Information Sharing and Oligopoly in Agricultural Markets: The Role of Bargaining Associations

Transcription

1 CARD Workng Papers CARD Reports and Workng Papers 9-00 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons Brent M. Hueth Iowa State Unversty Phlppe Marcoul Iowa State Unversty Follow ths and addtonal works at: Part of the Agrcultural and Resource Economcs Commons, Agrcultural Economcs Commons, and the Industral Organzaton Commons Recommended Ctaton Hueth, Brent M. and Marcoul, Phlppe, "Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons" (00). CARD Workng Papers Ths Artcle s brought to you for free and open access by the CARD Reports and Workng Papers at Iowa State Unversty Dgtal Repostory. It has been accepted for ncluson n CARD Workng Papers by an authorzed admnstrator of Iowa State Unversty Dgtal Repostory. For more nformaton, please contact dgrep@astate.edu.

2 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons Abstract We study ncentves for nformaton sharng (about uncertan future demand for fnal output) among frms n mperfectly compettve markets for farm output. Informaton sharng generally leads to ncreases n expected total welfare but may reduce expected frm profts. Even when expected frm profts ncrease, nformaton sharng does not represent equlbrum behavor because frms face a prsoner s dlemma n whch t s prvately ratonal for each frm to wthhold nformaton, gven that other frms report truthfully. Ths equlbrum can be overcome f frms commt to smultaneously reportng ther nformaton and f reports are verfable. We argue that agrcultural barganng assocatons serve both these roles. Keywords agrcultural markets, barganng, mperfect competton, nformaton sharng Dscplnes Agrcultural and Resource Economcs Agrcultural Economcs Industral Organzaton Ths artcle s avalable at Iowa State Unversty Dgtal Repostory:

3 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons Brent Hueth and Phlppe Marcoul Workng Paper 0-WP 313 September 00 Center for Agrcultural and Rural Development Iowa State Unversty Ames, Iowa Brent Hueth and Phlppe Marcoul are assstant professors of economcs at Iowa State Unversty. Prepared for presentaton at the Economcs of Contracts n Agrculture, Second Annual Workshop, July 1 3, 00, Annapols, Maryland. Supported n part by the Agrcultural Marketng Resource Center, Iowa State Unversty. Ths publcaton s avalable onlne on the CARD webste: Permsson s granted to reproduce ths nformaton wth approprate attrbuton to the authors and the Center for Agrcultural and Rural Development, Iowa State Unversty, Ames, Iowa For questons or comments about the contents of ths paper, please contact Brent Hueth, 371 Heady Hall, Iowa State Unversty, Ames, Iowa ; Phone: ; Fax: ; E-mal: bhueth@astate.edu Iowa State Unversty does not dscrmnate on the bass of race, color, age, relgon, natonal orgn, sexual orentaton, sex, martal status, dsablty, or status as a U.S. Vetnam Era Veteran. Any persons havng nqures concernng ths may contact the Offce of Equal Opportunty and Dversty, 1031 Wallace Road Offce Buldng, Room 101,

4 Abstract We study ncentves for nformaton sharng (about uncertan future demand for fnal output) among frms n mperfectly compettve markets for farm output. Informaton sharng generally leads to ncreases n expected total welfare but may reduce expected frm profts. Even when expected frm profts ncrease, nformaton sharng does not represent equlbrum behavor because frms face a prsoner s dlemma n whch t s prvately ratonal for each frm to wthhold nformaton, gven that other frms report truthfully. Ths equlbrum can be overcome f frms commt to smultaneously reportng ther nformaton and f reports are verfable. We argue that agrcultural barganng assocatons serve both these roles. Keywords: agrcultural markets, barganng, mperfect competton, nformaton sharng

