5. New evidence on market concentration
Using privately-held market research data, this chapter presents new estimates of concentration in seed markets covering a broad range of crops and countries, and analyses the determinants of market concentration levels in seed. In addition, this chapter provides evidence of “multimarket contact”, where the same firms face each other as competitors in several markets. The chapter concludes with a discussion of the available evidence on concentration in the market for GM technology.
Researchers working on market concentration in seed have often noted the difficulty of obtaining market share data at a sufficiently disaggregated level (Fernandez-Cornejo and Just (2007[6]), Mammana (2014[66])). For want of more detailed data, some analysts have used aggregate sales figures of firms and an estimate of the overall size of the global seed market to obtain estimates of global market concentration. Using this approach, Fuglie et al. (2011[9]) estimated that the combined market share of the four largest firms increased globally from 21% in 1994 to 54% in 2009, while ETC Group (2013[8]) puts this figure at 58% for 2011.
These aggregated approaches suffer from several shortcomings. First, as noted by Bonny ( (2014[25]), (2017[12])), estimates of the value of global seed sales tend to vary across different sources and methods, and some market research agencies tend to underestimate sales by small and medium-size enterprises. For instance, an estimate by Phillips McDougall of USD 35 billion in 2015 only appears to account for two-thirds of crops used in global agriculture. Estimates may differ in how they account for farm-saved seed; some estimates may only focus on commercial seed sales. When market concentration measures are calculated using such low estimates of total market size, the degree of market concentration is automatically overstated.
Moreover, any estimate of market concentration using global sales figures will automatically assign a greater weight to markets with a higher seed price. Correcting for differences in prices (for instance by using volume shares instead) can lead to very different results. To illustrate this, Figure 5.1 presents a rough estimate of the share of different geographies in global maize markets. The United States accounts for an estimated 40% of global maize markets in value terms (measured at average 2016 exchange rates). However, when measured in volume terms, the estimated share of the United States falls to below one-quarter of the global total, at about the same level as the People’s Republic of China (hereafter “China”). It is clear that when value-based measures are used estimates of global market concentration in maize seed markets will assign a much greater weight to firms active in the United States.
An approach which aggregates across all countries and crops using sales in value terms may thus be misleading about concentration levels in specific markets. For instance, using data from before the current mergers would typically conclude that Monsanto is the leading firm, followed by DuPont, and with Syngenta a distant third (Ragonnaud (2013[17]), Heisey and Fuglie (2011[7]), ETC Group (2013[8])). However, as noted earlier, about 60% of Monsanto’s sales originate in the United States, and more than 80% come from the Americas. Monsanto’s sales are limited in other regions, most notably Asia Pacific. Likewise, before its merger with Dow, agricultural sales of DuPont were overwhelmingly based in the Americas.1 Calculations of global market shares reflect the disproportionate size and value of seed markets in North and South America (in particular the United States). It is not clear how such global calculations can inform policy makers about market concentration in specific crop seed markets in specific geographies.
Note: Estimated share in global maize markets in value (at average 2016 exchange rates) and volume.
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database.
This chapter takes an alternative approach. Rather than providing global, regional or national estimates of overall market concentration, this chapter provides new estimates of market concentration at the level of specific crops in specific countries. The focus is on major field crops such as maize, soybeans, wheat and barley, rapeseed, sunflower, and to a lesser extent on sugar beet, potato, and cotton. This chapter also provides information on market concentration in GM traits using a variety of sources.
5.1. Seed market data and methodology
The data on market concentration in seed relies on the amis®AgriGlobe® database of the Kleffmann Group. The AgriGlobe database uses annually-recurring farmer surveys, complemented with bi-annual recurring distributor surveys in some markets, as well as additional input from experts. As a result, the country and crop coverage varies depending on the level of detail; in general, more data is available on market size than on market shares of plant breeders. For our analysis, we rely on market share data for 2016 unless noted otherwise. More information regarding the Kleffmann database is provided in Annex 5A.
The Kleffmann database distinguishes between the commercial seed market and the overall seed market, including farm-saved seed (in both cases defined as the market for seed for final use within the country). A choice needs to be made whether to calculate market shares relative to the commercial seed market or to the overall seed market. In what follows, all market shares have been calculated using the overall seed market. The rationale behind this approach is that if seed saving is important, using the market shares of the commercial seed market would overstate the degree of market power. Commercial plant breeders in those markets are in fact competing with farm-saved seed; calculating only their share of the commercial seed market would be misleading.2
Market shares used in the calculations refer to the plant breeder, i.e. the owner of the variety.3 Public sector plant breeding institutes are active in a number of markets. The Kleffmann database treats these in the same way as private sector firms.4 In some cases, more than one public sector institute is active in the same market. These are considered here as different suppliers in calculations of market concentration, as these institutes implicitly compete with each other. In some of the empirical analyses, an additional measure is included to control for the overall importance of public sector plant breeding in a market.
The choice of crops covered is based on data availability. In particular, detailed information on market shares is only available for maize, soybeans, wheat and barley, rapeseed, and sunflower. Additional information for a handful of countries is available for potato, sugar beet, and cotton. Data for some important markets are missing; where possible this has been complemented using estimates from the literature (Table 5.1).
This chapter presents the most detailed and complete estimates to date of market concentration in seed markets across crops and countries. Yet some gaps in coverage remain. The important market for vegetable seed is unfortunately not covered. Moreover, although the geographic coverage includes both OECD and non-OECD economies, much less is known about developing and least developed countries. Boxes 5.2 and 5.3 provide information on vegetable seed markets in Europe and market concentration in developing countries.
In the literature, two measures commonly used as indicators of market concentration are the four-firm concentration ratio (written as C4) and the Hirschman-Herfindahl Index (HHI). The C4 ratio is defined as the combined market share of the four largest firms in a particular market. This measure has the benefit of having an intuitive interpretation, but its main drawback is that it does not capture the relative size of different firms. For instance, a C4 ratio of 80% could reflect a market with four similar-sized firms, each with a market share of around 20%. However, this could equally reflect a market where the leading firm has nearly 80% of the market by itself, with the rest of the market shared among many smaller players.
The Hirschman-Herfindahl Index addresses this shortcoming. The HHI is defined as the sum of squared market shares. If a single firm has a 100% market share, the HHI would have a value of 10 000. By contrast, if the market was equally divided between 100 firms that each had 1% market share, the HHI would be 100; lower values are possible as the market becomes further fragmented. As the HHI is based on squared market shares, the index increases when a larger firm gains market share at the expense of a smaller firm.5
A typical challenge in analysing market concentration is market definition, i.e. correctly delineating the relevant market in which market shares of competitors can be analysed. The data available here implicitly defines the relevant market as the national market for seed of a specific crop.
In some cases, this geographic definition may be too broad, especially in large countries. For example, in the United States, there may be important differences in the maize markets between the central Corn Belt (the region in the Midwest of the United States where most maize production is situated), the neighbouring “fringe corn belt”, and the rest of the country. Agro-ecological characteristics will differ between such regions, meaning that what appears as a single market in the data (as presented in this study) may in reality represent separate markets. For other cases, however, the national definition of a market may be too narrow, especially in small countries. For example, in some EU Member States the agro-ecological conditions may be similar to those of neighbouring countries, in which case the relevant market may span an area greater than the national market.
Moreover, the crop dimension can also be called into question as it could be either too broad or too narrow. For example, the data available here groups together information on wheat and barley; but it could be argued that different types of wheat (e.g. hard red winter, durum) represent different seed markets. On the other hand, it could also be argued that from the point of view of the farmer different crops (e.g. maize, soybean) are to some extent substitutes, so that market concentration should be analysed in the context of a broader market which includes these substitutes.
These questions of market definition are important and central to much of the detailed analysis performed by competition authorities when evaluating mergers. In the context of the present study, the available data do not make it possible to go much beyond the national crop seed market definition, but it is important to keep in mind the limitations mentioned here. Despite these limitations, the data presented constitute a major improvement over what was previously available in the public domain.
5.2. Concentration in seed markets
Maize
Table 5.2 shows concentration measures for maize seed markets in 32 countries using the two most commonly used concentration measures (the Hirschman-Herfindahl Index, HHI, and the four-firm concentration ratio, C4), and using both volume and value measures. Figure 5.2 shows our preferred measure, the value-based HHI (hereafter referred to as the HHI).6 There is a considerable variation in concentration levels across countries in our sample, ranging from 933 in Belarus to almost 4 700 in Denmark. There is no obvious link between concentration and geography, or levels of development.7
There is a remarkable difference between the value-based and volume-based measures of market concentration in the Mexican maize seed market. As the centre of origin of maize, a large share of the seed market by volume (64%) is accounted for by farm-saved seed, with a low imputed value. By contrast, the commercial seed market is dominated by two multinational firms. This “dual” structure of the seed market explains the discrepancy between value- and volume-based measures.
There is a remarkable difference between the value-based and volume-based measures of market concentration in the Mexican maize seed market. As the centre of origin of maize, a large share of the seed market by volume (64%) is accounted for by farm-saved seed, with a low imputed value. By contrast, the commercial seed market is dominated by two multinational firms. This “dual” structure of the seed market explains the discrepancy between value- and volume-based measures.
