farmers behavioral

8/4/2019 Farmers Behavioral

1/13

Agricultural & Applied Economics Association

The Role of Farmers' Behavioral Attitudes and Heterogeneity in Futures Contracts UsageAuthor(s): Joost M. E. Pennings and Raymond M. LeutholdSource: American Journal of Agricultural Economics, Vol. 82, No. 4 (Nov., 2000), pp. 908-919Published by: Blackwell Publishing on behalf of the Agricultural & Applied EconomicsAssociationStable URL: http://www.jstor.org/stable/1244529

Accessed: 28/08/2009 11:28

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

http://www.jstor.org/action/showPublisher?publisherCode=black.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the

scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that

promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

Blackwell Publishing andAgricultural & Applied Economics Association are collaborating with JSTOR to

digitize, preserve and extend access toAmerican Journal of Agricultural Economics.

http://www.jstor.org
http://www.jstor.org/stable/1244529?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/1244529?origin=JSTOR-pdf


2/13

THE ROLE OF FARMERS' BEHAVIORALATTITUDES AND HETEROGENEITY

IN FUTURES CONTRACTS USAGEJOOST M.E. PENNINGS AND RAYMONDM. LEUTHOLD

The relationship between farmers' behavioral attitudes and use of futures contracts is examined,taking into account non-directly observable variables and the heterogeneity of farmers.The relation-ships are tested on a stratified data sample of 440 farmers. Cluster analysis and covariance structureequation models are used to validate the relationships. Farmers are found not to be homogenousregarding the factors influencing their use of futures. Heterogeneity at the segment level maskedimportant effects at the aggregate level, notably risk attitude. Furthermore, several psychologi-cal constructs for farmers related to market orientation, risk exposure, market performance andentrepreneurial behavior play important roles in their use of futures contracts.Key words: covariance structure model, measurement error, segments, futures usage.

Developing and maintaining viable agri-cultural futures contracts is an expensiveand time-consuming process.1 Therefore, itseems valuable to gain insight into theJoost M.E. Pennings s an associateprofessorat WageningenUniversity in the Departmentof Marketing& ConsumerBehaviorin the Netherlandsand a visiting professorat theOffice for Futures and Options Research at the Universityof Illinois at Urbana-Champaign.aymondM. Leuthold isthe ThomasA. HieronymusProfessor n the DepartmentofAgriculturaland ConsumerEconomicsat the UniversityofIllinois at Urbana-Champaign.he authors are grateful forthe generousparticipation f the 440 farmers n the personalcomputer-assistednterviews. inancial upportprovidedby theAmsterdamExchanges AEX), ChicagoMercantileExchange,Foundationor Research n AgriculturalDerivatives,he Officefor FuturesandOptionsResearch, nd the Niels StensenFoun-dationmade t possible o conduct he large-scalenterview. heauthorswould like to thankJ.A. Bijkerkfor buildinga user-friendly nterface or the computer-assistedersonal nterviews.The authorsexpressspecial thanksto W. Brorsen,M. Candel,C.Ennew,P.Garcia,F terHofstede,S.Irwin,M.T.G.Meulenberg,M. Rockinger,F. Verhees,A. Smidts,J-B.E.M.Steenkamp,B. Wierengaand the participants f the 1999 NCR-134meet-ing held at the ChicagoMercantileExchange or helpfulcom-mentson the researchprojectand preliminary ersionsof thismanuscript.'With an almostexponentialgrowth n the last decadeand2.2 billion futuresand optionscontracts radedthroughoutheworld n 1998,competitions stiffin the financial ervices ndus-try (FuturesIndustryAssociation).The derivatives ndustry scomposedof competing irms uchas exchanges, anksandbro-keragehousesofferingandfacilitating ver-the-counterrading.Withcommodityderivatives,he risk of failure s considerable(Carlton).n theperiod1994-1998 totalof 140newcommodityderivativeswere introduced round he world.Twelveof thesederivativeswere de-listedwithin hisperiod.TheLondonInter-nationalFinancial utures& OptionsExchange nd the ChicagoMercantileExchangewere the leaders n introductions ithfif-teen andfourteen, espectively, uring1994-98. f we followthecriteria ormulated y Silber, ifty-eight ercentof the introduc-tionshave failed.

characteristics of farmers who use futuresmarkets. In the finance literature, several fac-tors such as a firm's risk exposure, its growthopportunity, level of wealth, managerial riskaversion, financial distress costs, tax liabil-ity, and the accessibility to financing havebeen identified as influencing the decision ofa corporation to adapt derivatives to theirrisk management toolbox (Geczy, Minton,and Schrand; Graham and Smith; Koski andPontiff; Mian; Nance, Smith, and Smithson;Smith and Stulz; Tufano). In the agriculturaleconomics literature Asplund, Foster, andStout; Goodwin and Schroeder; Makus et al.;Musser, Patrick, and Eckman; Shapiro andBrorsen; and Turvey and Baker, among oth-ers, identified such factors as experience, edu-cation, farm size, off-farm income, expectedincome change from hedging, age, farm orga-nization meetings, leverage, risk management,and marketing seminar participation, as influ-encing farmers' use of futures contracts.These studies have increased our under-standing of how firm characteristics influencethe use of futures and hence the viabilityof futures trade. In this paper, we recog-nize that farmers make decisions based on,among others, their beliefs that are formed byperceptions.2For example, the perceived riskreduction performance may differ from the

2 Beliefs pertain to the degree to which an object (e.g., futurescontracts) may have particular consequences, and perceptionsreflect the interpretation of these consequences.Amer. J. Agr. Econ. 82(4) (November 2000): 908-919Copyright 2000 American Agricultural Economics Association


