. Proposton 3 ndcates tat tese ndvduals are apt to be te lower sklled members of te fleet. Because tey coordnate on fsng locaton and publc goods provson bot of wc lower costs) ter calculus s somewat dfferent, but t stll nges on ow θ compares toθ c. fθ > θ, te most sklled members of te cooperatve are actually dsadvantaged by te c fact tat t forms. Consder te most gly sklled joner. n te lmt, f te number of fsermen s large ts ndvdual earns te same proft as te least sklled ndependent. We establsed above tat all ndependents are strctly worse off n te presence of te cooperatve wenθ > θ, so te same s true for te gest skll joner. c 33 n a stuaton were all are fsng te same amount of tme per season, as was te case wt ndependent fsng before te co-op was allowed to form, ts rato would equal te rato of average catces for co-op joners to te average catc for te entre fleet. t follows tat te crtcal parameter θ c can be estmated from nformaton on average catc sares of joners and ndependents n a pre-co-op perod. 18

19 We next sow tat fθ = θ, a cooperatve stll forms and all wo jon are made better c off by te opportunty to jon. A suffcent condton for te formaton of a cooperatve s tat te lowest-skll fserman can earn ger proft by formng a one-person cooperatve tan by fsng n a completely ndependent fsery. Revenue n te two stuatons s te same wenθ = θ because a one person cooperatve s catc allocaton equals wat te ndvdual c would ave caugt by completely ndependent fsng. Cost for te one person cooperatve s lower tan wt ndependent fsng, owever, because te cooperatve coordnates on fsng locaton; so ts ndvdual would beneft by formng a one-person co-op. How does te gest skll joner n a mult-person co-op fare? Gven te decson to jon, ts person s proft as a coop member s at least as great as wat e/se could ave earned by optng nto te ndependent fleet. n turn, snceθ = θ te proft tat would ave been earned by coosng to fs as an c ndependent equals wat ts ndvdual would ave earned n a completely ndependent fsery. Tus, all joners are at least weakly advantaged by te ablty to jon a cooperatve. Fnally, f θ < θ c a cooperatve may or may not form. Clearly, f θ s suffcently near zero te loss from a low catc allocaton more tan offsets te gans from coordnaton for a coop of any sze, so no co-op wll form. Letθ < θ be te lowest value of θ for wc a cooperatve of some sze wll form. Ten for θ values n te nterval θ θ < θ a cooperatve forms and all fsermen, ncludng ndependents and joners, beneft from ts formaton. To see ts, frst note tat f te ndependent fleet contans N) fsers, ten even te least sklled of tese ndvduals s advantaged by co-op s formaton because an allocaton satsfyng θ < θ c L advantages tose wo opt nto te ndependent fleet. Next, consder te gest sklled joner. Gven te decson to jon, te ndvdual s co-op proft necessarly exceeds wat e/se could ave earned by optng nto te ndependent fleet. Ts potental ndependent fleet proft, n turn, necessarly exceeds wat e/se would ave earned n a completely ndependent fsery becauseθ < θ. Terefore te gest sklled joner s better off from te co-op s formaton. All c lower-sklled joners earn te same proft as te gest sklled joner and would ave earned less n a completely ndependent fsery, so tey are all advantaged as well. We summarze all of tese fndngs n te followng proposton: Proposton 4 Te formaton of a self-selected cooperatve as te followng dstrbutonal consequences: c L c 19

20 ) fθ L θ θ c te nsttutonal desgn s Pareto mprovng fsermen of all skll levels are made weakly better off by allowng te cooperatve to form. ) fθ > θ c te nsttutonal desgn s not Pareto mprovng all would-be ndependents and some would-be cooperatve fsermen are made worse off by allowng te cooperatve to form. ) f θ < θ L ten no cooperatve forms. 3. Emprcal Evdence We test te model s predctons wt data from te Cgnk sockeye salmon fsery and negborng fseres. Cgnk s one of Alaska s oldest commercal salmon fseres, datng to te 1880s. Pror to 2002 approxmately 100 ndependent purse sene permt olders competed for a sare of a fsery-wde lmt on allowed catc, wc was enforced by closng te season wen te target was met. 34 n 2002 te Alaska Board of Fseres approved a request by a group of permt olders to form annual cooperatves for voluntary joners and ts arrangement contnued troug Te number of fsermen wo joned ranged from 77 n 2002 and 2003 to 87 n 2004, wt te total number of permts equalng 100 trougout te perod. Sortly before te 2005 fsng season, te Alaska Supreme Court sut down te co-op rulng tat t volated an Alaska law probtng permt olders wo dd not actvely fs from accrung fsery profts. 36 Ts story generated tree dstnct perods tat we explot n our emprcal models te years before te co-op, te co-op years, and te years after te co-op was sut down. Te fact tat te co-op treatment swtc was turned on, and ten off, elps us solate te casual effects of te cooperatve from fsery specfc tme trends Meyer 1995, p. 158), and t elps elmnate seral correlaton n our panel regresson models Bertrand et. al. 2004, p. 251). 34 Purse sene fsng uses a large net tat cnces from te bottom to prevent scools of fs from escapng. Te Appendx sows maps of te Cgnk fsng area. 35 Te co-op s proposal explaned te motvaton for formng te cooperatve as follows: Te current fsng fleet s overcaptalzed and te compettve arvest system does not allow for real mprovements n produc[t] qualty or flexblty n competng wt farmed salmon Grunert v. State 2005, p. 927). 36 A constraned verson of te co-op was allowed to operate durng Te constrant requred all co-op members to spend tme at sea aboard fsng vessels Stcert 2007). Because our model mples tat ts requrement elmnated te man effcency gans of te co-op, we consder 2004 to be te fnal year of coop operaton. 20

