This chapter reviews theoretical and empirical literature on the potential effect of mergers on prices, innovation, and product choice. In addition to insights from the general economics literature, empirical evidence on the market for seed and GM technology is examined..
Concentration in Seed Markets
Potential Effects and Policy Responses
4. Theory and evidence on the potential effects of mergers
Abstract
Mergers can have negative impacts on consumers through higher prices, a lower rate of innovation, and reduced product choices. However, it is by no means certain that a given merger will have these negative effects, or that the effects will be large. Depending on the circumstances, negative effects may be offset by positive effects through efficiency gains.
This chapter discusses the possible impact of mergers in more detail, using economic theory and empirical evidence. The possible effects on prices are discussed first as this topic has been developed the most in the literature. The effects on innovation and on product choices are then discussed.
4.1. Potential effects on prices
The literature on mergers typically distinguishes between horizontal and non-horizontal mergers.1 In a horizontal merger, the merging firms are both active in the same market, and the merger therefore increases market concentration. All other types are by definition non-horizontal mergers (also known as conglomerate mergers); typical examples are mergers between firms that have a supplier-customer relation (a vertical merger) or between firms selling complementary products. As Chapter 3 demonstrated, this distinction is relevant in the seed industry as consolidation combines both horizontal and non-horizontal aspects.
Horizontal mergers
Horizontal mergers remove one competitor from the field and create a larger firm which may have more market power. This is known as a non-coordinated or unilateral effect of a merger.
Non-coordinated effects are more likely when consumers have fewer alternatives, e.g. when the merging firms have large market shares or when other firms in the industry are not supplying close substitutes. The merger itself may also make it less likely that alternatives emerge; for instance, when the merged firm controls important input sources or distribution channels and uses this power to make it more difficult for rivals to compete.2 A merger may also eliminate a firm that played an important role in keeping the market competitive, for instance because it was traditionally a price-cutter.
In addition to these non-coordinated effects, a merger can make it easier for the remaining firms in the industry to coordinate their behaviour to raise prices. This is known as a coordinated effect.
In a concentrated industry, firms may (implicitly or explicitly) coordinate to keep prices high, to limit production, and/or to limit capacity expansion. Firms can also try to reduce competition by agreeing to split the market, e.g. by geography or by customer segment. However, such (implicit or explicit) agreements are vulnerable to deviation by one or more firms. If several firms agree to keep prices high, one firm could profit by deviating from the agreement to undercut its rivals. To prevent this, firms could threaten to punish deviations, for instance by starting a price war. Whether an attempt at coordination will succeed depends on how easy it is for firms to monitor each other’s behaviour and whether they can credibly threaten to punish deviating firms.
A merger can make successful coordination among firms more likely by making it easier to agree, monitor or punish. First, a merged firm may have a disproportionately high market share in one geography or customer segment. This can make it easier for firms in the industry to implicitly agree to split the market, e.g. because other firms expect that the merged firm would react especially strongly against any intrusions in its “stronghold” market. Second, as the merger reduces the number of firms, it becomes easier to check whether a firm is deviating from an agreement. Third, a merged firm may be better able to survive a price war, e.g. thanks to lower costs or greater financial resources; this in turn can increase the likelihood that the merged firm can successfully punish a deviant firm through a price war.
Efficiencies
Firms often motivate mergers by pointing to expected efficiency gains from combining their operations, such as lower variable production costs from economies of scale or lower fixed costs by combining legal, marketing and other “overhead” functions.
From a strict social welfare point of view, a merger is welfare-enhancing if total efficiency gains outweigh the total efficiency losses associated with anti-competitive effects. In theory, a merger could therefore increase social welfare while simultaneously hurting customers. In reality, competition authorities typically use customer welfare as relevant benchmark. For this reason, efficiencies are usually only considered as a countervailing factor in a merger analysis if they benefit the customer, if they are merger-specific, and if they are verifiable.