3/13

Farmers' Heterogeneity and Futures Contracts Usage 909actual performanceas reflected in hedgingeffectiveness measures. Moreover, farmersmay very well evaluate the hedging serviceprovided by futures exchanges using crite-ria other than just performance.Therefore,we takepsychological onstructs nto account(Thaler).3An importantvariablefrom a theoreticalpoint of view related to hedgingbehaviorisrisk attitude.Some studies assume that man-agersare risk averse.However,riskattitudesmay differ across managers.Brockhaus andMarch and Shapirafound large differencesin risk attitudesamongmanagersof corpora-tions and farmers.Elsewhere,Goodwin andSchroeder found that farmerswith a statedpreferencefor risk are more likely to adoptforwardpricing(i.e., hedge) than risk-aversefarmers,a puzzlingresult.One of the reasonsfor these contra-intuitive indings s the dif-ficultyin measuringrisk attitudes.Variablessuch as risk attitude are not alwaysdirectlyobservable (so-called latent variables).4Weshow that farmersmake decisionsregardingthe use of futures based on, among others,latent variables.We refer to these latent vari-ables as attributesas well to indicate thatthese variables can be seen as the criteriafarmersuse when decidingto use futures.Two empirical problems may arise whentaking latent constructsinto account.First,we have to make sure that we have reliableand valid constructs.5We thereforemeasure

3The present study focuses on farmers, that is, owner-managersof small and medium-sized enterprises. An important differencebetween farmers and managers of a large enterprise lies in thefact that farmers do not have different functional departments,such as research and development, manufacturing-quality con-trol, sales and accounting. All these departments are combinedwithin the farmer. The decision process in the farmer's case isnot as rationalized as that of large enterprises. Some of the con-cepts used by farmers might be psychological constructs (suchas 'level of understanding') that are not directly measurable,and therefore remain absent in accounting data used in recentstudies about managers in large companies (Geczy, Minton, andSchrand). These psychological constructs may very well play apart in the farmer's decision to use futures.4A latent variable is a hypothesized and unobserved conceptthat can only be approximated by observable or measured vari-ables (indicators). An example of an unobserved concept (in thispaper also referred to as a construct) is farmers' risk attitudeand farmers' market orientation. In this paper, we use a set ofindicators, obtained in personal computer-guided interviews, tomeasure latent variables. In the psychometric and statistical liter-ature, methods are developed to test the reliability and validityof such variables (Nunnally and Bernstein).5Reliability refers to the extent to which a variable or set ofvariables is consistent with what it is intended to measure. Valid-ity refers to the extent to which a measure or set of measurescorrectly represents a concept (i.e., latent variable). Validity isconcerned with how well the concept is defined by the measures(i.e., indicators), while reliability relates to the consistency of themeasures. Ensuring validity starts with a thorough understand-ing of what is to be measured and taking the measurement as

latent variablesby a set of observable ndica-tors that are subjected o confirmatoryactoranalysisto assess theirproperties.Confirma-toryfactoranalysispermitsa rigorousassess-ment of the propertiesof the latent variables(Hair et al.). Second, relationshipsbetweenand amonglatent theoreticalconcepts (con-structs)that are not directlyobservablemayresult in biased coefficients when estimatedin a linear regression framework becauseof measurementerror.To account for mea-surementerror,a covariance tructuremodelis used as framework(often referred to asstructuralequation modeling),as it permitsthe explicitmodelingand estimationof errorsin measurement Bagozzi;BaumgartnerandHomburg;Bollen;Steenkampand vanTrijp).Often, the literature assumes enterprisesto be homogenous regardingfirms' use offutures. When estimating models, data aretreated as if they were collected from a sin-gle population.This assumptionof homo-geneity might be unrealistic. For example,farmers of different size or from differentregions may have different decision struc-tures. Hence, pooling data across farmersmightproducemisleading esults.Both issues,measurement error when using latent vari-ables and farmers'heterogeneity,willbe elab-orated on.The remainderof the paper is structuredas follows. Initially,we introduce a behav-ioral framework n which the characteristicsthat might be associated with futuresusageare discussed.Then we address how unob-servable constructs can be measured andtested for reliability using a confirmatoryfactor analysis framework, ollowed by thespecificationof a covariancestructuremodelthat simultaneously stimates the latent vari-ables from observedvariablesand the struc-tural relations between these variables andthe farmers'use of futures contracts.Afterthe researchmethod and the operationaliza-tion of the variables,differentrelationshipsbetween farmers'characteristics nd use offutures are estimated. Data obtained from440 farmersby means of computer-assistedpersonal interviewsconstitute the input forthispartof the research.Weconcludewith an"correct" and accurate as possible. However, accuracy does notensure validity.For example, the researcher could very preciselydefine total household income but still have an invalid measureof discretionary income because the "correct" question was notbeing asked. Reliability is the degree to which observed variablesmeasure the "true" value. If the same measure is asked repeat-edly, more reliable measures will show greater consistency thanless reliable measures.

Pennings and Leuthold


4/13

910 November 2000evaluation of the study and make suggestionsfor further research.

A Behavioral Conceptual FrameworkOne of the pillars in behavioral choice mod-els is the multi-attribute attitude theory, firstintroduced by Fishbein and Azjen. In thistheory the attitude towards an object (e.g.,futures contracts) will lead to the intentionto choose that object (e.g., the probabil-ity of choosing futures) and ultimately to achoice. In this paper we use this frameworkto study farmers' probability of using futuresand farmers' choice of using futures. We dis-cuss some constructs and identify characteris-tics that might influence farmers' probabilityof using futures.In most hedging theories Risk Attitude(RA) plays an important role in the deci-sion to engage in futures contracts (Carter).Recently, Tufano found that managerial riskaversion affects corporate risk managementpolicy in the North American gold miningindustry.Risk must first be perceived before afarmer is able to respond to it. Perceived RiskExposure (PRE) may be defined as a farmer'sassessment of the risk inherent in a situation.We expect these two constructs to play a rolein the farmers' probability of using futures.Tashjian and McConnell have shown thathedging effectiveness is a determinant inexplaining the success of futures contractsand, as a result, considerable attentionhas been paid to the hedging effectivenessof futures contracts (Ederington; Howardand D'Antonio). However, for farmers thePerceived Performance (PP) may differ fromthe performance as reflected by hedgingeffectiveness measures. We expect the farm-ers' perceived performance to influence theiruse of futures.In line with the broader definition of mar-ket orientation proposed by Jaworski andKohli, we consider farmers' efforts to obtaininformation about prices and volume tradeda central element of their Market Orientation(MO). If farmers are more market orientedin terms of tracing the market, we expectthem to be more inclined to use futures as amethod of pricing their products.Farmers often perceive futures as a com-plex financial service, which inhibits partic-ipation in futures trading (Ennew, Morgan,and Rayner). Costs associated with usingfutures include information gathering and the

efficiency of their adoption. A high level ofUnderstanding (UNDER) futures increasesthe ability of farmers to use futures contracts.We expect the level of understanding futuresto be positively related to farmers' use offutures.Sharpiro and Brorsen found that farmerswho are in a high debt position are morelikely to hedge since hedging can increasereturns and/or reduce risk. Furthermore,Turvey and Baker found a clear relation-ship between the use of price risk manage-ment instruments and the Debt-to-Asset Ratio

