GRADING, MINIMUM QUALITY STANDARDS,AND THE LABELING OF GENETICALLY

MODIFIED PRODUCTS

HARVEY LAPAN AND GIANCARLO MOSCHINI

We relate the labeling of genetically modified (GM) products to the theory of grading and minimum

quality standards. The model represents three stages in the supply chain, assumes a vertical product

differentiation framework, allows for the accidental commingling of non-GM products, and treats

regulation as a purity threshold for non-GM products. We find that a non-GM purity level that is

too strict leads to the disappearance of the non-GM product, and that some quality standard benefits

farmers. Indeed, the standard that is optimal from the perspective of producers is stricter than what is

optimal for consumers and for societal welfare.

Key words: biotechnology, food labeling, grading, identity preservation, minimum quality standards,

regulation, uncertainty.

DOI: 10.1111/j.1467-8276.2007.01002.x

The one-billionth acre of genetically modi-fied (GM) crop was planted in 2005, onlyten years after this technology was first in-troduced. Whereas this milestone exemplifiesthe widespread and fast diffusion of a tremen-dously promising and radical innovation, therecontinues to be public resistance and oppo-sition to GM food (Evenson and Santaniello2004). Nowhere is this more apparent than inthe European Union (EU) where public oppo-sition first led to a regulatory impasse with se-rious trade implications (Pew Initiative 2005),1

and finally resulted in a new, complex, andstringent regulatory system centered on thenotions of GM labeling and traceability. In theEU, all foods produced from GM ingredientsmust now be labeled, regardless of whether ornot the final products contain DNA or pro-teins of GM origin. Such labels will have tostate: “This product contains genetically mod-ified organisms,” or “This product has beenproduced from genetically modified [name oforganism].” To avoid carrying a GM label, ahigh level of purity is required: the tolerance

Harvey Lapan is university professor and GianCarlo Moschini isprofessor and Pioneer Chair in Science and Technology Policy, andboth are with the Department of Economics, Iowa State University.

The support of the U.S. Department of Agriculture, through aNational Research Initiative grant, is gratefully acknowledged.

1 The 2006 ruling of the WTO panel supported the U.S. posi-tion in its challenge to the moratorium that stalled GM productapprovals in the EU from 1998 to 2005.

level for the presence of “authorized” GMproducts is set at 0.9%.2

This drive toward GM labeling, mirrored inmany other countries, is emerging as a keypolicy response to the introduction of GMproducts. This effect contributes to the ongo-ing transformation of the agricultural industryfrom one that produces largely homogenouscommodities into one that eventually may becharacterized by differentiated goods. Meetingthe demand for differentiated food productsrequires a system that can credibly deliver suchdifferentiated products to end users. Previouswork in this area has centered on the notion of“grading” agricultural commodities. Gradingof products and government inspections havelong been used in agricultural markets in pur-suit of a variety of objectives (Dimitri 2003).In this setting it is useful to separate regula-tions that aim at improving the health safetyof the food supply from quality regulationsthat have mostly an informational role vis-a-visthe quality of the good as perceived by con-sumers (Gardner 2003). The latter are moregermane when considering the issue of GMlabeling (health safety considerations are ar-guably best dealt with at the approval stage ofGM products). Specifically, the introduction of

2 These regulations became effective in April 2004. The manda-tory labeling is supplemented by traceability requirements, meantto facilitate monitoring of unintended environmental effects andto help enforce accurate labeling (European Union 2004).

Amer. J. Agr. Econ. 89(3) (August 2007): 769–783Copyright 2007 American Agricultural Economics Association

770 August 2007 Amer. J. Agr. Econ.

GM products that some consumers deem un-desirable means that the corresponding non-GM preinnovation traditional product attains,for these consumers, the status of a “superior”product.

Tapping the emerging demand for non-GMfood presents challenges. If the superior non-GM product cannot be distinguished from theinferior GM one, the pooled equilibrium likelyto emerge would display the attributes of Ak-erlof’s (1970) “lemons” model; that is, it wouldcontain too high a proportion of low-qualityproduct (Gaisford et al. 2001). A credible la-beling and certification system, that distin-guishes GM from non-GM products in themarketplace, is clearly desirable. Whether thatshould take the form of mandatory labeling ofthe (inferior-quality) GM products, as with thenew EU regulation, is of course highly ques-tionable (Crespi and Marette 2003; Lapan andMoschini 2004). The problem here is not sim-ply one of asymmetric information (i.e., theseller has private information that may be valu-able to buyers, and labeling requirements mayforce disclosure of such information) but thefact that the information to be disclosed toconsumers needs to be produced through anad hoc process. This is because the product ishandled a number of times as it moves fromfarmers to consumers, and the possibility for(inadvertent) mixing of distinct products existsat each stage (Bullock and Desquilbet 2002).Thus, to satisfy the underlying differentiateddemand for GM and non-GM products, costlyidentity preservation (IP) activities are nec-essary, and such activities obviously need tobe carried out by the suppliers of the superior(non-GM) product.

Accepting the need for a GM labeling sys-tem still leaves open the question of what fea-tures it should have. In addition to the issueof whether labeling should be voluntary ormandatory, mentioned earlier, a critical ele-ment concerns what it means to be non-GM.Because of the aforementioned need for ex-tensive IP measures at various production andmarketing stages, it is becoming apparent that100% purity is simply not attainable.3 Thus,a critical element of emerging GM labelingregulations concerns the threshold or toler-ance level—i.e., the maximum level of impuritythat is admissible in food while still allowing aclaim of non-GM. No uniformity appears tobe emerging across countries on this matter.

3 Episodes of accidental GM contamination (e.g., the high-profile Starlink case) support this presumption.

As mentioned, the EU has set an extremelystrict threshold level of 0.9%. Australia andNew Zealand have an almost-as-strict toler-ance level of 1% (but, unlike in the EU, thesecountries exempt highly refined products, suchas vegetable oils, from the labeling require-ments). Japan and South Korea, on the otherhand, have opted for laxer standards (e.g.,Carter and Gruere 2003). Their tolerance lev-els are 5% and 3%, respectively, and only ap-ply to the main ingredients of a food item (topthree ingredients in Japan and top five ingre-dients in South Korea).

The question of the appropriate thresholdlevel for non-GM products can be viewed asthe establishment of a government-mandated“minimum quality standard” (MQS). The sem-inal work of Leland (1979) exemplified theconsequences of Akerlof’s (1970) lemon prob-lem and showed that an MQS can improve thewelfare attributes of an otherwise unregulatedcompetitive system. In agricultural economics,MQSs were analyzed by Bockstael (1984), whoshows that under perfect competition and withobservable quality an MQS would lead to wel-fare losses. This is because an MQS, in thatsetting, simply amounts to an arbitrary restric-tion on an otherwise undistorted competitivemodel (as also noted by Shapiro 1983). MQS is-sues also emerge in the context of studying theeconomics of grading agricultural commodi-ties, which identifies ranges of qualities forwhich different markets (and different prices)arise (Bockstael 1987). Some of the connec-tions between grading and MQSs, and theimportance of distinguishing between the con-texts of perfect information and of quality un-certainty, are discussed in Gardner (2003).

