CAN 1 _. '0 O· G I ULTURAL ECO OMICr. '1); LID ARY J~~~6 '1989 ~ Staff Pap-ers DEPARTMENT OF AGRICULTURAL ECONOMIC.§ , !yniversity of Kentucky • CoUege of Agriculture Lexington 40546 October 1988 DElWfD STUILITY rol HlATS: In AlOTHIR APPROACH RI\OZ&\ h.luAGen and Barry if. Bobst. Paper prosentlld At tbe .eting of the SOuthern baionel rood Demand Reaearch Teobnical Coslttee f K1nneapoU.• t October 3-.lf, 1988 statt Papera are ~~bli8b.d without toraal review. Vieva expressed are tho•• ot tbe author' aDd do not ~rloct the views ot the Un1veralty of Kentuck)" tbe Agricultural Experilllent Station, or the Cooperative Extenaion Service .. DEMAND STABILITY FOR KEAts: YET ANOTItER APPROACH by Reza Najar%4deh and Barry V. RobBt The agricultural economics literature is being lnundatedwlth demand meat hypothesis. studies concerned with the structural stability The question 18 whether consumer demand for meat. h .. shifted in recent years in favor of white ~ats (poultry and fish) at the expense of red meats (beef. veal, pork, and lamb)? Just this year at l~a.t three papers have been published on the subject and have come down on different sides of it. Evidence in favor of structural stability has been presented by Chalfont and Alston. Papers indicating different aspects of structural change hAve appeared from Eales and Unnevehr and from Dahlgran. Many. If not most, of the pre·198a studies are cited in these papers. The reviews show that respectable economists with respectable studies have come down on both sides of this issue. We hope our efforts will be regarded a$ aqual1y respectable. Whether they welcome us or not, our results place um in the structural stability camp. by a different Furthermore, we come to our conclusions train of reasoning concerning stability of meat demand. the structural In fact, our original goal was not the structural issue so much as it was to extend the Box-Cox functional form analysis conducted by Chang from the composite commodlcy "meat" to a jo1nt analysis of its constituents. In doing so we us6d a different data set from a different time period and used a maximum likelihood criterion function somewhat different from the one Chang ueed to det.rains vnat functionsl form b~~t fit his a®at deBand However, our results lndicated " f'.metlonal font that wu data. ••••ntially the 8 . . &S Chan,'.. Reproducibility of experiments 1. a hallaark of the natural seiencllla. To be considered valid, result. lIWIt be reproducible by OthOE'lI. Stability of un4erlylng phenoMnA 1s •• lcloa at iS8\14 in the natural Klenc•• , where.s it i. precis.,ly the 1.s. hero. Cb.ul&' •••• to POl.' WI this reason, the congruence of our re.ulu with to be -. atrona evidence of structural at&bility in th. demand for Deata, CIuma's study Chang uaed the lox-Cox nexible functional form to aMly:e per capita . a t dtaund on armual data for 1935·1974, excluding the 1942- 1947 val.' years. Hia model .pecified rod ..at (baef. v ••l, pork. laJab, and 8Utton) .s & functJ.on of the deflated retail price index of red &eats and deflated per capita income .. follows: Where the exponentiation notation is, for any variable X, (2) X{k) .... (Xk - l)/k. The likelihood function asaociate with this modal 1. where 82 1~ the e.timated regression variance for the ROdel liven value of k. a An iterative regression procedure vas used to auimize the l1kelihoocl function. obtained when k - -0.84. Th. CUrrent Study 4t The lUXi9!W!1 of thla function 11&8 The current gtudy applie. the Box~Cox flexible fora approach to the analysis of joint dea.ndG for ••veral ..ats. the object is to <Ii. aggregate Cbang'iI analy.1s. Essentially. The analyst. deals with jOint. qu.rterly demands for four seats: beef, pork. chicken, The aod$l 1••at_ted on quarterly data for the 1964- and turkey. 84 period. Full re.ults are pre••nted in Najarzadeh' 8 dis.ertation. Tho "8W1ption underlying joint .stiMation ia that d.euncl for . .au is govern.d by a branch of a weakly ••tt~z:able utility function . . who.. fOIill, though not directly •• t1llable. deterain.. the characteristic exponent k in the demand equations that derive from thi. utility function branch. Det.raining the likelihood function maxiaIz1ng value of k. therefore, provides an eaplrical estimate of the functional form of the deeand equations which best confora to the underlying utIlity function. Tbis argument lIU uaed by Gerkina and 84y.s in their analysb of housing demand. . An adaptation of the single··equation likelihood function used 1n previous studies was 116cesaary to aceoUlJlOdate our 1N1tlple· product approach. Hean expenditure weights were used to weight the likelihood function, as follow.: where Wi - expenditure weight for the lth •552 weight. were Pork: .309 Chicken: .109 Turkey: .030. 