This article was downloaded by: [Dalhousie University]On: 04 October 2013, At: 06:37Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Food Products MarketingPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/wfpm20

Consumer Buying Behavior of GeneticallyModified Fries in GermanyThea Nielsen aa University of Hohenheim, Stuttgart, GermanyPublished online: 18 Dec 2012.

To cite this article: Thea Nielsen (2013) Consumer Buying Behavior of Genetically Modified Fries inGermany, Journal of Food Products Marketing, 19:1, 41-53, DOI: 10.1080/10454446.2013.739552

To link to this article: http://dx.doi.org/10.1080/10454446.2013.739552

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Journal of Food Products Marketing, 19:41–53, 2013Copyright © Taylor & Francis Group, LLCISSN: 1045-4446 print/1540-4102 onlineDOI: 10.1080/10454446.2013.739552

Consumer Buying Behavior of GeneticallyModified Fries in Germany

THEA NIELSENUniversity of Hohenheim, Stuttgart, Germany

A purchasing experiment in which genetically modified and con-ventional fries were offered for sale at mobile fast food stands inGermany was conducted to identify factors influencing the willing-ness of consumers to purchase genetically modified fries. In total,331 purchasing decisions were made: 56.5% decided to purchaseconventional fries, 22.4% genetically modified fries, and 21.1%no preference. A logistic regression model analyzing question-naires found that worry about pesticides, frequency of organic foodpurchases, the acceptability of genetically modified foods with envi-ronmental benefits, and perceptions of health risks from geneticallymodified foods significantly impact the willingness to purchasegenetically modified fries.

KEYWORDS consumer acceptance, genetically modified foods(GMFs), logistic regression model, purchasing experiment, revealedpreferences

INTRODUCTION

This study analyzes buying behavior of consumers in Germany of fries madefrom genetically modified (GM) potatoes resistant to late blight, the mostcommon fungal disease affecting potato species worldwide (Song et al.,2003). Although not yet available on the market, scientists at BASF Plant

This article is based on Thea Nielsen’s master thesis at the Institute of AgriculturalPolicy and Agricultural Markets, Chair of Agricultural Markets and Marketing, University ofHohenheim, Stuttgart, Germany. She is a PhD candidate at the Department of AgriculturalEconomics and Social Sciences in the Tropics and Subtropics, Chair of Rural DevelopmentTheory and Policy, University of Hohenheim, Stuttgart, Germany.

Address correspondence to Thea Nielsen, Wollgrasweg 43, 70593 Stuttgart, Germany.E-mail: Thea.Nielsen@uni-hohenheim.de

41

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

42 T. Nielsen

Science, the John Innes Center, and Wageningen University are developingGM potatoes with a built-in resistance against late blight, significantly reduc-ing fungicide use on potato fields and resulting in several benefits for bothfarmers and the environment. Whether consumers in Germany are willing topurchase fries made from these GM potatoes has been unknown.

Although GM foods (GMFs) have existed for more than two decades andare allowed to be sold in Germany upon regulatory approval, labelled GMFsremain generally unavailable. This stems from a variety of reasons. First,major retailers and discounters have been effectively pressured to create self-imposed bans of GMFs (Dannenberg, Scatasta, & Strum, 2008). Second—asrevealed in surveys and bidding experiments—many Germans are not onlyopposed to GMFs, but are also unwilling to purchase such foods (Dialego,2009; Springer, Mattas, Papastefanou, & Tsioumanis, 2002; Dannenberg et al.,2008). The results from these surveys confirm for retailers and critics ofGMFs that Germans neither agree with GMFs nor intend to buy GMFs. Third,studies on consumer purchasing behavior of GMFs in non-hypothetical, non-laboratory settings are lacking in Germany as well as other countries. Thus,retailers and others within the food supply chain cannot accurately predictthe share of consumers willing to purchase GMFs.