It is common to presume that grading en-visions marketing of all product qualities,whereas an MQS excludes some qualities fromthe market. MQSs thus narrowly construedclearly do not apply to a number of realisticcases in agriculture. Organic standards set bythe USDA, for example, identify the minimumquality (percent of organic ingredients) neces-sary to belong to one of three organic prod-uct categories (“100% organic,” “organic,”and “made with organic ingredients”). Clearly,products failing an upper category can be mar-keted in the lower one, and products failing thelowest organic category standard can still bemarketed as conventional products (and thusare not excluded from the market). The caseof non-GM labeling that we posit in this arti-cle is of the same nature: a product failing thenon-GM standard can still be marketed as a

Lapan and Moschini Grading, Minimum Quality Standards, and GM Labeling 771

GM product. That standards do not necessar-ily prohibit marketing of lower qualities wasnoted by earlier studies (e.g., Bockstael 1984).In what follows, we will maintain this featureas an integral part of the model. Indeed, whatgives rise to the potential for product differen-tiation in our setting (i.e., heterogeneous pref-erences for quality) also directly specifies thenature of the market for the lower standardproduct and specifies the substitution relation-ship between the higher quality and the lowerquality products.

In this article, we pursue in detail the ques-tion of setting an MQS for non-GM food, asattempted by the GM labeling regulations dis-cussed in the foregoing. Our analysis relieson an explicit market equilibrium model thatcaptures the stylized attributes of GM inno-vation. Specifically, we develop a model thathas the following basic elements: (i) hetero-geneous consumers with preferences over thedifferentiated goods (GM and non-GM prod-ucts); (ii) producers (farmers) for whom GMproduct provides an efficiency gain; (iii) mid-dlemen, who purchase from farmers, grade andlabel goods, and resell them to consumers; and(iv) a government, which identifies the opti-mal grading system (i.e., the GM tolerancelevel). In the framework of analysis that wedevelop, uncertainty plays a critical role, andthe need for IP activities to keep GM and non-GM products sufficiently segregated in theproduction and marketing system is explicitlyaddressed.

The Model

We consider three market stages: (1) the farmlevel, where agricultural output of either GMor non-GM type is produced; (2) the mar-keting level, which uses agricultural productsas inputs in a chain that involves assembly,transportation, processing, and distribution,yielding food products that can be sold toconsumers; and (3) the consumer level, wherefinal users have the choice (in general) of GMand non-GM products. In this setting, there-fore, there are two output products at the farmlevel, and two output products at the market-ing level, so that we need to distinguish fourprices. The superscripts 0 and 1 will denote thefarm and consumer levels, respectively, and thesubscripts g and n will denote GM and non-GMproducts, respectively. Thus, p0

n is the farm-gate

price of a non-GM product; p0g is the farm-gate

price of the GM product; p1n is the consumer

price of the good certified as non-GM; and, p1g

is the consumer price of the GM good.The model that we develop envisions a com-

petitive farm sector with a standard upward-sloping supply curve. The marketing level isalso modeled as competitive and operatingunder constant returns to scale and, in addi-tion to the standard marketing services (e.g.,storage, transportation, processing, . . .), hereit also provides the IP activities necessary fornon-GM product. The final consumption leveldisplays differentiated demand for GM andnon-GM products, which is modeled as aris-ing from preference heterogeneity. The con-sumer level displays the property that the GMproduct is a weakly inferior substitute for thenon-GM products, as in Lapan and Moschini(2004).

Farm Level

We consider a sector in which many com-petitive farmers produce both GM and non-GM products. The GM product is appealing tofarmers because it decreases production costs.This is a property of so-called first-generationGM traits, as illustrated by herbicide-resistantcrops (Falck-Zepeda, Traxler, and Nelson2000; Moschini, Lapan, and Sobolevsky 2000).To represent this process in the most efficientway, we postulate that the GM product offers aconstant unit cost savings equal to � > 0. Thus,if Qg and Qn denote the aggregate productionof GM and non-GM products, respectively, theaggregate cost function of the farm sector iswritten as C(Qg + Qn) + �Qn, where C(·) isa convex function (i.e., the marginal cost C′(·)is increasing). Let S(·) denote the inverse ofthe marginal cost function, that is, the func-tion satisfying S−1(·) = C′(·). Then the aggre-gate supplies of the two farm-level products,arising from competitive profit maximizationat the farm level, can be written as

{Qg = 0

Qn = S(

p0n − �

)(1)

{Qg = S

(p0

g

)Qn = 0

(2)

⎧⎪⎨⎪⎩

Qn + Qg = S(

p0g

)Qn ∈ [

0, S(

p0g

)]and

Qg ∈ [0, S

(p0

g

)](3)

772 August 2007 Amer. J. Agr. Econ.

where equations (1) apply if (p0n − p0

g) > �,

equations (2) apply if (p0n − p0

g) < �, and equa-

tions (3) apply if (p0n − p0

g) = �.

The cost savings � is taken as exogenouslygiven. Thus, in any equilibrium where both GMand non-GM are produced and consumed, itwill be the case that p0

n = p0g + � (i.e., the

farm-level premium for the non-GM productsimply compensates for a production cost dif-ference). Also, note that, in such an equilib-rium, the supply of either product is infinitelyelastic, although clearly total supply is upwardsloping.

Marketing Sector

We consider a generic middleman, referred toas processor, who performs all the relevantmarketing functions between the farm leveland the consumer level. The processor buysproduct of a declared type from the farmer,moves it through a distribution chain, and sellsit to consumers. Because farmers can produceGM and/or non-GM products, any one proces-sor may be buying either product from any onefarmer.

A fundamental part of the problem at handis the possibility of the unintended comming-ling of GM and non-GM products, which ne-cessitates the use of IP activities that can con-trol such contamination. As the good movesthrough the production and marketing sec-tors, there is a positive probability of contami-nation. This contamination can occur duringproduction, storage, transportation, or else-where along the chain. It may occur becauseof cross-pollination during primary produc-tion, because employees are careless duringthe postharvest handling process, because con-tainers are not perfectly cleaned, and so forth.Our model is agnostic as to where contamina-tion takes place. We simply presume that someIP activities need to be carried out before anon-GM product can be sold as such to theconsumer, and for simplicity we model IP aspart of the marketing level.

Specifically, we assume that any lot of non-GM farm product that the processor purchaseswill, during the processing and distributionprocess, become contaminated with some (per-haps only trace amounts of) GM product, andthat this impurity level has a given distributionfunction. Thus, for each non-GM lot that theprocessor purchased and then processed, wedefine as si ∈ [0, 1] the impurity level of lot

i (i.e., the fraction of GM material in the finaloutput). The density and distribution functionsof si, which are assumed to be independentlyand identically distributed (i.i.d.), are writtenas f (s) and F(s), respectively. Thus F(s) rep-resents the probability that a given lot has animpurity level not higher than s, and given alarge number (continuum) of i.i.d. lots, it alsorepresents the proportion of non-GM outputthat has an impurity level not higher than swhen it reaches the marketing stage.