Note: HHI is the Hirschman-Herfindahl Index. All calculations refer to the shares in the overall seed market (including farm-saved seed) in value terms.
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database.
Soybeans
The database covers a much smaller number of countries for soybean seed (seven compared with 32 for maize). However, the countries covered here jointly account for an estimated 86% of the global soybean seed market by value. Data on market concentration are provided in Table 5.3 and Figure 5.3 Market concentration is lowest in the Ukrainian soybean seed market, and highest in Uruguay and South Africa. The highest observed concentration level for soybean seed is lower than that for maize markets. GM varieties for soybean are widely used in these countries. In Ukraine, the use of GM soybean is not permitted but glyphosate-tolerant varieties appear to be in use (Kleffmann Group, 2016[135]).
Compared with the other countries included in Table 5.3, a much greater share of soybean seeds in Ukraine derives from public research institutes and/or unknown sources (possibly farm-saved seed). Moreover, since several public research institutes are active, the market is quite fragmented.
Note: HHI is the Hirschman-Herfindahl Index. All calculations refer to the shares in the overall seed market (including farm-saved seed).
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database.
Wheat and barley
Public institutes play a greater role for wheat and barley as compared to other crops.8 They are important sources of cereal varieties in Bulgaria, Latvia, Mexico, Romania, the Russian Federation, Turkey, and Ukraine where they contribute 20% or more to total volumes. In addition, a large share of wheat and barley seed by volume could not be allocated to a plant breeder in several markets, indicating a reliance on farm-saved seed. In Romania, Latvia, Hungary and Poland this share was between 14% and 39% of the overall market by volume. Together, the presence of farm-saved seed and the role of public institutes imply that the share of seed accounted for by the commercial sector varies considerably between countries, as shown in Figure 5.4 for the 15 countries for which data is available in the Kleffmann database.
The commercial market dominates in the United Kingdom, Denmark, Austria and Slovakia. By contrast, the role of the commercial sector is relatively small in the Russian Federation, Mexico and Romania, among other countries.
Table 5.4 and Figure 5.5 present estimates of market concentration in wheat and barley seed markets.9 Overall, the levels of market concentration appear low, with the HHI index ranging from less than 600 in the Russian Federation to more than 4 000 in Mexico. As a comparison of Figure 5.4 and Figure 5.5 indicates, there is no clear link between the measured level of market concentration and the role of the commercial sector, as in some countries a single public institute dominates (e.g. Mexico) while in other countries several smaller public institutes are active (e.g. Belarus).10
Note: Estimated share of commercial sales in total, by volume; remainder of the market is either supplied by public institute or farm-saved seed. Data for Mexico refers to winter wheat only.
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database.
Note: HHI is the Hirschman-Herfindahl Index. All calculations refer to the shares in the overall seed market (including farm-saved seed). Data for Mexico refers to winter wheat only.
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database.
Rapeseed/canola
The database covers 16 markets for rapeseed, although it does not include information on Canada, the largest market for rapeseed. Instead, estimates for Canada have been calculated using data from the Canadian Grain Commission on the acreage of insured varieties. The Canadian rapeseed (canola) seed market is by far the largest in the world, and developments in this market are discussed in more detail in Box 5.1. Measures of market concentration are shown in Figure 5.6 and in Table 5.5
Note: HHI is the Hirschman-Herfindahl Index. All calculations refer to the shares in the overall seed market (including farm-saved seed).
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database except Canada based on Canadian Grain Commission, “Grain varieties by acreage insured,” www.grainscanada.gc.ca (consulted 4 July 2018).
Observed levels of market concentration vary from around 1 200 in Poland to more than 4 200 in Denmark, where the market leader serves more than half of the market. Other markets seem to have more moderate levels of market concentration. In the case of Belarus and the Russian Federation, public research institutes provide a large share of the seed by volume. In Belarus, this also translates into a high market share by value, and hence a high level of market concentration as measured here.
Until the 1960s, rapeseed played a minor role in Canadian agriculture. Less than half a million hectares were planted in 1968 (representing less than 5% of Canadian cropland). Since 1970, however, this area has grown strongly and by 2014 more than 8 million hectares were planted with rapeseed (more than 30% of cropland), as the industry was stimulated by several important developments.
In the late 1970s, publicly-funded research led to the introduction of canola – a specific type of rapeseed with lower levels of erucic acid and glucosinolates. In the following years, the introduction of intellectual property rights in the form of plant breeders’ rights and patents on traits for genetically-modified canola led to a dramatic increase in private R&D efforts. While public R&D spending had dominated historically, the private sector took the lead in the early 1990s, and by the mid-1990s most canola varieties originated in the private sector. Incentives for the private sector were further strengthened by the development of hybrid canola which reduced the role of farm-saved seed.
A by-product of these developments has been the increasing market concentration of seed and biotech firms. Table 5.6 shows estimated market shares of main plant breeders in canola seed for 2008, 2013 and 2017. In 2008, Bayer accounted for more than half the total canola area; the area seeded with Bayer’s most popular variety, InVigor 5020, was roughly equal to the combined area of the next three firms. While market shares fluctuate over time, Bayer remained the leading firm in 2017.
These numbers only capture market concentration in seed (germplasm), but market concentration is also high in GM traits. Figure 2.4 shows the shares of different types of herbicide tolerance traits in canola. Over time, Bayer’s LibertyLink technology has captured the majority of the market for such traits. In 2014, LibertyLink traits had 65% of the market (as measured in hectares); followed by 25% for Monsanto’s Roundup Ready and 10% for BASF’s Clearfield (a non-GM herbicide tolerance trait). These numbers imply a four-firm concentration ratio of around 100% and an HHI of almost 4 900 for genetic traits, considerably higher than the corresponding measures for seed itself.
It is clear that a combination of Bayer and Monsanto would have led to a near-monopoly in traits and a dominant position in seeds. As part of its global divestitures to secure regulatory approval for the merger, Bayer has transferred its canola hybrid business and LibertyLink portfolio to BASF.
The merger of Dow and DuPont, approved by the Canadian Competition Bureau in June 2017, did not involve any remedies in the canola market. As shown in Table 5.6, Dow and DuPont’s combined market share in 2017 was considerably below Bayer’s, while neither company has a presence in the market for canola traits.
Note: Seeded area in millions of hectares by herbicide tolerance traits: conventional (non-GM) canola, Bayer’s LibertyLink, BASF’s Clearfield, and Monsanto’s Roundup Ready.
Source: Brewin and Malla (2017).
* This box is based on Brewin and Malla (2012[135]), (2017[136]) and Malla and Brewin (2015[137]), as well as on data from the Canadian Grain Commission (www.grainscanada.gc.ca).
** A factor that may help explain the popularity of LibertyLink relates to the process by which hybrid canola seed is produced. Although several methods exist, one procedure relies on a genetically modified line which is both male-sterile and herbicide-tolerant, and which is crossed with a second line which is neither male-sterile nor herbicide-tolerant. The two lines are planted near each other. As the first line is male-sterile, it is pollinated exclusively by the second line. The second line is then eliminated by spraying the whole field with herbicide. Only plants of the first line remain, producing herbicide-tolerant seed which is a hybrid of the two lines. Hence, herbicide tolerance facilitates the production of hybrid seed. Bayer’s InVigor with LibertyLink (now owned by BASF) is a prime example of this method (Hawkes et al., 2011[138]). The OECD wishes to thank Robert Duncan and Derek Brewin of the University of Manitoba for clarifying this point.
Sunflower
Table 5.7 and Figure 5.8 contain data on market concentration in sunflower seed in the 11 countries included in the database. These countries jointly account for three-quarters of the global market for sunflower seed (by value). Ukraine and the Russian Federation together account for almost half of the global total.
Note: HHI is the Hirschman-Herfindahl Index. All calculations refer to the shares in the overall seed market (including farm-saved seed).
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database.
The lowest level of market concentration in the sample is found in France and the highest in South Africa. In South Africa, only a limited number of firms are active, with two firms accounting for nearly the entire sunflower seed market.
Sugar beet, potato, and cotton
There is only limited information available in the Kleffmann database regarding concentration in seed markets of sugar beet, cotton seed, and potatoes.11 This information, along with three additional data points from the available literature, is shown in Table 5.8 and Figure 5.9.
Note: HHI is the Hirschman-Herfindahl Index. All calculations refer to the shares in the overall seed market (including farm-saved seed).
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database except as noted in Table 5.8.
Market concentration appears especially high in cotton seed. The level of market concentration in Mexico is the highest level observed in the database (across all crops), and the level in Brazil is similarly high. Concentration in US cotton seed is lower than in Brazil and Mexico, but still fairly high. In all three cases, the market is dominated by GM cotton.
Market concentration for seed potatoes is low in the Netherlands and Germany. In Ukraine and Poland, market concentration for sugar beet seed is also low. However, the market concentration for sugar beet seed in Germany is exceptionally high. The leading firm in this country has more than half of the market, and the two leading firms account for around 90% of the market. Data for France similarly seem to show a high market concentration in sugar beet seed.
