Introduction

In October 2024, the South African Supreme Court of Appeal overturned the approval of MON87460, a genetically modified (GM) drought-tolerant (DT) maize event that had been commercially available since 2016 and was previously approved by the South African Executive Council (EC). The decision was based on the court’s opinion that there were regulatory failures, including failing to properly assess environmental impacts and failing to apply the precautionary principle. The 2024 judicial decision is the first of its kind concerning GM decision-making in South Africa. Importantly, MON87460 was considered a tool to effectively tolerate drought and protect maize producers from severe yield loss under drought stress1. This ruling highlights the ongoing controversy surrounding GM crops in South Africa and could potentially limit or delay the adoption of tools needed to combat the increasing frequency and severity of droughts, affecting food security in the country.

Although food insecurity has multiple dimensions, drought remains a persistent problem in South Africa, affecting both national and household food security. The World Bank estimated that there have been 11 major drought events in South Africa since 1990, causing over USD 2.6 billion in damages2. Specifically, a severe El Niño-induced drought across South Africa during the 2015/2016 season prevented farmers from planting nearly one million hectares of maize, resulting in a 50% decline in maize production compared to the previous five-year average3. Between November 2014 and November 2015, an estimated 22% of South African households lacked sufficient income to buy food4. The situation was especially severe in certain provinces, with food insecurity affecting 41% of households in the North West, 32% in the Eastern Cape, 31% in the Northern Cape, and 26% in the Free State. The higher rates of hunger were driven by increased cereal prices, mainly maize, which is fed to both humans and livestock, rising by an estimated 53.7%5.

Traditionally, GM crops have focused on traits like disease and pest resistance, herbicide tolerance, and biofortification. The Water Efficient Maize for Africa (WEMA) project, parts of which involve GM technology, was launched in 2008 as a public-private partnership led by the African Agricultural Technology Foundation (AATF) to reduce yield variability caused by increased drought frequency and severity across Africa6. The WEMA project aimed to develop and distribute GM DT maize varieties across accepting African countries. These varieties have been shown to yield between 7 and 15%7, 20 to 35%8, and 24 to 35%9 more under moderate drought conditions compared to currently adopted commercial maize varieties in Africa7. The project’s long-term goal is to make DT maize available royalty-free to small-scale farmers in Africa through African seed companies, thereby mitigating drought risk and stabilizing crop yields10. The varieties used by African farmers are royalty-free7. Notably, the recent South African Supreme Court of Appeals ruling pertains to MON87460, which is part of the WEMA project. Some varieties of the WEMA project that are now commercially available in South Africa are non-GM and go by the trade name TEGO, while products from the TELA project are called TELA and are Bt-traited. Given the fluid nature of naming subcategories of varieties, this study uses the umbrella term “WEMA” and specifically denotes that the varieties in question are genetically modified.

As of 2019, about 94.6% of white maize and 85.6% of yellow maize were produced under dryland conditions in South Africa11. White maize (for human consumption) accounts for 85% of total production in South Africa, with the remaining 15% being yellow maize (for livestock). Given the uncertainty of future rainfall patterns and amounts, it is crucial to explore sustainable solutions to reduce the impacts of climate change on maize production in South Africa and the broader Southern African region. Droughts and extreme heat events will continue to threaten food security through maize (the staple crop) and beef (which relies on maize). This challenge is likely to worsen as weather and climate volatility increase in Southern Africa, with both higher temperatures and drier conditions expected due to global climate change12.

GM crops are crucial for addressing hidden hunger and food insecurity in Africa13 and have been shown to have a positive impact on the global economy, environment, and human health14. The WEMA project has the potential to significantly enhance food security and resilience in Africa, particularly in regions prone to drought. However, a significant obstacle to the adoption of GM crops across Africa is political uncertainty and consumer resistance. Although the various benefits of GM foods may be recognized from the scientific point of view, from the consumers’ preferences and perspectives, the issue is more nebulous15. Therefore, it is crucial to have a clear understanding of consumer concerns and acceptance of GM crops, particularly drought-resistant varieties, during drought periods.