We conceive of marketing activities as sup-plied by a competitive sector displaying con-stant returns to scale at the aggregate level.This condition is defensible if there are no ma-jor barriers to entry in the supply of marketingand processing services. Also, with this as-sumption the results on the effects of an MSQfor GM labeling can be related directly to theprimitives of interest (e.g., heterogeneous con-sumer preferences, stochastic commingling,and costly IP). With that, we represent thecompensation of the activities incurred at themarketing level (except for IP) in terms of aconstant unit cost � > 0 that is incurred forany unit of farm output that is handled (re-gardless of its type). Furthermore, for any unitof farm-level non-GM output that is handled,processors also need to supply segregation andIP, and thus also incur a constant unit IP cost� > 0.4

Modeling segregation and IP costs as con-stant (per unit of output) may be seen asa simplification because some of these costs,such as storage costs or traceability require-ments, might entail marketing firms investingin some fixed input. It is apparent, however,that greater non-GM output moving throughthe marketing system requires more storagecapacity and more production lots whose ori-gins can be traced. Moreover, if these fixedcosts are on a per firm basis, and if each mar-keting firm has U-shaped average costs, thenwith free entry the standard competitive anal-ysis (with constant unit costs at the industrylevel) would apply. If fixed costs lead to declin-ing average cost for a firm, on the other hand,then the competitive market structure wouldhave to be abandoned. Our model focuses oncomparing producer and consumer attitudes

4 In this model, therefore, it is the labeling of the (superior) non-GM product that incurs the additional IP cost. As noted earlier,the EU labeling regulations require labeling of GM products. Butclearly, to escape the need for such a GM label, purveyors of non-GM products would need to undertake the (costly) activities thatensure that commingling does not take place.

Lapan and Moschini Grading, Minimum Quality Standards, and GM Labeling 773

toward standards in a typical competitive set-ting.

Consumer Demand

Underlying the need for costly IP, and for gov-ernment regulation, there must be willingnessto pay for non-GM product on the part of someconsumers. As discussed earlier, the premiseis that, whereas consumers never preferthe GM product when the equivalent non-GMgood is available at the same price, some con-sumers are willing to pay something to avoidthe GM product. Thus, as in previous studies(e.g., Giannakas and Fulton 2002; Lapan andMoschini 2004), we model GM and non-GMproducts as vertically differentiated products.A demand specification that has proven use-ful in this context is the unit demand model ofMussa and Rosen (1978), where consumers’differing valuation of quality is captured by anindividual taste parameter. To adapt this de-mand framework to our MQS setting, wherethe purity of the superior product is the essenceof the problem, we proceed as follows.

An individual consumer is assumed to buy,at most, one unit of the product, which comesin two varieties, the GM product and the non-GM product. We posit that either variety pro-vides the same basic level of utility u, but theGM product provides a disutility that differsacross consumers. Furthermore, because of thepresumption that 100% purity is not feasi-ble, the non-GM product at the consumptionlevel has an impurity level s, and this impu-rity also causes a disutility that differs acrossconsumers. That is, the consumer gets utilitylevels Un, Ug, or U0 when, respectively, a unitof non-GM is bought, a unit of GM is bought,or neither variety of this product is bought.Specifically,

Un = u − a�s − p1n(4)

Ug = u − a� − p1g(5)

U0 = 0(6)

where � ∈ [0, 1] indexes the type of the con-sumer.

The parameter � captures the heterogene-ity of consumers, vis-a-vis their preferences forthe non-GM attribute, and the parameter a >0 captures the intensity of consumers’ aver-sion to GM ingredients (due, for example, tothe subjective perception of the harm that may

derive from consuming GM products).5 Notethat if 100% purity were feasible, so that s = 0,the preference structure that we have assumedreduces to that used by Giannakas and Ful-ton (2002).6 The heterogeneity of consumerpreferences is represented by assuming thatthe individual preference parameter � is dis-tributed in the market according to the abso-lutely continuous distribution function H(�).As noted earlier, it is common in this setting tofurther assume that the distribution of typesis uniform, which clearly restricts the resultingdemand model. Thus, we derive most of ourresults for a general distribution of types (al-though we do resort to the uniform distributionassumption to derive unambiguous compara-tive statics analysis).

Given this preference specification, the con-sumer of type � will not consume the GMproduct if p1

n(�) < p1g(�); he will not buy the

non-GM product if p1n(�) > p1

g(�); and he

will be indifferent between the two productsif p1

n(�) = p1g(�), where

p1n(�) ≡ p1

n + a�s(7)

p1g(�) ≡ p1

g + a�(8)

can be interpreted as the “personalized prices”of the non-GM and GM products, respectively.It may be useful to observe that Tirole (1988,pp. 96–97) shows that the standard setup ofMussa and Rosen (1978), in which quality pro-duces a different utility level for each individ-ual, can be reformulated as the case in whichquality produces the same surplus from thegood but individuals face a personalized pricewhich, in that setting, reflects the impact ofincome distribution (i.e., the preference pa-rameter is isomorphic to the reciprocal of themarginal utility of income). Similarly, in oursetting the consumer’s aversion to the GM con-tent is conveniently reflected as augmentingthe effective price of the two products, withthe augmenting factor being proportional to

5 Thus, the highest preference for the non-GM product is ex-pressed by the consumer with � = 1, whereas for consumers with� = 0, GM and non-GM goods are perfect substitutes.

6 Giannakas and Fulton (2002) also consider the case of “misla-beling,” wherein the non-GM unit purchased by the consumers iseither a pure non-GM with some probability or GM otherwise. Thepreference structure that they use for that case is precisely what weare maintaining, with the purity level s replacing their probabilityof mislabeling.

774 August 2007 Amer. J. Agr. Econ.

β β

100β

• • •

•

•1nu p−

1n nU u p a sβ= − −

1gu p− 1

g gU u p aβ= − −

Figure 1. Heterogeneous preferences for quality and unit demand

the individual consumer preference parameter�.7

Based on the foregoing, it follows that, forgiven consumer prices p1

n and p1g, there exists a

consumer type � such that consumers of type

� < � will not purchase the non-GM product,

whereas consumers of type � ≥ � will not pur-chase the GM product, where8

� ≡(

p1n − p1

g

)a(1 − s)

.(9)

Thus, � denotes the type of consumer who is in-different (at given prices) between consumingthe GM and the non-GM product. Similarly, let� denote the type of consumer who would beindifferent (at given prices) between consum-ing the non-GM product or nothing, and let �0

denote the type of consumer who would be in-different between purchasing the GM productor nothing:

7 One advantage of the formulation that we propose is that thepersonalized prices are linear in the taste parameter �, whereas inTirole’s (1988) reformulation of the standard unit demand modelof VPD the personalized price is nonlinear in the taste parameter.

8 Note that the consumer will make a purchase only if u ≥Min{p1

n(�), p1g(�)}. Thus, for � < � the consumer will buy the GM

product if u > p1g(�), whereas for � ≥ � he will buy the non-GM

product if u ≥ p1n(�). We assume that u > p1

g(0) = p1g, guaranteeing

that the market exists.

