1. Background on GM Approval Process

Even if a GM event developer submits an approval request at the same time in different countries, it is very unlikely that a new GM event is approved simultaneously in all countries. This circumstance, defined as AA, is the outcome of dissimilar approval processes across countries (Stein and Rodríguez-Cerezo, Reference Stein and Rodríguez-Cerezo2010). Thus, AA depends on the politically optimal approval process of both importer and exporter.

Many situations may arise when comparing the GM approval processes between countries. The pace at which these processes take place depends mostly on political and economic factors. Political determinants can impact the environment in which GM approval processes occur within each country and thus contribute to AA. The quality of regulation can be a powerful political force in determining the magnitude of the AA, particularly in cases where countries have similar regulatory frameworks, and given that the quality of a regulation can reduce the overall time required for obtaining GM event approval.

The introduction of GM crops has driven countries to adopt domestic regulations concerning the production and consumption of food and feed products originating from these technologies to ensure their safe use for the environment and for human and animal health. The first regions to establish GM regulations – the United States and the European Union – followed different scientific approaches to GM products. While the EU invoked the precautionary principle and consumers’ ‘right to know’, the US approach to regulating GM products was based on the principle of substantial equivalence, where regulation should focus on the nature of the products, rather than on the process by which they are produced (Gruère, Reference Gruère2006).

Gruère (Reference Gruère2006) characterized eight groups of countries according to their regulatory framework, where one of the eight groups was GM product-free countries. Vigani and Olper (Reference Vigani and Olper2013) characterized 15 clusters based on an index from Vigani et al. (Reference Vigani, Raimondi and Olper2012), which takes into account six dimensions of the GM regulatory framework.Footnote ² Faria and Wieck (Reference Faria and Wieck2016) divided countries into three clusters based on national GM regulation components. That is, the initial differences in scientific approach between the EU and the US with respect to the GM regulation process have persisted over the years, and have become a benchmark for many countries (Vigani and Olper, Reference Vigani and Olper2015). These differences are themselves a potential source of AA. In addition, the approval process may be influenced by other issues within each country, and therefore heterogeneity across countries becomes the main potential source of asynchrony. Figure 1 presents some basic issues that may influence the GM approval process.

Figure 1.

Major issues influencing regulation for the GM approval process

Source: Own compilation.

The production process is directly impacted by GM technology that increases productivity and/or decreases production costs. Thus, once a new technology shows potential to increase farmer profits, it is expected that farmers become willing to adopt it and support more lax regulations regarding that particular technology. However, this behavior may differ across countries, depending on their comparative advantage or disadvantage in producing that crop. Farmers in countries with a comparative advantage are expected to have an interest in adopting the GM technology once it is proven to increase production and export supply. Research on the determinants of GM regulation supports the view that trade and comparative advantage influence GM regulatory policies. Additionally, there is evidence that differences in comparative advantage in GM production may explain the lax regulation of GM products in the US and stringent GM regulation in the EU (Anderson et al., Reference Anderson, Damania and Jackson2004) .

A rational decision in an importing country with a comparative disadvantage in GM production would be for farmers to lobby against GM crops, or at least not support their adoption (Anderson et al., Reference Anderson, Damania and Jackson2004). According to Curtis et al. (Reference Curtis, McCluskey and Swinnen2008), countries with fewer comparative advantages in agricultural production might use GM regulation as a protectionist measure, and thus adopt more stringent regulatory standards. In the EU moratorium scenario at the end of 1990s, Anderson et al. (Reference Anderson, Damania and Jackson2004) found that food-importing countries had an interest in foregoing GM product adoption in order to avoid increasing their comparative disadvantage to the exporting countries and to protect their own production.

In certain situations, the decision to adopt GM technology is based on the balance between potential gains and losses. Tothova and Oehmke (Reference Tothova and Oehmke2004) and Vigani and Olper (Reference Vigani and Olper2013) found that countries of similar bargaining power may synchronize their standards, if there are sufficient gains from trade. Regions with the most stringent requirements (the EU and Japan) encourage exporting partners to adopt similar regulation. Specifically, countries interested in exporting to the EU and Japan may have to comply with the restrictive GM regulations of these markets and synchronize their regulation with those of the EU and Japan if they want to maintain their market share in these regions (Gruère, Reference Gruère2006).

