Economic and Ecosystem Impacts of GM Maize in South Africa

White maize in South Africa is the only staple crop produced on a widespread commercial basis for direct human consumption using genetically modified (GM) cultivars. Using a combined economic and environmental approach, we estimate the total welfare benefits attributable to GM white maize in South Africa for 2001-2018 are $694.7 million. Food security benefits attributable to GM white maize in South Africa also manifest through an average of 4.6 million additional white maize rations annually. To achieve these additional annual rations using conventional hybrid maize, the additional land required would range from 1,088 hectares in 2001 to 217,788 hectares in 2014. Results indicate that GM maize reduces environmental damage by $0.34 per hectare or $291,721 annually, compared to conventional hybrid white maize. Acknowledgments I would like to express my gratitude to Lanier Nalley, Ph.D. for allowing me to conduct this multifaceted analysis on the effects of genetically modified maize adoption in South Africa. I am thankful for his patience, encouragement, and practical support over the past two years. I would also like to acknowledge and thank Marty Matlock, Ph.D., P.E., B.C.E.E. for his continued support. To Lanier Nalley, Ph.D., Marty Matlock, Ph.D., P.E., B.C.E.E., Petronella Chaminuka, Ph.D., Marijke D’Haese, Ph.D., Aaron Shew, Ph.D., and Jesse Tack, Ph.D. thank you for reading earlier versions of this manuscript and providing me with insightful commentary and improvements. I acknowledge and thank the South African Agricultural Research Council Grain Crops Institute (ARC-GC) for allowing access to the data used in this study. Thesis submitted in partial fulfilment of the requirements for the degree of Master of Science in Agricultural Economics issued by the University of Arkansas (United States of America) and the joint academic degree of International Master of Science in Rural Development from Ghent University (Belgium), Agrocampus Ouest (France), Humboldt University of Berlin (Germany), Slovak University of Agriculture in Nitra (Slovakia), University of Pisa (Italy) and University of Córdoba (Spain) in collaboration with Can Tho University (Vietnam), China Agricultural University (China), Escuela Superior Politécnica del Litoral (Ecuador), Nanjing Agricultural University (China), University of Agricultural Science Bengaluru (India), University of Pretoria (South-Africa) and University of Arkansas (United


Introduction
Contributing authors: Lawton Lanier Nalley, Aaron M. Shew, Jesse B. Tack, Petronella Chaminuka, Marty D. Matlock, and Marijke D'Haese White maize is an important field crop in South Africa, serving as the staple food for the majority of its population, particularly for low-income households (Abidoye and Mabaya, 2014;Gouse, 2013). Much of the research evaluating the impacts of transgenic crops (herein subsequently called genetically modified (GM) crops) has focused on the producer benefits (increased yields, reduced costs, or both) of input traits (Shi et al., 2013;Xu et al., 2013). Other findings conclude GM input traits have second-order socioeconomic impacts such as laborsavings and environmental benefits (Ahmed et al., 2020;Brookes and Barfoot, 2018;Gouse et al., 2016;Klümper and Qaim, 2014;Lusk et al., 2017;Qaim and Zilberman, 2003;Smyth et al., 2015;Xu et al., 2013). GM white maize in South Africa, typically produced as food for direct human consumption, provides a testable medium for the impacts of GM on the country's direct food supply. Critics suggest GM crops have not contributed to increases in yields nor led to reductions in pesticide usage (Gurian-Sherman, 2009;Hakim, 2016). Opponents have also highlighted a point in a National Academy of Sciences of the United States report, which stated that there was little evidence the introduction of GM crops in the United States led to increased yield potential beyond those of conventional crops (National Academies of Sciences Engineering and Medicine, 2016).

