Athur Mabiso, James Sterns, Lisa House and Allen Wysocki

Size: px
Start display at page:

Download "Athur Mabiso, James Sterns, Lisa House and Allen Wysocki"

Transcription

1 Estmatng Consumers Wllngness-To-Pay for Country-Of-Orgn Labels n Fresh Apples and Tomatoes: A Double-Hurdle Probt Analyss of Amercan Data Usng Factor Scores Athur Mabso, James Sterns, Lsa House and Allen Wysock Authors are Masters Student, Assstant Professor, Assocate Professor and Assocate Professor respectvely, Food and Resource Economcs Department IFAS, Unversty of Florda 1183 McCarty Hall A P.O. Box , Ganesvlle, FL E-mal: jasterns@fas.ufl.edu Selected Paper prepared for presentaton at the Amercan Agrcultural Economcs Assocaton Annual Meetng, Provdence, Rhode Island, July 24-27, 2005 Copyrght 2005 by Athur Mabso, James Sterns, Lsa House and Allen Wysock. All rghts reserved. Readers may make verbatm copes of ths document for noncommercal purposes by any means, provded that ths copyrght notce appears on such copes.

2 Estmatng Consumers Wllngness-To-Pay for Country-Of-Orgn Labels n Fresh Apples and Tomatoes: A Double-Hurdle Probt Analyss of Amercan Data Usng Factor Scores Athur Mabso, James Sterns, Lsa House and Allen Wysock Abstract Data are collected from prmary shoppers n Ganesvlle Florda, Atlanta Georga and Lansng Mchgan usng a Vckrey (ffth-prced sealed bd) expermental aucton and a survey questonnare to provde a sample of 311 observatons useable for analyss. The average wllngness to pay (WTP) for country of orgn labelng (COOL) Grown n the U.S. n apples and tomatoes are calculated then tested for equvalence to assess f WTP s produce specfc. A double-hurdle probt model s then estmated to ascertan the promnent determnants of WTP for COOL. Independent varables nclude demographcs, food safety and factor scores derved from a factor analyss of food qualty and food preference varables. Results show that on average consumers are wllng to pay $0.49 and $0.48 for COOL n apples and tomatoes respectvely wth 79% of the consumers wllng to pay more than $0.00 for apples labeled Grown n the U.S. and 72% n the case of tomatoes. Premums are found to be statstcally equvalent suggestng that WTP for COOL s not produce specfc. The double hurdle probt estmaton fnds most ndependent varables nsgnfcant wth the excepton of the food qualty factor scores and consumer trust levels for nformaton they receve from U.S. government agences. Locaton, age and ncome also turn out to be sgnfcant factors n the case of the truncated part of the estmaton as do food qualty and food safety concerns. Key words: Wllngness to pay, Country of Orgn Labelng, Vckrey Aucton, Apples and tomatoes 2

3 Introducton As consumer demand for agrcultural food-products becomes more complex and dynamc, food labelng s takng an ncreasngly mportant role n the food marketng system (McCluskey and Lourero, 2003). Consumers are constantly obtanng dfferent knds of nformaton about food-product attrbutes va food labels and ther purchasng decsons are nfluenced by these. Theoretcally, consumers demand food-product attrbutes (e.g. food qualty or taste) not the food-product per se and the food-product s consdered to be merely a bundle of these ndvdual attrbutes that gve rse to utlty. Thus purchasng decsons made by consumers are based on specfc food-product attrbutes emboded n a food-product (Lancaster, 1966). An understandng of ths s mportant f one s to recognze the sgnfcance of food labelng. Ths s because food labels present nformaton about specfc food-product attrbutes, whch potentally can affect consumer wllngness to pay (WTP) and n turn the effectve demand for the foodproduct. Recent studes show that consumers wll pay premums for certan attrbute clams made by dfferent food labels, e.g. Envronmentally-frendly clams made by eco-labels, organcally produced product clams n organc labels, non-gm food clams n non-gm labels and U.S. Certfed n the case of Country of Orgn Labelng (COOL) (Lourero, McCluskey and Mttlehammer, 2001; Burton et al 2001; Umberger et al, 2002 respectvely). These studes allude to credence attrbutes, whch consumers have no reasonable means of verfyng for themselves. In the U.S. fresh produce ndustry ths s a topcal ssue, partcularly COOL. The man reason s recent Mandatory Country of Orgn Labelng (MCOOL) legslaton n the 3

