Agricultural & Applied Economics Association

The Farm-Retail Price Spread in a Competitive Food Industry Author(s): Bruce L. Gardner Reviewed work(s): Source: American Journal of Agricultural Economics, Vol. 57, No. 3 (Aug., 1975), pp. 399-409 Published by: Oxford University Press on behalf of the Agricultural & Applied Economics Association Stable URL: http://www.jstor.org/stable/1238402 . Accessed: 13/12/2011 03:13
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Agricultural & Applied Economics Association and Oxford University Press are collaborating with JSTOR to digitize, preserve and extend access to American Journal of Agricultural Economics.

http://www.jstor.org

Articles

The

Farm-Retail Price

Food Competitive
Bruce L. Gardner

Spread Industry

in

Consistency with market equilibrium places constraints on the pricing policies of food marketing firms in a competitive industry. This paper examines the implications of simultaneous equilibrium in three related markets: retail food, farm output, and marketing services. From equations representing the demand and supply sides of each market, elasticities are generated which show how the farm-retail price spread changes when retail food demand, farm product supply, or the supply function of marketing services shifts. Implications for the viability of simple markup pricing rules and the determinants of the farmer's share of the food dollar are discussed. Key words: farm-retail price spread, marketing margin, market equilibrium, competition.

This study examines the consequences of competitive equilibriumin product and factor marketsfor the relationshipbetween farm and retail food prices. The investigation is based on a one-product,two-inputmodel developed by Allen and Hicks and since appliedto many issues at the industry level. Notable agriculturalexamples are the papers of Brandowand Floyd. The model is used in this paper to generate quantifiablepredictions about how various shifts in the demandfor and supply of food will affect the retail-farmprice ratio and the farmer'sshare of retailfood expenditures. The results have implicationsfor the viability of simplerules of markuppricingby marketing firms. In general, the markup must change whenever demandor supply shifts in order to be compatible with market equilibrium. Moreover, the markup will be forced to change in different ways depending on whether price movements originatefrom the retail demand or farm supply side. Related implicationsconcern the consequences of retail price ceilings and farm price floors, the elasticity of price transmission, and the determinantsof changes in the farmer'sshare of the food dollar.
BruceL. Gardner an associateprofessorof economicsat North is CarolinaState University. The authoracknowledgeshelp and criticismfrom Ronald A. J. Schrimper, B. Bullock, GeraldA. Carlson,JohnIkerd,PaulR. Johnson, E. C. Pasour, and the Journal reviewers.

The Model

Considera competitivefood marketingindustry using two factors of production,purchased commodities(a) and other marketagricultural ing inputs (b), to produce food sold at retail (x). The marketing industry'sproductionfunction is
(1) x = f(a,b).

It is assumed to yield constant returns to scale. The retail food demandfunction is

(2) x = D(Px, N),

where Px is the retail price of food and N is an arbitraryexogenous demand shifter which for purposes of specificity will be called population. The model is completedby equationsrepresenting the markets for b and a. On the demand side, firms are assumed to want to buy the profit-maximizing quantities of b and a, which implies that value of marginalproduct equals price for both
(3)
Pb = Px

fb

and
(4) Pa = Px 'fa,

wherefb andfa are the partialderivatives of x with respect to b and a.

400 August 1975

Amer. J. Agr. Econ.

The input supply equations are

(5) Pb = g(b,T),

(8)

dfa _ dN

da dN

db dN fab

Substitutingequation (8) into (7) yields the supply functionof b to the food marketing P da industry, and (9) ha da
(6)
Pa = h(a,W),

dN

dN
+

the supplyfunctionof agricultural output.The shifters of marketing input and exogenous farmproductsupply are representedby T and W. For purposes of specificity, W may be thought of as a weather variable for which higher values increase Pa (e.g., an index of drought),and T as a specific tax on marketing inputs which makes them all more expensive. This system contains six equations in six endogenous variables(x,b,a,PX,Pb,Pa). Under normal conditions (where the demand function has negative and the supply functions have nonnegative slopes), there will be a unique equilibrium for given values of the exogenous variables. At this equilibrium,the values of the six endogenous variables, and hence the farm-retailspread, are determined. This price spread may be measured by the difference between the retail and farm price,
Px - Pa, by the ratio of the prices, Px/Pa, by

Prf

db dP + A dN +fa dN

Next, analyze the b market by combining equations (3) and (5) and differentiating:
db (10) gb dN 60dNV db fbb dd N
+ Pxf

da dN

dPx dN

Equation (10) holds the b market in equilibrium while the relationshipbetween dPx and dPa is examined. Similarly,the thirdequation specifies equilibriumin the x market by differentiatingequations (1) and (2) combined:
()f da db dN +fb dN D x dP+D dN DN.

the farmer'sshare of the food dollar,aPa/xPx, or by the percentagemarketingmargin,(Px Pa)/Pa. This paper focuses on the retail-farm price ratio, the closely related percentage (12) farmer's share of retail food expenditures. Effect of a Food DemandShift on the Retail-FarmPrice Ratio The effects of a shift in retail demandon market equilibrium analyzedby differentiating are to (6) with respect to N, while W equations(1) and T are held constant. The six equationscan be immediatelyreducedto three (one equation for the final product marketand one for each input)by equating(1) and (2) to eliminatex, (3)
(13) margin, Px/Pa 1, and the

Equations(9)-(11)can be solved for da/dN, dbldN, and dPx/dN. The solution is made more intelligibleby convertingall derivatives to elasticities. Details of the necessary manipulationsare presentedin an appendix.The result is the three equation system:
0= _( S; Ea
+
0-

+ EbNg EpXN

Sa

EaN
,

aeb

and
(14) 7N= Sa EaN + Sb EbN - l EPXN.