� ≡ u − p1n

as(10)

�0 ≡ u − p1g

a.(11)

Assuming p1n > p1

g, the critical types {�,

�0, �} must be ordered either as � < �0 < �or as � ≥ �0 ≥ �. In the former case, no non-GM product will be sold and the marginalconsumer who buys the GM good is given by

Min{�0, 1}. In the latter case, if � ≥ 1, then themarket will be covered, and again no non-GM

product will be sold, whereas finally if � < 1,then some non-GM good will be demanded,and the marginal consumer will be Min{�, 1}.Figure 1 illustrates the case of � < 1. Then, theaggregate demand functions Dn(p1

n, p1g) and

Dg(p1n, p1

g) for non-GM and GM products for

the case pictured in the figure are

Dg(

p1n, p1

g

) =∫ �

0

h(�) d� = H(�)(12)

Dn(

p1n, p1

g

) =∫ �

�

h(�) d� = H(�) − H(�)

(13)

where H(�) is the distribution functionof consumer types, h(�) ≡ H′(�) is the

Lapan and Moschini Grading, Minimum Quality Standards, and GM Labeling 775

corresponding density function, and (withoutloss of generality) the mass of consumers in themarket is normalized to one.

Regulation

Government regulation in this setting takes theparticular simple form of a scalar R, which de-notes the maximum impurity (threshold) levelbelow which a good can be sold to consumersas non-GM. Thus, for example, the 0.9% EUstandard discussed earlier would be equivalentto R = 0.009 in our model, whereas the 5%Japan standard would be equivalent to R =0.05. Throughout we assume that this regula-tion can be enforced costlessly. Still, a givenstandard R affects the expected impurity s ofmarketed non-GM product, given the purifica-tion technology discussed earlier.9 Specifically,s is the mean of s truncated at R, that is

s =∫ R

0yf (y) dy

F(R)≡ s(R).(14)

Pooling and Separating Equilibria

Before considering the effects of regulation itis important to articulate the conditions that,in our model, govern whether a “pooling” or a“separating” equilibrium emerges.

Pooling Equilibrium

Consider first the equilibrium in which no la-beling of products occurs. In this case, becausesellers of the non-GM product have no way ofdifferentiating their superior (but more expen-sive) product from that of the GM product, inthe resulting pooling equilibrium only the GMproduct is sold. Because only the GM productis sold, the marketing sector’s profits are

�Mg = (

p1g − p0

g − �)Qg.(15)

Assuming free entry in the marketing sector,this implies

p1g = p0

g + �.(16)

Using the supply and demand functions forthe GM good developed earlier, the market

9 We note that the stochastic production of the high-quality goodin our setting is similar to the setup analyzed by Stivers (2003).He considers a standard monopoly quality setting within the VPDmodel. Production of quality is stochastic, so that a distribution ofquality is harvested (he provides the examples of timber produc-tion and diamond mining).

clearing condition that equates excess supplyZ(p0

g) ≡ S(p0g) − D(p0

g + �) to zero requires

S(

p0g

) −∫ Min{�0,1}

0

h(�) d� = 0.(17)

The willingness to pay for the highest �type is p1

g = (u − a) → p0g = (u − a − �); if

supply at that price is at least equal tothis maximum demand, then the market willbe “covered.” Using (17), if S(u − a − �) ≥∫ 1

0h(�) d�, then the market is covered. Be-

cause maximum market size is normalized to 1,and supply is the inverse marginal cost curve,this means the market is covered if (u − a −� − C′(1)) ≥ 0.

If the market is covered for this case, it mustalso be covered if the non-GM product is alsomade available, in which case total sales—andthe farm price—will be independent of govern-ment policy. Because one of our (and society’s)concerns is how regulation affects producers,we assume that, absent non-GM sales, the mar-ket is not covered. Specifically:

ASSUMPTION 1. C ′(1) > (u − � − a).

Given this assumption, from (17) the equilib-rium price p0,e

g solves

S(

p0,eg

) −∫ �0,e

0

h(�) d� = 0(18)

and the marginal buyer �0,e satisfies

�0,e = u − � − p0,eg

a< 1.(19)

Separating Equilibrium

We now analyze the case in which a labelingregime makes it possible for the processingsector to consider selling the higher qualitynon-GM good. For any lot handled at thismarketing stage, the prices that are relevant tothe processor’s decisions are the input pricesp0

n (the farm-gate price of a non-GM product)

and p0g (the farm-gate price of the GM prod-

uct), and the output prices p1n (the consumer

price of the non-GM good) and p1g (the con-

sumer price of the GM good). These pricesare endogenous to the system but are takenas given by an individual processor.

Suppose that the marketing sector handlesquantities Qn and Qg of farm-level output of

776 August 2007 Amer. J. Agr. Econ.

non-GM and GM products, respectively. Asdiscussed earlier, farmers will supply non-GMproduct only if p0

n > p0g, specifically,

p0n = p0

g + �.(20)

Consequently, (expected) profit-maximi-zing middlemen will handle Qn only with theintention of selling it as (certified) non-GM toconsumers, provided p1

n > p1g. But owing to the

government-set standard R, only a fractionF(R) of non-GM farm product is expected tobe sold as non-GM product to the consumer,and the remaining fraction [ 1 − F(R) ] thatexceeds the threshold level R must be soldat the lower consumer GM product price.Hence, the marketing sector’s profits �M

g and

�Mn for handling GM and non-GM products,

respectively, are

�Mg = (

p1g − p0

g − �)Qg(21)

�Mn = (

p1n F(R) + p1

g[1 − F(R)]

− p0n − � − �

)Qn.

(22)

In a competitive equilibrium, the marketingsector’s expected profit vanishes. Thus, if �M

g =�M

n = 0, in an equilibrium with both productsproduced in strictly positive amounts, it mustbe that

p1g = p0

g + �(23)

p1n F(R) + p1

g[1 − F(R)]

− p0n − � − � = 0.

(24)

By using (20) and (23), the equilibrium condi-tion in (24) can be rewritten as

p1n = p1

g + � + �

F(R).(25)

Hence, in an equilibrium with positive pro-duction of both GM and non-GM products,the four prices of our model are linked bythree arbitrage relations. Equation (20) spec-ifies that the farm-level price premium fornon-GM product must exactly equal the cost-efficiency gain � provided by the new GM crop.Equation (23) specifies that, for GM product,the difference between the consumer price andthe farm price must exactly equal the unit mar-keting cost �. And, equation (25) links the

retail price premium (p1n − p1

g) to the effec-

tive unit efficiency handicap (� + �) (which in-cludes the IP cost �) via the standard-dictatedamount F(R), which is the fraction of farm-level non-GM product that can actually be soldas such to consumers.

Given the three price arbitrage relations justdiscussed, and for a given level of the regula-tion standard R, the equilibrium value of theremaining price, p0∗

g , solves the market equi-

librium condition that is obtained by equatingtotal supply with total demand. Several possi-ble situations may arise. In particular, as is stan-dard in models of this type with a finite numberof heterogeneous consumers, it is necessary todistinguish the case in which all consumers buysome variety of the good (the market is “cov-ered”) from the case in which some consumersdo not buy either variety (the market is said tobe “uncovered”). Furthermore, here we needto distinguish the case in which both varieties(GM and non-GM) are provided from that inwhich only one variety is provided.10

Both GM and non-GM products are pro-duced and consumed when, for the given R,

the equilibrium price is such that 0 < � <Min{�, 1}. This situation can, of course, takethe form of either an uncovered market (i.e.,in equilibrium � < 1) or of a covered market(i.e., in equilibrium � ≥ 1). The most interest-ing equilibrium situation, for our purposes, iswhen both goods are produced and the mar-ket is uncovered, in which case the equilibriumprice p0∗

g solves

S(

p0∗g

) = H

(1

as(R)

[u − p0∗

g − � − (� + �)

F(R)

]).