Food insecurity and climate change in South Africa
Although the World Bank classifies South Africa as an upper-middle-income country, food insecurity is an ongoing concern for a large segment of its population. In 2018, 11% of individuals and 10% of households in South Africa were vulnerable to hunger (STATSA, 2020).
Moreover, there has been a marginal increase in the prevalence of undernourishment from five percent (2.8 million people) in 2014 to six percent (3.5 million people) in 2017 (FAO, 2019). In 2014-2015, 22% of households experienced food insecurity due to a severe drought and subsequent staple food price shocks (STATSA, 2016). The price of white maize more than doubled between January and December of 2015 due to the drought (Stoddard, 2016). Household food insecurity, as a consequence of the drought, reached as high as 41% in North West province and 32%, 31%, and 26% in Eastern Cape, Northern Cape, and Free State provinces, respectively (STATSA, 2016).
Climate change threatens much of sub-Saharan African agriculture, and maize production specifically, through increased frequency and severity of droughts (Conway et al., 2015;Lobell et al., 2011Lobell et al., , 2008Rippke et al., 2016;Serdeczny et al., 2017;Travis, 2016). The degree to which climate change will reduce maize yield is uncertain (Conway et al., 2015;Lobell et al., 2008); however, even conservative estimates signal considerable food security implications in the region, as Botswana, Eswatini, Lesotho, and Namibia depend on importing maize from South Africa (Southern African Development Community, 2020). Producers, researchers, and policymakers alike are considering a wide range of options to reduce food insecurity in the region, including GM crops (Mtui, 2011;Muzhinji and Ntuli, 2021;Zilberman and Lefler, 2021).
Although HT and Bt traits do not necessarily maximize yield potential, these traits have been associated with narrowing the yield gaps through improved weed control, insect resistance, and more timely planting (Fisher and Edmeades, 2010); each of which constitutes an important adaptation strategy for mitigating the effects of climate change (Ortiz-Bobea and Tack, 2018).

Impacts of GM crop adoption
Since the United States commercialized GM crops in 1996, the global area of GM crops increased 112-fold, making GM crops the fastest adopted crop technology in recent times (ISAAA, 2019). The predominate traits in GM crops globally are herbicide tolerance (HT) for herbicides such as glyphosate, Bacillus thuringiensis (Bt) insecticidal traits, or both HT and Bt traits (stacked traits). Globally, HT crops account for 47% of the total GM area, Bt account for 12%, and stacked account for 12% (ISAAA, 2017). Systematic reviews of the literature largely confirm the producer and environmental benefits associated with the adoption of GM crops from individual studies. Klümper and Qaim (2014) completed a meta-analysis of 147 studies on the agronomic and economic impacts of GM crop adoption, finding that the profit gains of GM crops are 60% higher in low-income countries than in high-income countries (Klümper and Qaim, 2014). Impoverished farmers in low-income countries have benefited the most from GM technology where there are fewer options for pest management and crop vulnerability tends to be higher (Klümper and Qaim, 2014;Zilberman et al., 2018). Producer benefits associated with the adoption of HT crops can include increases in yield and reductions in costs as a result of improved weed control and reduced labor (Brookes and Barfoot, 2018;Gouse et al., 2016). Producer and environmental benefits associated with the adoption of Bt crops have been well documented (Barrows et al., 2014;Bennett et al., 2003;Kouser and Qaim, 2011;Qaim, 2014;Subramanian and Qaim, 2010;Vitale et al., 2014;Yorobe and Smale, 2012). Improved insect control has resulted in increases in yield and less insecticide applications, which in turn has resulted in, cost savings and reductions in pesticide toxicity (Kouser and Qaim, 2011;Qaim, 2014;Subramanian and Qaim, 2010). Another indirect impact of Bt maize adoption, has been the reduction of mycotoxin (e.g., fumonisin and aflatoxin) contamination which has resulted in economic benefits and improvements in human health (Wu, 2006;Yu et al., 2020). The improvements to human health as a result of mycotoxin reductions also disproportionately benefit low-income countries where fumonisin and aflatoxin levels are often high, and maize serves as a staple food (Wu, 2006). GM adoption has created economic and environmental improvements across the agricultural sector (Brookes and Barfoot, 2018;Gouse et al., 2016;Klümper and Qaim, 2014;Lusk et al., 2017;Smyth et al., 2015;Xu et al., 2013). In 2016, the direct global producer income benefit from GM crop adoption was estimated at US$18.2 billion with more than half of these benefits attributable to GM maize varieties (Brookes and Barfoot, 2018). In South Africa, the producer income benefit from GM crops for 1998-2016 was estimated at US$2.3 billion with 97% of these benefits attributable to GM maize varieties (Brookes and Barfoot, 2018). In 1996-2016, the adoption of HT maize and Bt maize resulted in eight percent and 56% reductions in herbicide and insecticide usage globally, respectively (Brookes and Barfoot, 2018). Klümper and Qaim (2014) found on average, GM technology has increased crop yields by 21%, while simultaneously reducing the amount of pesticide usage by 37% and pesticide costs by 39% (Klümper and Qaim, 2014 (Abidoye and Mabaya, 2014;Eicher et al., 2006;Smyth, 2017;Zepeda et al., 2013).