With sales of USD 4.7 billion, the market for vegetable seed accounts for some 10% of the global seed market by value (Figure 2.6). The database used in this study does not contain market share information on vegetable seed, but some insights can be obtained from data on plant breeders’ rights in Europe. The analysis here uses the PLUTO database (Chapter 6).
Table 5.9 shows the ownership of EU-wide plant breeders’ rights for cucumber, carrot, lettuce, and tomato, respectively, as well as the four-firm concentration ratio. In all segments, the four largest firms account for most plant breeders’ rights: 72% in tomato, 79% in lettuce, 94% in cucumber, and 95% in carrot. Monsanto holds a strong position in cucumber (46%), while Bayer (through its Nunhems brand) holds a strong position in carrots (52%). Interestingly, several specialised firms have strong positions in vegetable seed, such as Rijk Zwaan and Enza Zaden, both privately held family-owned businesses from the Netherlands.
In its decision on the Bayer-Monsanto merger, the European Commission noted that Monsanto is the largest global supplier of vegetable seeds, and Bayer the fourth-largest player. To secure approval for the merger, Bayer divested its global vegetable seeds business to BASF. A combined Bayer-Monsanto vegetable seeds business would have held around 80% of plant breeders’ rights in cucumber in the European market and 71% in carrot, with a strong position in tomato (39%).
However, plant breeders’ rights are not a perfect measure of market shares. A single “star” variety may have a disproportionate market share while many other varieties may have little or no sales. Many successful varieties were introduced before the period under consideration (2013-2017), including some no longer covered by a plant breeders’ right. Moreover, firms may differ in their strategy regarding protection of their varieties. For these reasons, the correlation between market shares and plant breeders’ rights is far from perfect.
5.3. Determinants of concentration in seed markets
As the database varies in coverage across countries and crops, it is not easy to provide a clear assessment of overall trends. For some countries and crops, only a small number of observations exist. Some patterns can nevertheless be discerned.
Figure 5.10 plots all observations of market concentration for the different crop seed markets. Using the median observation per crop seed market, markets have been ranked from high to low market concentration levels. Despite the limited number of observations available, cotton appears to have the highest levels of market concentration, followed by several crops with intermediate levels (soybean, sunflower, sugar beet, maize). Rapeseed, potato, and wheat and barley appear to have much lower levels of market concentration.
Figure 5.11 plots these data points for the different countries and indicates with which crop each data point corresponds. Fewer observations are available per country than are available per crop. For 14 of the 38 countries there is only one observation; for seven countries, there are only two observations; for eight countries, there are three observations; and for the remaining nine countries, there are at least four observations. Hence, the ranking of countries needs to be interpreted with caution.
Figure 5.10 and Figure 5.11 show data without correcting for composition effects. For instance, the high score for cotton in Figure 5.10 could be due to biological or technological characteristics of cotton, or to the fact that data is only available from countries with higher levels of market concentration overall. Conversely, in Figure 5.11 the high score for South Africa could be due either to the country’s institutional characteristics or to the higher concentration levels overall for soybean, maize and sunflower. In other words, it is important to disentangle “crop-specific” and “country-specific” effects. Given the limitations of the dataset, it will not be possible to perfectly distinguish between the two here, but an attempt can be made through a regression of market concentration levels on crop and country indicator variables and a number of other control variables (Table 5.10).
Strong differences are found between crops. Compared to the reference category of wheat and barley, markets for sugar beet, sunflower and cotton seed on average have an HHI index which is more than 1 000 points higher, a sizeable difference. Similarly, maize and rapeseed have, on average, higher market concentration than wheat and barley. By contrast, the markets for potato and soybean seed do not appear to be systematically more concentrated than wheat and barley, after correcting for other factors including country-specific effects. These crop differences are visualised in Figure 5.12 (using the estimates in the third column of Table 5.10).
In addition, the regression analysis controls for other factors which could in theory affect market concentration, although none of these show a clear effect. In theory a smaller market could lead to higher concentration (as fewer firms are willing to enter the market), a market where GM technology is present could be more concentrated (because of economies of scale and complementarity effects), and a large role for public breeders or a large share of farm-saved seed could intensify competition among commercial firms. The data, however, do not support any of these hypotheses.12
Note: Shaded area refers to the 95% confidence interval around the estimated crop effect. Estimates shown are based on specification (3) in Table 5.10.
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database.
Figure 5.13 and Table 5.11 show show the country effects corresponding to the specification in the last column of Table 5.10. In this case, country effects are expressed as differences in the HHI compared with the reference category, Argentina. For most countries there is no clear evidence of systematic differences, but for some countries the estimated effects tend to be large, with HHI differences of 1 000 points or more compared to Argentina.
It appears from this exercise that Greece, Uruguay and South Africa have systematically higher levels of market concentration whereas countries such as Slovakia, Poland and Belarus appear more competitive across different crop seed markets.
These results need to be interpreted with caution given the low number of observations in the database. For instance, the United States appears to have relatively competitive seed markets in part because the US market for soybean and cotton seed is more competitive than in the other countries in the dataset. If additional markets were included, rankings might look different. Moreover, high levels of market concentration as measured here do not necessarily mean a high level of concentration in the commercial seed sector; in Mexico, for instance, the market for wheat and barley seed appears highly concentrated due to the important role played by public institutes.
Note: Estimates for Canada not available due to small number of observations. Shaded area refers to the 95% confidence interval around the estimated country effect. Estimates shown are based on specification (3) in Table 5.10.
Source: OECD analysis using the Kleffmann amis®AgriGlobe® database.
5.4. Multimarket contact
Seed firms are typically active in several geographies and several crops. For this reason, it is important to study the degree of multimarket contact or overlap between the set of players in different markets. A high degree of overlap in the firms active in different markets could increase the likelihood of collusion between firms as well as reduce the contestability of markets by diminishing the likelihood of entry by outsider firms.
Table 5.12 shows the presence of major maize firms across different countries in the database. The maize firms were selected as they were the ten firms present in the greatest number of countries. For each country, a company is indicated as being active if it is one of the top ten firms in that market. Of the 32 countries for which data on the maize seed market is available, Syngenta and DuPont are active in 31 countries, Monsanto is active in 29, and KWS and Limagrain in 25 countries. The degree of overlap is considerable.
This degree of overlap reflects both the presence of truly multinational firms such as Syngenta, DuPont and Monsanto, as well as firms with a strong European presence, such as Caussade or RAGT. As the Kleffmann database focuses on European countries, by construction firms with a strong European presence appear in the table as major firms; a database with a different geographical focus would presumably identify a different set of major firms. Hence, it is difficult to generalise to all seed markets, but for the markets covered in the Kleffmann database the degree of multimarket contact appears considerable. Even in non-European maize markets such as Argentina, Indonesia, Mexico, the Philippines, Thailand and the United States, there is clear evidence of multimarket contact as Syngenta, DuPont and Monsanto are present in all these markets.
The same pattern is found for the other crop seed markets. For soybean seed (Table 5.13), four firms (Monsanto, Syngenta, Nidera, and Asociados Don Mario) are active in at least five of the seven seed markets in the database. For wheat and barley (Table 5.14), KWS, Nordsaat and Limagrain are competing against each other in more than half of the countries in the database. Similar situations are found for rapeseed (Table 5.15) and sunflower (Table 5.16).13
However, significant differences can be found across markets too. For instance, the degree of multimarket contact seems to be greatest for maize, rapeseed, and sunflower seed but lower for wheat and barley. The set of firms tends to differ by crop. For instance, NPZ is active in a large number of countries, but focuses on rapeseed and is not a major multinational player in other seed markets. Other firms such as RAGT and Euralis are active across several countries and crops, although not necessarily to the same extent as the major firms. There are some notable absences. For instance, despite a strong position in rapeseed in European countries, neither NPZ nor DSV are active in the Canadian market (where GM technology is used, in contrast to the European market), nor is Monsanto a major player in European seed markets for wheat and barley or sunflower. Monsanto sold its European wheat and barley assets to RAGT in 2004 and its global sunflower assets to Syngenta in 2010.
As the database does not have complete coverage of geographies and crops, the findings here almost certainly underestimate the degree of multimarket contact. In general, then, there does appear to be a considerable degree of multimarket contact, both across geographies and across crops.
5.5. Concentration in the market for GM traits
The preceding sections have focused on market concentration in seed (germplasm). This section presents available information on market concentration in markets for GM traits, using a variety of sources. Data on concentration in GM traits can be found in publicly available information for US cotton and Canadian canola markets. For some other markets, the Kleffmann amis®AgriGlobe® database can provide indirect information on market concentration in GM traits. For some markets, the Kleffmann database lists not only sales by firm but includes data on the ten best-selling varieties. Technical documentation (e.g. on company websites) can be used to infer which GM traits are incorporated in these varieties. The resulting view of market concentration in GM traits is incomplete, as only data for the top ten varieties is available. For some markets, these account for the majority of sales; for others, the share of the market covered by the best-selling varieties can be as low as 10% (Table 5.17). Even in these markets, however, the data can provide some useful insights.