There have been just two studies analyzing consumer buying behaviorof GMFs in Germany in retail settings. One of these studies, published inNature Biotechnology, offered “pretend” GM, conventional and organic fruitfor sale at roadside stands (Knight, Mather, Holdsworth, & Ermen, 2007):when equally priced, 50% of consumers chose organic, 28% conventional,and 22% “spray-free” GM fruit. The other study, published in the newspa-per Südwestrundfunk, offered “pretend” GM bread and GM fries for sale(Südwestrundfunk in Dannenberg et al., 2008): four times more of the “pre-tend” GM bread and over 20 times more of the “pretend” GM fries were soldcompared to the more-expensive conventional bread and fries, respectively.Neither of these purchasing experiments included questionnaires to analyzedeterminants of the willingness to purchase GMFs.

Our study aims to fill an existing knowledge gap—the lack of informa-tion on consumer purchasing behavior of GMFs in retail settings in Germany.Although there have been several surveys and a few bidding experimentswhich have attempted to elucidate consumer purchasing behavior of GMFsin Germany, it is our belief—based on an extensive literature review—thatdue to possible differences between stated and revealed preferences, thesestudies cannot accurately predict whether consumers are indeed willing topurchase GMFs.1 Thus, to determine whether consumers in Germany are

1 See: Belk (1985); Cummings, Harrison, & Rutstrom (1995); Dannenbrg et al. (2008);Kalaitzandonakes, Marks, & Vickner (2005); Knight, Mather, & Holdsworth (2005a); Knight, Mather, &Holdsworth (2005b); Knight et al., (2007); List (2006); Lusk & Norwood (2009); Lusk et al., (2006);Noussair, Robin, & Ruffeux (2004); and Shogren, Fox, Hayes, Roosen (1999).

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

Consumer Buying Behavior in Germany 43

willing to purchase GM fries labelled “environmentally friendlier geneticallymodified,” a purchasing experiment was conducted at eight already exist-ing fast food stands in the Rhein-Neckar region in Germany. Furthermore,to better understand significant determinants of the willingness to purchaseGM fries, questionnaires that had been filled out by participants immediatelyafter they had made their purchasing decisions were analyzed using a logisticregression model.

This article proceeds as follows: the second section describes themethodology, including the experimental set-up and the statistical method-ology; the third section presents the results from the purchasing decisions,questionnaire responses, and the logistic regression model; and the fourthsection discusses the results and their implications.

METHODOLOGY

This section begins with an explanation of the experimental setup and closeswith a description of the statistical methodology used to analyze the data.

Experimental Setup

The purchasing experiment took place at eight already existing mobile fastfood stands in the Rhein-Neckar region. The eight locations included twolarge city centers and three adjacent suburbs, as well as two smaller citiesand a village. At each fast food stand, non-GM fries and “pretend” GM frieswere offered for sale at the same price (C1.60 for 250 grams). This sale offerwas indicated by a small display located on the counter of the booth. On thedisplay was written: Today there are fries from 2 types of potato varieties—with the same taste! 1. From conventional potatoes 2. From environmentallyfriendlier genetically modified potatoes (much less sprays). . . Which frieswould you like?

When consumers stated that they would like to purchase fries but didnot specify which type, the researcher pointed to the display and said,“Today there are fries from two types of potatoes: environmentally friendliergenetically modified potatoes or conventional potatoes. Which fries wouldyou like?” After consumers had decided which type of fries they would liketo purchase or said that they had no preference between the two varieties,they were made fully aware that they were participating in an experiment,that GM fries were not yet for sale and—if they had chosen GM fries—that they could purchase conventional fries. Furthermore, they were askedto fill-out a short questionnaire designed to take approximately three min-utes. Consumers who decided to purchase GM fries were given a slightlydifferent questionnaire from others. People filled out the questionnaires ontheir own; however, the researcher was on hand to answer any questions.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

44 T. Nielsen

In the situation in which consumers chose not to fill out the questionnaire,the researcher wrote down their purchasing decision along with their genderand approximate age.

