Consumer Valuation of and Attitudes towards Novel Foods Produced with NPETs: A Review

We review the emerging international body of evidence on attitudes and willingness to 10 pay (WTP) for novel foods produced with New Plant Engineering Techniques (NPETs). NPETs in11 clude genome/gene editing, cisgenesis, intragenesis, RNA interference and others. These novel 12 foods are often beneficial for the environment and human health and more sustainable under in13 creasingly prevalent climate extremes. These techniques can also improve animal welfare and dis14 ease resistance when applied to animals. Despite these promising attributes, evidence suggests that 15 many, but not all consumers, discount these novel foods relative to conventional ones. Our system16 atic review sorts out findings to identify conditioning factors which can increase the acceptance of 17 and WTP for these novel foods in a significant segment of consumers. International patterns of ac18 ceptance are identified. We also analyze how information and knowledge interact with consumer 19 acceptance of these novel foods and technologies. Heterogeneity of consumers across cultures and 20 borders, and in attitudes towards science and innovation emerges as key determinants of acceptance 21 and WTP. Acceptance and WTP tend to increase when beneficial attributes—as opposed to pro22 ducer-oriented cost-saving attributes—are generated by NPETs. NPETs improved foods are sys23 tematically less discounted than transgenic foods. Most of the valuation elicitations are based on 24 hypothetical experiments and surveys and await validation through revealed preferences in actual 25 purchases in food retailing environments. 26


Introduction
We review the emerging and fast-growing international body of empirical evidence 31 on consumers' attitudes and limited willingness to pay for/consume novel foods pro- 32 duced with inputs generated using New Plant Engineering Techniques, or NPETs. 33 NPETs include genome/gene editing, cisgenesis, intragenesis, non-transgenic RNA inter-34 ference, and others. 1 These novel foods often feature traits introduced via NPETs to ben-35 efit the environment and human health and to increase sustainability in the face of cli- 36 mate extremes. Water savings, reduced pesticide applications, reduced food waste, re- 37 sistance to pests and diseases, and more nutritious food are among the benefits created 38 using NPETs. When applied to animals, these techniques can also improve animal wel- 39 fare and disease resistance. Improving disease resistance in plants and animals may mit- 40 igate antimicrobial resistance [2], which can arise with the (over)use of antimicrobials. 41 Despite the benefits that NPETs confer, public (e.g., governmental) and private (in- 42 dividual) opposition to these technologies may limit their development by disincentiviz- 43 ing researchers and firms from investing in them [3]. Particularly relevant in the context 44 of our review, existing studies suggest that consumers discount these novel foods rela- 45 tive to conventional foods on average. Our systematic review sorts out findings to iden- 46 tify conditioning factors that can influence and increase the acceptance of these novel 47 foods in a significant segment of consumers. We also examine international patterns of 48 acceptance. NPETs, like genetically modified organisms (GMOs) twenty years ago, offer 49 the potential to efficiently introduce desirable traits into organisms but also appear to 50 face issues of consumer distrust, leading to decreased valuation of the new technology 51 despite its potential to improve sustainable agricultural practices [4,5]. Issues related to 52 distrust-including labelling, scientific knowledge, risk perception, and perception of 53 naturalness-are present with NPETs, just as they were with GMOs. Our investigation 54 points out the key differences in perceptions and willingness to pay (WTP) for NPET- 55 based foods relative to GMO-based foods and conventional and/or organic substitutes. 56 We also identify conditioning determinants of WTP, namely, the tangible benefits con- 57 sumers are interested in and those they discount. 58 As private firms and associated supply chains are increasingly focused on improv- 59 ing their sustainability and social engagement with environment, sustainability and gov-60 ernance (so-called "ESG") criteria [6], it is critically important to understand consumer 61 behavior towards biotechnology and new foods relying on NPETs. These new foods 62 could be misperceived and rejected even though these new biotechnologies hold much 63 promise to improve sustainable food supply chains and foster better health outcomes for 64 consumers and the environment.
