Transportation Costs in Econometric Models of State Agricultural Sectors: The Case of Beef in Hawaii

Size: px
Start display at page:

Download "Transportation Costs in Econometric Models of State Agricultural Sectors: The Case of Beef in Hawaii"

Transcription

1 Trnsporttion Costs in Econometric Models of Stte Agriculturl Sectors: The Cse of Beef in Hwii Rolnd K. Roberts Econometric models designed to show how ntionl policies ffect stte griculturl sectors often use ntionl prices s proxies for stte prices. Consequently, they ignore the influence of freight rtes on stte production. An ppliction to the Hwii beef industry demonstrtes tht both freight rtes nd ntionl beef prices hve importnt impcts on Hwii beef prices nd production. By using stte prices rther thn ntionl prices, error from chnges in freight rtes might be reduced, nd the model's cpcity for policy nlysis might be brodened. Interest hs grown in developing stte econometric models for policy nlysis (Knpp et l.). For stte griculturl model to be useful for wide rnge of policy nlyses, it should be ble to indicte stte-level impcts of chnges in both stte nd ntionl policies. Bum et l. employed such model to nlyze the impcts of U.S. beef import policy on the Virgini beef nd pork sectors. However, they used ntionl prices in their beef nd hog production equtions insted of Virgini prices. Such n pproch cn bis stte model nd subsequent impct nlyses. If prices re not totlly trnsmitted Rolnd K. Roberts is n Associte Professor, Deprtment of Agriculturl Economics nd Rurl Sociology, University of Tennessee. This reserch ws conducted while the uthor ws t the University of Hwii. The uthor is grteful to Peter Grrod nd John Roecklein for their contributions towrd the development of this pper. The suggestions of the nonymous reviewers re lso pprecited. This reserch contributes to Western Regionl Project W-145. Hwii Institute of Tropicl Agriculture nd Humn Resources Journl Series Number within the current period, or if there is trnsporttion cost differentil between ntionl nd stte prices, the use of stte prices estimted through price trnsmission equtions might reduce tht bis. Furthermore, the impct informtion obtinble from stte model would increse s the rry of policy vribles is ugmented by trnsporttion costs. An econometric model of the Hwii beef industry is used s cse study to demonstrte the potentil improvement in impct nlysis by stte econometric models tht use stte rther thn ntionl beef nd feed prices. The specific objectives of this pper re 1) to demonstrte the importnce of trnsporttion costs s determinnts of Hwii beef nd feed prices; 2) to illustrte tht introducing trnsporttion costs my eliminte specifiction bis nd render greter relibility to subsequent impct nlysis; nd 3) to demonstrte the ugmented policy nlysis cpbilities of stte griculturl commodity model tht includes trnsporttion costs nd ntionl prices in its stte price trnsmission equtions. Hwii provides unique setting for exmining the importnce nd usefulness Western Journl of Agriculturl Economics, 1(1): by the Western Agriculturl Economics Assocition

2 July 1985 of trnsporttion costs in stte commodity models. It is locted bout 2,5 miles from the rest of the United Sttes. This isoltion leds to richness of dt in tht trnsporttion costs re typiclly higher thn for other sttes nd more esily identified. Hwii cn be thought of s the extreme cse. If trnsporttion costs re not importnt in determining Hwii beef nd feed prices, they re unlikely to be importnt in determining those prices in other sttes. Whether trnsporttion costs re importnt price determinnts is n empiricl question tht depends on ech stte's or region's loction reltive to mjor mrkets. For exmple, the Omh utility cow price ws used in the Bum et l. model to determine Virgini beef cttle slughter. Whether there ws significnt trnsporttion cost differentil between Omh nd Virgini utility cow prices ws not ddressed. In this nlysis, Hwii beef nd feed prices re estimted s function of Los Angeles beef nd corn prices nd Los Angeles-to-Honolulu freight rtes. Although the dt nd results re specific to Hwii's beef production sector, this pper's pproch cn be dpted to ny stte or region in which prices re determined exogenously nd freight chrges re reltively importnt. A similr method my be pproprite for number of developing ntions tht import significnt proportion of commodity. The importnce of trnsporttion costs s determinnts of Hwii beef nd feed prices is stressed through regression nlysis, nd the Hwii beef model is briefly outlined. Three versions of the model re simulted to emphsize differences in estimted impcts when trnsporttion costs re not considered. The model's improved policy nlysis cpbilities re lso demonstrted under vrious ssumptions bout chnges in trnsporttion cost vribles. Finlly, implictions nd generl conclusions re drwn. 94 Western Journl of Agriculturl Economics Trnsporttion Costs s Determinnts of Hwii Prices The Hwii Agriculturl Reporting Service estimted tht 198 beef imports from the Minlnd United Sttes nd foreign sources (Austrli nd New Zelnd) ccounted for 53 nd 16 percent, respectively, of totl Hwii beef mrket supply. However, imports to Hwii from the Minlnd were only bout.2 percent of Minlnd production nd 1.3 percent of Austrlin nd New Zelnd exports to the entire United Sttes (Schermerhorn et l.). Becuse of Hwii's insignificnce in ntionl nd interntionl mrkets, Hwii beef prices re ssumed to be determined exogenously. The demnd for beef in Hwii hs little impct on locl beef prices. The difference between the Hwii demnd for beef nd locl production, t the exogenously determined price, cn esily be ugmented by imports from the Minlnd nd foreign sources. Becuse of the dominnce of imports in the Hwii mrket nd the influence of the United Sttes s pricemker in interntionl beef mrkets (Simpson, p. 1), it follows tht wholesle prices of Hwii-produced beef should be closely relted to Minlnd prices nd trnsporttion costs. Similrly, the price of feed in Hwii is ssumed to be determined by Minlnd price nd trnsporttion costs. On the other hnd, exogenously determined prices influence production decisions in Hwii, llowing trnsporttion costs to ffect Hwii beef production. Four equtions were specified to reflect beef nd feed price trnsmission from the Minlnd United Sttes to Hwii. Explntory vribles included current nd lgged Los Angeles prices nd ocen freight rtes from Los Angeles to Honolulu. Qurterly dummy vribles were lso included becuse sesonl vrition in stte, ntionl nd interntionl beef mrkets were expected to influence price trnsmission. Ech eqution ws first es-