GM crops produced for direct human consumption
Field-to-plate GM crops that could have large implications for global food security have historically struggled to find traction globally. In 2010, India's environmental minister declared a moratorium on the commercial release of the Bt eggplant. This "moratorium" overturned the Genetic Engineering Approval Committee's decision-India's biotechnology regulatory panelwhich approved the Bt eggplant for commercial production (Bagla, 2010;Gupta et al., 2015).
Bangladesh approved the Bt eggplant for commercial production in 2017 and has been rapidly adopted by producers who have benefited from cost savings associated with reduced pesticide usage (Shelton et al., 2018). In 2013, anti-GM groups in the Philippines received worldwide attention after vandalizing test plots of Golden Rice in the Bicol region (McGrath, 2013).
Currently, Golden Rice is pending approval for commercial production in the Philippines (Wu, 2020).
To our knowledge, the only GM crops commercially produced for direct human consumption are the papaya, squash, apple, potato, eggplant, and white maize (ISAAA, 2017).
Notably, white maize is the only staple crop produced on a widespread commercial basis using GM varieties, and in 2017, South Africa commercially produced approximately 1.1 million hectares (85% adoption rate) of GM varieties for direct human consumption.

GM adoption in South Africa
In 1998-1999, Bt yellow maize was commercially adopted in South Africa. In 2001-2002, the adoption of Bt white maize established South Africa as the first GM subsistence crop producer in the world (Gouse, 2012). The commercial adoption of HT maize and stacked traits soon followed in -2004and 2007-2008, respectively (Gouse, 2012. In 2016, 74% of the country's total maize crop used HT cultivars, while 91% of the country's total maize crop used Bt cultivars (Brookes and Barfoot, 2018).
Given the criticisms that GM has not contributed to increased yields resulting in improved food security and increased producer profitability, the main objectives of this study are to estimate additional annual rations attributable to GM white maize adoption and to estimate producer profitability (both from a breakeven and relative profitability sense) between GM and conventional white maize in South Africa from 2001-2018. Further, this study compares the environmental impacts per hectare of GM and non-GM white maize production using Life Cycle Assessment (LCA). The study focuses on quantifying the GM impacts for field-to-plate white maize because few GM crops produced commercially are for direct human consumption and the literature is void of this type of analysis. The results of this study are unique in that they address both the large criticisms of GM crops, inability to increase food security and lack of producer profitability, in a medium that few GM studies have analyzed before, a field-to-plate crop. These results are important for policymakers, producers, consumers, NGOs, plant breeders, and other agricultural scientists addressing global food security. Grain South Africa (Grain SA) was founded in 1999 as a nonprofit organization to provide commodity specific support and services to South African producers. The area devoted to white maize in each of the nine South African provinces was obtained from Grain SA. Table   A1 shows the area grown to white maize by province in South Africa from 1999-2018.

