Measuring Preferences for Genetically Modified Food Products By Charles Noussair, Stephane Robin and Bernard Ruffieux* Abstract In this chapter, we review the results of three experimental studies we have conducted to investigate consumers’ willingness to pay for food products containing genetically modified organisms (GMOs). Participants in the experiment are a demographically representative sample of French consumers. We observe that about 65% of our sample is willing to purchase products containing GMOs if they are sufficiently inexpensive. The results contrast sharply with surveys that indicate overwhelming opposition to GM foods. The data suggest that there is substantial surplus to be gained from the segregation of the market for food products into a GMO-free segment and a segment that allows the inclusion of GM ingredients. We also find that individuals do not read the labels for GM content when they make purchases, suggesting that they do not believe that the products they are purchasing contain GMOs. Finally, a study comparing two different systems of eliciting willingness-to-pay information is described, and evidence is provided that suggests that the Vickrey auction is more effective at eliciting willingness to pay information that the Becker-DeGroot-Marschak mechanism.
1. Introduction The introduction of genetically modified organisms (GMOs) into food products has been a major political issue for over a decade in many parts of the world. Regulatory authorities such as the FSA in the United Kingdom, the FDA in the United States and the DGAL in France, on the basis of recommendations from the scientific community, have recognised that the GMO
*
Noussair: Department of Economics, Emory University, 1602 Fishburne Dr., Atlanta, GA 30322-2240, USA,
[email protected]. Robin: GATE-CNRS, Université Lyon 2, 93, chemin des Mouilles, 69131 ECULLY, France,
[email protected]. Ruffieux: Ecole Nationale Supérieure de Génie Industriel, 46 Avenue Félix Viallet, Grenoble Cedex 1 38031, France,
[email protected]. This chapter is a summary of three articles (Noussair et al. 2002, 2004a, 2004b). The program “ Pertinence économique et faisabilité d’une filière sans utilisation d’OGM”, as well as The French National Institute of Agronomic Research (Program on Consumer Behaviour) provided research support for this project. We would like to thank Isabelle Avelange, Yves Bertheau, Pierre Combris, Sylvie Issanchou, Egizio Valceschini, and Steve Tucker for valuable comments and assistance.
1
products currently available are safe for the consumer and the environment. Moreover, there is a consensus among scientists that biotechnology has the potential to create products that will enhance nutrition, increase crop yields, and reduce the use of toxic pesticides and herbicides. However, polling of consumers consistently indicates a high degree of hostility to the presence of GMOs in the food supply. For example, Noussair et al. (2001) report that 79% of French respondents either agreed or mostly agreed with the statement “GMOs should simply be banned”. 89% were opposed to the presence of GMOs in food products, 89% in livestock feed, 86% in medicine, 46% in food packaging, and 46% in fuels. In the UK, surveys show a similar pattern. Moon and Balasubrimanian (2001) report the results of a survey conducted of 2,600 consumers in the UK. 38% of their respondents indicated that they were in support of agrobiotechnology and 46% were opposed. A poll of Americans conducted by ABC News in June 2001 found that 35% believed that GM foods were safe to eat, while 52% believed that they were not. The results of the fifth Eurobarometer survey on biotechnology and the life sciences (Eurobarometer, 2003) indicate that a majority of Europeans would not buy or eat GM foods. Between 30% and 65% percent of the respondents in every EU country reject every reason for buying GM foods listed in the survey. Greece, Ireland, and France are the countries in which the highest percentage of respondents rejects GM foods. Survey responses indicate that aversion to GMOs is based on both private considerations, such as potential health risk and a preference for natural foods, as well as social dimensions, such as environmental effects and ethical concerns. The unfavourable view has been exacerbated by the spread of the “mad cow” epidemic, the lack of benefit that the first generation of GMOs provides to the consumer, and the initial introduction of GMOs without the public’s knowledge. The dichotomy between scientific recommendations and public opinion has complicated the formulation of government policy with respect to GMOs, since in a democratic system public opinion must be taken into account in addition to the scientific merits of the policy and the market pressures in the economy. However, there is reason to question whether the anti-GMO sentiment expressed in surveys would be reflected in actual purchase behavior. It is known that individuals’ decisions can differ drastically between when they are hypothetical, as in a contingent valuation study or other survey, and when they involve a real commitment to purchase (see for example Neill et al. 1994; Cummings et al. 1995; Brookshire and Coursey, 1987; List and Shogren, 1998; or List and Gallet, 2001). Furthermore, most surveys do not inquire about actual purchase decisions at specific prices, while contextual cues or small changes in information provided to survey respondents may change results dramatically (Aizen et al., 1996).
2
Surveys about preferences over public goods, such as the preservation of GMO-free crops, may be particularly suspect. Sagoff (1988), Blamey et al. (1995) and Nyborg (2000) argue that survey and hypothetical contingent valuation measurement techniques for public goods do not accurately reveal participants’ willingness-to-pay. Surveys place respondents in the role of citizens, who make judgements from society’s point of view, rather than consumers, who make actual purchase decisions. Thus the two instruments, surveys and purchase decisions, measure different variables. In addition, even if provision or preservation of a public good is valuable to an individual, it may not be reflected in his willingness-to-pay because of the free rider problem (Stevens et al., 1991; Krutilla, 1967). A well-documented example of a dichotomy between surveys and consumer behaviour was observed during the introduction of recombinant bovine somatropin (rbST), a bovine growth hormone, into milk production in the United States in 1993. Surveys indicated that a majority of consumers had a negative opinion of the technique, primarily on ethical grounds. On the basis of the survey data, analysts predicted a 20% decline in total milk consumption. However, there was no decrease in actual milk consumption after the introduction of the technique (Aldrich and Blisard, 1998). The focus of the first study described in this chapter (Noussair et al., 2004a) is to consider, using experimental methods, the extent that actual decisions to purchase food products are affected by the presence of GMOs. We study purchasing behaviour of consumers using a laboratory experiment designed to elicit and compare the willingness to pay for products that are traditional in content and labelling, that are explicitly guaranteed to be GMO-free, and that contain GMOs. We also consider buyer behaviour with respect to different thresholds of maximum GMO content. The second study surveyed here (Noussair et al., 2002) is motivated by the fact that, despite the hostility toward GMOs that is ubiquitous in survey data, sales do not decrease when the label reveals that the product contains GMOs, for those few GM products that have been put on the European market, where GM content must be indicated on the product label. We use an experiment to consider whether the absence of a reaction in demand to the current labeling of products is due to the fact that most customers do not notice the labeling, and thus do not realize that the product they are purchasing contains GMO’s. The experimental approach is particularly appealing here because of the absence of field data. The current policy of most major European retailers not to carry GM foods, which has resulted from pressure of activists and the media, means that it is very difficult to estimate product demand for foods containing GMOs using field data from European countries. For the few GM products that are available, there is reason to believe that consumers are unaware of the
3
labelling of GM content. Furthermore, in the US, where the vast majority of GM food is sold, demand for GMOs cannot be inferred from market data since GM content is not indicated on the labelling. We are unaware of any previous estimates of consumer demand for the GMO-free characteristic in food products other than those obtained from experimental studies (see Lusk et al., 2001; Huffman et al., 2001; Lusk et al., 2005). However, previous work (see for example Shogren et al., 1999) suggests that experiments provide a good alternative method to study product demand in general, and that the artificial setting of the lab does not drastically alter consumer behaviour. Moreover, experimental methodology provides an environment to measure individual preference by controlling for noise and other confounding factors. In particular, researchers in the lab are able to precisely control the information communicated about product characteristics, which is not possible in the field.