In most studies reviewed, the average consumer discounts these NPET-based novel 66 foods relative to conventional ones, although the discount is not as pronounced as for 67 transgenic (GMO) foods, when comparative results are available. However, consumers 68 are heterogeneous in their preferences and valuations, as documented by many studies. 69 Heterogeneity of consumers within and across cultures and borders, heterogeneity in 70 attitudes towards science and innovation and in risk perceptions-which are related to 71 1 We closely follow Sticklen [1] to define NPETs. Genome or gene editing (GE) refers to a technique that adds, deletes or modifies precisely and site-specifically genes from the genome of a plant or animal. GE "genetic scissor" methods include CRISPRcas9, TALEN, and zinc finger nuclease (ZFN). When introducing a gene belonging to the same or cross breedable species, the resulting crop is called "cisgenic" or sometime ingenic. Cisgenic introduction includes the gene cassette with its regulatory sequences integrated in the host plant. Intragenic inserts are close to cisgenic, but the gene coding sequence is regulated by promoters and terminators of different genes from the same cross-breedable gene pool. RNA interference (RNAi) is a technique used to regulate or silence the transcription of a specific native gene in the host plant. Here we restrict RNAi to non-transgenic modifications.
objective knowledge about biotechnology [7]-emerge as key determinants of ac- 72 ceptance and WTP. Acceptance and WTP are higher when consumers perceive the at- 73 tributes generated by NPETs as beneficial. Tangible benefits include improvements in 74 nutritional value or taste and more sustainable processes such as reduced pesticide or 75 water use. Superficial improvements are discounted. 76 Most of the valuation elicitations are based on hypothetical experiments and sur-77 veys in standard research setups (e.g., lab experiment, online survey), in part because 78 few NPET-based novel foods have been commercialized. These hypothetical valuations 79 await validation through revealed preferences in actual purchases in the food retailing 80 environments when these novel foods will become widely available. 81

82
The article relies on a systematic review of the emerging literature on NPETs, con-83 sumers' attitudes and willingness to pay for NPET-based food. We first undertook a sys-84 temic search for available articles written in the English language, published or not, using 85 Google Scholar searches with the following keywords: gene/genome editing, CRISPR, 86 Talen, cisgenic, intragenic, ingenic in addition to consumer acceptance (or attitudes), or 87 consumer willingness to pay (purchase, eat, consume). This process yielded more than 50 88 references. Next, during the reading process we collected additional references to articles 89 that we did not identify in our initial literature searches. We complemented our list of 90 search-based candidate studies with these additional articles. Finally, the reading process 91 also revealed that some studies did not cover NPETs or had been erroneously cited as 92 addressing consumer behavior towards NPETs. The final process yielded 53 useable stud-93 ies, two of which were review articles (not generating any new data) and 51 of which were 94 based on original data collected for their respective investigation. Several studies on con-95 sumer behavior and NPETs yielded more than a single article. Additionally, a number of 96 investigations were international in nature and yielded WTP estimates for multiple pop-97 ulations.