Equation (12) pertains to the market for a, (13) to b, and (14) to x; Sa and Sb are the relative shares of a and b, e.g., Sa = aPa/xPx; and (5) to eliminate Pb, and (4) and (6) to oa is the elasticity of substitution between a and b; -r is the price elasticity of demandfor eliminate Pa. with the marketfor a, equations x; ea and eb are the own price elasticities of Beginning supply of a and b; r_) is the elasticity of de(4) and (6), the new equation is
(7) h

a _ df dP. dN LdN fadN" The dfa term of equation (7) must be expanded further. It is not simply the second partial derivative ofx with respect to a (which will be written faa). It also brings in the amount of b that a has to work with as

mand for x with respect to N; and EaN, EbN, and EPN are total elasticities which tell how the first subscripted variable responds to a change in the second.'
The capital E's are elasticities which take into account equilibrating adjustments in all three markets simultaneously; ea, eb, and rT are partial elasticities which refer to movements along the input supply and product demand functions. All the elasticities

Gardner

The Farm-RetailPrice Spread 401

The question to be investigated is how uct and bread as the retail product, what is PIPa changes when the demand for food the probable sign of equation (15)? Since shifts. The answer can be expressed as the wheat is a specific factor to the x industry, elasticity of PI/P' with respect to N. This while the components of b (labor, transportaelasticity is equal to the difference between tion, packaging, etc.) generally are not, and EPN and EPaN both of which can be obtained since a is land intensive, it seems likely that ea from the system of equations (12) to (14). The < eb. In this case, when the demand for food shifts to the right, Px/Pa falls. Therefore, the result (derived in the appendix) is retail-farmprice ratio is expected to decline - eb) when population(or any other exogenous food = (15) -NuSb(ea EP, demand shifter) increases. An interestingspecial case arises when ea = where D is a function of a-, r, ea, eb, and Sa.2 eb. In this case Px/Pa is unchanged when the The denominator has no intuitively clear demandfor food shifts. Thus, a fixed percenmeaningbut is positive in all normalcases (-r tage markuprule used by marketingfirms is < 0 and ea and eb > 0). Therefore, the viable in the sense that competitiveforces will numerator normally determines the sign of not requirethe markupto change when retail equation (15). 4 Because of the way the originalmodel was food demandshifts. In general,however, ea and a fixed percentage markupwill not be constructed,equation(15)will be more readily eb viable in this sense. applicable to some situations than others. In reality, of course, there are many marketing Equation (15) also helps in understanding activities and many marketing inputs. The the role of r-,the elasticity of substitutionbeindustry. Suppresent model assumes that these can all be tween a and b in the marketing into a single productionfunc- pose N increases, and ea < eb. Then the price lumpedtogether tion with a single marketinginput, b. Follow- of raw farm product relative to marketingining the usual requirementsfor aggregation, puts increases, creatingan incentive to substithis assumption should cause no analytical tute the latter for the former. In the wheat difficultiesso long as the relative prices of the example, additionallabor may be used to recomponentsof b are constant. Thus, equation duce grain wastage in processing operations, how shifts and the use of pest and spoilage control may (15) will be helpfulin understanding in food demand affect agriculturalproduct increase.4 However, in many marketingconprices relative to all marketing inputs as a texts the opportunitiesfor substitutionappear group, but will not be helpful in situations limited. This would be reflected in equation where substantial relative price changes (15) by a small value of a-. Since a- appears withinthe set of marketing inputsare induced. only in the denominatorand with a positive There may also be an aggregationproblem sign, the smaller a- is the more volatile the with the quantity of retail food, x, depending retail-farmprice ratio. The economic reason for this result can be on the context in which the model is applied. illustratedwith referenceto an increase in reIf x is taken to be an aggregateof all food, it must be assumedthat the relativeprices of the tail demand for food. The demand shift invarious food products are held constant. creases the derived demand for both farm inputs used Thus, the exogenous shift in demand should products and the nonagricultural in the food marketingindustry. But so long as be thought of as one applying to all forms of the two elasticities of supply are different (ea food. On the other hand, if the context in which # eb), their relative prices must change. How the model is applied is a relatively narrowly muchPa/Pb will change depends on the degree defined product, say, wheat, the aggregation to which a and b can be substituted in the problems for both x and b may be less serious.3 For a case like wheat as the farm prodare partial in the sense that the exogenous variables T and W are held constant. 2 D = -,q(Sbea + Saeb + a) + eaeb + oT(Saea + Sbeb). 3 The model applied separately to each of a set of narrowly defined retail products would still have implications for the average or aggregate farm-retail price spread. For example, a shift in demand towards relatively b-intensive products (such as TV dinners) would reduce the aggregate farmer's share even if the share for any particular product remained unchanged.
4 Analytically troublesome issues are raised by the possibility of changing the nature of the product when Pa/Pb rises, for instance, economizing on wheat use by milling poorer quality wheat or even introducing a bit of sawdust into the cracked wheat bread. One may question in such a case whether our observations are of movements along a well-defined demand curve and production function. This is not, of course, a difficulty peculiar to the present model. It pertains to almost any situation in which substitution in production is possible. Moreover, if we were purist enough to say that the nature of a food product changed whenever its farm level price changed, it could also put a quick end to empirical studies of retail food demand.