2. Background 2.1. Policy Issues: Segregation and Thresholds In response to the tension between scientific and public opinion on the issue of GM foods, the policy adopted by most European governments has been to declare a moratorium on approval of new GM products for cultivation and sale. For the few products that have already been approved, their policy has been to segregate GM and GMO-free products at all stages of production, to require labelling of products containing GMOs, and to allow the market to determine how much of each type of product is sold. Any food product sold in the European Union for human consumption that contains an ingredient that consists of more than 0.9% GMOs must be labelled “contains GMOs”. There is no GM produce currently sold in Europe and the only GM products for sale appear as ingredients in processed foods. Currently in France, three types of corn are authorised for cultivation. One type of corn and one type of soybean are authorised for importation. In the UK, in addition to corn and soybeans, one type of GM tomato is authorised for importation and use in tomato puree. No GM crops are grown commercially in the UK. In contrast, in the United States, as of early 2002, about two-dozen different GM fruits, vegetables, and grains were being cultivated. In the US, there are no specific regulations for biotech products, which are subject to the same regulations as other products. See Caswell (1998, 2000) for a discussion of policy issues relating to the labelling of GM products. In Europe, when a product is classified under current law as containing GMOs, it must carry in its list of ingredients the statement “produced from genetically modified …”. A note at the end of the list of ingredients, specifying the genetically modified origin, is also considered sufficient, as long as it is easily legible. The size of the letters must be at least as large as those in 4
the list of ingredients. To account for the incentive of producers to make the labels indicating their products’ positive characteristics as prominent and those revealing the unfavorable characteristics as discreet as possible, regulators have imposed strict conditions on the size, color, and positioning of information on packaging. Although the current policy of segregation and mandatory labelling is free-marketoriented in that it offers consumers a choice, some economists might view it as an inefficient policy. Segregating the entire process of production is costly to farmers and firms throughout the production chain, especially in the upstream part of the chain, which consists of the seed producers, farmers, and primary processors. In the United States, according to the US Department of Agriculture, segregation costs have been estimated at 12% of the price of corn and 11% of the price of soybeans in the year 2000. Buckwell et al. (1999) find that in general, identity preservation for speciality crops increases final costs by between 5% and 15%. Since there is no hard evidence that the GMOs that regulatory authorities have approved are harmful either to health or to the environment, it can be argued that the expenditure represents a deadweight loss. The two main alternatives to this segregation of the market are (a) to ban GM varieties entirely, or (b) to ban labelling, effectively making GM and non-GM varieties indistinguishable from each other from the viewpoint of a consumer. Both of these policies have potential downsides. Banning new GMOs may be inefficient if there are welfare gains from the adoption of biotechnology that are foregone. Indeed, the studies that have estimated the gains from the adoption of biotechnology in farming in the United States have found then to be considerable. Anderson et al. (2000) estimate the gains from the introduction of biotechnology at $1.7 billion per year for cotton, $6.2 billion per year for rice, and $9.9 billion per year for coarse grains. For soybeans grown in the American midwest, savings to farmers from the adoption of herbicide tolerant soybeans have been estimated at 60 million dollars annually (Lin et al., 2000, 2001). Lin et al. also estimate that the 60 million dollars in savings constituted 20% of the overall welfare gain to all parties (US farmers, rest of world farmers and consumers, the gene producers, and the seed companies). Falk-Zepeda and al., (2000) estimate that the welfare gains from the adoption of Bt corn in the US for the year 1996 equalled 240.3 million dollars. Of this total, 59 % went to U.S. farmers. The gene developer received 21%, U.S. consumers 9 %, producers and consumers in other countries 6%, and the seed producer 5%. Traxler et al. (2000) find that the surplus from the use of Round up Ready soybeans in the US in 1997 was distributed in the following manner: 50% to US farmers, 8% to US consumers, 22% to the gene developer, 9% to the seed companies, and 12% to foreign consumers and producers. Lence and Hayes (2001), using simulation techniques, provide estimates of potential welfare gains and anticipated costs for the United
5
States from the cultivation of GM crops. Under many of their parameterizations, both overall consumer and producer welfare is greater after the introduction of GM technology. On the other hand, if the production tracks are not segregated or labelling of GMO content is interdicted, as it is in the United States, a “lemons” scenario may result (Akerlof, 1970). The GMOs currently on the market were introduced for agronomic reasons and the foods containing them are indistinguishable from conventional foods to the consumer in the absence of labelling information. Since GMOs lower production costs, producers have an incentive to insert them into the food supply. If consumers value foods containing GMOs less than foods that do not contain GMOs, they will be unwilling to pay more for an unlabelled product than an amount that reflects the presence of GMOs. This would cause the market for non-GMO varieties to disappear, reducing social welfare by eliminating potential gains from trade. Furthermore, it could potentially cause a market collapse for entire products. If a firm cannot disclose that its product uses no ingredients that contain GMOs, it might replace ingredients that consumers believe may contain GMOs with those that cannot contain GMOs. This could eliminate the entire market for many products, such as soy lecithin, corn syrup, and cornstarch. From an economist’s point of view, the appropriate policy depends in part on the relative sizes of consumer and producer surpluses and the costs of implementing different policies. The surplus calculation hinges on whether the actual purchase behaviour of consumers corresponds to the polling data. If, as suggested by the polls, a large majority of consumers is unwilling to purchase products containing GMOs, banning GMOs is probably the best option, as the expense of creating two tracks of production would not be justified. On the other hand, if the large majority of consumers behave as if they are indifferent to GMOs, or would purchase products made with GMOs if they sold at lower prices, the production tracks could be safely integrated with little social cost. However, if a considerable segment of the market refuses to purchase products containing GMOs at any price, but another large segment would purchase GM products if they were cheaper, separation of the production tracks and the enforcement of mandatory labelling of products containing GMOs would be worth the expense. Under a policy of segregation, the threshold level of GMO content, above which a product is considered to be bioengineered, must be specified. Because of the ease of contamination throughout the production chain, it is impossible to intentionally make any product, in whose manufacture GMOs are already authorised, without any trace of GMOs. This technological constraint requires the specification of a threshold above zero, below which a product is to be considered as GMO-free, and above which the product must be labelled as containing GMOs. The lower the threshold, the greater is the cost of production of GMO-free
6
products. The increase involves the cost of producing very pure seeds, isolating parcels of land, and cleaning storage and transportation containers. The marginal cost of lowering the threshold may be justified if consumers have a strong preference for a low threshold.