The experimental set-up was based on Knight et al. (2007), althoughthere are several key differences. Unlike the Knight et al. (2007) study,this study: administered questionnaires to participants after their purchas-ing decisions were made; took place at stands already well-established inthe area; did not include an organic alternative; did not rotate prices; andwas conducted beside other businesses selling food.

Statistical Methodology

Data from the questionnaires was examined using both basic statistical meth-ods and a binary logistic regression model to analyze the willingness topurchase GM fries as well as its determinants. In the binary logistic regres-sion model the dependent variable, WTBGM, represents the willingness topurchase GM fries, equalling 1 if the participant is willing to purchase GMfries and equalling 0 if the participant is not willing to purchase GM fries.Participants were labelled as being willing to purchase GM fries if they eitherchose GM fries or had no preference. On the other hand, participants werelabelled as not willing to purchase GM fries if they chose conventional fries.The probability that WTBGM equals 1 is defined as p. To estimate p from theindependent variables (or “predictors”) we first transform probabilities intoodds, which equal p/(1 – p) (Rodríguez, 2007). Then, we take the logarithmof the odds:

Zi = log(odds) = ln(p/(1 − p)) (1)

where Zi is the log-odds of a participant being willing to purchase GM fries(Rodríguez, 2007). Often, it is clearer to discuss probabilities rather than oddsand thus we can solve for p using Equation 1:

p = exp(Zi)/(1 + exp(Zi)) (2)

Furthermore, we can define the logistic regression model more specificallyas being a function of predictors:

Zi = log(odds) = ln(p/(1 − p)) = ß0 + ß1∗x1+ . . . + ßk∗xk + ε (3)

where x1 to xk are the predictors, ß0 is the intercept, ß1 to ßk are theparameter coefficients and ε is the error term (Stats Direct, 2009). A parame-ter coefficient of a predictor indicates the amount of increase in the predictedlog-odds that a person is willing to purchase GM fries that would be pre-dicted by a one-unit increase in that predictor, holding all other predictors

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

Consumer Buying Behavior in Germany 45

TABLE 1 Definition of Predictors Used in the Logistic Regression Model

Predictor Description

WPE = 0 if more worried about genetic engineering or worried aboutneither pesticides nor genetic engineering

ST = 1 if more worried about pesticides or equally worried aboutpesticides and genetic engineering

HRISK = 0 if thinks GMFs are neither risky nor beneficial, beneficial, or verybeneficial for health

= 1 if thinks GMFs are very risky or risky for healthBIOOFT = 0 if purchases organic food less than often

= 1 if purchases organic food often or alwaysGMENV = 1 if thinks GMFs with environmental benefits are acceptable

= 2 if thinks GMFs with environmental benefits are not acceptable= 3 if does not know if GMFs with environmental benefits are

acceptableAGE = 1 if between 12 and 29 years of age

= 2 if between 30 and 49 years of age= 3 if over 50 years of age

GENDER = 0 if female= 1 if male

constant (Chen at al., 2010). Predictors were based on responses in thequestionnaire on worry about pesticides versus genetic engineering, healthrisk/benefit perceptions of GMFs, frequency of organic food purchases,acceptability of GMFs with environmental benefits, as well as the partici-pant’s age group and gender. Predictors, which are listed and defined inTable 1, are coded to equal 0, 1, 2 or 3, yet most are binary variables.

To obtain the empirical form of the binary logistic regression model, wecan substitute x1 to xk in Equation 3 with the predictors listed in Table 1:

Zi = log(odds) = ln(p/(1 − p)) = ß0 + ß1∗WPEST + ß2∗HRISK

+ ß3∗BIOOFT + ß4∗GMENV + ß5∗AGE + ß6∗GENDER + ε(4)

Once the parameter coefficients in Equation 4 are estimated based onthe questionnaire data, we can analyze significant factors determining thewillingness to purchase GM fries.

