Measuring Market Power in the UK Retail Salmon Industry

Size: px
Start display at page:

Download "Measuring Market Power in the UK Retail Salmon Industry"

Transcription

1 LAND ECONOMY WORKING PAPER SERIES Number 12. Measurng Market Power n the UK Retal Salmon Industry Correspondng Author: Abdula Fofana Land Economy Research Group Research Dvson SAC Ednburgh EH9 3JG Tel: E-Mal: Abdula.Fofana@sac.ac.uk 1

2 Measurng Market Power n the UK Retal Salmon Industry ABDULAI FOFANA (correspondence), Land Economy Research Group, Research Dvson, Kngs Buldngs, SAC, Ednburgh, EHG 3JG Emal:Abdula.Fofana@.sac.ac.uk, Tel: , Fax: SHABBAR JAFFRY, Department of Economcs, Unversty of Portsmouth Rchmond Buldng, Portland Street, Portsmouth PO1 3DE,UK Emal: Shabbar.Jaffry@port.ac.uk Abstract Ths paper presents an nvestgaton nto the market structure for three product types of salmon (smoked, fresh and whole salmon) n the UK retal market. Evdence of the potental for market power and prcng conduct s analysed usng structural smultaneous system equatons based on the Bresnahan (1982) model. The mportance of the retal market s recognsed gven the domnance of supermarket chans whch accounted for 1.6 bllon sales of seafood and the share of about 87% of all seafood retal sales n 2004 as compared wth only 16% n The results ndcate that the system s well represented by the models and that the market s compettve for fresh fllets and whole salmon but retalers exert some market power for smoked salmon. The hypothess that market power s the same for all three products n the study was rejected; further ndcatng that retalers may be exercsng market power for smoked salmon. Keywords: Market power, Error correcton model, Dynamc demand systems, salmon JEL Classfcaton: JEL-1, JEL-J 2

3 Introducton Rapd growth of the UK retal sector n the last 20-years has fueled an enormous change n the organzaton of relatonshps along the seafood marketng structure. Changes n consumer buyng habts wth the ncreased demand for convenence food and the consumpton of more meals away from home are frequently cted n the lterature as the natural causes of these transformatons. These changes have shfted the market from beng unsophstcated nto beng hghly technologcal and concentrated. Ths concentraton has been brought about partly by the substantal growth n the producton of ntensvely farmed salmon n Scotland, and partly by the ncreased domnance of supermarkets n retal sales. Mergers among supermarket chans and take-overs by multnatonal retal frms have led to supermarket domnance n the retal trade. Supermarket sales of seafood at 1.6 bllon for 2004 was up 9.6% on 2003 fgures (Seafsh Industry Authorty, 2004), and account for about 87% of all seafood retal as compared wth only 16% n 1988 (Murray and Fofana, 2002). Growth n supermarket sales has occurred at the expense of market stalls, fshmongers, retal vans and other ndependent retalers as depcted n Fgure 1. Consequently, the supermarket s now the outlet of choce for most consumers, rather than the ndependent small retaler (e.g. fshmonger). [Insert Fgure 1 here] The motvatons of ths paper are closely lnked to the developments that have occurred n the fsh retalng ndustry n the last two decades and the domnance of supermarkets n retalng of seafood products n the UK. These developments have rased concerns that supermarkets may be exercsng market power. Ths allegaton has been expressed by consumer groups and newspapers such as Fsh Farmng Today, No. 180 November, (2003) and Sunday Tmes, 23 August (1998)). A number of recently publshed books 1 1 Shopped by Joanna Blythman and Not on the Label by Felcty Lawrence New Economcs Foundaton Clone Town Brtan 3

4 about multple retalers growth and domnance also document the growng publc concern of the sprallng power of the supermarket retal chans. Concerns over the domnance and potental market power was one of the factors that led the Offce of Far Tradng (OFT) n the UK prompted the Competton Commsson (CC) to nvestgate UK supermarket sector, the report of whch was released n The CC s analyss covered both the supermarkets relatonshps wth supplers and the extent to whch the downstream market was compettve. The nvestgaton ncluded an analyss of prce trends over tme, nternatonal prce comparsons, proftablty, and consumer satsfacton. The CC concluded that there was no evdence of excess prces or profts, and that the market was broadly compettve. In early 2006, the OFT agan referred the retal ndustry to the CC for more detaled nvestgaton due to further consoldaton n the retal market snce The OFT s partcularly concerned that ncreased consoldaton and the move by supermarkets nto the convenence sector could reasonably be suspected to dstort competton and harm consumers. The queston of competton and market power come together n consderatons of domnant frms and n the chan they operate. Hgh market concentraton and shares are commonly assocated wth the exercse of market power (Burt and Sparks (2003)). Therefore a concept central to the ndustral organzaton lterature s that ncreased concentraton leads to ncreased market power, whle less concentraton ncreases the dscplnes on ndustry s prcng strategy. Dsadvantagng compettors by offerng lower prces to supplers provdes a beneft that allows domnant frms to ncrease market share or prces. In the lterature a number of possble ways of rasng costs can be dentfed, ncludng supply agreements, exclusve dealng, wage deals, complance costs, advertsng, development servce enhancements and vertcal ntegraton. All or some of these tactcs have been ascrbed to the operaton of multple retalers and supermarkets (Smth (2002), Clarke et al. (2002)). 4