5 INFORMATION SHARING AND OLIGOPOLY IN AGRICULTURAL MARKETS: THE ROLE OF BARGAINING ASSOCIATIONS Introducton Many markets for farm output are plausbly characterzed by some degree of mperfect competton. Ths s certanly true n most frut and vegetable markets where growers are numerous, and where ntermedaton (e.g., processng or shppng/packng) s relatvely concentrated. Processng or packng cooperatves, and cooperatve barganng among farmers, may n some nstances be nsttutonal responses to these market mperfectons n both cases, farmers can enhance ther barganng poston n prce negotatons wth noncooperatve ntermedares (e.g., Sexton 1990). Ths perspectve emphaszes the effect of cooperaton on the dstrbuton of surplus among partcpants n agrcultural markets. An alternatve vew the one we explore n ths paper s that cooperaton s an nsttutonal response affectng the effcency of market transactons and hence ndrectly the dstrbuton of surplus among partes. Brefly, we consder an mperfectly compettve market for farm output n whch nformaton sharng among ntermedares (about uncertan future demand for fnal output) potentally leads to hgher expected aggregate surplus. In ths context, we show that a barganng assocaton can solve a prsoner s dlemma among frms where all partes (frms, consumers, and growers) are better off when nformaton s fully shared, but where each frm s domnant strategy s to not reveal ts nformaton. The barganng assocaton serves two roles. Frst, t nvests costly resources n verfcaton of frm reports (frms can choose not to reveal ther nformaton, but f they do reveal, t s mpossble to le), and second, t provdes a mechansm by whch all partes commt to smultaneously revealng ther nformaton. We wll argue that these two functons are reasonable descrptons of what barganng assocatons actually do (among other thngs), and that they serve to solve the prsoner s dlemma.

6 / Hueth and Marcoul In what follows, we begn wth a descrpton of barganng n agrcultural markets. We then develop a model of nformaton sharng based on the work of Vves (1984), Rath (1996), and L (1985) (see Vves 1999, chap. 8, for an excellent summary of ths lterature), and demonstrate how a barganng assocaton can lead to effcency gans. The fnal secton concludes and dscusses the emprcal mplcatons of our model. Barganng n Agrcultural Markets Our ntent n ths secton s not to provde an exhaustve overvew of agrcultural barganng but rather to pont out ways n whch descrptons of the nsttutonal features of barganng assocatons seem consstent wth the noton that barganng can have effcency consequences. To begn, t s noteworthy that barganng occurs prmarly n markets for processng fruts and vegetables. Ths partcular set of markets comprses only a small porton of all agrcultural markets, and t s natural to ask why barganng assocatons aren t more wdespread. If the success of barganng as an nsttuton hnges on delverng hgher prces to growers, we should expect to observe barganng n a larger class of commodtes. At least two features of markets for processng fruts and vegetables seem relevant to ths pont. Frst, processors obtan ther output prmarly through forward contracts, so that tradtonal modes of prce dscovery are absent. Moreover, procurement decsons are typcally made n the context of uncertanty about the state of future demand (e.g., pror to plantng). To the extent that prce negotatons durng barganng facltate ndustry-wde communcaton about future demand, barganng can thus be vewed as a sort of ndrect prce dscovery mechansm. Second, processng of fruts and vegetables s hghly concentrated (both spatally and n numbers of frms), so that t s reasonable to expect some degree of olgopoly behavor by frms. Ths fact possbly ncreases the potental benefts of collectve barganng and (spatal concentraton) reduces organzatonal and communcaton costs among growers. Although the effcency consequences of barganng n the context of olgopolstc markets are not mmedately apparent, n the next secton we show how centralzed prce dscovery (or nformaton sharng ) effectvely ncreases competton among olgopsony buyers, thus leadng to a more effcent outcome. 1

7 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons / 3 Results and dscusson from two studes of barganng seem consstent wth the noton that barganng s more than smply prce enhancement. Frst, n a natonal survey of (processng) frut and vegetable barganng assocatons, Iskow and Sexton (199) note that the majorty of assocatons felt ther role was not only to mprove the well beng of grower members, but also to provde servces to processors. Of the servces provded, ncreased prce stablty, mproved nformaton, and mproved prce dscovery process were mostly frequently cted. Lackng smlar responses from processng frms, t s dffcult to know whether n fact such servces are provded and valued. Nevertheless, the fact that nearly all respondents vew prce dscovery and mproved nformaton as mportant servces provded by ther respectve assocatons s certanly consstent wth the hypothess that barganng can ncrease effcency. Bunje (1980) offers a comprehensve descrpton of barganng n U.S. agrcultural markets; 3 n summarzng the role of farm barganng he smlarly notes: Barganng assocatons can fll the needs of the market as well as the needs of the ndvdual producer. They can serve a supply coordnatng functon for the market and furnsh market ntellgence for the producer. They can operate as a prce dscovery vehcle, establsh market prces, and establsh unform terms of trade that serve the producer and the marketplace. Whle such a quote mght be vewed as self-servng comng from a representatve of barganng assocatons, t agan conveys the dea that, at least n the mnds of those who operate barganng assocatons, barganng can lead to effcency gans. In what follows, we formally analyze one possble source of such a gan. We develop an olgopoly model of n frms who produce substtute fnal goods, and who obtan ther raw farm nput from a group of homogeneous growers (represented by an aggregate supply relaton). Before procurement, each frm s uncertan about the true state of future demand but receves an mperfect sgnal of demand. We study prvate ncentves for frms to share (or pool) ther sgnals, and the correspondng welfare mplcatons. In ths context, we nterpret the ntensve communcaton that occurs durng the annual barganng process as a means of mplementng nformaton sharng.

