toward these new types of food technologies, as well as how consumer ... insecticides (Feldmann , Moris, and Hoisington, 2000; CSU Transgenic ... and benefit perceptions of GM food on information-seeking or purchasing behavior ..... income levels and a higher percentage of people that have attained a Bachelor's Degree.
A RISK PERCEPTION ANALYSIS OF GENETICALLY MODIFIED FOODS BASED ON STATED PREFERENCES
Marcia J. Bugbee and Maria L. Loureiro Department of Agricultural and Resource Economics Colorado State University Fort Collins, CO-80523
Paper Prepared for presentation at the American Agricultural Economics Association Annual Meeting, Montreal, Canada, July 27-30, 2003
Copyright 2003 by Marcia Bugbee and Maria Loureiro. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies.
A RISK PERCEPTION ANALYSIS OF GENETICALLY MODIFIED FOODS BASED ON STATED PREFERENCES
Abstract Most of the existing literature deals with consumer willingness to pay for GM-free food, or consumer willingness to accept for GM food. However, it is well know that consumers have mixed views about GM foods. In this research, we do not presume that all consumers may just have positive or negative preferences about GM products. Rather, heterogeneous preferences are considered. This paper presents a contingent valuation questioning sequence and associated modeling approaches that allow identification of both positive and negative preferences. Willingness to pay (WTP) for the GM product is contrasted with willingness to accept (WTA) compensation to buy it.
1
Introduction The study of consumer response toward genetically modified (GM) crops and foods is becoming increasingly important. While there is considerable economic literature on general attitudes toward GM products, international differences in policies regarding GM processes and labeling, and willingness to pay for foods labeled as “non-GM”, there are still many unanswered questions surrounding this topic. To our knowledge, there is a still a need to identify the role played by different risk factors in the consumer response toward these new types of food technologies, as well as how consumer response may change if the GM product offers a direct personal benefit to the purchaser. The impact of perceived risks and benefits associated with new food technologies has important economic and food safety implications. Therefore, better understanding of consumer attitudes and behavior toward genetically modified (GM) food products is essential for designing new market strategies for the second-generation of GM products which, unlike the first-generation of producer and environment-friendly GM, offer benefits to consumers. There are several known potential risks associated with GM crops and foods. These include, but are not necessarily limited to: health issues such as allergenicity, increased toxins, antibiotic resistance, and unknown consequences to humans that may exist, and environmental issues such as effects on non-target organisms, crop to crop cross pollination, crop to weed pollination, and development of pest resistance to insecticides (Feldmann , Moris, and Hoisington, 2000; CSU Transgenic Crops website). However, potential risks do not always translate into perceived risks by consumers. On the other hand, benefits associated with GM manipulation are also various,
2
depending on the final use or goal of the product. Some of the most common benefits include the increase in nutritional qualities, the reduction of pesticides applications, and increase in shelf life of the product. Though there are many studies about the economics of risk perceptions, and there is a growing body of literature about consumer risks and benefit perceptions associated with new food products, few economic studies have looked at the impact of those risk and benefit perceptions of GM food on information-seeking or purchasing behavior. The current paper will add to the economic literature about GM products in the context of heterogeneous preferences about risks and benefits associated with this new type of food technology. The valuation methods will account for the individuals that will pay for the benefits of GM food and those that will not be willing to pay anything due to perceived risks associated with the technology. The analysis includes two products, a genetically modified tomato and a genetically modified beef product. Heterogeneous preferences were considered, thus the analysis includes both WTP and WTA regressions and, repectively, WTP and WTA mean value calculations.. Based on the results, interesting comparisons can be made between the two product categories analyzed, crops and meats.