RESULTS

This section will first present results from the overall purchasing decisions aswell as responses in the questionnaires that were not included in the logis-tic regression model. Afterwards, results from the logistic regression modelanalysis will be presented.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

46 T. Nielsen

Results from the Purchasing Decisions and Questionnaire Responses

Purchasing decisions were made by a total of 331 customers in the springof 2010 at the eight fast food stands hosting the experiment: 187 (56.5%)chose conventional fries, 74 (22.4%) chose GM fries and 79 (21.1%) had nopreference. Those who were recorded as having no preference had told theresearcher when asked which type of fries they wanted to purchase: “I donot care”; “I have no preference”; or “It does not matter to me.” Two-thirds ofthe participants were men. Furthermore, nearly half of the participants werebetween 30 and 39 years of age. While the sociodemographics of the partic-ipants were not representative of that of Germany, this was not an intentionof the study. Questionnaires were filled out by a total of 235 consumers,yielding a response rate of 71%. The purchasing decisions of participantsfilling out the questionnaire are similar to the purchasing decisions of allparticipants: 53.3% of those filling-out the questionnaire chose conventionalfries, 26% chose GM and 18.7% had no preference.

Reasons why participants chose either conventional or GM fries variedgreatly depending on their product choice. In the questionnaire, participantswere asked why they had chosen to buy GM or conventional fries and weregiven a set of answer choices from which they could select more than one.Over 60% of those choosing GM fries (hereafter referred to as “GM buyers”)chose GM fries out of curiosity while only about 43% did so because ofthe GM fries’ perceived environmental benefit. Moreover, 13% of GM buyersdecided to purchase GM fries because they thought they were healthierthan conventional fries. Nearly half of the buyers selecting conventional fries(hereafter referred to as “conventional buyers”) did so because conventionalfries are produced without the use of genetic engineering, while 27% choseconventional fries out of habit and 18% because they thought conventionalfries are healthier than GM fries.

Although prices could not be rotated because all but one of the mobilefast food stands hosting the experiment were present at each location onceper week with mostly regular customers, participants were asked whetherthey would purchase GM fries sold at a discount/premium. Conventionalbuyers and buyers expressing no preference (hereafter referred to as “nopreference buyers”) were asked in the questionnaire if they would buy GMfries if they were sold at a C0.20 (12.5%) discount. On the other hand, GMbuyers were asked if they would still purchase GM fries if they were soldat a C0.20 (12.5%) premium. Nearly three-quarters of conventional buyersindicated that they would not purchase discounted GM fries. Unexpectedly,a slight majority of GM buyers indicated that they would still purchase GMfries if sold at a premium, whereas 23% said they would not and 26% didnot know.

Because two of the three late blight resistant potato varieties currentlybeing developed rely on genes from a wild Mexican potato variety, the finalquestion about GMFs given to conventional and no preference buyers first

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

Consumer Buying Behavior in Germany 47

provided additional information about these GM potato varieties (that theGM potato was created from genes of a wild Mexican potato, significantlyreducing the need for pesticides) and then asked these buyers if they wouldchange their purchasing decision and buy GM fries after learning this infor-mation. Surprisingly, after knowing this information, nearly one-fourth ofconventional buyers and three-fourths of no preference buyers respondedthey would purchase GM fries. Nevertheless, the responses to this questionas well as the questions asking participants if they would purchase GM friesif sold at a discount/premium are stated preferences rather than revealedpreferences. The purchasing decisions these customers would actually makein a purchasing experiment varying prices and information may differ fromtheir questionnaire responses.

Results from the Logistic Regression Model

An analysis of the purchasing decisions and questionnaire responses usinga binary logistic regression model was undertaken to determine factors sig-nificant in explaining the willingness to purchase GM fries. The softwareSTATA (Version 10.1) was used for the analysis (StataCorp LP, 2009). Thebinary logistic regression model includes 219 observations—all 235 ques-tionnaires were not included as some were not complete. The estimationresults are shown in Table 2. Having affirmed that the selected logistic regres-sion model appropriately fits the data and does not have specification errorsor multicollinearity problems through goodness-of-fit and multicollinearitytests2, we can interpret the results.