Public resistance to GM crops in Africa has persisted, with some opposition stemming from fear of the unknown rather than from scientifically justified concerns16. The acceptance of novel food technologies, such as GM maize, depends not only on consumer perceptions of the risks and benefits but also on the uncertainty and fear surrounding their effects on human health17. Consumers often obtain information about GM foods from various channels, including the internet, where such information varies in scientific quality and can be false18. Consumers have been shown to face difficulties assessing the risks associated with new technologies, such as GM crops, and the benefits arising from their use19. However, aversion to GM crops in Africa has softened during droughts. In November 2012, the Kenyan government banned all imports of GM crops, citing public health concerns16. In October 2022, Kenya lifted its 10-year ban on GM crops in response to East Africa’s worst drought in 40 years, hoping that GM maize would improve yields and enhance food security20.

This study aims to determine whether consumers in South Africa perceive all GM maize (herbicide-tolerant, insect-tolerant, DT, etc.) as identical, regardless of the trait, or if they have different preferences for each trait. Specifically, we are interested in whether consumer acceptance of GM maize varies when the benefits of WEMA are highlighted. This is especially important given the 2024 decision by the South African Supreme Court of Appeals to overturn the commercial release of the GM DT maize variety MON87460. While there is extensive literature on consumer acceptance or reluctance toward traditional GM traits, research is limited on traits like drought tolerance, which are explicitly bred to improve food security and stabilize maize prices. It is crucial to understand whether consumers perceive a GM trait intended to stabilize regional food security differently from an agronomic trait with more limited consumer benefits. As scientists work to adapt to the impacts of climate change, WEMA is seen as a potential tool that could have significant, widespread effects. However, if consumers cannot distinguish between different GM traits and instead view GM as a single, unified tool, the overall adoption and benefits of WEMA could be restricted.

In 1997, South Africa became the first African country to commercialize GM crops, starting with Bt cotton, followed by Bt maize in 1998, and herbicide-tolerant cotton and soybeans in 200016,21. Other African countries have been held back by the lack of legislation or policies to authorize the adoption of biotechnology, such as GM. Among the 54 African nations, as of 2024, only six countries: Egypt (maize), Ghana (cowpea), Kenya (maize), Nigeria (cowpea, maize, soybean, and wheat), South Africa (maize, rice, and soybean), and Zambia (maize) have approved cultivation of GM food crops22. The only GM crops commercially produced for direct human consumption are papaya, squash, apples, potatoes, eggplants, and white maize23. Notably, white maize is the only staple crop produced commercially using GM varieties in a field-to-plate system. In 2017, South Africa cultivated approximately 1.1 million hectares of GM varieties for direct human consumption (an 85% adoption rate).

The Bill & Melinda Gates Foundation, the U.S. Agency for International Development (USAID), and the Howard G. Buffett Foundation jointly funded the WEMA project to protect low-income consumers from the adverse effects of drought. The AATF collaborated with the International Maize and Wheat Improvement Center (CIMMYT), Bayer Crop Science, and National Agriculture Research Systems (NARS) in South Africa, Mozambique, Uganda, Tanzania, Ethiopia, and Kenya to distribute this technology to maize farmers across Africa.

Monsanto (now part of Bayer Crop Science) provided maize varieties from their global collection, including DT maize technology (MON 87460) and insect-resistant maize event (MON 810)23,24,25. Unlike other first-generation GM traits that focus on agronomic aspects and primarily benefit producers (such as Bt and herbicide tolerance), WEMA maize aims to stabilize regional food security. Monsanto applied to the South African Executive Council (EC) in 2014 for a permit to commercially release MON87460, which had been genetically modified to reduce yield loss in water-limited conditions. MON87460’s DT trait is achieved through the expression of the inserted Bacillus subtilis cold shock protein B (CSPB), which facilitates adaptation to environmental stresses, such as water scarcity, by binding to secondary RNA structures, thereby helping to preserve normal cellular functions26. In June 2015, the EC of South Africa approved the general release of MON87460.