3 Roberts timted by ordinry lest squres. The residuls were used to clculte the Durbin-Wtson sttistic (DW) nd similr sttistic (D4) designed to test for fourth order utocorreltion (Wllis). Where significnt first order utocorreltion ws indicted, Cochrne-Orcutt utoregression procedure ws used to obtin more efficient prmeter estimtes. Lg structures were not specified priori. Therefore, in equtions where lgs in price trnsmission were hypothesized, the number of lgs ws determined by including successively longer lgs until the coefficient of the finl lg becme negtive or negligible reltive to its stndrd error. The finl price trnsmission equtions (Equtions 1A-4A) re presented in Tble 1. The R2 re ll greter thn.96, suggesting tht the explntory vribles provide good fit (Kment, p. 234). In no cse does the D4 sttistic indicte significnt fourth order utocorreltion t the 5 percent level nd sesonl effects re only significnt in the grss-fed steer nd heifer price trnsmission eqution (Eqution 3A). 1 Ocen freight rtes re used in Equtions 1A-4A becuse time-series on totl trnsporttion costs for beef nd feed from Los Angeles to Honolulu re not redily vilble. Although ocen freight costs represent significnt portion of totl trnsporttion costs, other logistics costs such s whrfge fees, lnd trnsporttion costs for huling to nd from the docks, nd storge cn ccount for perhps s much s one-hlf of the totl cost (Grrod). Ocen freight rtes cn be viewed s proxies for totl trnsporttion costs becuse ll trnsporttion costs, whether for 1Sesonl dummy vribles were retined in Equtions 1A, 2A nd 4A for comprison. Reestimtion without qurterly dummy vribles did not pprecibly ffect the coefficients of the remining vribles nd the R 2 vribles only incresed slightly. For exmple, the R 2 of Eqution 2A incresed from.9847 to.985 when dummy vribles were dropped. Trnsporttion Costs in Stte Models lnd or se trnsporttion, re highly correlted with energy nd lbor costs. The freight rte vribles re ll highly significnt, with coefficients rnging from 2.27 in the Honolulu choice beef price eqution (Eqution 1A) to 2.57 in determining the price of grss-fed beef (Eqution 3A). These coefficients pper lrge t first glnce, but they re cceptble when one ccounts for nonocen trnsporttion costs. If the trnsporttion cost vribles in Equtions 1A-4A were totl trnsporttion costs rther thn ocen freight rtes, the expected size of the coefficients would be bout 1.. Two conditions increse the expected size of the coefficients. First, if ocen freight costs were one-hlf s much s totl trnsporttion costs nd if other logistics costs were highly correlted with ocen freight rtes, then n increse in ocen freight rtes by $1. per hundredweight would be ccompnied by n increse in totl trnsporttion costs by $2.. Hence, the price of beef or feed in Hwii would increse by bout $2. per hundredweight. Second, the ocen freight rtes used in this nlysis ssume tht continers re full, which is not lwys true. Prtilly full continers re chrged higher rte per hundredweight. Therefore, the ctul rtes re probbly higher thn the rtes used, further incresing the expected size of the coefficients. Given tht certin relevnt trnsporttion cost vribles re omitted, it should be cler tht Equtions 1A-4A re not presented s the true price trnsmission models. The coefficient on the Los Angeles steer price (LAGFBPRQ) in Eqution 1A nd the sums of the Los Angeles price coefficients in Equtions 2A-4A re sttisticlly different from unity. Divergence from unity might result from number of things such s differences in products nd levels of mrketing, imperfect price trnsmission or specifiction error cused by the omission of relevnt vribles. Becuse of locl pricing mechnisms, Eqution 1A most closely fits the Min- 95

4 July 1985 Western Journl of Agriculturl Economics LL 'L c -t o" co co co CD ) LO I CM - CIJ CI Cf) T C=; c ci6-1 CC-J 6 6 Cc - CO) CO I I- I I - <:!- LO CO CM CM 1- CV4 6 o oo oc r- s. L (.COLO 't CM o o- T- CM C o I-, 66i 6 o (UM Q) n E z r Cu c w:3 I V C co ) n Cu.u C c ) QC C) > i I CL CE LL z I m CO C) ce t r -CD cd t O C OC COi CM Co o o (D CO C) C O C) O (D O CO O O CD! ) 4 Oi N -) YO) IC CMJ cdc o cc c; C ; Co cm C C-- I C O co CO CmO O O C) c c 1-1 d6c 6 I.- C 1- c c 1- _ CO CM C; CE o I- Co o o66oo -. I.., L) NC CO co _ 1 co _c LO ) (D NT- CD C\i o o- -t '. r. co r- ) - - ) LeV) M 14t ) 14t co CM Q) r CO COO)i ' -4I.-CO- u, LU c,.i E cl C C) m < CM C) r- O Or- -O CO C CC - O ) h : C5 C C LO O cci I. -I- OC O O CM CY) LO CM O 1- CdC; c; o -1 d o oc I- (U IL _. C Cu - C X > i^ CE. W m CE LL. C3 4.1 T- c 1 CM I r. 14 C I or I n- O1^ z cr 15 ^ CMI Nl _- _I z z Ce Cc m o O < 4 ^ c z CEH o o o 96