The markets covered here represent some of the most important GM markets worldwide and cover two-thirds of the global GM area (Table 5.18). Coverage is greatest for canola (where Canada by itself accounts for most of global GM canola) and maize (where the United States accounts for more than half). Coverage is lowest for cotton as the data does not include major GM cotton producers such as India and Pakistan.
Analysis of market concentration in GM traits raises a conceptual problem because of the frequent practice of combining several GM traits in stacks. For instance, the AcreMax stack for maize (popular in the United States) contains Monsanto’s YieldGard and Dow’s Herculex HX1 as insect resistance traits, as well as Bayer’s LibertyLink and Monsanto’s Roundup Ready 2 herbicide tolerance traits. In a hypothetical scenario where 100% of the market used this stack, it is not clear whether LibertyLink and Roundup Ready should be seen as each simultaneously having 100% of the market for herbicide tolerance traits (since each trait is present in all seeds in this scenario) or as each having 50% of the market. The conceptual problem also exists for insect resistance traits, where different traits may also target different pests while not all pests are present in every sub-region of the market.14 A pragmatic approach is used here: rather than calculating explicit concentration measures, the following paragraphs provide a description of the competitive situation in GM traits.
Maize
The US maize market is the largest GM market in terms of acreage across all crops. The ten best-selling varieties only cover around 10% of the total market for US maize varieties, which makes it difficult to draw strong conclusions about overall market shares for GM traits. Nevertheless, the clear pattern that emerges is that stacks combining several herbicide-tolerant and/or insect resistance traits play an important role in the market. Two popular systems are AcreMax (including variants such as AcreMax Xtra and AcreMax XTreme) and SmartStax. All of these stacks combine traits of several firms, and all of the best-selling varieties include Monsanto’s Roundup Ready and insect resistance traits by Dow. Most varieties also include Bayer’s LibertyLink and Monsanto insect resistance traits; some varieties also include Syngenta’s Agrisure insect resistance trait.
Based on this information, the competitive situation in maize traits appears complex. A small number of stacks dominate the best-selling varieties, and just four firms supplied all the traits included in these stacks. At the same time, the stacks combine traits by competitors. Some of these traits are complementary (e.g. insect resistance traits for different pests) but the combination of two competing herbicide tolerance traits is also common. The merger between Bayer and Monsanto would have removed competition on herbicide tolerance traits, as none of the best-selling varieties included traits by other firms (as noted, Bayer sold its LibertyLink assets to BASF to obtain regulatory approval). The analysis also shows that the DowDuPont merger did not affect the situation in the market for GM traits as DuPont did not have a strong position in GM traits.15
In Brazil, the top-selling maize varieties similarly contain stacked traits such as Dow’s PowerCore, which combines insect resistance traits of Monsanto and Dow, as well as Roundup Ready and LibertyLink herbicide tolerance traits. At the same time, single-firm offerings were also present, such as Monsanto’s VT PRO and Syngenta’s Viptera. As in the United States, Monsanto traits appear to be ubiquitous, whether alone or in combination with Dow traits.
In Argentina, most of the top-ten varieties included Monsanto’s VT Triple Pro technology (a stack which combines insect resistance traits with Roundup Ready herbicide tolerance). Dow’s PowerCore trait stack was also present.
In South Africa, eight of the ten top-selling varieties are GM, and all of these include Monsanto traits, including non-Monsanto maize seed varieties.
The four maize markets covered here jointly account for 87% of the global GM maize area. Stacked traits are used widely. In the United States and Brazil, these stacks often combine traits by competing firms, including competing herbicide tolerance traits. In all markets, GM traits by Monsanto are ubiquitous, either in single-firm stacks or in combination with traits of other firms.
Soybeans
The available data for Brazil, Paraguay and Uruguay show that all of the top-ten soybean varieties in these countries incorporated a Monsanto-only GM trait stack. Both Monsanto’s Roundup Ready stack and the newer Intacta Roundup Ready 2 PRO stack were present. Monsanto’s GM traits also played an important role in Argentina. Trait stacks from other firms were available, in the form of Syngenta’s Plenus system and DuPont’s STS herbicide tolerance trait; these captured a smaller share of the market, however.
Other sources confirm Monsanto’s strong position in soybean traits in Latin America. During the 2013-14 season, Roundup Ready traits were present on an estimated 84% of the Brazilian soybean acreage (Bonato, 2016[142]). Intacta, a newer Monsanto offering, was planted on 24% of the planted area in Brazil, Argentina, Paraguay and Uruguay in the 2015-16 season and was expected to capture 31 to 38% of the area in the 2016-17 season. However, Monsanto did not expect Intacta to gain the same prominence as Roundup Ready had previously, as competing offers would soon become available (Bonato, 2016[142]).
In South Africa, all of the best-selling varieties used Monsanto’s Roundup Ready technology as this is the only soybean GM trait approved in the country (see further for an overview of GM approvals by crop and by country).
The soybean markets covered here jointly account for 63% of the global GM soybean area. As with maize traits, Monsanto’s GM technology appears widespread. In contrast with maize, where stacks sometimes included traits by competitors, the data suggests most soybeans have single-firm trait stacks, with Monsanto capturing most of the market.
Cotton
The Kleffmann database covers the cotton seed markets of Brazil and Mexico, while data from the United States Department of Agriculture is available on the US cotton seed market (Chapter 3.4). Together, these countries only cover some 4.7 million hectares out of an estimated 22 million hectares of global GM cotton area.
For the United States, Monsanto’s Roundup Ready and BollGard traits dominate the market. In 2011, 94% of cotton acreage included a variant of Monsanto’s Roundup Ready technology, while 73% included Monsanto’s BollGard insect resistance traits. Until 2007, nearly all GM technology in the cotton sector was Monsanto-only. After the Department of Justice required Monsanto to allow stacking Monsanto traits with those of competing firms in 2007, mixed stacks grew strongly and accounted for some 35% of total cotton acreage in 2014 and 2015. In parallel, the growth in sales of PhytoGen and Americot cotton varieties also allowed for a greater market share of Bayer (LibertyLink, TwinLink and GlyTol) and Dow (WideStrike) traits, although these remain much smaller than Monsanto.
The Brazilian and Mexican cotton seed market show contrasting patterns. In Brazil, several competing trait stacks are available in the market; the top-selling varieties include trait stacks such as Dow’s Widestrike, Bayer’s GLT (GlyTol, LibertyLink, TwinLink) and Monsanto’s BollGard2 RoundupFlex; Monsanto’s market share is smaller than that of its competitors. In Mexico, however, Monsanto appeared to be the only provider of cotton GM traits, with even Bayer’s cotton seed offerings relying on Monsanto GM traits.
The markets covered here only account for 21% of global GM cotton area, as no data is available for large producers such as India, China, and Pakistan. However, indirect information suggests Monsanto provides nearly all cotton GM traits in India, the largest GM cotton market in the world with 11 million hectares of GM cotton in 2017. Between 2002 and 2006, the only traits approved for commercialisation were Monsanto-owned (Pray and Nagarajan, 2012[143]). Expert estimates put Monsanto’s share of Bt cotton at 90-95% in 2009-10, and Monsanto’s share may have grown even further thanks to the success of its BollGard II varieties (Carl Pray, personal communication).
Rapeseed/canola
As discussed above, the available data for Canada shows that the market for GM traits in canola is considerably more concentrated than the market for seed (Figure 5.7). In 2014, around two-thirds of the canola area in Canada contained Bayer’s LibertyLink traits, while 25% contained Monsanto’s Roundup Ready, and 10% contained BASF’s Clearfield (a non-GM herbicide tolerance trait).
Indirect evidence from GM approvals
The International Service for the Acquisition of Agri-biotech Applications (ISAAA) maintains a global database of GM approvals, which includes information on crop type, developer, trait type, and approval type. Information on approvals of GM events for cultivation can provide an indirect view of concentration on GM trait markets, since without approval no sale can take place. A company’s share of GM approvals should therefore provide a rough indication of its potential share of the GM traits market.16
For maize, the number of GM events approved for cultivation is similar in the United States, Brazil, and Argentina (Table 5.19). Despite these similarities, the concentration of GM events follows somewhat different patterns. In the United States, Monsanto holds 14 of the 42 approved events (33%), with another two developed in collaboration with others. Syngenta, which is in second place, only holds eight approvals (19%). By contrast, in Argentina Syngenta has a considerably larger share of GM approvals than Monsanto, while in Brazil the two firms have the same number of approvals (once Monsanto’s jointly developed events are taken into account). The number of approved GM events for maize is lower in South Africa; Monsanto holds the majority of approvals. Canada has approved the highest number of maize GM events; although Bayer-Monsanto holds the highest number of approved events, Syngenta and DowDuPont hold a considerable share of the total.
A smaller number of events have been approved for soybeans (Table 5.20). Across all markets, Monsanto holds the highest number of approved GM events. Its relative share is highest in South Africa (where Roundup Ready is the only approved GM event for soybeans), Paraguay and Uruguay. In the United States, Brazil, Canada, and Argentina other firms such as Bayer, Dow and DuPont also hold an important share of approved events.