2.2. A Methodological Issue: How to measure willingness-to-pay? Willingness-to-pay information is typically elicited with a demand revealing bidding mechanism, such as the Vickrey auction or the Becker-DeGroot-Marschak mechanism, both described later in this section. The dominant strategy of truthful bidding and the commitment of real money create an incentive to truthfully reveal limit prices, regardless of the risk attitude of the bidder and the strategies other participants use. A demand-revealing auction has the advantage over the study of purchase decisions with field data that it allows an individual’s limit price to be measured directly. Observing only whether or not an individual purchases a product at the current market price in a store merely establishes whether or not his limit price exceeds the current market price. Accurate willingness-to-pay information is particularly useful for new products because other sources of demand estimates on which to base profit or cost-benefit calculations are not readily available. Experimental economists have employed demand revealing auctions to study limit prices for goods as varied as consumer products (see for example Hoffman et al., 1993; Bohm et al., 1997; List and Shogren, 1998; and List and Lucking Reiley, 2000), food safety (Hayes et al., 1995; Fox et al., 1998; Busby et al., 1998; Huffman et al., 2001; and Lusk et al., 2001), and lotteries (Grether and Plott, 1979; Cox and Grether, 1996). In this research, we use the second price sealed bid auction, also called the Vickrey auction (Vickrey, 1961), and the Becker-DeGroot-Marschak (BDM) mechanism (Becker et al., 1964). In a second price sealed bid auction, each member of a group of potential buyers simultaneously submits a bid to purchase a good. The agent who submits the highest bid wins the auction and receives the item, but pays an amount equal to the second highest bid among the bidders in the auction. In a BDM, each subject simultaneously submits an offer price to purchase a good. Afterwards, a sale price is randomly drawn from a distribution of prices with support on an interval from zero to a price greater than the anticipated maximum possible willingness-to-pay among bidders. Any bidder who submits a bid greater than the sale price receives a unit of the good and pays an amount equal to the sale price. There is a substantial literature studying the behavior of the two mechanisms in the laboratory when university student subjects are bidding for goods. Some of this research has used the technique of induced values (Smith, 1982) to create limit prices for fictitious goods. The experimenter offers a guarantee that bidders can resell goods at prices that are specified in 7
advance, should they purchase the items in the auction. Several authors, including Coppinger et al. (1980), Cox et al. (1982), Kagel et al. (1987) and Kagel and Levin (1993), have studied the behavior of the Vickrey auction, and other authors, including Irwin et al. (1998) and Kellar et al. (1993) have studied the BDM process using goods with induced values. These studies reach a variety of conclusions about bids relative to valuations, and some suggest that average bids are biased away from valuations. For example Kagel et al. (1987) and Kagel and Levin (1993) find that most winning bids in the Vickrey auction are higher than valuations. Irwin et al. (1993) find that the BDM process is more successful at eliciting true valuations for certain distributions of sale prices than others. Furthermore, all of the studies show that there is heterogeneity in bidding behavior that leads to a dispersion of bids relative to valuations. In the case of auctions for goods with homegrown (and therefore unobservable) valuations, such as consumer products, the evidence that bids tend to differ from valuations is indirect. Bohm et al. (1997) find that bids in the BDM are sensitive to the choice of endpoints of the distribution of possible transaction prices. List and Shogren (1999) find that bids in the Vickrey auction tend to increase as the auction is repeated, which suggests a bias in bidding either in the early or in the late periods. Rutstrom (1998) finds that the two mechanisms generate different mean bids for the same objects, indicating that at least one of the two must be biased. The research we have conducted on preferences for GM products allows us to compare these auctions within a similar environment. We evaluate and compare the auctions according to the following criteria. (1) Does either or both of the systems contain a bias toward under or overrevelation of true valuations? (2) Under which system do individuals on average bid closer to their true valuations? (3) Under which system is convergence by repetition toward demand revelation, if it occurs, more rapid? We pose these questions under specific conditions, when the population considered is a diverse sample of the population, when the goods considered have induced valuations, and when specific training procedures are in effect that our experience and intuition suggest would enhance the demand revelation performance of the mechanisms (see Noussair et al., 2004b, for more detail).
3. Methodology 3.1 The Participants The participants in the experiments were a demographically representative sample of 209 consumers in the Grenoble, France area. 97 subjects participated in experiment 1 and 112 participated in experiment 2. Twenty-six sessions comprised the two experiments, and each session took approximately two hours. The ages of the subjects ranged between 18 and 75 years, 8
and averaged 33 years. 53% were female. The socio-economic level of the sample was representative of the French urban population. At the time of recruitment, subjects were invited to come to the laboratory to sample food products for a government research project. Only individuals who made the food purchasing decisions in the household were permitted to participate. We recruited only individuals who were regular consumers of the products we used in the experiment. At the time of recruitment, subjects received no indication that the experiment concerned GMOs or potential risks to the food supply.
3.2. The BDM mechanism and the Vickrey auction In experiment 1, we used the Becker-DeGroot-Marschak (BDM) mechanism to elicit willingness-to-pay information. As mentioned earlier, in the BDM there is an optimal strategy for a bidder to bid his valuation, regardless of his risk attitude. Therefore in principle, the mechanism has the ability to reveal bidders’ valuations. The rules of the BDM mechanism are simple. Each subject simultaneously submits a bid to the experimenter in a closed envelope, indicating a price at which he offers to purchase one unit of the good offered for sale. The experimenter then randomly draws a sale price from a pre-specified interval, from zero to a price greater than the maximum possible willingness to pay among bidders. Any subject who submits a bid greater than the sale price receives an item and pays an amount equal to the sale price. The others do not receive units and make no payment. In experiment 2, we used Vickrey auctions to elicit willingness-to-pay information. In a Vickrey auction, each subject simultaneously submits a bid to purchase a good. No communication between subjects is allowed during the bidding process. The agent who submits the highest bid wins the auction, and pays an amount equal to the second highest bid among the bidders in the auction. The other bidders do not receive items and pay zero. Each bidder has a dominant strategy to truthfully bid an amount equal to his willingness-to-pay (Vickrey, 1961).