3 ~at. These e~p.nditur• Per capita demand functions were specified as functioN of deflated prices. deflated per capita disposable inca.... and "easonal dummies with the firat quarter deleted. These functions had the following general form: (5) ~ for the ith meat. Quantities, prices, and ~, ... lnco~ are exponentiated by & common value of k for all meats. An iterative estimation process was used. in which the regression DIOdel in equation (5) was estimated for alternative values of k to determine the value which maximized the likelihood function in equation (4). Estimation precision vas carried to 2 decimal places in k. The l1ke1ihood-maxiaizing value for k was found to be ·0.91 for the four-meat model over the 1964-84 sample period. e.timat~s of the demand aodel parameters at k - ·0.91 are shown in Table 1. Slutsky-constrained estimates vere UnconStrained also obtained. but are not shown here. A test of the restrictions could not reject the hypothesis that the data were consistent with the existence of the Slutsky conditions. The nonlinear data transformation makes it difficult to interpret coeff1cients directly. so elasticity measures are even more useful than usual. Elasticities computed at sample means are shown in Table 2 for all parameters with t-ratios exceeding +/- 1.0. All owu prlce coefficients meet thh criterion, although the t-ratio N for turkey price 1s quite low. Ineome 4 e~astlcitie8 are le88 than unitary for beef and pork, 4na more than unitary fQr chicken and turkey. Cross-price elasticities are positive in all but one case and, except for that case, indicate plausible competitive relationships. Cross-price elasticities for beef price in the pork. chicken, and Cross -price elasticities turkey deJlUlnd function. are positive. between beef and pork are nearly the SaDe, indicating approxiaata .,,-.etry between these _ate. Crosu-p:dce red seats and the poultry meats are ~t ~.la.tlcitie. symmetrical. between the Chicken and turkey dell4nds are affected by beef prices. while beef delllWld appears to be respoNlive only to turkey prices. Neither chicken nor turkey are responsive to pork prices, but pork demand responds to turkey prIces. These cross-price relationships of beef and pork demand with turkey price is consistent with riSing competition of processed turkey products vith processed red Hat products. No such convenient explanation exists for the relationships of beef and pork demand with chicken price. Beef deil4nd is indicated to be independent of chicken price, and pork demand Is shown to be weakly comple_ntary. The seasonal demand shifters seea reasonable. The weak seasonality of beef demand and the increase in chicken demand in the spring and sUIIl1JI.er qua.rters are borne out by simple inspection of the data. LikeWise, the strong fourth quarter shifts in demand for pork and turkey are as one would expect. The Structural Stability Issue Similarity of this estimate of the likelibood function· 5 maximizing funetional form with Chang's previous work has already been .entioned, but some issues about the preBent &stimates must be answered in order for this comparison to be taken serIously. First, 1s the aodel ignoring a structural change in deaand from red meats to poultry taking plece during the study period? Second, doe. the rather poor fit for beef dema.nd suggest mis!Specification of th. DIOde!? Theae issues are examined by aean. o~ analy.1. of goodne •• - of-fit of the model over sequential subperiods. are obtained for each of 15 sequential subperiods beginning with Predicted values 7-year (28 quarter) and ending with 19;8-84. 1964~10 Mean square errors are calculated for each subper10d and compared. Since comparative values rather than absolute numbers are important here, Table 3 displays these goodness-of-fit result. as rankings of the mean square errors. hnkings are in descending order, with the highest aean square errors ranked number one. Ideally. these rBnkings should appear in no particular order. indicating no change in the model's predictive capability over the sample period. Inspection of the mean square error rankings for beef indicate a clustering of ranks rather than a random ordering. A sequence of rising mean square errors occurs from the 1967-73 period through 1971-77. They decline after 1971, and in fact the model seems to fit best as it moves into the 19808. These results do not suggest any deterioration of the model' 8 predictive abUity during the sample period, but they do mirror the economic disturbances of the early 19708. The rankings for pork appear to be the most randomly ordered 6 of any of the meats. No ••qu0nce. or ether orderings ere apparent. If the chicken deund