Price Pooling and the Gains from Hedging: Application to a Swedish Grain Cooperative. AAEA Selected Paper Chicago, Illinois August 5-8, 2001

Transcription

1 Price Pooling and he Gains from Hedging: Applicaion o a Swedish Grain Cooperaive AAEA Seleced Paper Chicago, Illinois Augus 5-8, 00 Demcey Johnson, Tomas Nilsson, and Hans Andersson * Subjec code: 3 Absrac Opimal hedging sraegies are analyzed for a cooperaive operaing a price pooling sysem in he presence of price and quaniy risk. A hree-period model, accouning for defaul risk and sorage, is developed. Hedging allows he cooperaive o increase he pool price offered o he farmers by.8-4% for moderae risk parameers. * Johnson is Chief of Field Crops Branch, Marke and Trade Economics Division, USDA/ERS; Nilsson is graduae suden, and Andersson is Professor in he Deparmen of Economics, Swedish Universiy of Agriculural Sciences, Uppsala.

2 . Inroducion ODAL is a farmer-owned cooperaive, represening abou 30,000 farmers in cenral Sweden. As a resul of recen mergers, ODAL now markes abou 70 percen of Swedish whea. ODAL operaes a price pooling sysem on behalf of is members: all farmers who commi whea o he pool earn he same price, based on he prospecive average of sale prices. Price pooling is a feaure of oher grain-markeing sysems, noably Canada s. However, he Swedish case differs in wo imporan aspecs. Firs, paricipaion is volunary, as Swedish whea-farmers have access o oher marke oules. Second, ODAL announces is pool price before he bulk of is whea has been sold. This is in conras o he pracice of he Canadian Whea Board, which fixes is final pool price up o 8 monhs afer he sar of a markeing year. Alhough ODAL has no ye incorporaed whea fuures in is rading sraegy, i is beginning o examine he poenial gains from hedging in offshore fuures markes (i.e., CBOT, LIFFE and MATIF). Fuure liberalizaion of he CAP, induced by expanding EU membership and budge consrains, is likely o bring abou higher levels of volailiy in European grain markes (Brassley, 997). In his environmen, gains from hedging are likely o become more pronounced. This paper analyzes opimal hedging sraegies in he conex of a price pooling sysem. A concepual model is used o derive opimal markeing sraegies, wih and wihou hedging. Using empirical price daa, he model is used o quanify he poenial impac of hedging on he pool price offered by ODAL o Swedish farmers. The analysis is based on a hree-period opimizaion problem. The cooperaive can marke cash whea in each period. In he firs period, corresponding o pre-harves, he quaniy of whea handled by he pool is unknown; in he second period his uncerainy is resolved and he cooperaive announces is pool price. Hedge posiions can be esablished in eiher of he firs wo ODAL was founded in 996 by a merger of hree small-sized cooperaives in he middle par of Sweden. In January, 00, ODAL merged wih seven oher farmer-owned cooperaives ino he Swedish Farmers Cooperaive, (Svenska Lanmännen). Their core compeency remains he same, which is o supply parons wih producion inpus (seeds, ferilizers, feed, ec) and o marke grains and oilseeds. This sudy focuses on he grain inake marke area originally served by ODAL.

3 periods. In he hird period hedge posiions are closed and remaining grain invenories are liquidaed a prevailing cash prices. The cooperaive seeks o maximize he pool price offered o farmers subjec o a risk consrain. This limis he chance ha he cooperaive will defaul on is obligaion o farmers due o adverse price movemens. The plan of he paper is as follows. The nex secion provides some brief background on ODAL s price pooling sysem. The hird secion presens he concepual model of markeing and hedging decisions. Daa used in he analysis are described in he fourh secion. Model resuls are presened in he fifh secion. The paper concludes wih a shor discussion of implicaions.. Background on Price Pooling ODAL operaes hree differen pricing sysems for whea: a weekly spo price sysem, various grower conracs, and he pool sysem (Sinorn, 997). Spo prices are announced on a weekly basis and are paid for grain delivered immediaely. However, spo prices are usually lower han prices offered in he oher wo sysems. Grower conracs allow specific business arrangemens beween he cooperaive and is parons, wih prices arrived a hrough negoiaion. Conracing gives ODAL some laiude in is dealings wih large producers. The pool sysem has accouned for nearly wo hirds of he grain handled by ODAL in recen years (Table ). Wihin a markeing year, ODAL can offer several pools in succession. The firs pool is announced during he growing season and is closed a a predeermined dae afer harves. When he firs pool is closed, a second pool is opened; and when he second pool is closed (some monhs hence), a hird is opened. In pracice, mos grain handled by ODAL is commied in he firs pooling period, alhough farmers may have an incenive o defer sales if hey expec higher prices in laer periods. Table. Grain Handled by ODAL, Grain Toal quaniies Share of oal quaniy Share of oal Share of oal 3