(26)

For the covered market case, the equilibriumprice simply solves S(p0∗

g ) = 1.11 The other pos-

sible equilibria in this model include the casesQg = 0 (i.e., no GM product is produced) andQn = 0 (i.e., no non-GM product is produced).Because farmers have a ceteris paribus incen-tive to produce the (more efficient) GM prod-uct, and because consumers with low enough� will always want to consume this productwhen p1

n > p1g, it is clear that the possibility

of Qg = 0 in equilibrium can safely be ignored

10 Assumption 1 rules out the case of a covered market if regula-tions are such that only GM goods are sold.

11 Since each consumer buys, at most, one unit, when all con-sumers participate, total demand is equal to the mass of consumers,which we have normalized to equal one.

Lapan and Moschini Grading, Minimum Quality Standards, and GM Labeling 777

(given standard regularity conditions on thedistribution of consumer types). We similarlycan ignore the possibility that, in equilibrium,neither of the two goods is produced. The equi-librium with Qn = 0, on the other hand, in ourmodel is a real possibility, and we analyze thatnext.

Regulatory Standard and EquilibriumOutcomes

We have already analyzed the case in whichnon-GM goods were not supplied to the mar-ket because there was no labeling standard thatallowed consumers to differentiate them fromthe (weakly) inferior GM good. However, thepresence of a labeling standard is not sufficientto guarantee that the non-GM good will besupplied in equilibrium. In fact, a regulatorystandard that is too stringent is just as bad, forthe purpose of bringing diversity to the mar-ketplace, as a nonexistent system.

PROPOSITION 1. There exists a standard levelR¯∈ (0, 1] such that, in equilibrium, for R < R

¯the

non-GM product is not supplied to the market,whereas for R > R

¯both goods are marketed in

equilibrium.

A detailed proof of this result is provided inthe Appendix. A critical element of the proofis that F(0) = 0, i.e., perfect purity is not at-tainable ex ante. This condition reflects theoften-made argument, by agricultural and foodindustry operators who deal with the emergingGM regulation, that zero tolerance is not possi-ble. Proposition 1 thus provides the importantpolicy conclusion that GM labeling standardsmay go too far. Setting a standard that is toostrict (i.e., a threshold level R that is too low)may not help the consumer at all if the equi-librium outcome is that no non-GM productis supplied. The root of this result, of course,is that providing increased levels of purity isincreasingly costly in this setting.

Effect of the Purity Standardon Farmers’ Returns

In the rest of this article we consider the sce-nario in which it is feasible to set the regulatorystandard such that both goods will be marketedand the regulator chooses to do so:

ASSUMPTION 2. �(1) ≡ � + �

a[1 − s(1)]

< �0,e and R > R¯.

Even though we have previously assumed thatwithout labeling the market was uncovered,it is possible that with labeling there may bea regulatory standard that would increase de-mand sufficiently so that the market would becovered. To illustrate, consider the personal-ized price for the non-GM good of the highesttype (� = 1):

p1n(1) = p1

n + as(R) ≡ p0g + � + V (R)(27)

where

V (R) ≡ � + �

F(R)+ as(R).(28)

Given R > R¯

, it follows that we knowp1

n(1) < p1g(1). It is also readily shown that there

is a unique interior value of R that minimizesV(R). Hence, define:

�

R≡ arg minR∈[0,1]

V (R).(29)

Thus,�

R represents the best standard for themost GM-averse consumer.

PROPOSITION 2. If u − V (�

R) > C ′(1) + � >u − a, then there exist values of R such that themarket is covered when labeling occurs but isuncovered otherwise.

To derive this result, recall that C′(1) is thesupply (farm-gate) price needed to produceenough output to cover the market, and � isthe processor’s handling cost for the GM prod-uct. By Assumption 1, the second inequalityholds. The personalized supply price for type� =1, when the market is covered and the regu-latory standard R is used, is (C′(1) + �+ V(R)).If this price is less than consumers’ willingnessto pay (u), then all agents of type � ≤ 1 face apersonalized price less than their willingness topay, and in equilibrium, the market will be cov-

ered. If [u − V (�

R)] < [C ′(1) + �], then there isno R that can lead to a covered market.

In what follows we focus on the more in-teresting case in which the market is uncov-ered (so that aggregate demand is downwardsloping). We therefore make the followingassumption:

ASSUMPTION 3. C ′(1) + � + V (�

R) > u.

Thus, for any R ∈ (R¯

, 1 ] consumers in the inter-

val [0, �(R)] will buy the GM good, consumers

778 August 2007 Amer. J. Agr. Econ.

in the interval [�(R), �(R, p0g)] will buy the

non-GM good, and those in [�(R, p0g), 1] will

purchase neither good.From the earlier specification of farm-level

production, aggregate supply S(p0g) is an in-

creasing function of p0g, and so producer

surplus is also an increasing function of p0g.

Relaxing the purity standard (i.e., increasingR) lowers the equilibrium market price pre-mium for the non-GM product, which, becauseof (25), satisfies p1

n − p1g = (� + �)/F(R). This

effect will have offsetting impacts on total de-mand and hence on the equilibrium farm pricep0∗

g . An increase in R has an ambiguous im-

pact on demand: by lowering the premium, itincreases demand, but by lowering the puritylevel, it decreases demand if it raises the per-sonalized price for the marginal (high � type)buyer. The overall impact will depend on theindividual weights put on purity and on thestandard R. Thus, the equilibrium farm priceshould be a nonmonotonic function of R. Infact, we have the following result.

PROPOSITION 3. In the uncovered marketcase, there exists a critical standard level Rps

such that below this level farmers gain from re-laxing the standard while above this level farm-ers lose.

To derive this result we note that, givenAssumptions 1–3, the market will be uncov-ered, but positive non-GM sales will occur forR > R

¯. Furthermore, since producer surplus

is increasing in p0g, it is clear that producers

want to maximize p0g, which entails choosing

the standard that maximizes sales. The equi-librium price p0∗

g (R) is determined from

Z(

p0∗g , R

) ≡ S(

p0∗g

) − H(�(R, p0∗

g

)) = 0(30)

where

�(R, p0

g

) ≡(u − p0

g − �) − �+ �

F(R)

as(R).

As noted, S(p0g) = �′(p0

g), where �(p0g) is total

farmers’ profits. Further,

dp0∗g

dR= −∂ Z/∂ R

∂ Z/∂p0∗g

(31)

where ∂ Z/∂p0∗g > 0 (supply and demand have

their conventional slopes). Also:

∂ Z

∂ R= −h

(�(R, p0∗

g

)) ∂�

∂ R(32)

where

∂�

∂ R≡ f (R)

F(R)s(R)

[�(R)(1 − s(R))

− �(R, p0

g

)(R − s(R))

].

At (R¯, p0,e

g ), �(R¯

) = �(R¯, p0,e

g ) = �0(R¯, p0,e

g ),

so that (∂�/∂ R)|R¯

> 0 ⇒ (∂p0,∗g /∂ R)|R

¯> 0.