3.3. The Training Phase Both experiments 1 and 2 began with a training phase to help subjects to learn to use the dominant strategy for the mechanism employed. This training was similar for each mechanism and proceeded in the following manner. At the beginning of a session, each subject received 100 francs (roughly US$14) in cash. Subjects then participated in several BDM or Vickrey auctions, depending on which mechanism was in effect for the session, in which they bid for fictitious items. The fictitious items had induced values (see Smith, 1982, for an exposition of induced value theory). Before the auction took place, each subject received a sheet of paper that indicated 9
an amount of money, for which he could redeem a unit of the fictitious item from the experimenter, should he purchase it in the auction. The induced value differed from subject to subject and was private information. The ability to redeem an item from the experimenter induced a limit price in the auction, since a subject’s payoff if he won the auction equalled the induced value minus the price he paid. The inclusion of the auctions with induced values had three objectives: (a) to teach the subjects, and verify their comprehension of, the rules of the auction, (b) to reduce the biases and noise that tend to arise in bidding behaviour, and (c) to show subjects that the auction involved transactions where real money was at stake. The dominant strategy of bidding one’s valuation in the auctions is not at first obvious to most subjects. We chose not to directly inform the subjects of the dominant strategy. Instead, we used a technique intended to encourage subjects to come to understand the strategies that constitute optimal behaviour on their own. After subjects submitted their bids (and the experimenter drew a selling price in the case of the BDM), the experimenter wrote all of the valuations on the blackboard, and asked subjects if they could identify their own valuations and to predict which subjects would be receiving units of the good based on the valuations displayed. Then the experimenter recorded the submitted bids on the blackboard next to the corresponding valuations. He posed the following questions to the group of subjects, who were free to engage in open discussion on the topics. a) Which subjects received units in the auction? b) How much did the winners pay? c) Did anyone regret the bid he submitted? After the discussion, each of the winners received, in full view of all participants, an amount of money equal to his induced value minus the price he was required to pay. The cash was physically placed on the desk in front of the subject after the auction. A series of identical auctions was conducted using the same procedure. The valuations in each period were randomly drawn from a uniform distribution whose endpoints differed in each period. The auctions continued until at least 80% of the bids were within 5% of valuations. We ended the training phase of each session with an auction of an actual consumer product, a bottle of wine, whose label was visible. After bidding, all of the bids were posted, but there was no discussion as in the earlier induced value auctions. However, as in the induced value auctions, the sale price was drawn, the winners were announced, and the transactions were implemented immediately. There are two reasons that we added this auction to the training phase. The first reason is that it made subjects aware that others' valuations for goods could differ from their own. The second reason was to provide an easier transition into the next phase of the experiment, where subjects would be placed in a situation that is different in three ways from typical market purchases. They buy products whose labels and packaging have been removed,
10
they taste products without knowledge of the information displayed on the label, and they buy products without knowing the sale price beforehand. The auction for the actual consumer product also serves to illustrate to subjects that they are spending real money for real products that they can keep after the experiment, and that they are not in a hypothetical simulation. To render this transparent, a bottle of wine is given to each winner, who is required to immediately pay the price determined in the auction from his current cash total.
3.4. Experiment 1: GMO phase After the training phase of each session described above, the GMO phase of the session, the phase of primary interest, was conducted. In the GMO phase, we simultaneously auctioned four products, which we referred to as S, L, C, and N during the sessions. All four products were biscuits that are typically available in grocery stores and supermarkets throughout France, and we informed subjects of that fact before bidding began. The products were different from each other, but were close substitutes. The GMO phase of the experiment consisted of five periods, as outlined in table 1. At the beginning of this phase, subjects received a sample of each of the four products to taste, without it’s packaging or labelling. Before bidding in the first period, subjects were required to taste each product. They then indicated how much they liked the product on a scale where “I like it very much” and “I don’t like it at all” were at the extremes of the rating scale (see Combris et al., 1997, or Noussair et al., 2004c). Then the auction for period 1 took place. The four products were auctioned simultaneously. Each of the following periods consisted of the revelation of some information about some or all of the products, followed by four simultaneous auctions, one for each product. The sale price was not drawn for any period until the end of period 5, and no information was given to participants about other players’ bids at any time.
[Table 1: About Here]
Table 1 shows the information made available to subjects at the beginning of each period. At the beginning of period 2, we informed the subjects that product S contained GMOs and that product N was GMO-free. No information was given about products L and C in period 2. At the beginning of period 3, we informed the subjects that no ingredient in L contained more than 1% GMOs and that no ingredient in C contained more than 1/10 of one percent GMOs. We also indicated to subjects that no ingredient in N had any detectable trace of GM content, and that S did contain a GM ingredient, soy, that was authorised in France. At the beginning of period 4, 11
subjects received a four-page handout containing background information about GMOs. The information consisted of a) the definition of a GMO, b) the criteria for classifying a product as containing GMOs, c) the list of GM plants authorised in France, d) the food products sold in France that contain GMOs, and e) the current French law regarding GMOs. We took care to provide an unbiased characterisation and provided only facts without comment. Before the last period, we revealed the brands of the four products and the label indicating that product N was organic.
3.5. Experiment 2: GMO phase In experiment 2, the GMO phase, the phase of interest, consisted of three periods, in which we revealed information about the products and then conducted an auction for the products. Four chocolate bars were auctioned each period, including two identical bars, called S and U. The products are made by a world leader in the food industry and are widely available in grocery stores and supermarkets in Europe. At the beginning of period 1 of the GMO phase in experiment 2, subjects each received a sample of each of the four products to taste, without its packaging or labeling. Then a simultaneous Vickrey auction for each of the four goods took place. At the beginning of period 2, we distributed one unit of each of the products to each subject in its original packaging (with the price removed, but with the list of ingredients visible). Subjects then had three minutes to study the products. A second auction was then conducted for each of the goods. At the beginning of period 3, we magnified and projected the list of ingredients of each product, exactly as it appeared on the packaging, and invited subjects to read the list of ingredients. Subjects then bid in the final round of auctions.
4. Results 4.1. Experiment 1: The impact of GMO information Figure 1 graphs the evolution of the average normalised bid over all subjects over the five periods of the GMO phase for the 4 products. The data in the figure are normalised by taking each individual’s actual bid in period 1 as the base equal to 100, tracking that individual’s bids over time relative to his bid in period 1, and averaging across all individuals in each period. Only the data from those who bid greater than 0 for the product in period 1 are included in the figure (no subject who bid zero for a product in period one ever submitted a positive bid for that product in later periods).