4 Markeing Year handled by ODAL each year (in million meric ons) delivered during he firs pool period: Delivery from Aug. o Oc. in M meric on. quaniy delivered under he remaining pool periods 997,75 63% 7% 0% 998,40 64% 6% 0% 999,0 64% 6% 0% Source: Karlsson 999. quaniy delivered spo and hrough misc. conracs In he analysis ha follows, we focus on he operaion of he firs pool offered during a markeing year. I should be borne in mind ha ODAL does no know he quaniy ha will be markeed unil he pool is closed. The pool price is fixed when he pool is closed, and in advance of mos grain sales. If he proceeds from grain sales exceed he amoun guaraneed o farmers, he exra revenue is reurned o coop members (and no limied o paricipans in a paricular pool) in he form of paronage refunds. 3. Concepual Model To assess he poenial impac of hedging on ODAL s pooling sysem, we frame a hreeperiod opimizaion problem. The firs period is pre-harves, when quaniies commied o he pool are no ye known. In he second period, which is pos-harves, pool quaniies are known, and he price paid o farmers is fixed by ODAL. Markeing and hedging decisions are made in each of he firs wo periods. In he final period, ODAL liquidaes is remaining posiions in grain and fuures, and he profi or loss on pool operaions is deermined. ODAL s objecive is o maximize he expeced price paid o producers (SEK per on), subjec o a risk consrain. Le Z denoe he expeced price in period. In period, his is simply a planning price; in period, ODAL acually fixes he price o producers. In boh periods (,) ODAL solves: Max Z () Subjec o { R Z Q 0 } α Pr ob () 4

5 where R is markeing revenue for he pool (million SEK), Q is he quaniy of grain markeed by he pool (million ons), and α is he chance of he pool defauling on is obligaion o farmers. The risk consrain () has a deerminisic equivalen E (R) Z E (Q) K [V (R Z Q)] / α 0 (3) where E ( ) is he expecaion operaor condiional on period- informaion; V ( ) is he variance operaor condiional on period- informaion; and K α is he number of sandard deviaions associaed wih a specified probabiliy of defaul. Markeing revenue is defined R δ P X + δ P X + P X + δ H (F F ) + H (F F ) (4) where δis a compounding facor; P is he cash price (SEK/on) in period ; X is he quaniy sold (million ons) in period ; H is he hedge placed in period (million ons), wih H < 0 implying sale of fuures; and F is he fuures price (SEK/on) in period. Marke revenue (valued in period 3) includes he proceeds from cash grain sales as well as profis or losses on fuures ransacions. The difference beween marke revenue and he amoun promised o producers is he paronage refund. If he cooperaive pays oo high a pool price, pool members will have o refund money o he cooperaive. For he sake of simpliciy, we assume ha ODAL does no charge a handling fee. This is a varian of chance-consrained or sochasic programming, see Taha (976). Chance consrained programming is described in Taha in conex of operaions research (pp ). 5

6 Grain sales can occur in each of he hree periods, Q X + (5) + X X3 alhough he quaniy available for sale, Q, is no known unil period. There are hree sources of uncerainy in he model: cash prices, fuures prices, and he pool quaniy. By assumpion, cash prices evolve according o P + b 0 + bp e (6) where b 0 and b are coefficiens and e is a random disurbance. In line wih Kamara (98) and Myers and Hanson (996), fuures prices are assumed o follow a random walk, F F + f (7) wih disurbance f. Uncerainy abou he pool quaniy is represened by Q E (Q) + (8) u where u is a forecas error revealed in period. The errors (e, f, and u ) are assumed o be mulivariae normal wih zero mean, and are uncorrelaed across ime. Conemporaneous (posiive) correlaions exis beween e and f, he errors for cash and fuures prices. Correlaions may also exis beween u and he price errors, for reasons discussed below. The soluion o he overall problem involves backward inducion. Firs, opimal decision rules mus be derived for period, when pool quaniy is known. Then decision rules for period can be embedded in he opimizaion problem for period. In period, firs-order condiions for ODAL s opimizaion problem yield wo differen sraegies. ODAL could sore any unsold grain in period and place a fuures hedge (Sraegy A). Alernaely, he cooperaive could sell is grain immediaely and sore nohing unil period 3 (Sraegy B). 6

