3.1. Overview of Crops, Modifications and Countries
The crop most commonly studied in the papers reviewed was cotton (43 studies), followed by maize (20) and soybean (14). This differs from the global ranking of GM crops in terms of hectares planted, where soy bean dominates, followed by maize, cotton and rapeseed [39
]. This discrepancy may be because India appeared most frequently in the dataset (Table 2
) and cotton is the only GM plant commercialised in India [39
]. Rapeseed appeared rarely in the dataset. It is only planted in the US, Canada, Australia and Chile, countries rarely studied in the literature reviewed (Table 2
). Nineteen publications focused on other crops (e.g., aubergine, papaya, rice, rapeseed or wheat) and 32 did not specify a crop.
The most common modification to be studied was insect resistance through Bt (53 studies), which is the second most common modification globally. Herbicide tolerance (HT), the most widespread genetic modification in terms of hectares planted [39
], was the focus of 20 studies; 11 studies addressed other modifications (e.g., drought resistance or nutritional enhancement) and 37 studies did not specify the modification. Many of the studies that did not specify plant or modification dealt with regulations or trade agreements regarding GM crops, where specific crops or modifications are not the focus of attention (some examples include [44
Pooling modifications and crops appearing in the papers, Bt and cotton was the most common combination (43 studies), followed by Bt and maize (15), HT and cotton (12), Ht and maize (12) and HT and soybean (11). Scrutiny of the papers confirmed that Bt cotton was the GM crop most commonly addressed. However, the results from pooling data on modification and crop must be treated with caution, as each paper was analysed for crop and modification separately and many papers included several crops and modifications.
Seventeen different countries were mentioned in the studies (Table 2
). The majority of the studies (72/99) concerned countries in the Global South. Of these, 50 related to specific countries and 22 to the Global South in general. Only four studies focused on the Global North, all referring to specific countries (Australia, Canada, USA and Switzerland), 15 studies discussed both the Global North and Global South and eight studies did not discuss specific locations.
Countries planting the most GM crops in the world compared with countries appearing in the dataset.
Countries planting the most GM crops in the world compared with countries appearing in the dataset.
|Ranking of countries based on million hectares of GM crops planted [
39]||Ranking of countries based on appearance in the dataset (number of studies)|
|USA (73.1)||India (17)|
|Brazil (42.2)||South Africa (11)|
|Argentina (24.3)||Argentina (6)|
|India, Canada (11.6)||Philippines (4)|
|China, Paraguay (3.9)||Brazil, Burkina Faso, China, Pakistan (3)|
|Pakistan (2.9)||Australia, Mexico, USA, Canada (2)|
|South Africa (2.7)||Chile, Ethiopia, Nigeria, Paraguay, Switzerland (1)|
|Bolivia, Philippines, Australia, Burkina Faso, Burma, Mexico, Spain, Colombia, Sudan, Honduras, Chile, Portugal, Cuba, Czech Republic, Romania, Slovakia, Costa Rica, Bangladesh (1.0 and lower, descending order)|
All countries appearing repeatedly in the dataset (Table 2
) were countries where GM crops are planted commercially, indicating that the majority of studies presented empirical evidence. However, this was often indirect evidence, as seen in the number of studies without farm-level data (Table 3
). Ethiopia, Nigeria and Switzerland, which appeared in one study each in the dataset, do not have commercial plantations of GM crops [39
]. Surprisingly, the US, the country with the largest area planted to GM crops, only appeared in two studies in the dataset.
3.3. How Are Different Social Impacts Addressed?
Careful scrutiny of the papers provided a more in-depth understanding of the ways in which different social impacts were addressed. Economic impacts, by far the most commonly addressed group, almost invariably included studies on yield and farm finances. The combined results from such studies taken together indicated that on average, the GM crops available today have raised yields and improved farm finances (as highlighted in studies such as [53
]). However, a significant number of studies discussed yield and farm finances without own bottom-up data (52/83 studies). This meant that existing empirical data on yield and farm finances were repeated in many studies (e.g., [14
]). This could give the impression that claims of increased yields and profits have a more substantial empirical foundation than is actually the case.