Literature Review There are many studies dealing with willingness to pay (WTP) for differentiated products, although only a handful regarding willingness to pay for GM or GM-free food products. Existing studies have elicited WTP for GM-free products by using either contingent valuation (CV) methods or experimental auction methods. While CV methods
3
will be used for the purpose of this study, it is interesting and useful to look at the experimental methods and results as well. Using experimental economics methods of calibration and first- and second-price auctions, Lusk et al. (2001) determined consumer willingness to pay for non-GM corn chips. Results from the calibration, using scale-differential questions, indicated a high level of acceptance of GM products. However, results from the double-hurdle model bids indicated that 70% of participants were unwilling to pay for non-GM chips. Additionally, 20% of consumers bid $0.25/oz. or more for the opportunity to exchange their GM chips for non-GM chips. Another study, by Huffman et al. (2001), presented willingness to pay for foods with and without GM labels using laboratory auction experiments for three food items. The random nth-price auction format was used in these experiments, eliciting premiums of about 14 percent for foods that respondents perceived as non-GM. This premium was similar across the three different products, making it possible to conclude that consumer demand for GM foods is significantly lower than the demand for the non-GM counterpart. None of the sociodemographic characteristics appeared to alter consumer WTP for GM foods. Contingent valuation (CV) methods were used by Loureiro and Hine (2002), with the objective of determining consumer willingness to pay for a labeled value-added potato that could be marketed as organic, GMO-free, or Colorado grown. Perhaps the WTP research that may be most closely related to the objectives of the present paper is that of Moon and Balasubramanian (2001). The survey instrument used in their study included CV questions to assess consumer willingness to pay a premium for non-GM
4
breakfast cereals. They found that in both the UK and the US, risk and benefit perceptions clearly translated into purchasing intentions and behaviors as measured with willingness to pay. Equally important was the conclusion that risk perception exerts a greater impact on WTP than benefit perception. It should be noted, however, that consumer reception of GM foods may be significantly different if the food products can offer direct, personal benefits. Such products, which are referred to as neutraceutical products or functional foods, are the second generation of GM products and are included in the analysis of the present paper. The present study attempts to analyze consumer trade-offs between potential benefits and potential risks associated with the GM technology, analyzing the role played by subjective beliefs and risk perceptions on consumer acceptance of GM products. Willingness to pay (WTP) for the product is contrasted with willingness to accept (WTA) compensation to buy the product. Unlike most contingent valuation studies where WTA represents the amount of money that has to be offered to a respondent in order to forego the consumption of a specific good and remain at the same utility level (i.e., equivalent variation measure), in this application WTA corresponds to the minimum amount of money that has to be offered to the respondent to accept a less preferred situation associated with the program (i.e., compensating variation measure). In modeling heterogeneous preferences, we draw from existing environmental economics literature about lovers and haters (winners and losers). This prior research shows that ignoring negative preferences leads to considerable overestimation of willingness to pay (Huhtala, 2000; Clinch and Murphy, 2001). These researchers have demonstrated the importance of allowing for lovers and haters in contingent valuation,
5
asserting that the externalities of a program (or good) may be viewed costs by some people and benefits by others. Kristrom (1997) discusses the use of spike models for CV studies, allowing also for zero WTP (in addition to positive and negative WTP). The next sections of this paper present the empirical methods, the data, and concluding results.
Theoretical Background The consumer’s decision process is modeled using a random utility approach. Consumer utility, U ( y, x, m) , is assumed to have three arguments: genetic manipulations of food products that offer consumer benefits, y, consumer socio-demographic characteristics and personal beliefs that may affect choice, x, and the income level, m. The variable y is an indicator variable, which equals one if the product has been genetically modified, and zero otherwise. The consumer is willing to pay c dollars to switch to a GM product, which will make utility at least as great as it would be if the product was not genetically modified. Mathematically, c is represented as
(1)
U (0, x0 , m ) ≤ U (1, x1 , m − c ),
where the 0 and 1 subscripts denote the choice of non-GM and GM food products, respectivelyi. The consumer’s utility function is unknown since some components are unobservable and thus, can be considered random variables from the researcher’s standpoint. Therefore, utility is decomposed into an unobservable part and an error term,
εj
. Mathematically,
U ( y, x j , m) = V (y, x j , m ) + ε j
. The random error term
εj
is
assumed to be independently and identically distributed with a mean of zero. The
6
consumer’s decision to pay c dollars in terms of utility can be represented as:
(2)
V (0, x0 , m ) + ε ≥ V (1, x1 , m − c ) + ε 1 ,
which can be expressed in a probability framework as:
(3)
P(WTP ≥ c ) = P(V0 + ε 0 ≤ V1 + ε 1 ) = P(ε 0 − ε 1 ≤ V1 − V0 ).
This theoretical model sets the groundwork for the specific empirical models that follow. In the current study, a binary choice model approach is chosen to analyze the decision of paying a premium (WTP) or accepting compensation (WTA) for two genetically modified food products, a tomato and a beef product.