With regard to the demand side, consumer concerns exist around the potential health and environmental risks of GM foods. As food originated from genetically modified organisms have potential long-term risks that have not yet been fully demystified, consumers adopt various positions regarding demand for these products, and the income level of the population within a country can influence these stances. Depending on the consumer preferences and the income level, the price ratio may be sufficient to change the consumer's demand for a potentially risky product. While GM technologies directly impact crop production – decreasing costs and increasing productivity – they benefit consumers through lower food prices, since perceived product quality and characteristics usually seem the same. Thus, the total benefit of GM technology may not be sufficiently perceived by consumers, particularly if they have an established negative opinion regarding GM crops and if the price difference is not significant to the consumer (Vigani, Reference Vigani, Adenle, Morris and Murphy2017). Changes in consumer preferences concerning GM foods have an essential impact on reinforcing the lobbies for or against approval of GM events. Lusk et al. (Reference Lusk, Bruce, House, Valli, Jaeger, Moore and Morrow2006) deny the conventional assumption that consumer preferences are homogeneous by highlighting differences in preference for GM food between the US and the EU.

The media can also influence consumer preferences. Several papers highlight the prominent role of the media in shaping consumer attitudes toward GM technology. According to McCluskey and Swinnen (Reference McCluskey and Swinnen2004), consumers generally prefer to consume negative news because this allows them to avoid negative outcomes and income losses. Media companies choose between their own preferences regarding the stories they publish and their profits, but are partially driven to promote stories that follow consumer preferences. Therefore, potential risks and consumer concerns about biotechnology are likely to be emphasized in the news.

According to Curtis et al. (Reference Curtis, McCluskey and Swinnen2008), people in developing countries hold less negative attitudes against GM food than people in developed countries and at times even hold positive opinions about GM foods. This is partially explained by less exposure of consumers in developing countries to negative news due to the higher opportunity cost of keeping themselves well-informed, in addition to generally less media freedom in these countries.Footnote ³ In developing countries where increasing productivity and reducing food prices are important issues, and where governments may exert some influence over the media, consumers can be less exposed to negative news. In addition, media tends to focus on influential groups within the country. In agricultural-based economies, for example, agricultural innovation is more likely to be viewed more positively than in developed countries where consumers have more purchasing power and are more concerned with food safety issues (Vigani, Reference Vigani, Adenle, Morris and Murphy2017).

Internet access has played an important role in giving consumers access to sources of information other than traditional mass media, despite the fact that much internet news is not trustworthy (Vigani, Reference Vigani, Adenle, Morris and Murphy2017). Literature on the displacement effects of the internet abounds with examples where the internet displaces traditional mass media such as television, newspapers (Dimmick et al., Reference Dimmick, Chen and Li2004), and radio (Lee and Leung, Reference Lee and Leung2008). The time-displacement theory suggests that the use of one particular media may reduce the time spent with others, which is to say that the increase in internet usage may displace that of television, radio, and newspapers (Dutta-Bergman, Reference Dutta-bergman2004; Nie and Hillygus, Reference Nie and Hillygus2002). However, internet users may seek out various sources on the internet and try to form their own opinion with respect to a certain topic, including GM products.

The arguments presented so far show substantial differences in opinion regarding GM products between the two main interest groups of farmers and consumers, and across countries. Farmers and consumers may either fight to achieve their own, opposing interests regarding GM product approval, or work together to influence government with respect to the approval process. The speed of the approval process therefore depends on the interactions among interest groups inside a country and on the overall country sentiment toward GM products. Swinnen and Vandemoortele (Reference Swinnen and Vandemoortele2011) drew a political economy model of food standards to analyse how the government simultaneously considers both the contribution of various lobbies, as well as domestic welfare, to determine a political equilibrium. In this model, consumers and producers are politically organized, and their lobbies contribute to influencing governmental decisions. According to the model, domestic welfare is defined as the sum of producer profits and consumer surplus. The government's objective function depends on the truthful contribution schemes of lobbies, weighted by the lobby's strength and by domestic welfare. Lobbies may either take a third-party view when farmers and consumers have differing opinions (such as the case of an exporting country that has a comparative advantage but consumers there present negative attitudes toward GM products), or lobbies may take a stance in agreement (when an importing country has lower comparative advantage and its consumers also have negative attitudes toward GM products, for example).

An important factor with regard to the GM regulation process is the quality of a country's regulatory authority. On the one hand, Kalaitzandonakes (Reference Kalaitzandonakes2011) suggested that the capacity of regulatory agencies, the accuracy of the information submitted by the applicants for GM event approval, and the speed at which this information is submitted can affect the speed of approving GM events and can ultimately cause AA. Governments can therefore decrease asynchrony by designing schemes to reduce the time lag between domestic and foreign approval. On the other hand, literature has shown that corruption is likely to decrease the effectiveness of regulation, to decrease the stringency of environmental policy, and can also be associated with environmental degradation in developing countries (Fredriksson and Svensson, Reference Fredriksson and Svensson2003; Immordino and Pagano, Reference Immordino and Pagano2003; Wilson and Damania, Reference Wilson and Damania2005). Concerning biotechnology, its opponents perceive that private interests corrupt public regulation (Newell and Glover, Reference Newell and Glover2003); hence, the stringency of the approval process may be lower when corruption is high.