Table 5.21 shows the number of approved GM events in cotton. Across the four markets, Monsanto again has the largest share of GM events. Notably, Monsanto’s share of approved events is smallest in India, where several local players have developed GM events. As mentioned previously, however, Monsanto’s BollGard and BollGard II insect resistance traits are by far the most used in the Indian market. In the other markets, Monsanto holds more than half of all GM events, while Bayer is the second-largest firm in the United States and Brazil.
Finally, for canola in Canada, 19 GM events have been approved for cultivation.17 Of these, twelve are from Bayer, five from Monsanto, and two from DuPont. As noted earlier, BASF’s Clearfield technology is a non-GM herbicide tolerance trait and is therefore not included in these numbers.
Across the different markets, some regularities stand out. First, the total number of approved GM events is relatively low, considering that ISAAA’s database contains all approvals since the 1990s (which means several approved events have probably been superseded by newer technology or have become obsolete because of resistance build-up). A small number of approved events naturally limits the amount of competition that can be expected in a market. Second, markets with smaller numbers of approved events tend to show a greater prominence of Monsanto, which probably reflects Monsanto’s first-mover advantage in developing GM technology. Third, a comparison with the market share data presented earlier shows that the GM events database understates in many cases the degree of concentration.18 For instance, in US cotton, only 10-20% of total area planted does not contain Monsanto traits (a figure which includes non-GM cotton). Yet Table 5.21 shows that Monsanto holds 56% of GM events. Likewise, Monsanto holds two of the six approved GM events in India, but reportedly nearly all of the GM traits market. The available data for other markets similarly suggests that the true degree of market concentration is higher than what is implied by GM event approvals.
Indirect evidence from patents
Another indirect way of evaluating concentration comes from assessing the concentration in ownership of intellectual property rights on biotechnology traits and tools.19
Heisey and Fuglie (2011[7]) estimated that DuPont was responsible for 20.7% of US patents for agricultural biotechnology issued between 1976 and 2000, followed by Monsanto at 16.8%, Dow at 9.9%, and Syngenta at 9.8%. These four firms accounted for 57% of patents.20 These numbers include patents on plants and biotechnologies.
Graff et al. (2003[144]) analysed data on US agricultural biotechnology patents granted between 1982 and 2000. They found that Monsanto (14%), DuPont (13%), and Syngenta (7%) were the main private-sector firms holding such patents.21
Louwaars et al. (2009[133]) provide data for both the United States and the European Union, focusing on private-sector patents on biotechnology processes and technologies. Figure 5.14 shows the evolution for the United States from 1980 to 2004. Both the number of firms and the number of patents granted per year showed a strong increase in the 1980s and 1990s, but with an apparent drop in the early 2000s.22 The share of patents granted to the top firm increased from around 10% of the total during the 1980s to 56% in the early 2000s, while the share of the top ten firms increased from around 50% to almost 90%.
Note: Data refers to patents granted by USPTO for IPC classes A01H1 to A01H4 (which includes processes for changing genotypes and phenotypes, as well as plant reproduction via tissue culture techniques) and C12N15/82, 83 and 84 (which includes recombinant DNA/RNA and other technologies used for the genetic modification of plants), for which full information was available about the identity of the patent holder/applicant. Data excludes patents for products (such as plant varieties), as well as patents from the public or non-profit sector or from individual patent holders.
Source: Analysis by Louwaars et al. (2009[133]) using data from US Patents and Trademark Office (USPTO).
Figure 5.15 shows similar data for the European Union, where trends are somewhat different. The number of firms and patents shows strong growth in the 1990s before reaching a plateau in the early 2000s. The European data show lower and more stable levels of concentration of patents. 23
Note: Data refers to EPO patent applications for IPC classes A01H1 to A01H4 (which includes processes for changing genotypes and phenotypes, as well as plant reproduction via tissue culture techniques) and C12N15/82, 83 and 84 (which includes recombinant DNA/RNA and other technologies used for the genetic modification of plants), for which full information was available about the identity of the patent holder/applicant. Data excludes patents for products (such as plant varieties), as well as patents from the public or non-profit sector or from individual patent holders.
Source: Analysis by Louwaars et al. (2009[133]) using data from European Patent Office (EPO).
Louwaars et al. (2009[133]) also present the main firms applying for biotechnology patents in the 2003-2007 period in the United States and Europe (Table 5.22). In contrast to Monsanto’s strong position in GM traits and approved GM events, the firm with the largest number of patent applications in both the United States and the European Union was DuPont Pioneer. The data confirms the higher level of concentration in the United States, where DuPont and Monsanto jointly accounted for more than half of patent applications. In the European Union, the three leading firms (DuPont, BASF and Monsanto) each accounted for only 8-9% of patent applications.
More recent data on ownership concentration of biotechnology patents is provided by Jefferson et al. (2015[145]), who constructed a unique database of genetic sequences of maize, rice, and soybean referenced in the claims of patents issued globally from 1993 to 2014.24 Their data on sequences and patents for maize and soybean are shown in Figure 5.16.25
For maize, DuPont holds around 400 patents, with Monsanto holding less than 100. For soybean, DuPont holds almost 200 patents while Monsanto holds less than 100. In both cases, DuPont appears to hold a larger gene patent portfolio. Data on the sequences referenced in issued patent claims show an even greater degree of concentration for maize, with DuPont holding more than 1 000 sequences and Monsanto less than 200.26 For both maize and soybean, it is clear that DuPont and Monsanto have a considerable lead over other multinational firms in terms of ownership of biotechnology-related patents.27
Note: Figures show the holdings of plant-related patents issued globally during the period 1993-2014 and of the genetic sequences referenced in the claims of these patents.
Source: Jefferson et al. (2015[145])
Recent technological developments, in particular CRISPR-Cas gene editing technology, could considerably change which patents are relevant to biotechnology. It is therefore important to understand the ownership pattern of these patents. A recent analysis by Egelie et al. (2016[146]) identified all patents filed around the world referring to any aspect of CRISPR and Cas9 technologies, removing double-counting of equivalent patents filed around the world. Figure 5.17 shows the top ten patent holders.
Of 604 inventions, MIT accounted for the greatest share (45, or 5.3%) followed by Harvard (44, 5.2%) and the Broad Institute (30, 3.6%). These three institutes collaborate intensively on CRISPR-Cas9. Moreover, Editas is a commercial spin-off from MIT and the Broad Institute. Aggregating these four organisations puts their total at 131 inventions, or 16% of the total. When the numbers for Dow and DuPont are combined, this firm becomes the second largest holder, with 33 inventions (4%). The University of California has only a small share of the inventions (14, or 2%), but these may include some essential aspects of the technology. Remarkably, other firms such as Bayer, Monsanto or Syngenta do not have a strong position in these patents.
Note: Patents counted in terms of “inventions” (i.e. removing double-counting of equivalent patent applications worldwide). Top ten holders account for 28% of total inventions.
Source: Egelie et al. (2016) using data from Thomson Innovation
Patent holdings on CRISPR-Cas9 thus appear considerably less concentrated than is the case for agricultural biotechnology, and academic institutes dominate the patent portfolio. A more detailed analysis by Egelie et al. (2016[146]) shows that industry players such as Dow and DuPont seem to be more active in agricultural and food applications. At the same time, the prominent role of academic institutions does not guarantee easy access to the intellectual property underlying CRISPR-Cas9. While most academic institutions allow free use of their technologies for academic research purposes, they do not necessarily commit to providing broad and non-exclusive commercial licensing to make technology widely available for commercial applications.
Concluding remarks
Assessing market concentration in GM traits is difficult both because of the limited availability of data and because of conceptual difficulties. Yet, the evidence reviewed here strongly suggests that the degree of market concentration in GM traits is higher than that of seeds across the different markets surveyed.
Whereas independent local or regional firms often play a role in specific seed markets (e.g. Americot in US cotton; Don Mario in soybean seed in Latin America), only the major firms (Monsanto, Bayer, Syngenta, Dow, DuPont) are generally active in the markets for GM traits. In maize markets, trait stacks often include traits from competing firms, but this seems to be less common in other markets. Overall, traits by Monsanto appear to be ubiquitous. By contrast, patents for CRISPR-Cas9 appear much less concentrated and are mostly held by academic institutes. While DowDuPont holds several patents, other firms such as Bayer-Monsanto or Syngenta were notably absent from the list of top 10 patent holders of this new technology.
The analysis here also suggests that indirect methods of assessing market concentration may give a misleading picture. First, a comparison of data on best-selling varieties with data on approved GM events suggests that the latter tends to understate the true extent of concentration in the GM traits market. At the same time, the data on patents consistently shows DuPont as either the leading firm in terms of patents on agricultural biotechnology or a close second after Monsanto. Yet DuPont is not a major player in terms of either approved GM events or market share of GM traits. Indirect measures of market concentration based on approved GM events or patent ownership are hence imperfect proxies for actual market concentration in sales.
5.6. Conclusion
This chapter has reviewed a large amount of new empirical evidence on market concentration in seed and GM traits, covering a broad set of countries and crops. Several insights emerge from the data.