TABLE 2 Estimation Results from the Binary Logistic Regression Model

Variable ß Mean S.E. z P > |z| Odds Ratio

WPEST∗ .6258147 .6849 .3370331 1.86 0.063 1.869769HRISK∗∗∗ −1.008141 .5799 .3413816 −2.95 0.003 .3648969BIOOFT∗ .6429699 .2511 .3582072 1.79 0.073 1.902122GMENV∗∗∗ −.7114759 1.6849 .2465858 −2.89 0.004 .4909191AGE −.030975 1.9954 .2093961 −0.15 0.885 .9700599GENDER −.0100268 .6849 .3290809 −0.03 0.976. 9900233_cons 1.011703 − −.6860412 1.47 0.140 2.356319

Notes: ∗indicates the variable is significant at the 90% level and ∗∗∗ at the 99% level. The log-likelihoodintercept-only equals -150.5892; the log-likelihood full model equals -130.1788; and the likelihood ratiochi-square equals 40.82 with 6 degrees of freedom. The Adjusted Count R-squared equals 0.337. TheHomer and Lemeshow’s goodness-of-fit chi-square test statistic equals 10.96 with 8 degrees of freedomand a p-value of 0.2042. The goodness-of-fit test using linear predictors to rebuild the model found that_hat P > |z| = 0.000 and _hatsq P > |z| = 0.429. All Variance Inflation Factors are between 1.02 and1.36, with a mean VIF of 1.14. The condition number for the matrix of predictors is 6.15.

2 Three goodness-of-fit tests were employed: the likelihood ratio test, the Hosmer and Lemeshow’sgoodness-of-fit test and a test using linear predictors to rebuild the model. Multicollinearity was tested

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

48 T. Nielsen

The odds ratios of WPEST and BIOOFT are greater than one, indicatingthat when these variables increase by one-unit, the odds that a person iswilling to purchase GM fries increase with all other predictors held constantat certain values. On the other hand, the odds ratios for the variables HRISK

and GMENV are less than one, indicating that when these variables increaseby one unit, the odds that a person is willing to purchase GM fries decreasewith all other predictors held constant at certain values. For example, theodds ratio for the variable BIOOFT indicates that the odds that a person iswilling to purchase GM fries increases by 90.2% when the variable BIOOFT

increases from 0 to 1. Thus, people who purchase organic food often oralways have much higher odds of being willing to purchase GM fries com-pared to people who purchase organic food less often. The odds ratio of thevariable HRISK means that when the variable HRISK increases from 0 to 1, theodds that the person is willing to purchase GM fries increase by a factor of0.365 (or, equivalently, decrease 63.5%) with all other predictors held con-stant at certain values. In other words, people who perceive GMFs as riskyor very risky for their health have lower odds of being willing to purchaseGM fries than those who think otherwise.

The odds ratio for the variable GMENV is more difficult to interpret thanthe other predictors, as unlike the other significant predictors it is not a binaryvariable. The odds ratio for the variable GMENV equals 0.491, indicating thatfor every one-unit increase in the variable GMENV, the odds of being willingto purchase GM fries increase by a factor of 0.491 (or, equivalently, decrease50.9%) with all other predictors held constant at certain values. For a two-unitincrease in the variable GMENV, the odds of being willing to purchase GMfries increase by a factor of 0.491 squared (or, equivalently, decrease 75.9%)with all other predictors held constant. The influence of this predictor on thewillingness to purchase GM fries will be discussed in further depth below.

The impact of individual predictors on the willingness to purchase GMfries can also be demonstrated by varying one predictor while holding allother predictors at their mean values and then calculating the predicted prob-ability that an individual with these specifications is willing to purchase GMfries. This can by achieved by first incorporating the estimated parametercoefficients from Table 2 into Equation 4, which results in the followingequation:

Zi = log(odds) = ln(p/(1 − p)) = 1.011703 + .6258147∗WPEST

− 1.008141∗HRISK + .6249699∗BIOOFT − .7114759∗GMENV

− .0303975∗AGE − .0100268∗GENDER

(5)

using a Variance Inflation Factor and by examining the conditioning of the matrix of predictors. Resultsfrom these tests are in the notes of Table 2.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

Consumer Buying Behavior in Germany 49

TABLE 3 Predicted Probabilities of Being Willing to Purchase GM Fries

Value of Predictor WPEST HRISK BIOOFT GMENV

0 .3368 .5831 .3988 −1 .4870 .3379 .5578 .55932 − − − .38383 − − − .2342

Note: All other predictors are held at their mean values.

