A Case Study of Consumer’s Risk Attitudes to the Use of Plant Biotechnology in Food, Industrial and Medical Applications

A Case Study of Consumer’s Risk Attitudes to the Use of Plant Biotechnology in Food, Industrial and

Medical Applications

Michele Veeman, Yulian Ding, Yu Li, Wiktor Adamowicz

Department of Resource Economics and Environmental Sociology, University of Alberta,

Edmonton Alberta T6G 1X1, CanadaContact: [email protected]

Why is PMF of interest?

• The use of plants as food, medicines, and industrial products dates from prehistory

• Modern biotechnological methods to use plants as production platforms for vaccines & pharmaceutical drugs, specialized industrial products & functional foods date from 1992

• Plant molecular farming (PMF) promises both potentially important benefits and potential risks and costs……..public acceptance is necessary!

Some factual examples:

• Production of plant-based pharmaceutical drugs (eg, insulin expressed in safflower plants or animal and human vaccines produced in tobacco plants) continues under development

• Initiatives to improve nutritional content of particular foods (eg, Golden Rice)--continue to face considerable regulatory and commercialization challenges.

• Industrial products such as biofuel and other biomaterials obtained from modified plants.

• Contamination by plants modified to express medical or industrial compounds may lead to accidental contamination of food and feed crops,

• Associated potential issues of food and environmental safety

• Very considerable financial costs from accidental contamination of non-GM crops

Potential costs include:

Rationale for study of individuals’ risk perceptions

Better understanding of public/consumers’ attitudes to these bio-economy innovations can aid development of effective policy, risk management, and risk communication

Conceptually, the study of risk perceptions:

• recognizes the influence of family and society on individual’s risk preferences and beliefs

• recognizes that risk attitudes may change with new information and experience (Branson et al. 1996, Viscusi 1989)

• trust is also increasingly viewed as an important influence on people’s views and behavior relative to risky situations and actions (eg., Uslaner 2002)

• early psychometric studies established the importance for risk perceptions of whether a risk is undertaken voluntarily, involves lack of choice, and /or is poorly understood or invokes dread, amongst other features (eg., Slovic 1987, 2000, 2010)

Empirical model: We apply ordered probit models based on individual’s risk assessments

where: y*mn is an unobserved continuous dependent latent variable (the extent of concern attributed by the nth individual to the mth risk situation), Xn specifies the socio-economic and demographic characteristics of individual n, andβ is the parameter vector.

Following Greene (2003), for estimation:

• y * mn is replaced by observed categorical values of respondent’s risk ratings (ymn).

• Parameters µ are to be estimated and specify thresholds between category rankings (0 ˂ μ₁˂ μ₂) . Considering four risk rating levels, which we give respective values of 0, 1, 2, and 3, gives:

• , , , .

Again, following Greene (2003):

Assuming that ϵmn are normally distributed, the probabilities of ymn = 0,1,2,3 are calculated, as are the marginal effects.

Marginal effects indicate the probabilities of change from one risk rating level to another, based on the significant estimated parameters associated with respondents’ characteristics

DATA: Two large scale cross-Canada consumer surveys, conducted in 2009 and 2005

Each survey focused on consumers’ attitudes and stated choices re agricultural biotechnology innovations.

Each was developed & tested with randomly recruited focus groups.

Each survey (in English and French) was administered to sampled respondents drawn from large internet-based consumer panels. For the 2009 survey sample N= 1009; the 2005 survey sample N= 1575.

By design, features of each sample are reasonably consistent with major demographic characteristics of the Canadian population; this may not apply for unobserved population characteristics.

Each survey queried risk ratings for four biotechnology innovations & seven other food risk issues

Innovations/risk issues were random ordered:“Use of genetic modification/engineering in crop production”

“Drugs (i.e. medicines) made from plant molecular farming through genetic modification/engineering”

“Genetically modified/engineered crops to produce industrial products like plastics, fuel or industrial enzymes”

“Genetically modified/engineered crops to increase nutritional qualities of food.”

Respondents were asked to rated each of these as: “High Risk”, “Moderate Risk”, “Slight Risk”, “Almost No Risk” or “Don’t Know/Unsure”

Explanatory variables: socioeconomic and demographic information

Much similar data are available for each sample. But data on different measures of trust attitudes & family health status differ2005: no family health information; trust in information from different sources is simply queried (Y/N) 2009: family health is queried two concepts of trust are queried--- generalized trust (GSS) “Most people can be trusted” ---institutional trust (following De Jong 2008) we develop a measure of system trust for each respondent, based on ratings for different dimensions of trust in food producers, processors, retailers and government)

Results and Discussion, Qualitative Issues:Table1. Summary statistics of risk ratings for four plant biotechnology applications, 2009 survey (Number of

respondents=1009)