12
[Figure 1: About Here]
We observe that consumers, on average, value the absence of GMOs. In period 2, we revealed that product N did not contain GMOs and product S did contain GMOs. The GMO-free guarantee raised the limit price for product N of the average consumer in our sample by 8%. 41 of the 83 subjects, who bid more than zero for product N in period one, raised their bid in period 2, and only 7 lowered it. A sign test (eliminating the ties in which bids were the same in both periods) rejects the hypothesis that a bidder is equally likely to lower as to raise his bid at the p < .001 level. A pooled variance t-test also rejects the hypothesis that the mean bid for product N is equal in periods 1 and 2 at p < .01, indicating that, on average, consumers increased his bid for product N in period 2. In contrast, revealing that product S contains GMOs lowered its average limit price by 39%. Only 4 participants increased their bid for S after learning that it contained GMOs while 64 lowered their bid. Both a sign test and a pooled variance t-test reject the hypothesis of equality at the p < .001 level. The relatively small increase for the GMO-free product suggests that in the absence of any information about GM content, consumers typically act as if there is a low probability that products contain GMOs. The average premium for the GMO-free product over the product containing GMOs was 46.7%. Period 3 is designed to measure the impact of GMOs content thresholds. Our subjects appear to view a guarantee that no ingredient contains more than 0.1% GMOs as consistent with the typical GMO content of conventional products (the unlabeled product historically available). They value a 0.1% guarantee more highly than a 1% guarantee, and the 1% threshold appears to be seen as a higher level of GMO content than that of a conventional product. Furthermore, the 1% guarantee is viewed differently than the label “contains GMOs” and the 0.1% guarantee is interpreted differently than “GMO-free”. In period 3, we revealed that no ingredient in product L contained more than 1% GMOs and no ingredient in product C contained more that 0.1% GMOs. We observed no significant change in the median willingness to pay for product C between periods 2 and 3 (p = .38 for the sign test), but the average bid for product L declined by 10%, and the decline was statistically significant (p < .05 for the sign test). A pooled variance t-test of the hypothesis that the mean normalised bids for products L and C are equal rejects the hypothesis at a significance level of p < .01. There was no consensus among the participants about whether a product meeting the 0.1% threshold was valued more or less highly than a conventional product. 33% increased their bid (by an average of 28%) after learning the maximum possible GMO content was 0.1% of any ingredient, while 27.9% reduced their bid. 4.4% reduced their bid to
13
zero. The bidding behaviour for product L reveals that a product meeting a 1% threshold is viewed much differently than a product labelled as containing GMOs. 17.9% of subjects increased their bid when informed of the 1% threshold, and 40.5% left their bid unchanged. Thus over half of our participants considered a product satisfying the 1% threshold as no worse than the conventional product. The 1% guarantee was viewed as different from the 0.1% guarantee. The mean normalised bids in period 3 for products N and C, as well as for products L and S, were significantly different from each other at p < .01. The distribution of background information about biotechnology in period 4 led to a slight increase in average limit prices, which was significant at p < .05 for three of the four products. The increase was greatest for the GMO-free product N. The information did not bring the prices of L, with a 1% threshold, or S, which contained GMOs, to their levels before any information was revealed. For all four products at least 57% of the bids were unchanged between periods 3 and 4. For product N, the GMO-free product, 20 bidders increased their bid while 9 lowered it, and we can reject the hypothesis that an individual was equally likely to raise and to lower his bid at the p < .05 level. However, we cannot reject the analogous hypotheses for the other three products. Thus, for each of the products, though the pooled variance t-test indicates that the information increased the average bid, the more conservative sign test is not significant, and the majority of participants did not change their bids. Revealing the brand names of the products in period 5 raised the average prices for three of the four products. The effect was significant at p < .01 for L and S. The average bid for product C was significantly lower in period 5 than in period 4 at p < .01. However, for all four products, we fail to reject the hypothesis that an equal number of bidders raised and lowered their bids in period 5 relative to period 4. There was no increase in price for product N from revealing that it was organically produced, perhaps because revealing its label exerted an offsetting negative effect. Our consumers can be classified into four categories. Unwilling consumers bid zero for product S after learning that it contained GMOs. They comprised 34.9% of our subjects. Specifying a threshold did result in a lower incidence of zero bidding than the announcement “contains GMOs”. 10.7% of the subjects bid zero for the product with a maximum of 1% percent GMO content in any ingredient, and only 4.4% bid zero at 0.1%. That means that over 95% of our participants were willing to accept a level of GMO content that typically results from inadvertent co-mingling if the product is sufficiently inexpensive. 18.1% of our consumers did not change their bid for product S upon finding out that it contained GMOs. We classify them as Indifferent consumers. Another 4.9% of participants were
14
Favourable, demonstrating behaviour consistent with having a preference for GM foods. Thus a full 23% of bidders were willing to accept GMOs in their food at the same price as the conventional product. Despite the current unpopularity of GMOs in food, there is still a large group of consumers willing to buy them at the same price as conventional products and to allow them to establish a foothold in the marketplace. 42.2% lowered their bid for product S when they found out that it contained GMOs, but did not go so far as to bid zero. The average percentage of the decrease was 28.3%. We call this group the Reluctant consumers. This group places negative value on GMO content or a claim for an equitable share of the surplus created by GMOs adoption. They will lower (raise) their bid prices when faced with products with higher (lower) GMO content. They are willing to trade off GMO content and the price they pay. Indeed, 36.1% of the Reluctant consumers exhibited a willingness to pay that was monotonic in the strength of the guarantee of the maximum GMO content.
4.2. Results from the second experiment: Do individuals read labels? Figure 2 shows the normalized average bid over all subjects for each of the three periods of the GMO phase of experiment 2. Before bidding in period 2, subjects observe the products as they are seen in the supermarket. Presumably, we have created more favorable conditions for the subjects to read and study the labels than exist in the supermarket. Subjects are seated and have no alternative activities for 3 minutes other than to study the labels. Nevertheless, we observe that average bids do not change between periods 1 and 2. A pooled variance t-test fails to reject the hypothesis that the normalized average bids are different between periods 1 and 2 (t = .071 for product S, which does not contain GMOs and t = .070 for product U, which does contain GM corn). We also cannot reject the hypothesis that the bids are different from each other (t = 1.53). We thus obtain the result that the labeling of products as containing GMO’s does not affect the willingness to pay of consumers. However, the data change radically in period 3, in which subjects bid while able to view the list of ingredients on large overheads. The average willingness to pay for the product labeled as “containing GMO’s” decreases by 27.3% compared to the previous period. The decrease is statistically significant (a pooled variance t-test for a difference in sample means between periods 2 and 3 for product U yields t = 2.40). In contrast, an identical product without any indication of GMO content (product S) experiences an insignificant average decrease of 3% from the previous period (t = .271). The bids for the two products, S and U, are significantly different from each other in period 3 (t = 10.37). Upon learning that product U contains GM corn, 22% of our 15
subjects boycott the product entirely by bidding zero, and 60% lower their bid by at least 5%. Thus the labeling “contains GMO’s”, when it is actually noticed, induces a substantial decrease in willingness to pay that is specific to that product.
[Figure 2: About Here]
4.3. Comparison of the BDM and Vickrey processes The data from the training phase of the two experiments provides an opportunity to compare the BDM and Vickrey processes with regard to their demand revelation properties. We use two measures for our comparison. The first measure is the overall average bias of the mechanisms in period t, normalized by the valuation. It indicates the extent to which average bids are higher or lower than valuations. The bias for period t is calculated as Σj[bjt – vjt]/vjtnt, where bjt denotes player j’s bid in period t, vjt is her valuation in period t, and nt is the total number of bidders in period t. The second measure is the average dispersion, defined for period t as Σj|[bjt – vjt]|/vjtnt. The average dispersion is equal to the average absolute value of the difference between bids and valuations, normalized by the valuation. For an individual bid, the dispersion is the absolute value of the bias. Table 2 illustrates the average value of each measure over the course of the first four periods, as well as the last period of a session (which never exceeded period 6), under both the Vickrey and the BDM processes. The standard deviations are indicated in parentheses.