First, seed markets show a large variation in market concentration. The Hirschman-Herfindahl Index varies from values below 1 000 (equivalent to a market with ten equally-sized firms) to values above 5 000 (equivalent to a market with two equally-sized firms).
Second, within this heterogeneity there are clear patterns. Seed markets for sugar beet, cotton, sunflower, maize, and rapeseed are typically more concentrated, while seed markets for potato, soybean and wheat and barley appear much less concentrated. There is also evidence of differences between countries, as some have systematically higher degrees of market concentration across different crop seed markets. However, the statistical analysis did not show a clear relationship between market concentration and market size, the presence of GM technology, the volume share of public breeders, or the volume share of farm-saved seed.
Third, there is a considerable degree of multimarket contact across seed markets, with similar firms competing against each other in different geographies and different crops. Not surprisingly, the large multinational firms are found in many markets. However, there are important differences between markets. There is more multimarket contact for maize, rapeseed, and sunflower, and less for wheat and barley. The set of firms also differs: some firms are active across geographies in a limited number of crops (e.g. NPZ in rapeseed), while large firms may be absent in some markets (e.g. Monsanto in European markets for wheat and barley, and sunflower).
For GM traits, market concentration appears much higher than for seed markets. While several medium-sized regional players are active in seed markets, the market for GM traits is dominated almost exclusively by large multinational firms (Monsanto, Bayer, Syngenta, DowDuPont). Traits by Monsanto appear particularly prominent, especially in markets where a smaller number of events has been approved. Data on patents for CRISPR-Cas9, however, suggest that this new technology is mostly dominated by academic institutes, with some presence of DowDuPont, but no strong position for Bayer-Monsanto or Syngenta.
An overarching conclusion from the analysis in this chapter is the need for precise data on market concentration issues, and an understanding of the context of specific markets. In the case of GM traits, data on patents or approvals appear to understate the real degree of market concentration. As noted earlier, discussions of market concentration in seed and GM traits have often relied on aggregate data, such as the four-firm concentration ratio for the global seed and GM traits market as a whole. Such measures are not very useful for policy makers. For instance, the commercial seed market is less important as a source of seed in developing regions, and within the commercial seed market some regions and crops dominate in terms of market size. Global figures will therefore tend to reflect the situation in the largest commercial seed markets. Furthermore, as the detailed analysis here has shown, there are important differences by crop and by country. Using aggregate figures can therefore result in misleading conclusions. By contrast, disaggregated data make it possible to conduct more precise analyses on the effects of market concentration, as shown in Chapter 6.
The data presented in this chapter covers mostly developed countries and some emerging economies. For most developing and least developed countries, little to no information is available regarding market concentration or the potential impact of recent mergers.
In many of these countries, farmers rely on informal systems (e.g. farm-saved seed, seed exchanged with neighbours and informal local markets) rather than on the commercial seed sector. In a survey of farmers in five Sub-Saharan African countries and Haiti, McGuire and Sperling (2016[147]) found that 90% of seed was sourced through such informal channels (mostly informal local markets). Likewise, the formal commercial sector in India is estimated to account for only a quarter of all seed (see Spielman et al. (2014[148]), (2014[149])). Small, resource-poor farmers in particular tend to rely on informal channels. In turn, this means that mergers and market concentration are likely to have only a limited impact on these farmers.
Yet the commercial sector appears to be growing in many developing countries. In India, the commercial seed sector is growing more than 10% per year (Spielman et al., 2014[148]), while in Sub-Saharan Africa the number of private seed firms has been increasing in recent years. In addition to local and regional seed firms, many of the leading multinationals are active in the emerging commercial seed markets in Africa (African Centre for Biodiversity, 2015[150]).
In addition to local and regional firms, many of the large global firms are active in these developing countries. In 2013, DuPont acquired Pannar, a South African seed company with a presence in 24 countries in Sub-Saharan Africa. Prior to this acquisition, DuPont had already established a presence in the region. Monsanto has a strong presence in South Africa, with an estimated market share of 45% for field crops through its ownership of Sensako (acquired in the 1990s); it is also present in several other eastern and southern African countries, although with much smaller market shares. Syngenta is similarly active in the region; in 2013, for instance, it acquired MRI, the largest private seed company in Zambia (African Centre for Biodiversity, 2015[150]).
In developing countries, the commercial sector tends to be prominent in maize, in part because hybrid maize offers a biological means to protect intellectual property. In India, private-sector investment has focused on cotton, maize, pearl millet, sorghum, and horticulture crops where hybrids are possible (Spielman et al., 2014[148]). Data on market concentration among these commercial firms is scarce. For India, Pray and Nagarajan (2014[151]) report a four-firm concentration ratio of 34% in 2009 for the private seed firms in their sample, down from 77% in 2000. Similarly, Murugkar et al. (2007[152]) estimate that the top five firms selling proprietary hybrid seed in India saw their share decline from 84% in 1996-7 to 59% in 2004-5, although the market for GM traits is more concentrated. Spielman and Kennedy (2016[153]) report data for Nepal in 2012, finding a four-firm concentration ratio of 64% for rice, 82% for wheat, and 91% for maize. For other countries and seed markets, public data is generally lacking, although proprietary data may exist in the private sector.
Given the growing importance of commercial seed sectors in the developing world, there is a clear need for more public data on concentration in these markets, in addition to several other indicators necessary to enable better policymaking (Spielman and Kennedy, 2016[153]).
The available data allows for a “pro forma” assessment of how the recent mergers of Dow and DuPont and of Bayer and Monsanto would have affected market concentration in the absence of any divestitures.
For the countries covered here, the mergers would have no effect on wheat and barley. In 2016, Bayer did not have any wheat or barley varieties on the market.28 Monsanto likewise had little presence in wheat and barley. In 2004, the firm sold its European wheat and barley breeding programmes to the French firm RAGT as part of a global reorganisation. In other markets included in the dataset, Monsanto’s role in wheat and barley is negligible. Dow had an active presence in wheat in Australia, but limited activity in other markets; DuPont sells wheat seed in the United States, but has a limited presence in Turkey (as well as in Italy, Greece and Portugal, which are not included in the database).
Likewise, the mergers would have no effect on markets for seed potatoes and sugar beet seed, as none of the four firms (Bayer, Monsanto, Dow, DuPont) were active in these markets.29 In sunflower seed, the dataset shows that mergers would affect some of the countries covered, although the effect on market shares would be small.30
Maize
Table 5.A.1 shows the impact of the mergers on relevant maize seed markets using a “pro forma” approach, i.e. combining the market shares of 2016 without taking into account divestitures. As Bayer was not active in maize, the impacts reflect the merger between Dow and DuPont.
As a guide to interpreting the results, the Horizontal Merger Guidelines of the US Department of Justice indicate that mergers “potentially raise significant competitive concerns” when they change the HHI by at least 100 points and result in a HHI of more than 1 500 points. Mergers which raise the HHI by more than 200 points and lead to a highly concentrated market (with a HHI above 2 500 points) are “presumed to be likely to enhance market power.”
By far the largest impact would have been in Brazil, with an increase in the HHI of more than 1 000 points, leading to a post-merger HHI of 3 900. The mergers would also have led to a large increase in the HHI in the United States and Slovakia. In the United States, the impact of the merger was apparently not considered a concern by the competition authorities; while DuPont was only required to divest certain crop protection chemicals, no divestitures were required in seed markets.
Given the large potential effects in the Brazilian maize seed market, the Brazilian Competition Authority (CADE) imposed a set of measures as a precondition for allowing the DowDuPont merger. Dow was required to transfer a copy of its germplasm bank, part of its maize hybrid varieties (including some still under development), some production and research facilities as well as associated brands and employees. These assets were acquired by the Chinese investment fund CITIC for USD 1.1 billion in 2017 and renamed LP Sementes.
Several maize seed markets are included in the database but not shown in Table 5.A.1 as the mergers would not affect the market. This is the case for Austria, Belarus, Belgium, Bulgaria, Croatia, Denmark, France, Germany, Greece, Indonesia, Italy, Mexico, the Netherlands, the Philippines, Poland, Romania, the Russian Federation, Serbia, Slovenia, South Africa, Thailand, and Turkey. DuPont is active in all these markets (except Belarus) but Dow is not, or has only a minor presence. Similarly, although Monsanto is present in some of these markets, Bayer is not active in any. Hence, a large number of maize seed markets are not affected by the mergers.
Soybean
For soybean seed, the available data points to a noticeable impact of the mergers in the United States, where the HHI would increase from 1 683 to 2 041, an increase of 358 points. However, the resulting HHI is fairly low, which probably explains why again no remedies were requested from Dow and DuPont in this regard.