Then we can replace parameters with either their mean values or with aspecific value and calculate p. The predicted probabilities of being willingto purchase GM fries when specifying particular predictors at certain valuesand holding all other parameters at their means are shown in Table 3.

The predicted probabilities for the variable GMENV are particularly inter-esting because they show that people who do not know if GMFs withenvironmental benefits are acceptable (GMENV = 3) are less likely to bewilling to purchase GM fries compared to people who do not think suchfoods are acceptable (GMENV = 2). This result and others are discussed inthe following section.

DISCUSSION OF RESULTS

The overall purchasing decisions indicate that although the majority of con-sumers did not prefer GM fries, 43.5% were willing to purchase GM fries.Furthermore, half of GM buyers stated that they would be willing to pur-chase GM fries at a premium, indicating that a niche market for GM friesmay exist in Germany.

It is surprising that 13% of GM buyers chose GM fries because theyconsidered them to be healthier than conventional fries. Exactly whathealth aspect these GM buyers believed to be better in GM fries comparedto conventional fries remains unknown. Nevertheless, it is reasonable tohypothesize that these GM buyers believed that potatoes produced withless pesticides are healthier. Even though potatoes are highly processed andpeeled before making french fries, thereby significantly reducing pesticideresidues (Geetanjali, Santosh, & Naik, 2009; Soliman, 2001), it is possi-ble that some participants in this study harbored fears that any pesticidesincorporated into the developing potato tuber would persist in the finalproduct—french fries. Thus, it follows that some people may have chosenGM fries produced with “much less sprays” because of their concern for thepossible adverse effects of these pesticides. Furthermore, estimated predictedprobabilities of being willing to purchase GM fries when varying the variableWPEST (representing worry about pesticides versus genetic engineering) alsosupport the idea that people may have chosen GM fries over conventionalfries because of pesticide concerns.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

50 T. Nielsen

There are several explanations why such a high proportion of conven-tional and no preference buyers responded that they would change theirpurchasing decision after receiving the additional information about the GMpotato variety (that it was produced with the genes of a wild Mexican potatovariety, significantly reducing pesticides). One explanation is that these buy-ers found it appealing that the GM potato variety was created using genesfrom another potato variety. These perceptions of the “naturalness” of theGM potato variety may also have been augmented by the word “wild.”Another explanation is that participants did not fully comprehend the pes-ticide reduction when making their purchasing decision, but then becamemore aware of this when reading the question. Thus, it is possible thatthe environmental benefit of reduced pesticide usage was the determinantfor why these buyers indicated that they would change their purchasingdecision after knowing the additional information. The responses to thisquestion indicate that more information given to the public on the origin andenvironmental benefit of GM potatoes resistant to late blight could increaseacceptance in Germany of GM fries made from these potatoes.

The results from the logistic regression model produced two surprisingfindings. The first is that the variable GMENV demonstrates that people whodo not know if GMFs with environmental benefits are acceptable to themhave a lower probability of being willing to purchase GM fries compared topeople who think that GMFs are not acceptable. Therefore, if one wants toincrease the number of consumers in Germany willing to purchase GM fries,more communication about the environmental effects of GMFs should beprovided so that less people are undecided about the acceptability of GMFswith environmental benefits. This finding also advances the rationale thatproviding more information about GMFs could help increase their accep-tance. The second unexpected finding from the logistic regression model isthat people who buy organic food often or always are more likely to be will-ing to purchase GM fries than those who buy organic food less often. Thismay be because organic food buyers in Germany may be more concernedabout pesticide residues on foods and the environment, and the GM frieswere advertised as “environmentally friendlier (much less sprays).” In fact,a study conducted in 1999 found that for 74% of organic food customers inGermany a key influence for purchasing organic food was health reasons,while for 51% a key influence was environmental considerations (CMA citedin Fuchshofen & Fuchshofen, 2000).