98 99 We tabulated the 53 studies in searchable spreadsheet format to catalogue the follow-100 ing characteristics: the name of the authors, year of appearance; the full reference; the topic 101 (attitude/acceptance, WTP, framing effects, etc.); the commodity(ies) or food items; what 102 was estimated (WTP, attitude or acceptance); comparative study of more than one tech-103 nology; traits covered by the innovations; methodology/approach (choice experiment, 104 auction, survey, statistical methods, qualitative, etc.); the sample size; estimated val-105 ues/key results; technologies covered (GMO, GE gene/genome editing, other 106 NPETs/NBTs (cisgenic, intragenic, ingenic), conventional/hybrids, and organic); coun-107 try(ies); population sampled; and additional remarks. These key attributes are presented 108 in Appendix Table 1. Then, we used descriptive statistics (counts and frequencies) to char-109 acterize the key attributes of these studies. We then evaluated the estimated results and 110 findings in a more qualitative way to obtain stylized facts on discounts and premia in 111 WTP, and treatment effects influencing the acceptance of and attitudes toward NPETs. 112 While falling short of undertaking a full-blown meta-analysis, we go much beyond the 113 typical literature survey. The tabulated folder is posted online and searchable by NPET 114 type, country/region, and commodity. Appendix Table 1 presents the studies with key 115 attributes. 116  Among the 53 identified studies investigating consumer attitudes/behavior with re-125 spect to NPETs, 30 focused on genome/gene editing, while 23 examined other NPETs (16 126 cisgenic/ingenic; three intragenic; four RNAi) covering the period 2004-2021. The earliest 127 investigations predominantly focused on goods generated with cisgenic or intragenic 128 modifications relative to standard (transgenic) GMO substitutes [8][9][10][11]. The more recent 129 papers focus on GE, RNAi, and other newly developed NPETs. Among these 53 studies, 130 36 address consumer attitudes and acceptance and willingness to eat or consume; 29 stud-131 ies provide WTP or willingness to purchase information. These two sets of studies include 132 a number of comparative, multiple-country studies, and all WTP studies include some 133 version of variables that capture attitudinal information of participants in their surveys 134 The studies cover a wide range of countries, though coverage is predominantly fo-135 cused on two regions. European countries (24 studies) and North America (USA and Can-136 ada) (20 studies) have received the most attention, while the number of studies examining 137 consumer attitudes/valuation in Asia (5), Latin America and the Caribbean (4), and Africa 138 (2) are limited. Information about the specific country or region for which data were col-139 lected in the studies we survey is included in Appendix Table 1. Although the majority of 140 the investigations use experiments and questionnaires that involve participants making 141 choices, several of the studies are framed in terms of consumers' perceptions and attitudes 142 regarding NPETs, and associated perceived risks and benefits, without asking partici-143 pants to make explicit choices. Further, 39 investigations involve comparative analysis of 144 technologies-a combination of conventional, GMO, and/or organic versus NPETs. 145 Among these comparative studies, 31 cover conventional technologies/hybrids, 36 involve 146 GMO, and ten deal with organic goods. 147 Most investigations and experiments involve hypothetical or fictitious choices, since 148 very few NPET-based goods have been commercialized with the exceptions of soybean 149 and canola oil, and apples. Even those products that have been commercialized are not 150 widely available and, due to regulatory issues, have not been approved for produc-151 tion/commercialization in many countries or regions, such as the EU [3]. Two articles that 152 used real-rather than hypothetical-choices elicited data on WTP through an experi-153 mental auction with real food products [10,11]. However, even though real transactions 154 occurred, the goods in the auction were not actually produced using NPETs; rather, pur-155 chasers were given a conventional version of the product. Another set of studies attempts 156 to incorporate non-hypothetical data by combining store scanner data and NPET survey 157 data for the same subjects in an effort to condition the responses to the survey with scan-158 ner data (the revealed preferences of shoppers through their purchases of organic milk 159 and rye bread) [12,13]. 