South Africa commercialized insect-resistant (Bt) TELA maize (a derivative of WEMA) in 2016, and South African farmers are growing royalty-free seeds27. The launch of WEMA, also known as TELA maize, in South Africa in 2016 marks a significant milestone. However, in August 2015, the African Center for Biodiversity appealed the commercial release of MON87460, arguing that Monsanto had violated the precautionary principle. This principle requires that when an activity could significantly impact the environment, decision-makers adopt a risk-averse, cautious approach that considers the limits of current knowledge about potential impacts28.

Results

Demographics

Table 1 compares the sample’s demographic data with the South African population data from the most recent census, conducted in 2011. Both surveys indicate a higher proportion of females and a lower proportion of males than the general population. The Gauteng province (Johannesburg and Pretoria) is overrepresented relative to the general population. Additionally, both surveys have an underrepresentation of Black/African individuals and an overrepresentation of White, Colored, and Indian individuals compared to the population. The sample is overeducated relative to the South African population.

Table 1 Socio-demographic characteristics in percentages
Full size table

Mixed logistic regression results

Table 2 displays the results of mixed logistic regression models examining the impact of independent variables on purchasing behavior in both the “non-drought” and “drought” surveys. In non-drought conditions, white maize meal had the highest coefficient (9.962), followed by GM white maize meal, GM yellow maize meal, and price. The positive coefficient for white maize meal indicates that respondents are significantly more likely to choose white maize meal over not buying maize meal. Similarly, the positive coefficients for GM white maize meal and GM yellow maize meal suggest a preference for these options over the “no buy” option. However, the likelihood of choosing these options decreases from white maize to GM white maize and from GM white maize to GM yellow maize. The price coefficient was statistically significant and negative, which aligns with economic theory (price has a negative impact on utility).

Table 2 Mixed logistic regression results for drought and non-drought surveys
Full size table

Both the “non-drought” and “drought” models use the “no buy” option as the base product to estimate the alternative specific constants. Group effects are measured throughout the interaction coefficient for each alternative, where the group that received no information and GM white maize meal (GM no info) was used as the base experimental group. While only a few interaction terms were significant at p < 0.05, a simultaneous test for group effects was significant, suggesting that including interaction effects enhanced model fit. The significant (P < 0.05) interaction terms were associated with the groups that received information about the benefits of WEMA (both WEMA info and GM info). Groups receiving GM information significantly prefer GM white maize when provided with information compared to those not receiving it (p < 0.01). Similarly, respondents in the WEMA info group have a white GM coefficient greater than that of the GM info group at the same level of statistical significance (p < 0.01). This suggests that respondents are more likely to choose GM maize in the WEMA information group than in the GM information group.

Respondents did not receive a “no buy” option in the drought survey (to mimic the necessity of purchasing something during a drought). Hence, M2 uses GM yellow maize as the base maize product to estimate the ASCs, and group effects are measured throughout the interaction coefficient for one alternative, where the group that received no information (WEMA no info) is used as the base group. The respondents who received information (WEMA info) had a negative coefficient for white maize meal, indicating that they were less likely to choose white maize meal compared to those who did not receive information (p < 0.05).

Pooled non-drought conditional and unconditional

Figure 1 displays conditional and unconditional predictive margins from the non-drought survey. The unconditional predictive margins for white maize meal in the WEMA no info group were estimated at 44%, while margins for WEMA maize meal were 32%. In contrast, these percentages were reversed for the group provided with information about WEMA (WEMA info), with white maize meal at 32% and WEMA maize meal at 44%. The change in percentages between the two groups highlights the influence of relevant information on purchasing decisions.

Fig. 1: Pooled predictive margins (%) from the drought and non-drought surveys by treatment groups.
Fig. 1: Pooled predictive margins (%) from the drought and non-drought surveys by treatment groups.The alternative text for this image may have been generated using AI.
Full size image

Source: Authors’ depiction based on data analysis.