21 Te model n te precedng secton essentally mrrors te fsery s key pyscal attrbutes and ts actual mode of operaton. Regardng pyscal attrbutes, sockeye salmon mgrate towards only one rver n te Cgnk system Cgnk R.) and are funneled toward tat rver as te mgraton extends from open ocean, troug Cgnk Bay, nto Cgnk Lagoon, and fnally nto Cgnk Rver. Processng facltes are located near te destnaton and te stock becomes more concentrated as t mgrates. Ts generates a tenson between te ndvdual ncentve to fs outsde to contact te stock before oters do, and te collectve effcency of arvestng nsde near te stock s ultmate destnaton. Regardng te mode of operaton, te co-op was allocated a sare of te total allowable catc TAC) eac year to arvest as t saw ft, wt te remander desgnated for tradtonal, compettve arvest by te ndependent sector. Eac co-op joner sgned a 1-year contract before fsng began, agreeng tat te cooperatve would manage s/er fsng effort for te comng season. Te two sectors fsed at dfferent tmes, determned by te regulator, and eac sector s season was closed wen ts TAC sare was reaced. Eac year te co-op s TAC sare was determned by te followng rule: ) f less tan 85 percent of permt olders joned, te co-op receved an allocaton equal to nne-tents of a per capta sare for eac joner; and ) f 85 percent or more of permt olders joned, te co-op receved a full per capta sare for eac joner. n terms of te model, tese two allocatons correspond toθ = 0. 9 andθ = 1. 0, respectvely. Furter, tere s sgnfcant varaton n fsermen s skll levels as measured by sare of catc story n years pror to te co-op, ndcatng te potental for an nteror soluton were some elect to jon te cooperatve wle oters reman ndependent. Two features not specfcally addressed by our model mply tat te co-op s formaton could affect prce. Te frst s te presence of monopsony power n Cgnk under tradtonal, non-cooperatve fsng wt 100 fsermen, and only one or two processors n te perod we study. t s wdely beleved tat processors extract most of te rents from negotaton wt ndependent fsermen; presumably, a coordnated arvester group could weld ts own market power. Te second s a potental prce premum for ger qualty product; ndeed, te possblty of exercsng greater care n arvestng n order to delver a ger qualty product was promnent n ntal dscussons on formng a co-op. Bot consderatons ndcate tat te co-op s formaton mgt lead to ger prces to co-op fsermen We cose not to ncorporate te market power feature explctly n te model, n part because ts effect seems obvous and n part because ts seems specfc to Cgnk and not a feature of common pool resources more generally. Te co-op s ncentve to coordnate to guarantee ger product qualty s smlar to ts ncentve to provde club goods, and n tat sense s consstent wt our model. Te dfference s tat enanced product qualty rases prce, wle we treat te effect of club goods as decreasng costs. 21

22 Allocaton of Fsng Effort Peraps te most compellng reason to form a cooperatve n Cgnk was to reduce te number of vessels actually fsng Prop.1.), n order to catc ts TAC allocaton wc depended on membersp) more effcently. Tus, we expect te proporton of permts actually fsed to declne durng Fgure 5 compares te proporton of permts actvely fsed n Cgnk to te proporton fsed across fve adjacent purse sene fseres for years before, durng and after te co-op. 38 As te fgure sows, te proporton of permts actvely fsed n Cgnk fell from 0.94 n 2001 to 0.41 n 2002 wen te co-op frst operated, and ten ncreased after te co-op was effectvely dssolved n Te darkest bars sow te dfference between Cgnk and te average across te control-group fseres. Ts dfference was strctly postve before and after te co-op years, but approxmately zero durng Fgure 5 Proporton of Permts Fsed n Alaska's Purse Sene Fseres co-op years cgnk fsery average across oter fseres dfference Te adjacent fseres are Alaska Pennsula, Cook nlet, Kodak, Prnce Wllam Sound, and Souteast Alaska. 39 Te spke up to 0.98 n 2005 s wort explanng. n early 2005, sortly before te start of te fsng season and after te co-op was already formed for te 2005 arvest, te Alaska Supreme Court ruled tat te co-op volated an Alaska law probtng permt olders wo dd not actvely fs from accrung profts. Te state s remedy for te 2005 season was to allow te co-op to fs but to requre tat all co-op members actvely fs for a small part of te season. n 2006, te co-op was entrely dssolved. We dscuss te Court decson n more detal later. 22

23 A complementary panel regresson demonstrates tat te vsual evdence of te co-op effect s robust to addtonal controls Table 1). Te panel regresson employs 66 fsery-year observatons, from 6 fseres over 11 years ). Ts tme span allows four years of data before and after te co-op was actve. Te regresson ncludes fsery fxed effects, as well as year fxed effects to control for annual varaton n factors suc as te prces of fuel and farmrased salmon. We also control for te fsery-wde TAC. Te result ndcates tat te co-op polcy reduced te proporton of permts fsed by nearly Te drecton of te effect, a reducton, s consstent wt expectatons and te coeffcent estmate s economcally and statstcally sgnfcant. 40 Te result s partcularly strkng because t pertans to consoldaton across te entre fsery. Consstent wt our teory, annual Cgnk Area management reports ndcate tat almost all of te consoldaton occurred wtn te co-op; durng te proporton of permts actvely fsed was for te co-op and for ndependents. 41 Table 1 Panel Regresson on te Proporton of Actve Permts ndependent Varables 1) Y = proporton of permts fsed Constant 0.329* Co-op Polcy t-statstc Fsery-Wde TAC Fxed Effects Year Dummes Fsery Dummes * 3.69) 1.19e ) ncluded ncluded Observatons 66 Adjusted R Notes: 1) * Sgnfcant at 0.05 level for a one-taled t-test. 2) Year dummes span ) Te 5 control fseres are te nearby purse sene fseres: Alaska Pennsula, Cook nlet, Kodak, Prnce Wllam Sound, and Souteast. 4) Data are avalable at: ttp:// and ttp:// 40 To correct for possble seral correlaton of errors wtn eac fsery we conduct a robustness ceck recommended by Bertrand et. al. 2004). We collapse te data nto averages for eac fsery durng tree tme perods before, durng, and after te co-op years. We next run a panel regresson usng te 18 observatons 6 fseres and 3 tme perods) and nclude fsery and tme perod fxed effects along wt te average fsery-wde TAC. Ts generates consstent standard error estmates Bertrand et. al. 2004). n our case, te coeffcent on Co-op Polcy for te collapsed data s wt a t-statstc of Members wo fsed on bealf of te co-op were pad salares to compensate for ter costs. All co-op members were ten pad equal sares of te proft remanng after tese salares and oter co-op costs were deducted; Knapp and Hll 2003). 23