402 August 1975 Pa/Pb will

Amer. J. Agr. Econ.

-- 0, the Marshallian derived demand model

marketingprocess. The greatera is, the less change when P, is changing.In the extreme case when a- --, oc, equation (15) approaches zero and Px/IP,is constant. In the morerealisticlimitingcase in whichcapplies (Friedman,chap. 7). In this case the
propositions concerning ea, eb, and Yrin this

constraints in marketingactivities may make eb quite small, so that PIP, is especially volatile. Another extreme case would be (external) economies of scale in marketing activities, which would make eC < 0 and could

and the following sections can be derived graphicallyusing the methods of Tomek and Robinson (chap. 6).
Effect of a Farm Product Supply Shift on the Retail-Farm Price Ratio

A shift in equation (6) is analyzed by taking derivatives with respect to W, while dN and dT are held equal to zero. Whenthe results are converted to elasticities, a system of three equations identical to equations (12) to (14) results, except that all E's have W as their second subscript;'rNbecomes zero in equation (14), and e, (the elasticity of Pa with respect to W) replaces zero in equation (12). Solving this new system for the elasticity of P,iP, with respect to a change in W yields5
(16) e
E'PaW

even reverse the sign of equation (16). In this case, an increasein farmproductsupply could conceivably reduce PIPa by reducing the price of marketing services as output increases. A finalinterestingspecial case is that in which marketing inputs are perfectlyelastic in supply (a long-run,nonspecificfactor case). In this case, Pb remains constant, but an increase in farm supply will still increase P,/P,. This occurs because even thoughPb is absolutely unchanged,it is increasedrelativeto Pa. Hence the relative contributionto retail food costs accounted for by marketinginputs will increase.
Effect of a Marketing Input Supply Shift on the Retail-Farm Price Ratio

DwSbea(-

eb)

Equation (16) differs from equation (15) in that for all normalcases, EpxIp,,awnegative. is Thus, the percentage difference between Px and Pa will fall when P, rises as a result of a leftwardshift in the supply functionof agriculturaloutput. Conversely, an exogenous event that reduces P, by increasing a, such as a technical improvement in crop production. will widen the percentagedifference between Px and P,. The economic reason for this result can be explainedas follows. Whenfarm product supply shifts to the right, both P, and P,, will tend to fall. But the increase in x will inputs. So long as requireadditionalmarketing

A shift in equation (5) is analyzed by taking derivatives with respect to T, while dN and dW are equal to zero. In this case, the system of equations correspondingto equations (12) to (14) is changedas follows: %N becomes zero in equation (14), er (the elasticity of Pb with respect to T) replaces zero in equation (13), and all E's have T as their second subscript. Solving this system for the elasticity of PX/P, with respect to T yields (17) (17)
E Se.T erS,,es(e, - rl) _ EPx/a D

Equation(17) has the same formas equation (16) except that ea and eb are interchangedin and the sign is reversed. Equathe numerator tion (17) will be positive in all normalcases, so that the percentagemarginbetween Px andPa will increase when Pb rises as a result of a specific tax on marketinginputs. Thus, while 0 - eCb< 00, Pb must therefore rise, increasing an exogenous change that decreases agriculprice the cost of marketingrelative to farm inputs tural supply will decrease the retail-farm ratio, the same kind of change in the supply of and hence the ratio Px/P,. As was the case in equation (15), o- plays a marketinginputs will increase the ratio. Equation (17) seems more limited in its moderating role in that the larger a is, the less
a given shift in W will change Px/Pa. The responsiveness of PxPa to W varies substantially with the context being considered. For example, in a very short-run context for a narrowly defined product, capacity
5 Derivation outlined in appendix.

applicability than equations (16) or (15) due to the aggregation problem. It is difficult to think of exogenous shifters of marketing input supply that will affect all the components of b proportionally. Technical progress, for example, will typically be associated with a particular marketing input or activity. This will

Gardner

The Farm-RetailPrice Spread 403

Table 1. Elasticitiesof Px/Pa with Respect to Shifts in Retail Food Demand, Farm Product Supply, and MarketingInput Supply
Parameter Values r 0.5 0 0 0 0 0
ea eb
EPxlPaN

1.0 1.0 1.5 2.0 2.0 1.0

2.0 2.0 2.0 2.0 1.0 2.0

"0 -0.5 -0.5 -0.5 -0.5 -0.5 -1.0

Sb

eq. (15)a -0.13 -0.18 -0.06 0 0.18 -0.14

eq. (16)a -0.33 -0.46 -0.48 -0.50 -0.54 -0.43

EPx/Pa,w

EPxIPa,T

eq. (17)a 0.40 0.54 0.52 0.50 0.46 0.57

0.5 0.5 0.5 0.5 0.5 0.5

a The values of N, in eq. (15), ew in eq. (16), and eT in eq. (17) are set equal to 1. Thus, the elasticity of P/Pa with respect to N measures the percentage response in PI/Pa to a change in N sufficient to shift the demand for x by 1% at given prices.