8 4 / Hueth and Marcoul Model The Setup There are n frms who convert farm output nto a vector of fnal consumpton goods q = (q 1,, q n ), where q represent the quantty of fnal goods sold by frm. For smplcty, we suppose that each frm transforms q nto fnal output n Leontef fashon wth constant margnal cost (normalzed to zero), and moreover that a sngle unt of farm output yelds a sngle unt of fnal output. Thus, for gven output prce p (q), and farm prce r(q), frm s profts are gven by Π(q, q - ) = [p (q) r(q)]q, where q represents the n 1 vector of outputs other than s. Growers are represented by an aggregate (nverse) supply functon r = a + bq, where Q = q s the aggregate quantty of farm output purchased. 4 Fnal goods are dfferentated and valued by a representatve consumer wth utlty functon n = 1 Uq q 1 q qq n n ( ) = ( α+ε) β + γ j, = 1 = 1 j (1) where β > γ > 0, α > 0, and where ε s a normally dstrbuted, aggregate source of uncertanty. We suppose that all frms share a common pror of 0 for the mean of ths random varable, and that ts varance, σ ε, s known. For a gven vector of prces p = (p 1,, p n ), consumers choose quanttes to maxmze U(q) demand schedules for each frm s output gven by n = 1 pq,, yeldng nverse p (q, q ) = α + ε β q γ q. () The tmng of actons n our model s as follows. In perod 0, each frm receves an ndependent (and prvate) sgnal s = ε + ν, where ν s dstrbuted normally and ndependently of ε wth E[ν ] = 0, E[ ν ] = σ ν, E[ν ν j ] = 0 for j. Thus, each s represents mperfect, though unbased, nformaton on the state of future demand. Based on these sgnals, frms form expectatons n perod 1 about demand n perod and coordnate wth growers for delvery of some quantty of farm output that arrves n j j

9 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons / 5 perod. Expectatons depend on the nformaton avalable to each frm, and we consder two scenaros. In the frst, each frm keeps ts nformaton prvate and thus forms an expectaton based on s (for frm ). Alternatvely, frms pool ther nformaton and thus form expectatons based on the full vector of sgnals s = (s 1,, s n ). Fnally, n perod, frms noncooperatvely choose prces to maxmze ther ndvdual proft, gven the quanttes of output arranged for delvery n the prevous perod. We assume effcent ratonng (e.g., Trole 1989) of quanttes, so that equlbrum prces n perod are just those that form an equlbrum when all quanttes are delvered to the market. The structure of ths market s analogous to Bertrand compettors choosng capactes n an ex ante perod, where here capactes are gven by the quantty of output arranged for delvery durng perod 1. For the equlbrum we descrbed above, t s, of course, essental that no frm can obtan addtonal output n perod (relatve to what was arranged for delvery durng perod 1). Ths s a natural feature of the markets we study gven the tme nterval requred to produce most knds of farm output. 5 Market Equlbrum Wthout Informaton Sharng In perod 1, after each frm receves ts sgnal s, the frms play a Cournot game n choosng quanttes of output for delvery n perod. For gven q, the (condtonal) expected proft of frm s gven by Π ( q, q s) = α+ E[ ε s] βq γ Eq [ j s] q (3) j where α = α a, β = β + b, and γ = γ + b. Let ρ = σ ε /( σ ε + σ ν ) represent the correlaton between s and s j. Then frms update ther prors on ε wth the formula E[ε s ] = ρs, whch s a (varance) weghted average of the pror and s. Frm s reacton functon s then gven by α+ρs γ Eq [ j s] j q( q ) =. (4) β To fnd an equlbrum for ths game, we suppose that frms use strateges that are affne n ther sgnals and we then verfy that these strateges ndeed form an equlbrum.