5 In a recent artcle, Jaffry, Fofana and Murray (2003) carred out a study of the retal market for fresh salmon fllets n the UK usng data for the ndustry from Ther work was motvated by the recent questons about clams that supermarket chans use ther buyng power to obtan substantal dscounts from supplers but fal to pass these benefts on to consumers. Usng sngle demand and supply equatons, Jaffry et al analysed the chan between the retalers and consumers for the fresh salmon fllets usng the New Emprcal Industral Organzaton (NEIO) methodology. The study revealed that the retal market for fresh salmon fllet s compettve despte the market beng hghly concentrated. However, the study does not mply that market power s not beng appled to other product forms such as smoked salmon fllets or whole salmon at retal level. Snce the late 1980s, applcaton of the NEIO to estmatng the extent of market power has become very popular. Recent developments n tme seres economcs have further enhanced emprcal nvestgaton of market power usng the NEIO methodology. In ths methodology, correct specfcaton of the demand process s crucal. Whle most applcatons of ths method have been restrcted to sngle markets, Cotterll, Dhar and Putss (2000), Hausman, Leonard and Zona (1994) and Hyde and Perloff (1998), among others have used system demand equatons to overcome the theoretcal shortcomngs of sngle demand equaton n determnng market power. Hyde and Perloff (1998) developed cross-markets nvestgaton of market power for meat usng a lnearsed form of AIDS model. In ths paper an economc model of frm conduct smlar to that of Hyde and Perloff (1998) s used to measure explctly the degree of compettve behavour n the retalng of salmon products n the UK. Ths approach yelds precse, nterpretable statstcal tests to evaluate the degree of market power n salmon retalng. Whle the approach used n ths paper s smlar to that of Hyde and Perloff, t s more robust n that the tme seres propertes of the data are accounted for n the constructon of the LAIDS model. The margnal cost equaton n ths artcle also dffers from that of Hyde and Perloff n that we 5

6 ncluded a varable that captures prce of captal for retalng and tests are conducted for the legtmacy of the nstruments used n the estmaton. The modellng approach also dffers from prevous salmon retal market power studes n the UK that uses sngle demand models (e.g. Jaffry et al. (2003)). 1. Model formulaton The key to measurng market power usng the NEIO methodology s the consstent estmaton of demand and cost models. Therefore, we frst specfy a system demand model estmated and then derve the optmalty equatons followng Hyde and Perloff (1998). 1.1 Demand Systems The frst stage n the estmaton of market power s to specfy and estmate a model that descrbes consumer demand. The key to any demand estmaton s to specfy a model that s both flexble and consstent wth economc theory. The lnear form of the AIDS has enjoyed great popularty n appled demand analyss due to ts consstency and ease of estmaton. Startng from a specfc cost functon wth the basc assumpton that commodtes are weakly separable from non-related goods, the AIDS model gves the share equatons n a 3 good system as where z t p q 3 j1 ln p j jt xt ln pt (1) t t zt s the budget share of the th good, p jt are prces of the j th good n the xt bundle, x t s total expendture on all goods n the system and p t s the ndex of prces n tme perod t. The ndex of prces p t s assumed to be a functon of commodty prces and s defned as a translog prce ndex of the form: 1 ln pt 0 k ln pk 2 j j k jk ln p The lnear AIDS model uses the 'corrected' Stone s (1953) prce ndex: j (2) 6

7 ln p s t z p ln p t t 0 where 0 p s the mean of the seres used as the base perod. To keep the model consstent wth economc theory, the parameters are constraned such that the homogenety, addngup and symmetry condtons hold. These restrctons are mposed as follows: j 1 j j (Homogenety) 3 1 j j j (Addng up) (Symmetry) Therefore we estmate own prce and expendture elastctes of the three product forms ncluded n the study. The uncompensated own prce and ncome elastctes of demand (Green and Alston, 1990) are calculated as follows (Own prce) ε γ 1 β z (Expendture) η 1 β z Calculatng elastctes, the budget shares are deally the predcted shares at the estmaton pont. However, Chalfant (1987) ndcated that the use of the correspondng sample share closely approxmates the predcted shares, and that these can be used n emprcal work. 7

8 1.2 Optmalty condton As hghlghted by Hyde and Perloff (1998), Bresnahan (1982) showed that a sutable way of generatng an entre famly of possble equlbra s to equate margnal cost wth a measure of effectve margnal revenue. Ths s mathematcally gven as follows MR p λ p q q (3) where q s quantty, p s prce and the term s a parameter to be determned and captures the degree of market power n the ndustry for all goods = 1,..., n. Theoretcally, term takes a value between 0 and 1 ( 0 λ 1) and f equals zero perfect competton exsts. If s equal to 1 the ndustry behaves as a cartel. Intermedate values dentfy dfferent olgopoly regmes. In general, the optmalty condton n (3) can be rewrtten for good as p p q q MC q (4) where MC q s a formulaton for the margnal cost functon whch s the frst dervatve of the total cost functon. The procedure summarsed n equaton (4) has been appled n many sngle market studes (see for example Jaffry et al (2003) and Steen and Salvanes (1998) for specfc examples on the applcaton to salmon markets. Hyde and Perloff (1998) generalsed equaton (4) to study several markets smultaneously. Followng the same ven, we generalsed equaton (4) to study several product markets smultaneously. Some authors have nterpreted as conjectural varaton, n ths paper we have followed Perloff (1992) as the wedge between prce and margnal cost created by an unknown game. To extend Hyde and Perloff (1998) analyss, we ncluded a varable, k, whch s a proxy ndex that captures the prce assocated captal and other utltes faced by all retal outlets 8