Data and Methodology
This study uses yield (tons/hectare) data for white maize genotypes from the National Maize Cultivar Trials and estimated province level GM white maize yield premiums derived from Shew et al. (2021). Using a multivariate regression model, Shew et al. (2021) regressed yield for each cultivar in a province for a given year on an indicator variable for GM while controlling for location and year fixed effects, resulting in the estimated province level GM white maize yield premiums. The National Maize Cultivar data includes observations for both white and yellow maize, for a total of 58,952 observations across 106 locations ( Figure 1) and 491 cultivars across 28 years. Shew et al. (2021) define a "trial" by unique location-year combination of which there are 966 in the dataset. While the National Maize Cultivar Trials test both white and yellow maize, we focus on only white maize, as it is for direct human consumption (Shew et al., 2021). Although a gap between experimental and actual yields exists, Brennan (1984) concluded that the most reliable sources of relative yields are cultivar trials outside actual farm observations (Brennan, 1984). Although yields are often greater in experimental test plots as compared with producers' fields, the relative yield differences between varieties are assumed to be comparable (Shew et al., 2018).
GM maize adoption rates for South Africa were obtained from various sources (Abidoye and Mabaya, 2014;Esterhuizen, 2015;ISAAA, 2018ISAAA, , 2017ISAAA, , 2016. Table A2 provides an overview of GM white maize area in South Africa by province and the adoption rates for GM white maize in South Africa from 2001-2018. Although Bt yellow maize was first introduced in South Africa in 1999, this analysis is concerned with GM white maize production, and as such, begins in 2001 when Bt white maize was first commercially adopted. Due to limited data availability, for the years 2015-2018 the adoption rates were not disaggregated by maize color and are representative of both GM white and GM yellow maize. Adoption rates for the years 2001-2014 are disaggregated by white and yellow maize, and thus, are reflective of GM white maize adoption. It should be noted that maize production data (e.g., area and yield) and GM maize adoption rates used in this study are not disaggregated by commercial versus subsistence farmers (some estimates suggest that subsistence farmers contribute to approximately two percent of production (Lacambra et al., 2020)). Further research is warranted about the yield, income, and environmental effects of GM adoption between commercial and subsistence white maize producers in South Africa.

Estimation of additional annual rations
For the purposes of this study, annual ration was defined as the average annual consumption of white maize for one individual in South Africa. To estimate the number of additional annual rations attributed to GM white maize adoption in South Africa, the additional tons of white maize produced by province p for year t attributable to GM adoption (Q A P,t) were estimated as: Where Apt denotes the area in hectares devoted to white maize production in province p for year t, bpt denotes the percentage of white maize produced that was GM in province p for year t, and yp denotes the estimated yield premium associated with GM white maize adoption for province p in tons per hectare from (Shew et al., 2021).
The number of additional annual rations, R, attributable to GM maize adoption in year t is estimated as (Rt): Where åQ A pt denotes the summation of additional tons of white maize produced in each of the nine provinces of South Africa, p, attributable to GM maize adoption in South Africa for year t, and Q R t denotes the maize consumption in kilograms per capita per year in South Africa for year t. South Africa's per capita maize consumption between 2001-2017 was retrieved from FAOSTAT (Table 2) (FAO, 2020(FAO, , 2017.

Estimation of welfare gains
An equilibrium displacement model (EDM) was developed to quantify changes in producer and consumer surplus attributable to the adoption of GM technology in white maize production in South Africa (Wohlgenant, 2011). The EDM employed for this analysis was specified as: , , Where EQD and EQS are the relative changes in demand and supply, respectively, EP denotes the relative change in market equilibrium price, h denotes the own price elasticity of white maize, e denotes the supply elasticity of white maize, d denotes the relative change in demand, and k denotes the relative change in supply.
The relative change in demand, d, was assumed constant. The relative change in supply, k, was calculated as the average percent share of additional white maize attributable to GM adoption from the supply of white maize as a sum of domestic production and imports for years 2001-2018. The relative change in supply for the years 2001-2018, k, was estimated as a 7.05% shift upward.
Existing literature on the estimates of the elasticity of demand, h, and supply, e, of white maize in South Africa was sparse. For this analysis, an own price elasticity of -0.149 was used, since it was specific to white maize in South Africa (van Zyl, 1986). A long-run supply elasticity for maize of 0.36 was used since it was specific to maize in South Africa ) and provides a more conservative estimation of consumer welfare compared to other estimates Rosegrant et al., 1995). Further sensitivity analysis was performed using all elasticity of demand and supply combinations found in the literature.
The analytical solutions for changes in market price and quantity were specified as: Given these analytical solutions, the changes in consumer surplus (DCSt), producer surplus (DPSt), and net surplus (NSt) for year t were derived: ,