24 Te model furter predcts tat te co-op wll coordnate on te locaton of arvest n order to reduce costs. Because te co-op secures a guaranteed allocaton of catc, co-op arvesters sould wat untl fs mgrate nsde, at wc tme te arvest wll be more effcently executed Prop. 1.). n contrast, some or all of te ndependent sector s arvest s expected to take place outsde Prop. 2.). We use data on te spatal locaton of catc to test tese propostons n two dfferent ways. Frst, we examne fsery-wde annual tme-seres data to see ow te proporton of sockeye caugt nsde devated durng from longer annual tme trends. We ten use wtn-fsery cross-secton data to assess ow te proporton of nsde catc dffered between co-op and ndependent fsermen durng Our measure of locaton s te annual proporton caugt n Cgnk Lagoon see Fg. A2), te nsde locaton. Fgure 6 Proporton of Cgnk Area Sockeye TAC Caugt 'nsde' 1998 to coop years Fgure 6 sows te fsery-wde proporton of sockeye caugt nsde over a 10 year perod tat ncludes , te co-op s years of operaton, and ndcates tat te proporton caugt nsde peaked durng te co-op years. Te tme-seres regresson model n Column 1 of Table 2 uses annual data on te nsde catc proporton for , te entre perod for wc data are avalable. t accounts for te cyclcal nature of te tme-seres data by ncludng a 4 t - order polynomal tme trend and controls for varaton n arvest by ncludng a 4 t -order polynomal n te annual allowed catc. Allowng te cooperatve to form apparently ncreased te proporton caugt nsde by 0.27, wc presumably reduced costs. Note tat ts proporton apples to te entre fsery, ncludng bot co-op fsermen and ndependents. 24

25 Table 2 Tme-Seres Regresson Analyss of nsde Catc and Season Lengt ndependent varables 1) Y = proporton of catc from nsde 2) Y = number of days fsed Constant 0.773* Co-op Polcy t-statstc Fsery-Wde TAC Fsery-Wde TAC 2 Fsery-Wde TAC 3 Fsery-Wde TAC 4 Year Year 2 Year 3 Year 4 Observatons F-statstc Adjusted R * 3.48) 4.63e e e e e * 3.66) e e e * 7.107* 0.021* Notes: 1) * Sgnfcant at 0.05 level for a one-taled t-test 2) Te data used ere come from Cgnk area annual management reports. Column 1 uses avalable data for Column 2 uses avalable data for We lack data on season lengt pror to 1980 and te 2007 data are not yet publsed. Table 3 compares te locaton coces of co-op and ndependent fleets durng As te model predcts, te co-op arvested ts entre allocaton nsde Cgnk Lagoon n eac year. 42 By comparson, te ndependent fleet arvested from bot nsde and outsde n 2002 and 2003, wc s consstent wt te possblty of a mxed equlbrum. Durng 2004 wen tere were only 13 ndependents, all ndependent arvest took place nsde te lagoon. 42 Te followng account from a co-op founder makes clear tat fsng nsde was a conscous operatng polcy: We ad orgnally planned to employ a couple of large seners to fs out on te capes [outsde], but we realzed tat te extra runnng tme would ncrease costs and reduce product qualty. Harvestng n te close proxmty and concentrated arvest area of te Cgnk Lagoon [nsde] was smply te most effcent and qualty conscous metod to pursue. Ross 2002a). 25

26 Table 3 Proporton of Sockeye Caugt nsde by Co-op and ndependent Fleets on days reserved exclusvely for one of te two fleets) 2002 Number of sockeye arvested Proporton caugt nsde 2003 Number of sockeye arvested Proporton caugt nsde 2004 Number of sockeye arvested Proporton caugt nsde Cooperatve fleet 576, , , ndependent fleet 162, , , Te model also predcts tat te co-op s polcy wll result n a longer fsng season. Prop.1. ndcates tat te cooperatve s allocaton wll be arvested by a subset of ts most effcent members. n order to make maxmal use of ts most effcent arvesters te co-op lmts te number of members wo actually fs, wc slows te rate of fsng and lengtens ts season. By contrast, all ndependents are predcted to fs eac day ter season s open, causng te regulator to sorten te season n order to meet te TAC constrant. We test te season lengt predcton usng tme-seres data for Cgnk, employng te same strategy used to examne te proporton of catc from nsde. Column 2 of Table 2 uses annual data from te avalable years, , on te total number of days sockeye salmon fsng occurred at Cgnk. Te regresson estmates ndcate tat, on average, te presence of te co-op lengtened te season by 32 days, a 48 percent ncrease n season lengt from te long run average of 66 days n non co-op years. 43 We noted earler te co-op s potental to rase prce by delverng ger qualty fs or by exercsng ncreased market power n negotatons wt processors. We cannot separate tese two effects, but we can test for a prce ncrease usng a panel-regresson model wt te same controls as Table 1. Te results Table 4) ndcate tat formaton of te co-op was accompaned by an average prce ncrease of $0.21 per pound n te Cgnk fsery n 2008 dollars). Ts represents a 27.6 percent ncrease from te Cgnk average of $0.76 outsde of te co-op years Our teory predcts tat te co-op s practce of fsng only a small fracton of ts permts wll necessarly result n longer season lengts. Te regulator s polcy of splttng te season nto separate openngs could also ncrease season lengts. 44 Te prce data are nflaton adjusted and are n 2008 dollars. As before we estmated a verson of te regresson n Table 4 by collapsng te data nto averages for eac fsery durng tree tme perods 26

27 Note tat ts s a lower-bound estmate of any prce premum te co-op aceved because nearly one-trd of te sockeye caugt at Cgnk were arvested by ndependents durng Table 4 Panel Regresson of Gross Earnngs Per Pound n 2008 dollars) ndependent Varables Y = gross earnngs per pound Constant 0.526* Co-op Polcy t-statstc Fsery-Wde TAC Fxed Effects Year Dummes Fsery Dummes 0.214* 2.22) -1.17e-06* 2.14) ncluded ncluded Observatons Adjusted R Notes: 1) * Sgnfcant at 0.05 level for a one-taled t-test. 2) Year and fsery dummes are as n Table 1. 3) Te gross earnngs data are adjusted by te CP and are presented n 2008 dollars. 4) Data are avalable at: ttp:// and ttp:// Publc nput Provson Our evdence on provson of sared or publc nputs by te co-op s anecdotal, gleaned from trade press accounts and annual management reports of te Alaska Department of Fs and Game ADFG). Te most promnent sared nputs nstalled by te co-op were fxed leads, statonary nets placed along te fs mgraton route to funnel te stock toward watng purse seners. 46 Te fxed leads altered te style of fsng and dramatcally reduced te number of vessels requred to aceve a gven catc. Ts sort of sared nfrastructure was not employed by te ndependent fleet. 47 Oter actons we caracterze as publc good provson by te co-op amount to very precse coordnaton of members actons. An mportant form of coordnaton was a fnely tuned before te co-op years, durng te co-op years, and after te co-op years. Ts approac generates consstent standard error estmates Bertrand et. al. 2004). Te resultng coeffcent on te Co-op Polcy for te collapsed data s wt a t-statstc of We lack cross-secton data durng tat would allow us to compare output prces between te co-op and ndependent sectors. 46 See Pappas and Clark 2003). 47 Ross 2002a). 27