7 7 A) Sore grain and hedge ) / ( X H X Q X 0 X f ef 3 3 B) Sell grain, sore nohing H X X Q X The choice beween hese wo sraegies will depend on price relaionships observed in period and he expeced reurns o sorage. ODAL would be indifferen beween sraegies (A) and (B) under he following condiion: f e f ef 3 K P E P + δ (9) ha is, if he (compounded) period cash price equals he expeced period 3 price less a risk adjusmen. Here ef denoes he covariance of cash and fuures prices, f is he variance of fuures price, and e is he variance of he cash price. Subsiuing P b 0 + b P + e and E (P 3 ) b 0 + b P, equaion (0) solves for he criical value of e ha leaves ODAL indifferen beween sraegies (A) and (B) δ + δ f e f ef 0 0 * K P ) b )(b b ( b b e (0) The probabiliy ha ODAL will elec o sore and hedge (Sraegy A) in period is Φ e * e ob(a) Pr () where Φ denoes he sandard normal cdf. The probabiliy ha ODAL will elec o sell is remaining cash grain in period (Sraegy B) is Φ e * e ob(b) Pr ()

8 The larger is K, he risk parameer, he less likely is sorage wih deferral of grain sales o period 3. Now consider he period- decision problem. ODAL seeks o maximize Z subjec o risk consrain (3). Choice variables include X (cash grain sale), H (hedge posiion), and Z (expeced pool price). The risk consrain requires specifying he expeced value of pool revenue less payou (R Z Q), and variance of he same. These are given by E (R Z Q) Pr ob(a) E/ A (R Z Q) + Pr ob(b) E/ B(R Z Q) (3) and V (R + + Z Q) Prob(A) V Pr ob(a) [E Pr ob(b) [E / A / B (R (R / A Z Q) Z Q) (R Z Q) + E (R E (R Z Q)] Z Q)] Pr ob(b) V / B (R Z Q) (4) where E /A and E /B are condiional expecaions, and V /A and V /B are condiional variances, given indicaed price relaionships in period. Equaion (4) indicaes ha he variance of (R Z Q) equals he mean of condiional variances plus he variance of condiional means (Lindgren, 976). Formulas for condiional expecaions and variances given a runcaed normal disribuion are found in Greene (p.899 and 97). 4. Daa and Parameer Esimaes Daa required for he analysis include cash whea prices (P ) and whea fuures (F ). As here are no suiable official spo price quoes for whea from he Swedish grain marke, we sough prices from a relaed marke. A relevan markes for he Swedish grain rade is he French marke in Rouen, one of he larges grain markes in he EU (Tkaczyk, 999). Swedish milling whea is of higher qualiy han he sandard grade raded a Rouen, bu ranspor cos differenials are assumed o be relaively sable (Sinorn 999) 3. 3 Formally, we can describe he price relaionship as in (P S P R + Prem Trans); where P S is he spo price for Swedish grain, P R is he spo price for Rouen grain, Prem is he qualiy premium for Swedish grain and Trans represens he ransporaion cos parameer from Rouen o a Swedish harbor. 8