Ecosystem impacts of GM maize adoption
The LCA framework was employed to quantitatively compare cradle-to-farm gate environmental effects associated with conventional and GM maize production in South Africa.
Comparisons were made between one hectare of conventional white maize and one hectare of GM maize using the LCA modeling platform SimaPro 9.0.0.48 © (Pre' Sustainability, Amsersfoort, The Netherlands) and the ecoinvent database (Wernet et al., 2016 (Shew et al., 2021). Table 3 shows the recommended pesticides and herbicides used in one hectare of conventional and GM (stacked technology) maize production during one growing season (Grain SA, 2019a). Other inputs (e.g., fuel, fertilizer, land preparation, etc.) were assumed constant across maize seed technology due to a lack of data, however, the main differences in both production systems manifest from the differences in yield and pesticide requirements.
The Stepwise Life Cycle Impact Assessment (LCIA) method was utilized to provide a combined score for both human and environmental effects, in dollar terms (Weidema, 2009).  (Weidema, 2009). Based on the budget constraint, it is estimated the potential annual economic production per capita is 88,737 (2018 USD) (Weidema, 2005). The Stepwise method assigns a cost of 1/14 QALY per BAHY (Weidema, 2015). This allows impacts associated with resources and ecosystems to be expressed in the same units as impacts associated with human well-being.
Subsequent calculations were performed to derive the environmental externalities (E Externalities) differences from GM white maize adoption for year t and the net impacts of GM white maize adoption (Net Impacts) for year t: Stepwise LCIA for one hectare of conventional white maize production and one hectare of GM white maize production, respectively, and NSt denotes the change in net surplus for year t in USD 2018 (Equation 10). By using the stacked GM technology (instead of HT or Bt) this study likely overestimates total LCA benefits when aggregated up to South Africa in its entirety, as approximately 66% of the total GM maize production was under stacked technology in 2018 followed by HT (23%) and Bt (11%) (ISAAA, 2018). However, using the LCA metric of comparing one hectare of conventional maize to one hectare of GM (thus no aggregation needed) the difference would represent the largest potential benefit of GM adoption.

Profitability and profit margin differentials between GM and non-GM white maize
Given that GM crop production is typically associated with greater up-front costs to producers (mainly in seed expense), yield should not be the sole metric considered when evaluating GM white maize producer benefits. Conventional and Bt cultivars were compared in a head-to-head profitability comparison using cost of production, mean yield from the National and North West were used in this analysis due to the availability of their production budgets.
Unlike the LCA, where stacked GM technology was used in which there were no cost of production but detailed input amounts (Grain SA, 2019a), in the profitability analyses we use Grain SA (2020) Bt maize production budget which provides the costs associated with Bt maize but not the input amounts (Grain SA, 2019b, 2019c. Given the National Maize Cultivar Trials test many cultivars (i.e., some are older cultivars used as checks with low yield and no current producer adoption) only the top ten highest yielding conventional and Bt cultivars were compared. Table A6 shows cost of production for conventional and Bt maize for both provinces, which were provided by Grain SA (Grain SA, 2019c, 2019b. Profitability was simulated using @Risk © (Palisade Corp., Ithaca, NY) for conventional and Bt white maize across 18 locations in Free State and 18 locations in North West. From the 1,000 simulation iterations, a two-tailed t-test was used to test for statistical differences between the profitability of conventional and Bt white maize. Two levels of profitability were considered.
First, an estimation of the breakeven percentage producing conventional and Bt white maize was conducted. Second, the relative return on investment for conventional and Bt white maize by location was investigated. Given that Bt production is associated with greater up-front costs to producers, another measure of profitability is the return on investment, which is defined as the cost of production in this study. The profit margin in this sense is defined as profit per unit of cost. Profit margin estimates were obtained by the simulations described above. Thus, this study explores the percentage of times a producer would breakeven or earn a profit as well as the profit margin comparison between conventional and Bt maize for 36 locations across two provinces in South Africa.  Table 1 illustrates the differences in the GM white maize yield premiums across provinces. Estimated yield gains range from a low of 0.370 t/ha in North West to a high of 0.986 t/ha in Gauteng (P < 0.01). The estimated yield gain in the top maize producing province, Free

Estimation of additional annual rations
State, is 0.591 (P < 0.01) (Shew et al., 2021). Table 1 (Table 2). These results are important as they refute, at least in the South African context, an often-cited criticism of GM crops have ambiguous effects on food insecurity (Gurian-Sherman, 2009;Hakim, 2016;Heinemann, 2009;Nature, 2010;UNCTAD, 2013;WHO, 2005). As derived by the summation of additional tons of white maize produced in each province.
d Estimated yield premium associated with GM adoption in province p was statistically different (P < 0.01) from conventional maize yield.
e Calculated as the product of Table A2 and estimated yield premium on Table 1.