28 temporal allocaton of ts members effort Stcert, 2007). Durng low tdes n Cgnk Lagoon, te nsde locaton were te co-op arvested, srnks to a fracton of ts sze at g water. Ts concentrates te fs and reduces arvest cost. A promnent co-op member descrbed ow ts strategy worked: nstead of [a co-op member] makng four or fve sets durng te flood [g tde] for 200 to 300 [fs] a aul, e now could wat tll te Lagoon draned out. At low tde [te cannel] became a slow, meanderng rver of concentrated sockeye. And now, fsng for te entre co-op, e could make one gant drag for 3,000 to 5,000 fs. 48 Ts strategy requred tat co-op arvesters allow fs to escape up rver durng g tdes, even toug t s legal to catc tem. Gven te co-op s secure catc allocaton and ts ablty to coordnate, owever, te ncentve to do ts was present. We know of no nstances of ndependent fsermen ntentonally allowng fs to swm up rver. Te co-op also coordnated ts members actons to mprove te qualty of fs delvered to processors. t receved permts to old lve fs n net pens for up to tree days, wc allowed t to better matc delveres to processng capacty. On occason, te co-op even released lve fs from capture wen processng capacty was nsuffcent. 49 ndependent arvesters ave no ncentve to engage n suc practces and we are aware of no evdence ndcatng tat tey dd. Te co-op also coordnated nformaton on stock locatons from all of ts actve members and used ts nformaton to dspatc vessels and crews to te most advantageous locatons. We are aware of no evdence tat te ndependent fleet followed ts practce; ndeed, fsermen are notorous for dng suc nformaton from ter compettors. Fnally, te fsery manager also benefted from te co-op s ablty to coordnate, by enablng precse control of a day s catc n ways tat cannot be accomplsed wt ndependent fsng. Wt ndependent fsng te fsery manager must forecast te rate of catc and announce a closng tme calculated to meet te overall catc target, an mprecse process at best. On days te co-op fsed, te manager could t te target precsely, smply by requestng tat te co-op cease fsng wen te desred number of fs was caugt Pappas and Clark, 2003). Wo oned and Wo Fsed for te Co-op? Our model addresses tese two key questons and yelds te followng predctons: glners wll reman ndependent wle less-sklled fsermen wll opt nto te co-op Prop.3.) 48 Ross December 2002). 49 Te precedng two examples are from: Mark A. Stcert, 2004 Cgnk Management Area Annual Management Report. Alaska Department of Fs and Game, at: ttp:// 2007) 28

29 and te co-op wll deploy ts gest sklled members to fs on bealf of te entre enterprse Prop.1.). Our model suggests tat storc catc s a good proxy for te crtcal skll parameter, γ, so we test tese predctons wt data on catc sares durng te pre-co-op perod. Wle ndvdual catc sares are not dsclosed due Alaska confdentalty laws, we were able to obtan catc sare data aggregated to te tree frm level. 50 Te procedure for carryng out tese aggregatons was desgned to mnmze catc sare eterogenety among te observatons tat were grouped. Because some arvesters canged status durng te co-op perod, dfferent aggregatons were formed, usng te same procedure, for 2002, 2003 and For 2002 aggregatons, ndvdual fsermen were frst parttoned nto tree groups dependng on ter 2002 co-op status: co-op joners wo fsed, non-fsng co-op joners and ndependents. All fsermen n a gven group were ordered by average sockeye catc sare over te storc perod. 51 Successve fsermen were ten clustered nto groups of tree and te average storc catc sare wtn eac cluster was reported to us. Ts procedure was ten repeated for groups formed on te bass of 2003 and 2004 co-op status. Te end result was a set of rougly 100 observatons on co-op status eac year durng and average storc catc sare for 3-frm aggregates) durng We examne te model s predctons bot by testng for dfferences n mean storc catc sares across groups and by testng for stocastc domnance n te storc catc sare dstrbutons of dfferent groups. 52 Table 5 presents comparsons of mean storc catc sares. Te frst set of comparsons ndcates tat te mean storc catc sare of tose wo fsed ndependently durng sgnfcantly exceeded tat of co-op joners 1.29 percent compared to 1.00 percent). Te second 50 We are ndebted to te Alaska Commercal Fseres Entry Commsson for performng tese aggregatons for us. n a few cases t was necessary to aggregate over four frms. 51 We do not consder more dstant catc stores because vessel attrbutes and skll levels can cange over tme; we do not consder oter salmon speces because te co-op fsed exclusvely for sockeye. 52 Te premse underlyng tese emprcal tests, tat catc sares n te pre-co-op storc) perod are a good ndcator of skll, deserves furter dscusson. Te pre-cooperatve perod n te fsery corresponds to te ndependent sector s second stage subgame were all fsers coose dstance and fsng tmes ndependently and face a common season lengt. Te model ndcates tat all permt olders wll coose to fs te entre season n ts crcumstance. n a symmetrc Nas equlbrum, were all fs at te same dstance, all obtan te same catc per unt effort and n ts case te correspondence between skll and catc sares s precse. To see ts, let A ~ ndcate catc per unt effort andt ~ te season lengt; fserman s catc s ten A ~ ~ γ T, wc s proportonal to s skll. n a mxed equlbrum, were some fs nsde and some fs outsde, te correspondence s not exact unless all fsermen splt ter fsng tme between te two locatons n te same proporton. Ts s true because nsde effort aceves a lower catc per unt effort tan outsde effort. f skll were correlated wt coce of locaton, storc catc sares would partly reflect locaton coces and not just skll. We see no reason for a correlaton between skll and locaton coce correlaton, owever, because equlbrum proft per unt effort s te same at bot locatons. For ts reason we regard storc catc sare as an unbased but possbly mprecse ndcator of skll n te mxed equlbrum case. 29