(DAR) of farmers.Farmers often base their decisions on theopinions of the members of their DecisionUnit (DU), such as spouse, partner and advi-sors. We hypothesize that if a farmer believesthat relevant others expect him/her to makeuse of the hedging service of futures con-tracts, that (s)he will be more inclined to usefutures. Therefore, we included the farmer'sperception of the extent to which signifi-cant others think that (s)he should engage infutures trading.Working provided an alternative explana-tion for farmers' motivation to hedge thathas not yet been addressed empirically: usingfutures gives the farmer a greater freedomfor business action. Working argued that thefreedom gained could be used to make a saleor purchase that would otherwise not be pos-sible. This is in line with the recent findingsin management studies showing that man-agers value instruments that increase their"degrees of freedom of action" in the mar-ket place (e.g., Brandstatter). In the contextof our study, we may expect that farmersvalue using futures as a way to exploit theirEntrepreneurialFreedom (EF).We hypothesize that the constructsreviewed influence farmers' probability ofusing futures and ultimately their choice foror against futures. We do not expect farm-ers to be a homogeneous population, thatis, different variables influence farmers' useof futures and that common variables areweighted differently. Therefore, we segmentacross farmers, such that the choice processis similar within a segment and dissimilaracross segments.

Research Method and Empirical ModelsThe Dutch hog industry is examined empir-ically. It represents a domain in which the

Amer. J. Agr. Econ.


5/13

Farmers' Heterogeneity and Futures Contracts Usage 911

underlyingcommodityis homogeneous,theunderlyingcash market is broad, there aremany participants,and large, unpredictable,pricefluctuations xist.6A personalcomputer-guidednterviewwasdevelopedon the basisof the literature.Priorto the quantitativestudy,we conductedfourgroup discussionswith farmersabout priceriskmanagement.The goal of the groupdis-cussionswas to gain insightinto the processinvolved in selling hogs using futures.Morespecifically,we wanted to gain insight intothe criteria (the so-called attributes)farm-ers use when choosing futures.The groupsconsisted of ten farmers each. The groupdiscussionstook place in an informal atmo-sphereand each session lastedfor abouttwoand a half-hours.From the groupdiscussionsit became clear that the relevant price riskmanagementinstrumentis the hog futurescontract radedon theAmsterdamExchange.It also became clear that a number of cri-teria are used in deciding whether or notto use futures contracts,such as the (per-ceived) risk reductionperformanceand thepossibilityof exercisingentrepreneurialree-dom. The latter meant that futuresare per-ceived as an attractive nstrumentwhenevertheir use increases the degrees of freedomin the marketplace (that is, whenevertheyare perceivedas a tool whichcan be addedto the existingmarketing oolbox and henceincreasethe strategiesfarmerscan employ),confirming he suggestionmade by Workingmore thanforty years ago.The survey was computerized.Care wastaken to build a user-friendlynterface.Wecarefully designed the interview instrumentsuchthat it resembled the farmers'decision-making process within their own businesscontext.To ensure that the computerinter-face was well understoodand perceived as"veryuser-friendly"ndfitting"the realbusi-nesssetting", ifteentest interviewswere con-ducted. The large-scale nterviewtook placein the first half of 1998,on appointment,atthe farmer'senterprise.All the interviewershad prior interviewing experience and hadfollowed an extensive training programforthe assessment procedures.A total of 440farmersparticipated.

6 Dutch hog pricesfluctuatewidely.The coefficientof varia-tion (CV) is 0.19,which s relativelyhighwhencomparedo, forexample,US soybeans CV is 0.14),based on dailyobservationsover the period1990-97.

The interview consisted of several parts.After being asked severalbackgroundques-tions (pertainingto size of the enterprise,age, educationlevel and debt-to-assetratio,where the latter was measured on a 10point scale with 1 = debt-to-asset ratio1-9%, 2 = 10-19% etc.), the farmers wereconfronted with statements about futurescontracts.In selecting the measurementprocedure,we evaluated the different scaling proce-dures as first proposedby Guttman, Likert,and Thurstone and Chave. Because previ-ous scales based on the Likert scale pro-ceduredemonstratedgood performance,heLikert scaling procedurewas used to mea-surefarmers'characteristics. ikertscalesareeasy to constructand can easilybe tested ontheir reliabilityand unidimensionality.Fur-thermore,heydo not sufferfromsome draw-backs of the Guttman,and Thurstone andChave scaling procedures,as identified byNunnallyand Bernstein (see the Appendixfor a descriptionof the scales andthe statis-ticalpropertiesof the scales).The influenceof the decision unitwas mea-sured by asking the farmer to indicate theextent to whichpersonssurrounding im/herthoughtthat (s)he should or should not usefuturesas a hedgingtool by distributing100points acrossusingfuturesas a hedgingtoolor not usingthem as a hedgingtool.The probabilityof using futures contractswas measuredby asking the farmer to dis-tribute 100 points across using futures as ahedging tool or not using them as a hedg-ingtool.7Finally,he use of futureswas basedon past marketactivity,registeringwhetherfarmers had used futures in the last threeyears.8Measuring Unobservable Constructs:A Confirmatory Factor Analytical ModelBecause the unobservedvariables are mea-sured with a set of observable vari-ables (so-called indicators),we adhered tothe iterative procedure recommended byChurchill to obtain reliable and valid con-structs. First, a large pool of questions

7Putte van den, Hoogstratenand Meertensshowed that dis-tributing100 points across alternativesprovidesa more accu-rate measure:t forcesrespondentso make a trade-offbetweenalternatives, hereby not assuming a particularcomparisonmechanism.8 Because we have accounting ndsurveydata,we were ableto distinguishbetween futuresuse for speculativeand for riskmanagement easons.Our measureexclusivelyreflects futuresusage n a hedgingcontext.