Thus, demand and equilibrium price are in-creasing in R at R

¯. However, at R = 1,

∂�

∂ R

∣∣∣∣R=1

≡ f (1)

s(1)

[�(1)(1 − s(1))

− �(1, p0

g

)(1 − s(1))

]< 0.

(33)

Hence, the regulatory standard RPS that max-imizes producer surplus must satisfy R P S ∈(R

¯, 1). Indeed, the foregoing analysis estab-

lishes that RPS solves

�(RPS)(1 − s(RPS))

− �(RPS)(RPS − s(RPS)) = 0.

(34)

Thus, Proposition 3 establishes the interest-ing conclusion that some regulation, in theform of a MQS defining what can be identifiedas non-GM, may be desirable from the produc-ers’ perspective, even though in any equilib-rium with positive production of both goods,farmers are actually indifferent as to whichgood to produce. The quality standard herehelps to exploit optimally consumers’ prefer-ence for product differentiation. A corollaryto this result is that the absence of a standard(or, equivalently, R = 1) is generally not in theinterest of agricultural producers.

Welfare Effect of the Purity Standard

In the model that we have developed there isno profit at the marketing level, because mar-keting services are provided at constant unitcosts (i.e., by a constant returns-to-scale indus-try), but the purity standard has the potentialto affect the welfare of farmers and of finalconsumers. Having discussed the qualitativeimpact on producer surplus in the precedingsection, let us now turn to aggregate welfare.Summing producer and aggregate consumersurplus, for a given pair (p0

g, R) the welfare

function is

Lapan and Moschini Grading, Minimum Quality Standards, and GM Labeling 779

W(

p0g, R

) = �(

p0g

) +∫ �

0

Ug(

p0g, R

)h(�) d�

+∫ �

�

Un(

p0g, R

)h(�) d�

(35)

where �(p0g) is producer surplus; that is,

�(

p0g

) ≡∫ p0

g

0

S(p) dp(36)

and the individual utility functions that enteraggregate consumer surplus are

Ug(

p0g, R

) = u − p0g − � − a�(37)

Un(

p0g, R

) = u − p0g − �

− � + �

F(R)− a�s(R)

(38)

where we have used the arbitrage equilibriumrelations p0

n = p0g + �, p1

g = p0g + �, and p1

n =p1

g + (� + �)/F(R). Let R∗ denote the stan-

dard that maximizes the welfare function in(35). Then R∗ exists by virtue of Weierstrass’stheorem,12 and for an interior solution it sat-isfies ∂W (p0∗

g , R∗)/∂ R = 0, where p0∗g is the

equilibrium price, which therefore must satisfy∂W (p0∗

g , R∗)/∂p0g = 0.13

Differentiating the welfare function in equa-tion (35), we obtain

∂W

∂p0g

∣∣∣∣∣p0∗

g ,R∗= �′(p0∗

g

) − H

(1

as(R∗)

[u − p0∗

g

− � − (� + �)

F(R∗)

])= 0

(39)

∂W

∂ R

∣∣∣∣p0∗

g ,R∗= f (R∗)

F(R∗)

∫ �∗

�∗

(� + �

F(R∗)

− a�(R∗ − s(R∗))

)h(�) d� = 0

(40)

12 The problem involves the maximization of a continuous func-tion defined over a compact set (the unit interval).

13 Note that, for any given R, the competitive equilibrium priceminimizes the sum of producer and consumer surplus, for reasonssimilar to those articulated in Smith (1963).

where the limits of integration satisfy �∗ =�(R∗) and �∗ = �(R∗). Note that the conditionin (39) is equivalent to the equilibrium condi-tion in (26) because Hotelling’s lemma implies�′(p0

g) = S(p0g).

It is possible to characterize this welfare-maximizing standard relative to the standardRPS that is optimal from the producers’ per-spective. In particular, we find the following.

PROPOSITION 4. The welfare-maximizingregulatory level is such that R∗ > RPS. That is,consumers prefer a more lax regulatory stan-dard than do producers.

To derive this result, note that the conditionfor the welfare-maximizing optimal standardR∗ in equation (40) can be rewritten as

J(R∗) ≡∫ �∗

�∗(�(R∗)(1 − s(R∗))

− �(R∗ − s(R∗)))h(�) d� = 0.

(41)

Evaluating this function at the standard RPS

that maximizes producer surplus, which solvesequation (34), we obtain

J(RPS) ≡∫ �

�

(�(RPS) − �)

× (RPS − s(RPS))h(�) d� > 0

(42)

where the limits of integration here are � =�(RPS) and � = �(RPS). This result, togetherwith the fact that the sufficient conditions forwelfare maximization require J′(R∗) < 0, es-tablishes that R∗ > RPS.

The fact that producers prefer a stricterstandard than do consumers may be a bitsurprising, especially in light of the fact thatthe regulation of GM labeling is commonlyunderstood as a response to consumers’ con-cerns. To a certain extent this result reflects thespecial features of the unit demand assump-tion that underlies the Mussa-Rosen modelof demand for quality (but note that our re-sult has been derived without any restrictionon the distribution of consumer types).14 Still,especially when considered along with ourProposition 1, the result in Proposition 4 does

14 Insofar as that is the case, the result is a reminder that theMussa-Rosen model of differentiated demand remains a partic-ular specification (albeit a very useful one) of a general class ofpreferences (Neven 1986).

780 August 2007 Amer. J. Agr. Econ.

provide a further check on the simplistic ap-proach that seems to underpin much of thediscussion concerning GM labeling policies inthe EU and elsewhere, whereby the observa-tion that (some) consumers are averse to thepresence of GM content in food is invoked tojustify very strict regulations. But any reason-able analysis of this situation must grant thepremise that consumers are really heteroge-neous in their aversion to GM content, as re-flected in the model that we have developedhere. Hence, the notion of consumers’ welfarein this setting is really about some averageconsumer (see, e.g., Spence 1975). In par-ticular, in moving from RPS to R∗, not allconsumers benefit from such a relaxation ofthe purity standards, although aggregate con-sumer surplus does increase.

The reason producers prefer a tighter stan-dard than do consumers (on average) maybe understood as follows. As previously dis-cussed, an increase in R has two offsetting ef-fects. It increases the number of lots accepted,as raising R lowers (by the same amount foreverybody) the actual market price of the non-GM good. At the same time, this looseningof standards raises the expected GM contentof approved goods, which hurts high � typesthe most. Since each person buys, at most,one unit, the producer wants to choose R tomake the highest possible � type participate;thus, at RPS the standard is chosen so that,for this highest participating type �, the twoeffects offset and therefore the personalizedprice for type � is minimized. But this implies

that every consumer of type � ∈ (�, �) wouldbenefit by an infinitesimal weakening of thestandard; hence, R∗ > RPS. At the consumeroptimum, a marginal increase in R would ben-efit some consumers and hurt others, so that av-erage consumer surplus would be unchanged.At the welfare optimum, of course, aggregateconsumer surplus is still increasing in R, whileproducer surplus is decreasing in R.