Taking into account the issues discussed above, many situations can arise that would lead to AA. Interactions among variables, and countries’ particular interests can lead to different approval times for GM events. In Figure 1, the lateral dashed arrows illustrate this possibility. Finally, the stringency of the GM event approval process of the importing country is a critical issue in determining the potential impact of AA on international trade. On the one hand, a high level of AA may not be a risk to international trade if importing countries are less restrictive than exporting countries. On the other hand, even a low level of AA could lead to trade disruptions when the importer does not approve the GM events that are already approved by the exporters.

2. Empirical Framework

Figure 2 shows the empirical framework that is based on the literature review. We selected certain explanatory variables as proxies for the determinants of AA, grouped according to their political or economic characteristics. The political determinants are divided into ‘institutions’ and ‘political power’, while the economic determinants consist of ‘trade and comparative advantage’ and ‘media and development’.

In our paper, AA is measured through the protectionism index (PI) developed by Li and Beghin (2014) and calculated by Faria and Wieck (Reference Faria and Wieck2015).Footnote ⁴ The PI is defined on a bilateral basis (country pairs) by comparing the GM approval status for each and every GM event traded globally,Footnote ⁵ and is established by the importing and exporting country. The PI index takes into account both the dissimilarity and the stringency of the GM approval status between the importer and exporter. It ranges from 0 at the lower bound to e ≅ 2.72 at the upper bound. A score of one indicates that importing and exporting countries have the same GM approval status, which means the absence of AA. A score higher than one indicates that the GM approval status of the importing country is on average more stringent than that of the exporting country. Under this circumstance, AA exists and may create trade disruption. Lastly, scores lower than one indicate that the approval status of the importing country is less restrictive than that of the exporting country, resulting in a situation where the existence of AA does not necessarily translate into trade restrictions. The measurement of AA for each importer is obtained by calculating the average of the country pair scores for each importing country, considering all exporters in the sample. Appendix Table 1 shows the average PI index for the importers in our sample.

As highlighted by Faria and Wieck (Reference Faria and Wieck2015, Reference Faria and Wieck2016), calculating the PI indices involves taking certain critical issues into account. First, the regulatory approvals considered in this paper are those for importation, specifically. Second, we consider non-established GM regulation as the most restrictive approval status, since in our sample the majority of countries that lack GM regulation do not approve GM events. This issue translates into higher PI values for countries such as Bangladesh and Thailand, although many such countries import GM products, implying that the non-approval of a GM event is not the same as prohibiting its import. Instead of treating non-established regulation as the most restrictive approval status, another possibility would be to treat non-established regulation as the least restrictive approval status. This would then take into account situations where a lack of regulation is de facto a synonym for complete liberalization. The decision on how to treat non-established regulation is usually driven by the characteristics of the sample (Winchester et al., Reference Winchester, Rau, Goetz, Larue, Otsuki, Shutes, Wieck, Burnquist, Pinto de Souza and de Faria2012; Li and Beghin, 2014). Finally, only commercialized GM events are considered for the index calculation. As a result, the PI values for this calculation are considerably lower than the values found when considering all GM events.

The Worldwide Governance Indicators (WGI) of the World Bank define regulatory quality (RQ) as the ability of the government to formulate and implement policies and regulation that promote private sector development. We assume that the quality of regulations influences the efficiency of the approval process for GM events (i.e., the higher the quality of regulations, the shorter the time necessary to approve new GM events, and the lower the AA). In addition, based on Vigani (Reference Vigani, Adenle, Morris and Murphy2017), one may argue that if the level of the public trust in the ability and competence of the government to formulate and implement policies is high, the public becomes less demanding regarding the approval process.

The WGI also define corruption as the perception of the extent to which public power is exercised for private gain. It ranges from –2.5 (the lowest corruption level) to 2.5 (the highest corruption level). Theoretically, corruption could either decrease the stringency of the approval process, reducing AA, or reduce the efficiency of the approval process, increasing AA.

Data on rural population (Vigani and Olper, Reference Vigani and Olper2013) and employment in agriculture as a share of total employment (Fredriksson and Svensson, Reference Fredriksson and Svensson2003) have been used as proxies for the strength of the agricultural lobby. A non-linear relationship may exist between group size and the importance of votes, and the introduction of the rural population – both linear and squared – may capture the trade-off off between smaller group lobbying activities and the voting weight of the rural population (Vigani and Olper, Reference Vigani and Olper2013). Therefore, the expected effect of the lobby variable on AA is ambiguous.