4 2002 Farm Securty and Rural Investment Act. Subttle D of ths composte act specfes that currently market actors can voluntarly label ther produce wth COOL so as to nform shoppers at the fnal pont of purchase the orgn of the produce. Gudelnes for voluntary COOL whch were ssued by the USDA n October 2002, apply to meats (beef, pork, lamb and fsh) as well peanuts, fruts and vegetables the so-called covered products (VanSckle, 2003). These products were selected because they are food products most prone to food safety and health problems. The law, however, s set to change to MCOOL. Ths change was ntally set to be effectve on the 30 th of September 2004 but has been postponed by two years wth the passng of the Omnbus Appropratons bll n January The mmnent change n law brngs to the forefront several ssues of debate, whch surround justfcaton of the polcy. One of these major ssues concerns consumers WTP for COOL. There s lttle emprcal nformaton on how much consumers are WTP for COOL. Informaton on ths would be mportant n pontng polcy and/or decson makers n a partcular drecton n as far as mplcatons of MCOOL are concerned (Menkhaus, 2001). Ths study seeks to fll part of ths nformaton gap by analyzng f Amercan prmary shoppers are WTP a premum for fresh apples and tomatoes labeled Grown n the U.S. In addton, the study examnes whether premums for Grown n the U.S. labels are product specfc or not, n the context of apples and tomatoes. 4

5 Prevous Studes Food labelng n the Apples Market In the last decade several studes have been carred out on food labelng n the produce ndustry and apples n partcular. In 1999, Blend and Van Ravenswaay studed ecolabelng n apples and found consumers to be wllng to try eco-labeled apples. More than a thrd of Amercan consumers were wllng to pay a $0.40 premum for a pound of ecolabeled apples. Lourero, McCluskey and Mttelhammer (2001) also studed WTP for organc labels and eco-labels n apples. Usng contngent valuaton methods (CVM) they collected data from the Mdwest and estmated a maxmum-lkelhood multnomal logt model. Results found food safety, produce qualty and envronmental concerns to be sgnfcant and postvely affectng WTP. Presence of chldren under the age of 18 n the household was also found to be a sgnfcant factor ncreasng the probablty of an Amercan consumer to purchase organc labeled apples. Country of Orgn Labelng Studes Schupp and Gllespe (2001) are part of the frst researchers who turned the focus onto COOL n agrcultural products and specfcally beef. By analyzng a sample of Lousana households they found an average of 90.3 percent consumers supportng MCOOL. Also, they estmated a probt model and establshed that food safety concerns were a sgnfcant factor ncreasng the probablty of a consumer to support MCOOL. Consumer preference for locally produced beef also postvely affected the lkelhood to support MCOOL. 5

6 In Lourero and Umberger (2002) Colorado consumers were sampled from dfferent stores and shown to be wllng to pay for a MCOOL program. Consumers who had completed a hgh level of educaton and had hgh household ncomes were less lkely to pay a premum for U.S. Certfed labels n beef. Ths dsproved ntal hypotheses that more educated and wealther consumers would pay attenton to COOL and be more lkely to pay a premum for t. Female consumers were most lkely to pay a premum for COOL and to be more supportve of MCOOL. In 2004, Lourero and Umberger used expermental auctons to solct nformaton about U.S. consumers WTP for COOL n beef. They ascertaned that COOL n beef s a less mportant determnant of consumers WTP as compared to food safety nspecton labels, product qualty labels (tenderness) or traceablty of the beef. Smlar fndngs were made n Europe, (see Roosen, Lusk and Fox; Verbeke and Ward) mplyng that food qualty and food safety are very mportant factors that may overrde country of orgn. Whle all these studes provde evdence for a causal relatonshp between WTP for COOL, and the demographcs, food qualty and food safety varables, other varables may play a role on the nature of the relatonshp e.g. geographc locaton, or product under consderaton. Wth respect to fruts and vegetables the stuaton s unclear, snce the majorty of prevous studes focused on beef. Ths research delves nto these specfc ssues to address COOL n apples and tomatoes. Theoretcal Framework The random utlty approach s used to estmate the consumer s WTP decson process. Ths process can be broken down nto two parts, ) the partcpaton decson and ) the 6