It is also useful to note that the result inProposition 4 is distinct from a seemingly sim-ilar finding of the MQS literature, where italso emerges that a standard collectively set byproducers may be too strict. That result actu-ally reflects the exercise of monopoly power,whereby returns to producers may increasewith reduced marketed quantities, and in suchmodels one of the effects of a stricter standardis in fact that of reducing the quantity sup-plied by producers (e.g., Leland 1979). Here,by contrast, competitive conditions are main-tained and producers supply both goods to themarket.

Comparative Statics of Equilibrium

Having characterized the optimality condi-tions for the optimal purity standard, we wishto investigate how this level is affected by someof the model’s critical parameters. To get un-ambiguous results, however, here we have torestrict the analysis by assuming that the dis-tribution of consumer types H(�) is uniform,which means that h(�) = 1 , ∀� ∈ [0, 1].15

PROPOSITION 5. Assuming uniform distribu-tion of types, the welfare-maximizing puritystandard R∗ and equilibrium price p0∗

g satisfythe following comparative statics properties:

(i) ∂ R∗/∂� > 0 and ∂p0∗g /∂� < 0;

(ii) ∂ R∗/∂� > 0 and ∂p0∗g /∂� < 0;

(iii) ∂ R∗/∂a < 0 and ∂p0∗g /∂a < 0;

(iv) ∂ R∗/∂u < 0 and ∂p0∗g /∂u > 0; and

(v) ∂ R∗/∂� > 0 and ∂p0∗g /∂� < 0.

Details of the proof are given in Lapanand Moschini (2006). Thus, the optimal puritylevel R∗ should be sensitive to the costliness ofthe required segregation activities (even if, asin our formulation, the unit segregation cost� is itself independent of the required pu-rity level set by the government). Specifically,as segregation becomes costlier (� increases)the optimal impurity level R∗ increases (andthe equilibrium farm price declines). Exactlythe same qualitative effects apply to an in-crease in the size of the efficiency gains dueto the GM innovation (an increase in �). Onthe other hand, as the population of consumersbecomes more averse to GM content (as mea-sured by an increase in the parameter a), theoptimal standard R∗ should be tightened (andthe equilibrium farm price is reduced becauseof the negative demand effect).

An increase in the base willingness to payfor the two goods (the parameter u) obvi-ously increases the equilibrium farm price andshould also result in a tightening of the optimalstandard R∗. This comparative statics effectis amenable to an interesting interpretation ifwe note that, in this context, the parameter uis inversely related to the (absolute) value ofthe elasticity of total demand. Hence, this re-sult suggests that tighter purity standards (i.e.,lower R∗) should be associated with foods thathave a more inelastic demand. Finally, the com-parative statics effect of the parameter � (the

15 Whereas the assumption of uniform distribution of types issomewhat restrictive, it is routinely made in papers that study qual-ity with the Mussa-Rosen setup (e.g., Stivers 2003; Fulton and Gi-annakas 2004).

Lapan and Moschini Grading, Minimum Quality Standards, and GM Labeling 781

unit costs of marketing services) illustrates theintuitive conclusion that an increase in the costof marketing depresses the equilibrium farmprice and should also result in a laxer optimalstandard R∗.

Concluding Remarks

In this article, we developed a framework ofanalysis for a critical economic issue that arisesin the pursuit of a credible system of IP and la-beling for GM and non-GM products, namely,the setting of a standard for non-GM prod-ucts. The model represents three stages inthe supply chain: farm production, market-ing handlers, and final users. The possibility ofaccidental commingling of non-GM productsis modeled at the marketing stage. Regulationtakes the form of a threshold level of purity fornon-GM products. Uncertainty is modeled ex-plicitly, and the equilibrium solution relies ona demand specification of (vertically) differen-tiated GM and non-GM products.

Despite its simplicity, the rigorous charac-terization of the model required a fair amountof formal analysis. The results that we obtainedare quite interesting, however, and show thateven in a competitive setting where agentshave no scope for strategic behavior, govern-ment regulation of the labeling of GM prod-ucts still presents a meaningful problem. Weshowed that there exists a welfare-maximizingstandard for products that claim a non-GMstatus, and this welfare standard has intuitivecomparative statics properties. In particular,the lack of any standard leads to a poolingequilibrium whereby only the GM product isproduced, which is typically suboptimal froma welfare perspective. Similarly, a standardthat is too strict (i.e., high purity of the non-GM product) may also lead to a collapse ofthe market for the non-GM product. In ad-dition, we showed that the labeling standardthat is optimal from society’s viewpoint typi-cally differs from the standard that would bepreferred by farmers. Somewhat surprisingly,the standard that farmers would prefer is actu-ally stricter than what society would find opti-mal. At a minimum, this result suggests that theopposition to GM labeling displayed by someproducers’ organizations may be misplaced;it certainly suggests that support for such aposition ought to be sought outside the fac-tors considered in the market analysis of thisarticle.

Having characterized the qualitative proper-ties of the GM labeling problem of this article,there remain several interesting open issues.For instance, the issue of calculating the “opti-mal” R in an actual empirical setting was out-side the scope of this article, but of course hasimmediate interest from a policy perspective.Whether or not the EU standard of R = 0.009 ispreferable to Japan’s standard of R = 0.05, forexample, remains an unsettled issue. The anal-ysis presented in this article hopefully will pro-vide opportunities for empirical applicationsand extensions.

A major hurdle in the empirical implemen-tation of the analysis of this article is likely toconcern the quantification of consumers’ ac-tual preferences vis-a-vis GM products. Mar-keting surveys and experimental methods lookpromising in this setting (see Lusk et al.[2005] for an interpretative review and sum-mary of existing results). Noussair, Robin, andRuffieux (2004) elicit consumers’ willingnessto pay for food products that only differ intheir amount of GM content. Their resultsare broadly supportive of the assumption ofmodeling context pursued here. Consumers doturn out to have heterogeneous preferenceswith respect to GM content: about one-quarterof them are substantially indifferent, and theremainder displays demand that is decreas-ing in the GM content.16 Conversely, on thesupply side it is abundantly clear (consistentwith the structure of our model) that supply-ing non-GM products entail segregation and IPcosts throughout the production and handlingsystem that are directly related to the strin-gency of the purity level (Bullock and Desquil-bet 2002; Huygen, Veeman, and Lerohl 2003).Translating such qualitative and quantitativeelements in a coherent system-wide assess-ment of the welfare impacts of alternative GMpurity standards will also require the specifica-tion and calibration of a number of other re-lations in an explicit differentiated productionand marketing system (e.g., Moschini, Bulut,and Cembalo 2005). Such extensions and ap-plications of our analysis are left for futurework.

[Received January 2006;accepted October 2006.]

16 Rousu et al. (2004) also found that consumers’ mean bid differ-entials are higher with 1% GM products than 5% GM products,although not statistically different (this lack of statistical signifi-cance is likely due to their small sample size).

782 August 2007 Amer. J. Agr. Econ.

References

Akerlof, G.A. 1970. “The Market for Lemons: Qual-

ity Uncertainty and the Market Mechanism.”

Quarterly Journal of Economics 84:488–500.

Bockstael N. 1984. “The Welfare Implications of

Minimum Quality Standards.” American Jour-nal of Agricultural Economics 6:466–71.

——. 1987. “Economic Efficiency Issues of Grad-

ing and Minimum Quality Standards.” In R.L.