change the relative prices of the components of b, hence violatinga necessary conditionfor aggregation. To examine further the anatomy of equations (15), (16), and (17), they can be evaluated at hypothetical parametervalues. Let Sa = 0.5, -q = -0.5, ea = 1.0, ar = 0.5, and eb = 2.0. The resulting values of Px/Pa from equations (15) to (17) are shown in the firstline of table 1. The -0.13 elasticity means that a change in population sufficientto generate a 10%rightward shift in retail demand reduces Px/Pa by approximately 1.3%.6 Thus, the price ratio (and percentage marketing margin) fall, though quantitativelythe response is small.7 Keeping the other parameter values the same, let a- be zero. In this case the change in the marketingmarginis larger(line 2 of table 1). The economic reason for this result was discussed above with reference to a shift in retail demand. A crop like sweet potatoes, which uses a relativelysmallfractionof the land suitablefor it, may have ea largerthan 1.0, especially in a long-run context. Lines 3 and 4 of table 1 examine what happens when ea increases, holding the other parametersconstant. From line 5, when ea > eb, the percentage margin increases when Pa and P~ rise. In this case, when retail demandincreases, it is the nonagriculturalinputs in marketingthat become relatively scarce. However, when the increase in Pa and P, is induced from a shift in the agricultural supply function, equation (16),
Px/Pa
6

what ea is. In this case there is given a change in Pa, say, induced by drought. Of course, ea enters indirectly in that the larger ea is, the worse the droughtwill have to be to obtain a given increase in Pa. (For example, the effect on the price of chickens when several million contaminatedbirds were killed in Mississippi depended on the degree to which other chickenproducerscould increase supplyin response.) Line 6 shows the consequences of a more elastic demand curve at the retail level, the other parametersremainingthe same as in line 2. Price Supportsand Price Ceilings Price Controlon x If a price ceiling lower than the marketclearingprice is imposed on a food productat the retaillevel, but not at the farmlevel, what effect will this have on the retail-farmprice ratio? This question can be answered by inP, troducing as an exogenous variablein place of the demandequation(2). The resultingsystem can be solved to obtain
(18)

Ep (18)

a + eb

+ Saeb + Sbea

where Px equals the legal maximum price.8 If ea = eb, then Eapi = 1, and a legislated reduction in PX will reduce Pa by the same
a No derivation of equation (18) is given because this same result is presented in Floyd (p. 151) to show the effect of a farm level price support on the price of land and labor in agriculture. Even though Floyd considers a minimum price and equation (18) a

falls when prices increase no matter

because equations(15), (16), and (17) perAn approximation for tain to smallchanges.The approximation largechangeswould be better the closer equations (2), (5), and (6) are to constant
own-price elasticity, i.e., log-linear form, and the closer equation (1) is to a CES form.

legal maximumprice, the results are the same because in both cases the restrictionmoves us along the product supply curve. Floyd accomplishesthis for the minimumprice by having the marginis expressedas a percentagemarkup governmentbuy all that is offered at the supportprice. In the 7 If the marketing maximum price case, excess demand will exist at Ps, requiring over the farmlevel price, then PI/Pa and the marginare directly related as PI/Pa = 1 + marketing margin. nonpricerationing.

404 August 1975

Amer. J. Agr. Econ.