5 Roberts Trnsporttion Costs in Stte Models z C U) w C ( U c ) I. ) LL cr m z I IL r 3 r L m I m t VI m CO co CMJ CD ct - < ) Rd, co o o CO O O CO I -.O Lc TC 66 I ~ t1- I- CD r- CO C'4!> CD CD > z o cu N. CDT- (D T M ~ D -OD )) ' Q) CY) c; ; : -r ; I.- CN ) CM C O 7- T.: c LO ) C- ;o o c; ci ci - co O c (D CO C o ) ) t - O C) oo C- c\i o CO CC co -_COY) Q- C Q.. U) z co * ' c o ~ c CO U. C E CeM.o o Lo 4. X ) C 2N '. co E 2 U 2) U> Ew o co co ) Co Lc (D ) < ) ) : n ) ) o C C (u ) > - U)C.o lnd price plus trnsporttion cost model. Once week the mjor Hwii slughterhouses cll slughterhouses in Los Angeles for price quottions. Hwii grin-fed steer nd heifer prices re bsed on those quottions plus mrkup for trnsporttion costs. The coefficient for the Los Angeles wholesle choice steer price (LAGFBPRQ) is close to unity (.98) s expected. 2 The coefficient for TRANBQ nd the constnt term suggest tht the trnsporttion cost differentil between Los Angeles nd Honolulu is bout 2.27 times the ocen freight rte minus $7.18 per hundredweight. Trnsmission of cow prices from the Minlnd to Hwii is more complicted thn for choice beef. Pricing methods re not s well defined nd, becuse Hwii imports lrge quntities of cow beef from Austrli nd New Zelnd, price trnsmission from the Minlnd is indirect vi the Austrlin nd New Zelnd mrkets. Lgged Los Angeles cow prices re included in Eqution 2A to cpture price trnsmission delys cused by the gret distnces involved nd the time required for chnges in the U.S. cow price to work through the Austrlin nd New Zelnd mrkets to Hwii. The sum of the coefficients on the current nd lgged Los Angeles cow prices is.94, which is gin resonbly close to unity given differences in products nd the indirect trnsmission of prices through the Austrlin nd New Zelnd mrkets. Severl fctors complicte trnsmission of grss-fed steer nd heifer beef prices to -d U) C) m 4 I- o ) Q ci X > w I IS ' DC^ CM I In ' In o o' ) j. F ( E me I > U) (o c'4-u)u 2)U, c *5 re )c CD o *. t))z U) CO 2 The true prmeter estimted by this regression coefficient would be equl to 1. only if the Los Angeles nd Honolulu prices were for identicl product nd prices were perfectly trnsmitted. In this cse, the Los Angeles price is for steers, while the Honolulu price is for steers nd heifers. Therefore, the true prmeter should be close to unity, but not necessrily equl to unity. The estimted coefficient is sttisticlly different from 1.. However, whether it is close to unity in the sme sense tht the true prmeter is close to unity is mtter of judgment. 97

6 July 1985 Hwii. First, there re no wholesle grssfed steer nd heifer beef prices in Hwii or on the Minlnd. Second, dressed weight price received by frmers is recorded in Hwii but not on the Minlnd. Third, s with cow beef, the Hwii price is determined by the Minlnd mrket vi the Austrli nd New Zelnd mrkets. Becuse Hwii-produced grssfed beef competes with both cow nd grss-fed steer nd heifer beef imported from Austrli nd New Zelnd, it is hypothesized tht Minlnd steer nd cow prices re both highly influentil in determining the Hwii grss-fed steer nd heifer beef prices. Eqution 3A uses current nd lgged Los Angeles utility cow prices, nd current nd lgged differences between the Los Angeles choice steer price nd the utility cow price, to represent the influence of the Minlnd beef mrket on the Hwii grss-fed steer nd heifer price. The current nd lgged price coefficients of Eqution 3A suggest tht, if both Los Angeles prices incresed by $1., the Hwii grss-fed beef price would increse by $.64. An increse by $1. in the Los Angeles utility cow price or the Los Angeles choice steer price, holding the other price constnt, would result in increses in the Hwii grss-fed beef price of $.19 nd $.5, respectively. These coefficients seem resonble given differences in commodities nd levels of mrketing. Minlnd prices directly determine the Hwii cttle feed price pid by frmers. Most of the feed used is mnufctured in Hwii from feed stuffs imported from the Minlnd. Reltively little mnufctured feed is received from the Minlnd for use by cttle. Pricing methods re poorly defined. Therefore, current nd lgged Los Angeles wholesle corn prices re used in Eqution 4A to cpture delys in price trnsmission from the Minlnd to Hwii nd from one level in the mrketing chin to nother. The sum of the current nd lgged price coefficients is Western Journl of Agriculturl Economics.57. This is cceptble given differences in the levels of processing nd mrketing. Equtions 1B-4B of Tble 1 re identicl to Equtions 1A-4A except tht trnsporttion cost vribles re excluded. These equtions suggest tht positive bis might be present in the price coefficients resulting from omission of relevnt vribles. The estimted coefficients of the trnsporttion cost vribles re positive nd highly significnt in Equtions 1A- 4A. Therefore, if prices nd ocen freight rtes re positively correlted, this omission would likely produce positive bis in the price coefficients of Equtions 1B- 4B (Kment, pp ). Price vribles nd freight rtes re highly correlted. The simple correltion coefficients between the beef freight rte nd current nd lgged Los Angeles beef prices re ll greter thn.8. Similrly, ll correltion coefficients between the feed freight rte nd current nd lgged Los Angeles corn prices re greter thn.65. The consequences of omitting freight rtes re evident. Almost without exception the estimted price coefficients re lrger in Equtions 1B-4B thn in Equtions 1A-4A. This is not to sy tht Equtions 1A-4A re without specifiction bis. It is cler, however, tht one source of specifiction error is eliminted by including ocen freight rtes. These findings suggest tht if freight rtes were omitted from model of Hwii beef production, subsequent impct nlyses would be dversely ffected. Differences in simulted impcts re ddressed fter the Hwii beef model is briefly presented. The Hwii Beef Model The model used for this nlysis hs been described in detil (Roberts et l.). The version used in this pper consists of 23 equtions, 13 of which re behviorl reltionships estimted from qurterly nd nnul dt for 197 through 198. The 98