9 as part of the determnant of margnal costs. 2 The k varable n the margnal cost equaton emphasses the asymmetry n costs that can exst between the ncumbent large retal frms and potental entrants n the market. Supermarkets have used ths advantage to cut prces to lure customers and by so dong movng away from short run to long run proft maxmzaton objectves. These strateges are desgned to ncrease market share, protect market poston n the long run and act as a barrer strategcally to deter potental entrants n the ndustry. The evaluaton of compettve markets and market behavour often focuses on the extent to whch one or more frms can sustan a prce ncrease. Frms wll fnd t very dffcult to sustan sgnfcant long-term prce ncreases f t s easy for new frms to enter a market and provde a substtute product or servce. The exstence of barrers to market entry wll lmt such market responses. There are many types of barrer to entry n dfferent markets. Ban (1956) dentfed absolute cost advantage 3 and economes of large scale producton that requre large captal expendture as among the general sources of barrer to entry n an ndustry. Ban argued that a barrer to entry s a structural attrbute of a market mplyng that ncumbent sellers can earn more than a normal rate of return wthout attractng new entrants nto the ndustry. For example, the cost of land s a sgnfcant part of costs n the UK. The Competton Commsson (2000) reported that the fve man supermarkets n the UK pad more to acqure land for development than ther counterparts n other European countres. Along these lnes the Competton Commsson (2000) examned development costs of land n France, Germany, and the Netherlands. The analyss suggests that development costs n the UK were typcally 2.5 tmes hgher than n France; 1.6 tmes hgher than n Germany and 2.7 tmes hgher than Netherlands. Therefore k whch s an ndex of the prce of captal and other utltes s a key varable n the cost structure of large retal organzatons n the UK and t s vtal to uncover any antcompettve behavour. 2 Steen and Salvanes (1998), Park and Welwta (1999), Jaffrey, Fofana and Murray (2003) ncluded measures of captal cost n the margnal cost equatons n determnng olgopoly power. 3 Absolute cost advantage s used to descrbe a number of potental cost advantages that ncumbents may have over new entrants, other than economes of scale. 9

10 Followng Hyde and Perloff (1998), constant returns to scale s assumed so as to allow the possblty of detectng other forms of market structures. In ths ven, margnal cost for each good reflects constant returns to scale and s lnear n wholesale prces, wages and prce of captal as shown n the followng equaton. MC q a b v d w e k (5) where v s the wholesale prce of each fsh product; w s an ndex of retal wage costs common for all salmon products and k s a measure of captal cost common to all retal outlets n the ndustry. a, b, d and e are parameters to be estmated. Smlar margnal cost representatons have also been used by Deodhar and Sheldon (1995) and Hatrl et al (2003). Takng frst dfferental of the LAIDS model holdng total expendture (x) and other prces p j, p q j constant, the slope of each of the demand curve can be gven as follows j p q j 1 j z j j (6) z z where j refers to the Kronecker delta. In the salmon retal market retalers stock most of the product forms for salmon. In the case where each frm sells all three goods, as we assume that retal frms sell all three products, (4) can be generalsed to 3 p q j j 1 p j q MC q Substtutng (5) and (6) nto the optmalty condton (7), wth some mathematcal manpulatons yelds the followng equaton used n the estmaton. (7) p a b v λ z d w ek q j j p j q z j j λ 1 1 z 1 (8) 10

11 where p s the retal prce of salmon product, v s the wholesale prce, w s an ndex of the retal wage rates, q s the quantty sold, z s the budget share, and are coeffcents from the LAIDS model and a, b, c, d, e and are parameters to be estmated. As evdent from equaton (8), an ntegral part of the optmalty condton whch depcts the ndustry s prcng behavour s the elastcty concept whch s the conduct coeffcent ( ). The parameter measures a gap created by some unknown game as could be the case n a folk-theorem equlbrum that les between the collusve and Cournot equlbra (Hyde and Perloff (1998)).Thus the model nvolves a knd of pseudo-dynamcs depcted n a framework that s essentally statc. 4 More recently, the appearance of studes n the lterature have tested the valdty of statc olgopoly models (e.g., Nevo (2001); Hausman and Leonard (2000); Genesove and Mulln (1998); Wolfram (1999)). Most of these studes suggest that the seemngly statc olgopoly models yeld reasonably accurate predctons of prcng behavour. The conduct coeffcent wll reveal what knd of olgopolstc behavour that characterzes the market, and there s no need to mpose any a pror restrcton on t. That s, t s not necessary to assume a certan conduct beforehand, and test for ts proprety. Instead, any behavoural model s a pror plausble. 2.0 Data and Emprcal Modellng Salmon retal data for smoked, fllets and whole salmon were collected from Seafsh Industry Authorty and the producton and wholesale data were obtaned from Scottsh Qualty Salmon (SQS). 5 Quantty-value transformaton was used to derve unt prces per klogram. Asche et al (2001) argues the sgnfcant prce reductons of new aquaculture speces ndcates that the markets for these speces are not strongly lnked to the markets for other products. As perfect substtutes have a constant relatve prce and close substtutes have hghly correlated prces, ths suggests that farmed fsh does not compete too 4 Although derved n a statc context, these tests provde a vald emprcal test of dynamc equlbrum (Worthngton (1990)). Consequently we do not ntroduce any dynamcs n the optmalty model durng estmaton. 11