Interestingly, the unconditional and conditional predictive margins for WEMA white maize meal in the WEMA info group are 6% higher than those for GM white maize meal in the GM info group. This suggests that the specific benefits associated with WEMA maize may be more compelling or more straightforward for consumers to understand and appreciate compared to the broader category of labeling a product as made with GM white maize. The higher predictive margins for WEMA maize may also indicate that consumers do not view all GM maize products similarly.

The predictive margins for GM yellow maize across all groups (WEMA info, WEMA no info, GM info, and GM no info) for unconditional results remained robust, with an average of 18.5% and a range of 17–20%. All treatment groups for the conditional results of GM yellow maize meal were consistent, except for WEMA information. The WEMA info group showed a 10% decrease in preference for GM yellow maize meal compared to the WEMA no info group. This highlights that the information provided about WEMA maize shifted consumer preference away from GM yellow maize meal. The average of the GM yellow maize meal predictive margins for GM info, GM no info, and WEMA no info treatment groups was 19%, with a range of 18–20%.

Examining the conditional predictive margins, at least 80% of respondents were found to prefer GM white maize meal or conventional white maize meal over GM yellow maize meal. This is consistent with the preference for white maize over yellow maize, specifically in the black South African population across South Africa. The unconditional results in Fig. 1 indicate that at least 75% of respondents preferred GM white maize or conventional white maize over GM yellow maize meal or chose not to purchase any maize meal. Importantly, this likely indicates that imports of yellow maize may not curb food insecurity if there is a severe enough drought in the future to warrant maize imports from outside the Southern African region. Given the limited supply of white maize on the international market, it would be problematic should South Africa experience another drought like 2015/2016.

Pooled drought conditional

In the drought survey, 48% of respondents in the WEMA info group chose WEMA white maize meal, compared to 37% in the WEMA no info group. Conversely, 24% of respondents in the WEMA info group chose white maize meal, compared to 38% in the WEMA no info group. The significance of informing consumers is evident, with a difference of more than 10% between maize varieties across both groups, and WEMA consistently showing higher predictive margins among informed respondents. Therefore, we can assume that if more consumers received information about how WEMA maize improves yields under drought conditions, they would likely be positively influenced to purchase WEMA maize over non-GM white maize.

In the drought survey, consumers’ preference for WEMA white maize increased by 11% (from 37% to 48%) when they were given information about WEMA maize. This highlights the importance of effectively conveying the benefits of WEMA to South African consumers before the next large-scale drought. These results suggest consumers value this drought attribute more (through decreased prices) only after a drought occurs (comparing these results to the non-drought results). This is problematic given that producers will respond to consumer demand and cannot predict when a drought will occur or its severity.

Pooled drought and non-drought comparison

When provided information about WEMA, the preference share for WEMA is consistently higher than GM maize meal in the non-drought survey for both conditional and unconditional modeling. The preference increases from 34% and 32% in the WEMA no-info group to 46% and 44% in the WEMA info group. This trend is consistent with the drought increase from 37% in the WEMA no-info group to 48% in the WEMA info group. A notable difference is the increase in the selection of GM yellow maize from drought to non-drought conditions. The WEMA info treatment group in the drought survey has the highest predictive margin for GM yellow maize at 27%. This treatment group (WEMA info) in Fig. 1 for the drought survey exhibits the only instance where the margin for GM yellow maize exceeds that for white maize. In the drought survey, 83% of respondents were either middle- or low-income, consistent with the non-drought survey at 84%, making the lower prices of GM yellow maize appealing to this consumer group. However, the margin nearly doubles to 48% for WEMA maize, despite its higher price compared to GM yellow maize, indicating a strong preference for WEMA white maize, even with an approximately R20 price increase per 2.5/kg bag, between GM yellow maize and WEMA white maize.