7 Roberts Trnsporttion Costs in Stte Models model is divided into four sections. The first section dels with Minlnd-to-Hwii price trnsmission. For this nlysis, Equtions 1A-4A of Tble 1 replce the first seven equtions of Roberts et l. Exogenously determined Hwii beef nd feed prices nd n energy price index re used in Section 2 to determine Jnury 1 inventories of beef cows, heifers, heifers held for replcement, heifers not held for replcement, nd steers. The nnul clf crop is lso estimted in Section 2. Section 3 uses the prices from Section 1 nd the cttle inventories generted in Section 2 to estimte qurterly production of grinfed beef, grss-fed steer nd heifer beef nd cow beef. For completeness, bull beef production is lso estimted s function of cow beef production. Finlly, Section 4 links the other sections through period trnsition identities tht convert qurterly prices into nnul verges for use in Section 2. Jnury 1 cttle inventories lso re modified for use in the qurterly production equtions of Section 3. Exogenously determined prices gretly simplify estimtion procedures. The mtrix of endogenous vrible coefficients is tringulr, nd recursive system is ssumed (Johnston). Consequently, ordinry lest squres nd Cochrne-Orcutt utoregression procedures (White) were used to estimte the structurl equtions of the model. For equtions including lgged dependent vribles, prtil djustment ws ssumed (Nerlove), nd Grid Serch utoregression procedure ws used to verify tht the Cochrne-Orcutt procedure converged to globl mximum of the likelihood function (Betncourt nd Kelejin). Modifictions of the Model for Impct Anlysis Simultions of the model over the 1972 through 198 period re used to emphsize differences in simulted impcts nd other problems resulting from the exclusion of trnsporttion costs. Three versions of the model re used. Model I uses Equtions 1A-4A to trnsmit prices from the Minlnd to Hwii, while Model II uses Equtions 1B-4B for price trnsmission. All other equtions re identicl in Models I nd II. In Model III, ll inventory nd production equtions re reestimted, ccording to the specifiction of Models I nd II, using Los Angeles beef nd corn prices s regressors rther thn Hwii prices. Therefore, Model III includes no price trnsmission equtions. The reestimted equtions re presented in Tble 2 for comprison with those of Roberts et l. Symbols re defined in Tble 3. Using Los Angeles prices introduces n dditionl misspecifiction into Model III. Becuse there re no prices for Los Angeles grssfed steer nd heifer beef, the Los Angeles utility cow price is substituted s proxy in the grin- nd grss-fed beef production equtions. Additionl differences in impcts would result to the extent tht policy chnge or other exogenous shock ffected the Los Angeles utility cow price differently from the Hwii grss-fed steer nd heifer price. Historicl Simultions of Models I-III Before simulted impcts re mesured, the reltive bilities of the models in trcking historicl events re ssessed by compring Theil U 2 coefficients (Leuthold). 3 The U 2 coefficients re clculted for ech model by performing dynmic simultion over the 1972 through 198 period. Observtions for re excluded to ccommodte lgs. Ech simu- 3 The U 2 coefficient is defined in this nlysis s V[(Pt- Atl) - (A, - At-) U2 = 2 /V(A, - A,_) with Pt being the predicted outcome nd At being the ctul outcome for period t. 99

8 July 1985 Western Journl of Agriculturl Economics Tble 2. Estimted Equtions nd Identities of the Hwii Beef Econometric Model Using Ntionl Rther Thn Hwii Prices (Model III). Eqution numberb Eqution I. Qurterly Price Trnsmission Equtionsc II. Annul Cttle Inventory nd Clf Crop Equtions 5(9) BCI = LAGFBPR(-1)/LACFPI(-1) (35.429) (.6) -.9LP(- )/LACFPI(-1) +.769BCI(-1), (.1) (.342) R 2 =.767, DHd, OLS. 6(1) HI = CC(-1).46LAGFBPR(- 1)/LACFPI(-1) (8.549) (.95) (.16) +.12ILP(-1)/LACFPI(-1) +.759HI(-1), (.2) (.12) R2 =.7942, AUT, = (.261) 7(11) OHI = (HI - HHDCR) -.11LAGFBPR(-1)/LACFPI(-1) (2.785) (.7) (.8) -.9ILP(-1)/LACFPI(-1), (.1) R 2 =.9651, AUT, = (.272) 8(12) HHBCR = HI - OHI - HHDCR. 9(13) SI = CC(-1) -.45LAGFBPR(-1)/LACFPI(-1) (1.38) (.134) (.21) -.2ILP(-1)/LACFPI(-1) +.553S1(-1), (.4) (.189) R 2 =.866, AUT, = (.37) 1(14) CC = (BCI + DCI) +.784(HHBCR + HHDCR), (24.931) (.183) (.375) R 2 =.7275, AUT, = (.273) III. Qurterly Beef Production Equtions 11(15) GFBPQ = -5, TSOHIQ*D1Q TSOHIQ*D2Q (1,61.2) (27.69) (2.988) TSOHIQ*D3Q TSOHIQ (15.976) (26.428) TSOHIQ(-4)*D1Q RSOHIQ*D1Q (24.56) (56.1) + 1,346.2RSOHIQ*D2Q RSOHIQ*D3Q + 3,636.8RSOHIQ (726.57) (542.6) (834.32) LAGFBPRQ(-1)/LACFPIQ(-1) LACPRQ(-1)/LACFPIQ(-1) (11.835) (1.67) ILPQ(-1)/LACFPIQ(-1) (1.529) 1.52[LAGFBRQ(-1 )/LACFPIQ(- 1) - LAGFBPRQ(-2)/LACFPIQ(-2)] (6.49) 1

9 Roberts Trnsporttion Costs in Stte Models Tble 2. Continued. Eqution number b Eqution [OILPQ(-1)/LACFPIQ(-1) - OILPQ(-2)/LACFPIQ(-2)] (1.675) DM1Q DM2Q DM3Q WQ TQ, (215.38) (19.64) (33.21) (13.69) (14.823) R 2 =.9499, AUT, p =.328. (.151) 12(16) NFBPQ = TSOHIQ*D1Q TSOHIQ*D2Q (343.1) (1.34) (.713) TSOHIQ*D3Q TSOHIQ RSOHIQ (.999) (5.95) (156.92) -.141GFBPQ(-3) LAGFBPRQ(-3)/LACFPIQ(-3) (.31) (2.58) LACPRQ(-3)/LACFPIQ(-3) LPQ(-3)/LACFPIQ(-3) (2.611) (.314) WQ(-3) DM3Q +.629NFBPQ(-1), (31.89) (59.849) (.93) R 2 =.9164, AUT, = (.157) 13(17) TSHBPQ = GFBPQ + NFBPQ. 14(18) CBPQ= 3, CIQ*D1Q CIQ*D2Q +.457CIQ*D3Q (981.95) (.484) (.484) (.411) CIQ [LACPRQ/LACFPIQ - LACPRQ(-1 )/LACFPIQ(-1)] (9.282) (4.271) [LAGFBPRQ/LACFPIQ - LAGFBPRQ(-1 )/LACFPIQ(-1)] (3.532) TQ WQ, (3.216) (47.959) R2=.613, AUT, p =.38. (.141) 15(19) BBPQ = CBPQ D1Q D2Q (64.666) (.38) ( ) (15.92) D3Q +.459BBPQ(-1), (15.296) (.142) R 2 =.319, DH =-.851, OLS. 16(2) TBPQ = TSHBPQ + CBPQ + BBPQ. 4 17(21) LACFPI(L)e =.25 LACFPIQ(t). t=l 4 18(22) LAGFBPR(L) =.25 LAGFBPRQ(t). t=i 4 19(23) OILP(L) =.25 S OILPQ(t). t=1 2(24) CIQ(t = 1-4)= BCI(L) + DCI(L). IV. Period Trnsition Identities 11