12 closely wth other goods. Therefore data for substtute products were excluded from the analyss. Wage (w) n the margnal cost equaton s an ndex of wage costs of retalers, based on a contnuous surveys conducted by the UK Offce for Natonal Statstcs (ONS) n whch over 12, 500 samples are selected annually. Prce captal varable (k) n the margnal cost equaton s an ndex of rent, whch ncludes mantenance, nsurance, ground rent, councl tax, water rates, mortgage, buldng mantenance and an nterest payment ndex; collected from the Offce for Natonal Statstcs (ONS), UK. The data avalable for emprcal modellng are monthly fgures from January 1992 to December Table 1 summarses and presents descrptve statstcs of the varables used n the estmaton procedure. [Insert Table 1 Here] Although monthly data were avalable we follow Genesove and Mulln (1998) n aggregatng up to quarterly level. Genesove and Mulln argued that ths was to ensure that the estmated elastcty represents the long run elastcty as opposed to short run. In addton, the long run s consdered because under mperfect competton, retalers are more lkely to establsh a prce n the long-run rather than short-run profts n mnd and to mantan that prce for a consderable perod of tme (Juma, (2004)). Before specfyng the most approprate emprcal model the tme seres propertes of the data were revewed n order to nvestgate formally whether the long run demand relatonshps are economcally meanngful or merely spurous. 5 SQS s an ndustry body that represents Scottsh Salmon Growers Assocaton, the Scottsh Salmon Board and Scottsh Qualty Salmon. A new expanded trade assocaton, Scottsh Salomon Producer Organsaton, was created n 2006 that has taken over the operatons of SQS. 12

13 Frst, each of the ndvdual tme seres s tested for a unt root n the demand equatons. Unt root testng was mplemented usng the Augmented Dckey-Fuller (ADF) test. The tests are conducted wth and wthout a trend term and the results ndcated that the constructed seres for fresh fllets, whole and smoked fllets were all non-statonary n levels but statonary n frst dfferences, a necessary condton for contegraton. The second step was to nvestgate whether both z and the vector of explanatory varables n each demand equaton are contegrated. Ths test was mplemented usng Johanson s (1988) approach, a vector auto-regressve model. Ths test s mportant because the demand model specfed would be napproprate f the varables do not have a long run relatonshp or are not contegrated. The results show that there s at least one contegraton vector n the share equaton, z for each salmon product forms. The results are presented n table 2 below. [Insert Table 2 Here] Havng found that all varables n the LAIDS model are I(1) and there s at least one contegratng vector n each of the share equaton mples that the LAIDS model must be estmated n a way that accounts for these propertes. The estmaton of the LAIDS model usng ntegrated data has been addressed usng a number of methods. For example, Ng (1995) specfcally consders the ssue of testng the homogenety restrcton and uses a method n whch the emprcal dstrbuton of the relevant test statstcs are smulated by parametersng the data generatng process and usng ths as the bass for a Monte Carlo exercse. The most commonly used method n fsh demand analyss s the Anderson and Blundell (1982) technque as exemplfed by Asche et al. (1997). Therefore followng Asche et al. we specfy our dynamc LAIDS error correcton model (LAIDS-ECM) as follows: t1 k1 n Δ4 z α δδ4 z α s α t lnδ4 p β lnδ4 Εη (10) t t 3 k k t j1 j j t 13

14 whereδ 4 denotes fourth dfferenced flter (e.g. Δ 4 p p t pt4 ), x t E, p t p j... Δ 4 zt. s k are seasonal dummes, t a lnear trend, t s the error term. The use of a trend s qute common n demand analyss, for example see Burton and Young, (1996), Asche et al. (1997). The trend term n the equaton s assumed to capture several factors. Frst, marketng of salmon especally by SQS n the meda to promote salmon as a healthy det to counteract negatve publcty about salmon farmng has ntensfed snce the creaton of the ndustry body. The actvtes of the ndustry body and smlar organzatons that promote oly fsh have drawn attenton to salmon products. Second, the trend also captures structural shfts n demand and changes n taste. Lke Asche et al., we are not able to separate the trend and the constant terms n equaton (10) thus estmated are not nterpreted as mnmum expendture on each salmon product. The flter n equaton (10) has been argued to control for both determnstc and stochastc seasonal trends n the model (Asche et al., 1997). The fourth dfference flter s preferable n wth quarterly data to the normal frst dfference flter because t controls both for seasonal cycles and for unt roots such that the short-run dynamcs n the data seres are statonary. Followng Anderson and Blundell (1983) and Burton and Young (1996), the propertes of homogenety and symmetry are appled n the long-run segment of the model. Whle symmetry may not hold n the short run, homogenety holds by the vrtue of the addng up propertes. One lag was suffcent to account for the dynamcs and the effects of habt formaton n the model. 3.0 Emprcal Results To accommodate the large number of parameters n equatons (8) and (10) and the attendant multcollnearty problems the demand system equatons and the system of equlbrum condtons were estmated sequentally. 6 The demand system (8) was estmated us- 6 In all, we attempted to estmate fve equatons system wth demand theoretcal restrctons mposed usng non lnear three- stage least square to acheve maxmum effcency. Unfortunately we were not able to 14