Some economic studies without farm-level data assessed the kind of economic impacts that can be discussed without farm-level data. For example, the study by Parfitt [46
] (which was also classified as addressing access and distribution impact groups) studied the effects of the increasing economic concentration within the biotech industry and concluded that this leads to crop varieties not being developed for more marginal environments, higher crop and input prices, and increased difficulty for smaller farmers surviving on only farming. Dowd-Uribe [60
] studied the economic implications of the governance of Bt cotton production in Burkina Faso and concluded that high seed prices, corruption and late payments are important in dissuading farmers from producing cotton.
Distributional impacts were the second most common impact addressed. The type of distributional impacts covered was more varied than in the economic impact group and the studies concerned presented a less conclusive picture than the economic studies. At the farm level, the distributional impacts studied concerned mainly how adoption and economic benefits are distributed among farmers in different income groups or, more rarely, in relation to gender. Mutuc et al.
] presented results from the Philippines showing that poorer farmers without access to irrigation and living further away from formal seed suppliers are less inclined to adopt Bt maize. However, in another study from the Philippines, Sanglestsawai et al.
] showed that if they can adopt Bt maize, lower-yielding farmers, who also tend to be the poorest, experience higher yield increases than farmers with higher yields before Bt maize adoption. Distributional impacts with regard to Bt cotton (see Section 3.4
) support this inconclusive picture, indicating that the institutional environment in which GM crops are introduced has significant effects on their distributional effects. Indeed, Newell and Mackenzie [44
] argue that the effects of biotechnology in itself on poverty and food security are limited, and that it is the institutional framework in which the technology is introduced that largely determines its distributional effects.
Studies addressing distributional impacts between farmers and other actors included e.g., Rao and Dev [55
], who point out that, as a result of different political and regulatory contexts, there are significant differences between countries regarding the share of benefits distributed to farmers. Those authors present figures from 2005 showing that distribution of economic benefits between agrochemical companies and farmers adopting Bt cotton in all countries in the example (China, India, Mexico, South Africa) except Argentina is to the benefit of farmers. Chinese farmers benefited the most, taking 94% of the benefits but, at the other end of the spectrum, Argentinian farmers received only 21% [43
]. In the original source of these figures, farmer gains are calculated as the balance between additional input costs, determined by the price premium paid for Bt cotton seed and whether or not farmers can recycle seed or have to purchase new seed every year, and gains on the output side in the form of increased yields and savings in production costs due to reduced pesticide applications (cotton price to consumers is assumed to be constant in the model). Farmer gains are compared with the size of profits that the private sector can extract from their technologies in any given country, which is expected to depend on the strength of IPR protection and/or the availability of other measures to prevent farmers’ own reproduction of seeds [63
]. While the figures presented in Rao and Dev might well have changed since 2005, it can be noted that many articles in the review point out that agricultural policies and regulations surrounding GM crops have a significant influence on whether and how different farmers might benefit from GM crop adoption (e.g., [44
]). The fact that Argentina provides such a negative institutional environment for farmers with regard to GM crops may explain why all studies (6 in total) addressing distributional aspects in Argentina concluded that adoption of GM crops have led to increased inequality in the Argentinean agricultural sector [51
Rao and Dev [55
] and Carpenter [56
] concluded that, looking at the global average, farmers receive a significant share of the profits. As shown above, this does not mean that all farmers benefit equally. In fact, Rao and Dev [55
] noted that, even though more resource-constrained and marginalised smallholders have also benefited from GM crops, the fact that a few multinational companies control the majority of crop biotech R&D has resulted in products being biased towards the needs of large, capital-intensive farms. Several other studies reviewed also acknowledged that the global dominance of private industry within biotech R&D reduces the potential benefits for smaller and more marginalised farmers [46
]. Newell and Mackenzie [44
] study the role that international organisations such as the World TradeOrganization (WTO) and the Cartagena Protocol on Biosafety (CBP) have on how biotechnology is governed in the Global South and show how these organisations have not been able to provide a negotiation climate that secures the interests of countries in the Global South. The authors give many examples of where wealthier states and private interests have successfully pressured developing nations to change their biosafety and patent legislations to be more in line with the interests of the stronger parties. The authors argue that this inequality in the global governing of biotechnology might lead countries in the Global South to resent GM crops as such, without sufficiently considering the potential advantages of the technology.