7 consumpton decson, where the former refers to the choce whether or not the consumer s wllng to pay for COOL and the latter referrng to how much, f ndeed the consumer s wllng to pay. We assume that the ndvdual consumer can attan utlty from a specfc product attrbute, n ths case COOL n apples or tomatoes ( Grown n the U.S. ). Ths utlty s a functon of consumer characterstcs that nfluence consumer choce and the cost that the consumer s wllng to pay n order to obtan the attrbute. Thus, (1) U( c ) 0 c U = π, where the cost that the consumer wll pay to obtan the attrbute. π s a combnaton of consumer characterstcs and Utlty ganed from the attrbute s zero when the consumer s not wllng to pay anythng to obtan the attrbute, otherwse t s greater than zero. We dsregard the case of dsutlty (.e. U < 0 ) because we assume a ratonal consumer where the buyng decson ultmately must yeld postve utlty. The utlty functon s unobservable and cannot be measured by the researcher. However, a proxy measure of utlty can be estmated by the WTP. Smlarly, we assume that not all consumer characterstcs are drectly observable and quantfable, e.g. consumers perceptons about food qualty or consumers feelngs about food preferences. Instead, these are latent constructs whose phenomena are observed va other drectly quantfable proxy varables. Thus, we deconstruct the utlty functon n smlar fashon to Adamowcz et al (1998), wth the only dfference beng that we propose a drectly observable determnstc part λ, an ndrectly observable determnstc part ρ, and a stochastc error termε. The error term s assumed to be ndependent and dentcally dstrbuted wth a mean of zero and a constant varance. 7

8 U π c = V λ, ρ c + ε (2) ( ) ( ) We postulate that the varance of the ndrectly observable ρ can be better estmated by way of a factor analyss of the drectly observable and quantfable proxesθ, rather than by usng an ndvdual observable proxy varable. Mathematcally, (3) ρ = ψθ, whereψ s a vector of factor loadngs Thus the WTP decson can then be framed n lkelhood terms as (4) { PMT 0} = Pr{ V ( λ, ρ ) ε } Pr, even though ρ s unobservable drectly. The above forms the basc theoretcal framework for the double hurdle model estmaton wth factor scores, whch we propose n ths study. Methods Snce the decson process on WTP for COOL s a two-step process we model t usng Cragg s double hurdle model. Thus; (5) WTP * = Z ' β + u for the partcpaton equaton denotng the dchotomous wllng to pay or not wllng to pay part of the framework. Then, * ' (6) PMT = γ + ε for the quanttatve consumpton part of the framework. X In (5) the varable * WTP s the consumer wllngness to pay assumng 0 f not and 1 f wllng to pay. Ths dependent varable represents the underlyng utlty assocated wth the partcpaton decson; essentally whether or not the consumer derves utlty from the * attrbute. In (6) PMT s the actual premum that consumers are wllng to pay for the 8

9 * apples or tomatoes wth COOL, f n (5) WTP was equal to 1. Ths represents the magntude of the latent utlty assocated wth the COOL attrbute. ' Z and ' X are predctor vectors whle β and γ are parameter vectors to be estmated for ' ' the respectve predctor vectors. and can potentally be dentcal and nclude Z reduced varables n the form of factor scores derved from factor analyss. If X ' Z and X ' are equal and β and γ are also equal then the tobt model results nstead of a truncated tobt. u and ε are random error terms, normally ndependently dstrbuted. 2 u ~ NID(0, σ ) and ε ~ NID(0,1) Theoretcally the underlyng utlty whch s non-measurable can also be expressed as U = * f ( X *, Z * ) where * U s the ndvdual consumer s utlty. Equatons (5) and (6) are estmated separately wth the frst beng estmated frst because ts results are used n the estmaton of the second (.e. n estmaton of the censorng rule). A probt model can be estmated for the frst equaton usng the maxmum lkelhood functon: * * * * * (7) Pr( PMT = 0 Z, X ) = Φ( X γ / σ ) + Φ( X γ / σ ) Φ( Z β ) Then the second equaton can be estmated usng, (8) f ( PMT Z *, X *, PMT > 0) = exp{ ( PMT X γ ) * 2 2 2πσ 2 / 2σ } Φ( Z β ) * Where Φ sgnfes the standard normal cumulatve densty functon. 9

10 Data Data were collected n December 2003 and January 2004 usng a Vckrey expermental aucton (ffth-prced sealed bd) and a wrtten questonnare. The Vckrey aucton solcted data on WTP premums after whch wrtten questonnares collected data on numerous varables ncludng demographcs, food safety concerns, food qualty concerns and food preferences. Data were collected n Ganesvlle, FL, Atlanta, GA and Lansng, MI and partcpatng respondents were randomly recruted through local cvc organzatons, rangng from fath based organzatons to Parents Teachers Assocatons (PTA) at schools. The survey was conducted n each respectve organzaton s facltes and compensaton made for the use of the facltes. A total of 335 prmary shoppers were sampled and Table 1a shows the breakdown of these across locatons. In total, 311 observatons were useable for analyss; 175 n the tomato aucton and 136 n the apple aucton. The 24 observatons deleted were unusable due to mssng data. Fewer respondents were recruted from Lansng, MI, makng the data unevenly dstrbuted across locatons, partcularly n the case of the apples data. The respondents ages ranged from 25 to 65 years and only prmary shoppers partcpated n the survey. Research protocol had specfcally asked for prmary shoppers only. A prmary shopper was defned as an ndvdual responsble for at least 50 percent of food purchases n the household. Comparng the sample data to the U.S. populaton census revealed several dspartes. For nstance, 88.4% of the sample was female compared to 50.9% n the U.S. census. Ths was nonetheless expected consderng that most prmary shoppers are ndeed female and that ths was the target populaton. However, the observed dspartes call for 10