Estimation of welfare gains
To quantify changes in producer and consumer surplus attributable to the adoption of GM technology in white maize production in South Africa, a general EDM was employed (Equations 3-10). The resulting changes in consumer surplus, producer surplus, and net surplus (2018 USD  The total net surplus estimates resulting from the sensitivity analysis range from $388.4 million to $905.9 million (2018 USD) with an average total net surplus of $668.9 million (2018 USD) (Table A3). By comparison, the total net surplus as reported in Table 2 was $694.7 million (2018 USD), indicating that our results gravitate towards the average of all scenarios.  (van Zyl, 1986) and the elasticity of supply of 0.36 . The upward shift in supply was set at 7.05% assuming demand held constant. Prices and quantities varied with the year observed but elasticities remained constant (see Equations 3-10). f Maize consumption data for the year 2018 was not available, thus it was assumed maize consumption remained constant from the previous.

Ecosystem impacts of GM maize adoption
The ecosystem impacts of GM maize adoption are presented in the counterfactual argument. That is, we ask, how much more environmental damage would have occurred if GM white maize was not adopted in South Africa from 2001-2018? The main differences between conventional and GM maize production include the yield (i.e., yields associated with conventional dryland maize and GM dryland maize were estimated at 3.998 and 4.421 tons per hectare, respectively) and the pesticides requirements (Table 3). For example, conventional maize production uses seven different herbicides (e.g., glyphosate, atrazine, terbuthylazine, Smetolachor, mesotrione, 2,4-D, and terbuthylazine) while GM maize production solely uses glyphosate, albeit three times as much than conventional maize production. Both conventional and GM maize production use a pyrethroid compound as a means of pest control. Conventional maize production, however, uses about twice as much pyrethroid in a growing season compared to GM maize production. Commercial maize production in South Africa involves more inputs than those listed on Table 3, but all other inputs in maize production are assumed to be identical for conventional and GM maize, implying that the differences in ecosystem impacts would manifest themselves through the different inputs on Table 3.  (Grain SA, 2019a). b Includes pesticides that do not have built unit process in ecoinvent database (Wernet et al., 2016). 497.2 g/ha of terbuthylazine and 20.0 g/ha of mesotrione is accounted for as an unspecified pesticide in the conventional dryland maize unit process.
The functional unit for comparison is one hectare of production under both seed technologies (stacked GM and non-GM). Table 4 presents the results associated with each Stepwise impact category. The total cost reflects the sum of all impact categories in terms of monetary cost (2018 USD). The total costs for one hectare of conventional dryland white maize production and GM dryland white maize production are estimated at $9.11 (2018 USD) and $8.77 (2018 USD), respectively. The total costs can be viewed as the costs to mitigate the environmental damage associated with the production of a hectare of conventional and GM white maize in South Africa, respectively. Respiratory inorganics and effects associated with global warming from fossil fuels are the major contributors to the economic costs associated with both conventional and GM white maize production in South Africa. All other environmental costs combined accounted for under nine and ten percent of the total costs for conventional and GM white maize, respectively (Table 4).  (Grain SA, 2020). The yield for GM dryland white maize was estimated as 4.421 tons per hectare and was calculated as the sum of the mean yield for conventional dryland white maize (3.998 tons per hectare) (Grain SA, 2020) and the estimated yield premium for GM dryland maize (0.423 tons per hectare) (Shew et al., 2021).
To estimate the ecosystem externalities attributed to GM white maize adoption, the difference between the conventional dryland white maize total cost and the GM dryland white maize total cost, $0.34 (2018 USD), was multiplied by the area of GM white maize in South Africa annually (Equation 11). in the net impacts of GM white maize adoption (Equation 12). As indicated in Table 5, the total net impacts of GM maize adoption for 2001-2018 was $700 million (USD 2018). On average, the net impacts of GM maize adoption were $38.9 million (USD 2018). While $292,282 (USD 2018) in annual environmental benefits is marginal compared to the revenue gains, these findings are contrary to the claim that GM crops are detrimental to the environment (Hakim, 2016;Heinemann, 2009;UNCTAD, 2013).