Only 22% of the studies on access were based on farm-level data. However, such studies often addressed the effects of private industry dominance of the biotechnology sector, IPR or trade agreements, which may be studied in relevant ways drawing on national, regional or global level data (e.g., [44
]). Studies on access generally showed how the system in which GM crops are introduced, including the global private sector dominance, and IPR associated with GM crops in some countries, leads to limited access for certain groups of farmers or certain countries (e.g., [44
]). The study by Parfitt [46
] can be mentioned as one example of many pointing out the mutual shaping of private sector dominance and the IP system. This body of literature described how the way in which IPR are attached to GM crops in some jurisdictions facilitates corporate concentration in the seed industry, which in turn leads to a more limited choice of seed in the shops and high seed prices, and makes it illegal for farmers to recycle seed. Developing countries in general and small-scale farmers in particular are cited as losers in this, as the industry directs its production to capital-strong buyers, i.e.
, large farmers in the Global North [46
]. However, there are different ways of governing biotechnology in different jurisdictions. Parfitt [46
] mentioned India as a country where the state has retained extensive protection of farmers’ rights to seed. Rao and Dev [55
] confirmed this, but suggest that Chinese farmers receive an even larger share of the benefits than farmers in India, as IPR are less strictly enforced in China. However, Rao and Dev [55
], like Parfitt [46
], pointed out that even if GM seed in India and China is comparatively cheap and accessible to smallholders, the private sector dominance of the industry has so far resulted in there being very limited attention to crops and traits suitable for smallholders and more marginal environments. Thus, many studies mentioned a need for increased public sector investment and public-private partnerships to ensure that smallholders and the poorer sections of societies can access GM crops, and that the crops and modifications are relevant to them [44
]. Improving access to unbiased agricultural advice through public investment was also highlighted as essential in allowing smallholders to benefit from a GM crop like Bt cotton [55
]. Such investments could also reduce the negative ecological effects of GM crops, with potential secondary social impacts (as discussed by [55
Studies addressing wellbeing often focused on changes in exposure to toxins, allergenicity and improved nutrition due to changed diets or improved household income. Such aspects essentially need to be studied at individual or farm level. The many mentions of wellbeing in studies without farm-level data (37/48 studies) thus either relied on the limited number of empirically-based studies available or provided a more general discussion, or merely express future hopes or concerns, about potential wellbeing effects of GM crops (e.g., [49
]). Eight of the 11 bottom-up studies connected wellbeing with empirically observed or farmers’ experiences of effects on health. Seven of these related their empirical results to Bt cotton. The exception was Huang et al.
], studying IR rice. The results presented are contradictory, indicating that the empirical base is still too small to draw any generalisable conclusions. Six studies indicated various positive effects on wellbeing for farmers. Two of these six studies provided first-hand data relating to wellbeing: Bennett et al.
] reported declining hospital visits as a result of reduced pesticide use with Bt cotton in South Africa; Qaim and Kouser [82
] found a correlation between improved caloric intake and increased household incomes resulting from the introduction of Bt cotton in India. The remaining four studies reporting positive wellbeing impacts did not measure impacts directly, but relied on farmers reporting such impacts. These studies reported that farmers in Australia, India and China experience reduced health problems as a result of declining pesticide use with Bt cotton and rice [80
]. In contrast, studies by Bennett et al.
] and Debyani and Neeta [87
] from South Africa and India reported negative health implications for farmers as a result of adoption of Bt cotton, due to increased toxic exposure and allergic reactions, respectively. Both these studies made the connection between their data and wellbeing, without measuring wellbeing impacts directly. Bennett et al
] connected pesticide exposure to health effects without measuring health effects as such, and Debyani and Neeta [87
] reported that farmers have claimed allergic reactions, without independently testing whether such reactions are connected with exposure to Bt cotton. The three remaining bottom-up studies did not connect wellbeing effects with empirical observations at farm or household level. Krishna and Qaim [88
] predicted potential reductions in health treatment costs resulting from reductions in insecticide exposure with adoption of Bt aubergine in India. Botta et al.
] generally discussed the potential health effects of GM crops and glyphosate and also reported unconfirmed connections between ill-health and glyphosate sprayings on HT soybean in Argentina. Newell and Mackenzie [44
] examined the possibility to address health concerns in the regulations surrounding GM crops.