11 cauton n the nterpretaton of results from ths study. Extrapolatve generalzatons based on the study may be erroneous. Table 2 shows detals regardng the comparablty of the survey data to the natonal census data. Dspartes are evdent n terms of the ethncty, the hghest level of educaton attaned and the pre-tax household ncome varables. The greatest dsparty showed up n the hghest level of educaton attaned where the majorty of the sample had attaned a much hgher level of educaton than the general U.S. populaton. Moreover, the survey captured household prmary shoppers that were more affluent relatve to the census data. Nearly a thrd of the respondents had a pre-tax household annual ncome greater than $100,000 as compared to only 14 percent for the census data (U.S Census Bureau, 2000). Even though the data used n the study were somewhat devant from the U.S. populaton profle, they were useable because the devaton was ntally expected snce the target populaton of the study was Amercan prmary shoppers and not the general U.S. populaton. Also, some smlartes wth U.S. census data were found as shown n Table 2. Emprcal Model Specfcaton In analyzng the factors that nfluence the WTP for COOL, the followng double hurdle probt model was specfed. (9) WTP = β Age + β Gender + 13 = 11 β PC 1 = 3 + β Trust + β Safe β Edu 18 = = 5 β Pfr + β Loc 20 = 19 + β Qual = 7 9 β Inc + u + β Expose + 10 Where: WTP s the dchotomous wllngness to pay (.e. partcpaton dependent varable), expressed as a probablty 11

12 For the second hurdle PMT replaces WTP, where PMT s the quanttatve wllngness to pay (.e. consumpton dependent varable) and ε takes the place of u as the error term. Age = Age of respondent Gender = Gender of the respondent Edu = Hghest level of educaton completed by respondent Loc= Locaton (one of Atlanta, Ganesvlle, or Lansng) Inc = Income group Expose = Self ratng on exposure to food safety nformaton n fresh frut and vegetables PC = Presence of chldren under age of 16 n the household Trust = Extent of respondent s trust n nformaton about food producton obtaned from U.S. Government Agences, (e.g. USDA, FDA, EPA, etc.) Safe = Perceptons about food safety Pfr = Food preferences factor scores Qual = Food qualty factor scores Results and Dscusson Wllngness-To-Pay Estmates The mean WTP for both the apples and tomatoes aucton bds were calculated separately and sgnfcance tests performed on them. Four rounds of bddng had occurred n each aucton and the bddng progressons of these are presented n Fgure 1. The mean WTP for apples and tomatoes labeled Grown n the U.S. came out to be approxmately $0.49 and $0.48 respectvely. Standard devatons n both cases were hgh, 0.58 and 0.55 respectvely. Ths hgh level of dsperson suggests that dfferent consumers have 12

13 dstnctly dfferent levels of WTP (.e. ths can be vewed as an ndcator of consumer surplus for COOL). Unvarate hypothess testng of the mean WTP proved both means to be sgnfcantly greater than zero at 0.05 alpha levels. On calculatng the means for only those consumers who were WTP a premum for ether apples or tomatoes labeled Grown n the U.S. the expected ncrease was regstered, as shown n Table 3. Ths calculaton was done to gve further nsght on the exstng dfferentals between the sub-sample of consumers WTP and the whole sample. Overall, 79% of the consumers were WTP more than $0.00 for apples labeled Grown n the U.S. whle 72% were WTP n the case of tomatoes labeled Grown n the U.S. Followng procedures outlned n Wellek (2003) equvalence testng was done to assess premum equvalency. The crtcal regon for the unequal sample szes was computed usng the SAS program as recommended by Wellek. Usng an alpha level of 0.05 the two premums were found to be equvalent. Ths suggests that n the fresh produce ndustry the premum for COOL may not be product specfc. Possbly ths s due to smlar qualty and food safety condtons/standards across the produce sector. Double Hurdle Probt Estmaton The ndependent varables of the double hurdle probt model are descrbed n detal n Table 4. Of partcular nterest s the factor analyss used to derve food preference and food qualty factor scores (see Tables 7a and 7b). Intally the food safety factor score was also to be derved but results showed low correlatons (see Table 8) plus an extremely low Kaser-Meyer-Olkn measure of samplng adequacy of less than 0.6. Ths 13