10 6 / Hueth and Marcoul Lettng frm s equlbrum strategy be gven by q = c 0 + c 1 s, and notng that E[s j s ] = ρs, t s straghtforward to verfy that an equlbrum s obtaned settng c 0 = α/δ, and c 1 = ρ/δ ρ, where δ = β + (n 1)γ, and δ ρ = β + (n 1)γρ, and hence that the equlbrum quantty for frm s gven by q p α ρs = + δ δ. (5) ρ For future reference, we note that E[ q ] = α/δ, E[( q ) ] = E[ q ] + σ ε ρ/ δ and p p p, ρ p p E[ qq] = E[ q ] + j p σρ ε / δ for j., ρ The full nformaton equlbrum level of producton (when α s known for certan) s gven by α/δ, so that frms ncrease or decrease ther output relatve to ths benchmark dependng on whether the realzaton of s s greater than or less than zero. The varance of sgnal nose σ ν has an ambguous effect on the slope term ρ/δ ρ. On the one hand, as σ ν decreases, frms put more weght on ther sgnals relatve to ther pror, and ths makes frms more responsve. Ths effect s reflected n the numerator of the second term n equaton (5) where a decrease n σ ν ncreases ρ. However, a decrease n σ ν also ncreases the correlaton of the frms sgnals. Ths n turn mples that f some frm, say frm, receves nformaton suggestng hgh demand, t s lkely that other frms have receved smlar nformaton. Because the outputs of each frm are substtutes, an equlbrum response to ths s a reducton n frm s output. Ths effect s reflected n the denomnator, where a decrease n σ ν ncreases δ ρ. Changes n σ ε have a smlarly ambguous, though recprocal, effect on frm responsveness. A reducton n σ ε lowers the weght placed on each frm s sgnal, makng frms less responsve, but also reduces the correlaton of sgnals, and ths tends to ncrease responsveness. Expected proft for each frm pror to observng ther sgnal s, but antcpatng equlbrum behavor for any realzaton of s, s gven by Π p = E max Π( q, q s) q (6) p whch from (4) reduces to Π =β E[( q )]. Drect calculaton from (5) then yelds p

11 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons / 7 α σρ Π p = + δ δ. (7) ε ρ The frst term n ths expresson represents the profts each frm would receve f there were no uncertanty (σ ε = 0). From ths term, expected profts are hgh when aggregate demand and supply are hgh (hgh α or low a), or when the total prce decrease resultng from a small ncrease n each frm s output s small (low δ). One consequence of nformaton sharng s an ncrease n the precson wth whch frms estmate ε. Thus, before consderng the market equlbrum wth nformaton sharng, t s natural to consder how a reducton n the varance of the sgnal error σ ν (whch reduces the varance of each frm s estmate of ε) affects expected frm profts when there s no nformaton sharng. From (7), a reducton n σ ν has a smlar qualtatve effect on profts as on the equlbrum responsveness of each frm s output to ther sgnal (prevously descrbed). Frms beneft from a reducton n the varance of sgnal nose because ther output decson more accurately reflects actual demand condtons. However, because the sgnals of each frm become more correlated, equlbrum outputs also have greater correlaton, and ths tends to reduce expected profts. Thus, whether or not frms gan from nformaton sharng generally wll depend on a drect comparson of expected profts n each regme. In the next secton, we derve an expresson for expected frm profts when nformaton s shared and make ths comparson. Market Equlbrum Wth Full Informaton Sharng Here we suppose that some mechansm s avalable for frms to share ther nformaton. Later n the paper, we ll argue that a barganng assocaton can be one such mechansm. To focus on the potental benefts from nformaton sharng, we contnue to assume that frms act as olgopsonsts n the market for farm output. 6 When nformaton sharng occurs, frms receve the full vector of sgnals s and thus form common estmates of ε. Wth n ndependent sgnals, the best estmate of ε s gven by E[ε s] = ρ n s, where ρ n = nσ ε /(nσ ε + σ ν ), and s s the mean value of the vector s. Proceedng as n the prevous secton, frm s reacton functon s then gven by