A practical implication of this focus on the customer impact is that reductions in variable or marginal costs will be more likely to be counted as a relevant efficiency gain compared with reductions in fixed costs. Lower marginal costs are more likely to be passed on to customers in the form of lower prices and may outweigh the possible tendency to price increases. In contrast, reductions in fixed costs are more likely to be translated into higher profits for the firm, without direct benefits for the customers.3
Non-horizontal mergers
Compared to horizontal mergers, non-horizontal mergers are generally considered a lesser threat to competition. By definition, such mergers do not lead to changes in the level of concentration in the market, while typically offering greater scope for efficiencies. However, this does not mean that non-horizontal mergers are completely risk-free.
Potential efficiency gains
There are several sources of potential efficiency gains with non-horizontal mergers. In the case of a vertical (supplier-customer) merger, the combined firm can avoid costs related to finding suppliers or customers, negotiating contracts, monitoring whether contractual obligations are being fulfilled, etc. In the case of firms selling complementary products targeting the same customers, a merger can lead to cost savings for market research, marketing and advertising, and distribution costs.
For complex complementary products, mergers can facilitate coordination between technical experts by bringing the entire R&D or production process in-house. For instance, developing new herbicides and complementary herbicide-tolerant seeds may require considerable coordination between researchers, and could be facilitated by bringing the teams together in a merged firm.
An additional potential benefit from non-horizontal mergers pertains to resolving spillover problems. For instance, if one firm sells herbicides and another complementary herbicide-tolerant seed, any measure to stimulate the sales of herbicides (e.g. more marketing, lower prices) would also stimulate sales of the complementary seeds. Before a merger, the herbicide firm would not take into account this spillover effect, as it incurred the full costs but did not receive the full benefits of its promotional efforts. A combined firm would receive the full benefits, and hence would have an incentive to invest more in both products.
Foreclosure as a threat to competition
Despite these efficiency effects, non-horizontal mergers could limit competition when the merged firm makes it harder for potential or current competitors to access markets (customer foreclosure) or supplies (input foreclosure).
In the case of seed markets, customer foreclosure could occur if a plant breeder acquires a large distributor of seeds. The integrated firm could refuse to distribute seed of competing firms, or could make it more difficult to do so by imposing disadvantageous terms and conditions. To restore access to these markets, competing firms would need to incur significant costs, e.g. by setting up their own distribution networks. This raises the costs of rivals and discourages entry into the industry.
Input foreclosure in seed markets could occur if a seed firm acquires a supplier of a widely-used GM trait or technique and subsequently refuses to cross-license this technology to competitors. This would force competitors to invest in developing their own technology, a costly and lengthy process which raises the costs of competitors and discourages entry.
In the specific case of complementary goods, foreclosure can also occur through bundling and tying. With bundling, products are sold as a package (with or without the option to buy the products separately). With tying, customers buying one good (e.g. herbicide) are required to purchase another good (e.g. herbicide-tolerant seeds) from the same supplier either because of contractual requirements or because products have been designed in such a way that they only work when used together.
Both bundling and tying are common practices that do not necessarily lead to anticompetitive consequences. However, a combined firm with market power in one product (e.g. herbicide) may use bundling or tying to create a strong position in another product (e.g. herbicide-tolerant seeds). In turn, this may discourage entry of potential competitors as they would now be required to invest in both markets.
For all types of foreclosure, the theoretical possibility of foreclosure does not necessarily mean it will be in the combined firm’s best interest to do so. The merged firm will need to trade off the potential loss of sales in one market (e.g. by refusing to cross-license the technology) with the potential gains in the other market (e.g. the ability to charge higher prices). However, as the European non-horizontal merger guidelines point out, even the likelihood that a merged firm could engage in foreclosure may deter potential entrants.4
Removing a potential entrant
A firm which is active in a different but related industry may have the assets, know-how, distribution network, and other necessary resources to potentially enter the market. Even when this firm is not entering the market, the threat of entry may be sufficient to act as a restraint on market participants. This situation is known as a “contestable market” (Baumol, 1982[75]).