6/13

912 November 2000

(i.e., indicators) was generated. The indicatorswere based on the literature available. Carewas taken to tap the domain of the constructas closely as possible. Next, the indicatorswere tested for clarity and appropriateness inpersonally administered pre-tests. The farm-ers were asked to complete a questionnaireand indicate any ambiguity or other diffi-culty they experienced in responding to theindicators, and to make any suggestions theydeemed appropriate. Based on the feedbackreceived from the farmers, some indicatorswere eliminated, others were modified, andadditional indicators were developed. Theresulting set of indicators was administeredto the farmers in the large-scale personalinterview.Confirmatory factor analysis was used toassess the (psychometric) measurement qual-ity of our constructs (Hair et al.). The factoranalytical model assumes that the observedvariables are generated by a smaller num-ber of latent variables (called factors). Therelationship between the observed and latentvariables can be represented by the followingmatrix equation:

(1) x = AK + 8where x is the q x 1 vector of the n sets ofobserved variables (i.e., indicators), K is then x 1 vector of underlying factors, A is theq x n matrix of regression coefficients relat-ing the indicators to the underlying factors,and 8 is the q x 1 vector of error terms of theindicators. Because we wish to develop unidi-mensional constructs, a construct is hypothe-sized to consist of a single factor. The overallfit of the model provides the necessary andsufficient information to determine whether aset of indicators describes a construct. Hence,equation (1) describes a measurement model.In the Appendix the results for the con-firmatory factor analysis are given. All fac-tor loadings (i.e., the regression coefficientsin A in equation (1)) were significant (mini-mum t-value was 4.60, p < 0.001) and greaterthan 0.4. These findings support the conver-gent validity of the indicators (Anderson andGerbing). The composite reliabilities for theconstructs ranged from 0.70 to 0.76, indi-cating good reliabilities for the constructmeasurements (see Appendix). The selectedindicators are used in our structural modeldescribing the relation between the farmers'characteristics and the farmers' probability ofusing futures, and the relation between theexpressed probability of using futures andwhether or not farmers had used futures.

Covariance Structure ModelThe factor model in equation (1) estimatesthe latent variables from the observed vari-ables but does not provide information aboutthe structuralrelationship among these latentvariables and the farmers' probability ofusing futures. Because the latent variablesare measured with error, a covariance struc-ture model is introduced that simultaneouslyestimates the latent variables from observedvariables and estimates the structural rela-tions between these variables and the farm-ers' probability of using futures and thechoice for or against futures.Let qi be a r x 1 vector of the endoge-nous latent variables (i.e., probability of usingfutures and futures usage), and let 5 be as x 1 vector of the exogenous latent vari-ables (i.e, the farmers' characteristics).9Themodel assumes that the variables are relatedby the following system of linear structuralequations:(2) i = B + rF +where B is a r x r matrix of coefficients relat-ing the latent endogenous variables to oneanother, and r is a r x s matrix of coeffi-cients relating the latent exogenous variablesto the endogenous variables. In the equa-tions, s is a r x 1 vector of errors indicat-ing that the endogenous variables are notperfectly predicted by the structural equa-tions. Equation (2) can be written alterna-tively as BI- = rF + S, where B is definedas (I - B). It is assumed that: 1) all vari-ables are measured as deviations from theirmeans: E(rl) = E(S) = E(a) = 0 (this doesnot affect the generality of the model, sincethe structural parameters contained in B andr are not affected by this assumption), 2)none of the structural equations is redun-dant: B exists, and 3) the errors in equationsand the exogenous variables are uncorre-lated: E(Ss') = O.The covariance matrix forthe exogenous variables is: 4- = E(~'), thecovariance matrix for the errors in equationsis: I = E(ss') and the covariance matrix forthe endogenous variables is

COV(ig) =E(n')*-1 ?-1

= B (rDr' + g))B9Note that the variable measuring whether or not farmers usedfutures is not a latent variable, therefore this variable is not mea-sured with error, and hence, the measurement error of this vari-able is zero.

Amer. J. Agr. Econ.


7/13

Farmers' Heterogeneity and Futures Contracts Usage 913Unlike regressionmodels,X and5 are notrequiredto be observable variables.Rather,ir and5 are related to observablevariablesxandy by a pairof confirmatoryactormodelsas expressed n

(3) x=A x +y = Ay? + e

wherex is a q x 1 vectorof observedexoge-nous variables(i.e., the indicatorsmeasuredduring the interview) and y is a p x 1vectorof observedendogenousvariables i.e.,the farmers'probabilityof usingfuturesandchoice of futures).Ax is a q x s matrix ofloadingsof the observed x-variables on thelatent5 variables,Ay and is a p x r matrixofloadingsof the observed y variables on thelatent 7 variables.The vectors of disturbances6 (q x 1) and e (p x 1) representmeasure-ment errors (the so-called unique factors).Withineach factor model (e.g., equation (3))the unique factors may be correlated.Thatis, COV(6) = E(SS') = 0, and COV(e) =E(ee') = 0,.Sincethe variablesare measuredas devia-tions from theirmeans, he covariancematrixfor the observedvariablescan be written as

-AyTll'Ay +e's Ayvl'Ax'+eS'+AyTre'+e1'Ay' +Ay 8'+etA'Ax(4) n=E

Axtl'Ay +88e' Ax'Ax,' +88'+Axte'+8qn'Ay +Ax'(8'+tAx'_Using the assumptionsmade about the zerocovariances among variables and applyingthe definitionsof the covariancesamongvari-ables, the covariance can be expressed interms of the parametersof the structuralmodel:

AyB-l(FF)F'+) AyB'-lrFAx'(5) Q= xB'-Ay'+0,

AxrF'B' -A AxI Axx O _This equation relates the variances andcovariancesof the observedvariablesto theparametersof the model. Estimation of themodel involvesfindingvalues for the param-eter matrices that produce an estimate ofQi according o equation (5) that is as closeas possible to the samplematrixS (e.g., thecovariancematrixof the rawdata).