Kilmer and W.J. Armbruster, eds. EconomicEfficiency in Agricultural and Food Marketing.

Ames IA: Iowa State University Press, chapter

12.

Bullock, D., and M. Desquilbet. 2002. “The Eco-

nomics of Non-GMO Segregation and Identity

Preservation.” Food Policy 27:81–99.

Carter, C., and G.P. Gruere. 2003. “International

Approaches to the Labeling of Genetically

Modified Foods.” Choices, Second Quarter.

Crespi, J.M., and S. Marette. 2003. “‘Does Contain’

vs. ‘Does Not Contain’: Does It Matter Which

GMO Label Is Used?” European Journal ofLaw and Economics 16:327–44.

Dimitri, C. 2003. “Agricultural Marketing Institu-

tions: A Response to Quality Disputes.” Jour-nal of Agricultural & Food Industrial Organi-zation 1, article 17.

European Union. 2004. “Question and Answers

on the Regulation of GMOs in the EU.”

MEMO/04/85, Brussels, 15 April.

Evenson, R.E., and V. Santaniello. 2004. ConsumerAcceptance of Genetically Modified Foods.

Wallingford, UK: CABI Publications.

Falck-Zepeda, J.B., G. Traxler, and R.G. Nelson.

2000. “Surplus Distribution from the Introduc-

tion of a Biotechnology Innovation.” AmericanJournal of Agricultural Economics 82:360–69.

Fulton, M., and K. Giannakas. 2004. “Inserting GM

Products into the Food Chain: The Market and

Welfare Effects of Different Labeling and Reg-

ulatory Regimes.” American Journal of Agri-cultural Economics 86:42–60.

Gaisford, J.D., J.E. Hobbs, W.A. Kerr, N. Perdikis,

and M.D. Plunkett. 2001. The Economicsof Biotechnology. Cheltenham, UK: Edward

Elgar.

Gardner, B. 2003. “U.S. Food Quality Standards: Fix

for Market Failure or Costly Anachronism?”

American Journal of Agricultural Economics85:725–30.

Giannakas, K., and M. Fulton. 2002. “Consumption

Effects of Genetic Modification: What if Con-

sumers Are Right?” Agricultural Economics27:97–109.

Huygen, I., M. Veeman, and M. Lerohl. 2003. “Cost

Implications of Alternative GM Tolerance

Levels: Non-Genetically Modified Wheat in

Western Canada.” AgBioForum 6:169–77.

Lapan, H., and G. Moschini 2004. “Innovation and

Trade with Endogenous Market Failure: The

Case of Genetically Modified Products.” Amer-ican Journal of Agricultural Economics 86:630–

44.

——. 2006. “AJAE Appendix: Grading, Min-

imum Quality Standards, and the Label-

ing of Genetically Modified Products.” Un-

published manuscript. Available at http://

agecon.lib.umn.edu

Leland, H. 1979. “Quacks, Lemons and Licensing: A

Theory of Minimum Quality Standards.” Jour-nal of Political Economy 87:1328–46.

Lusk, J., M. Jamal, L. Kurlander, M. Roucan, and

L. Taulman. 2005. “A Meta-Analysis of Genet-

ically Modified Food Valuation Studies.” Jour-nal of Agricultural and Resource Economics30:28–44.

Moschini, G., H. Bulut, and L. Cembalo. 2005.

“On the Segregation of Genetically Modified,

Conventional and Organic Products in Euro-

pean Agriculture: A Multi-market Equilibrium

Analysis.” Journal of Agricultural Economics56:347–72.

Moschini, G., H. Lapan, and A. Sobolevsky. 2000.

“Roundup Ready Soybeans and Welfare Ef-

fects in the Soybean Complex.” Agribusiness16:33–55.

Mussa, M., and S. Rosen. 1978. “Monopoly and

Product Quality.” Journal of Economic Theory18:301–17.

Neven, D. 1986. “‘Address’ Models of Differen-

tiation.” In G. Norman, ed. Spatial Pricingand Differentiated Markets. London, UK: Pion,

pp. 5–18.

Noussair, C., S. Robin, and B. Ruffieux. 2004. “Do

Consumers Really Refuse to Buy Genetically

Modified Food?” Economic Journal 114:102–

20.

Pew Initiative. 2005. U.S. vs. EU: An Examination ofthe Trade Issues Surrounding Genetically Mod-ified Food. Pew Initiative on Food and Biotech-

nology, December.

Rousu, M., W.E. Huffman, J.F. Shogren, and A.

Tegene. 2004. “Are United States Consumers

Tolerant of Genetically Modified Foods?” Re-view of Agricultural Economics 26:19–31.

Shapiro, C. 1983. “Premiums for High Quality Prod-

ucts as Returns to Reputation.” Quarterly Jour-nal of Economics 98:659–79.

Smith, V.L. 1963. “Minimization of Economic Rent

in Spatial Price Equilibrium.” Review of Eco-nomic Studies 30:24–31.

Spence, M. 1975. “Monopoly, Quality and Regula-

tion.” Bell Journal of Economics 6:417–29.

Lapan and Moschini Grading, Minimum Quality Standards, and GM Labeling 783

Stivers, A. 2003. “Quality Standards with Exoge-

nously Distributed Quality.” Economic Letters80:131–136.

Tirole, J. 1988. The Theory of Industrial Organiza-tion. Cambridge, MA: The MIT Press.

Appendix

Proof of Proposition 1

Let �0,e denote, as earlier, the marginal buyer when

there is no labeling. For a given R, let �(R) denotethe consumer who is indifferent between the GMand non-GM good (i.e., p1

n(�) = p1g(�)). If �(R) <

�0,e, then introducing labeling, with this standard,will result in an equilibrium with both goods mar-keted and with a higher farm price. The latter con-

clusion follows because �(R) < �0,e ⇒ � > �0,e sothat, at the original farm price level, not only willsome non-GM good be sold, but more types willwant to buy the product. Hence, the introductionof labeling leads to excess demand at p0,e

g , and

price must rise to clear the market. However, if

�(R) ≥ �0,e, then allowing labeling (with the stan-dard R) will not affect the equilibrium outcomesince the introduction of the labeled good will alterneither quantity demanded nor quantity supplied at

that price. The equilibrium is unique, since supplyis positively sloped and demand is negatively (ornot positively) sloped. Given the personalized pricedefinitions in equations (7) and (8), and the marketarbitrage relations in (20), (23), and (25), it is readilyverified that

�(R) = � + �

aF(R) [1 − s(R)].(A.1)

It is also clear that �(R) is a decreasing function ofR because

�′(R) = − f (R)

F(R)�(R)

(1 − R

1 − s(R)

)< 0(A.2)

where we have used

s/(R) = f (R)

F(R)[R − s(R)] > 0.(A.3)

Now, if �(1) ≥ �0,e, then no consumer exists thatwould buy the non-GM product. Otherwise, be-

cause we have shown that �(R) is a decreasing

function of R, and because limR→0 �(R) → ∞ (asis apparent from equation (A.1)), there exists a R

¯∈ (0, 1) such that �(R¯

) = �0,e. Hence, if R ∈ [0,R¯

),

�(R) < �0,e, introducing the labeled good with astandard in this interval will result in the same equi-librium as when no labeling occurs.