For Brazil, available market data for 2016 seems to show little to no impact of the Bayer-Monsanto merger. Bayer had sold its soybean activities in Brazil to the Dutch firm Nidera in 2005 (which in turn was sold to Syngenta in 2017). However, in this case historical market shares understate the potential role Bayer might have played in soybean seed markets without a merger. Bayer had been following a deliberate strategy of increasing its investments in the Brazilian soybean seed industry to challenge Monsanto’s strong market position. In 2013, Bayer acquired the soybean seed company Wehrtec, as well as the soybean seed business of Agricola Wehrmann and a soy germplasm bank of Melhoramento Agropastoril-Cascavel (Birkett, 2013[154]). In 2015, Bayer acquired the seed business of CCGL-Cruz Alta (which included soybean seeds) (Birkett, 2015[155]). Bayer also acquired firms active in soybean seeds in Paraguay (Granar) and Argentina (FN Semillas), and in its 2015 Annual Report explicitly announced its intention to “establish competitive positions in soybeans and wheat” (Bayer, 2015, p. 60[156]). In 2016, Bayer announced plans to introduce genetically modified soybean seeds (Freitas, 2016[157]). In other words, a merger with Monsanto would have eliminated a strong potential competitor in the industry. For this reason, the Brazilian competition authorities required the divestiture of Bayer’s soybean seed assets (among others).
Rapeseed/canola
Table 5.A.2 shows the pro forma impact of mergers on the rapeseed seed market.
Monsanto is an important supplier of rapeseed seed for Bulgaria, the Czech Republic, Denmark, France, Germany, Hungary, Latvia, Poland, Romania, Slovakia, Sweden, the United Kingdom, as well as the Russian Federation and Ukraine. In some of these countries, Bayer is also active, although typically with a smaller presence. As Table 5.A.2 shows, in some countries the merger would increase the HHI considerably (in several of these countries, the HHI would increase by more than 100 points). Generally, the post-merger HHI would remain relatively low based on 2016 data. Nevertheless, the European Commission was of the opinion that the Bayer-Monsanto merger “would have eliminated competition in Europe between the largest supplier in Europe – Monsanto – and the largest supplier globally, Bayer, which is currently expanding into Europe” (European Commission, 2018[134]). Bayer’s global rapeseed plant breeding activities form part of the assets sold to BASF.
The largest impact by far, however, would have occurred in Canada. In 2017, Bayer had an estimated market share of 56% and Monsanto 12% (Box 5.1). A merger would have increased the HHI from an already-high value of almost 3 500 to more than 4 800, a level practically equivalent to a market dominated by two equal-sized firms.
Cotton
The available data shows a large potential impact of the mergers in cotton seed markets (Table 5.A.3). For the three countries for which data is available, the mergers would have led to a strong increase in competition; in the case of Mexico even leading to a near-monopoly, with a post-merger HHI practically equal to the theoretical maximum of 10 000. In none of these markets did DuPont have a presence; hence, the effects are due to the Bayer-Monsanto merger.31
In Mexico, competition authorities required the divestiture of Bayer’s cotton assets as a precondition for approving the Bayer-Monsanto merger (among other divestitures in vegetable seeds and crop protection chemicals). Bayer divested these assets to BASF. In Brazil, the divestiture of Bayer’s cotton seed business was one of the prerequisites for approving the Bayer-Monsanto merger.
While the dataset does not cover European countries, it is worth noting that the European Commission also indicated that a Bayer-Monsanto merger (without any divestitures) could have reduced competition in the market for cotton seed licensing in Europe.
Notes
In preparing this study, the OECD contacted three market research firms specialising in agricultural input markets: Kynetec, Informa – Phillips McDougall, and Kleffmann. After comparing the available information and pricing, the OECD chose Kleffmann as the data provider. This annex discusses the methodology behind the Kleffmann database, provides some validation checks of the data against external sources, and gives further methodological notes regarding the calculation of market concentration measures.
Methodology of the Kleffmann AgriGlobe database
The main data source in this report is the Kleffmann Group’s amis®AgriGlobe® database. The Kleffmann Group specialises in agricultural market research on seed, crop protection, agricultural machinery, livestock and animal health, and related markets.
Kleffmann data on seed markets covers 75 countries, corresponding to an estimated 95% of the global seed market. However, data coverage does not go into the same level of detail in every country. Data on market shares are available only for selected crops in a smaller number of countries. Market share data is obtained by Kleffmann through annual interviews with more than 100 000 farmers worldwide, covering over 60% of the data in value terms. Standardised questionnaires are used to ensure comparability of the data across years and across countries. Where such detailed farmer panels are not conducted, data are obtained through bi-annual distributor surveys in emerging markets and information from consultants and experts to obtain estimates of market size by crop. All Kleffmann data refers to 2016 unless noted otherwise.
Cross-validation with other data sources
The Kleffmann data is consistent with data obtained from other sources, as seen from a number of validation exercises. First, it is possible to compare Kleffmann’s overall estimate of the size of the global seed market with other estimates as presented in Bonny (2017[12]), as well as the estimate by Syngenta (2016[15]) and ISF estimates presented in Ragonnaud (2013[17]). As shown in Figure 5.B.1 below, Kleffmann’s estimate for the size of the global seed market is between 65 and 70 billion for 2016. Although somewhat on the higher end of the range of the available estimates, this is in line with the ISF and Syngenta estimates of the preceding years. (In addition to ISF and Syngenta estimates, the figure shows estimates from eleven other market research firms, collected by Bonny (2017[12]); several are similar to ISF estimates but some estimates are considerably lower).
A more detailed assessment is possible by comparing the Kleffmann data with the data in Syngenta (2016[15]), which were used to construct Figure 2.4. The Syngenta data represent estimates of the size of specific crop seed markets (e.g. maize seed) in broadly defined regions (e.g. Europe, the Middle East and Africa; Asia Pacific). To compare the two data sources, the Kleffmann data were aggregated bottom-up to a similar regional grouping. One challenge here is that definitions do not match exactly. For instance, it is not clear which countries are included in Syngenta’s regions, and the Kleffmann database includes data for “cereals”, i.e. wheat and barley, whereas Syngenta only shows data for wheat. Despite these discrepancies, the numbers from the two sources are broadly consistent, with the exception of estimates for the size of maize seed markets in Latin America and in Europe, the Middle East and Africa, where the Kleffmann numbers are lower. Part of this discrepancy is likely due to the fact that a bottom-up calculation was used for Kleffmann, while not all countries are represented in the Kleffmann dataset; part of this may also be due to definitional differences (e.g. some datasets may define “maize” to also include other coarse grains). The two data sources nevertheless seem to match fairly closely.
Source: Bonny (2017), Syngenta (2016), Ragonnaud (2013).
Note: In the first panel, each point corresponds with a regional market for a specific crop, e.g. maize in Latin America. In the second panel, each point corresponds with a country, e.g. the total seed market in France..
Source: OECD analysis using the Kleffmann database and Syngenta (2016) (first panel) and ISF data cited in ISAAA (2016) (second panel).
ISAAA (2016[16]) cites ISF estimates of the value of seed markets at the national level for 2012. These data are compared with the corresponding Kleffmann data for 2016 in the figure below. Despite the different years, there is a high correlation between the two datasets, with a correlation coefficient of 0.98. The main outlier is the Philippines, where the discrepancy seems due to a fairly low ISF estimate of USD 18 million. In other words, the Kleffmann estimates at national level again seem to be validated by other sources.
A final quality check on the Kleffmann data was conducted for the markets included in the empirical analysis of market concentration. The Kleffmann data reports market shares and total market sizes not only in volume (tonnes of seed) and value, but also in terms of hectares. These numbers were compared with publicly available data on the total area devoted to the corresponding crop, using a variety of data sources (including FAOSTAT, Eurostat, USDA GAIN reports, and websites of national statistical agencies). For most markets there was a close fit between these different data sources; where numbers deviated by more than 20%, the precise definitions were verified with Kleffmann to ensure a correct interpretation of results.
Example
For each market included in the analysis of market concentration, the Kleffmann database reports the ten leading suppliers of seed, with their corresponding volume, sales value, acreage, and overall size of the market, as shown in Table 5.B.1 using fictional data.
“Company” here denotes the ultimate global owner of the varieties, i.e. the plant breeders. For instance, when a plant breeder licenses out the multiplication or distribution of varieties to other firms, the company listed here will be the plant breeder, not the distributor. Similarly, the Kleffmann data uses the ultimate global owner; hence, a local subsidiary of Monsanto will be listed as Monsanto in the database. Public institutes are not identified as such in the Kleffmann data; private and public sector breeders are distinguished by manually verifying the names of the breeders.
Farm-saved seed is typically included as seed with “unknown” plant breeder. The monetary value of farm-saved seed is calculated by assigning an average price to a volume estimate. This average price tends to be lower than the prevailing average price in the market. The “Grand Total” refers to the overall market size, including the non-commercial part of the market (i.e. including farm-saved seed).
Calculating market concentration
The main measures of market concentration used in this chapter are the four-firm concentration ratio and the Hirschman-Herfindahl Index. The four-firm concentration ratio was found simply by adding up the market shares of the four largest firms. In case one of the top four firms was listed as “unknown” (i.e. farm-saved seed), the next largest firm was used.
The Hirschman-Herfindahl Index is defined as the sum of the squared market shares of all firms in a market. In theory, calculating the HHI requires information on the market shares of all firms in a market, while the Kleffmann data only includes information for the ten largest firms. The calculation of the HHI here uses an approximation for the unobserved firms. As is common in the industrial organisation literature, the assumption is that the unobserved firms are all as large as the smallest firm in the top ten. In other words, if the tenth-largest firm has a market share of 5%, and if the top ten firms together account for 85% of the market, the assumption is that there are three other firms with 5% each (3 x 5% = 15%, the remaining part of the market). This approximation slightly overstates the HHI (as it assumes all remaining firms are equally large), but the effect tends to be negligible.