14 necesstated the use of a sngle queston from the questonnare n the fnal model specfcaton. Table 5 presents the probt estmaton. It shows that both food qualty factor scores were sgnfcant at α = Consumers who were more conscous about food qualty (be t qualty n general or qualty assocated wth natural foods) were found to be more lkely to pay a premum for apples or tomatoes labeled Grown n the U.S. The level of trust that consumers have for nformaton they receve from U.S. government agences (e.g. USDA, FDA, EPA etc.) was the only other sgnfcant varable (at α = 0.1). Here we found that consumers who were more trustng of the nformaton they receve from U.S. government agences were more lkely to pay for COOL. Surprsngly, all demographcs turned out to be nsgnfcant n the partcpaton decson makng process, suggestng that t does not matter f one s male or female or f ther ncome s hgh or low. Perhaps the partcpaton decson s smply not a functon of demographcs. Overall, the model was sgnfcant at 0.1 sgnfcance level wth a 75.6% correct predcton rate. In our estmaton of the second stage of the double hurdle model we performed 2 the ch-squared specfcaton test, = 2( ln L + ln L ln L ) χ to evaluate f probt truncaton the truncated tobt estmaton was a better ft than the tobt. The truncated tobt proved to 2 be a superor ft ( χ = ). The truncated tobt estmaton s presented n Table 6. It shows that age and locaton sgnfcantly determne how much the consumers are WTP once they have decded they are WTP. Margnal effects also show consumer WTP as -0.6 cents per pound for every year older that the respondent s. Ths mples that, on average, older consumers wll pay less for COOL. In terms of locaton, the base used was Atlanta tobt 14

15 Georga and consumers n Lansng Mchgan were found to be WTP substantally less than those n Atlanta (49 cents per pound less). In contrast, consumers n Ganesvlle, Florda were WTP 4 cents per pound more than those n Atlanta Georga. A reason for ths could be that MCOOL polcy has been prevalent for the past 26 years at the state level n Florda. Thus, shoppers n Ganesvlle Florda could be accustomed to MCOOL and therefore are WTP for COOL. Conversely, Mchgan s geographcally far from ether Georga or Florda. Thus t could be case that Mchgan consumers are less exposed to COOL and are therefore less WTP for COOL. For more detals on locatonbased premum dfferentals refer to Table 1b. Income level s another demographc that seemed to have an mpact on the amount consumers are WTP. The greater than $100,000 ncome group was used as the base n the model. Fndngs suggest that consumers wth an ncome level of less than $50,000 are the only group wth a sgnfcantly greater WTP than the base (14 cents per pound more). Ths fndng s smlar to that n Lourero and Umberger (2002), mplyng that more affluent consumers consder t unmportant to know where ther apples or tomatoes come from and thus do not value COOL. The truncated estmaton also suggested that the food qualty factor score assocated wth natural foods would postvely ncrease the premum that consumers wll pay by 6 cents for every unt ncrease n the factor score. However, the factor score for general qualty had no sgnfcant mpact on the amount they were WTP. In contrast, the food safety varable was sgnfcant, wth consumers that thnk about food safety when purchasng fruts and vegetables beng found to be WTP more for the label Grown n the 15

16 U.S. Unexpectedly, ths food safety varable had not been sgnfcant n the partcpaton stage. Conclusons In ths paper we have estmated the WTP for the label Grown n the U.S. n apples and tomatoes to be approxmately $0.49 and $0.48 per pound respectvely, wth 79% of the consumers WTP more than $0.00 for the label n apples and 72% n the case of tomatoes. An obvous mplcaton of these fndngs s that we can safely assert that consumers do value COOL n apples and tomatoes. Ths adds to the justfcaton for MCOOL or at least COOL on a voluntary bass n the apples and tomatoes market. Also, the fndngs suggests that t may be possble for producers and marketers to use label Grown n the U.S. n order to garner a compettve advantage over mport substtutes n the market. A comparson of how the label Grown n the U.S. fars wth other country labels s, however, mperatve f any conclusve assertons are to be made. In addton, we have made the nterestng fndng, that premum equvalency exsts between the WTP for COOL n apples and n tomatoes. Ths mples that WTP for COOL s not product specfc, at least n the context of the apples and tomatoes markets. It may also suggest that there s potental to promote the generc label Grown n the U.S. for all produce to enhance overall demand for U.S. produce over mports. Addtonally, more research on ths s requred before generalzatons can be made. We have also used Cragg s double hurdle model to estmate the WTP for COOL and found that food qualty perceptons are crtcal factors n both the partcpaton and consumpton decson makng processes. Furthermore, the extent to whch consumers 16