The result that perception of personal health risks is a significant factordetermining the willingness to purchase GM fries confirms the ongoing riskanalysis people perform when forming attitudes toward GMFs. Perhaps thispersonal risk analysis plays such a significant role because people tend tooverweight low probability events and underweight intermediate to highprobability events (Kahneman & Tversky, 1979), while a risk perceived asinvoluntary is more threatening than a risk perceived as voluntary (Freweret al., 2004). Thus, people may be more risk-averse toward GMFs for which

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

Consumer Buying Behavior in Germany 51

there is a very low probability of a loss occurring from its consumption andfor which risks are of an involuntary nature. Nevertheless, the reality is thatrisks do impact purchasing decisions for GMFs. To placate risk concerns ofGMFs, broader communication should be provided concerning the lack ofscientific proof that GMFs pose risks to either the environment or humanhealth.

Applying the results from this study to other countries should be donecautiously as support for GMFs varies greatly from country to country. Forexample, the most recent survey completed for the European Commissionon European’s opinions of GMFs found that support – defined as outrightsupport and risk tolerant support—for GMFs in the EU-25 was between 12%and 74%, depending on the country (Gaskell et al., 2006). Moreover, theimplications of our experiment undertaken in Germany cannot be applied tothe United States and Canada where GMFs are less of a novelty item and donot even have to be labelled as such. To better understand the willingnessof consumers to purchase GMFs and the dynamics behind their purchasingdecisions in other countries, more purchasing experiments should be con-ducted in different countries as well as with a variety of GMFs, both freshand processed.

In conclusion, this purchasing experiment represents an addition to thesmall handful of other purchasing experiments analyzing the willingness ofconsumers to purchase GMFs. When french fry consumers in Germany weregiven a choice between purchasing equally priced GM and conventionalfries in an actual retail setting, 43.5% were willing to purchase GM fries.Analysis of responses in the questionnaire by both simple and econometricmethods yields several interesting conclusions about the willingness to pur-chase GMFs in Germany: (1) one-fourth of conventional and three-fourths ofno preference buyers indicated that they would buy GM fries when explic-itly provided information about the origin of the GM potato variety and itspesticide reduction; (2) people who do not know if GMFs with environmen-tal benefits are acceptable are much less likely to be willing to purchaseGM fries compared to people who do not think such foods are acceptable;(3) people who buy organic food more than often are more likely to bewilling to purchase GM fries than those who buy organic food less often.These findings suggest that disseminating more information about GMFs topotential consumers in Germany is vital to winning their approval.

REFERENCES

Belk, R. W. (1985). Issues in the intention-behavior discrepancy. Research inConsumer Behavior, 1, 1–34.

Chen, X., Ender, P., Mitchell, M., & Wells, C. (2010). Logistic regression with stata.Stata Web Books. Available at: http://www.ats.ucla.edu/stat/stata/webbooks/logistic/.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

52 T. Nielsen

Cummings, R. G., Harrison, G. W., & Rutström, E. E. (1995). Homegrown valuesand hypothetical surveys: Is the dichotomous choice approach incentive-compatible? The American Economic Review, 85(1), 260–266.

Dannenberg, A., Scatasta, S., & Strum, B. (2008). Does mandatory labeling of genet-ically modified food grant consumers the right to know? Evidence from aneconomic experiment (Working Paper 08–29). Mannheim, Germany: Centre forEuropean Economic Research.

Dialogo. (2009). Genmanipulierte nahrungsmittel: Eine befragung der Dialego AG.(Genetically modified food: A survey from Dialego AG), Aachen, Germany.

European Commission. (2008). Do European consumers buy GMFs? (ProjectNo. 518435. Framework 6). London, UK: King’s College.