15 ng a seemngly unrelated regresson (SUR) technque, whch accounts for cross-equaton contemporaneous correlaton and as a result takes nto account the optmzaton process that s essental n any demand system. The demand system s estmated by ncorporatng the theoretcal restrctons of addng-up, symmetry and homogenety. Followng Anderson and Blundell (1982), one equaton (we dropped the whole salmon share equaton) was dropped to avod sngularty durng estmaton and the addng-up restrcton was used to recover the equaton after estmaton. The estmated parameters are reported n Table 3[Insert Table 3 Here] The model specfcaton s robust and fts the observed data for both product forms as measured by system R-square 1 of 87%. The hypotheses of homogenety and symmetry condtons mpled n economc theory are tested wth a Wald test. The computed test statstcs for symmetry restrcton s 10.4 and the tabulated 2 statstcs s 3.84 for one degree of freedom, whch allowed the rejecton of the hypothess at the 5% level of sgnfcance. On the other hand, the homogenety restrcton was accepted at the 5% level of sgnfcance. The correspondng computed value of the test statstcs for homogenety s 4.60 and the tabulated 2 statstcs for 2 degrees of freedom s The rejecton of the symmetry condton found n ths study s not unusual. Ths mples that emprcal demand studes often found that some theoretcal restrctons do not hold (Deaton and Muellbauer 1980b, Cozzarn and Glmour 1998). Cozzarn and Glmour conducted a survey on these restrctons n emprcal demand studes and found homogenety (symmetry) was tested n 29 of the models and the restrcton was rejected 57% percent of the tme. A jont test for homogenety and symmetry were also conducted. The computed test statsacheve convergence n that case may be due to large number of parameters to be estmated. However a number of market power studes have used sequental estmaton method (e.g. see Steen and Salvanes, (1999); Gohn and Guyomard, (2000); Jaffry et al, (2003). 15

16 tcs s 6.12 and the correspondng crtcal value at 5% level of sgnfcance s 7.82, whch means that we can t reject the null of a jont homogenety and symmetry restrcton. We use the Lagrange Multpler (LM) to test for the presence of autocorrelaton. In the 0 LM test the null hypothess of H : 0 (no autocorrelaton) s tested aganst the alternatve hypothess of the presence of autocorrelaton. The LM test has a 2 dstrbuton wth degrees of freedom equvalent to the number of lagged resduals. Usng the aforementoned test the computed 2 statstcs are 0.13 and 0.22 and the tabulated crtcal s 5.99 for 2 degrees of freedom. The test suggests that there s no problem of autocorrelaton detected n both estmated share equatons. Followng from the results obtaned, the results ndcate that the system model estmated for retal demand for salmon possess both theoretcal propertes of homogenety and symmetry. The economc content of the LAIDS model s presented n Table 4 n terms of elastctes. The own prce elastctes have correct sgns (negatve) accordng to the economc theory. Estmates of elastcty n the retal market for dfferent salmon products n the UK are not commonly reported n the lterature except for fresh fllets. Compared wth earler studes of salmon market n the UK, the results of ths study appear n lne wth fresh fllets estmates reported n the past. For example Clay and Fofana (1999) and Jaffry et al. (2003) reported own prce elastcty estmate values of 0.33 and 0.62 for fresh salmon fllets. It s nterestng to note that our calculatons ndcate that whole salmon s relatvely more elastc compared wth fresh fllets or smoked salmon. Ths s typcal of unprocessed food, meanng that a slght ncrease n the prce of whole salmon trggers hgher than proportonate decrease n demand. Unlke smoked or fresh fllet whch are to some degree processed and hgher value goods and therefore a slght ncrease n prce would not change demand a great deal. Also worth notng s that the elastcty estmate for fresh salmon fllets appears to be ncreasng n value wth tme. Ths trend ndcates the ndependence of consumers and that the market demand curve s gradually dscplnng the 16

17 market. That s customers react more strongly n later years to changes n prce as compared wth earler years; ths s a sgn of compettveness n the ndustry. [Insert Table 4 Here] 17