9 Currenly, here are wo whea fuures conracs raded on European fuures exchanges, namely he LIFFE feed whea fuures and he MATIF milling whea fuures. However, lile hisorical daa are available for he MATIF conrac, which sared rading in 998. Therefore, his sudy uses fuures quoes from he neares-o-maure LIFFE feed whea conrac. The conrac size is 00 meric ons and he price is quoed in Briish pounds (GBP). Bridge in Sockholm provided he fuures daa (999). LIFFE operaes whea fuures wih five mauring monhs: January, March, May, July, Sepember, and November. Our model does no explicily accoun for flucuaions in exchange raes. Insead, we conver all prices o quarerly averages in Swedish currency (SEK/on) a prevailing exchange raes. 4 Using daa from Ocober 993 o Sepember 999, he following equaion was esimaed by OLS (-raios in parenheses): P P F (5) (.) (5.790)* (7.4)* * significan a % level R-squred: observaions where F is he firs difference of fuures prices. The coefficien on F can be inerpreed as a minimum-variance hedge raio under he mainained hypohesis ha fuures evolve as a random walk (Myers and Thompson). For purposes of forecasing, (5) collapses o (6) wih b and b Le Le ν denoe he residual from (5). Thus, he cash price error e is consruced as e v * F (6) for our sample. Wih f F, he correlaion beween e and f is.845 in his period. The variance-covariance marix of e and f is shown below. 9

10 e ef e f f The expeced pool quaniy, E (Q), is fixed a.5358 million ons. This is based on a 70 percen marke share for ODAL, planed hecares for 000 and rend yields (Saisics Sweden 999). The sandard deviaion of u (he quaniy forecas) is based on residuals from a rend yield model, scaled o reflec acual planed hecares and he assumed 70 percen marke share: u Correlaions beween u and he price errors, e and f, are unknown; however, negaive correlaions seem plausible. Farmers who observe a price increase beween periods and migh choose o defer heir sales o ODAL. 5 Conversely, if prices should fall beween periods and, farmers would have a greaer incenive o commi heir grain o he pool (o claim par of ODAL s higher average price). For purposes of sensiiviy analysis, correlaions beween u and price errors are varied in he analysis repored below. 5. Model Resuls The base case for our analysis reflecs a number of assumpions ha can be briefly summarized. The iniial cash price for grain sold by ODAL (P ) is fixed a 000 SEK/on. To ensure a posiive expeced reurn o sorage 6, an adjusmen is made o he inercep in price equaion (6). The adjused inercep is b 0 * b Wih a quarerly ineres rae of.5 percen, his implies an expeced real price increase of abou.6 percen in period and period 3. The risk 4 Thompson and Bond (987) presen a hedging model ha explicily accouns for exchange rae flucuaions. They conclude from he derived soluion ha i is no possible o deermine effecs of exchange rae flucuaions on he opimal hedge raio. 5 Recall ha ODAL can open several pools in succession during a markeing year. Subscripions o he second pool begin when he firs pool is closed, and so on. 6 If expeced reurns o sorage are insufficien, ODAL sells all is cash grain immediaely. Wih cash price risk largely eliminaed, here is lile incenive o hedge. 0

11 parameer is se a K. Given (0) and () and he price adjusmen, his implies Prob(A).053, or a 5.3 percen chance of sorage beween periods and 3. The correlaions beween u and e, and beween u and f, are fixed a -. in he base case. Resuls for he period- decision variables are shown in Table. Table. Base Case Resuls. Variable Unis Value Z Expeced pool price SEK/on 09. X Physical sales Million ons 0.87 H Sales of fuures Million ons The opimal soluion calls for an immediae cash sale of 0.87 million ons and a shor fuures posiion of 0.99 million ons. Since he expeced pool quaniy is.5358 million ons, he hedge raio (HR) for unsold grain is HR H E (Q) X Noe ha his is smaller han he minimum-variance hedge raio implied by regression equaion (5). Figures and show he effecs of alernaive parameer values on model resuls. The correlaion beween price and quaniy forecas errors are allowed o vary beween 0 and 0.8 (as compared o 0. in he base case). The figures also show he effec of a larger price adjusmen han assumed in he base case. Expecaions of higher cash prices (b 0 * b ) in periods and 3 resul in lower hedge raios (Figure ), as well as higher expeced pool prices (Figure ). Hedge raios also decline as he correlaion increases, in absolue value, beween price and quaniy forecas errors.