12 8 / Hueth and Marcoul yeldng the equlbrum quantty α+ρ q( q ) = n s γ Eq [ j s] j, (8) β q s α+ρ = δ n s, (9) wth E[ q ] = E[ q ] = α/δ, and E[( q ) s s ] = E[ qq ] = E[ q ] + σ ε ρ n /δ. s p s Thus, equlbrum expected output s the same regardless of whether or not frms share nformaton about ther common demand uncertanty. Frms are more responsve to ther aggregate sgnal s than to ther prvate sgnal s when j s ρ n /ρ δ/δ ρ, (10) and t s straghtforward to verfy that ths condton s always satsfed (for β > γ). Thus, the greater precson of the frms estmate of ε, and the correspondng ncrease n responsveness (measured by the term on the left hand sde of the above nequalty), outweghs the reducton n responsveness assocated wth ncreased (perfect) correlaton among the frms sgnals (measured by the rght hand sde of the nequalty). Expected frm profts wth nformaton sharng are gven by α σρ Π s = + δ δ ε n, (11) and are thus hgher than wthout nformaton sharng when ρ / ρ δ / δ. (1) n ρ The followng proposton summarzes the condtons under whch nformaton sharng leads to hgher expected frm profts. PROPOSITION 1. Informaton sharng ncreases expected frm profts when outputs are suffcently dfferentated and when aggregate farm supply s suffcently elastc (for a gven total quantty).

13 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons / 9 Intutvely, a hgh degree of product dfferentaton s analogous to each frm actng as a monopolst n the downstream market for farm output. Improved nformaton on future demand ncreases each frm s ablty to prce dscrmnate, and ths n turn ncreases expected proftablty. Smlarly, when supply s suffcently elastc, procurement decsons have a relatvely small effect on the upstream prce of farm output, and ths also enhances expected proftablty wth nformaton sharng. For the reasons dscussed earler, the magntude of ex ante uncertanty σ ε and varance of sgnal nose σ ν have ambguous effects on the expected proftablty of nformaton sharng. Welfare Comparson In ths secton, we evaluate the effect of nformaton sharng on total expected welfare and on the expected welfare of consumers and growers ndvdually. We evaluate ex ante welfare (pror to the frms recevng ther sgnals) but suppose, as n the prevous secton, that frms antcpate the equlbrum that wll obtan n ether scenaro and for a gven realzaton of s. Surplus for growers s gven by surplus s gven by 1 b (( rq) aq ) Q, = so that expected grower nb ( Eq [ ] + ( n 1) Eqq [ j] ). (13) Usng the expressons for E[ q ] and E[q q j ] obtaned n the prevous sectons, t s straghtforward to verfy that growers always beneft from nformaton sharng. Intutvely, both growers and frms gan from ncreased precson n estmatng aggregate demand. However, the ncrease n correlaton among frms outputs lowers expected frm profts, and ncreases expected grower surplus. Thus, the two effects assocated wth nformaton sharng ncreased precson n estmatng aggregate demand and ncreased correlaton among frms outputs are countervalng wth respect to frm profts but complementary wth respect to grower surplus. Consumer surplus s gven by Uq ( ) pq. Takng expectatons (assumng equlbrum behavor by frms) yelds n = 1

14 10 / Hueth and Marcoul n ( Eq n Eqq j ) β [ ] + ( 1) γ [ ]. (14) Thus, consumers also beneft from the correlaton among frms outputs, but only when there s some degree of product substtutablty. As wth grower surplus, the effects of nformaton sharng on consumer surplus tend to complement, though to a lesser degree snce E[q q j ] s weghted by γ < β. Usng the expressons for E[ q ] and E[q q j ] from the prevous secton, consumers gan from nformaton sharng whenever ρn δ β+ n γρ ( 1) ) (. ρ δ ( β+ ( n 1) γ) ρ (15) Because ρ < 1, f nformaton sharng leads to hgher expected profts, then expected consumer surplus also ncreases. When b = 0, ths nequalty wll always be satsfed snce δ = β + (n 1) γ and δ ρ = β + (n 1) γ ρ. For b > 0, condton (15) wll generally hold but can be volated. Thus, consumers generally gan from nformaton sharng, though we cannot rule out the possblty that expected consumer surplus falls. Addng up the expected surplus measures for each party, total expected surplus s gven by and s greater when nformaton s shared f n ( Eq n Eqq j ) 3 β [ ] + ( 1) γ [ ], (16) ρn δ β+ n γρ ( 1) ) (3. ρ δ (3 β+ ( n 1) γ) ρ (17) As wth consumer surplus, we cannot rule out the possblty that expected total surplus falls wth nformaton sharng, though n general t seems dffcult to volate the nequalty n (17). The followng proposton summarzes the effects of nformaton sharng on the grower, the consumer, and the total surplus.