As mentioned in the main text, this study chooses to calculate market concentration relative to the overall seed market, including farm-saved seed, rather than only the commercial seed market. In calculating the HHI, farm-saved seed is considered part of the market accounted for by firms outside of the top ten, and is thus subject to the same approximation.
Notes
← 1. Approximately 54% of DuPont’s 2016 agricultural sales originated in North America, with an additional 16% coming from Latin America. These figures refer to overall agricultural sales including crop protection chemicals (which accounted for 30% of total revenues).
← 2. The importance of farm-saved seed varies across countries and across crops. Farm-saved seed is less common for crops where hybrid varieties are used, as is notably the case for maize. In the dataset used here, farm-saved seed is especially important for cereals (wheat and barley), as discussed below. The Kleffmann database lists farm-saved seed as seed with an “unknown” plant breeder. In the calculations of market concentration, rather than interpreting farm-saved seed as a single “supplier”, the total share of farm-saved seed is interpreted as if it consisted of a large number of small firms, which is a better representation of the actual situation.
← 3. The owner of the variety is not necessarily the firm responsible for multiplication, distribution or sales of the seed. Owners of varieties often let other firms take care of these tasks in return for royalty payments; these firms tend to be more numerous and less concentrated than those engaging in plant breeding. In Argentina, for instance, the four-firm concentration ratio for soybean seed is 88% by volume using the final owner of the variety. For the actual production of seed, the corresponding figure is only 29% (Pedro Lavignolle, INASE, personal communication).
← 4. Market shares are based on ownership of varieties, even if multiplication and distribution are organised through private firms.
← 5. A drawback of the HHI is that the index is less intuitive than the C4 ratio. To help with interpretation, it is useful to keep in mind that with n equally-sized firms, the HHI index is (10 000/n). For instance, if four firms each had 25% of the market, the HHI would be 2 500; if two firms each had 50% of the market, the HHI would be 5 000.
← 6. The value-based HHI correlates strongly with the volume-based HHI and the C4-ratios. We use the value-based HHI here as it is the measure used in economic analysis.
← 7. A comparison with similar market share data for 2013 for 30 countries shows no clear pattern of increasing or decreasing market concentration. Across these countries, the median change in the HHI index (in value terms) is a modest decrease of 118 points; the median change in the C4 ratio (in value terms) is an increase of 1%. In terms of the HHI, 47% of the countries registered an increase versus 57% when using the C4 ratio.
← 8. The Kleffmann database refers to “cereals,” a category which does not include maize and hence refers mainly to wheat and barley. In the case of Mexico, the Kleffmann database normally includes sorghum under cereals. As this market has different characteristics, the analysis here uses separate estimates by Kleffmann for winter wheat only.
← 9. In calculating these figures, no distinction has been made between market shares of public institutes and of commercial firms. Seed with “unknown” plant breeder (typically referring to farm-saved seed) has been treated in the same way as unallocated sales more generally; the assumption used is that these sales are provided by a number of small firms with market shares equal to the smallest market share observed in the data. (See Annex 5.B). This is the common approach used in the literature to calculate the HHI index when some sales are unallocated. This approach tends to overstate the degree of market concentration, although the impact is likely to be small in the sample used here. In the wheat and barley seed data, the smallest market share by country is typically between 1% and 2.5%.
← 10. The concentration measures used here consider each public institute separately. In the statistical analysis presented later, the total share of seed coming from public institutes is used as an additional control variable.
← 11. In contrast with the other crops mentioned here, potatoes are not grown from seed but from tubers (which are, however, referred to as “seed potatoes”). Nevertheless, the terminology of seed markets will be applied to potatoes here.
← 12. The volume share of public breeders is the sum of the volume shares of all public sector plant breeders in a market. Using their combined market share in volume terms rather than in value terms gives a better representation of the importance of public sector breeders, as these sometimes sell seed at lower prices.
← 13. Data for sugar beet, potato and cotton are not presented given the limited number of countries in the database.
← 14. The approach of “trait acres” used by Heisey and Fuglie (2011[7]) measures the area sown to GM crops where stacked GM traits are counted multiple times based on the number of traits stacked. This measure gives an indication of the relative importance of firms in overall GM activity in a market. However, it does not necessarily provide an appropriate measure of market share. For instance, if an insect resistance trait is the only trait on the market protecting against a specific pest, it should logically have 100% market share. If stacked traits are the norm and if the firm supplying this trait does not supply many other traits, however, the trait acres approach could show a much smaller market share for the firm.
← 15. For this reason, the National Corn Growers Association pointed out that the Dow-DuPont merger could benefit farmers as “[t]he Dow-DuPont combination brings together Dow’s trait development expertise with Pioneer’s germplasm and distribution network – making the new company a far stronger competitor with the current industry leader [Monsanto]” (National Corn Growers Association, 2016[235]).
← 16. The number of approvals was used, for instance, as an indicator of GM trait market concentration in a joint statement on the Bayer-Monsanto merger by the American Antitrust Institute, Food & Water Watch and the National Farmers Union (2017[236]).
← 17. The discussion here focuses on Argentine canola (Brassica napus), the most commonly used variety. Four GM events have been approved for Polish canola (Brassica rapa), of which one by Bayer and three by the University of Florida.
← 18. A similar result was found by Heisey and Fuglie (2011[7]), who found that Monsanto accounted for around half of global approvals (1982-2007) but about 85% of global trait-acres in 2007. (“Trait-acres” represent the area sown to GM crops where stacked GM traits are counted as multiple acres).
← 19. The discussion here focuses on intellectual property for GM and other biotechnology (including recent techniques such as gene editing). For a review of broader questions of intellectual property in agriculture, see Clancy and Moschini (2017[238]).
← 20. The measure used here is based on the narrow definition of agricultural biotechnology in Heisey and Fuglie (2011[7]), focusing on patents “pertaining specifically to crops and to the suite of modern biotechnology techniques.”
← 21. Interestingly, 24% of US agricultural biotechnology patents were granted to public sector actors such as universities. Graff et al. (2003[145]) report the public sector also played a large role in Europe (25%) and Japan (14%). There were qualitative differences between the types of patents granted to public and private organisations, with public research focusing more on plant developmental processes and private research on specific applications.
← 22. It is not clear whether this decline reflects an actual decrease or whether it is due to a change in reporting or other data issues. When combining information on firms’ shares of the total, it appears that the decline in patents occurred across all firms, including the top firm. This would suggest either a widespread structural change or a change in reporting.
← 23. Louwaars et al. (2009[133]) also provide data for 2005-2006. The period shown here is restricted to 2004 to allow for a better comparability with the US data. Moreover, results for this two-year period are more likely to be skewed by exceptional results in a single year. The 2005-2006 data in Europe do appear to show a reduction in the number of patents per year, although much less than what the US data show for 2000-2004 compared to the preceding period. The 2005-2006 data show higher concentration levels for the top firm, top 5, and top 10 firms. These higher shares may simply be due to the shorter period: some smaller firms may not apply for a patent every year and thus not appear in the data for 2005-2006 while they would show up when a longer period was used.
← 24. The data underlying the paper by Jefferson et al. (2015[146]) are based on work by The Lens (www.lens.org), a project to provide free access to global data on scholarly works and patents, and gene and genome-based patents in particular. The Lens contains more than 110 million patent records from 95 jurisdictions. Its PatSeq platform for genetic sequences referenced in patents includes more than 294 million patent sequences linked to patent information and scholarly research. See Jefferson et al. (2015[244]) and (2015[245]) for a description of data and methods. The collection of issued patents in Figure 5.16 is available at https://www.lens.org/lens/collection/5464.
← 25. Data for rice, not included here, shows less concentration in patent ownership, but also less patenting overall compared to maize and soybean.
← 26. The data in Jefferson et al. (2015[146]) further confirm the findings of Graff et al. (2003[145]) on the important role of the public sector, reporting that the public sector accounted for 40-50% of plant gene patenting before 2000 and around 25% since.
← 27. In some jurisdictions (notably the United States), patents can be issued for new plant varieties obtained with or without genetic modification. By contrast, the data here refers to patents referencing genetic sequences in their claims, which implies the data are a good proxy for biotechnology-related patents.
← 28. Bayer did invest to develop new wheat varieties, for instance by opening a wheat breeding station near Paris (France) in 2013 and through a partnership with the Dutch biotech firm KeyGene.
← 29. In sugar beet, Monsanto has developed Roundup Ready GM sugar beet traits, but does not sell sugar beet seed. Rather, these traits are used in the North American market by other firms such as Betaseed, owned by KWS.
← 30. In none of the sunflower seed markets included here would the HHI increase by more than 100 points. The US Horizontal Merger Guidelines indicate that mergers are “unlikely to have adverse competitive effects” if the change in HHI is below this level.
← 31. DuPont’s cotton activities were concentrated in two countries: India and Greece.