Cultural heritage was the impact group least frequently studied in the papers reviewed. Overall, this topic appeared in 15 studies, of which only four [44
] were based on their own data. While in part relying on farm-level data, Newell and Mackenzie [44
] did not discuss cultural heritage impacts in relation to these farm-level data, leaving only three studies actually addressing cultural heritage impacts empirically. They presented results regarding the effects of adoption of GM crops on local/traditional farming systems, knowledge and crop varieties. Stone [76
] performed an empirically detailed case study of how introduction of Bt cotton in India is accelerating the current process of agricultural de-skilling. Rapid technological change, combined with lack of unbiased agricultural advice and widespread availability of cotton seed of unclear and sometimes dubious quality, has resulted in farmers being unable to use their agricultural knowledge to choose cotton varieties suited to their farming environment. The result is loss of farmers’ agricultural knowledge. Hall et al.
] concluded, based on empirical evidence from Brazil, that export-orientated soybean farmers have been able to integrate illegally imported and bred HT soybean with their existing farming practices and benefit from the new technology. They speculated that the benefits would not be the same if farmers had to pay the legally binding royalties to Monsanto. Furthermore, more marginalised farmers in the country, who have not yet adopted HT soybean, were not expected to benefit due to their limited access to education and capital. Factors such as agricultural knowledge, social conditions and adaptability were reported to be relevant for determining the ability of farmers to successfully adopt GM crops and integrate them into their farming system.
It is likely that the overall empirical dominance of the Global South in the literature reviewed here affected how social impacts were addressed. Looking in more detail at four studies in the dataset with an empirical and/or theoretical focus on the Global North [85
], one of these, although empirically based in Hawaii, in fact only discussed implications for adoption in the Global South [91
]. Thus only three out of the 99 studies reviewed focused exclusively on empirical settings in the Global North. Clearly this makes it impossible to make statistically sound statements about whether the way in which social impacts are discussed is affected by the focus on the Global South or North. However, it can be interesting to show what these studies actually address. Andrée [90
] used two biotech products (bovine growth hormone and HT wheat) in Canada to examine the interplay between civil society, state and industrial actors in shaping biotech outcomes. The conclusion was that despite its monopolistic power and significant support from the state, the biotech industry is sometimes forced to compromise its intentions significantly due to strong civil society and academic counter-powers. Russell [85
] used a case study of GM cotton farming in Australia to discuss how the outcome of GM technology depends on the local social context, while Speiser el al
] made an ex-ante assessment of different economic and ecological risks and benefits of introducing a number of GM crops into Swiss agriculture. A key conclusion reached by both Speiser et al.
] and Russell [85
] is that the interplay between type of plant, modification and socio-ecological context results in different outcomes in terms of risk and sustainability.
3.4. The Example of Bt Cotton
More in-depth analysis of the discussion around Bt cotton, the most commonly addressed GM crop in the studies reviewed, provides some additional insights into how social impacts are currently studied and discussed. Cotton is a crop strongly affected by pests and thus IR cotton has the potential to significantly lower pesticide inputs in areas with high pesticide use and to increase yields in areas with less comprehensive spraying regimes, by reducing crop losses to pests. Higher yields and/or reduced pesticide use can then have secondary social impacts such as improved household finances, better labour conditions and health improvements. Looking specifically at studies of Bt cotton relying on own farm-level data, there were 23 such studies in the dataset [53
]. The majority (12) were from India, followed by South Africa (5), Argentina (2), Burkina Faso (2), Mexico (1) and Australia (1) as the only country representing the Global North. (With regard to the countries represented in these studies, it should be noted that Pechlaner and Otero [70
] presented data from three countries, Canada, Mexico and the USA, but Bt cotton was only mentioned for Mexico and therefore only Mexico is mentioned here. In the overall classification of studies, Ezezika et al.
] was classified as not providing data from a specific country, but rather a region (sub-Saharan Africa). Thus when numbers for the whole dataset are presented, this study is included as not focusing on a specific country. However, with regard to Bt cotton, Ezezika et al.