Non-drought conditional by race

A study from 1999 to 2021 showed that the highest prevalence of hunger occurred in South African households of Black Africans (30%) and those of mixed ethnic origin (referred to as Coloureds by Statistics South Africa) (13.1%), followed by Indians (8.6%), while only 1.3% of White households experienced hunger29. Given that maize is the staple in black South African households, it is important to delineate preferences for maize by race. In the non-drought survey, predictive margins remained robust across treatment groups for WEMA white, GM white, and GM yellow maize meal among black respondents (see Fig. 2). White maize meal had an average predictive margin of 42.5%, with individual margins ranging from 39% to 45%. WEMA white maize meal had an average predictive margin of 41.5%, with individual margins ranging from 39% to 44%. While GM white maize meal averaged a predictive margin of 40%, with individual margins ranging from 39% to 41%.

Fig. 2: Conditional predictive margins (%) by race from the drought and non-drought surveys by treatment groups.
Fig. 2: Conditional predictive margins (%) by race from the drought and non-drought surveys by treatment groups.The alternative text for this image may have been generated using AI.
Full size image

Source: Authors’ depiction based on data analysis.

Black respondents selecting GM yellow maize meal had an average predictive margin of 17% across all treatment groups, with individual margins ranging from 16% to 18%, reflecting a 23% decrease from the lowest average among other maize meal options. These results are consistent with a study of black South Africans, who reported that the color, taste, and smell of yellow maize were unpleasant. During a drought, importing yellow maize meal would not effectively address hunger in South Africa, as the Black South African population has a prevailing preference for white maize meal.

Drought conditional by race

In the drought survey, 51% of Black respondents in the WEMA info treatment group opted for WEMA maize. Figure 2 shows a 14% increase from black respondents in the WEMA no info group to the WEMA info group. However, respondents from the other racial groups (excluding Black and White) in the survey showed slight variations in their preferences between the WEMA no info group and the WEMA info group. Instead, their preference for GM yellow maize increased by 12%, from 23% to 35%, and their preference for white maize decreased by 13%, from 38% to 25%, when comparing the WEMA no-info group to the WEMA info group. This suggests that the black population may prioritize WEMA maize over GM yellow maize, despite a price difference of R20 (per 2.5 kg bag), as WEMA is more expensive, while the other racial groups seem to prioritize price over maize color. These results for non-black consumers’ purchasing preferences are supported by a study conducted in Kenya, which found that consumers, on average, require a 37% price discount to accept yellow maize (De Groote and Kimenju, 30). The prices for yellow GM maize in our drought survey were discounted by 39%. However, with black South Africans making up 79.2% of the population, importing yellow GM maize would only feed a small portion of the population in South Africa. This underscores the importance of educating consumers about the benefits of GM white maize varieties such as WEMA.

Discussion

Despite being classified as an upper-middle-income country, South Africa continues to face the threat of significant food insecurity. Climate change, crop yield variability, sporadic power cuts, and poverty influence this ongoing issue. Over the past 50 years, Southern Africa has experienced heightened drought severity and frequency due to climate change, negatively impacting maize yields. The 2014/2015 drought saw white maize prices more than double, resulting in food insecurity affecting 22% of households across South Africa. One climate change mitigation effort is the adoption of water-efficient maize (WEMA), which embodies a GM drought-tolerant trait. Despite the widespread adoption of GM maize (herbicide-tolerant, Bt, and stacked technologies) in South Africa, consumer skepticism of GM persists. Misconceptions about genetically modified (GM) food, likely stemming from misinformation or a lack of information, influence consumers’ thinking and decision-making. Unlike most countries outside Africa, South Africa relies on white maize for consumption, which limits import availability during shortages, given the small percentage of global area planted with white maize. This cultural factor makes GM yellow maize, even at a lower price point, a less viable solution to improving food security during times of white maize scarcity.

In the drought survey, WEMA maize’s market share increased at the expense of non-GM maize after receiving information about the benefits of WEMA. The market share of WEMA maize meal (48%) was twice that of non-GM white maize (24%). This suggests that when faced with food insecurity (due to higher prices) and information about the benefits of WEMA, consumers were less averse to GM maize. In the non-drought survey, the results when respondents were given information were similar but less drastic, with an increase in WEMA market share (34–46%) and a decrease in non-GM maize market share (46–33%). This would suggest that if widespread acceptance of WEMA were to occur, its benefits would need to be highlighted to South African consumers.