percentage. In this case, the percentage the x market and to equation (19) when the marketing margin is unchanged. It seems change originates in the a market (see likely, however, that ea < eb, which implies equations[A.10-11]). Since equations(18) and that Ep ,p > 1. In this case, Pa falls by a (19) are different, the value of the elasticity greaterpercentagethanP, so that the percen- of price transmission is obviously not indetage margin widens when price controls are pendent of whether the exogenous changes that generate our observations come from imposed. The sign of equation(18) is positive, imply- the demand for x or the supply of a. If the ing that an effective price ceiling on retailfood supply of a is the source of observed price will always reduce farmlevel prices. The only changes, then equation (19) applies, and c. exception would be if ea ---> In this case, a EpxP is less than one. But if shifts in food price ceiling on P, would leave Pa unchanged. demand are responsible for observed price The reason for this result is that the price changes, equation (18) applies and EJ.P(, ceiling on x always reduces the derived de- will be closer to unity, and will exceed it if ea mand for a, even though consumers want to > eb, i.e., if marketinginputs are more nearly buy morex at the lowerPX.Deriveddemandis fixed in supply than are farm products. reduced because competitive marketingfirms A function such as George and King's cannot afford to pay as much for a with the (p. 57), price ceiling as they could without one. The + Pa = P+P/ , exception when ea -- o arises because when a is perfectly elastic in supply, its price is unaf- even if it fits perfectly conditionsgeneratedby fected by a shift in the derived demandfor a. farm supply shifts, would not yield estimates of ac and p applicableto conditions generated Price Control on a by retaildemandshifts. Estimationwhen both farm supply and retail demand are shifting If the price of a is kept at a legislatedlevel by would yield an elasticity of price transmission means of productioncontrols, what effect will that is a hybrid of equations (18) and (19).10 this have on the retail-farmprice ratio? This George and King use the elasticity of price question can be answered by leaving out the transmission to derive farm-level elasticities supply equation (6) and introducinga as an from retail price elasticities of demand. In the exogenous variable. The resultingsystem can terminologyof this paper, theirresult(p. 61) is be solved to obtain This equationcan be misleading.The problem where Pl is the legislated price supportlevel. is that George and King, following Hildreth In order for a percentage marketingmargin and Jarrett (p. 108), do not distinguish bemust equal 1. tween quantities of product at the farm and to remain unchanged, retail level. They assume that x - a. This EpTx normal As long as eb > -q, that is, in all cases, assumption is of no great analytical sigequation(19) will be less than 1.9Therefore,a nificance in the case of fixed proportions, productioncontrol programthat raises Pa will since a can be transformed into x by means of raiseP, by a smallerpercentage,and the per- a constant production coefficient. Although centage marginwill narrow. fixed proportionsmay not be an unreasonable contexts, there assumptionin many marketing are several commoditiesexamined by George The Elasticity of Price Transmission
and the Elasticity of Derived Demand The percentage change in P, associated with a change in Pa is equal to the reciprocal of equation (18) when the change originates in
9 The condition for equation (19) being less than 1 is saO + Saeb < eb + SaO - SbrO. + eb from both sides and dividing by -Sb yields eb
10 There is one other relationship that arises inEPxpa directly when the exogenous event that changes both Px and Pi is a shift in marketing input supply, T in equation (5). This elasticity is

(19)

+ eb) E -+ Sao -eb Sa(<" - SbS'7 (19) E _

(20)

EaP, = (7))(Eppa).

Ep

+ e (N, W constant).

Subtracting > r). Sa,

This case is especially interesting in that it is the only one that can account, within the confines of this model of this paper, for a simultaneous fall in Pa and a rise in Px. While equations (18) and (19) are positive for all normal parameter values, this elasticity is not. It will be negative whenever a < [rvl.

Gardner

The Farm-RetailPrice Spread 405

and King for which the ratio of a to x may vary. A more general statement of the relationship between the retail elasticity of demand for food, -q, and the farm level elasticity of demand, EaPa, is readily obtainablefrom the originalsystem of equations(1)-(6). As Floyd (p. 153) shows, the elasticity of demandfor a, which is identical to the elasticity of factor demand found by Hicks (p. 244), is
(21) Eaga

The Farmer's Share of the Food Dollar

o + eb(Sar7Sbo) eb + Sacr - Sbo

Whether EaPa is greater than or less than -q depends on the relative size of o- and (the absolute value of) -q. The derived demandfunction for a will be less elastic than the retail demandfunction if and only if o- < 1|I.If o = I|1, then equation (21) yields Eapa = -. The retail and farm level elasticities are equal. If is more elastic than is the demandfunctionfor In the finalproduct.11 the case of fixed propora>
),

then the derived demand function

tion of this paper).The two are the same in the US DA publicationsbecause the quantitiesof farm product are adjusted by means of estimated productioncoefficients to obtainequivalent units for a andx. Thus, Pa is the value of farmproductper unit of x. For example, in the case of pork, the farm price for 1969is multipliedby 1.97on the groundsthat 1.97poundsof "live hog equivalent" yields 1 pound of pork sold to consumers(Scott and Badger, p. 115). This substitution of units of x for units of a is strictlycorrect only in the fixed proportions case.12 In general, the farmer's share of the food dollar is conceptually quite different from the farm price as a percentage of the retailprice of food. This sharecan be analyzed by the same methods used above to analyze

relative share aPa/xPx (which is Sa in the nota-

The data generatedby the U.S. Departmentof Agricultureon farm-retail price spreadsdo not distinguishbetween the price ratio Pa/Px and

Therefore, in this case, farm level demand is always less elastic than retaillevel demand. To show how this general approachfits in with the elasticity of price transmission as used by George and King, replace the lefthand side of equation (20) by Eapa from equation (21). Replace the right-handside of
equation (20) by equation (19) times -. These

It turns out (derivation in appendix) tions, since o- = 0, o- is always less than [1.o Px/Pa. that (22)
EsaN =
NSb(ea

- eb) (0- -

1)

where the parametersand D are as defined in determinesthe sign of equation(22). Thereare constant. A shift in demand for food at the retail level will have no effect on the farmer's
share. (b) If eb > ea and ar < 1 or if eb < ea and o- > 1, then Sa increases with N. An increase three interesting cases. (a) If either eb = ea or 0- = 1 (the Cobb-Douglas case), then Sa is equation (15). Since D > 0, the numerator

substitutions yield
+ 7)oeb(Salq

- SbCr)

eb + SaO- - SbO

4 7

Sa(eb + o-) eb + SaT - SbO)

in demandfor food will increase the farmer's

share. (c) If eb > ea and o > 1 or if eb < ea and o- < 1, then an increase in the demandfor food