Estimation ProcedureThe covariance structure in equation (5)can be estimated by one of the full infor-mation methods:unweighted least squares,generalized east squares,and maximum ike-lihood. The fitting function measures showhow close a given Q is to the samplecovari-ance matrixS. Becauseof its attractive tatis-tical properties,we use maximum ikelihoodprocedures (Bollen). The maximum likeli-hood estimators minimize the followingfit-ting function:(6) S -f) + (gFML(S;) = tr(n S) + (log InI

-log ISI)-(r + s),wherelog IQIIs the log of the determinant fQf,r is the numberof endogenousvariables(in our case r = 2) and s is the numberofexogenous variables(in our case s = 8). Iff and 17have a multivariatedistribution,hemaximumikelihoodprocedurehas desirableasymptoticproperties, s scale invariantandhas desirableproperties or statistical esting(e.g.,Anderson andGerbing;Bagozzi;Bollen;GerbingandAnderson).Priorto the estimation,we tested whetherthe underlyingassumptionsof the covari-ance structuremodel had been satisfied.Wescreenedthe data for coding errors and thepresenceof outliers,and testedfor univariateand multivariatenormalityof the observedvariables. The coefficient of relative mul-tivariate kurtosis was 1.09, indicating thatthe assumptionof multivariatenormalityistenable (Steenkamp and Van Trijp).As ameasure of associationwe used the covari-ances.As pointed out by BaumgartnerandHomburgthe use of covariances or corre-lations has no effect on the overall good-ness of fit indices.However,standarderrorsmay be inaccuratewhen using correlations.Therefore,we used covariances.The modelcan be estimatedin the maximum ikelihoodLISREL frameworkdeveloped by Joreskogand Sorbom.ResultsFirst,we model the farmersuse of futuresacrossthe whole sample.That is, we do nottake heterogeneity into account, assumingthe sampleto be homogeneous.Table1 showsthat,in this case,the decisionunit,perceivedperformance,xercisingentrepreneurialree-dom,and level of understandingignificantly



8/13

Table 1. Results of Covariance Structure Models for the Whole Sample and Different Segments (equProbability of Using Futures

Explanatory Variables

DU PP EF MO RA PRE DARTotal sample (N = 440) r 0.254 0.195 0.188 0.042 0.156 0.123 0.004t-value 5.632 4.028 3.872 1.126 1.567 0.943 0.426

Fit statistics X2/df = 4.3 p = 0.01 RMSEA = 0.09 GFISegment 1 (N = 120)Spot market:cooperative r 0.212 0.201 0.084 0.062 0.306 0.212 0.091t-value 2.203 2.488 1.422 1.246 3.246 2.228 2.066

Fit statistics X2/df = 1.0 p = 0.23 RMSEA = 0.03 GFISegment 2 (N = 320)Spot market: trader F 0.264 0.189 0.258 0.111 0.172 0.148 0.036t-value 4.996 3.966 5.348 2.003 1.636 1.171 0.412

Fit statistics X2/df = 2.0 p = 0.01 RMSEA = 0.05 GFINote: Beta is the standardized regression coefficient that shows the relationship between the probability of using futures and the latent constructs. DU is the decision unadvisors on the probability of using futures), PP perceived performance, EF exercising entrepreneurial freedom, MO market orientation, RA risk attitude, PRE perceived riunderstanding futures. RMSEA is the root mean square error of approximation, GFI the goodness-of-fit index, and TLI the Tucker Lewis Index (Joreskog and Sorbom). Se


9/13

Farmers' Heterogeneity and Futures Contracts Usage 915

influencethe probabilityof usingfuturesandconsequently farmers' choice. Surprisingly,risk attitude and perceived risk exposureare not significantlyrelated to the farmer'suse of futures (Makus et al.; Shapiro andBrorsen).We suspectthatthe sampleis not homoge-nous. That is, we expect different groupsof farmersuse differentcriteria(attributes)when decidingto use futures.If this is thecase, we might find that different factorsinfluence their choice behavior, and thatthe common factors are weighted differ-ently (i.e., the coefficients in matrix F inequation (2)).To test for heterogeneitywe use clusteranalysis using squared Euclidian distancesand Ward's method (error sum of squaresmethod) (Punjand Stewart).10We found twodistinctsegments.Segment 1 consists of 120farmerswho have a relativelylow probabil-ity of usingfutures.Segment2 consists of 320farmerswho have a relatively high proba-bility of using futures.In order to explorethe differencesbetween these two segments,we analyzedthe characteristics f the farm-ers regarding he probabilityof usingfuturesusing the Mann-WhitneyU test (Hinkle,Wiersma, ndJurs).Thetwo segmentsmainlydifferedon theircash-trading ehavior("sell-ing to a cooperative", z = 4.88 [p = 0.00]and "selling to a trader",z = 5.28 [p =0.00]). These two segments do not signifi-cantlydifferregardingage andeducation.So,whileboth segments may appearsimilar,dif-ferentfactors nfluence heuse of futurescon-tracts.Hence,we divide the sampleinto twosegments and estimate the model for eachsegment.From table 1 it is clear that risk atti-tude and perceived risk exposure do playa role in segment 1. Also, the debt-to-assetratio plays a role for this segment,which isin line with the recent findingsof Sharpiroand Brorsen, and Turvey and Baker. Thevalue of taking heterogeneity into accountis shown in this case. Had we treated thesampleas homogenous,we would have con-cluded that risk attitude and perceivedriskexposuredo not influence the use of futurescontracts,which would not have compliedwith financial theory. In segment 2 it wasfound that market orientation and exercis-ing entrepreneurialreedom are factors thatinfluence the probability of using futures,10For a detailed explanation on cluster analysis, see Hair et al.