10 July 1985 Western Journl of Agriculturl Economics Tble 2. Continued. Eqution numberb Eqution 21(25) TSOHIQ(t = 1-4)= SI(L) + OHI(L). 22(26) RSOHIQ(t = 1-4) = S(L)/OHI(L). In the utoregressive equtions (AUT), R 2 is viewed only s mesure of goodness-of-fit (Kment, p. 234). Numbers in prentheses below coefficients re estimted stndrd errors (symptotic stndrd errors for AUT equtions). Numbers in prentheses following vrible nmes indicte lgs. Vribles re defined in Tble 3. b Numbers in prentheses following eqution numbers indicte the corresponding eqution in Roberts et l. c Price trnsmission equtions re not used in this version of the model. d Clcultion of Durbin's h sttistic ws not possible. e L refers to the current yer nd t refers to the qurter of tht yer. ltion is performed by llowing the model to iterte, with exogenous vribles set t historicl levels nd lgged endogenous vribles set t predicted vlues. These simultions re used s bses for clculting simulted impcts in the next section. The U 2 coefficients clculted from the models re presented in Tble 4. When compring models, lower U 2 coefficients indicte greter ccurcy. The U 2 coefficients re lower for Model I thn for Models II nd III, with the exception of bull beef production (BBPQ), for which the U 2 coefficient is lowest for Model II. In no cse is the U 2 coefficient smller for Model III thn for Models I or II. These comprisons demonstrte (1) tht the exclusion of trnsporttion costs from the price trnsmission equtions substntilly decreses the ccurcy of the model nd (2) tht elimintion of price trnsmission equtions further reduces the model's goodness-of-fit to historicl dt. 4 4 The inventory nd production equtions of Models I nd II were theoreticlly specified nd djusted ccording to goodness-of-fit criteri (eg., ir 2, stndrd error nd signs of the coefficients) nd dt vilbility. Hd the sme criteri been used to djust the inventory nd production equtions of Model III, it is possible tht better fitting model, with different structure, would hve been developed. However, given the exclusion of freight rtes, the lck of Los Angeles grin-fed steer nd heifer price, nd other dt limittions, it is unlikely tht such model would hve U 2 coefficients lower thn those of Model I. 12 Differences in Impcts Severl dditionl simultions re conducted to emphsize differences in simulted impcts mong the models. Subsequent simultions for prticulr model re compred to their respective bses. Impcts re defined s devitions from the bse. Three dditionl simultions re performed with ech model. Ech dditionl simultion increses one of the three Los Angeles prices by 1 percent bove nnul historicl levels, while other vribles re kept t bse vlues. Tble 5 contins the simulted impcts of key vribles, verged over the simultion period (1972-8) to conserve spce. Before impcts re compred mong the models, it is helpful to describe briefly the impcts produced by Model I. Generlly, the verge impcts for Model I re s expected. An increse in the Los Angeles choice steer price (LAGFBPRQ) of $8.1 for the period results in increses of $8. nd $3.7 in the Honolulu choice beef price (HCFBPRQ) nd the Hwii grss-fed steer nd heifer price (HNFBPRQ), respectively. Thus, HGFBPRQ increses reltive to HNFBPRQ, resulting in n increse in grin-fed beef production (GFBPQ) of 1,59, pounds nd decrese in grssfed steer nd heifer beef production (NFBPQ) of 277, pounds below bse levels for Cow beef production (CBPQ) lso decreses n verge of

11 Roberts Trnsporttion Costs in Stte Models 195, pounds s rnchers, responding to incresed profit incentives, build the cow herd (BCI) by reducing the culling rte. The net result is 671, pound increse bove bse levels in Hwii beef production (TBPQ) for the period. A $6.5 verge increse in the Los Angeles utility cow price (LACBPRQ) results in n increse in the Hwii grss-fed beef price (HNFBPRQ) reltive to the Honolulu choice beef price (HGFBPRQ). Consequently, Hwii grin-fed beef production (GFBPQ) decreses by 214, pounds below bse levels nd grss-fed beef production (NFBPQ) increses 228, pounds bove bse levels for Cow beef production (CBPQ) decreses only slightly, nd the effect on totl beef production (TBPQ) is negligible (24, pound increse verged over the simultion period). Beef cow inventory is not ffected by the increses in the Los Angeles cow price becuse the Honolulu utility cow price ws found not to be significnt fctor in explining the size of the Hwii cow herd. Therefore, it ws excluded from the model's beef cow inventory eqution. The Hwii cttle feed price (HCFPRQ) increses n verge of $.5 bove bse levels when the Los Angeles corn price increses n verge of $.8. Thus, the feed price increses reltive to ll beef prices, cusing chnges in the composition of beef production. The signs of the verge impcts on grin-fed (GFBPQ) nd grss-fed (NFBPQ) beef production pper incorrect. However, the dynmics of the Hwii beef industry explin the result. When the feed price increses, Hwii rnchers reduce the size of the cow herd (BCI) by incresing the culling rte (CBPQ) nd reducing the replcement rte. The smller cow herd produces fewer feeder clves vilble to be plced on feed or grss. The number of feeder clves plced on grss decreses, ccounting for most of the reduction in the clf crop, nd the number plced on feed remins firly constnt. A possible explntion is tht few lrge rnches in the stte lso own the feedlot nd slughter fcilities on Ohu, giving them vested interest in mintining feedlot nd slughter volume. Compring the impcts of Model II with those of Model I revels the effects of excluding ocen freight rtes from the price trnsmission equtions. The verge impct for on the Honolulu choice beef price (HGFBPRQ), resulting from 1 percent increse in the Los Angeles choice steer price (LAGFBPRQ), is 14 percent greter for Model II thn Model I, nd the verge impct on the Hwii grss-fed beef price (HNFBPRQ) is 24 percent higher. The higher price impcts filter through the system, cusing lrger impcts on beef production nd cow inventory in Model II thn Model I. For exmple, in Model II the verge impct on grss-fed steer nd heifer beef production (NFBPQ) is 16 percent lrger thn in Model I, nd the verge impct on grin-fed beef production is 13 percent lrger. Similr events occur when the Los Angeles utility cow nd corn prices increse by 1 percent. When the Los Angeles utility cow price (LACPRQ) increses by 1 percent, the verge impcts in Model II on the Hwii grss-fed price (HNFBPRQ) nd the Honolulu utility cow price (HCPRQ) re 18 nd 19 percent lrger, respectively, thn in Model I. The exclusion of the feed freight rte results in 4 percent difference in the verge impcts on the Hwii cttle feed price (CFPQ) when the Los Angeles corn price (LACORNPQ) increses by 1 percent. The verge impcts from Model III, in which Los Angeles beef nd corn prices re used directly, re mrkedly different from those of Model I. Agin, the cuse is differences in model specifiction. Los Angeles prices re not the prices Hwii rnchers fce. Nor re they good proxies, becuse trnsporttion costs nd lgs in 13