0- -- 0, then

In general, the two sides are not equal. But if will decrease the farmer's share. It seems most likely for any particularfood Saeb) commodity or for an aggregateof such comSaebo) modities that eb > ea (the elasticity of the eb - SbO _ eb - Sbo supply curve of agricultural outputis less than and the George and King approachis correct. that of nonagricultural inputs used in the food marketing industry)ando- < 1. These are case 11 For instance, if( = -0.2, ar = 0.5, S, = 0.5, and eb = 1, the (b) conditions, suggesting that the farmer's value of equation (21) is
Ea
-

0.5(-0.2) + 1(0.5(-0.2) - 0.5(0.5)) = -0.33. 1 + 0.5(0.5) - 0.5(-0.2) The farm level demand elasticity is substantially greater. To obtain the general condition for > , note that this Eap occurs iff Eapa/'q > 1. From equation (21), this occurs when qor + eb(Saqn Sba) > q(eb + SaO Sby). + Saeb from both sides, dividing by Sb, and Subtracting a on factoring out Sao the left-hand and q on the right-hand side yields ar > -71.

share should increase in the presence of an exogenous increase in food demand, such as has been created for U.S. farm products by increasing export demand in recent years. Equation (22) is distinct from, though closely related to, the effect of a change in N
12 Pork does not seem to constitute a fixed proportions case since the 1.97 figure changes from year to year.

406 August 1975

Amer. J. Agr. Econ.

on Px/Pa as given by equation (15). Equation (22) and the negative of equation (15) are the same if and only if o- = 0.13 Similar methods can be used to analyze the effects of supply as well as demand shifts. The right-hand-side elasticities are different in this case, being derived by differentiating with respect to W instead of N. The resulting equation is (23) 23 EsaW = eweaSb(7)- eb) (0 - 1) D

The sign of equation (23) is determined by obeing less than, equal to, or greater than 1. If o< 1, then a shift in the supply function of a which increases Pa, for example, a drought, will increase the farmer's share. The economic sense of this result can be explained as follows. A drought reduces the food supply and hence tends to increase the price of food at both the farm and retail levels. The drought also makes agricultural output scarce relative to marketing inputs. The price of the latter rises by a smaller amount than does Pa. Therefore, the price of retail food rises by a smaller percentage than does the farm level price. If o-a 0 the ratio b/a will increase. The larger o- is, the more the demand for b will shift to the right, and consequently the larger the nonfarm input into food, which implies a smaller relative share of a in retail food costs. The elasticity of supply of b enters because although substitutability of b for a generates a shift in demand for b, the amount of additional b used depends also on its elasticity of supply to the marketing industry. The preceding discussion is intended to bring out analytical differences between the farmer's share of the food dollar Sa and the price ratio PaIP,. The USDA publications on farm-retail price spreads use the share approach by adjusting Pa such that the units it pertains to are units of x. Whether data on Sa orPa!Px are more desirable depends, of course, on the use to which they are to be put.14 The point of this discussion is that one has to be careful in interpreting farmer's share data in price ratio terms when o- > 0. For example, consider the historical data on price spreads in vegetable shortening. The farmer's share has decreased from 0.43 in 1947-49 to 0.30 in
'a Because EPaxV = - EPx PaN.
14 Actually, it is hard to see that either one has much significance for agricultural policy or welfare issues. For most purit would seem more pertinent to look at relative farm income poses than relative prices or shares.

1967-69 (Scott and Badger, p. 174), a decline of about 30%, while P,/PX has increased about 17% over the same period.15 How is this possible? It is possible because while Pa (the price of soybeans) increased relative to Ps, other inputs have replaced soybeans to such an extent that Sa has actually fallen. Indeed, an estimate of o- can be obtained by dividing the percentage change Sa by the percentage change in Pl/Pa, since they differ only in being multiplied by (o- - 1).16 Thus, o- - 1 = -0.30 and o- 2.8. This is a very crude -0.17 estimate and implicitly includes alternative farm products to soybeans in b. This may account for the high value of o-. That Pa/PX increased while Sa decreased itself implies o- > 1.

Summary and Conclusion Consistency with market equilibrium in a competitive food industry puts constraints on the pricing policies of food marketing firms. This paper has investigated the consequences of these constraints for the retail-farm price ratio and the farmer's share of the food dollar. One implication of the results is that no simple markup pricing rule-a fixed percentage margin, a fixed absolute margin, or a combination of the two-can in general accurately depict the relationship between the farm and retail price. This is so because these prices move together in different ways depending on whether the events that cause the movement arise from a shift in retail demand, farm supply, or the supply of marketing inputs. Some more specific results concerning the retail-farm price ratio are as follows. (a) Events that increase the demand for food will reduce the retail-farm price ratio (and percentage marketing margin) if marketing inputs are more elastic in supply than farm products, but increase PX/Pa if marketing inputs are less elastic in supply than farm products. (b) Events that increase (decrease) the supply of farm products will increase (decrease) Pr/P,.
'5 For Px, the data are the retail price figures of Scott and Badger (p. 174); for P,, the price of soybeans as reported by the USDA. 16 This is true whether the observed changes in S, and PX/Paare generated by shifts on the supply side or the demand side, since both equations (15) and (22), and (16) and (23) differ only by the term ar - 1. The same result holds for elasticities with respect to T.