[Table 2: About Here]
The table reveals the following patterns. Period 1 was a practice period that did not count toward participants’ earnings. The table shows that both auctions are biased in period 1, with bids tending to be below valuations. This bias is larger and the dispersion is greater under the BDM mechanism. Overall, 90% of subjects bid less than their valuations and only a very small percentage bid more than their valuations. 2.4% of participants bid more than their valuations under the BDM process and 6% did so under the Vickrey auction. The percentage bidding an amount equal to their valuations is also small in both auctions, between 6 and 7 percent of subjects under both systems, though 17% bid within 2% of their valuations in the Vickrey auction. On average, under the BDM mechanism, bids are 39.9% lower than valuations, with a standard deviation (of the percentage difference between bid and valuation) of 28.9%. In the 16
Vickrey auction, the period 1 average bid is 30.2% less than the corresponding valuation with a standard deviation is 32.5%. Pooled variance t-tests indicate that the bias is significantly different from zero at the p < .01 level for both mechanisms. The proportion of participants bidding less than their values is greater in the BDM than in the Vickrey auction. The magnitude of the average underbid is less severe in the Vickrey auction. The average underbid is 44.6% of the valuation for the BDM and 36% for the Vickrey auction. The average absolute difference between bids and valuations, our measure of dispersion, is 41.7% in the BDM compared to 32.6% in the Vickrey auction. Both the average bias and the average dispersion are significantly greater than in the BDM than the Vickrey auction at the p < .05 level. Thus, in the practice period, the Vickrey auction is less biased, exhibits less dispersion, and has a greater percentage of agents bidding within 2% of values than the BDM process. In period 2, the first auction that counted toward subjects’ earnings, both auctions remain biased, but less so than in period 1. The introduction of monetary payments, as well as repetition, appears to improve decisions. Nonetheless, 87.8% of bids in the BDM and 76.1% of those in the Vickrey auction are less than valuations. The bias is –28.1% of valuation for the BDM and – 11.5% for the Vickrey auction. The decline in the bias between periods 1 and 2 is steeper in the Vickrey auction than in the BDM. The bias in the BDM decreases by 29.6%, whereas in the Vickrey auction the decrease is 63.3%. The decline is mainly due to a reduction in the amount that agents underbid, and not to a decrease in the percentage of agents underbidding. The percentage bidding equal to valuations increases to over 10% overall and is slightly higher in the Vickrey auction than in the BDM. The overall dispersion shrinks in both systems but the decrease is steeper in the Vickrey auction (51.5% versus 31.4%). Thus, the overall data from periods 1 and 2 suggest that the Vickrey auction is less biased, exhibits lower dispersion, induces a greater percentage to reveal their exact valuations, and improves its performance more quickly over time. These trends continue in subsequent periods. The overall bias decreases in each subsequent period for both processes, reaching zero in the Vickrey auction and 6% in the BDM mechanism in the final period. In each period, the bias in the BDM is significantly greater in magnitude than in the Vickrey auction at p < .05 (according to a pooled variance t-test). Beginning in period 4, the bias is no longer different from zero at conventional significance levels in the Vickrey auction. However in all of the periods, the bias is significant at the 5% level in the BDM. The percentage of agents bidding an amount equal to their valuations increases from period to period under both processes, reaching 41.5% for the BDM and 68.4% for the Vickrey auction in the last period. In the Vickrey auction, 77% of bids are within two percent of valuations and 90% are within ten percent of valuations in the last period. The dispersion between
17
bids and valuations decreases in each period of the BDM. Though the measure does increase between periods three and four in the Vickrey auction, the overall trend is clearly downward. The average absolute difference in the last period is 3.9% in the last period of the Vickrey auction compared to 11.8% in the BDM. In the Vickrey auction, the dispersion is significantly less than in the BDM at p < .01 in all periods except for period 4. Therefore, over the entire time horizon, the Vickrey auction generated data much closer to truthful bidding than did the BDM.
5. Discussion This chapter surveyed three studies. The first two consider empirical questions related to the willingness to pay for food products with genetically modified content, and the third compares two different techniques to elicit willingness to pay. Our results for the first experiment show a sharp contrast to the predominantly negative views of French survey respondents toward genetically modified organisms in food products. In our experiments, we observe a wide range of revealed preferences. Whereas 35% of our subjects absolutely refused to purchase a product containing GMOs, the remaining 65% of our subjects were willing to purchase a GM product if it was sufficiently inexpensive. Nearly one-quarter of participants showed no decrease in their willingness to pay in response to learning that a product contained GMOs. The two different thresholds, 0.1% and 1%, generated significantly different bids and were thus were clearly perceived as meaningfully different. Furthermore, the 0.1% threshold was not considered equivalent to GMO-free, and the 1% threshold generated higher bids than the classification “contains GMOs”. This indicates that market demand is decreasing in GMO content. 89% of our participants were willing to purchase a product satisfying the 1% threshold, the maximum content that the European Union exempts from labelling. Lowering the threshold to 0.1% would make another 7% of participants willing to purchase products satisfying the threshold, as 96% of our participants were willing to purchase a product, in which no ingredient contained more than 0.1% GMOs, if it were sufficiently inexpensive. The policy options available to address the arrival of biotechnology in food production can be grouped into three types. The first option is to ban the use of GMOs in food products. The second is to integrate conventional and biotech varieties into one production chain. The third is to create two production tracks and introduce a labelling system (which could be voluntary or mandatory) to allow the consumer to identify the two varieties. Based merely on polls, we would have concluded that the only policy action that would be feasible in France given current public opinion would be the complete interdiction of GMOs in food, at least for the time being. However, our experimental results indicate that only slightly more than a third of the population 18
would be unwilling to purchase GM foods at any price. The remainder is willing to purchase GMOs even when no threshold is specified, and could receive a welfare gain if GMOs make products cheaper. The data thus argue against the banning of GMOs, which would cause gains from trade to be foregone. The data also reveal potential welfare costs to consumers from integrating the two production streams. The consumers who are willing to purchase GMOs if they are sold at a discount might be made better off. However, the segment that refuses to purchase GM-products at any price (35% of participants in our sample) would experience a decrease in their welfare, and would have to switch to products with ingredients that have no GM varieties. Therefore, in our opinion, our results weigh in favour of segmenting the market between products containing GMOs and products that are GMO-free. In this way, the Unwilling consumers could be assured of GMO-free varieties, while price sensitive Reluctant, as well as Indifferent and Favourable, consumers could benefit from the cost reductions that the first generation of GMOs provides. As long as the segregation costs are not greater than the welfare gains from market segmentation, the sizes of each of the markets appear to justify the establishment of two separate production tracks. The separation and labelling policy gives the market the role of transmitting information about the safety of GM products, by providing an opportunity and an incentive for consumers to sample the lower cost products made with GMOs voluntarily. Our data suggest that a large fraction of consumers would do so. Our comparison of the valuation elicitation systems indicates that, given the training methods and the procedures we have used in our study, the Vickrey auction is preferable to the BDM mechanism as an instrument for the elicitation of the willingness-to-pay for private goods. We observe that the BDM is subject to more severe bias, greater dispersion of bids, and slower convergence to truthful revelation than the Vickrey auction. With our techniques, neither auction could be made into a perfect tool to reveal valuations with our subjects, at least not during the time horizons that were available to us. However, the Vickrey auction performs better than the BDM by the three criteria we have set for it. Our experimental protocol was effective in debiasing the Vickrey auction over several periods, but less effective on the BDM. Of course, it remains unknown whether unbiased bidding for goods with induced values carries over to subsequent bidding for goods with homegrown values. Our research supports the proposition that the Vickrey auction can be an effective tool for demand revelation with non-student subject pools, but also cautions that sufficient practice and appropriate training in the rules of the auction is important.