Methodology Based on the survey format, it is plausible to estimate the willingness to pay and willingness to accept measures using a sample selection framework. As stated previously, the survey used a pattering format very similar to the double bounded model in which respondents face an initial preference question about GM products (at no premium). Based on their initial preferences, a follow-up question asked consumers willingness to pay or willingness to accept. That is, only consumers who give a positive answer to the first question receive the WTP follow-up question. The opposite occurs for the WTA question, since only those that answered “no” to the initial preference question were asked in the follow-up question their willingness to pay. Using a general
7
framework, we can state that the sample equation that determines the selection process is the first initial question asked in the survey, z * i = γ ′wi + u i . Let the equations of primary interest be:
(4)
WTPi = β ′xi + ε i , if z i > 0, and
(5)
WTAi = β ′xi + ε i , if z i 0 = β ′xi + ρσ ε λi (α u ),
where α u = −γ ′wi / σ u and λ (α u ) = ϕ (γ ′wi / σ u ) / Φ (γ ´wi / σ u ) .
Notice that the same applies for the WTA (after recoding the response of the selection question), such that:
(7)
[
]
E WTAi | zi* < 0 = β ′xi + ρσ ε λi (α u ),
These selection equations will be estimated using LIMDEP.
8
Data The data was gathered using a mail survey in the Western States of the United States. A sample of 1000 participant households was drawn from a mailing listing purchased from Survey Sampling, Inc., a leader in the science of sampling methodology and research quality. This listing is compiled from the white page directories, and supplemented with a variety of other sources such as Department of Motor Vehicles (DMV) information, voter information, census data, and school records. Thus, we expect that this listing is representative of the current U.S. Census. Upon receiving the listing of 1,000 households, scripted calls were placed to each household. The purpose of the call was to determine if someone would be willing to participate in the study by completing a mail survey. Based on the respondents’ willingness to cooperate, the survey was sent to a total of 680 households. Before the survey was sent off, a pretest was conducted in two different locations in the states of Colorado and California. After making slight modifications using the information gathered in the pre-test, the final survey was sent out in a six-page, booklet format, with a cover letter explaining the project, and a pre-paid return envelope. The survey included four different sections. In the first section, different warm-up questions related to general knowledge and information about risks and benefits associated with genetically modified foods were presented to the respondents. The level of consumer concern with social/ethical, health, and environmental issues surrounding genetic modifications was obtained in section two. The third section contained the elicitation of willingness to pay for different genetically modified processes in both animals and crops, and with and without the association of any risk for humans or for the environment.
9
Finally, the last section contained questions related to socio-demographics variables. The third section of the survey is of particular importance to the purpose of this paper because it allows for the comparison of WTP and WTA measurements for both crop (tomato) and animal (beef) products. In both cases, the initial question presents survey respondents with a direct consumer benefit of higher vitamin content for the tomato and lower fat content for the beef product. These potential benefits that are associated with the GM food are expected to elicit a positive WTP value. The survey questions used to analyze the net WTP for the GM products were the following:
Assume that there is a new GM tomato with higher nutritional value and both GM and non-GM tomatoes of comparable appearance are available at the same price, would you be willing to buy the GM tomatoes? ___1. Yes ___2. No If you responded NO, would you be willing to buy those same GM tomatoes at a price 10% lower per pound ($2.06) than that of the non-GM tomatoes (regular price=$2.29/pound)? ___1. Yes ___2. No If you responded YES, would you be willing to spend 10% more per pound ($2.52) if the GM tomatoes have higher vitamin content than the non-GM tomatoes (regular price=$2.29/pound)? ___1. Yes ___2. No Assume that there is a new type of GM beef with higher nutritional content and less calories, if both GM and non-GM beef of comparable appearance were available for the same price, would you be willing to buy the GM beef? ___1. Yes ___2. No If you responded NO, would you be willing to buy that same GM beef at a price 10% lower per pound ($4.31) than that of the non-GM beef (regular price=$4.79/pound)? ___1. Yes ___2. No If you responded YES, would you be willing to spend 10% more per pound ($5.27) if the GM beef had lower fat content than the non-GM beef (regular price=$4.79/pound)? ___1. Yes ___2. No
10
A total of 164 responses were received The respondents' average age was between 45-49 years, with a mean education that included junior college, 64.6% were married, and 49% of all respondents had children under the age of 18 years old living in their household. The average household size was about 2.5 members, and average household income was between $50,000 and $59,000 for the 2001 fiscal year. Summary statistics of the socio-demographics are presented in Table 1. When comparing our socio-demographic figures with the U.S. Census (U.S. Census Bureau), as in Table 2, we see that our sample is considerably older (approximately 15 years older), with higher income levels and a higher percentage of people that have attained a Bachelor’s Degree or higher. The sample also has a lower percentage of females and a lower percentage of households with children under 18 years of age. It is clear that this sample population had considerably different socio-demographic characteristics than the broader U.S. population, however, it is difficult to assess the effects that may be associated with these differences in our results. While a representative sample is always of concern to a researcher, it is likely that we encountered some degree of sample selection bias in which respondents who were more interested in the topic of GM crops and food products elected to participate in the survey. In the current study, participation was estimated to be about 24% of the total solicited population. Research conducted by Edwards and Anderson (1987) found significant differences between the characteristics of survey respondents and nonrespondents. Finally, Messonnier et al. (2000) examined sample nonresponse and selection biases, finding out that unit nonresponses seriously affected welfare measures.