30 set ndcates tat te mean storc catc sare of tose wo fsed for te co-op exceeded te mean for non-fsng co-op members 1.11 percent compared to 0.90 percent). Bot results agree wt predctons. Table 5 Comparson of Mean Catc Hstores for Ranked and Sorted Clusters of Fsermen # of Obs. Mean Catc Sare t-stat for dff. abs. value) ndependents v. All oners ndependents * All co-op members Fsng vs. Non-fsng Co-op oners Co-op members wo fsed * Co-op members wo dd not fs Notes: * statstcally sgnfcant at 0.05 level for a one-taled test. Te data used ere are pooled for Tests for frst-order stocastc domnance n te emprcal dstrbuton functons also agree wt predctons. Fgure 7a plots te arvest sare cumulatve densty functons for joners and ndependents usng te ranked and clustered data just descrbed. From vsual nspecton, te emprcal CDF for ndependents stocastcally domnates tat for joners,.e., te fracton of observatons wt catc sare value less tan or equal to a gven value s greater for c-op joners tan for ndependents, for all observed catc sare values. n Fgure 7b we see tat, except for a sngle excepton near te rgt tal, te emprcal CDF for co-op fsers stocastcally domnates tat for non-fsng co-op members. Kolmogorov-Smrnov tests confrm tat te dfferences n te CDFs are statstcally sgnfcant by conventonal standards Te test results are avalable from te autors. 30

31 Fgure 7a: CDF of Catc sare 1 For Co-op oners and ndependents durng Catc sare Coop oners ndependents Fgure 7b: CDF of Catc sare For oners wo Fsed and joners wo dd Not Fs durng catcsare Coop Non-Fsers Coop Fsers Value of te Fsery A remanng queston s: ow were profts n te fsery affected by allowng te co-op to form? We lack data on ndvdual frm-level profts, but we do ave data on te value of fsng permts. Te value of a Cgnk fsng permt sould reflect te expected present value proft tat a margnal low skll) fserman could earn n ts fsery. Te margnal fserman s proft s relevant, rater tan te glner s proft, because gnorng dfferences n non-pecunary returns) te margnal fserman would ave te lowest reservaton prce for sellng a permt, and would terefore determne te transacton prce to potental buyers. 31

32 Table 6 sows our estmate of te effect of te co-op polcy on permt value usng te same panel-regresson model and controls used n Tables 1 and 4. Te result ndcates tat te co-op polcy ncreased te value of a permt by $48,814 n 2008 dollars. Ts s a 26.3 percent ncrease relatve to $185,806, wc was te mean value of a Cgnk permt over 1998 to 2008 excludng te coop years. 54 Table 6 Panel Regresson of Permt Value n 2008 dollars) ndependent Varables Y = permt value Constant 7,962 Co-op Polcy t-statstc 48,814* 2.32) Fsery-Wde TAC ) Fxed Effects Year Dummes Fsery Dummes ncluded ncluded Observatons Adjusted R Notes: 1) * Sgnfcant at 0.05 level for a one-taled t-test. 2) Year and fsery dummes are as n Table 1. 3) Te permt value data are adjusted by te CP and are presented n 2008 dollars. 4) Data are avalable at: ttp:// and ttp:// We translate te permt value effect nto an annual proft effect, as follows. Te permt value dfference n Table 6 presumably reflects te co-op s effect on te present value of expected future proft for te margnal arvester. Wle t was operatng, owever, te co-op s lfe span was unknown. We deal wt ts uncertanty by estmatng a range of values for te mpled annual proft effect, eac based on a dfferent assumpton about te co-op s expected lfe span. Te lawsut tat eventually ended te co-op was fled n Aprl 2002 Grunert v. State 2005, p. 928), just before ts frst year of operaton. We terefore set te lower bound lfe expectancy at 54 As before we estmate a verson of te regresson n Table 6 by collapsng te data nto averages for eac fsery durng tree tme perods before te co-op years, durng te co-op years, and after te co-op years. Ts approac generates consstent standard error estmates Bertrand et. al. 2004). n our case, te coeffcent on te Co-op Polcy for te collapsed data s 49,024 wt a t-statstc of

33 3 years, ts actual perod of operaton. We set te upper bound at nfnty, correspondng to an expectaton tat t would persst n perpetuty. Te estmated proft effect s calculated as follows. Let π ndcate te expected annual proft before te co-op formed, and assume t s constant; let V ndcate te pre-co-op lcense value and let r be te nterest rate. Assumng lcense values observed before te co-op formed dd not ncorporate expected profts from te co-op s possble formaton, te precedng varables are lnked byv = π / r. Let ΔV be te cange n lcense value resultng from te co-op s formaton, wc we estmate, and let T ndcate te number of years te co-op was expected to operate. We ws to estmate te proportonate cange n proft resultng from allowng te co-op to form, Φ. T + 1 Te approprate present value formula gves { / } {1 1/1 ) 1 ΔV = Φπ r + r }. Te term of nterest, Φ, can now be found by combnng te two precedng expressons: Φ = Δ T V / V {1 1/1 + r) }. Table 7 Proportonate proft ncrease from allowng co-op to form ncrease n lcense value $48,814 Baselne lcense value $185,806 Coop operatng orzon years) Proportonate proft gan Φ) r=.10) Proportonate proft gan Φ) r=.07) Applyng ts formula to te data yelds te results n Table 7. Te lower-bound estmate of te annual gan n te margnal fserman s proft due to te co-op s formaton s 26 percent. f partes bddng for Cgnk lcenses tougt te co-op would last for 5 years, te mpled proportonate effect on annual proft s percent and oter entres n Table 7 ave smlar nterpretatons. Pareto mprovements, te TAC Allocaton Rule and te Co-op Lawsut Our emprcal evdence on te queston of Pareto mprovements conssts of data on te storc catc of co-op joners and ndependents, te regulator s TAC allocaton rule and te lawsut tat callenged te co-op. Our model Prop. 4) ndcates tat dvdng te TAC between 33

34 te co-op and ndependent sectors n proporton to aggregate skll, correspondng toθ =, would make tose wo coose to jon te co-op better off and leave tose wo coose to fs as ndependents ndfferent. Ts s a knfe-edge Pareto mprovement, owever; even a slgt devaton from ts TAC dvson tat dsfavors te ndependents θ > θ ) would make all ndependents worse off, and presumably cause tem to oppose te co-op s formaton. Te allocaton rule set fort wen te co-op was frst autorzed descrbed at te begnnng of ts secton) resulted n a TAC sare for te co-op of n 2002, ts frst year of operaton. Ts sare resulted from avng 77 joners and a nne-tents per capta sare θ = 0. 9 ) for eac =. 693). Our model ndcates tat storc, pre-co-op, catc s a natural measure of skll. Te aggregate storc ) catc sare of tose wo cose to jon n 2002 was wtn 1 percentage pont of te co-op s assgned TAC sare. Based on te model, terefore, t appears tat ts outcome and te essentally dentcal outcome n 2003 were rgt on te knfe s edge for a Pareto mprovement. By 2004, owever, te number of co-op joners ad ncreased from 77 to 87. To ensure a Pareto mprovng outcome as te sze of te ndependent fleet declned, te TAC allocaton granted for eac ndependent permt older would need to be ncreased.e. θ would need to declne). Ts s true because tose leavng te ndependent sector to sgn on wt te co-op would be te least sklled ndependents Prop. 3); ence, tose remanng would be te most sklled ndependents. Te actual allocaton formula dd just te opposte. Once te number of co-op members reaced 85 n 2004, te allocaton rule reduced te ndependent sector s TAC sare to concde wt te proporton of permt olders tat cose to fs ndependently. Ts corresponds to an allocaton based onθ = 1wc, accordng to our model, dsadvantages all ndependents. Roug calculatons ndcate tat t would ave been necessary to ncrease te ndependent sector s per capta TAC allocaton by at least 10% to ensure a Pareto mprovement; nstead t was reduced by 40%. Te lawsut callengng te co-op polcy was fled by Mcael Grunert and Dean Anderson. Consstent wt te model s predctons, bot were among te gest earnng Cgnk permt olders and neter joned te co-op. Te fact tat Grunert and Anderson fled te lawsut n 2002 suggests tat tey assgned a postve probablty to te number of joners growng over tme to te pont were glners would become dsadvantaged, wc clearly seems to be wat appened by c θc 34