17 trust nformaton they receve from U.S. government agences such as the USDA, FDA and EPA was found to be a sgnfcant determnant of the partcpaton decson. Ths s lkely due to the fact that the agences are responsble for regulatng and enforcng produce labelng laws. Snce COOL s a credence attrbute label ths would be mportant, and f consumers trust the nformaton they are gettng from agences then they are more lkely to value COOL because they would beleve that there s a trustworthy labelng verfcaton system n place. We have also establshed that for the consumpton decson (.e. how much to pay), some demographcs (age, locaton and ncome level) are sgnfcant determnants. Also, we found that consumers who take food safety concerns nto consderaton when makng the decson to purchase fruts and vegetables wll pay more money for the label Grown n the U.S. Clearly the mplcaton here s that consumers regard U.S. produce to be safer. In concluson, we report that food qualty s a more mportant determnant of WTP for COOL n apples and tomatoes durng the partcpaton decson stage together wth trust levels for nformaton from U.S. government agences. Food safety and demographc varables start to have an mpact at the consumpton decson level, makng them mportant varables f nterested n nfluencng the amount consumers wll pay. 17

18 References Adamowcz, W., R. Boxall, M. Wllams, and J. Louvere Stated Preference Approaches for Measurng Passve Use Values: Choce Experments and Contngent Valuaton. Amercan Journal of Agrcultural Economcs 80(1998):64-75 Blend, J.R. and E.O. Van Ravenswaay Consumer Demand for Eco-labeled Apples: Results from Econometrc Estmaton, Amercan Journal of Agrcultural Economcs, 81(1999): Burton, M., D. Rgby, T. Young, and S. James Consumer Atttudes to Genetcally Modfed Organsms n Food n the UK European Economc Revew of Agrcultural Economcs 28(2001): Cragg, J.G. Some Statstcal Models for Lmted Dependent Varables wth Applcaton to the Demand for Durable Goods. Econometrca 39(1971): Lancaster, K. A New Approach to Consumer Theory. Journal of Poltcal Economy 74(2) (1966): Lourero, M.L, J. J. McCluskey and R. C. Mttelhammer, Assessng Consumer Preferences for Organc and Eco-labeled and Regular Apples, Journal of Agrcultural and Resource Economcs 26(2) (2001): Lourero, M.L., and Umberger, W.J., A Choce Experment Model for Beef Attrbutes: What Consumer Preferences Tell Us. Selected Paper Presented at the Amercan Agrcultural Economcs Assocaton Annual Meetng, August 1-4, 2004, Denver CO Lourero, M.L., and Umberger, W.J., Estmatng Consumer Wllngness-to-Pay for Country-of-Orgn Labels for Beef Products. Selected Paper presented at the Amercan Agrcultural Economcs Assocaton, Annual Meetng, July 28-31, 2002, Long Beach, CA Lusk, J.L., J. Roosen, and J.A. Fox. Demand for Beef from Cattle Admnstered Growth Hormones or Fed Genetcally Modfed Corn: A Comparson of Consumers n France, Germany, the Unted Kngdom, and the Unted States. Amercan Journal of Agrcultural Economcs 85(1) (2003): McCluskey J.J. and M.L. Lourero, Consumer Preferences and WTP for Food Labelng: A Dscusson of Emprcal Studes. Journal of Food Dstrbuton Research 34(3) (2003): Menkhaus, D.J., Expermental Auctons: New Theoretcal Developments and Emprcal Fndngs- Dscusson Selected Paper presented at the Western Agrcultural Economcs Assocaton, Annual Meetng, July 8-11, 2001, Logan, Utah 18