19
References Aizen, I., T. Brown and C. Rosenthal (1996) “Information Bias in Contingent Valuation: Effects of Personal Relevance, Quality of Information and Motivational Orientation”, Journal of Environmental Economics and Management, 30, pp. 43-57. Akerlof G. (1970) “The Market for Lemons: Qualitative Uncertainty and the Market Mechanism”, Quarterly Journal of Economics, 89, pp. 488-500. Aldrich, L., and N. Blisard (1998) “Consumer Acceptance of Biotechnology: Lessons from the rbST Experience”, Current Issues in Economics of Food Markets, December, pp. 1- 6. Anderson, K., C. Neilson and S. Robinson (2000) “Estimating the Economic Effects of GMOs: the Importance of Policy Choices and Preferences”, CIES Policy Discussion paper 35, Adelaide University, Adelaide, Australia. Becker G., M. DeGroot, and J. Marschak (1964) “Measuring Utility by a Single-Response Sequential Method”, Behavioural Science, 9, pp. 226-232. Blamey, R., M. Common and J. Quiggin (1995) “Respondents to Contingent Valuation Surveys: Consumers or Citizens?” Australian Journal of Agricultural Economics, 39, pp. 263¯288. Bohm P., J. Lindén, and J. Sonnegård (1997) “Eliciting Reservation Prices: Becker-DegrootMarshak Mechanisms vs. Markets” The Economic Journal, 107, pp. 1079-1089. Brookshire, D. and D. Coursey (1987), “Measuring the Value of a Public Good: An Empirical Comparison of Elicitation Procedures”, American Economic Review, 77, pp. 554-566. Buckwell, A., G. Brookes and D. Bradley (1999), “Economics of Identity Preservation for Genetically Modified Crops”, Food Biotechnology Communications Initiative Report. Busby J., J. Fox, R. Ready, and S. Crutchfield (1998) “Measuring Consumer Benefits of Food Safety Risk Reduction”, Journal of Agricultural and Applied Economics, 30(1), pp. 69-82. Caswell J. (2000) “Labelling Policy for GMOs: to Each his Own?” AgBioForum, 3(1), pp. 305309. Caswell J. (1998) “Should use of Genetically Modified Organisms be Labelled?” AgBioForum, 1(1), pp. 22-24 Combris P., S. Lecocq, and M. Visser (1997) “Estimation of a Hedonic Price Equation for Bordeaux Wine: Does Quality Matter?” The Economic Journal, 107, pp. 390-402. Coppinger, V. M., V. L. Smith, and J. A. Titus (1980), “Incentives and Behavior in English, Dutch and Sealed-Bid Auctions,” Economic Inquiry, 18:1, pp. 1-22. Cox, J., and D. Grether (1996), “The Preference Reversal Phenomenon: Response Mode, Markets, and Incentives”, Economic Theory 7:3, pp. 381-405. Cox J., B. Roberson and V. Smith (1982), “Theory and Behavior of Single Object Auctions’, In Research in Experimental Economics, (ed. V.L. Smith), Vol. 2, Greenwich, CT : JAI Press. 20
Cummings, R., G. Harrison, and E. Rutstrom (1995) “Homegrown Values and Hypothetical Surveys: Is the Dichotomous Choice Approach Incentive Compatible”, American Economic Review 85, pp. 260-266. Eurobarometer, (2003), Europeans and Biotechnology in 2002, Eurobarometer 58.0, General Directorate for Research, European Union, Brussels. Falk-Zepeda, J., G. Traxler and R. Nelson, R. (2000) “Surplus distribution from the introduction of a biotechnology innovation,” American Journal of Agricultural Economics, 82, pp. 360-369. Fox J., J. Shogren, D. Hayes, and J. Kliebenstein (1998) “CVM-X: Calibrating Contingent Values with Experimental Auction Markets”, American Journal of Agricultural Economics, 80, pp. 455465. Grether D., and C. Plott (1979), “Economic Theory of Choice and the Preference Reversal Phenomenon”, American Economic Review, vol. 69, pp. 623-38. Hayes D., J. Shogren, S. Shin, and J. Kliebenstein (1995) “Valuing Food Safety in Experimental Auction Markets“, American Journal of Agricultural Economics, 77, pp. 40-53. Hoffman, E., D. Menkhaus, D. Chakravarti, R. Field, and G. Whipple (1993) “Using Laboratory Experimental Auctions in Marketing Research: A Case Study of New Packaging for Fresh Beef”, Marketing Science, 12, pp. 318-338. Huffman, W., J. Shogren, M. Rousu, and A. Tegene (2001) “The Value to Consumers of GM Foods in a Market with Asymmetric Information: Evidence from Experimental Auctions”, mimeo, Iowa State University. Irwin J., G. McClelland, M. McKee, W. Schulze, and N. Norden (1998) “Payoff dominance vs. cognitive transparency in decision making” Economic Inquiry, 36, pp. 272-285. Kagel, J., R. Harstad, and D. Levin (1987) “Information Impact and Allocation Rules in Auctions with Affiliated Private Values: A Laboratory Study”, Econometrica, 55, pp. 1275-1304. Kagel J. and D. Levin, (1993), “Independent Private Value Auctions: Bidder Behaviour in First-, Second- and Third-Price auctions with Varying Numbers of Bidders”, Economic Journal, 103, pp. 868-879. Kellar, L. R., U. Segal, and T. Wang (1993) , “The Becker-DeGroot-Marschak Mechanism and Generalized Utility Theories: Theoretical Predictions and Empirical Observations”, Theory and Decision, 36, pp. 83-97. Krutilla, J. (1967), “Conservation Reconsidered”, American Economic Review, 57, pp. 777-786. Lence, S., and D. Hayes (2001), “Response to Asymmetric Demand for Attributes: An Application to the Market for Genetically Modified Crops”, mimeo, Iowa State University. Lin, W., W. Chambers and J. Harwood (2000), “Biotechnology: US Grain Handlers Look Ahead”, Agricultural Outlook, Economic Research Service/USDA, pp. 29-34.