160 161 Many of the articles-29 out of 55-estimate valuation of NPETs. The three main 162 approaches used to elicit data for WTP estimation in these studies are choice experiments, 163 experimental auctions, and multiple price lists (MPLs). While each of these techniques is 164 designed to estimate valuation of products or product attributes, the approach used by 165 each method-as well as situations in which each method is most beneficial-differs. In 166 choice experiments, respondents view choice sets that contain a few product alternatives 167 (typically two) along with an option to indicate they would not purchase either option, 168 yielding binary data on choices, which are associated with variations in prices and attrib-169 utes. Choice experiment investigations of WTP rely on a Random Utility Model (RUM) 170 and some form of binary (logit or probit) regression model with various degrees of so-171 phistication to address latent variables and estimate preference heterogeneity, or deal 172 with other statistical challenges like zero willingness-to-pay for boycott/protest consum-173 ers and data censoring. Some investigations directly estimate WTP whereas other derive 174 it by using ratios of estimated marginal utility of attributes divided by the negative of the 175 price response. Both are standard ways to derive WTP. Choice experiments are well-176 suited for situations in which the researchers wish to evaluate multiple attributes of the 177 products. 178 Experimental auction approaches directly elicit WTP measures by having partici-179 pants bid directly on food products with varying attributes. These WTP measures can 180 then be used in simple statistical tests (such as t-tests to evaluate whether, say, WTP elic-181 ited under two conditions significantly differs) or in linear regression models, depending 182 on the design of the research. Experimental auction studies are typically used when there 183 is a single focal attribute (or condition) that researchers wish to estimate WTP for. Auc-184 tions also require real purchases due to greater threat of hypothetical biases [14]. In the 185 context of NPETs, these studies evaluate differences in WTP between conventional and 186 modified product variants. As noted previously, the lack of commercialized NPET-based 187 products limits the use of methods that rely on non-hypothetical choices; few studies on 188 consumer valuation of NPETs have used experimental auctions [10,11]. 189 MPL-based studies present respondents a list of prices for two products (at a time). 190 One of the products' prices incrementally changes in each row of the list. The respondent 191 makes a choice between each product in each row. The approach captures when the re-192 spondent switches from one product to the other or to none. These studies frequently use 193 interval regression to analyze the data derived from MPL studies [15,16]. 194 The novelty of NPETs means that, unless researchers trade out NPET-based products 195 for conventional products at the end of the experiment (after presenting choices as real) 196 [10], most studies are by necessity hypothetical. While there are widespread concerns 197 about biased valuation estimates resulting from hypothetical decisions, hypothetical 198 choices-and consequences of hypothetical studies, such as hypothetical bias-have been 199 widely studied [14]. Researchers have developed methods to reduce overestimates of val-200 uation stemming from the hypothetical nature of these choices, including the use of cheap 201 talk scripts-which remind participants to think about budget constraints or other de-202 mands on their money, certainty follow-ups that ask how sure they are about their deci-203 sion, and honesty priming tasks, as well as valuation calibration techniques, among others 204 [14,17,18]. While hypothetical bias has been widely documented, multiple studies in con-205 sumer choice settings have noted that the bias affects the WTP level-that is, the total 206 amount the consumer is willing to pay for the good-but not marginal WTP for attributes 207 [19,20]. 208 A few studies complemented quantitative methods to understanding consumer percep-209 tions with qualitative approaches. Qualitative studies (or components of studies) included 210 interacting with small numbers of participants in focus groups [21] and face-to-face inter-211 views [21,22], as well as eliciting open-ended responses to questions from large numbers 212 of participants in online surveys [23]. This qualitative research identified themes related 213 to consumer attitudes towards NPETs, including concerns about risks of the use of these 214 novel technologies for human and environmental health, perceptions of unnaturalness of 215 the NPET-derived organisms, distrust in firms' use of NPETs to modify organisms, and 216 misperceptions about the food production system (e.g., concerns that modifying dairy 217 cattle to eliminate horns would prevent them from fighting off predators) [21-23]. 218 219 The first key-and quite robust-finding is that consumers on average discount food 220 goods generated using NPETs relative to foods produced using traditional breeding tech-221 niques. All studies reflect this discounting of NPET-based goods relative to conventional 222 goods (or NPET-based improvements relative to similar improvements generated from 223 conventional breeding techniques), when averaging over all consumers surveyed or sub-224 jects in experiments. However, when compared to WTP for (transgenic) GMOs, NPET-225 based innovations and goods tend to be valued more highly than their GMO counterparts, 226 provided they embody improvements beneficial to the environment or human and animal 227 health. This finding is also robust. 