15 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons / 11 PROPOSITION. Informaton sharng always benefts growers. Expected consumer surplus and total surplus ncrease whenever expected frm profts ncrease and may ncrease even as expected frm profts fall. Because the expressons for changes n expected proft and consumer surplus resultng from nformaton sharng yeld ambguous results, we evaluate these measures (and expected grower surplus) for a partcular specfcaton of our model. We set n = 5, α = 1, β = 0.3, a = 0, b = 0.1, σ ε = 0.3, and σ ν = 0.1. Wth ths specfcaton, we then let γ range from 0 to β and evaluate dfferences n expected surplus wth and wthout nformaton sharng. The results are dsplayed n Fgure 1. When outputs are suffcently substtutable, expected frm profts fall when nformaton s shared, though by a relatvely small amount. Growers gan most from nformaton sharng when outputs are hghly Proft Consumer Surplus Grower Surplus FIGURE 1. Dfference n expected surplus wth and wthout nformaton sharng as frm outputs become ncreasngly substtutable n consumer preferences γ

16 1 / Hueth and Marcoul dfferentated. Interestngly, the change n expected consumer surplus wth nformaton sharng s ntally ncreasng wth the degree of product substtutablty, then decreasng. Fgure dsplays the results of a smlar comparatve statc, but where we hold γ constant at 0.05 and let n range between and 10 frms. Agan, nformaton sharng leads to a decrease n expected frm profts but now for n suffcently large. Informaton sharng benefts growers (and, to a lesser degree, consumers) by a larger amount, as the number of frms ncrease. Though not reported, a decrease n b (makng supply more elastc for any gven quantty of aggregate output) ncreases expected consumer surplus wth nformaton sharng and reduces expected surplus for growers. In all cases analyzed, expected total surplus ncreases from nformaton sharng, and the beneft to frms s relatvely small (and sometmes negatve). It s also noteworthy that growers seem to gan substantally 0.06 Proft 0.05 Consumer Surplus Grower Surplus n FIGURE. Dfference n expected surplus wth and wthout nformaton sharng as the number of frms ncreases

17 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons / 13 from nformaton sharng relatve to frms, thus addng further potental beneft from barganng beyond what mght be acheved through changes n market structure (e.g., more compettve prcng of farm output; see endnote 6). Prvate Incentves to Reveal Informaton and the Role of Barganng Assocatons We have seen that nformaton exchange among frms can lead to a market equlbrum that Pareto-domnates the equlbrum wth no nformaton exchange; however, t turns out that when we examne each frm s prvate ncentve to share nformaton, a frm ncreases ts expected profts by not reportng, gven that all other frms have reported truthfully. More formally, suppose that frms play a two-stage game n whch each frm can truthfully report ts sgnal or report nothng n the frst stage, and then frms choose quanttes and prces noncooperatvely n the second stage, condtonal on equlbrum reports n the frst stage. The followng proposton (from Rath 1996) summarzes the frst stage equlbrum of ths game. PROPOSITION 3. In the two-stage game n whch frms frst decde whether or not to report ther sgnal to other frms and then choose quanttes and prces noncooperatvely (condtonal on the vector of equlbrum frst stage reports), each frm s domnant equlbrum frst-stage strategy s to not report ts sgnal. In other words, gven that all frms j report ther sgnals truthfully, frm gans by devatng and reportng nothng. Intutvely, gven that all other frms report ther sgnals, frm obtans the full benefts from ncreased precson n estmatng aggregate demand and, by wthholdng ts sgnal, reduces the correlaton among equlbrum outputs. Ths unambguously rases expected profts for frm, relatve to the equlbrum n whch t also reports ts sgnal. Thus, frms potentally face a prsoner s dlemma n whch all partes gan from nformaton sharng but equlbrum behavor s to not share. Moreover, as we ve seen n the prevous secton, ths equlbrum generally leads to lower welfare for consumers and growers. It s thus natural to consder the knds of nsttutons that mght lead to an effcent outcome. Vves (1990) and Krby (1988) suggest that trade assocatons are such an nsttuton n markets where frms outputs are strategc complements. Wth