12 80% Hedge raio 75% 70% 65% 60% b b % Correlaion beween price and quaniy Figure. Impac of Price Expecaions on Hedge Raio

13 Expeced Pool Price b b Correlaion beween price and quaniy Figure. Impac of Price Expecaions on Expeced Pool Price To measure he impac of hedging on expeced pool price, i is necessary o compare model resuls wih and wihou hedging. In he consrained model, hedging is no allowed (H H 0). Firs-order condiions for he period- problem are modified accordingly. The erm under he radical in (0) is replaced by e, and probabiliies of period- sorage () and sales () are revised o reflec he greaer risk associaed wih cash grain posiions. Sorage is discouraged, and expeced pool price is lowered, relaive o he unconsrained case. Opimal values for his case are displayed in Table 3: Table 3. Resuls for Period in he Case of No Hedging. Variable Unis Value X Physical sales Million ons.494 H Sales of fuures Million ons 0 Z Expeced pool price SEK/on

14 Table 4 shows he impac of hedging for differen values of K, he measure of risk sensiiviy. The base case corresponds o K equal o one. Wih higher values of K, he probabiliy of period- sorage goes o zero, 7 and he conribuion of hedging o ODAL s expeced pool price is inconsequenial. However, a lower levels of risk sensiiviy (K0.5), hedging increases he expeced pool price by.8 percen. Table 4. Impac of hedging wih differen levels of risk sensiiviy. K: Risk Sensiiviy Prob(A): Probabiliy of Period- Sorage wih hedging No hedging Z (SEK / on): Expeced Pool Price % in Z due o hedging wih hedging no difference hedging The impac of hedging is more pronounced when here are higher expeced reurns o sorage. In Table 5, wo differen price scenarios are compared. The firs (+35) corresponds o he base case, and he second (+ 50) represens a larger expeced price increase in periods and 3. In he laer case, hedging resuls in a 4% higher expeced pool price. Table 5. Impac of hedging wih differen expeced reurns o sorage. Price Adjusmen Prob(A): Probabiliy of Period- Sorage wih hedging No hedging wih hedging Z (SEK / on): Expeced Pool Price no hedging difference % in Z due o hedging To avoid compuaional errors, a lower bound of 0.00 is placed on Prob(A). 4

15 Thus far we have only discussed opimal values of choice variables in period. However, he disribuion of oucomes in he second and hird periods is also of ineres and can be evaluaed hrough simulaion echniques. Specifically, we ake 5,000 random drawings of he disurbance erms in equaions (6) hrough (8) and simulae he impacs of opimal sraegies by ODAL. The procedure for simulaing error erms is described by Johnson and Wichern (998). Parameers for he base case are unchanged: he risk parameer K is kep a uniy, and he add-facor for price adjusmens remains a 35. Opimal sraegies for period are he same as hose shown in Table. For period, he opimal sraegy 8 depends on he realizaion of price disurbances. Sraegy A (sore grain and hedge) exposes ODAL o coninuing price risk, while Sraegy B (sore nohing) involves liquidaing he remaining grain invenories. If Sraegy A is pursued, here is a chance ha marke revenues will no be sufficien o cover ODAL s price commimen (in which case, paronage refunds will be negaive in period 3). If Sraegy B is pursued, he pool price can be se equal o he average of marke revenues, and paronage refunds are zero. The resuls are compared wih hedging and wihou hedging. Resuls from he 5,000 ieraions on he (period ) pool price and (period 3) paronage refund are shown in Tables 6 and 7 below. 8 Relevan decision rules for X (cash grain sales), H (hedge posiion), and Z (pool price) under each sraegy are derived from he period- opimizaion problem. Deails are available from he auhors. 5

16 Table 6. Simulaed Period- Pool Prices and Paronage Refunds. Wih Hedging Pool price Z Paronage Refund PR Wihou Hedging Pool Price Z Mean Sandard Deviaion Kurosis Skewness Minimum Maximum When hedging is allowed, he average period- pool price is.4 percen higher han oherwise. When hedging is no allowed, he paronage refund is zero because no risks remain in period 3. Noe ha he period- pool price has a lower sandard deviaion when hedging is no allowed. The reason is ha in he absence of fuures markes, ODAL markes a larger volume in he firs period when he cash price is known. When hedging is allowed, ODAL assumes greaer risk in order o realize a higher expeced price (hrough sorage). The period- pool price is no normally disribued in eiher case (wih hedging allowed, or hedging no allowed). 9 The posiive skewness value for Z (wih hedging) indicaes ha is disribuion has an exended righ-hand ail. This is in conras o Z (wihou hedging), which is characerized by negaive skewness. 9 The Anderson-Darling es of he null hypohesis of normaliy is rejeced a a high level of significance. 6