Even a highly concentrated market could behave competitively if existing firms take into account the threat of entry by potential competitors. A merger with a non-competitor can reduce the contestability of the market by removing this threat. Mergers between competitors can also make entry by other players more difficult by increasing the costs of entry for new firms.
In the context of the seed industry, a “different but related” industry could be interpreted in various ways. For instance, a large agrochemical firm could invest to develop its own breeding programme and GM traits; a seed and biotech firm active in one region could invest to enter another region; a firm active only in maize seed and traits could invest to develop a product offering in soybean seeds and traits. In each case, the firm is a potential entrant, although the barriers to entry might differ by scenario. As these examples show, both horizontal and non-horizontal mergers could eliminate a potential entrant.
Multimarket contact
Mergers between firms active in different regions or different products could facilitate collusion by increasing the degree of “multimarket contact”. Firms competing across several markets may find it easier to collude, as both the benefits of collusion and the costs of deviating are greater (Bernheim and Whinston, 1990[76]). A merger between companies A and B which are active in different markets may lead to higher prices by making it easier for the combined firm to collude with a competitor C active in both markets.
In the context of the seed industry, multimarket contact can occur across several dimensions. Firms active in the seed industry can also compete in agrochemicals (or even in non-agricultural products); firms competing in one geography may also face each other as competitors in other geographies; firms competing in the seed market for one crop may also compete in the seed market for other crops.
The scope for multimarket contact is even greater when several mergers are taking place at the same time. A situation of four independent firms split across two markets could be transformed into a situation with two firms competing with each other in both markets. In each market separately, the degree of market concentration may not have changed; but by creating fewer players with multimarket contact, these mergers could facilitate collusion, leading to higher prices for consumers in both markets.
Empirical evidence
Market power versus efficiency effects of mergers
Despite their economic importance, there is surprisingly little recent empirical evidence on the actual effects of mergers – or on the impact of competition policy to prevent negative effects from mergers (Ashenfelter et al. (2009[77]); Ashenfelter and Hosken (2010[78]); Carlton (2009[79])).5
A recent contribution is provided by Sheen (2014[80]). Using data from Consumer Reports, an independent magazine evaluating the price and quality of consumer goods, Sheen (2014[80]) analyses 9 000 products in 20 categories (ranging from interior paint to washing machines) sold by 372 firms between 1980 and 2009. In this large database, 88 mergers are identified and their impact analysed at the level of individual products. Sheen (2014[80]) finds that after a merger, the quality of products of merging firms converges on their average quality level. However, prices for both the acquiring and the target brand fall relative to the competition. This effect is strongest when both companies were active in the same category, and price declines are greater when the merging firms have higher market shares. At the same time, these mergers lead to higher stock market valuations, indicating that lower prices did not come at the expense of profits for the merging firms. The most likely explanation is that merging firms can exploit economies of scale and/or buyer power to lower their variable production costs, and pass part of these savings on to consumers. These results suggest that even in horizontal mergers, efficiency effects may outweigh the incentive to charge higher prices to consumers.6
A more recent analysis, however, casts doubt on the existence of efficiency effects after a merger. Blonigen and Pierce (2016[81]) study productivity and mark-ups (an indicator of market power) using plant-level data of the entire US manufacturing industry from 1997 to 2007, covering 187 000 plants, and look at plant-level effects of mergers and acquisitions. They find little evidence for efficiency gains but rather a strong increase in the market power exerted by firms after the merger.7
Kwoka (2015[82]) reviews 19 studies investigating mergers in various industries. Although comparison of effects from different study contexts is not straightforward, on average these studies seem to show a price increase of 5-6% following a merger, although there is evidence of a reduction in costs of around 8.5%. Effects on efficiency measures appear to vary widely. While most estimated effects appear to be negative, on average the estimated efficiency effect appears to be modestly positive (between 0.3% and 2.2% depending on how results are aggregated). The studies for which cost and efficiency measures are reported, however, mostly focus on one sector (US hospitals).