High risk

Moderate risk

Slight risk

Almost no risk

Don't know

Use of genetic modification/engineering in food production 225 265 257 141 121 Drugs (i.e. medicines) made from plant molecular farming through genetic modification/engineering 184 234 254 160 177 Genetically modified/engineered crops to increase nutritional qualities of food 194 237 263 187 128

Genetically modified/engineered crops to produce industrial products like plastics, fuel or industrial enzymes 260 238 222 153 136

T

Table 2. Summary statistics of risk rating rank order, by percentage of respondents citing issue as “high risk” and associated percentages of respondents choosing this response, 2009 and 2005

  2009 order & (%)

2005order & (%)

Bacteria contamination of food 1 (45%)9 (18%)

Pesticide residuals in foods 2 (44%) 4 (29%)

Fat and cholesterol content of food 3 (38%) 3 (31%)

BSE (mad cow disease) 4 (35%) 7 (25%)

Use of hormones in food production 5 (31%)1 (33 %)

Use of antibiotics in food production 6 (28%) 2 (31%)

Genetically modified/engineered crops to produce industrial products like plastics, fuel or industrial enzymes 7 (26%)

 10 (15%)

Use of genetic modification/engineering in food production 8 (22%)

  5 (29% )

Genetically modified/engineered crops to increase nutritional qualities of food 9 (19%)

 8 (19%)

Use of food additives 10 (18%) 6 (25%)Drugs (i.e. medicines) made from plant molecular farming through genetic modification/engineering 11 (18%)

 11 (15%)

Results and Discussion, Quantitative Models:

The following two tables give estimated coefficients for two different versions of the ordered probit models for each of the four biotechnology innovations: generalized trust (table 3) and food system trust (table 4) for 2009. Estimated coefficients based on the 2005 data are in the paper. Marginal effects based on significant coefficients from these estimates are in Tables 5 and 7

Table 3. Coefficients and standard errors of ordered probit models for four plant biotechnology applications based on

generalized trust, 2009 survey data  Food

productionMedicines Nutritionally

enhanced foods

Industrial Products

Constant 1.0352*** 1.2473*** 1.0203*** 1.2083***  (0.1764) (0.1820) (0.1784) (0.1813)Male 0.1790** 0.1387* 0.2260*** 0.2304***  (0.0751) (0.0777) (0.0754) (0.0764)QC -0.1571* -0.1154 -0.1388 -0.2380***  (0.0867) (0.0903) (0.0876) (0.0890)University -0.0500 -0.0072 -0.0073 -0.0773  (0.0835) (0.0865) (0.0839) (0.0856)Income 0.0832* 0.1184*** 0.0429 0.1184***  (0.0365) (0.0375) (0.0367) (0.0372)Age -0.0077*** -0.0083*** -0.0067** -0.0120***  (0.0026) (0.0027) (0.0027) (0.0028)Urban 0.0304 -0.1186 0.0461 -0.0120  (0.0892) (0.0908) (0.0894) (0.0894)Kid 0.0186 -0.0809 0.0028 -0.1521*  (0.0848) (0.0872) (0.0851) (0.0864)Family health -0.1790* -0.1217 -0.01405* -0.1443*  (0.0738) (0.0763) (0.0741) (0.0750)Generalized Trust 0.1938** 0.2429*** 0.1962*** 0.1568**  (0.0755) (0.0778) (0.0760) (0.0767)Sample size 888 832 881 873Log likelihood -1184.806 -1119.434 -1194.411 -1163.377µ1 0.8208*** 0.7949*** 0.7660*** 0.7398***  (0.0376) (0.0383) (0.0366) 0.0368µ2 1.7156*** 1.6851*** 1.6084 1.5287***  (0.0516) (0.0515) (0.0484) (0.0506)

Table 4. Coefficients and standard errors of ordered probit models for four plant biotechnology applications based on trust in the food system, 2009 survey data

  Food production Medicines Nutritionally enhanced foods

Industrial Products

Constant 0.4066* 0.83315*** 0.4040* 0.6872***  (0.2123) (0.2176) (0.2125) (0.2171)Male 0.1808** 0.13675** 0.2294*** 0.2321***  (0.0752) (0.0776) 0.0755) 0.0765QC -0.1592* -0.1248 -0.1328 -0.2399***  (0.0862) (0.0901) (0.0870) (0.0884)University 0.0208 0.0543 0.0649 0.1274***  (0.0838) (0.0867) (0.0841) (0.0857)Income 0.0902* 0.12475*** 0.0494 -0.0110***  (0.0365) (0.0374) (0.0367) 0.0028)Age -0.0062** -0.00675** -0.0050* -0110***  (0.0026) (0.0027) (0.0027) (0.0028)Urban 0.0334 -0.1175 0.0442 -0.0224  (0.0895) (0.0909) (0.0895) (0.896)Kid 0.0161 -0.0878 -.0067 -0.1613*  (0.0849) (0.0872) (0.0852) (0.0866)Family health -0.1759** -0.1303* -0.1450* -0.1459*  (0.0740) (0.0763) (0.0742) (0.0752)Most trusting 0.7475*** 0.4969*** 0.7585*** 0.5787***  (0.1371) (0.1404) (0.1364) (0.1411)Second trusting 0.6745*** 0.4843*** 0.6595*** 0.6052***