Frewer, L., Lassen, J., Kettlitz, B., Scholderer, J., Beekman, V., & Berdal., K. G. (2004).Societal aspects of genetically modified foods. Food and Chemical Toxicology,42, 1181–1193.

Fuchshofen, W. H., & Fuchshfen, S. (2000). Organic Trade Association’s export studyfor U.S. organic products to Asia and Europe. Retrieved from http://www.ota.com/organic/mt/export_form.html

Gaskell, G., Allansdottir, A., Allum, N., Corchero, C., Fischier, C., . . . Wagner,W. (2006). Europeans and biotechnology in 2005: Patterns and trends.Eurobarometer, 64(3).

Geetanjali, K., Santosh, S., & Naik, S. N. (2009). Food processing a tool to pesticideresidue dissipation—A review. Food Research International, 42, 26–40.

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision underrisk. Econometrica, 47(2), 263–292.

Kalaitzandonakes, N., Marks, L. A., & Vickner, S. S. (2005). Sentiments and actstoward GMOs. International Journal of Biotechnology, 7(1–3), 161–177.

Knight, J. G., Mather, D. W., & Holdsworth, D. K. (2005a). Consumer benefits andacceptance of genetically modified food. Journal of Public Affairs, 5, 226–235.

Knight, J. G., Mather, D. W., & Holdsworth, D. K. (2005b, December). Acceptanceof genetically modified food with consumer benefits. Paper presented at the2006 ANZMAC Conference on Consumer Behaviour, Bisbane, Australia.

Knight, J. G., Mather, D. W., Holdsworth, D. K., & Ermen, D. F. (2007). Acceptanceof GM Food—An experiment in six countries. Nature Biotechnology, 25(5),507–508.

List, J. A. (2006). Field experiments: A bridge between lab and naturally occurringdata. Advances in Economic Analysis and Policy, 6(2), Article 8.

Lusk, J. L., & Norwood, F. B. (2009). Bridging the gaps between laboratory experi-ments and naturally occurring markets: An inferred valuation method. Journalof Environmental Economics and Management, 58(2), 236–250.

Lusk, J. L., Traill, W. B., House, L. O., Valli, C., Jaeger, S. R., Moore, M. M., &Morrow, B. (2006). Comparative advantage in demand: Experimental evidenceof preferences for genetically modified food in the United States and EuropeanUnion. Journal of Agricultural Economics, 57(1), 1–21.

Noussair, C., Robin, S., & Ruffieux, B. (2004). Do consumers really refuse to buygenetically modified food? The Economic Journal, 114, 102–120.

Rodríguez, G. (2007). Generalized linear model: WWS509 [Lecture notes]. Princeton,New Jersey: Princeton University, Woodrow Wilson School of Public andInternational Affairs.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013

Consumer Buying Behavior in Germany 53

Shogren, J. F., Fox., J. A., Hayes, D. J., & Roosen J. (1999). Observed choices for foodsafety in retail, survey, and auction markets. American Journal of AgriculturalEconomics, 81, 1192–1199.

Soliman, K. M. (2001). Changes in concentration of pesticide residues in potatoesduring washing and home preparation. Food and Chemical Toxicology, 39(8),887–891.

Song, J., Bradeen, J. M., Naess, S. K., Raasch, J. A., Wielgus, S. M., . . . Jiang, J.(2003). Gene RB cloned from Solanum bulbocastanum confers broad spectrumresistance to potato late blight. Proceedings of the National Academy of Sciences,100(16), 9128–9133.

Springer, A., Mattas, K., Papastefanou, G., & Tsioumanis, A. (2002, August).Comparing consumer attitudes toward genetically modified food in Europe.Paper presented at the 10th European Association of Agricultural Economists’Congress, Zaragoza, Spain.

StataCorp LP. (2009). Stata Statistical Software (Release 10.1). [Computer software].College Station, Texas.

StatsDirect. (2009). Logistic regression. Retrieved from www.statsdirect.com/help/regression_and_correlation/logi.htm

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

06:

37 0

4 O

ctob

er 2

013