] used interviews made in Burkina Faso and is therefore counted here as providing data on Bt cotton from Burkina Faso. As in the total dataset, economic impacts were most frequently addressed (20/23 studies), followed by distributional impacts (12), wellbeing and access (7) and cultural heritage (2).
The unanimous conclusion from bottom-up studies of Bt cotton, focusing on economic impact, was that on average, it provides economic benefits for farmers. These economic benefits were described as combinations of direct and secondary impacts of raised yields and/or reduced costs for pesticides [53
]. A commonly cited example of secondary impacts related to changes in labour demand. Some studies pointed out reductions in labour as a cost-saving effect [81
], or as a negative effect for those previously employed for spraying [85
]. Labour needs were also shown to increase in some cases. For example, Subramanian and Qaim [101
] showed that increased labour needs, resulting from higher harvests with adoption of Bt cotton, were positive for women in their study due to increased employment opportunities.
Regarding impacts of access and distribution, the results are inconclusive. Subramanian and Qaim [100
] presented results from India showing that farmers across income levels benefit from adopting Bt cotton. However, Subramanian and Qaim [100
] also showed that economic benefits are greater for larger farms. On the other hand, some South African studies showed that smaller-scale farmers do not benefit less than large producers [81
]. There were also reports from South Africa and India that smallholders obtain greater economic benefits from growing Bt cotton than farmers with more land [77
]. In contrast, Arza et al.
] and van Zwanenberg and Arza [67
] reported negative effects on distribution with the adoption of Bt cotton, due to the co-evolution of Bt cotton in Argentina with a political and regulatory environment favouring large farms. Large, capital-intensive farms have the possibility to control larger parts of the highly industrialised cotton production chain and can therefore secure higher returns, whereas smallholders become increasingly reliant on middlemen with the introduction of Bt cotton, reducing their benefits. Pechlaner and Otero [70
] describe how the Mexican economic and regulatory environment has similar negative effects for smallholders, making them particularly negative to GM crops. Russell [85
] showed how Australian farmers, although largely positive to the environmental benefits of Bt cotton, have expressed concerns about the monopolistic seed industry and the risks this poses through reduced variety in supply and pricing of GM products [85
]. Based on evidence from Burkina Faso, Ezesika et al.
] concluded that, as women are excluded from post-harvest sales of Bt cotton, they have limited possibilities to control the benefits of increased harvests. In summary, the accumulated evidence on Bt cotton indicates that a key reason why economic distribution and access impacts on the ground vary with context is that outcomes on the ground are largely affected by different political and regulatory contexts.
Studies from Argentina and India make it relevant to question whether the economic benefits of GM crops will persist over time [66
]. A key reason for this is the widespread lack of unbiased, accessible and sufficiently detailed agricultural advice in both countries, resulting in farmers not knowing how to cultivate Bt cotton in a way that ensures good yield and delays insect resistance. Stone [76
] presented results from Warangal in India, where the overall income gains with Bt cotton might be undermined by the widespread lack of understanding and practical implementation of refugia amongst farmers and retailers (the main source of information on Bt cotton for farmers in the region). Refugia are parts of Bt cotton fields planted with conventional hybrids for the sake of preventing or delaying resistance to the Bt toxins in target insects. In addition to the inadequacy of available agricultural advice, in both Argentina and India uncertified Bt cotton seed of dubious quality is circulating on the market. The outcomes of this are farmers planting seed that might not perform as expected and an increased risk of insect resistance development [76
]. In contrast, Russell [85
] presented evidence from Australia, where Bt cotton has been successfully adopted as part of a wider Integrated Pest Management (IPM) strategy. This indicates the importance of introducing Bt cotton in a functional agricultural advisory context, and the potential gains if this is done.
Shankar et al.
] examined the role of production risk in determining the economic impacts of Bt cotton, an aspect which they argued is rarely investigated. Studying Bt cotton production in Makathini flats, South Africa, Shankar et al.
] concluded that Bt cotton performs well when agricultural conditions are good, but that there are indications that under certain, less optimal agricultural conditions, production risk increases with Bt cotton. They also concluded that risk effects with Bt cotton are complex and difficult to predict and that more investigations into production risk are needed.
As for the whole dataset, the seven empirical studies addressing wellbeing gave an overall inconclusive picture (see Section 3.3