11
In our study, since we do not have any information regarding the nonrespondents, we cannot assess the impact of sample selection biases on our WTP and WTA estimates. Given the preceding observations, we acknowledge that our findings are limited in their ability to be applied to a fully generalized broader population.
Model Specification and Variable Definitions The first stage, probit models are specified as follows:
(8) WTPGMTomato / GMBeef =
β 0 + β 1 MidAge + β 2Older + β 3Child + β 4 Female + β 5 Middle + β 6Upper + β 7 Manipulati on + β 8GMRisky
The results from this first stage are held and then the selection models depicted in equation (4) and equation (5) are estimated for each of the four sub-groups: GM tomato WTP, GM tomato WTA, GM beef WTP, and GM beef WTA. The final specifications of the equations are as follows:
(9) WTP / WTA GMTomato =
β 0 + β 1 Bid + β 2 SomeInf + β 3 PoorInf + β 4 Manipulati on + β 5 GMRisky + β 6 Lambda and (10) WTP / WTAGM Beef =
β 0 + β 1 Bid + β 2 SomeInf + β 3 PoorInf + β 4 Manipulati on + β 5 GMRisky + β 6 Lambda
where Bid is the random percentage premium (or discount) a consumer was faced with
12
(ranging from 5-50%); MidAge is a dummy variable that represents ages in the interval of 35-54 years; Older is an indicator variable representing those over 55 years of age; Child is a dummy variable that represents the presence of children in the household; Female is an indicator variable that represents a female respondent; Middle is a cross product of a dummy variable that represents a household income within the range of $30,000 to $69,999 per year and a dummy variable that represents an education level of “some college” or “junior college graduate”; Upper is a cross product representing those with a household income greater than $70,000 per year and the minimum education level of a 4year university graduate; SomeInfo is an indicator variable representing those who consider themselves “somewhat informed” about GM food; and PoorInfo is a dummy variable representing those who consider themselves either “poorly informed” or “not at all informed” about GM food; Manipulation is a scale variable ranging from 1 to 10, with 1 representing the preference of preserving natural species at all costs, and 10 representing manipulating natural species in order to get a benefit at all costs; GMRisky is an indicator variable representing those that believe GM food carries both food safety and environmental risks.
Results The probit model results presented in Table 4 are the first stage of our model. This first stage was based on the initial dichotomous choice question, which addressed whether or not respondents would be willing to purchase a GM product (tomato and beef) with higher nutritional content if it was of comparable appearance and price. The results from the GM tomato probit regression suggest that age, gender, social
13
status, views on manipulation of natural species, and risk perceptions associated with GM foods are the significant determinants of whether or not one would purchase a GM tomato. Positive effects are associated with being middle aged or older (35-55, >55) and with those that agree with manipulation of natural species to obtain a benefit. Alternatively, negative effects are associated with being female and from the middle class, as well as believing GM food carries both food safety and environmental risks. Also reported in Table 4 are the results from the GM beef probit regression, which indicates that the presence of children in the household, views of species manipulation, and risk perceptions are the major determinants of whether or not one would purchase a beneficial GM beef product. The results from this beef regression, as well as those obtained from the GM tomato regression, were then retained and factored into the selection models presented below.