18 The own prce elastctes for smoked and fresh fllets of 0.88 and 1.02 are not sgnfcantly dfferent from (mnus) one, mplyng that consumers almost make proportonal responses n the quantty demanded when the prces of these products change. However, the own prce elastcty for whole salmon (-1.49) depcts elastc demand. All other thngs remanng unchanged, f the prce of whole salmon n the retal market ncreases by 10% consumers cut back on the quantty demanded by almost 15%. Ths adjustment by consumers for whole salmon s not surprsng for a product that s unprocessed and mostly dsplayed on wet fsh counters n supermarkets, thus competng drectly wth other unprocessed fsh products. The expendture elastctes are statstcally sgnfcant ndcatng that total expendture on salmon products s a sgnfcant determnant of the demand for all the product forms of salmon that were nvestgated. However, we note the elastctes estmates are not sgnfcantly dfferent from one for smoked fllets and whole salmon. The expendture elastcty for fresh fllets s, however, greater than one ndcatng that demand s expendture elastc. Earler studes of salmon n the 1980s to md 1990s typcally estmated hgher expendture elastcty values as compared those found n ths study. Our estmates are dssmlar from earler studes due to recent changes n the market and the demand structure for salmon. These changes are fuelled by marked shfts n food consumpton patterns due to changng lfestyle of the UK consumer, along wth economc, socal and demographc changes. Furthermore, the demand system specfcaton and the estmaton technques used n ths artcle are more robust n contrast to many earler studes when the market was mmature. Thus, the results n ths study are smlar for ncome to comparable elastctes of Asche et al. (1997), Clay and Fofana (1999) and Jaffry et al. (2003). On the whole, the elastcty estmates are plausble and meanngful and further lend support to the satsfactory nature of the LAIDS-ECM model estmated. Due to the non-lnearty of the system of equlbrum condtons (10), the non-lnear three-stage least squares (NL3SLS) procedure wth nstruments s used for estmaton. Intal startng value estmates were obtaned n a two-stage estmaton procedure usng 18

19 an nstrumental varable method to avod smultanety bas. The nstruments used were GDP, tme trend, UK money supply (M3), quarterly average of UK Banks' base nterest rates, UK CPI, ndex of producton by Standard Industry Classfcaton (92) and an ndex of expendture on food and non-alcoholc beverages. 7 Our estmates of the margnal cost equaton for each product form and the correspondng market power estmates are presented n Table 5. A key problem wth nstrumental varable s that nstruments are requred whch are at least asymptotcally uncorrelated wth the error term n the regresson equaton. In order to test the valdty of the nstruments we used a Sargan (1964) valdty test of nstruments. Under the null hypothess, the Sargan statstc s asymptotcally dstrbuted as wth p-k degrees of freedom; and s wrtten down as: S Δεˆ W n 1 W Δεˆ Δεˆ W Wεˆ (11) 1 where W s the chosen matrx of nstruments, p ndcates the number of columns n W and k the number of parameters to be estmated. The test statstcs were 1.98, 1.95 and 8.14 for smoked fllet, fresh fllet and whole salmon margnal cost equatons respectvely. From the results, we accept the valdty of nstruments for all the margnal cost equatons when compared wth the crtcal value of 9.49 for 4 degrees of freedom and 5% level of sgnfcance. 2 [Insert Table 5 Here] 7 GDP, CPI, ndex of productvty, and Index of expendture on food and non-alcoholc beverages were collected from the Offce for Natonal Statstcs. M3 and base nterest rate were obtaned from the Bank of England. 19

20 The model fts the data well wth satsfactory R 2 values. An autocorrelaton test (DW test) was performed to check for the exstence of autocorrelaton n each of the margnal cost equaton. The results also suggest that no autocorrelaton n these equatons. Consderng the effects of the retal wage rate on retal margnal cost, these were found to be of mnmal mpact wth those for salmon steak and whole salmon not beng statstcally sgnfcant. Ths effect was as expected n that the retal wage rate s generally close to the mnmum wage. More mportantly however s the fact that employees n the retal sector have lttle nfluence on ther wage rates and are therefore prce-takers. Wth respect to the market power parameters,, these were found not to be sgnfcant at the 5% sgnfcance level for fresh fllet and whole salmon but sgnfcant for smoked salmon. More mportantly however, s that the market power estmates for fresh fllet and whole salmon are close to zero for us to conclude that the market for these products s near perfectly compettve. However, n relatve terms the market power estmate for smoked salmon s sgnfcant and of hgher magntude as compared wth other two products. Ths suggests that retalers exercse some degree of market power for smoked fllet. Furthermore, the correspondng Lerner ndex 8 for smoked salmon s whle the estmated ndxes for fllets and whole salmon are and respectvely. It can be seen that the Lerner ndexes for fresh fllet and whole salmon were neglgble and suffcently close to zero to ndcate that they are compettve whle the value for smoked salmon was dfferent from zero suggestng some relatve degree of market power. We also tested whether the retal margnal cost functon and the market power parameter for each salmon product are dentcal. The Wald test statstcs for dentcal margnal cost functon s and the correspondng tabulated value s for 6 degrees of freedom and 5% level of sgnfcance. The test clearly rejects the hypothess that the margnal cost 8 L p MC λ ; where MC s the margnal cost for salmon product. p ε 20