function has a improve in pr~dict:ive tr~nd. it seou to be to capability over the swaple period. The three poorest fits Appear 1n the first three subperiods. Turkey. on the other hand. structural change. display. ela.sie syaptou of Tho poore.t fits occur in the early and late .ubperiods, and the beat fits are clustered in the aiddle, when. presumably, dependent and independent their 8aurple Mans. v.ri~bl&. vere clos.st to This pattern indicate. a structural chanao 1. occurring that 1. not accounted for by the model'. specification. However. the ujor component of the change appear. to ~ in seasonalIty rather than in the economic variable relationship•. Seasonal variation in turkey -consumption is declining over tiae, but this is not taken into account in tho current specification. Year .. to-year changes in quantities deaanded ••em to be vell accounted for by the price and income components of the specification. Concluding Remarks It is well-known that it i5 difficult to validate a structural specificatlon on the same data set uaed to estimate ita parameters. For this reason, our conclu.ion that the structure of meat demand 1. stable, with the exception of seasonality in turkey, may not be wholly persuasive. That is why the agreement with Chang's results sre so interesting. If one accepts the argument that the functional fom of demand equations reflects the form of their underlying utility function (which 11 done routinely in the clUe of linear expenditure functions). then the inference of preference stability 7 follow. fro. the apparent atabillty of functional fora. 8 Chalfont, J. A., and J. M. Alston. "Accounting for Chana.. in T•• tell." Jouma1qf I'QUtlcal !1;llDOJU. 96 (April 1988) :391-340. ChanS. H. S. "FUnctional Forms and the De.and for Meat." Royi•• of i&onomlcs And StAtiltie •• 59(Auguat 1977):355-359. Dahlsran, R. A. "Changing Meat Derund Structure in the United State.: Evidence froll a Price Flexibility Analys! •• " Barm Central JQYmAl of Agricultynl Economics. 10(July 1988) :165- 176. lal•• , J. S.. and L. J. Unnevehr. "Deund for Beaf and Chicken Products: Separab:Uity Structural and Change. " _"iCAU JRurnAl of Asriculty[al iQonOlicl' 70(August 1988):521-532. Ger'klng, S. D. , and W. J. Bayes. "The lole of Functional Form in Estimating Elaaticitt•• of Housing Expenditure•• " Southern Icggpmic JourDA1. 47(1980):287-302. Najarzadeb. ReZ4. "Functional Foma and Deund.e for Different Type. of Keats in the United States. 1t Unpublished Dissertation. University of Kentucky, 1981. 9 Ph.D. ~ ~ff1ct.nt UccOlP8trd.DMl Tab10 1. Ie .. "'~' 91 btlutela at for hef t Pork, Chlekeft, &1.lCl 'l'urby. Pork Ie.f. Chlcbn 'luray --1305." liltercept -71.00 .. 61.2' ..'2:U.05 IMf Price .. 1.81 (-5.40) 1.12 (3.03) 1.70 (2.74) 25.69 (1.78) 0.29 .. 0.04 (·0.01) 0.54 ·3.26 (2.71) (-14.7) . (0. 'U). 0.03 (0.18) .. 0.23 (·1.50) .. 1.61 .. 2.56 (¥6.30) (-0.44) hrk Prlc. Cblcbn Price 0.56 1.07 .. 0.11 (2.99) (5.13) (-0031) .. 10.56 (-1.32) 66.27 (6.82) 57.99 (5.39) 210.91 (11.8) l1i6.8l (2.85) Turkey hiea lccOlH quartet' 2 -0.0003 (-0.34) Quutsr 3 0.0018 (1.95) Qu,artlltll;' 4 0.0007 Is Likelihood-Value -r-r4t1~. .. 0.0031 (·3.15) -0.0045 (~4.4S) (O.72) 0.003' (3.18) 0." 0.S7 21.469 are in parenthos4e. 10 0.0109 (6.58) 0.0128 0.1554 (4.07) 0.4444 (7.65) (11.6) -0.0001 ( .. 0.47) (11.6) 0.96 0.11 0.6868 Tabl. 2. hie. am lneoae Ilutlclty IiIt:tuto. at Sapt. . . . . leef Pork Chicken Turkey leef Price ... 49 .23 .24 .94 hrk Pric. .19 ...13' -.. -.11 ....52 ._ .. ... -.74 Cb1cJr,eD mc. Turkey hie. .29 .42 IDCoae . 70 .46 11 1.16 ...,.. ,. ~ 1.69 Table 3. Ranks of Prediction Error Mean Squares for Beof. Pork. Chicken. and Turkey Demand in Fifteen Subs ample Periods. Pork Chicken Turkey 10 8 1 1 1965-71 9 9 2 2 3 1966-72 7 15 3 5 4 1967·73 5 13 8 8 5 1968-74 4 11 5 10 6 1969-75 ) 6 4 13 7 1970-76 2 4 14 12 8 1971·77 1 7 9 15 9 1972·78 6 5 13 14 10 1973-79 8 1 12 11 11 1974-80 11 3 7 9 12 1975·81 13 10 6 7 13 1976~82 14 14 15 6 14 1977-83 15 12 10 4 15 1978-84 12 2 11 3 Period Years 1 1964-10 2 Beef 12 ... ",eo .., The College of Agriculture Is an Equal Opportunity OrganizatIon wIth respect to education and employment and authorization to provide research. education Information and other services only to IndIviduals and Institutions that function without regard to race. color. national origin. sex. religIon. age and handicap. Inquiries regarding compllanes with Title Viand TItle VII of the Civil Right Act of 1964. Title IX ofthe EducatIonal Amendments. Section 504 of the Rehabilitation Act and other related matter should be directed to Equal Opportunity OffIce. Col/ege of Agriculture. Umverslty of Kentucky. Room S-105. Agricultural SclenclI Building-North. Lexington. Kentucky 40546.