GM Tomato—WTP Estimate About 63% of the survey respondents said that they would be willing to purchase a genetically modified tomato. The regression presented in Table 5 indicates the determining factors that impact the probability that a person from this group of consumers will pay a premium for the GM tomato. The regression results show that the significant variables impacting the probability that a consumer will pay a premium for the GM tomato include: Bid, the random percentage premium a consumer was faced with (ranging from 5-50%), and SomeInfo, an indicator variable representing those who consider themselves “somewhat informed” about GM food. These variables are both significant and negative, which means that a consumer presented with a higher premium
14
price is less likely to pay a premium for the product, and those that consider themselves “somewhat informed” are also less likely to pay a premium with respect to the omitted category. Looking at the marginal effects of these variables, as presented in Table 6, we can expand on the impact of these two variables on the probability of paying a premium. As the bid amount (Bid) increases by 1%, there is an approximate 0.7 decrease in the probability that a consumer will be willing to pay a premium for the GM tomato. The probability also decreases by about 0.3 if the consumer considers themselves “somewhat informed” about GM food. Based on these regression results the estimated mean WTP value for GM tomatoes was calculated. The results, as reported in Table 7, show that consumers from this group (which consisted of 63% of our solicited population) are willing to pay, on average, a premium of approximately 25% for these genetically modified tomatoes.
GM Tomato—WTA Estimate Approximately 36% of our respondents said initially that they would not buy the GM tomato. Thus, each of these consumers was asked if they would accept the tomato at a discount. The estimated mean willingness to accept value, reported in Table 7, is a discount of about 9% the original tomato price. The probability of consumers from this group accepting compensation for the GM tomato is impacted by two significant variables, Bid and PoorInfo. The positive sign on the Bid variable is an indication that the greater the amount of compensation offered, the more likely it is that consumers will accept the modified tomato. Alternatively, the
15
negative sign of the PoorInfo coefficient indicates that those respondents that consider themselves “poorly informed” or “not at all informed” about GM food are less likely to be willing to accept the GM tomato. In fact, based on the marginal effect measure, the probability that a consumer will accept the GM tomato decreases by about 0.2 if the consumer is poorly informed.
GM Beef—WTP Estimate Those consumers that were willing to buy the GM beef consisted of about 55% of the solicited population. The estimated mean WTP for GM beef is a premium of about 46%. The only variable that impacts the probability that a consumer in this group will pay a premium for the GM beef product is the bid amount. As expected, Bid has a negative and statistically significant coefficient. The marginal effect associated with the bid value indicates that a 1% increase in the premium value results in a 1% decrease in the probability that a consumer will pay the premium.
GM Beef—WTA Estimate Approximately 44% of the respondents said that they would not purchase the GM beef product. In estimating the probability that consumers from this group will purchase the product at a discounted rate, there was only one significant variable, Lambda. The significance of this variable justifies the use of the Heckman estimator. If we were to have omitted the initial question from the CV questioning sequence of the survey, we would have lost some valuable information about the WTP/WTA estimates.See Table 6 for the indirect effects.
16
Conclusions The results from this analysis include two products, a genetically modified tomato and a genetically modified beef product. Additionally, heterogeneous preferences were considered, thus the analysis includes WTP regressions and WTP mean value calculations, as well as WTA regressions and WTA mean value calculations. Results indicate that the main influencer of mean WTP or WTA measures is the bid amount a respondent is presented with. The higher the premium, the less likely it is that the consumer will pay it. Also, the probit models suggest that the across-the-board determinants of whether a person is a lover or hater of GM foods are their views on manipulation of natural species (Manipulation) and their perceptions of the risks associated with GM technology (GMRisky). Other socio-demographic variables were also found to be significant contributors. Interesting comparisons can be made between the two product categories analyzed, crops and meats. First, we note that it seems a higher percentage of respondents prefer tomatoes than GM beef. It is possible then, that the general population is more accepting of plant modifications than of GM animal products. However, the ones preferring GM beef also seem to be willing to pay higher premiums than the “lovers” of GM tomatoes. Those people that enjoy this beneficial GM beef product, which offers higher nutritional content and fewer calories, are willing to pay premiums of approximately 46%. Future research should continue to explore the differences in perceptions and WTP/WTA measures of GM crop and GM animal products.