21 functons of each salmon product are equal. Ths further shows that there are physcal as well as cost dfferences between the dfferent products of salmon that affects retal margnal costs. The cost dfferences are reflected n the degree to whch each product form s processed,.e. smoked fllet s hghly processed whle whole salmon s the least processed. Ths leads to the exstence of dfferent varable cost structures assocated wth dfferent market structures, hgher costs beng assocated wth non-competton. The Wald test statstcs for equal market power measure for each product form s 8.44 and the correspondng tabulated value s 7.82 for 3 degrees of freedom at the 5% level of sgnfcance. These results further show that the market power estmates for each salmon product are not dentcal, lendng support to the noton that retalers mght be exercsng some degree of market power for smoked fllet. Indeed there are reasons n the ndustry to suggest that retalers of smoked salmon fllet mght have some degree of market power. Salmon or ndeed fsh processng n the UK ndustry s a mature and conservatve one based on tradtonal practces wth some methods of processng stll unaltered by recent technologcal nnovatons. A survey conducted by the UK Sea Fsh Industry Authorty n 2000 revealed that close to 40% of prmary processors ether have no nternet web ste or do not use computers even n basc tasks n ther busnesses, such as book-keepng or nventory purposes. Moreover, there has been more consoldaton, lqudatons and ratonalzaton of salmon smokng frms or actvtes wthn frms than any other frms specalsng n other forms of processng of salmon. 4.0 Conclusons Salmon products have now become affordable to the ordnary consumer despte the hghly concentrated channels by whch the supples are obtaned. In order to gan an understandng of the transformaton of the retal sector for salmon products, we appled a new approach n Industral Economcs to the salmon retal market. The methodology, unlke ts predecessors, s theoretcally more robust. Modelng the formaton of prces n the market, we estmated an ECM LAIDS model that s clamed to approxmate con- 21

22 sumer behavour. Ths s the frst study to use such a model wthn a structural econometrc framework of frms to test for market power for salmon products n the UK. Applyng these models to salmon products at the retal level yelds robust demand and margnal cost models wth theoretcally acceptable model dagnostcs, suggestng that the system s well represented. The emprcal results were derved wth data for and the results show compettve prcng behavour n general but that the retaler mght have some degree of market power for smoked salmon fllet. Largely the fndngs from ths study support prevous fndngs (Jaffry et al (2003); Steen and Salvanes, (1999)). In addton, our fndngs support the clams made by supermarkets n the UK, that, regardless of ther huge market share and hgh concentraton, retal sales for salmon reman compettve. 22

23 References Anderson, G.J., and Blundell R.W., Estmaton and hypothess testng n dynamc sngular equaton systems. Econometrca, vol 50 pp , Anderson, G.J., and Blundell R.W.,. Testng restrctons n a flexble dynamc demand system: an applcaton to consumers expendture n Canada, Revew of Economc Studes vol.50 pp , Asche, F., and Wessells, C. R.,. On Prce Indces n the Almost Ideal Demand System. Amercan Journal of Agrcultural Economcs vol. 79 pp , 1997 Ashe, F., K. G. Salvanes and Steen, F.,. Market Delneaton and Demand Structure, Amercan Journal of Agrcultural Economcs vol.79 pp , Breshnahan, T. F., The olgopoly soluton concept s dentfed, Economcs Letters vol.10 pp 87 92, Burt, S., and Sparks, L. Power and Competton n the UK Retal Grocery Market, Brtsh Journal of Management vol. 14 pp , Burton, M., and Young T., The structure of changng taste for meat and fsh n Great Brtan, European Revew of Agrcultural Economcs vol.19 pp , Burton, M., and Young, T., The Impact of BSE on the demand for beef and other meats n Great Brtan, Appled Economcs vol. 28pp , 1996 Chalfant, J.A. A Globally Flexble, Almost Ideal Demand System, Journal of Busness and Economc Statstcs vol. 5pp , Clay P., and Fofana, A., Prce and ncome elastctes for the demand for farmed salmon, Unpublshed, Report for the Scottsh Executve, Clarke, R., Daves S., Dobson, P. and Waterson, M., Buyer Power and Competton n European Food Retalng, Cheltenham, UK, and Northampton, MA, USA: Edward Elgar, Competton Commsson Supermarkets: A report on the supply of groceres from multple stores n the Unted Kngdom. Cotterll, R. W., W. P. Putss Jr. and Dhar R., Assessng the Compettve Interacton between Prvate Labels and Natonal Brands. Journal of Busness vol.73 pp ,

24 Cozzarn, B., and Glmour, B., A methodologcal evaluaton of emprcal demand systems research, Canadan Journal Agrcultural Economcs 46: , 1998 Deaton, A., and Muellbauer, J. An Almost Ideal Demand System, Amercan Economc Revew, vol. 70 pp , Deaodhar, S.Y. and Sheldon, I.M., Is foregn Trade (Im) Perfectly Compettve? Journal of Agrcultural Economcs, vol 46 pp , Fsh Farmng Today. November, No Genesove, D., and Mulln W.P., Testng statc olgopoly models: Conduct and cost n sugar Industry Rand Journal of Economcs vol. 29pp , Gohn, A. and Guyomard, H. Measurng Market Power for Food retal Actvtes: French Evdence, Journal of Agrcultural Economcs vol. 51 pp , Green, R., and J. M. Alston. Elastctes n AIDS Models, Amercan Journal of Agrcultural Economcs vol. 72pp , Hausman, J., G., Leonard and J. D. Zona. Compettve Analyss wth Dfferentated Products, Annales D Econome et de Statstque vol 34 pp , Hausman, J. and Leonard, G. The Compettve Effects of a New Product Introducton: A Case Study, Mmeo, MIT, Hatrl, A.S., Jones, E. and Aktas, A.R., Measurng the Market Power of the Banana Import Market n Turkey, Turksh Journal of Agrculture and Forestry, vol. 27 pp Hyde, C., and Jeffrey M. Perloff.. Multmarket market power estmaton: the Australan retal meat sector, Appled Economcs, vol 30 pp , Jaffry, S., A. Fofana, and A. D. Murray. Testng for market power n the UK retal sector, Aquaculture Economcs and Management, vol 7pp Jumah, A., The long run, market power and retal prcng, Emprcal Economcs, vol. 29 pp Murray, A. and Fofana A., The changng nature of UK fsh retalng, Marne Resource Economcs, vol 17 pp