228 Another important result common to many investigations is that there exists multi-229 dimensional heterogeneity among consumers with respect to their acceptance of and WTP 230 for NPETs. Forty-three investigations find some form of heterogeneity, either by identify-231 ing a segment of consumers who heavily discount the novel foods or are not willing to 232 consume or purchase them at any price; or through statistically significant standard devi-233 ations of estimated parameters capturing the range of WTPs in the sampled population. 234 Consumers show heterogeneous levels of knowledge about NPETs, have various atti-235 tudes towards food innovations and technology, have variable ethical concerns about nat-236 uralness of NPET-based foods, and have varying concerns about the risk the use of NPETs 237 presents for health and the environment. These multiple aspects influence the willingness 238 to consume and WTP for NPET-based novel foods, including products that feature im-239 proved attributes with clear, tangible benefits to the consumer or society. This also means 240 that there is a market segment for these novel foods when they offer additional health, 241 taste or environmental benefits, appealing to consumers who are open to food innovations 242 [24][25][26]. 243 An important source of heterogeneity seems to arise from consumers' country of res-244 idence, which may reflect varying regulatory approaches or cultural values; for instance, 245 trust in the regulatory bodies of one's home country is associated with attitudes towards 246 approved technologies [21]. All but one study find marked differences in WTP or willing-247 ness to consume among countries. The exception (Ferrari et al. [27]) compares young con-248 sumers in Belgium and the Netherlands, neighboring countries with a common culture, 249 who are "millennials" or members of Generation Z, who may be more accepting of the 250 use of NPET technology than older generations [28]. The range of attitudes, concerns and 251 attitudes gets amplified with geographic and cultural distance, which reflects findings 252 from the literature on GMO-based agriculture and food [29,30]. In particular, the divide 253 between the European continent and North America is as striking as it was for GMO-254 based foods. For example, French consumers have lower acceptance and/or WTP for 255 NPET-based foods than U.S. and Canadian consumers do (see, for instance, Lusk and 256 Rozan [9] on vegetables; Marette et al. [24,25] for apples; Narh et al. [31] on rice; and Shew 257 et al. [15] on acceptance of CRISPR rice). In addition, in many WTP studies based on dis-258 crete choices, the standard deviations of most relevant parameters are significant, indicat-259 ing that the valuation of attributes is heterogeneous. Within Europe, perceived risks and 260 concerns about NPET-derived food are much lower than they were for transgenic food 261 but they remain highly heterogeneous across countries [32][33][34][35][36]. 262 The heterogeneity of acceptance and valuation of NPET-derived foods extends to the 263 type of food item and the process level, which is reminiscent of findings for GMO-based 264 food [29,30]. The lowest levels of acceptance are for meat and milk [37]. The relative WTP 265 for NPET-derived fresh tomato and spinach is higher than the WTP in processed form 266 (pasta sauce, frozen spinach). The opposite is true for bacon and pork produced using 267 NPETs. WTP for NPET-derived bacon-a more highly processed product-is higher than 268 the WTP for pork [38]. 269 WTP for NPET-derived foods increases with tangible improvements such as tastier 270 grapes, improved nutritional value, or environmental benefits (reduced pesticides, water 271 use) or improve animal welfare. Marginal improvements such as color of grapes or bene-272 fits accruing to farmers (more muscle mass on animals) tend to be discounted in NPET 273 valuation experiments. However, the premium over conventional substitutes lacking the 274 tangible improvements is limited in all these experiments. Unless some superlative attrib-275 ute is added, the improvements brought about by NPETs are likely to result in incremen-276 tal increases in WTP rather than drastic changes yielding higher valuations for NPET-277 derived products. 278 Knowledge-in various forms-also appears to be an important factor in consumer 279 response to NPETs. Higher levels of knowledge about science and technology promote 280 acceptance/WTP for the use of NPETs and NPET-derived products [21,27]. Greater 281 knowledge about the product being modified-specifically, in this case, wines-also pro-282 motes greater WTP for NPET-based products [28]. Interestingly, basic familiarity with 283 products that contain modified ingredients may also promote attitudes. A study of atti-284 tudes towards GMOs in the US found that residents of Vermont-which implemented the 285 first GMO labeling policy in the US-became more positive towards GMOs after the im-286 plementation of the labeling policy relative to residents of other states [39]. 