along with the decision unit and the per-ceived performance.The role of understand-ing futures does not play a significantroleany more when analyzingthe two segmentsseparately.In this paper, the model fit is evalu-ated using different types of fit indicesrecently developed in the literature. Weuse the likelihood-ratioChi-squarestatistic,Goodness-of-FitIndex (GFI), TuckerLewisIndex (TLI), and the Root Mean SquaredError of Approximation RMSEA) to evalu-ate the model fit." The fit statisticsshowthatthe model thatcoversthe wholesamplehas amodestfit,while the model that describes heuse of futurescontracts or segments1 and 2shows a very good fit.It appears that farmers in segment 1use "financialstructure"characteristics asimbedded n the debt-to-assetratio,riskatti-tude and the perceivedriskexposure) n theirdecision to engage in futures,whereas thefarmers n segment2 use "marketing"har-acteristics(imbeddedin market orientationand exercisingentrepreneurial reedom) intheir decision to engage in futures.Farmersin segment 1 (cooperative farmers) attachhigh value to "continuing he firm'sopera-tion for successors"whereasfarmerswho sellto traders segment2) attachvalue to "keep-ing up with markets and tryingto get highprices".Also, we estimated the models with mul-tiple regression, using the sum scores ofthe latent constructs n the regression.Thiswas done to test whether taking measure-ment error into account using a covariancestructure model does indeed contribute toour understandingof farmers'behavior. Inthe multiple regression,the latent variables

1The likelihood-ratio Chi-square statistic (X2) tests whetherthe matrices observed and those estimated differ. Statistical sig-nificance levels indicate the probability that these differences aredue solely to sampling variations.Low x2 per degree of freedom(value lower than 2.5) indicates that the actual and predictedinput matrixes are not statistically different. The likelihood-ratioChi-square statistic is heavily (negatively) influenced by sam-ple size (N > 200) (Bentler). Because of this problem, otherfit indices have been developed, such as the Goodness-of-FitIndex (GFI), which represents the overall degree of fit, that is,the squared residuals from prediction compared with the actualdata. The measure ranges from 0 (poor fit) to 1.0 (perfect fit).The Tucker Lewis Index (TLI) is an incremental fit measurethat combines a measure of parsimony into a comparative indexbetween the proposed and null model. A recommended value is0.9 or greater. The Root Mean Squared Error of Approximation(RMSEA) estimates how well the fitted model approximates thepopulation covariance matrix per degree of freedom (Steiger).Browne and Cudeck suggested that a value below 0.08 indicatesa close fit (see Baumgartner and Homburg; Bentler; Hair et al.).



10/13

916 November 2000risk attitude, market orientation, and exercis-ing entrepreneurial freedom were not signif-icant, thereby substantiating the importanceof taking measurement error into account.These results show that by taking measure-ment error into account we are able to detectrelevant factors influencing farmers' use offutures that otherwise (i.e., in a regressionframework) would not have been identified.The proposed modeling procedure is a moregeneral modeling framework than the mul-tiple regression framework as it can includedirectly observable variables and non-directlyobservable variables (constructs). This frame-work allows researchers to include farm-ers' perceptions when trying to understandfarmers' behavior. Farmers' perceptions andbeliefs appear to have relationships withrevealed behavior.Our results show that farmers are a het-erogeneous group, suggesting that futuresexchanges might consider offering differenttools to different segments. Identifying thedifferent segments is a challenge. With thisinformation, the futures exchange could tar-get their marketing efforts. Based on thecharacteristics of the different segments, theywould be able to select a group of potentialcustomers to whom they would offer a riskreduction service designed to match the cus-tomer's choice profile. This implies differen-tiation of the services offered by exchanges.In our empirical study, the segments couldeasily be observed by the exchange, becausethey relate to the farmers' cash trading pat-tern. In this regard, a challenge for furtherresearch would be to develop a method thatsimultaneously estimates all parameters, suchthat a set of parameters identifies the seg-ments to which farmers belong, and thatalso represents the structural equation modelwithin the segments. The recent findings byJedidi, Jagpal, and DeSarbo on general finitemixture structural equation modeling relateto this.

Discussion and ConclusionsThis study expands earlier work on farmer'suse of futures by including their market ori-entation, perceived risk reduction and marketperformance, farmers' entrepreneurial behav-ior and perceived risk exposure as factorsthat influence futures usage. These factorsare all non-directly observable variables and

hence must be measured by a set of observ-able indicators. By introducing a modelingtechnique that is able to account for themeasurement error of these variables, wewere able to identify their effect on farm-ers' decision making. Not taking account ofmeasurement error (e.g., using a multipleregression framework) would have led usto conclude that these factors do not playa role in farmers' decision to use futures.The proposed modeling method allowed usto confirm empirically a suggestion made byWorking more than forty years ago, namelythat futures usage is positively related withthe extent to which farmers believe thatusing futures gives them greater freedomfor business action. These results show theimportance of farmers' perceptions, and psy-chological constructs in general, and theimportance of taking measurement errorinto account, when attempting to understandfarmers' futures usage behavior.In this study we also included the influ-ence of the members of the farmer's deci-sion unit, such as husband, wife and succes-sor, on futures usage. It appears that the opin-ions of these decision-unit members aboutfutures have an influence on the farmer'sfutures usage. Hence, the decision to usefutures is not solely made by the farmer,but it is also influenced by the opinions ofothers.

Special attention is paid to the funda-mental driver of risk management: farmers'risk attitude. Although theory suggests thatrisk attitude should be related to futuresusage, empirical research in agricultural eco-nomics does not always confirm that rela-tionship (Goodwin and Schroeder). Here,risk attitude is an important determinant offutures usage for a particular segment offarmers. In this group of farmers (segment1 in our study), other financial factors alsoplay an important role, such as the debt-to-asset ratio, which is in line with the find-ings of Shapiro and Brorsen, and Turvey andBaker.Because some of the theoretical constructsrelated to futures usage can not be mea-sured directly, but must instead be estimatedfrom multiple indicator measures, it is recom-mended to use a covariance structure mod-eling framework. This framework providesus with a method for estimating structuralrelationships among unobservable constructsand for assessing the adequacy with whichthose constructs have been measured. Such

Amer. J. Agr. Econ.