Gardner

The Farm-Retail Price Spread 407

(c) Events that increase (decrease) the supply of marketinginputs will decrease (increase) P/IPa. (d) An effective price ceiling on retail food will reduce the price of farm products (unless the supply of farm products is perfectly elastic);PIPa will increase (decrease)if the elasticity of supplyof farmproductsis less (greater) than that of marketing inputs. (e) Supportingthe price of farm products above the unrestrictedmarket equilibriumlevel will reduce PI/Pa. All the preceding propositions can be derived by graphical methods like those of Tomek and Robinson (chap. 6) under the assumptionof fixed proportionsin food marketing (a- = 0). The advantageof the mathematical model is that it allows the treatmentof the more general case in which ar > 0 and it provides quantifiableresults.17 Other related results are as follows. (f) The farm level demand for agriculturalproducts will be more or less elastic than the retail demandfor food as ar Ir . (g) The percentage price spread is analytically distinct from the farmer's share of the food dollar, and the two will behave differentlyunder changingmarket conditionsunless ar= 0. If- = 1, the farmer's share is constant. If a-> 1, an increase in the marketingmarginwill be accompaniedby an increase in the farmer's share of the food dollar. Otherwise,lower marginsgo togetherwith an increasedfarmer'sshare. (h) The elasticity of substitution between farm products and marketinginputs in producingretail food can be estimated by dividingobserved changes in the farmer's share of the food dollar by observed changes in the ratio of farm to retail food prices. Two limitationsof the model are that it assumes competition and that it aggregates all marketingactivities into one productionfunction and all nonfarmmarketing inputsinto one quantity. In relaxing the assumption of competition, althoughthe constraintsimposed by competition would disappear, the behavior of the marketingmarginwould still not be arbitrary. For example, the price behavior of a profitmaximizingretail food seller with monopoly power could be analyzed by replacingmarginal producttimes input price by marginalrevenue product in equations (3) and (4). Then
17 In the strictfixedproportions case, marginal productscannot be calculatedand the originalsystem of derivativesbreaksdown. The correctprocedureto get quantitative predictionsin this case is to take the limit of equations(15) to (23) as (r --- 0.

elasticities such as equations (15), (16), and (17) could be solved from the new system. Similarly, monopsony in the purchase of a farmproductcould be introducedby replacing input price by marginalfactor cost. The aggregationproblemis serious in some contexts but negligible in others. It is most serious when the changes being considered have large effects on the relative prices of different marketing inputs. In order to examine particular relative price changes within the set of marketinginputs, a threeinput model along the lines of Welch might prove a useful alternative approach. A possible further extension would be to add separate production functions and profit-maximizationequations for different marketing activities. This approach would provide more realismfor investigatingcertain problems but would be costly in terms of complexity and intelligibility, and it seems doubtful whether it would yield any basic changes in the results from the simple model of this paper as expressed in propositions (a) through(h). But this remains to be seen. Finally, it mightprove interestingto investigate the consequences of technical progress in the marketing industry by introducing exogenous shifters of equation (1). This also could follow the approachof Welch. [Received October 1974; revision accepted March 1975.] References
Allen, R. G. D. Mathematical Analysis For Economists. London: Macmillan & Co., 1938. Brandow, G. E. "Demand for Factors and Supply of Output in a Perfectly Competitive Industry." J. Farm Econ. 44 (1962):895-99. Floyd, John E. "The Effects of Farm Price Supports on the Return to Land and Labor in Agriculture." J. Polit. Econ. 73 (1965):148-58. Friedman, Milton. Price Theory. Chicago: Aldine Publishing Co., 1962. George, P. S., and G. A. King. Consumer Demand for Food Commodities in the United States with Projections for 1980. Giannini Foundation Monograph 26, Mar. 1971. Hicks, J. R. The Theory of Wages. Gloucester, Mass.: Peter Smith, 1957. Hildreth, Clifford, and F. G. Jarrett. A Statistical Study of Livestock Production and Marketing. N.Y.: John Wiley & Sons, 1955. Scott, Forrest E., and Henry T. Badger. Farm-Retail Price Spreads for Food Products. USDA ERS Misc. Publ. 741, Jan. 1972.

408 August 1975 Tomek, William G., and Kenneth L. Robinson. Agricultural Product Prices. Ithaca and London: Cornell University Press, 1972. U.S., Department of Agriculture. Agricultural Statistics, 1972. Washington, 1972. Welch, Finis. "Some Aspects of Structural Change and the Distributional Effects of Technical Change and Farm Programs." Benefits and Burdens of Rural Development, pp. 161-93. Ames: Iowa State University Press, 1970.

Amer. J. Agr. Econ.