This study’s findings robustly reveal that providing information to consumers about WEMA and its ability to help mitigate maize price volatility during droughts enhances market acceptance relative to conventional white maize. Despite its stable price across both surveys due to the ability to import, yellow maize has only a marginal market share, posing a food security challenge due to the restricted availability of white maize internationally in the event of another drought akin to 2015/2016. Given the cultural attachment to white maize (with yellow maize typically associated with livestock feed), this study suggests that an attempt to combat food insecurity through the import of yellow maize during droughts would likely falter.

Achieving higher adoption of WEMA maize to stabilize maize prices during a drought requires enhanced consumer awareness of WEMA’s benefits in South Africa through information (either via campaigns or on-packaging labeling). Given that the WEMA yield benefits are primarily expressed under adverse agricultural conditions, the benefits in terms of lower prices may not be evident until another severe drought occurs. This is problematic because if demand is low or aversion exists, maize producers in South Africa may respond by not planting WEMA, given its lower yield potential under ideal climate conditions. When another drought occurs, it will be too late to market WEMA for that given year, which would increase the likelihood of price instability in the maize market. Thus, educating consumers before the next drought about WEMA may alter consumer thinking on GM maize and lead to higher acceptance and widespread adoption amongst producers. Specifically, because most people lack knowledge about GMOs and their traits, in this case, drought tolerance, the common attitude is likely to be a general aversion31. Previous research has shown that increasing the effectiveness of messaging via GM labeling, general education, and campaigns targeting individuals with lower levels of education is necessary32.

What is not known is whether the increased market share (for WEMA) in our survey represents a snapshot of a new and novel product or a long-term trend. Consumers may be excited to try a new product, but their institutional memory of benefits, specifically those that occur infrequently (such as drought), may deteriorate rapidly. Further complicating issues is the heavily publicized political debate over GM and GM labeling, and how that may have reshaped the landscape in South Africa. Given its high level of food insecurity, the increased frequency and severity of droughts, and its reliance on maize as a staple crop, WEMA may have significant market potential in South Africa. Studies like this aim to shed light on how the GM maize market may evolve in the future.

These studies’ findings are pertinent in light of the October 2024 South African Supreme Court of Appeal ruling that overturned the approval of MON87460. Outside of the approval/safety arm of commercialization of GM crops in South Africa, this study highlights that consumers in South Africa would be more willing to consume a GM-WEMA trait than traditional GM traits. This study has shown that consumers in South Africa value GM traits heterogeneously, which is important because as new GM breeding techniques are introduced (such as CRISPR) to combat issues like drought and heat stress this could help expand maize production areas currently considered marginal and, more importantly, can become an adaptation practice to climate change and the increasing occurrence of droughts and heat shocks. Further, GM maize can also help combat undernourishment and malnutrition through biofortification. Previous research has shown that laws (directives) on GMOs are often too restrictive and suffer from over-regulation that prevents any attempt to come to a science-based approach on genetically modified (GM) plants33. While regulation, such as the overturning of the approval for MON87460, is crucial for ensuring safety and environmental stewardship, overly restrictive or cumbersome regulatory frameworks can inadvertently hinder the adoption and development of genetically modified crops, thereby impacting food security and nutrition efforts in South Africa.

Methods

Non-drought survey

An online choice experiment focusing on preferences for WEMA maize was conducted in 2023 with 2598 South African maize consumers. Respondents were presented with three choices of maize meal (the most common way to consume maize in South Africa), each priced differently, alongside the option to abstain from purchasing. There were two screening questions for survey respondents. The respondents had to be at least 18 years old and have purchased maize meal (mielie meal, imphumphu, bupi) within the last month. The survey was terminated if a respondent did not meet these two criteria.