18 14 / Hueth and Marcoul strategc complementarty, nformaton sharng unambguously ncreases expected frm profts, and t s a domnant strategy for frms to report ther nformaton. Thus, by collectng ndustry-wde nformaton and reportng aggregate statstcs, t s argued that these assocatons effectvely mplement an nformaton-sharng outcome. However, t s mportant to recognze that mplementng the sharng outcome from the earler secton requres a hghly detaled nformaton-gatherng effort by the assocaton, even when frm outputs are strategc complements. In partcular, we noted earler that frms reports of ther sgnals must be verfable n the sense that frms are unable to msreport ther sgnals (though they can choose to not report at all). Zv (1993) studes nformaton sharng when frms can strategcally dstort ther sgnals and fnds that frms always wll choose to report nontruthfully. Thus, n practce, verfablty s lkely to be a substantal nformatonal barrer, and t s not clear from exstng theoretcal work whether trade assocatons actually overcome ths barrer or whether ther prmary servce s on other dmensons (e.g., lobbyng and promotonal actvtes). In the context of agrcultural markets, barganng assocatons represent a mechansm for sharng and verfyng nformaton among frms. The annual prce negotaton that occurs wth barganng s an opportunty for explct consderaton of future demand condtons (perhaps even the prmary actvty) and s arguably a much more ntensve nformaton-gatherng effort than s the reportng of aggregate ndustry condtons (as a trade assocaton mght do). Moreover, the structure of barganng legslaton effectvely forces nformaton revelaton, snce frms are requred to engage n prce negotaton. Thus, even when t s prvately ratonal for frms to not report ther nformaton, or to msreport, barganng may effectvely force (truthful) nformaton sharng. As we ve seen, n some markets ths wll lead to an ex ante Pareto mprovement, whle n others, frms may collectvely receve lower expected profts. It s not mmedately clear how to test whether barganng assocatons are ndeed playng ths role, but the model presented above s suggestve. Suppose we take as the null hypothess that the prmary role of barganng s prce enhancement for growers. In prncple, we could dentfy dfferences n the predctons of such a model (regardng the condtons that are conducve to the formaton of barganng assocatons) wth those of the nformaton-sharng model and examne whch set of predctons best explans the

19 Informaton Sharng and Olgopoly n Agrcultural Markets: The Role of Barganng Assocatons / 15 ncdence of barganng assocatons across commodtes and regons or whether predctons assocated wth nformaton sharng add explanatory power. The key dfference n these alternatve explanatons of barganng les n the potental gan for frms from nformaton sharng. If nformaton sharng were mportant, then, ceters parbus, we d expect barganng to emerge f ts effect on expected frm profts s relatvely large (postve, or not too negatve). In contrast, f prce enhancement were the prmary functon of barganng assocatons, then we d expect the effect of barganng on frm profts to be less mportant n explanng the ncdence of barganng. Concluson We provde a ratonale for the exstence of barganng assocatons n agrcultural markets that s entrely ndependent of the role they may play n countervalng market power. In markets wth a large proporton of contracted producton, and a correspondng absence of spot markets, tradtonal modes of prce dscovery are absent. One possble substtute for prce dscovery va markets s drect communcaton among competng frms concernng expected future supply and demand condtons (pror to the annual procurement decson). In the sprt of work by Vves (1984), L (1985), and Rath (1996), we model ths communcaton as a Bayesan game among olgopolsts n whch each of n frms receves a sgnal of future demand (for smplcty we gnore supply uncertanty) and evaluate the welfare mplcatons of frms sharng ther respectve sgnals wth other frms. Informaton sharng tends to beneft consumers and growers but to have ambguous consequences for expected frm profts. Informaton sharng allows frms to ncrease the precson of estmated future demand, but because the sgnals are postvely correlated (a natural assumpton, gven the nature of the markets we study), nformaton sharng also tends to ncrease the correlaton among frms equlbrum strateges. In markets where fnal outputs are substtutes, frm strateges are strategc substtutes, so that a postve correlaton of strateges reduces expected proft. Thus, the effects of nformaton sharng tend to counterval wth respect to expected profts, and to complement wth respect to consumer and grower surplus. Even when expected profts for frms ncrease because of nformaton sharng, frms face a prsoner s dlemma n whch the equlbrum behavor of each frm s to not report