(A.1) Ep,
SqN(Sbee + Saeb + a) --q(Sbea + Saeb + o-) + eaeb + o-(Saea + Sbeb)"

The denominator of this expression, in the text, and henceforth in this appendix is denoted by D. Next, from the same system of equations, solve for EaN. (A.2)
Nea(eb EaN= + 0-)

Appendix To get from Mathematical Derivations Derivation of Empa,N. Starting with equations (9) to (11), first convert all derivatives into elasticities. For example, from equation (5), dPb db Multiplying by b/b and Pb/Pb,
b( dPb
EaN to EPaN,

divide EaN by

ea,

since

(A.3)

EaN/ea

da =( dN
dPa dN

N a
N Pa

da
dPa

P,
a

Finally, to get EP?P,N, note that (A.4) d(PxIPa) dN

EP/PaN = P/ PadPx - PxdPa

N ' NPa Px

Pa2 dN NP 2dPx Pa2PxdN

= EPxN EPauN

I Pb
eb

NPxdPa
PxPadN

where eb is the own-price elasticity of supply to the industry. Second, use the assumption of constant returns to scale to eliminate all second partials (Allen, p. 343), since fab and fa b .faf
a o-x

Substituting (A. 1), (A.2), and (A.3) into (A.4) yields

(A.5)
EPPaN _ "NSb(ea eb)

fafb

o-x

which is text equation (15). Derivation of Eppa,w. After making the changes in equations (12) - (14) described in the text, solve for

Third, eliminate fa and fb wherever they appear by substituting Pa/Px and Pb/Px from equations (3) and (4). Making these substitutions and rearranging terms yields equations (12) to (14). From the system of equations (12)-(14), first find EN by means of Cramer's Rule. Expanding the appropriate determinants, E SbSo +Sb S EP=N( eIeb
+

(A.6)
and

=Ep

ExD

eweaSa(eb

+ 0-)

(A.7)

Eaw =

+ eb(Sa"q Sb0.)) ewea(eqo0D

I + (Tea eaeb 2Sa+_

__V

SbSa Sa2 S2 0"2V

/(Sb +

Ia

1 +

2+

+ S

eb

+ Sb ea

The second bracketed term of the denominator equals 1/o. (since Sb = 1 - Sa). Multiplying the numerator and denominator by yields
0eaeb

To get from Ea, to EPaW,it is again necessary to divide Eaw by the elasticity of a with respect to Pa. But the appropriate elasticity is the elasticity of demand for a, not the supply elasticity as was used in equation (A.3). In the preceding section the demand for x was shifting, which generated movement along the supply curve of a. In this section the supply curve of a is shifting, which generates movement along the demand curve for a. Therefore, to get EPaw, divide Eaw by EaPa where EaPa is the elasticity of demand for a. Using the formula for EaPa given as text equation (21) yields

Gardner (A.8) EPaw = Eaw/Eapa _ ewea(eb + Sao0 - Sbr1) D

The Farm-Retail Price Spread 409

E (A.13)

dx
dN Pa PX
= SaE

N
x da dN N x

da
dN

N
x

db
dN db dN

N
x N x b b

Subtracting EPw from EPw to get EP

(A.9)

yields /paw

Pb a + P1

Ep

eweaSb(n - eb)

+ SbEbN.

which is text equation (16).

Derivation of EpPa when price changes are caused by a shift in product demand. This elasticity can be obtained by dividing EpNl/EPaN.Using equations (A.1),

Substitutingequation (A.13) into equation (A.12) yields

(A.14)
ESaN = EPaN - EP + Sb(EaN - EbN).

(A.2), and (A.3),

(A.10)

Ep

(xPa

= "N(Sbea

+ a)

The new elasticity in (A.14) is EbN. It is the third variable in the system, equations (12) to (14), which has already been solved for EaN and Ep Returningto the N. equationsystem for the firstpartof the appendix.
(A. 15) D(A) Neb(ea
+
0r)

DSaeb

'qN(eb +

r)
EbN =

Sbea + Saeb + eb + o

This is the reciprocalof text equation(18).

Derivation of Epxpa when price changes are caused by a shift in the supply curve of a. This elasticity can be

Combiningequations (A.3), (A.1), (A.2), and (A.15) accordingto equation(A. 14) yields
(A.16) E
= (eb + O - Sbea - r + Sbeaeb

obtainedby dividing(A.6) by (A.8),

+ (A.11) E xp = eweaSa(eb r) D

+ Sbeacr Sbebea - SbebOr) D Sa0r SA')

-

)N(Sb(ea

- eb) (

1)),

ewea(eb +

Sa(eb + ra) eb + Sar - Sb'

which is text equation(22).

Derivation of Esaw. Again, all the elasticities are

which is text equation (19). Derivationof ESaN. First consider the total differential
of Sa
dS= da

availableexcept Ebw:
(A. 17) + Ew = eweaebSa(qr 0r)

xPx(adPa + Pada) - aPa(xdPx + PAdx) (xPx)2

Combiningequations (A.8), (A.6), (A.7), and (A.17) accordingto equation(A. 14) with W replacingN yields
Es
=

Dividingthroughby a changein population,dN, to get an exogenous influenceon the system fromthe demandside, yields (after convertingto elasticities)
(A. 12) EsaN =
EpaN + EaN EN

ewea

[eb + SaOr -

SAbe

Saeb

SaOr

SAb'r

Ex-N.

+ Sb(Saeb7e SbebOc Saebe7 - Saeba)]

These elasticities were all discussed earlier except for

ExN. Using the facts thatfa = PaIP, fb = Pb/Px, and dx = fada + fbdb, this elasticity is analyzed as follows:

Sewea[(r

- eb) (0r 1)],

which is text equation (23).