35 4. Conclusons We ave modeled te potental gans from coordnatng te actons of common pool extractors and presented emprcal evdence on tese gans n te context of te fsery. Muc of te receved lterature on fsery management empaszes te ablty of market-based nstruments suc as ndvdual catc sares TQs) to elmnate redundant fsng unts and end wasteful races tat result from te rule of capture. We extend ts work by demonstratng tat te value of a common pool resource can be furter enanced by coordnatng te actons of extractors. Wle ndvdual rgts olders lack ncentves to coordnate, a arvestng group wt te contractual autorty to drect ts members nputs can be structured to capture tese gans, essentally by actng as a sngle frm. Coordnaton gans result from provdng publc nputs suc as sared nfrastructure and sared nformaton on stock locatons and from coordnatng arvesters actons over space and tme n cases were te value of stock vares over tese dmensons. Our emprcal results from te Cgnk case ndcate tat te effcency gans can be substantal. Allocatng a porton of te allowed catc to a group of arvesters to manage as tey see ft wtn broad constrants s a growng trend n fsery management. Examples from te U.S. are recently formed sector allocatons for groundfs n te New England regon, allocatons to cooperatves for arvestng Alaska pollock and Pacfc Wtng and te Cgnk cooperatve examned ere. Te reasons cted for ts trend nclude ponts empaszed ere: te relatve ease of assgnng rgts among a few sectors and te gans from coordnatng effort. 55 Te lawsut tat ended te co-op glgts a consderaton seldom mentoned n te lterature on common pool management te value and dffculty of desgnng polcy n a way tat enables reform wtout losers. n te Cgnk case, te queston of weter or not te co-op s formaton would lead to a Pareto mprovement was determned by tree factors: te fact tat jonng te co-op was voluntary, te regulator s rule for dvdng te allowed catc between co-op and ndependent sectors, and te co-op s nternal rule for sarng profts. Te co-op s voluntary nature was advantageous because t provded a vecle for lmtng ndvdual losses by allowng any dssentng partes to contnue under a regme tat resembled te status quo. To aceve actual loss avodance, owever, requred a precse dvson of te allowed catc between sectors and ts 55 Sullvan 2000) comments on te ease of assgnng catc sares among members of two mportant fsng cooperatves. Evdence from New Zealand demonstrates tat coordnatng enttes can be layered onto exstng ndvdual catc sare systems, by allowng quota olders to form assocatons of arvesters to coordnate ter actons. n New Zealand s paua abalone) and sea scallop fseres, arvester groups ave formed assocatons or a sngle frm n te case of scallops) tat allocate effort spatally, sare nformaton on stock denstes, support researc and stock enancement efforts and carry out oter actons tat beneft te fsery as a wole. None of tese actons s n te nterest of any sngle arvester. 35

36 was complcated by te fact tat ndvduals self-selected nto te two sectors on te bass of skll. n te Cgnk case, ger skll fsermen cose to reman ndependent, necesstatng tat ts sector receve a dsproportonately large catc allocaton. Wle te actual dvson ncorporated ts to a degree, t apparently dd not go far enoug. Fnally, t s clear tat te co-op could ave altered ts equal-proft sare rule n a way tat would ave ganed more support from g skll fsermen. 56 Te co-op s founders consdered alternatve proft sare rules n ntal delberatons, but tese negotatons proved dffcult. n te end a smple equal dvson rule was adopted. Despte evdence of potental gans from varous forms of rgts-based management, 57 less tan 2% of te world s fseres currently employ te most promnent rgts-based regme, te ndvdual catc sare. At least tree factors account for ts deart of mplementaton. Frst, ncumbent fsermen often vocally oppose catc sares on te grounds tat tey elmnate free access to te resource. Second, te ntal allocaton of rgts nvtes rent-seekng contenton. Trd, te ndvdual transferable quota TQ) model tat as aceved some success n Alaska, celand, New Zealand, and elsewere, may stll leave sgnfcant rents on te table by falng to aceve potental gans from coordnatng te actons of ndependent quota olders. Te rgts-based approac exemplfed by te Cgnk experment, wt catc rgts assgned to groups formed voluntarly, makes progress toward overcomng eac of tese mpedments. t elps to defuse te rgt to fs argument by offerng all partcpants te rgt to opt nto a sector governed by te status quo management regme. Wle t does not elmnate wasteful struggles over te ntal allocaton, t arguably reduces te magntude of te problem by assgnng to voluntary groups te task of negotatng catc sares among ter members, wle requrng te regulator only to make te gross dvson of catc between sectors. Fnally, by enablng coordnaton among ndvduals ts approac can substantally ncrease rents n te fsery, makng te often contentous process of reform a more lucratve postve sum game tan t oterwse would be. 56 n fact, one of te two glners wo fled te sut tat ended te co-op as argued n favor of a fsery management approac n wc all arvestng s by cooperatves and co-op profts are sared n proporton to storc catc sares. See Anderson 2002). 57 Evdence n favor of rgts-based management s extensve. Hannesson 2004) provdes a broad revew of teoretcal and emprcal work. Grafton et al. 2000) found large ncreases n rents assocated wt ntroducng a rgt-based system for Canadan Halbut. Leal 2002) surveys several US and Canadan TQ programs and fnds smlar effects. Lnn et al. 2008) estmate tat ntroducng a rgts-based system for west coast groundfs n te U.S. could reduce costs by 60 percent, some $18-$20 mllon annually. Newell et al. 2005) document sgnfcant annual rents n New Zealand s many TQ programs. Costello et al. 2008) document reductons n collapse from TQs. 36