11/13

Farmers'Heterogeneitynd FuturesContractsUsage 917framework allows agricultural economists toinclude farmers' perceptions in their analy-sis. The proposed modeling approach is nota substitute but rather a complement to theregression framework used in previous stud-ies on futures usage. These studies could havebenefited from this modeling frameworkbecause it allows for explicit managementof measurement error of the independentvariables. As shown in this study, variablesthat seemed to play no role in explainingfutures usage in a regression framework didplay a role when measurement error wasaccounted for. Moreover, taking farmers' het-erogeneity into account showed that differ-ent groups of farmers have different decisionstructures. Such information can be valuablefor a futures exchange, especially when theexchange wishes to customize its hedgingservice.Some caveats of our analysis should bementioned. First, this paper focused exclu-sively on the level of the individual farmer,thereby not taking into account the farmer'scommercial environment, that is, the mar-keting channel (s)he is in. Although Dutchhog farmers do not have as many alterna-tive risk management instruments (there isvirtually no forward trading) as their U.S.colleagues, we might expect the farmer'schoice for futures contracts to interact withthe risk management decisions made by thewholesaler or processor whom (s)he sup-plies. Second, we do not include transactioncosts in our analysis. Recently, Lence showedthat hedging costs are important determi-nants of hedging behavior. In the contextof our model it would be interesting toinclude the farmers' perceived transactioncosts. Third, the reason farmers use futurescontracts is often not to reduce the price riskof a single commodity, but rather to reducethe risk which remains after all price riskshave been offset against one another, theso-called residual risk (Anderson and Dan-thine; Fackler and McNew). For a futuresexchange, it may therefore be interesting toadd other types of futures contracts to thecontracts already listed. This raises an impor-tant question that needs to be solved: is itbeneficial to add new futures contracts to theexisting ones? Further research that includesthese elements are interesting avenues toexplore.

Appendix.Descriptionof the ScalesDescribingFarmer'sCharacteristicsUsingConfirmatory actorAnalysisFarmers were asked to indicate their agree-ment with each item through a nine-pointscale ranging from "strongly disagree" to"strongly agree".

PerceivedPerformanceConstruct eliability= 0.76*1. I think thatselling my hogs by means of futurescontractswill enable me to reduce the fluctua-tions in my revenues.2. I thinkthat a futurescontract anhelpme man-age risk.3. I thinkthat usingfuturescontractswill reducepricerisk.Modelis saturatedresulting n a perfectfit(X2= 0; df = 0; p = 1).

EntrepreneurialreedomConstruct eliability= 0.751. I think hatby usingfuturescontracts canfullyexploit my spiritof free enterprise.2. I think that the use of futures contractsgivesme the opportunityto receive an extra highprice.3. I thinkthat usingfuturescontractsgives me alarge freedomregardingactions in the marketplace.Model is saturatedresulting n a perfectfit(X2= 0; df = 0;p = 1).PerceivedriskexposureConstruct eliability= 0.741. I am able to predicthog prices.2. The hog market s not at all risky.3. I am exposedto a largeamountof risk when Ibuy/sell hogs.Model is saturatedresulting n a perfectfit(X2= 0; df = 0; p = 1).Risk attitudeConstruct reliability = 0.731. I like to "play t safe."2. With respect to the conductof businessI amriskaverse.3. Withrespectto the conduct of businessI liketo take the sure thinginstead of the uncertainthing.4. WhenI am sellinghogs I like to take risks.X2/df = 1.0 (p = 0.37); RMSEA = 0.0; GFI =0.99; AGFI = 0.99; CFI = 1.



12/13

918 November 000Market orientationConstruct reliability = 0.701. I think it is important to understand the wishesof my customers.2. I think it is important to know how my cus-tomers evaluate my product.3. I adapt to changes into the market place.4. I track the market prices of the products I pro-duce.X2/1.1 (p = 0.31); RMSEA = 0.01; GFI = 0.99;AGFI = 0.99; CFI = 0.99.Level of understandingConstruct reliability = 0.741. I know how the futures market is functioning.2. There is sufficient information on the function-

ing of futures markets.3. I understand the way I can hedge my risk onthe futures market.4. I keep informed about futures prices.X2/df = 3.1 (p = 0.04); RMSEA = 0.06; GFI =0.99; AGFI = 0.97; CFI = 0.98.*The value of the construct reliability ranges between 0 and 1, with highervalues indicating higher reliability (see Hair et al.)

[Received July 1999;accepted February2000.]

ReferencesAnderson, R.W., and J.P.Danthine. "Hedging andJoint Production, Theory and Illustrations."J. Finan. 35(May 1980):487-98.Anderson, J.C., and D.W. Gerbing. "Structural

Equation Modeling in Practice:A Review andRecommended Two-step Approach." Psych.Bull. 103(May 1988):411-23.Asplund, N.M., D.L. Foster, and T.T. Stout. "Farm-ers' Use of Forward Contracting and Hedg-

ing." Rev. Fut. Mkts. 8,1(1989):24-37.Bagozzi, R.P. "Structural Equation Models inMarketing Research: Basic Principles." inR. Bagozzi, ed. Principles of MarketingResearch. Cambridge MA: Blackwell, 1994,317-85.

Baumgartner, H., and C. Homburg. "Applicationsof Structural Equation Modeling in Mar-keting and Consumer Research: A Review."Internat. J. Res. Mktg. 13(April 1996):139-61.Bentler, P.M. "Comparative Fit Indexes inStructural Models." Psych. Bull. 107(March1990):238-46.Bollen, K.A. StructuralEquations with Latent Vari-ables. New York: John Wiley, 1989.Brandstatter, H. "Becoming an Entrepreneur: AQuestion of Personality Structure."J. Econ.Psych. 18(April 1997):157-77.

Brockhaus, R.H. "Risk Taking Propensity ofEntrepreneurs." Academy Mgt. J. 23(Septem-ber 1980):509-20.Browne, M.W., and R. Cudeck. "Single Sam-ple Cross-Validation Indices for CovarianceStructures." Multiv. Beh. Res. 24(0ctober1986):445-55.Carlton, D.W. "Futures Markets: Their Purpose,Their Growth, Their Successes and Failures."J. Fut. Mkts. 4(Fall 1984):237-71.Carter, C.A. "Commodity Futures Markets: A Sur-vey." Austr. J. Agr. Resour. Econ. 43(June1999):209-47.Churchill, G.A. "A Paradigm for DevelopingBetter Measures of Marketing Constructs."J. Mktg. Res. 16(February 1979):64-73.