Either way, efficiency effects that can offset price increases would more likely occur in some industries than in others. For many manufactured goods (such as those covered by Sheen (2014[80])), variable production costs represent a considerable share of the total cost. Economic theory predicts that a firm with market power will pass on part of the reduction in marginal costs to consumers in the form of lower prices. It is not clear that similar effects would hold in industries where variable costs represent only a minor part of the total cost such as software, telecommunications, or pharmaceuticals. These sectors are characterised by large up-front investments (e.g. in R&D or infrastructure), but variable production costs are typically low (or even practically zero, as in the case of software).
Multimarket contact
Empirical research has demonstrated that multimarket contact is indeed associated with higher prices. Such effects have been found for a wide range of industries, including movie theatres (Feinberg, 2014[83]), cement (Jans and Rosenbaum, 1997[84]), hotels (Fernández and Marín, 2003[85]), and airlines (Ciliberto and Williams, 2014[86]). In addition to finding evidence that multimarket contact increases prices, there is direct evidence from the airline industry showing that mergers can lead to higher prices by increasing multimarket contact (Singal (1996[87]), Bilotkach (2011[88])). In contrast, Waldfogel and Wulf (2006[89]) found no effect of increased multimarket competition in the radio broadcast industry on advertising prices.
Merger policy
An important limitation of analyses such as those by Sheen (2014[80]) is that any post-merger analysis can only look at mergers that have been approved by competition authorities. It is possible that all mergers which would have led to higher prices were blocked by the authorities, so that only mergers with considerable efficiency effects were allowed to proceed. In other words, the results showing modest or no harmful price effects after mergers could indicate that competition authorities correctly identified and prohibited all potentially harmful mergers. The question is what the price effect would have been of mergers which were not allowed to go through – and hence, whether merger policy is effective in blocking harmful mergers.
Ashenfelter and Hosken (2010[78]) shed some light on this by investigating the price effects of five mergers “on the enforcement margin” – i.e. mergers which ex ante seemed potentially anti-competitive but were nevertheless allowed to proceed by the US competition authorities. Ashenfelter and Hosken (2010[78]) compare price evolutions of brands of the merging firms with price evolutions of private-label brands and competitor brands, and conclude that prices increased by 3% to 7%.8 Given the methodology used, these effects should be considered an upper bound on price effects for mergers that were permitted, and a lower bound for mergers that were blocked. As the authors note, this result suggests that merger enforcement, to a first approximation, is not overly strict or overly lax.
A related question is whether merger policy achieves its objectives, in particular where remedies are used. The term “remedies” refers to measures used by competition agencies to resolve and prevent harm to the competitive process that may result from a merger (OECD, 2011[90]). Typically, a distinction is made between structural and behavioural remedies. A structural remedy requires the divestiture of an asset, while a behavioural remedy imposes an obligation to engage in, or refrain from, a certain conduct. The sale of a part of the business or the transfer or licensing of intellectual property rights are considered structural remedies. Examples of behavioural remedies include non-discrimination obligations, transparency provisions, or limitations on what firms may include in contracts with customers.
The strictness and effectiveness of merger policy in the United States has been examined by Kwoka (2015[82]). Highlighting the relative scarcity of empirical research on the impact of merger policy, Kwoka (2015[82]) constructs a dataset of 47 studies on the effects of individual mergers, as well as a dataset of 19 studies investigating groups of mergers. Some results from this meta-analysis were discussed earlier.9 In addition, Kwoka (2015[82]) analyses trends in behaviour of US competition agencies using publicly available data. This leads to three main conclusions.
First, merger policy in the United States is relatively lenient and has become increasingly so over time. Only a small fraction of mergers are investigated, and an even smaller fraction is challenged. Competition agencies are also more lenient than the Horizontal Merger Guidelines would suggest, except at very high levels of market concentration.
Second, mergers typically lead to price increases. In the dataset assessing individual mergers, around 80% of mergers resulted in price increases after controlling for other factors; 31% of mergers resulted in price increases exceeding 10%.