17 6. Concluding Remarks This paper presens a muli-period model of price pooling ha accouns for boh price and quaniy uncerainy. The analysis focuses on he poenial gains from hedging in he conex of a cooperaive s price pooling sysem. When hedging is allowed, a lower share of grain is sold a harves in he cash marke. Simulaion resuls show ha he disribuion of pool prices also changes considerably when he cooperaive is allowed o hedge is cash grain posiion. As discussed in Carer (984), price risk exposure a he farm level originaes as early as he ime of planing in he spring. Therefore, o ake full advanage of hedging wihin a price pooling sysem, i is desirable o exend he planning horizon backwards o spring. An alernae roue would involve forward conracs beween he cooperaive and farmer-parons, signed early in he planing season; his would also reduce he cooperaive s uncerainy abou quaniies o be markeed. Incorporaing hedging in ODAL s markeing sraegy will pose a challenge o managers, who o his poin have lile direc experience wih fuures markes. Equally imporan, parons mus accep he new cooperaive markeing sraegy a poin emphasized by Fulon, Popp and Gray (998). Buccola and Subaei (985) add anoher dimension o he problem on how he cooperaive should develop pricing sysems. If members are heerogenous in erms of heir risk preferences, feaures of some pooling arrangemens migh no be accepable o all parons. The 7

18 laer aspecs form an ineresing opic for fuure research of how o opimally manage price pools for subses of member caegories. References Bäckius, J. BRIDGE. Provider of fuures price daa. Sockholm: Sweden, 999. Brassley, P. Agriculural economics and he CAP: an inroducion. Oxford : Blackwell Science, 997. Buccola, S., and A. Subaei. Opimal Marke Pools for Agriculural Co-operaives. American Journal of Agriculural Economics. 67(February 985): Carer, C.A. Pooling Sales versus Forward Selling and he Managemen of Risk: The Case of Whea. In: Inernaional Agriculural Trade: Advanced Readings in Price Formaion, Marke Srucure, and Price Insabiliy. Edied by Sorey, G. G., A. Schmiz, and A. H. Sarris, Inernaional Agriculural Trade. Boulder, Colorado: Wesview Press, 984. Fulon, J. R., Popp, M. P., and Gray, C. Evolving Business Arrangemens in Local Grain Markeing Cooperaives. Review of Agriculural Economics. 0(998): Greene, W. H. Economeric Analysis, Fourh Ediion. Upper Saddle River, NJ: Prenice Hall, 000. Johnson, R. A. and D.W. Wichern. Applied Mulivariae Saisical Analysis, Fourh ediion. New Jersey: Prenice Hall, 998. Kamara, A. Issues in Fuures Markes: A Survey. Journal of Fuures Markes. (Fall 98):6-94. Karlsson G. ODAL Grain Division. Provider of ODAL daa. Norrköping: Sweden, 999 and 000. Lindgren, B. W. Saisical Theory, Third Ediion. New York: MacMillan Publishing Co., 976. Myers, R. J, and S. R. Thompson. Generalized Opimal Hedge Raio Esimaion. American Journal of Agriculural Economics. 7(989):

19 Myers, R.J., and S.D. Hanson. Opimal Dynamic Hedging in Unbiased Fuures Markes. American Journal of Agriculural Economics. 78(February 996):3-0. Nilsson, T. Opimal Hedging Sraegies for Swedish Grain Agens. Maser Thesis. Uppsala: Swedish Universiy of Agriculural Sciences, 00. Sinorn, J. Prissäning skörd 97. Odalaren Mellansvenska Lanmännen Odals Medlemsidning. 4(July 997):5-6. Sinorn, J. ODAL Grain Division. Provider of ODAL daa. Norrköping, Sweden, 999 and 000. Saisics Sweden, Yearbook of Agriculural Saisics. Bread grain in Toal yields, meric ons. 5 percen waerconen. Official Saisics of Sweden. Saisics Sweden, various issues Taha, H. A. Operaions Research, Second Ediion. New York: MacMillan Publishing Co., 976. Thompson, S.R. and G.E. Bond. Offshore Commodiy Hedging under Floaing Exchange Raes. American Journal of Agriculural Economics. (February 987): Tkaczyk, L. MINTEC. Provider of spo price daa. London: Unied Kingdom, 999. Walldén, H. Ökad marknadsanpassning måle. Odalaren Mellansvenska Lanmännen Odals Medlemsidning. 4(July 997):4-5. 9