287 An experiment that educated consumers about the function of genetic modification 288 technology in food production via a five-week course suggests a causal role for knowledge 289 [7]. Participants in the course developed more positive attitudes, greater willingness to 290 consume the foods, and decreased perceived risk of the foods during the course in three 291 countries: the US, the UK, and the Netherlands. A recent finding on knowledge and sup-292 port for GMOs highlights the importance of objective (i.e., measurable)-as opposed to 293 subjective (self-reported)-knowledge [40]. Those individuals who were the most op-294 posed to the use of GMOs had the lowest levels of objective knowledge, but believed that 295 they had high levels of knowledge about GMOs [40]. Several investigations focus on in-296 formation and communication strategies implications to increase acceptance of these 297 NPETs, building on lessons learned with GMOs (see De Marchi et al. [41], Marette et al. 298 [24], Edenbrandt et al. [42]). However, consumers can get confused by conflicting mes-299 sages and these cancel out [10,11]. 300 A time aspect: for GMO, EU consumers were much more worried in 2010 than they 301 were in 2019 about GMO in their food supply. The concern for GE is already small relative 302 to GMOs, so NPET-based foods may have an easier transition to acceptance. 303 In experiments addressing labelling of NPETs derived foods, labelling is preferred, 304 especially in European countries [10][11][12]21,27,36,43]. To the extent that consumers may 305 feel deceived if not informed about the use of NPETs in the development of ingredients 306 or foods they purchase, there is a legitimate reason to add a label, including on imported 307 goods [44]. However, consumers may pay less attention to attributes-including the use 308 of NPETs-in real buying/retailing environment when information and sensory overload 309 is heightened.

311
In summary and with the appropriate qualifiers spelled out in the previous sections, 312 the accumulated evidence suggests that large segments of consumers, but not all, are will-313 ing to consume and pay for NPET-derived foods, especially if they embody useful traits 314 that the consumers perceive as beneficial for human and animal health and the environ-315 ment. However, these foods tend to be discounted relative to close substitutes obtained 316 through conventional breeding methods. In most situations when informed about these 317 useful traits, consumers discount NPET-derived foods to a lesser extent than their trans-318 genic (GMO) substitutes. They also find them more "natural" although their knowledge 319 about and familiarity with NPETs are limited. 320 The major limitation of current knowledge on consumers' behavior vis a vis NPETs is that 321 most of these elicited WTPs and attitudes are based on hypothetical choices and/or in ar-322 tificial settings of lab experiments, experimental auctions, or online surveys. The limited 323 commercialization of NPET-based foods precludes study of consumer preferences for 324 these products under more natural, or at least incentivized, conditions. Future validation 325 or falsification of these findings in real retailing situations will be possible once these novel 326 foods become widely available. 327 Labelling is probably preferable as consumers are concerned by process attributes and 328 want to know the improved characteristics of the novel food and how they have been 329 derived. It remains to be seen how consumers will react in real shopping environments 330 when a deluge of information signals might cancel each other and might not be as instru-331 mental as declared in hypothetical choices. Colson's work suggests this possibility in an 332 auction setting [10,11]. However, the incorporation of NPET-based ingredients may also 333 promote acceptance of the technology if labeling is present to help consumers make the 334 connection, as apparently occurred with GMO-labeling [39]. 335 We assessed the promising demand side of the market for NPET-derived foods. How will 336 the supply side shape up and how will specialized markets develop for NPET-derived 337 foods? NPETs do not require the scale of transgenic biotechnology as they are much less 338 expensive in the R&D stage, especially for emerging techniques like CRISPR [45]. These 339 technologies, initially driven by non-profit research institutions, have led to an unusual 340 number of patents globally, and many startups [46,47]. Nevertheless, scale is useful for 341 marketing and distribution aspects of food and food retail markets are typically competi-342 tive environments. It would be useful to assess commercialization efforts of these novel 343 foods. The current regulatory uncertainty on NPETs may also inhibit the emergence of 344 these markets [3,48,49].

345
Supplementary Materials: The following Table S1 is available online at www.mdpi.com/xxx/