The export share to the EU and Japan represents trade dependence, when a country has the EU and Japan as its major importers of GM products; i.e., the higher the export share to both markets, the higher the level of AA, as Japan and countries in the EU may establish more stringent approval processes. We assume that if a country has a comparative advantage in the production of cotton, maize, and soybeans, it also has comparative advantage in GM production, since a significant share of cotton, maize, and soybean production is GM. We hypothesize that countries with comparative advantage may exhibit a less stringent GM approval process, and hence may approve GM events faster, lowering the AA. The analysis applied the revealed comparative advantage index (Balassa, Reference Balassa1965) to proxy for comparative advantage in GM production. The Normalized Revealed Comparative Advantage (NRCA) ranges from –1 to 1. The lower bound represents a situation where the importing country does not have a comparative advantage in GM trade, while the upper bound represents a situation where the importing country has a comparative advantage in GM trade.

Although negative news regarding GM products is relatively common on the internet, the user plays an active role in searching among online news and has the opportunity to research information on the benefits of GM foods, whereas television viewers absorb the information that is chosen by the particular news program. As argued by Vigani (Reference Vigani, Adenle, Morris and Murphy2017), internet users are not only consumers of news, but they also have the opportunity to make comments and engage with a given social network. According to the time displacement theory, we expect that the higher the number of internet users, the less the exposure to negative news about biotechnology from the traditional mass media – particularly in developed countries. In developing countries, this behavior could be reversed since the media there can be biased toward the benefits of biotechnology. For example, in developing countries with a large agricultural sector, the media tends to focus on the livelihood of farmers. If the government there is pro-GM products, less exposure to traditional mass media in such countries would make consumers less willing to support GM products.

Additionally, consumers in more developed economies may display significant risk aversion to GM products and have higher demand for quality. Therefore, the level of economic development may positively affect the stringency of GM regulation, and slow the approval of GM events. We use per capita income (in current USD) as a proxy for the level of economic development.

The econometric equation to test the political and economic determinants of AA follows:

(1)$$\eqalign{PI_{k,j,t} = & \, \alpha _j + \beta _1PI_{k,j,t-1} + \beta _2RQ_{\,j,t} + \beta _3Corrup_{\,j,t} + \beta _4RuralPop_{\,j,t} + \beta _5RuralPopsq_{\,j,t} \cr & + \beta _6ExpShareEU\_ \,JPN_{k,j,t-1} + \beta _7NRCA_{k,j,t} + \beta _8IntUsers_{\,j,t} \cr & + \beta _9GDPpc_{\,j,t} + u_{k,j,t}}$$

where k denotes the type of commodity (cotton, maize, and soybeans), j denotes the importing country, t denotes the year, α _j are the unobserved country effects, and β ₁ through β ₉ are the coefficients to estimate. Following Faria and Wieck (Reference Faria and Wieck2015), we consider 40 countries that have a significant share in the international trade of GM cotton, maize, and soybeans.Footnote ⁶ We assume a three-year interval as adequate for evaluating changes in the stock of GM events over time. Thus, we take data for every three years for the period 2000–2015.

Since PI values are higher for countries that lack GM regulation, we verify the robustness of our results by estimating the baseline sample, which includes all the importing countries, and a GM adopter sample that comprises the importing countries that allow imports only of GM products whose events are pre-approved by their competent authorities.

We obtained data on regulatory quality and corruption from the WGI. We used data on rural population as a share of total population, the number of internet users as a share of total population, and per capita income from the World Development Indicators (WDI) of the World Bank. We calculated the PI index as proposed by Faria and Wieck (Reference Faria and Wieck2015). Finally, we calculated the share to the EU and Japan of total exports of commodity k and calculated the normalized revealed comparative advantage (Balassa, Reference Balassa1965) using data from the United Nations Commodity Trade Statistics (UN-COMTRADE).

Serial correlation may be present in our sample, as the AA of year t depends on its previous observation in t − 1. In addition, as noted by Faria and Wieck (Reference Faria and Wieck2015), where the AA of GM events was found to have an impact on international trade, some endogeneity bias may be present in trade-related variables such as export share to the EU and Japan.

We use the generalized method of moments (GMM) estimator of Arellano and Bover (Reference Arellano and Bover1995) and Blundell and Bond (Reference Blundell and Bond1998). The system GMM estimator is based on the instrumental variables method. Specifically, it associates a basic set of equations in first differences with appropriately lagged levels as instruments, with another set of equations in levels with appropriately lagged first differences as instruments, and thus depends on the validity of the moment conditions. The validity of these additional instruments can be tested using the Sargan test of over-identifying restrictions or by using Difference Sargan or Hausman comparisons between the first-differenced and the system GMM results (Bond et al., Reference Bond, Hoeffler and Temple2001). Under the null hypothesis of the above tests, the instruments are valid.

The consistency of the system GMM estimator depends on not having a second-order serial correlation in the residuals of first-differenced equations. It can be checked by the Arellano-Bond test for AR(1) and AR(2) in first differences.

To assess potential simultaneous causality bias between trade-related variables and the AA, the export share to the EU and Japan is assumed to be endogenous, which means that the export share may be a determinant of AA and in turn, AA may affect the export share.