21
Lin, W., G. Price, J. Fernandez-Cornejo (2001) “Estimating Farm Level Effects of Adopting Herbicide-Tolerant Soybeans”, Oil Crops Situation and Outlook, Economic Research Service, USDA, pp. 25-34. List, J., and C. Gallet (2001) “What Experimental Protocol Influence Disparities Between Actual and Hypothetical Stated Values? Evidence From a Meta-Analysis”, Environmental and Resource Economics, 20, pp. 241-254. List, J. and J. Shogren (1998), “Calibration Between Actual and Hypothetical Bids in a Field Experiment”, Journal of Economic Behaviour and Organisation, 37, pp. 193-205. List, J. and D. Lucking-Reilly (2000), “Demand Reduction in Multi-Unit Auctions: Evidence from a Sportscard Field Experiment,” American Economic Review, 90:4, pp. 961-972. List, J., and J. Shogren (1999) “Price Signals and Bidding Behavior in Second-Price Auctions with Repeated Trials,” American Journal of Agricultural Economics, 81, pp. 942-949 (1999). Lusk, J.L., M.S. Daniel, D.R. Mark, and C.L. Lusk (2001) “Alternative Calibration and Auction Institutions for Predicting Consumer Willingness-to-Pay for Non-Genetically Modified Corn Chips,” Journal of Agriculture and Resource Economics, 26, pp. 40-57. Lusk, J., J. Mustafa, L. Kurlander, M. Roucan, and L. Taulman (2005), “A Meta Analysis of Genetically Modified Food Valuation Studies” Journal of Agricultural and Resource Economics, 30, 28-44. Moon, W., and S. Balasubrimanian (2001) “A Multi-Attribute Model of Public Acceptance of Genetically Modified Organisms”, mimeo, Southern Illinois University. Neill, H., R. Cummings, P. Ganderton, G, Harrison, and T. McGuckin (1994) “Hypothetical Surveys and Real Economic Commitments”, Land Economics, 70, pp. 145-154. Noussair, C., Robin, S. and Ruffieux, B. (2001) ‘Genetically modified organisms in the food supply: public opinion vs. consumer behaviour’, working paper, Purdue University. Noussair, C., S. Robin, and B. Ruffieux (2002), “Do Consumers Not Care About Biotech Foods or Do They Just Not Read the Labels?” with Stephane Robin and Bernard Ruffieux, Economics Letters, Vol. 75, no 1, March 2002, pages 47-53. Noussair, C., S. Robin, and B. Ruffieux (2004a), “Do European Consumers Really Refuse to Buy GM Food?” Economic Journal, Vol. 114, No. 492, January 2004, pages 102-120. Noussair, C., S. Robin, and B. Ruffieux (2004b), Revealing Consumers’ Willingness to Pay: A Comparison of the BDM Mechanism and the Vickrey Auction”, Journal of Economic Psychology, December 2004, Vol. 25, No .6, pages 725-741. Noussair, C., S. Robin, and B. Ruffieux (2004c), “A Comparison of Hedonic Rating and Demand Revealing Auctions”, with Stephane Robin and Bernard Ruffieux, Food Quality and Preference, Vol. 15, No. 4, June 2004, pages 393-402. Nyborg, K. (2000) “Homo Economicus and Homo Politicus: Interpretation and Aggregation of Environmental Values,” Journal Of Economic Behaviour And Organisation, 42, pp. 305-322 22
Rutstrom, E. (1998) “Homegrown Values and Incentive Compatible Auction Design”, International Journal of Game Theory, 27, pp. 427-441. Sagoff, M. (1988) The Economy of the Earth. Cambridge University Press, Cambridge. Shogren, J., J. Fox, D. Hayes, and J. Roosen (1999) “Observed Choices for Food Safety in Retail, Survey, and Auction Markets”, American Journal of Agricultural Economics, 81, pp. 192-1199. Smith, V. (1982) “Microeconomic Systems as an Experimental Science”, American Economic Review, 72, p. 923-955. Stevens, T., J. Echeverria, R. Glass, T. Hager, and T. More (1991) “Measuring the Existence Value of Wildlife: What Do CVM Estimates Really Show?” Land Economics 67, pp. 390¯400. Traxler, G., J. Falck-Zepeda and R. Nelson (2000) “Rent Creation and Distribution From Biotechnology Innovations: The Case of Bt Cotton and Herbicide-Tolerant Soybeans in 1997" Agribusiness Vickrey, W. (1961) "Counterspeculation, Auctions, and Competitive Sealed Tenders", Journal of Finance, 16, 8-37.
23
Table 1: Sequence of events in GMO phase of an experimental session, experiment 1 Period 1
Period 2
Period 3
Period 4 Period 5
Transactions
- Information: Blind tasting of the four products S, L, C and N - Recording of hedonic rating of the four products - Auction - Additional Information: “S contains GMOs” and “N is GMO free” - Auction - Additional Information: “No ingredient in L contains more than 1% GMOs”, “No ingredient in C contains more than 1/10 of 1% GMOs”, “One ingredient in S (soy) is derived from an authorised genetically modified product,” and “No ingredient in N contains any detectable trace of GMOs”. - Auction - Additional Information: General information about GMOs - Auction - Additional Information: the brand names of the four products, and the designation “organically grown” for product N. - Auction - Random draw of the auction that counts toward final allocations. - Implementation of transactions for the period that counts
24
Table 2 – Deviations of Bids from Valuation, Induced Value Phase of Both Experiments Period 1
Period 2
Period 3
Period 4
Last Period
Average bias
-39.87%
-28.06%
-12.76%
-8.19%
-6.33%
BBDM =Σj[bjt – vjt]/vjtnt
(28.89%)
(22.32%)
(23.38%)
(23.60%)
(20.95%)
Average dispersion
41.65%
28.59%
16.86%
13.94%
11.75%
DBDM =Σj|[bjt – vjt]|/vjtnt
(26.24%)
(21.62%)
(20.58%)
(20.69%)
(18.43%)
Average bias
-30.16%
-11.50%
-5.57%
+1.33%
-0.06%
BV =Σj[bjt – vjt]/vjtnt
(32.53%)
(27.76%)
(11.94%)
(26.49%)
(11.13%)
Average dispersion
32.57%
16.79%
6.25%
9.27%
3.89%
DV =Σj|[bjt – vjt]|/vjtnt
(30.10%)
(24.89%)
(11.59%)
(24.96%)
(10.42%)
BDM
Vickrey
25
Figure 1: Average Bids for the Four Biscuits in Each Period of GMO Phase, Experiment 1 120 107,7
110 100
100
99,2
110,9 100,9
114.3
114,9
104.9
102 94,6
90
89,9
Threshold of 0.1% (Product C)
80
GMO Free (Product N)
70 61,1
60
60,5
66,2 63.6
O s M G
3, Th 4, re sh Ba ol ck ds gr ou nd In fo rm at io n Pe ri o d 5, Br an ds
Pe rio d
O or G M
d Pe rio 2, W ith
Pe rio d
Fr ee
1, Bl in d
50
Pe rio d
Contains GMOs (Product S) Threshold of 1% (Product L)
90.1
26
Figure 2: Average Bids for the Two Identical Chocolate Bars in Period 1-3, Experiment 2 (Corn corresponds to Product S, GM Corn to Product U) 110.0 100.0
100.0 95.9
97.0
Product S, GM free
72.7
Product U, Containing GM Corn
97.0 90.0 80.0 70.0 60.0 Period 1: Taste Period 2: Product Observe Package and Labeling
27
Period 3: Ingredients Displayed