  (0.1240) (0.1258) (0.1227) (0.1269)Sample size 888 832 881 873Log likelihood -1171.284 -1116.514 -1180.785 -1153.763µ1 0.8360*** 0.8011*** 0.7820*** 0.7490***  (0.0381) (0.0385) (0.0372) (0.0371)µ2 1.7396*** 1.6910*** 1.6359*** 1.5465***  (0.0520) 0.0514 (0.0489) (0.0510)

Table 5. Marginal effects of significant coefficients (!%, 5%, 10%) for “high risk” and “almost no risk” ratings, 2009 data

  Food production

Medicine Nutritionally enhanced foods

Industrial Products

High risk        Male -0.0559 -0.0401 -0.0650 -0.0779QC 0.0507 ns ns 0.0836University ns ns ns nsIncome -0.0263 -0.0346 ns -0.0409Age 0.0024 0.0024 0.0019 0.0041Urban ns ns Ns nsKid ns ns ns 0.0528Family heath 0.0560 ns 0.0407 0.0491Generalized trust -0.0604 -0.0697 -0.0565 -0.0532Most trusting -0.2024 -0.1301 -0.1854 -.0.1790Second trusting -0.2216 -0.1478 -0.2023 -0.2134Almost No risk        Male 0.0426 0.0374 0.0654 0.0581QC -0.0356 ns ns -0.0561University ns ns ns nsIncome 0.0195 0.0317 ns 0.0294Age -0.0018 -0.0022 -0.0019 -0.0030Urban ns ns ns nsKid ns ns ns -0.0367Family health -0.0425 ns -0.0404 -0.0362Generalized trust 0.0462 0.0661 0.0567 0.0394Most trusting 0.2038 0.1462 0.2421 0.1611Second trusting 0.1418 0.1212 0.1713 0.1360

Table 7. Marginal effects of significant coefficients (1% and 5% levels) for “high risk’ and “almost no risk” ratings, 2005 data

  Crop production

Medicine Nutritionally enhanced foods

Industrial Products

High risk        Male -0.095 -0.056 -0.056 -0.068QC 0.107 ns ns nsUniversity ns ns ns nsIncome -0.008   -0.007 -0.007Age ns ns ns nsUrban ns ns ns nsKid 0.018 ns ns nsTrustUniversity ns ns ns -0.038                 -Almost No risk        Male 0.059 0.060 0.052 0.085QC -0.058 ns ns nsUniversity ns ns ns nsIncome 0.005 ns ns 0.008Age ns ns ns nsUrban ns ns ns nsKid -0.011 ns ns -0.0367TrustUniversity ns ns ns 0.047

Summing Up

• Quebec residents are more averse to GM/GE in crop/food production than are other Canadians

• Women see more risk than do men in these GM/GE applications. Older people perceive more risk than others. This is consistent with results commonly found in numbers of studies of food risk perceptions

• Our most striking results are the influence of trust (ie those who trust & those who trust the food system) in mitigating high risk perceptions.

Implications: need to maintain trust & need for gender awareness in risk communication!

Extrapolating from our previous work and other studies regarding food bio-fortification:

Although many individuals place value on nutritive or environmental benefits, this value component is typically less than the discount in WTP to accept identified GM/GE-based foods. Stigmatization and regulatory lags and costs hinder approval of food bio-fortification through GM/GE (eg Golden rice) &, with targeting by activists, these have heightened barriers to research and commercialization of such products. Thus, where genetic diversity allows, efforts to produce foods with nutritionally improved components are much more readily accepted if these are not directly based on GM/GE techniques.

As regulatory barriers have grown, what other effects have these influences tended to have on bio-economy innovation?

Dramatic reductions in costs to identify, sequence and analyze genes enables plant scientists and breeders to use of new molecular biological tools to identify molecular “markers” of desired plant traits such as drought resistance and some—but not all---desired nutritional components.

Thus “traditional” breeding techniques, allied with molecular biology techniques may be pursued for many (not all) crop innovations.

This approach involves: # Capital costs and genetic diversity in target plants, and# Considerable commitment to both basic and applied research,

But these are under major financial pressures in public and university research centers. The implications of these pressures surely deserve further assessment by economists and other policy analysts.