17
References Baker, G.A. and Burnham, T.A. (2001). “Consumer Response to Genetically Modified Foods: Market Segment Analysis and Implications for Producers and Policy Makers.” Journal of Agricultural and Resource Economics, 26(2), 387-403. Boman, M., G. Bostedt, and B. Kriström (1999). “Obtaining Welfare Bounds in DiscreteResponse Valuation Studies: A Non-Parametric Approach.” Land Economics, 75(2):28494. Carson, R.T., L. Wilks and D. Imber (1994). “Valuing the Preservation of Australia’s Kakadu Conservation Zone.” Oxford Econom. Papers 46: 727-749. Clinch, P. and A. Murphy (April 2001). “Modeling Winners and Losers in Contingent Valuation of Public Goods: Appropriate Welfare Measures and Econometric Analysis.” The Economic Journal 111: 420-443. Transgenic Crops: An Introduction and Resource Guide, Colorado State University Website, Available at: http://www.colostate.edu/programs/lifesciences/TransgenicCrops/ Feldmann, M.P., Morris, M.L., and Hoisington, D. (2000). “Why All the Controversy?” Choices, 15(1), 8-12. Haab, T. and K. McConnell (1997). “Referendum Models and Negative Willingness to Pay: Alternative Solutions.” Journal of Environmental Economics and Management 32: 251-271. Hanemann, M (1989). “Welfare Evaluations in Contingent Valuation Experiments with Discrete Response Data: Reply.” American Journal of Agricultural Economics 71: 10571061. ---------------------and B. Kanninen. 1999. The Statistical Analysis of Discrete-Response CV Data. In: Valuing Environmental Preferences, 302-441. I. Bateman and K. Willis Eds.
Huang, J., Haab, T.C., and Whitehead, J.C. (2001). “Absolute Versus Relative Risk Perception: An Application to Seafood Safety.” Working paper: AEDE-WP-0007-01, Department of Agricultural, Environmental, and Development Economics, The Ohio State University. Huffman, W.E., Shogren, J.F., Rousu, M., and A. Tegene (2001). “The Value to Consumers of GM Food Labels in a Market with Asymmetric Information: Evidence From Experimental Auctions.” Paper presented at the annual American Agricultural
18
Economics Association, Chicago, August 5-8. Huhtala, A (2000). “Binary Choice Valuation Studies with Heterogeneous Preferences Regarding the Program Being Valued.” Environmental and Resource Economics, 16: 263-279. Hutchinson, G., R. Scarpa, S. Chilton, and T McCallion (January 2001). “Parametric and Non-Parametric Estimates of Willingness to Pay for Forest Recreation in Northern Ireland: A Discrete Choice Contingent Valuation Study with Follow-Ups. Journal of Agricultural Economics, 52(1):104-122. Kriström, B. (May 1990). “A Non-Parametric Approach to the Estimation of Welfare Measures in Discrete Response Valuation Studies.” Land Economics, 66: 135-139. Kriström, B. (1997). “Spike Models in Contingent Valuation.” American Journal of Agricultural Economics, 79: 1013-1023. Loureiro, Maria L. and Susan Hine, December 2002. “ Discovering Niche Markets: A Comparison of Consumer Willingness to Pay for Local (Colorado Grown), Organic and GMO-free Products,” Journal of Agricultural and Applied Economics 34(3): 477-487. Lusk, J.L., Daniel, M.S., Mark, D.R., and Lusk, C.L. (2001). “Alternative Calibration and Auction Institutions for Predicting Consumer Willingness to Pay for Nongenetically Modified Corn Chips.” Journal of Agricultural and Resource Economics, 26(1), 40-57. Moon, W. and Balasubramanian, S.K. (2001). “Public Perceptions and Willingness-toPay a Premium for Non-GM Foods in the US and UK.” AgBioForum, 4(3&4), 221-231. Available on the World Wide Web: http://www.agbioforum.org. Smith, K.V. and Johnson, F.R. (1998). “How Do Risk Perceptions Respond to Information? The Case of Radon.” The Review of Economics and Statistics, 70(1): 1-8. Turnbull, B. W., (1976). “The Empirical Distribution Function with Arbitrarily Grouped, Censored and Truncated Data.” J. Roy. Statist. Soc. Ser. B. 38: 290-295. U.S. Census Bureau, United States Census 2000. Available at: http://www.census.gov/