Third, and perhaps most surprising, Kwoka (2015[82]) finds that enforcement actions of competition authorities do not appear to have much impact in preventing negative effects of mergers. For all cases where the agencies challenged mergers, prices nevertheless increased on average by 7.7%. Cases where a divestiture was required saw price increases of 6% on average, while conduct remedies were associated with price increases of 13%.
Some of these findings have been challenged. Vita and Osinski (2016[91]), two researchers at the Federal Trade Commission, have highlighted what they see as methodological shortcomings in this analysis, in particular the small number of observations on which the analysis of divestitures and (especially) conduct remedies is based. Moreover, as Kwoka (2015[82]) himself indicates, it is possible that prices would have increased even more in the cases where divestitures and conduct remedies were applied. Nonetheless, given the high standards of empirical rigour imposed by Kwoka (2015[82]) in selecting the studies for his meta-analysis, these results do suggest that remedies often fail to protect or restore competitive conditions.
In the European Union, the Directorate-General for Competition (DG Competition, 2005[92]) has reviewed 96 remedies imposed between 1996 and 2000 using a combination of case files and interviews. More than 85% of the remedies studied involved a divestiture or a similar measure. Overall, the report concluded that 81% of the divestitures were fully or partially effective, while 7% were ineffective and 12% unclear.
A very different empirical approach was used by Duso et al. (2011[93]). To isolate the impact of merger remedies, they compare how the stock market reacts to initial announcements of mergers with how the stock market reacts when the European Commission makes its final decision on the merger and the required remedies. To be able to separate efficiency effects from anticompetitive (market power) effects, Duso et al. (2011[93]) look also at how share prices of competitors of the merging firms react. If a merger is expected to lead to higher prices through market power, this is likely to increase the scope for competitors to also raise prices. Conversely, if a merger would lead to efficiency gains but no market power effects, the merged firm could reduce prices and put pressure on the profits of competitors. Hence, the behaviour of the share price of competitors can be used as an indication of whether a merger is pro- or anti-competitive.
After correcting for the market’s expectation of the EU decision, Duso et al. (2011[93]) find that remedies are only partially effective, as they do not fully remove anti-competitive rents. However, remedies appear to be more effective when applied to an industry with which the European Commission has experience. Overall, Duso et al. (2011[93]) conclude that remedies might be a good policy tool when the anti-competitive concerns of a merger are not too serious, but that outright prohibitions may be the only way to restore effective competition for complex mergers with serious risks to competition.
Evidence on seed markets
For the United States, several studies have explored the determinants of prices of seed (in particular GM seed) and the link with market concentration. Particularly important in this regard are a number of studies by Guanming Shi, Jean-Paul Chavas and Kyle Stiegert on soybeans (Shi, Chavas and Stiegert, 2009[94]), maize (Shi, Chavas and Stiegert, 2010[95]), and cotton (Shi, Stiegert and Chavas, 2011[96]).
The results across these three crops are similar. First, there is evidence that a higher degree of market concentration contributes to higher seed prices. For instance, moving from a hypothetical market with perfect competition to a monopoly would increase conventional maize seed prices by USD 15 per bag. (As a comparison point, the average price of conventional maize seed in 2007 was USD 94 per bag.)
There is also evidence of cross-market effects. For example, a higher degree of market concentration for maize seeds with traits protecting against European corn borer appears to reduce the price of maize seeds with traits protecting against root worm, and vice versa. This is perhaps due to complementarity effects on the supply side. For some other combinations, however, the cross-market effect is to increase prices even further. This may be the case when two products are substitutes.