19 Roosen, J., J.L. Lusk and J.A. Fox, Consumer Demand for and Atttudes towards Alternatve Beef Labelng Strateges n France, Germany and the UK. Agrbusness: An Internatonal Journal 19(1) (2003): Schupp, A. and J. Gllespe, Consumer Atttudes towards Potental Country of Orgn Labelng of Fresh and Frozen Beef, Journal of Food Dstrbuton Research 32(November 2001): Umberger, W.J., D.M. Feuz, C.R. Calkns, and K. Kllnger. U.S. Consumers Preference and Wllngness-to-Pay for Domestc Corn-fed versus Internatonal Grass-Fed Beef Measured through an Expermental Aucton. Agrbusness: An Internatonal Journal 18(4) (2002): VanSckle, J.J., Country of Orgn Labelng- A COOL Update, Internatonal Agrcultural Trade and Polcy Center, Polcy Bref Seres, PBTC December 2003, Unversty of Florda. Verbeke, W. and R.W. Ward, Importance of EU Label Requrements: An Applcaton of Ordered Probt Models to Belgum Beef Labels A selected paper presented at the Amercan Agrcultural Economcs Assocaton Annual Meetng, July 27-30, 2003, Montreal, Canada Wellek, S. (2003), Testng Statstcal Hypotheses of Equvalence Chapman and Hall/CRC, Boca Raton Webstes U.S. Census Bureau. Unted States Census, 2000 Internet address: Retreved March 12,

20 Table 1a. Tomato and Apple Data across Locaton Product TOMATOES APPLES Observatons Deleted Observatons Retaned Percent Cumulatve Percent GNV, FL LAN, MI ATL, GA Total (n) GNV, FL LAN, MI ATL, GA Total (n)

21 Table 1b. Mean WTP across Locaton Ganesvlle, FL ($/Lb) Lansng, MI ($/Lb) Atlanta, GA ($/Lb) Apples Tomatoes

22 Trend/Progresson of Mean bds $0.55 $0.50 $0.45 $0.40 Tomatoes Apples $0.35 $0.30 Round1 Round2 Round3 Round4 Round of bddng Fgure 1. Lne graph showng the trend of bds n both tomato and apple auctons, averaged over all locatons 22

23 Table 2. Demographc profle of respondents Category U.S. Census Sample Average (%) Average (%) Apples Tomatoes Age Race Whte Black or Afrcan Amercan Asan Other Ethncty Hspanc Income <$15, $15,000 to $24, $25,000 to $34, $35,000 to $49, $50,000 to $74, $75,000 to $99, $100,000 or above Educaton Bachelors Degree or Hgher Some College Hgh School Dploma (or equvalent) Less than Hgh School Total Useable Responses

24 Table 3. Average WTP for Apples (n = 108) and Tomatoes (n = 126): Samplng only those consumers who were WTP more than $0.00 Mean Standard Devaton All four bds 3 rd and 4 th round bds All four bds 3 rd and 4 th round bds Apples $0.60 $0.61 $0.56 $0.59 Tomatoes $0.64 $0.68 $0.53 $

25 Table 4. Descrpton of Independent Varables used n the Double Hurdle Probt Model Varable AGE GENDER Descrpton Numercal age of the respondent A dchotomous Gender varable: Whether the respondent was female (1) or male (0) (Male dropped) Hghest level of educaton completed : Some College or less Hghest level of educaton completed : Unversty undergraduate degree Hghest level of educaton completed : Unversty postgraduate degree (dummy dropped) Locaton: Ganesvlle, Florda Locaton: Lansng, Mchgan Locaton: Atlanta, Georga (dummy dropped) EDU1 EDU2 EDU3 LOC1 LOC2 LOC3 INC1 Pre-tax Household Income: Less than $50,000 INC2 Pre-tax Household Income: $50,000 to $74,999 INC3 Pre-tax Household Income: $75,000 to $99,999 INC4 EXPOSE PC1 PC2 PC3 PC4 SAFE TRUST PFR1 PFR2 PFR3 QUAL1 QUAL2 Sgma Pre-tax Household Income: $100,000 and above (dummy dropped) Self ratng on level of exposure to nformaton about food safety n fruts and vegetables: Lkert scale of 1-4 treated as a numercal varable. 1 s a great deal and 4 s nothng at all Presence of Chldren under 16 years n the household: 0 present Presence of Chldren under 16 years n the household: 1 chld present Presence of Chldren under 16 years n the household: 2 chldren present Presence of Chldren under 16 years n the household: 3 or more present Lkert scale ratng on statement I thnk about food safety when purchasng frut and vegetables 1-6 scale treated as a numercal varable. 1 s strongly dsagree and 6 s strongly agree Lkert scale ratng on level of trust that consumer has on nformaton about food producton from U.S. government agences, (e.g. USDA, FDA, EPA) 1-6 scale treated as a numercal varable. 1 s strongly dstrust and 6 s strongly trust Frst numercal Food Preference factor score: The more postve and hgher t s the more preference for varous foods and more open to dfferent foods. (Open to unfamlar foods) Second numercal Food Preference factor score: The more postve and hgher t s the less preference for unfamlar foods and rsk (Choosey about foods) Thrd numercal Food Preference factor score: The more postve and hgher t s the less preference for unfamlar foods and rsk (afrad of unfamlar foods) Frst numercal Food Qualty factor score: The more postve and hgher t s the more conscous about food qualty n general Frst numercal Food Qualty factor score: The more postve and hgher t s the more conscous about food qualty assocated wth natural foods Dsturbance standard devaton (Included n the truncated model) 25