19
Table 1. Summary Statistics: Main Socio-Demographics Sample Characteristics Description Mean St. Dv. Cases Age
Gender Education
Income
Employment
Household Members Children Under18 at Home Marital Status
1=Under 20 2=20-24 3=25-29 4=30-34 5=35-39 6=40-44 7=45-49 8=50-54 9=55-59 10=60+ years 1=Female 0=Otherwise 1=Elementary school or less 2=Some high school 3=High school graduate 4=Some college 5= Junior college graduate 6=4-year university graduate 7=Post graduate work 8=Any other education 1=Under $20,000 2=$20,000-$29,999 3=$30,000-$39,999 4=$40,000-$49,999 5=$50,000-$59,999 6=$60,000-$69,999 7=$70,000+ Student (1.25%) Full-time (51.25%) Part-time (8.75%) Stay at home (4.38%) Retired (31.25%) Not Employed (3.12%) Continuous
2.509317
1.346839
161
Continuous
0.490683
0.981822
161
Married (64.6%) Single (11.18%) Separated/Divorced (8.7%) Domestic Partnership (6.21%) Widowed (9.32%)
7.975000
1.999843
160
0.416149
0.494457
161
5.354037
1.586702
161
5.137255
1.936692
153
160
161
20
Table 2. Comparison of Sample Socio-demographic Versus U.S. Population U.S. Populationa 50.9%
Socio-demographics
Sample
% Female
41.6%
% Household with children under 18 years of age
25.2%
36.0%
% Bachelor’s degree or higher
52.8%
24.4%b
Median income
5 ($50,000-$59,999)
$41,994
a b
Median age 8 (50-54) 35.3 Source: Consumer Survey and U.S. Census Bureau, Census 2000. Persons of 25 years and over, 2000.
21
Table 3. Consumer Concern with Social/Ethical, Health, and Environmental Issues Surrounding Genetic Modification
Social/Ethical Issue 1. Patenting life/Playing God 2. Accelerating growth of multinational corporations 3. May lead to human genetic engineering 4. Transferring genes between plants and animals 5. Increase income inequalities between rich and poor countries Health Issue 1. Allergies 2. Increased toxins 3. Lower nutrient content in the food 4. Unknown consequences to humans Environmental Issue 1. Effect on non-target organisms 2. Crop to crop cross pollination 3. Crop to weed pollination 4. Development of pest resistance to insecticides
1= 2= 3= 4= 5= Extremely Very Somewhat Not Very Not At All Concerned Concerned Concerned Concerned Concerned 18.47% 5.73% 22.93% 22.29% 25.48% 14.10% 17.95% 18.59% 25.64% 16.67%
6= Don't Know n= 5.10% 157 7.05% 156
19.48%
12.99%
23.38%
19.48%
17.53%
7.14% 154
20.00%
13.75%
26.25%
16.88%
13.75%
9.38% 160
12.90%
13.55%
20.65%
23.87%
20.00%
9.03% 155
27.85% 27.85% 23.72% 43.75%
19.62% 22.98% 16.67% 24.38%
30.38% 27.33% 27.56% 18.75%
11.39% 9.32% 15.38% 5.62%
5.06% 4.97% 9.62% 3.75%
5.70% 7.45% 7.05% 3.75%
N= 158 161 156 160
9.49% 8.92% 11.54% 6.92%
N= 158 157 156 159
27.85% 21.66% 19.87% 30.19%
17.09% 19.75% 21.79% 32.70%
30.38% 29.94% 30.13% 19.50%
8.86% 10.83% 11.54% 5.03%
6.33% 8.92% 5.13% 5.66%
22
Table 4. Consumers Willing to Purchase Genetically Modified Foods—Probit Results Variable
GM Tomato
GM Beef
Coefficient -0.767 CONSTANT 1.001** MIDAGE 0.789* OLDER 0.447 CHILD -0.531** FEMALE -0.613** MIDDLE 0.300 UPPER MANIPULATION 0.187*** -0.837*** GMRISKY LOG LIKELIHOOD -75.222 50.537 CHI-SQUARED 153 N=
P-Value Coefficient P-Value 0.111 -0.784 0.114 0.019 0.597 0.179 0.071 0.592 0.194 0.159 0.898** 0.006 0.026 -0.323 0.179 0.048 -0.400 0.198 0.311 0.011 0.970 0.000 0.179*** 0.000 0.001 -1.115*** 0.000 -77.871 52.837 152 ***,**, and* represent statically significant coefficients at α = 0.001, α = 0.05 , and α = 0.1 , respectively.