12 July 1985 Western Journl of Agriculturl Economics TABLE 3. Vrible nd Symbol Definitions. Vrible or Symbol BBPQ BCI CBPQ CC CFPQ Cl CIQ D1Q D2Q D3Q DCI DLAGFCPRQ DM1Q DM2Q DM3Q GFBPQ HCPRQ HGFBPR HGFBPRQ HHBCR HHDCR HI HNFBPRQ LACFPI LACFPIQ LACORNPQ LACPRQ LAGFBPR LAGFBPRQ NFBPQ OHI OILP OILPQ RSOHIQ SI TBPQ TQ TRANBQ TRANFQ TSHBPQ TSOHIQ WQ R 2 Definition Vribles Bull beef production (dressed weight, 1, pounds). Beef cow inventory (Jnury 1, 1, hed). Cow beef production (dressed weight, 1, pounds). Clf crop (1, hed). Cttle feed price (pid by Hwii rnchers, $/1 pounds). Beef plus diry cow inventory (Jnury 1, 1, hed). Beef plus diry cow inventory (Jnury 1 inventory for ech qurter of the current yer, 1, hed). Equls 1 in the first qurter nd otherwise. Equls 1 in the second qurter nd otherwise. Equls 1 in the third qurter nd otherwise. Diry cow inventory (Jnury 1, 1, hed). Equls LAGFBPRQ - LACPRQ. Price freeze dummy, equls 1 for 1973 (11)-1973 (III). Pre-triler freight regultion dummy, equls 1 for 1976 (I)-1977 (II). Post-triler freight regultion dummy, equls 1 for 1978 (IV)-198 (IV). Grin-fed steer nd heifer beef production (dressed weight, 1, pounds). Honolulu cow price (wholesle, ll crcsses, utility, $/1 pounds). Honolulu grin-fed beef price (nnul verge of HGFBPRQ). Honolulu grin-fed beef price (wholesle, 5-9 pound crcsses, choice feedlot steers nd heifers, $/1 pounds). Heifers held for beef cow replcement (Jnury 1 inventory, 1, hed). Heifers held for diry cow replcement (Jnury 1 inventory, 1, hed). Heifer inventory (Jnury 1, 1, hed). Hwii grss-fed beef price (dressed weight, steers nd heifers, $/1 pounds). Los Angeles cttle feed price index (nnul verge of LACFPIQ). Los Angeles cttle feed price index (LACORNPQ converted to n index with 198 = 1). Los Angeles corn price (wholesle, $/1 pounds). Los Angeles cow price (wholesle, 35-7 pound crcsses, utility, $/1 pounds). Los Angeles grin-fed beef price (nnul verge of LAGFBPRQ). Los Angeles grin-fed beef price (wholesle, 6-7 pound crcsses, choice steers, $/1 pounds). Grss-fed steer nd heifer beef production (dressed weight, 1, pounds). Other heifer inventory, i.e., heifers not held for beef or diry cow replcement (Jnury 1, 1, hed). U.S. crude oil wholesle price index (nnul verge of OIIPQ). U.S. crude oil wholesle price index (1967 = 1.). Rtio of steer to other heifer inventory (Jnury 1 inventories for ech qurter of the current yer). Steer inventory (Jnury 1, 1, hed). Totl beef production (dressed weight, 1, pounds). Time, equls 1 in 197 (I) to 44 in 198 (IV). Cost of trnsporting beef from the U.S. West Cost to Hwii in continers ($/1 pounds). Cost of trnsporting niml feeds nd feed ingredients from the U.S. West Cost to Hwii in continers ($/ton). Totl steer nd heifer beef production (dressed weight, 1, pounds). Steer plus other heifer inventory (Jnury 1 inventories for ech qurter of the current yer, 1, hed). Wether dummy, equls 1 in qurters when droughts occurred. Other Symbols One minus the rtio of the sum of squres residul to the sum of squres totl (clculted from untrnsformed dt for utoregressive equtions). 14