26 Table 5. Probt Model for Combned Apples and Tomatoes data Varable Coeffcent Standard Standardzed Margnal p-value Mean of Error Coeffcent Effects Regressor AGE E E E GENDER E E EDU E E EDU LOC LOC E INC E E INC E INC E E EXPOSE E PC E PC E E PC E SAFE E E E TRUST E E E PFR E E E-16 PFR E E E E-06 PFR E E E E-07 QUAL E E E-06 QUAL E E E-06 Restrcted log lkelhood value, ln L 10 = Maxmum unrestrcted log lkelhood value, ln L 1 = Log lkelhood χ 2 (df=19)= (p = ) R 2 (McFadden, 1973) = R 2 (Estrella, 1998)= % of correct predctons = 75.6 (Note: E+nn or E-nn means multply by 10 to + or -nn power.) 26

27 Table 6. Truncated Tobt Model for Combned Apples and Tomatoes data Varable Coeffcent Standard Error Standardzed Margnal p-value Mean of Coeffcent Effects Regressor AGE E E E GENDER EDU E E EDU E E LOC E LOC INC INC E INC E E EXPOSE E PC E E PC E E PC E SAFE E E TRUST E E E PFR E E E-01 PFR E E E E-02 PFR E E E E-01 QUAL E E-01 QUAL E E-01 Sgma E Number of observatons = 311 Observaton after truncaton = 233 Log lkelhood functon = Threshold values for model: Lower = 0 Upper = + (Note: E+nn or E-nn means multply by 10 to + or -nn power.) 27

28 Table 7a. Rotated Component Matrx(a) for Food Qualty Proxy Varables Factor Proxy Varable 1 (General Qualty conscous) 2 (Natural qualty conscous) I usually am to eat natural food I am wllng to pay somewhat more for a product of better qualty Qualty s decsve for me n purchasng foods I always am at the best qualty When choosng foods, I try to buy products that do not contan resduals of herbcdes and antbotcs I am wllng to pay somewhat more for food contanng natural ngredents For me, wholesome nutrton begns wth the purchase of foods of hgh qualty Note: Extracton Method: Prncpal Component Analyss. Rotaton Method: Varmax wth Kaser Normalzaton. a Rotaton converged n 3 teratons. 28

29 Table 7b. Rotated Component Matrx(a) for Food Preference Proxy Varables Factor Proxy Varable I lke foods from dfferent countres Ethnc food looks too werd to eat I lke to try new ethnc restaurants At partes, I wll try a new food I am very partcular about the foods I wll eat I am constantly samplng new and dfferent foods 1 (Open to unfamlar foods) 2 (choosey) 3 (afrad of unfamlar foods) I don't trust new foods I wll eat almost anythng If I don't know what s n a food, I won't try t I am afrad to eat thngs I have never eaten before Extracton Method: Prncpal Component Analyss. Rotaton Method: Varmax wth Kaser Normalzaton. a Rotaton converged n 5 teratons

30 Table 8. Correlaton Matrx For Food Safety Proxy Varables I thnk about food safety when I purchase fresh fruts and vegetables Fruts and vegetables produced n the U.S. are more lkely to be safe I thnk about food safety when I purchase meats The U.S. agrcultural food ndustry provdes the safest, most affordable food supply n the world I beleve there are currently too many chemcal pestcde resdues on fresh fruts and vegetables Correlaton Sg. (1-taled) I thnk about food safety when I purchase fresh fruts and vegetables Fruts and vegetables produced n the U.S. are more lkely to be safe I thnk about food safety when I purchase meats The U.S. agrcultural food ndustry provdes the safest, most affordable food supply n the world I beleve there are currently too many chemcal pestcde resdues on fresh fruts and vegetables I thnk about food safety when I purchase fresh fruts and vegetables Fruts and vegetables produced n the U.S. are more lkely to be safe I thnk about food safety when I purchase meats The U.S. agrcultural food ndustry provdes the safest, most affordable food supply n the world I beleve there are currently too many chemcal pestcde resdues on fresh fruts and vegetables