37 References Anderson, Dean, Tere s More tan One Way to Craft a Co-op. Alaska Fserman s ournal, December: Barrett, Scott, Self-enforcng nternatonal Envronmental Agreements, Oxford Economc Papers, Vol. 46: Beddngton, R, Agnew, D, Clark, CW, Current problems n te management of marne fseres., Scence, 2007, Vol: 316, Pages: Bertrand, Maranne, Duflo, Ester and Sendl Mullanatan How Muc Sould we Trust Dfference-n-Dfference Estmates? Quarterly ournal of Economcs: Bouwens, Kennet and Aaron Poetter Cgnk Management Area Annual Management Report. Alaska Department of Fs and Game. Avalable at: ttp:// Bouwens, Kennet and Mark. A. Stcert Cgnk Management Area Annual Management Report. Alaska Department of Fs and Game. Avalable at:ttp:// Boyce, ndvdual Transferable Quotas and Producton Externaltes n a Fsery. Natural Resource Modelng 64): Coase, Ronald H. November 1937). "Te Nature of te Frm". Economca 4 16): Cornes, Rcard, and Todd Sandler, Te Teory of Externaltes, Publc Goods and Club Goods, 2 nd Ed., Cambrdge: Cambrdge Unversty Press. Costello, Crstoper, Steven D. Ganes and on Lynam, Can Catc Sares Prevent Fseres Collapse? Scence Vol. 321 September 19, 2008): Costello, Crstoper and Robert T. Deacon Te Effcency Gans from Fully Delneatng Rgts n an TQ Fsery, Marne Resource Economcs Vol. 22: d Aspremont, Claude, Alexs acquemn, ean askold Gabszewcz, and on A. Weymark, On te Stablty of Collusve Prce Leadersp, Canadan ournal of Economcs, Vol. XV, No. 1: Deacon, Robert T., Domnc P. Parker and Crstoper Costello, mprovng Effcency by Assgnng Harvest Rgts to Fsery Cooperatves: Evdence from te Cgnk Salmon Coop, Arzona Law Revew Vol. 50 Summer 2008): da Slva, P. P, and A. W. Ktts, Collaboratve Fseres Management n te Norteast US: Emergng ntatves and Future Drectons. Marne Polcy Gordon, H.S Te teory of a common property resource: te fsery,. Polt. Economy 62: Grafton, R. Quentn, et al, ncentve-based Approaces to Sustanable Fseres, Canadan ournal of Fseres and Aquatc Scence Vol. 63: Grafton, R, D. Squres, and K. Fox. Prvate property and economc effcency: A study of a common pool resources. Te ournal of Law and Economcs. 43: Grfft, Davd Te ecologcal mplcatons of ndvdual fsng quotas and arvest cooperatves. Fronters n Ecology and te Envronment. 64):

38 Grunert v. State, 109 P.3d 924, Alaska 2005, p Halpern et al A global map of uman mpact on marne ecosystems. Scence ): Hamlton, Barton H., ack A. Nckerson and Hdeo Owan, Team ncentves and Worker Heterogenety: An emprcal Analyss of te mpact of Teams on Productvty and Partcpaton. ournal of Poltcal Economy, Vol. 111,No. 3 une 2003): Hannesson, R Te Prvatzaton of te Oceans. MT Press. Hlborn, R.,. M. Orensanz, and A. M. Parma nsttutons, ncentves and te future of fseres. Plosopcal Transactons of te Royal Socety of London Seres B-Bologcal Scences, 360:47-57 ackson et al., 2001.B.C. ackson et al., Hstorcal overfsng and te recent collapse of coastal ecosystems, Scence ), pp onson, R. and G. Lbecap Contractng problems and regulaton: te case of te fsery. Amercan Economc Revew. 725): Ktts, A. W., and S. F. Edwards, fortcomng. K Cooperatves n US Fseres: Realzng te Potental of te Fsermen s Collectve Marketng Act. Marne Polcy. Knapp, G. and L. Hll Effects of te Cgnk Salmon Cooperatve: Wat te Permt Holders Say. UA Researc Summary. une. Knapp, Gunnar; Sver, Darla; Deroce, Pat, and Alexandra Hll Effects of te 2002 Cgnk Salmon Cooperatve: A Survey of Cgnk Salmon Permt Holders. SER Report, Unversty of Alaska Ancorage. Avalable at: ttp:// Knapp, Gunnar and Lex Hll Effects of te Cgnk Salmon Cooperatve: Wat Permt Holders Say. UA Researc Summary No. 1. SER, Unversty of Alaska Ancorage. Avalable at: ttp:// Kolstad, Carles, and Alastar Ulp, Learnng and nternatonal Envronmental Agreements. Clmatc Cange Vol 89: Lau, Lawrence., Yngy Qan, and Gerard Roland, Reform wtout Losers: An nterpretaton of Cna s Dual-Track Approac to Transton, ournal of Poltcal Economy, Vol. 108, No. 1: , Leal, Donald R. Ed Evolvng Property Rgts n Marne Fseres. Rowman & Lttlefeld Publsers, nc. Lanam, MD. Leal, Donald R., Fencng te Fsery: A Prmer on Endng te Race for Fs, Property and Envronment Researc Center, Bozeman, MT. Lbecap, Gary, and Steven N. Wggns, Contractual Responses to te Common Pool: Proratonng of Crude Ol Producton, Amercan Economc Revew, Vol. 74, No. 1 Marc 1984): Lnn, Sng, Wennger Placeolder <Estmate effects on costs of west-coast TQ> Matulc, Scott C., Murat Sever and Fred naba, Fsery Cooperatves as an Alternatve to TQs: mplcatons of te Amercan Fseres Act. Marne Resource Economcs 16)