Respondents were randomly assigned to receive either one set of options containing exclusively maize labeled as “GM” or another containing maize labeled as WEMA maize. One group received three maize meal options: conventional white maize, GM white maize, and GM yellow maize. Another group received three maize meal options, including conventional white maize meal, WEMA white maize, and GM yellow maize, as shown in Fig. 3. In South Africa and across Southern Africa, white maize is traditionally grown for human consumption, and yellow maize is produced for livestock feed. We included yellow maize in this study due to its implications for food security. Most of the world’s maize traded internationally is GM yellow maize. As such, during regional droughts across Southern Africa, yellow maize may be the only option when sourcing supplemental maize through imports from countries or regions such as Argentina, Brazil, the United States, and Canada.

Fig. 3: Example respondent choice set in survey.
Fig. 3: Example respondent choice set in survey.The alternative text for this image may have been generated using AI.
Full size image

The left hand figure illustrates the information choice set and the right hand figure illustrates the no-informatin choice set options.

Three price levels (i.e., low, medium, and high) were determined for each option based on the price ranges observed in South Africa during the summer of 2023. The price levels for 2.5 kg of yellow GM maize were R20.99 [1.13USD], R23.99 [1.30 USD], and R26.99 [1.46 USD]. Prices for 2.5 kg of white maize meal, the most expensive option, were R38.99 [2.11 USD], R41.99 [2.27 USD], and R44.99 [2.44 USD]. GM white maize and WEMA white maize meal prices were R29.99 [1.62 USD], R32.99 [1.79 USD], and R35.99 [1.95 USD] for 2.5 kg. Data were collected from nine choice sets, which allowed prices to be balanced (i.e., each price level appeared three times for a product) and orthogonal (i.e., price levels were uncorrelated across products).

The study employed a between-subjects design to investigate the impact of the information on consumers’ preferences for maize meal. Respondents were randomly assigned to one of four treatment groups (GM no info, GM info, WEMA no info, and WEMA info). Depending on the treatment group, respondents received additional information before answering the questions in the choice experiment. In the GM no info and WEMA no info, respondents were given no information about the benefits of GM or WEMA maize besides the label on each choice option. Information about WEMA and its associated benefits was provided to those in the GM info and WEMA info groups. Those in the GM info and WEMA info groups received the information in Table 3.

Table 3 Information about the benefits of WEMA
Full size table

All respondents were also instructed to imagine themselves shopping for maize meal in a grocery store, and a cheap talk (a set of instructions provided before bidding begins, designed to reduce behavioral biases that may otherwise distort willingness-to-pay estimates) was provided to reduce hypothetical bias34.

The data were collected through a nationwide online survey among South African maize consumers from April to May 2023. The survey was programmed in Qualtrics, and respondents were recruited by a market research agency (Dynata) using a quota-based sampling approach, which provided a representative sample of the South African population in terms of gender, age, race, and income. Any incomplete surveys or those completed in under a minute and a half were discarded from the analysis, resulting in the omission of 296 responses. A total of 2598 completed responses were collected and used in the study for the non-drought survey.

Drought survey

A second online choice survey focused on consumer preferences for the same maize meal types during a simulated drought, which would result in higher prices. The prices in the drought survey reflected the market price levels of white maize meal types during the 2015/2016 drought in South Africa. The price of conventional white maize meal increased by 100% from the non-drought survey price, mirroring the actual price increase from the 2015/2016 drought13. The price of WEMA white maize meal increased by 76% from the non-drought survey. This increase reflects the enhanced yield potential of WEMA maize during droughts, which yields 24–35% more grain under moderate drought conditions9. While in practice, maize meal in South Africa would likely not be identity preserved so that consumers could select 100% WEMA maize meal, this scenario was undertaken to see if attitudes towards WEMA would change if it resulted in lower prices during a period of drought.

The following price ranges were used for each: R51.99 to R61.99 [2.82 to 3.36 USD] for 2.5 kg of WEMA white maize meal and R76.99 to R88.99 [4.17 to 4.82 USD] for 2.5 kg of conventional white maize meal. The price of the GM yellow maize meal remained the same (R20.99, R23.99, and R26.99), reflecting the possibility of importing GM yellow maize from non-Southern African countries during periods of drought. Respondents were given the same maize options as those in the non-drought survey, except for the price increases for the white maize meal options. Again, data were collected from nine choice sets using a balanced and orthogonal design.