Ederington, L.H. "The Hedging Performance ofthe New Futures Markets."J. Finan. 34(March1979):157-70.Ennew, T.C.,W. Morgan, and T. Rayner. "Role ofAttitudes in the Decision to Use Futures Mar-kets: The Case of the London Potato FuturesMarket." Agribus.: An Internat. J. 8(Novem-ber 1992):561-73.Fackler, P.L., and K.P. McNew. "Multi-productHedging: Theory, Estimation, and an Appli-cation." Rev. Agr. Econ. 15(September1993):521-35.Fishbein, M., and I. Azjen. Belief, Attitude, Inten-tion and Behavior: An Introduction to Theoryand Research. Reading MA: Addison-Wesley,1975.

Futures Industry Association. 1999 VolumeReports.

Geczy, C., B.A. Minton, and C. Schrand. "WhyFirms Use Currency Derivatives." J. Finan.52(September 1997):1323-54.

Gerbing, D.W., and J.C. Anderson. "An UpdatedParadigm for Scale Development Incorporat-ing Unidimensionality and its Assessment."J. Mktg. Res.25(May 1988):186-92.Goodwin, B.K., and T.C. Schroeder. "Human Cap-ital, Producer Education Programs, and theAdoption of ForwardPricing Methods."Amer.J. Agr. Econ. 76(November 1994):936-47.Graham, J.R., and C.W. Smith. "Tax Incentivesto Hedge." J. Finan. 54(December 1999):2241-62.

Guttman, L.A. "Basis for Scaling QuantitativeData." Amer. Soc. Rev. 9(April 1944):139-50.Hair, J.F, R.E. Anderson, R.L. Tanham, andW.C. Black. MultivariateData Analysis. Engle-wood Cliffs NJ: Prentice Hall, Inc., 1995.Hinkle, D.E., W. Wiersma, and S.G. Jurs.AppliedStatisticsfor the Behavioral Sciences. Boston:

Houghton Mifflin Co., 1998.

Amer.J. Agr.Econ.


13/13

Farmers'HeterogeneityndFuturesContractsUsage 919Howard, C., and L. D'Antonio. "A Risk-returnMeasure of Hedging Effectiveness." J. Finan.

Quant. Anal. 19(March 1984):101-12.Jaworski, B.J., and A.K. Kohli. "Market Ori-entation: Antecedents and Consequences,"J. Mktg. 57(July 1993):53-70.Jedidi, K., H.S. Jagpal, and W.S. DeSarbo."Finite-Mixture Structural Equation Mod-els for Response-Based Segmentation andUnobserved Heterogeneity." Mktg. Science16,1(1997):39-59.

Joreskog, K.G., and D. Sorbom. LISREL 8: Struc-tural Equation Modeling with the SIMPLISCommand Language. Chicago: Scientific Soft-ware International, Inc, 1993.Koski, J.L., and J. Pontiff. "How Are Deriva-

tives Used? Evidence from the MutualFund Industry." J. Finan. 54(April 1999):791-816.Lence, S.H. "Relaxing the Assumption ofMinimum-Variance Hedging." J. Agr. Resour.Econ. 21(July 1996):39-55.Likert, R. "A Technique for the Measurementof Attitudes." Archives of Psych. No. 140

(1932).Makus, L.D., B.H. Lin, J. Carlson, and R. Krebill-Prather. "Factors Influencing Farm Level Useof Futures and Options in Commodity Mar-keting." Agribus.: An Internat. J. 6(November1990):621-31.March, J.G.,and Z. Shapira. "Managerial Perspec-tives on Risk and Risk Taking."Mgt. Science33(November 1987):1404-18.Mian, S.L. "Evidence on Corporate Hedging Pol-icy." J. Finan. Quant. Anal. 31(September1996):419-39.Musser,W.N.,G.F Patrick, and D.T. Eckman. "Riskand Grain Marketing Behavior of Large-Scale Farmers." Rev. Agr. Econ. 18(January1996):65-77.Nance, D.R., C.W. Smith, Jr., and C.W. Smithson."On the Determinants of Corporate Hedg-ing."J. Finan. 48(March 1993):267-84.

Nunnally, J.C., and I.H. Bernstein. Psychomet-ric Theory, 3rd ed. New York: McGraw-Hill,1994.

Punj, G., and D.W. Stewart. "Cluster Analysisin Marketing Research: Review and Sugges-tions for Application." J. Mktg. Res. 20(May1983):134-48.Putte van den, B., J.Hoogstraten, and R. Meertens."A Comparison of Behavioral AlternativeModels in the Context of the Theory of Rea-soned Action." British J. Social Psych. 35(June1996):257-66.

Shapiro, B.I., and B.W. Brorsen. "Factors Affect-ing Farmers' Hedging Decisions." NCJ. Agr.Econ. 10(July 1988):145-53.Silber, W.L. "Innovation, Competition, and NewContract Design in Futures Markets."J. Fut.Mkts. 1(Summer 1981):123-55.Smith, C.W., and R. Stulz. "The Determinantsof Firms' Hedging Policies." J. Finan. Quant.Anal. 20(December 1985):391-405.

Steenkamp, J-B.E.M., and H.C.M. van Trijp."TheUse of LISREL in Validating Marketing Con-structs." Intern. J. Res. Mktg. 8(November1991):283-99.

Steiger, J.H. "Structural Model Evaluationand Modification: An Interval EstimationApproach." Multiv. Behav. Res. 25(April1990):173-80.

Tashjian, E., and J. McConnell. "Requiem for aMarket:An Analysis of the Rise and Fall of aFinancial Futures Contract." Rev. Fin. Studies2,1(1989):1-23.Thaler, R.H. Advances in Behavioral Finance.New York: Russell Sage Foundation, 1993.Thurstone, L.L., and E. Chave. The Measurementof Attitude. Chicago: University of ChicagoPress, 1929.

Tufano, P. "Who Manages Risk? An Empiri-cal Examination of Risk Management Prac-tices in the Gold Mining Industry."J. Finan.51(September 1996):1097-137.

Turvey, C.G., and T.G. Baker. "A Farm-LevelFinancial Analysis of Farmers' Use of Futuresand Options under Alternative Farm Pro-grams." Amer. J. Agr. Econ. 72(November1990):946-57.

Working, H. "Futures Trading and Hedging."Amer. Econ. Rev. 43(June 1953):314-43.

PenningsndLeuthold

farmers behavioral

Documents