13 Roberts Trnsporttion Costs in Stte Models TABLE 3. Continued. Vrible or Symbol Definition DW Durbin-Wtson sttistic. DH Durbin h sttistic. OLS Ordinry lest squres. AUT Autoregression procedure (Cochrne-Orcutt or Grid Serch). ip Estimted first order utoregressive prmeter. Q t the end of vrible nme denotes qurterly observtions. All other vribles re nnul. price trnsmission re not considered. Also, the Los Angeles utility cow price is used in the grin- nd grss-fed beef production equtions s proxy for the return to producing grss-fed steers nd heifers. Under this specifiction, Model III overestimtes the bsolute impcts from n increse in the Los Angeles choice steer price becuse no offsetting increse in the Hwii grss-fed beef price exists tht is nlogous to the increses implicit in Models I nd II. Likewise, when the Los Angeles utility cow price increses, the effects re lrger in Model III thn if Equtions 3A nd 3B were used for price trnsmission. Differences in verge impcts over the simultion period cused by excluding trnsporttion costs from the Minlnd-to- Hwii price trnsmission equtions seem substntil. Those differences re even lrger when the price trnsmission equtions re excluded nd Los Angeles prices re used directly. 5 These findings cst serious doubt on the relibility of Models II nd III in evluting how chnges in ntionl griculturl policies concerning Minlnd beef nd feed grin prices ffect the Hwii beef industry. The relibility of such models in evluting the impcts of chnges in stte-level policy instruments lso would be questionble. 5 It is importnt to remember tht differences in verge impcts re not relted to how ccurtely ech model explins historicl dt. Rther, they re direct result of differences in the mgnitudes of the estimted coefficients cused by the omission of freight rtes or the substitution of Los Angeles prices for Honolulu prices. Augmented Policy Anlysis Potentil The results of two dditionl simultions of Model I re presented in Tble 6 to demonstrte the model's dded potentil for policy nlysis if trnsporttion costs re included in price trnsmission equtions. A 1 percent increse in beef freight rtes (TRANBQ) bove ctul levels for ech yer between 1972 nd 198 cuses ll Hwii beef prices to increse. However, both the Hwii grssfed steer nd heifer price (HNFBPRQ) nd the Honolulu utility cow price (HCPRQ) increse reltive to the Honolulu choice steer nd heifer price (HGF BPRQ). Consequently, grin-fed beef production (GFBPQ) decreses in , while grss-fed beef production TABLE 4. Theil U 2 Coefficients for Models I- III, 1972 (I)-198 (IV). Model I Model II Model III HGFBPRQ HNFBPRQ HCPRQ CFPQ GFBPQ NFBPQ TSHBPQ CBPQ BBPQ TBPQ BCI Cl HI OHI HHBCR SI CC

14 July 1985 Western Journl of Agriculturl Economics TABLE 5. Averge Simulted Impcts for Models I-III Resulting from 1 Percent Increse in One Los Angeles Price, Holding Others t Historicl Levels. Model II Model III Percent- Percent Model I ge ge Historicl Chnge Chnge Averge Impct Impct from Impct from Vrible (units)b (units)b (units)b Model I (units)b Model I Increse LAGFBPRQ 1% ($8.1) HFGBPRQ HNFBPRQ GFBPQ 18,657 1,59 1, , NFBPQ 5, , CBPQ 5, TBPQ 3, BCI Increse LACPRQ 1% ($6.5) HNFBPRQ HCPRQ GFBPQ 18, NFBPQ 5, CBPQ 5, TBPQ 3, BCI Increse LACORNPQ 1% ($.8) CFPQ GFBPQ 18, NFBPQ 5, CBPQ 5, TBPQ 3, BCI Vribles re defined in Tble 3. bthe units re dollrs per hundred pounds for HGFBPRQ, HNFBPRQ, HCPRQ, pounds for GFBPQ, NFBPQ, CBPQ, nd TBPQ, nd thousnds of hed for BCI. nd CFPQ, thousnds of (NFBPQ) increses. After 1975, the impct on GFBPQ becomes positive, nd the positive impct on NFBPQ becomes lrger s the incresing cow herd (BCI) produces more clves. The cow herd grows with decrese in the culling rte, s indicted by the decline in cow beef production (CBPQ). The impct on totl beef production (TBPQ) is negtive in the first two yers but reches level of 1.46 percent in 198, reflecting the increse in the clf crop. A 1 percent increse in the cost of trnsporting corn to Hwii cuses the Hwii feed price (PCFQ) to increse reltive to ll beef prices. As result, cows re culled t fster rte, cusing cow 16 beef production (CBPQ) to increse nd beef cow inventory (BCI) to decrese. The decrese in BCI eventully cuses reduction in the clf crop. As the number of feeder clves decreses, the production of grss-fed steer nd heifer beef (NFBPQ) decreses more thn production of grinfed beef (GFBPQ), reflecting the vested interest of rnchers in mintining feedlot nd slughter volume. The importnce of these simultions is tht the impcts result from chnges in Los Angeles-to-Honolulu trnsporttion costs. Such n nlysis would hve been difficult if Los Angeles prices hd been used directly or if trnsporttion costs hd been excluded from the price trnsmission

15 Roberts equtions. These simultions demonstrte the model's potentil usefulness to beef producers, stte policymkers, nd others interested in the effects of trnsporttion costs on the Hwii beef industry. For exmple, this model could esily be modified to evlute the possible consequences of freight rte increses proposed by mjor freight crriers. Also, the impcts of deregultion could be simulted under vrious ssumptions bout freight rte djustments resulting from such ction. An exmple of this type of nlysis ws done by Roberts et l. who evluted the impcts of energy price increses on the Hwii beef industry. Summry nd Conclusions This study demonstrtes tht trnsporttion costs re importnt in determining beef nd feed prices in Hwii. Beef nd feed trnsporttion costs vribles re highly significnt when used in conjunction with Los Angeles beef nd corn prices in Minlnd-to-Hwii price trnsmission equtions. Becuse of their importnce in price trnsmission nd their high positive correltion with Los Angeles beef nd feed prices, exclusion of trnsporttion costs leds to positive bis in the Los Angeles price coefficients. Lrger coefficients yield lrger bsolute impcts, putting in question the usefulness of such model (Model II) for policy impct nlysis. The mgnitudes of the simulted impcts increse even further when price trnsmission equtions re eliminted nd Los Angeles prices re used (Model III) rther thn Hwii prices. The inclusion of freight rtes nd Hwii beef nd feed prices in the Hwii beef model (Model I) is not purported to eliminte ll specifiction bis. Obviously, the unvilbility of certin trnsporttion cost vribles, nd other dt limittions, restrict the model's structure. However, in the cse of the Hwii beef industry, more ppropritely specified Trnsporttion Costs in Stte Models price trnsmission equtions improve the ccurcy of the model nd confidence in its results. The usefulness of the model is lso enhnced s the number of exogenous vribles is incresed to include freight rtes. Thus, by including trnsporttion cost vribles, chnges in trnsporttion policy or proposed rte chnges by mjor crriers could be evluted. Dt limittions constrin specifiction nd estimtion of most econometric models. Therefore, the results presented here should be qulified by recognizing tht Model I is not without error nd tht Models II nd III were estimted ccording to different criteri thn Model I. Model I ws specified ccording to theory, but respecified nd estimted with n cceptble structure tht provided good fit to the limited dt. On the other hnd, Models II nd III were specified nd estimted with the sme structure nd sttisticl techniques s Model I, except for the deletion nd substitution of certin vribles. Therefore, the differences in impcts presented here should be interpreted s prtil results becuse they show differences cused by the deletion of trnsporttion cost vribles or the substitution of Los Angeles prices for Honolulu prices, holding model structure nd estimtion techniques constnt. If the models hd been specified nd estimted independently, the totl difference in impcts would hve been the difference resulting from deletion or substitution of certin vribles plus the difference resulting from chnges in model structure nd estimtion techniques. Independent specifiction nd estimtion probbly would hve produced models fitting the dt better thn Models II nd III. It is unlikely, however, tht such models would hve performed better thn Model I given the exclusion of trnsporttion cost vribles tht hve been shown to be significnt determinnts of Hwii beef nd feed prices. Notwithstnding these qulifictions nd the specificity of the results to the 17