Third, seeds with stacked GM traits are typically cheaper than would be expected based on the prices of the single-trait seeds (a finding consistent with the data in Figure 2.14). GM seed is sold at higher prices than conventional seed, reflecting a price premium charged for GM traits. When looking at prices of stacked seeds (with two or more GM traits), however, this price premium appears to be consistently less than the sum of the price premiums of the same traits in single-trait seeds. This may reflect economies of scope on the supply or demand side. The difference is considerable. For example, the price premium for the trait protecting against European corn borer was around USD 26 per bag in a single-trait maize seed, while that for root worm was USD 46 per bag, or a combined premium of around USD 72 per bag. Maize seed with both traits would sell only for a premium of around USD 61 per bag, or 16% less than would be implied by simple additive pricing.
In short, these studies confirm the theoretical expectations that higher market concentration appears linked to higher prices, although efficiency effects may mitigate this effect to varying extents.
Regarding the likely effects of the recent mergers, the only published simulation of price effects is the analysis by Bryant et al. (2016[97]) for the US maize, soybean and cotton seed markets (also see the non-technical summary in Maisashvili et al. (2016[98])). Bryant et al. (2016[97]) adapt a methodology developed by Hausman et al. (1994[99]), Hausman and Leonard (1996[100]) and Hausman (2010[101]), which combines data on market shares, own-price elasticities and cross-price elasticities to simulate post-merger prices. Bryant et al. (2016[97]) report the following market shares for US seed markets:
Table 4.1. Seed market shares in the United States, 2014-15
Note: Data for maize and soybeans refers to 2014; data for cotton refers to upland cotton for 2015.
Source: Bryant et al. (2016[97])
In the years considered in their study, Bayer was not active in maize and soybean markets, whereas DuPont Pioneer was not active in cotton markets. Hence, the relevant analyses conducted by Bryant et al. (2016[97]) look at the expected impact of the DowDuPont merger in maize and soybean markets, and the expected impact of the Bayer-Monsanto merger in cotton. The methodology used by Bryant et al. (2016[97]) allows to predict the price changes for each market player separately. In addition, Bryant et al. (2016[97]) construct a market-share weighted expected price increase representing the overall change in prices.
The simulations predict that Dow would increase its prices for maize seed by 6.3% and for soybean seed by 5.8%, while DuPont Pioneer would increase its prices by 1.3% for soybean seed and 1.6% for maize seed. Overall, the expected price increase for the maize seed market as a whole would be around 2.3%, and for the soybean seed market around 1.9%.
The strongest simulated effects are for the cotton seed market, where prices were expected to increase by around 18% on average. The Bayer-Monsanto merger (using 2015 data and assuming no divestitures) would have brought together the two leading firms in the US cotton market, with a combined market share of around 70%. (Bayer sold its US cotton seed business to BASF to obtain regulatory approval of its acquisition of Monsanto).
Merger simulations have their limitations, and these apply to the analysis by Bryant et al. (2016[97]).10 The analysis is done at a high level of aggregation (the United States as a single market), assumes that only the merging firms will increase their prices, and assumes that other factors (e.g. product choice or innovation) remain unchanged.11 Overall, the analysis suggests that price increases will be greater where combined market shares are higher, and that the biggest price increase will occur for the firm with the smaller initial market share. This could be used as a rule of thumb in other markets.
4.2. Potential effects on innovation
Traditionally, the evaluation of mergers has been mostly concerned with the potential impact on prices. In recent years, however, competition authorities have increasingly focused on the potential harmful effects on innovation. For instance, in 2016 the US Department of Justice sought to block Monsanto’s proposed sale of Precision Planting LLC, a firm producing high-speed planting machines, to the farm equipment manufacturer John Deere (as discussed in Section 3.3). John Deere was the only other producer of such equipment, and the Department of Justice argued that it was the intense rivalry between the two firms which had led to the rapid introduction of innovative new features (MacDonald, 2017[102]).
However, the link between (a lack of) competition and innovation is not straightforward. Reviewing theoretical and empirical research, Aghion and Griffith (2005[103]) argue for an inverted U-shaped relationship, where both very low and very high levels of competition discourage innovation (see Aghion et al. (2005[104]), Hashmi (2013[105])). Where there is little competition, firms face no pressure to invest in innovation as there is no threat to their profits. But iIn a highly competitive market, profits from a successful innovation are competed away quickly, thus making it hard to recover the investments in R&D.