Any incomplete surveys or those completed in under a minute and a half were excluded from the analysis. Under these criteria, 100 responses were omitted from the analysis. A total of 606 completed responses were collected and used in the analysis for the non-drought survey (Table 4).

Table 4 Treatment groups from non-drought and drought surveys
Full size table

Econometric model

Discrete choice experiments (DCEs) follow the random utility theory assumption that when individuals assess the utility of a set of products, they choose the product that maximizes utility35. Utility cannot be determined directly; thus, indirect utility is derived from preferences for the attributes of a product. The utility function can be specified by:

$${{\boldsymbol{U}}}_{{\boldsymbol{ij}}}={{\boldsymbol{V}}}_{{\boldsymbol{ij}}}+{{\rm{\varepsilon }}}_{{\boldsymbol{ij}}}$$
(1)

where \({U}_{{ij}}\) denotes the utility of respondent i from choice alternative j, \({V}_{{ij}}\) denotes the observable component of utility and \({{\rm{\varepsilon }}}_{{ij}}\) denotes the unobservable component of utility. In this study, the j alternatives are associated with maize meal products.

Two random parameter logit (RPL) models were estimated for the observable utility component to allow for preference heterogeneity across respondents36,37. The estimated coefficients from the RPL model describe the proportion of individuals that derive utility from selecting the product37,38. A base model without group effects was first estimated, which can be specified by:

$${{\boldsymbol{V}}}_{{\boldsymbol{ij}}}={{\boldsymbol{ASC}}}_{{\bf{1}}{\boldsymbol{j}}}+{{\boldsymbol{\beta }}}_{{\bf{1}}}{{\boldsymbol{P}}}_{{\boldsymbol{j}}}+{{\rm{\varepsilon }}}_{{\boldsymbol{ij}}}$$
(2)

where \({{ASC}}_{1j}\) is the alternative specific constant (ASC) for the jth alternative, \({P}_{j}\) denotes price, and \({{\rm{\varepsilon }}}_{{\boldsymbol{ij}}}\) is an extreme value error term that is independently and identically distributed.

A model that included group interactions with the ASCs was then estimated to determine the effects of price levels, pre-purchase information, and point-of-purchase labeling on purchase decisions, which can be specified by:

$${{\boldsymbol{V}}}_{{\boldsymbol{ij}}}={{\boldsymbol{ASC}}}_{{\bf{2}}{\boldsymbol{j}}}+{{\boldsymbol{\beta }}}_{{\bf{2}}}{{\boldsymbol{P}}}_{{\boldsymbol{j}}}+{{\boldsymbol{\eta }}}_{{\boldsymbol{j}}}\left({{\boldsymbol{ASC}}}_{{\bf{2}}{\boldsymbol{j}}}{{\boldsymbol{T}}}_{{\boldsymbol{ig}}}\right)+{{\rm{\varepsilon }}}_{{\boldsymbol{ij}}}$$
(3)

where \({{\boldsymbol{T}}}_{{\boldsymbol{ij}}}\) denotes indicator variables equal to 1 for the gth group that respondent i was randomized to, and \({{\boldsymbol{\eta }}}_{{\boldsymbol{j}}}\) are the coefficients of interest that estimate group effects on purchasing decisions.

Predictive margins

The conditional (conditional on buying an option) and unconditional predictive margins are calculated to show the average proportion of respondents who selected a given choice option. Predictive margins are derived from the coefficients estimated by the RPL models and provide a more straightforward interpretation than the estimated coefficients. Predictive margins integrate over the full distribution of heterogeneous preferences and translate estimated utilities into changes in choice probabilities38. These can be used to make comparisons between products, groups, and across demographic characteristics39,40, making them particularly appropriate for evaluating product adoption and group effects.

Each option of maize meal had three different prices (low, middle, or high), balanced for each choice option and orthogonal across the choice options. This systematic variation in prices ensured equal representation of all possible combinations of price differences across the choice options.