16 July 1985 Western Journl of Agriculturl Economics E ) ( (U ) N r CC CD O LOJO C- C il C- C 6 CD O C Cm o _ om : ci c_ n _ CM ) - CD N t 1 'It C In - CM.) ) L. (U IL ) ) - rnd ) CO D C N V N o -_ C%1 C CDO CJ C )O ')' i. _ t t ) N ) CD t _- CJ CMJ _.- :.- : ci -.C.- CM ) C )oo r I I I N CO L) O _ C') C N C) s C CM o o cl t ci CM ) CD CO D t O O C O C) - C') co~i~incqcdcd~inco~o Co ooc lq q t co t O() Ci N f l O OM O IO CMO CM O CM ) CD CD o S CSJ OCOO<O'rtOC~i Oi '- CV) ) O ) m CI U) CD N) ) N ) C) C) CD Co) ' C - r-o oi',l', cd n * CD T- LCO C(D C) CD O CD COCDoNlc C D N ) CD c t Ct t Moq ) CN ur- CO _* CDO _ C _It O CD CD Wo C CD o C cmco coddddd*dtjddd ) I-- ) r. T CO " t ) O QV) CD D LO CD C ) CM - I- I II I o) i c ) o 2 c c c) C' CMC '- CJ O CJ) O O O CV O O _ O '-LJCc tt c oco c J C.. )..O. O... O. _ O O If IM IM I _. cm V) CD CD CD Cd CD N CD '- C) _ ) O O IC m m m wtc cc m o m o Em m m m. Er.. o. IL C ( LL LL - E m Co CL cl -L LL CL M CO EL CL. Co Co C. - ( z O(.LLL Co m Oz o LL LC m tflo LL L Co oom IzzOO (3 z O m IIM( zz O ZmO~ m( O Z UFm mt ir cn o E CD, ) o m OC U) o C I QCI) HC* c r U) o <1 H.. 18

17 Roberts Trnsporttion Costs in Stte Models Hwii beef industry, the results suggest tht reserchers be cutious in using ntionl rther thn stte price vribles in stte commodity models. When stte prices re exogenously determined by ntionl or mjor regionl mrket prices, it is n empiricl question whether trnsporttion costs re importnt in price trnsmission. These results might encourge other stte econometric modelers to try to improve the ccurcy nd usefulness of their griculturl models by including trnsporttion cost vribles in their models where pproprite. References Bum, K., A. N. Sfyurtlu, nd W. Purcell. "Anlyzing the Economic Impct of Ntionl Beef Import Level Chnges in the Virgini Beef nd Pork Sectors." Southern Journl of Agriculturl Economics, 13(December 1981): Betncourt, R. nd H. Kelejin. "Lgged Endogenous Vribles nd the Cochrne-Orcutt Procedure." Econometric, 49(July 1981): Cliforni Federl-Stte Mrket News Service. Livestock nd Met Prices nd Receipts t Certin Cliforni nd Western Are Mrkets. Cliforni Deprtment of Food nd Agriculture, Scrmento, vrious issues, Grrod, P. V. Interislnd Ocen Freight Services in Hwii, Deprtmentl Pper No. 48, University of Hwii Agriculturl Experiment Sttion, Hwii Agriculturl Reporting Service. Sttistics of Hwiin Agriculture. Hwii Deprtment of Agriculture, Honolulu, vrious issues, Hwii Mrket News Service. Honolulu Prices: Wholesle Eggs, Poultry, Pork, Beef nd Rice. Hwii Deprtment of Agriculture, Honolulu, vrious issues, Johnston, J. Econometric Methods. 2nd edition, McGrw-Hill, New York, Kment, J. Elements of Econometrics. Mcmilln, New York, Knpp, J. L., T. W. Fields, nd R. T. Jerome, Jr. A Survey of Stte nd Regionl Econometric Models. Virgini Tyloe Murphy Institute, Chrlotte, Leuthold, R. M. "On the Use of Theil's Inequlity Coefficients." Americn Journl of Agriculturl Economics, 57(1975): Mtson Nvigtion Compny. Triffs 14-B through 14-G. Honolulu, Hwii. Nerlove, M. "Distributed Lgs nd the Estimtion of Long-Run Supply nd Demnd Elsticities: Theoreticl Considertions." Journl of Frm Economics, 4(1958): Roberts, R. K., G. R. Vieth, nd J. C. Noln, Jr. "An Anlysis of the Impct of Energy Price Escltions During the 197's on Hwii Beef Production nd Prices." Western Journl of Agriculturl Economics, 9/1(1984): Schermerhorn, R. W., P. V. Grrod, nd C. T. K. Ching. A Description of the Mrket Orgniztion of the Hwii Beef Cttle Industry. Informtion Text Series No. 11, Hwii Institute of Tropicl Agriculture nd Humn Resources, Simpson, J. An Assessment of the United Sttes Met Import Act of Economic Informtion Report No. 152, Food nd Resource Economics Deprtment, University of Florid, Wllis, K. F. "Testing for Fourth-Order Autocorreltion in Qurterly Regression Equtions." Econometric, 4(July 1972): White, K. J. "A Generl Computer Progrm for Econometric Methods-SHAZAM." Econometric, 46(Jnury 1978):