A drawback of the “inverted U” literature is that it focuses on the overall degree of competition in a market, but says little about the effects of specific mergers. Focusing on antitrust policy, Shapiro (2012[106]) reviews the literature on competition and innovation and sets forth three principles on innovation:
● The contestability principle: Innovation is stimulated if a firm has the prospect of gaining, or protecting, profitable sales by providing greater value to customers
● The appropriability principle: Innovation is stimulated if a firm can appropriate a greater share of the social benefits resulting from its innovation
● The synergy principle: Innovation is stimulated if firms can combine complementary assets, leading to greater innovation capabilities
While the contestability and appropriability principles relate to firms’ incentives, the synergy principle relates to firms’ ability to innovate. A fourth principle, which is implicit in Shapiro (2012[106]) but worth making explicit, can be added:
● The parallel paths principle: All else equal, having multiple firms seeking to innovate in the same domain increases the likelihood of discovery and introduction of new processes and products.
These principles are discussed in turn.
Contestability
As Shapiro (2012[106]) points out, the contestability principle operates if a successful innovator can expect to win profitable sales from competitors. Contestability is undermined to the extent that market shares are “sticky,” for instance because of brand loyalty among consumers or high switching costs. In such a context, firms’ incentives to innovate are reduced, as innovation would not lure away consumers from competing firms.
Similarly, an innovative product developed by a firm with a large market share may negatively affect the sales of its existing products rather than winning market share from competitors, an effect known as “cannibalisation.” This effect will be more prominent as the firm’s initial market share is greater. Again, this would mean a successful innovator could not expect to win profitable sales from competitors. If two firms merge, they would have more to lose from such cannibalisation effects and the incentive to innovate could be reduced.
The threat that a potential competitor might enter the market, however, can induce the incumbent to invest in innovation to protect its market share.12 To the extent that smaller firms could challenge the incumbent through innovation, the market is contestable and the incumbent will have an incentive to continue investing in innovation. Hence, a high market share by itself says little about incentives to innovate; rather, the contestability of the market is what drives innovation.13 Conversely, a small market share is consistent with strong incentives to innovate as long as there is a reasonable chance of capturing a large amount of profitable sales in the future.
Appropriability
Contestability is necessary, but not sufficient to ensure that a firm benefits from its innovative activities. If competing firms can freely and quickly imitate a new process or product, the innovating firm will not be able to enjoy its competitive advantage for long. Appropriability is therefore clearly linked to intellectual property rights, but imitation is not the only way in which the benefits of innovation can be dissipated. If innovation increases the demand for a complementary product, part of the benefit of innovation will spill over to producers of the complementary product. Such spillovers also reduce the appropriability of innovation. From a market structure point of view, appropriability can be increased if suppliers of complementary products merge. Such a merger could therefore increase innovation.14
Synergy
The synergy principle draws attention to the fact that innovation does not occur in a vacuum. Merging firms with complementary assets can increase innovation capabilities, for instance by improving coordination between R&D teams working on complementary products. However, as Shapiro (2012[106]) points out, “merger synergies are far easier to claim than to achieve.”
In the case of the seed industry, R&D synergies could arise from closer collaboration between teams working on plant breeding, traits, and crop protection chemicals. For instance, it may be easier to develop GM seed with stacked traits if more of these traits are developed within the same firm, as firms may be reluctant to share sensitive data with competitors. As another example, the development of crop protection chemicals requires a good understanding of local agronomic conditions. This expertise might in turn help plant breeders and geneticists in their R&D efforts.
Parallel paths
As the parallel paths principle indicates, all else equal, it is useful to have firms working in parallel on innovations. Parallel paths increase the likelihood that at least one firm will make a useful discovery. Although in theory firms can pursue multiple approaches in parallel in-house, this may prove difficult in practice due to organisational obstacles (Shapiro, 2012[106]). Moreover, if a firm is developing different approaches in-house, the firm will still have an incentive to release only one product innovation at a time.