Profitability and profit margin differentials between GM and non-GM white maize
Given that GM white maize production is associated with greater up-front costs to producers, yield gains alone are not a sufficient metric when evaluating the producer benefits of GM maize adoption. The mean yield and yield variance for conventional and Bt cultivars in 18 locations in Free State and North West were derived from Shew et al. (2021). Only the top ten yielding dryland conventional and dryland Bt maize cultivars for 2000-2017 (Table A5).     Table A5 assuming an average price of 147.12 (2019 USD) and average total cost of 17.06 (2019 USD) for Free State and 14.17 (2019 USD) for North West simulated 1000 times using @Risk. b Breakeven percentage for Bt maize cultivars in location l was statistically different (P < 0.05) from conventional maize cultivars. y The relative profit margin for Bt maize cultivars in location l was statistically different (P < 0.05) from conventional maize cultivars.
In Free State, Bt adopters, on average, have a greater relative profit margin compared to their conventional counterparts at 27.96% and 21.27%, respectively (P < 0.05). These results suggest, on average, Bt cultivars return 0.07 Rand and more profit for each Rand invested than conventional cultivars return in Free State. In North West, it was found that there is not a statistical difference (P > 0.1) between the relative profit margin of Bt and conventional maize production ( Table 6). The results presented in Table 6 suggest that the higher upfront costs associated with Bt white maize are offset by the ability for producers to breakeven more frequently, and for producers in Free State, offset by higher relative profit margins. These findings are contrary to the frequent criticism of GM crops that producers' higher yields are offset by higher input costs (Heinemann, 2009).

Conclusions
Contributing authors: Lawton Lanier Nalley, Aaron M. Shew, Jesse B. Tack, Petronella Chaminuka, Marty D. Matlock, and Marijke D'Haese Despite South Africa's upper-middle-income country classification, food insecurity is an ongoing concern for a large segment of the population, as evident from 2014-2015 when over a fourth of households experienced food insecurity due to severe drought and subsequent food price shocks (STATSA, 2016). Concerns surrounding food security are also amplified by the threat of climate change and its subsequent effects on sub-Saharan agricultural production and maize production particularly. Given the present and future concerns, producers, agricultural scientists, and policymakers alike are considering a wide range of options to reduce the present, as well as mitigate future food insecurity in the region, including GM technology adoption.
Three of the most common criticisms of GM adoption are that GM crops do not increase the food supply, do not make producers more profitable, and do not reduce the environmental impact of agricultural production. Using a combined economic (province-level yield benefits of GM and adoption rates) and environmental (LCA) approach, we estimate the total welfare benefits attributable to GM white maize adoption in South Africa for 2001-2018 are $694.7 million. Food security benefits attributable to GM white maize also manifest through an average of 4.6 million additional rations annually. To achieve these additional rations using conventional maize, the production area in South Africa would have to increase by up to 217,788 hectares.
The LCA results indicate that GM maize reduces environmental damage by $0.34 per hectare or $291,721 annually, compared to conventional hybrid white maize. Our analysis of producer profitability focuses on the main production regions in the North West and Free State provinces, and we find that GM hectares breakeven more often than non-GM hectares. Given that GM is often associated with higher upfront costs, relative profitability was also compared, and we find that GM adopters in Free State, but not in North West, had higher relative profit margins. While the results of this study indicate that GM maize adoption in South Africa can increase maize rations, it is naive to think that increasing food supply is the only element of food security.
Markets, incomes, purchasing power, and international trade all factors in food security. While this paper only analyzes one aspect of food security, its results indicate that adoption of GM maize in South Africa has contributed to additional maize supply which may have improved local and regional food security.
Unlike previous studies, we focus on one of the only field-to-plate GM crops, which has direct food security implications. Studies such as this provide important information for consumers, producers, NGOs, and agricultural policy makers about what GM crops can and cannot (e.g., completely alleviate food insecurity) achieve in South Africa. As we face a hotter and drier future, agricultural technologies such as GM may be one of the most salient ways to combat food insecurity while simultaneously reducing the environmental impact of agricultural production. Without metrics and effective communication about what and who the benefits and benefactors are, public confidence and trust surrounding GM technology is likely to remain low.
Global food security necessitates an interdisciplinary approach among economic, scientific, and technical disciplines as demonstrated in this study.  Parenthesis values represent the percentage of area devoted to growing white maize out of the total area grown (white and yellow maize).    As estimated from (Shew et al., 2021). Production budgets obtained from (Grain SA, 2020).