20 16 / Hueth and Marcoul ts nformaton (not reportng when other frms report reduces the correlaton of strateges, wth no effect on the precson of estmated future demand). We argue that collectve barganng represents one possble means of overcomng ths equlbrum and, more generally, of ncreasng expected total welfare.

21 Endnotes 1. Of course, f nformaton sharng leads to a collusve outcome, effcency may be reduced. Though t may be n the nterest of both processors and growers to collusvely set output, t s doubtful that n practce such an outcome can be sustaned. As we ll demonstrate, nformaton sharng wthout colluson generally leads to hgher expected surplus for consumers and growers (even when growers do not have countervalng market power), and possbly even hgher expected profts for frms.. Of the 36 assocatons sampled, 31 cted ncreased prce stablty, 3 cted mproved nformaton, and 5 cted mproved prce dscovery. When quered about servces offered to growers, only prce negotaton and tme and method of payment were smlarly cted by more than 30 assocatons. 3. Ralph Bunje was a leadng spokesman and proponent of farm barganng for over 30 years durng hs tenure as manager of the Calforna Cannng Peach Assocaton (see forward n Bunje 1980). 4. Ths specfcaton of the farm sector gnores grower heterogenety, whch may be mportant n consderng the ncentves for growers to form a barganng assocaton. We d lke to consder the ndustry wde ncentves to form a barganng assocaton, ndependent of the organzatonal and admnstratve dffcultes created by grower heterogenety. 5. It s also worth notng that we take as a gven each frm s desre to coordnate wth growers n perod 1 (rather than to compete for aggregate output n perod ). Ths s consstent wth the noton that frms contract wth growers, rather than purchasng output on some knd of spot market. Understandng why frms choose to contract s an nterestng queston, but one that les beyond the scope of ths paper. Interestngly, as noted n the prevous secton, the absence of spot markets (and the correspondng prevalence of contracted arrangements) seems to be a necessary condton for the establshment of barganng assocatons. 6. Barganng that leads to compettve prcng for farm output would, of course, generate effcency gans, but we d lke to evaluate the benefts from nformaton sharng ndependent of changes n market structure.

22 References Bunje, R.B Cooperatve Farm Barganng and Prce Negotatons. Techncal Report 6, Agrcultural Cooperatve Servce, U.S. Department of Agrculture, Washngton, D.C. Iskow, J., and R. Sexton Barganng Assocatons n Grower-Processor Markets for Fruts and Vegetables. Techncal Report 104, Agrcultural Cooperatve Servce, U.S. Department of Agrculture, Washngton, D.C. Krby, A.J Trade Assocatons as Informaton Exchange Mechansms. Rand Journal of Economcs 19(1): L, L Cournot Olgopoly wth Informaton Sharng. Rand Journal of Economcs 16(4): Rath, M A General Model of Informaton Sharng n Olgopoly. Journal of Economc Theory 71: Sexton, R.J Imperfect Competton n Agrcultural Markets and the Role of Cooperatves: A Spatal Analyss. Amercan Journal of Agrcultural Economcs 7(3, August): Trole, J The Theory of Industral Organzaton. Cambrdge, MA: MIT Press. Vves, X Duopoly Informaton Equlbrum: Cournot and Bertrand. Journal of Economc Theory 34: Trade Assocatons, Dsclosure Rules, Incentves to Share Informaton, and Welfare. Rand Journal of Economcs 1(3): Olgopoly Prcng: Old Ideas and New Tools. Cambrdge, MA: MIT Press. Zv, A Informaton Sharng n Olgopoly: The Truth Tellng Problem. Rand Journal of Economcs 4(3):