25 Nevo, A., Measurng Market Power n the Ready-to-Eat Cereal Industry, Econometrca, vol. 69 pp , Ng, S., Testng for homogenety n demand systems when the regressors are nonstatonary, Amercan Journal of Agrcultural Economcs, vol 10 pp , Park, T. and Welwta, A., Compettve Behavour n the US Food Retalng Industry, Canadan Journal of Agrcultural Economcs, vol 47 pp45-55, Perloff, J.M., "Econometrc Analyss of Imperfect Competton and Implcatons for Trade Research," n Ian M. Sheldon and Denns R. Henderson, eds., Industral Organzaton and Internatonal Trade: Methodologcal Foundatons for Internatonal Food and Agrcultural Market Research, NC-194 Research Monograph Number 1, July Sargan, J.D, Wages and prces n the Unted Kngdom: A Study n econometrc methodology, London; Butterworth. Reprnted n D.F. Hendry and K.F. Walls, Econometrcs and Quanttatve Economcs. Oxford:Basl Blackwell, Seafsh Industry Authorty, UK Sea Fsh Processng Industry key features. Sea Fsh Industry Authorty Press, Seafsh Industry Authorty. Retal Market Overvew. Sea Fsh Industry Authorty Press, Smth, H, Supermarket Choce and Supermarket Competton n Market Equlbrum, Revew of Economc Studes, vol 71pp , Steen, F and Salvanes G. K., Testng for market power usng a dynamc olgopoly model, Internatonal Journal of Industral Organzaton, vol 17 pp , Stone, J. Lnear Expendture Systems and Demand Analyss: An Applcaton to the Pattern of Brtsh Demand, Economc Journal, vol 64: pp , Sunday Tmes. 23 rd August, 1998 Wolfram, C., Measurng Duopoly Power n the Brtsh Electrcty Spot Market, Amercan Economc Revew, vol. 89 pp ,

26 Worthngton, P., Strategc Investment and Conjectural Varatons, Internatonal Journal of Industral Organzaton, vol 8 pp ,

27 Tables Table 1: Descrptve Statstcs Varable descrpton Coeffcent Mean Standard devaton Demand equaton: Budget Share Smoke salmon Salmon fllet z 1 t z Whole salmon z 3 t Prce Smoke salmon Salmon fllet Whole salmon Expendture Margnal cost equaton: Wholesale prce of smoked salmon v Wholesale prce of salmon fllet v Wholesale prce of whole salmon v Prce of labour (wage ndex) w Captal cost (ndex) e t Table 2: Statonarty and Contegraton Tests for Salmon Retal Products Varable Constant ncluded Level Unt root test Frst dfference Constant + trend Included Constant ncluded Constant + trend ncluded Contegraton test Max Trace Z smoked fllet a a 71.8 a a Z fresh fllet a a 65.0 a a Z whole a a 38.1 b a LnP smoked a a lnp fresh fllet a a LnP whole a a Ln(Exp) a a a sgnfcant at 1% level, b sgnfcant at 5%. 27

28 Table 3: Estmated Parameters of an LAIDS-ECM for Retal Salmon Products Share Equaton Smoked Fllets Fresh Fllets Varable Parame- t rato Parame- t -rato α (constant) z b b Smoked Fllet t c b Fresh Fllet t c c Whole t b Expendture t Smoked Fllet t b Fresh Fllet t c Whole t b b Expendture t c b 2.18 Smoked Fllet t-5 (z t Fresh Fllet t-5 (z t-5 ) c Whole t-5 (z t-5 ) c Smoked Fllet t Fresh Fllet t b 3.01 Whole t b b Expendture t c 1.75 S S S b 2.12 System R R b sgnfcant at 5% level; c sgnfcant at 10% level Table 4: Uncompensated Elastcty Estmates for Retal Salmon Products Estmate Coeffcent t-rato Own prce Smoked Fllet a Fresh Fllet a Whole a Expendture Smoked Fllet 0.98 a 3.75 Fresh Fllet 1.27 a 8.21 Whole 0.93 a 5.44 a sgnfcant at 1% level 28

29 Table 5: Parameter Estmates for Margnal Costs Equatons Con- V W K R 2 DW Smoked Fl a a b (0.67) (1.51) (1.12) (1.30) (-2.18) Fresh Fllet b b b (-0.11) (3.59) (2.26) (2.82) (0.69) Whole a b b (1.48) (1.25) (2.32) (2.32) (-0.92) V=wholesale prces; W=ndex of retal wage; =market power measure, K= ndex of captal cost, DW=Durbn-Watson test statstcs; student t-test statstcs n parenthess. b sgnfcant at 5%, c sgnfcant at 10% level Fgure 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Supermarkets Market stalls Mongers Other outlets Fgure 1: Value of fsh sales by outlet 29