23
Table 5. WTP/WTA for GM Tomatoes and GM Beef Variable
GM Tomato WTP GM Tomato WTA
Coefficient 0.450** CONSTANT -0.706** BID -0.286** SOMEINFO -0.120 POORINFO 0.027 MANIPULATION -0.093 GMRISKY 0.133 LAMBDA 97 N=
GM Beef WTP
GM Beef WTA
P-Value Coefficient P-Value Coefficient P-Value Coefficient 0.028 0.055 0.688 0.556** 0.039 -0.125 0.008 0.455* 0.077 -1.000*** 0.001 0.069 0.017 -0.016 0.882 -0.089 0.533 0.085 0.276 -0.200* 0.056 0.085 0.496 -0.153 0.241 0.012 0.572 0.024 0.405 0.004 0.440 -0.053 0.546 0.139 0.410 0.062 0.460 0.023 0.859 -0.129 0.572 0.299* 55 84 66
P-Value 0.442 0.766 0.366 0.122 0.859 0.595 0.083
***,**, and* represent statically significant coefficients at α = 0.001, α = 0.05 , and α = 0.1 , respectively.
24
Table 6. Marginal Effects of Variables
Variable
Direct effects in the regression
Indirect effects in LAMBDA
Total effect for variables in both parts
GM Tomato WTP
GM Tomato WTA
GM Beef WTP
GM Beef WTA
CONSTANT BID SOMEINF POORINF MANIPULATION GMRISKY
Effect 0.450 -0.706 -0.286 -0.120 0.027 -0.093
P-value 0.028 0.008 0.017 0.276 0.241 0.440
Effect 0.055 0.455 -0.016 -0.200 0.012 -0.053
P-value 0.688 0.077 0.882 0.056 0.572 0.546
Effect 0.556 -1.000 -0.089 0.085 0.024 0.139
P-value 0.039 0.001 0.533 0.496 0.405 0.410
Effect -0.125 0.069 0.085 -0.153 0.004 0.062
P-value 0.442 0.766 0.366 0.122 0.859 0.595
CONSTANT MIDAGE OLDER CHILD GENDER MIDDLE UPPER MANIPULATION GMRISKY
0.065 -0.085 -0.067 -0.038 0.045 0.052 -0.025 -0.016 0.071
0.832 0.754 0.810 0.851 0.767 0.792 0.893 0.622 0.660
-1.14E-02 1.49E-02 1.17E-02 6.65E-03 -7.90E-03 -9.12E-03 4.46E-03 2.78E-03 -1.25E-02
9.70E-01 9.56E-01 9.66E-01 9.74E-01 9.59E-01 9.63E-01 9.81E-01 9.31E-01 9.38E-01
-6.46E-02 4.92E-02 4.88E-02 7.40E-02 -2.66E-02 -3.30E-02 8.70E-04 1.48E-02 -9.19E-02
8.38E-01 8.62E-01 8.66E-01 7.20E-01 8.62E-01 8.68E-01 9.96E-01 6.41E-01 5.73E-01
-1.49E-01 1.14E-01 1.13E-01 1.71E-01 -6.14E-02 -7.62E-02 2.01E-03 3.41E-02 -2.12E-01
6.37E-01 6.88E-01 6.97E-01 4.08E-01 6.88E-01 7.00E-01 9.91E-01 2.81E-01 1.93E-01
MANIPULATION 0.011 GMRISKY -0.022
0.778 0.912
1.48E-02 7.00E-01 3.83E-02 3.67E-01 3.83E-02 3.31E-01 -6.53E-02 7.22E-01 4.66E-02 8.42E-01 -1.51E-01 4.52E-01
25
Table 6. Mean WTP Calculations MEAN WTP MEAN WTA (% PREMIUM) (% PREMIUM) GM TOMATO
24.84
9.11
GM BEEF
46.38
9.53
i Note that the same framework is valid to analyze WTA, although it requires to make a change in all the signs.
26