39 Meyer, Bruce D., Natural and Quas-Experments n Economcs, ournal of Busness and Economc Statstcs, Vol. 13, no Myers RA, Worm B 2003) Rapd worldwde depleton of predatory fs communtes. Nature 423: Newell, R.G.,.N. Sancrco and S. Kerr., Fsng Quota Markets, ournal of Envronmental Economcs and Management, vol ), pp Ostrom, E Governng te Commons. Cambrdge Unversty Press. Pappas, George and Kevn Clark Cgnk Management Area Commercal Salmon Fsery, Stock Status, and Purse Sene Cooperatve Fsery Report. Report to te Alaska Board of Fseres. Avalable at: ttp:// Provencer, Bll, and Oscar Burt, A Prvate Property Rgts Regme for te Commons: Te Case for Groundwater, Amercan ournal of Agrcultural Economcs, Vol. 76 November 1994): Ross, ame, 2002a. nsde te Co-op, Part : Radcal Surgery Prevents Slow Deat, Alaska Fserman s ournal, November Ross, ame, 2002b. nsde te Co-op, Part : Te Cgnk Fsery and te Aftermat, Alaska Fserman s ournal, December Scelle, Kurt, Carlson, Stefane, and Nancy Free-Sloan Cgnk Salmon Purse Sene Fsery: Summary Data on ssues Related to te 2002 Cooperatve Fsery. Report Prepared for e Alaska Board of Fseres. Avalable at: ttp:// Scott, A Te fsery: te objectves of sole ownersp,. Polt. Economy 63: Scott, Antony, Movng Troug te Narrows: From Open Access to TQs and Selfgovernment. n Ross Sotton ed., Use of Property Rgts n Fseres Management, FAO Fseres Tecncal Paper 404/1, at Stcert, Mark A Cgnk Management Area Annual Management Report. Alaska Department of Fs and Game. Avalable at: ttp:// Sullvan,. M Harvestng Cooperatves and US Anttrust Law: Recent Developments and mplcatons. Accessed at: ttp://oregonstate.edu/dept/fet/2000/papers/sullvan.pdf. Tetenberg, Tom, Te Tradable Permts Approac to Protectng te Commons: Wat Have We Learned? n Ostrom, Elnor, et al, Te Drama of te Commons. Wasngton DC: Natonal Academy Press. Townsend, R. E Producer Organzatons and Agreements n Fseres: ntegratng Regulaton and Coasean Barganng. n Evolvng Property Rgts n Marne Fseres, Ed. D. Leal. Rowman & Lttlefeld Publsers, nc. Lanam, MD. Townsend, R. E., Fseres Self-Governance: Corporate or Cooperatve Structures? Marne Polcy 191) Wggns, Steven N., and Gary Lbecap, Ol Feld Untzaton: Contractual Falure n te Presence of mperfect nformaton, Amercan Economc Revew, Vol. 75, No. 3 une 1985):

40 Wlen, ames TURFs and TQs: Coordnated vs. Decentralzed Decson Makng. Paper presented at te Worksop on Advances n Rgts-based Fseres Management, Reykjavk, celand, August Wlen, ames E., Property Rgts and te Texture of Rents n Fseres. n Donald R. Leal, ed. Evolvng Property Rgts n Marne Fseres Oxford: rowman & Lttlefeld Publsers, nc. Wlson, Fsng for knowledge. Land Economcs: 661) Worm B, Barber EB, Beaumont N, Duffy E, Folke C, Halpern BS, ackson BC, Lotze HK, Mcel F, Palumb SR, Sala E, Selkoe K, Stacowcz, Watson R 2007b) Response to comments on 'mpacts of bodversty loss on ocean ecosystem servces'. Scence 316:1285 Worm, B., R. Hlborn,. Baum, T. Branc,. Colle, C. Costello, M. Fogarty, E. Fulton,. Hutcngs, S. ennngs, O. ensen, H. Lotze, P. Mace, T. McClanaan, C. Mnto, S. Palumb, A. Parma, D. Rcard, A. Rosenberg, R. Watson, and D. Zeller Rebuldng global fseres. 325: Yandle, Tracy, Te Callenge of Buldng Successful Stakeolder Organzatons: New Zealand s Experence n Developng a Fseres Co-management Regme, Marne Polcy

41 Appendx A.1 Co-op s optmal polcy Te co-op s optmal allocaton mnmzes mn d, T ; x F ) Z α + F β ) Z d F β ) { Z G x x } + β ) φ T A.1) subject to 1 γ T = F β ) Z, d { 0,d}, T [ 0, T ] for all, andt T. Snce A.1) s strctly ncreasng n d te optmal polcy sets d = 0 for eac member. Te term n brackets s te net beneft tat te publc nput provdes. Gven assumed propertes of G x ) te followng frst-order condton s necessary and suffcent for mnmzng A.1) wt respect to x : ) Z G ) 1 = 0 1 F β x. A.2) Ts s te Samuelson condton for effcent publc nput provson. t remans to fnd an assgnment of member fsng tmes tat mnmzes te fourt term n A.1), subject to te catc constrant. Te catc constrant for group mples te followng constrant on effort: γ 1 T Z F β ). A.3) ndex co-op members n ncreasng order of te ratoφ γ. Snceφ andγ are s cost per unt tme and effort per unt tme respectvely, ts rato s 's cost per unt effort. Consder a polcy, denoted Λ, wc assgns fsng tmet to successve co-op members, n order of ter ndex, untl te constrant A.3) s volated or satsfed wt equalty. f A.3) s volated, let î ndcate te gest ndexed member n ts low ndexed subset and assgn ts member a fsng tme tat satsfes A.3) exactly; all ger ndexed members are assgned zero fsng tme. Ts assgnment satsfes te catc constrant by constructon. To see tat ts assgnment s cost φ mnmzng, wrte te fourt term n A.1) as γ T γ. Te termγ T s te fsng effort 41

42 42 assgned to and te rato s 's cost per unt effort. Any alternatve to polcy Λ would requre reducng T γ by a lower ndexed member and ncreasng T γ n te same amount by a ger ndexed member. Snce te ndex orders members n terms of te rato γ φ, ts alternatve assgnment would necessarly result n ger total cost. Terefore te assgnment of fsng tmes n polcy Λ s cost mnmzng. A.2 Publc nput beneft per member ncreases wt co-op sze Te publc nput confers a net beneft, per co-op member, of ) ) = x x G K n n Z F n n NB ) ) ) ) 1 ) 1 β. A.4) Dfferentatng ts wt respect to ) n yelds ) ) ) ) ) 1 ) ) ) ) 1 ) ) ) ) ) 1 ) ) n x x G K n n Z F n x x G K n n Z F K n Z F n n n NB = β β β A.5) Makng use of te frst-order condton A.2) and smplfyng, yelds te desred result ) 0 ) ) ) 2 > = n x n n NB. A.5)

43 Fg. A1 Map of Cgnk Management Area on te Alaskan Pennsula Source: Stcert 2007). Fg. A2 Cgnk Management Area wt Dstrct Boundares and Statstcal Areas Source: Stcert 2007). 43

44 Fg. A3 Map of Cgnk Bay and Near Vcntes Source: Stcert 2007). 44