Comanor and Scherer (2013[107]), studying the pharmaceutical industry, argue that firms underinvest in parallel R&D approaches, and that mergers often lead to shutting down “duplicative” R&D efforts. They provide evidence from pharmaceutical trials in the United States indicating that companies typically only have one molecule in trial for a disease. Comanor and Scherer (2013[107]) argue this may indicate that companies fail to appreciate the value of parallel R&D approaches, viewing these as wasteful duplication. One implication of this argument is that technological progress is maximised when there are multiple firms in an industry pursuing R&D.
Empirical evidence
General literature
The empirical literature on the effect of mergers on R&D has tended to produce mixed evidence (Cassiman et al., 2005[108]). One explanation is that earlier studies did not always distinguish between different characteristics of merging firms. Recent research, which does make this distinction, has tended to find clearer results.
Cassiman et al. (2005[108]) use an in-depth analysis of 31 mergers and acquisitions to explore the impact of firms’ technological and market relatedness. They find that where firms have complementary technologies, their R&D efforts and efficiency increase after the merger. However, where merging firms’ technologies are substitutes, there is a decrease of R&D efforts after the merger. The reduction in R&D for “substitute” firms is especially pronounced if the firms were direct rivals before the merger. The reduction in R&D occurs through various channels, including greater turnover of key employees, a more focused R&D portfolio, a shorter R&D horizon, and less internal funds provided for R&D.
Bena and Li (2014[109]) emphasise that innovation is a key motivation for mergers and acquisitions as the firms involved in mergers and acquisitions are disproportionately more likely to be involved in innovation. For instance, in their sample of public mergers in the United States between 1984 and 2006, almost two-thirds of firms were engaged in patenting activities. In contrast, when looking at all firms (merging and non-merging), only 30% report R&D expenses and less than 10% regularly delivered patentable output. Bena and Li (2014[109]) show that the likelihood of a merger is higher if firms have “technological overlap.” By comparing successful mergers with a control group of firms where an acquisition bid was withdrawn, Bena and Li (2014[109]) also demonstrate the effect of mergers on innovation. In particular, for mergers where there was a technological overlap prior to the merger, there is an increase in the combined firms’ post-merger innovation output. This result shows that where technological complementarities are driving a merger, the impact on innovation is expected to be positive.
These results confirm the principles regarding mergers and innovation proposed earlier and based on Shapiro (2012[106]). When merging firms have substitutive technologies, the tendency is to cut costs by reducing redundant efforts – hence closing down parallel paths. This effect is especially pronounced if firms were product market rivals, as the merger also removes an incentive to innovate (cfr. the contestability principle). When the technologies of merging firms are complementary, the synergy effect and the appropriability effect dominate and lead to improvements of R&D efforts and outcomes.
The pharmaceutical industry
Further insights on the effect of mergers on innovation are found by looking at the experience of the pharmaceutical industry. Similar to the seeds and biotech industry, the pharmaceutical industry is highly R&D intensive and has experienced merger waves. In an essay in Nature Reviews: Drug Discovery, John L. LaMattina, the former head of Pfizer’s global R&D organisation, has argued in an essay in Nature Reviews: Drug Discovery that these mergers are an important driver behind a fall in innovative activity (Lamattina, 2011[110]). He argued this was not always the case; the 1989 merger between Bristol Myers and Squibb did not lead to major R&D cuts. However, during Pfizer’s acquisitions in the early 2000s, it closed numerous research sites, including those where major discoveries had been made. LaMattina (2011[110]) also points out that before Pfizer’s merger with Wyeth in 2009, the two firms had a combined R&D budget of USD 11 billion, whereas the firm post-merger had the goal to reduce the R&D budget to USD 6.5-7 billion. LaMattina (2011[110]) concludes that in pharmaceuticals, “industry consolidation has resulted in less competition and less investment in R&D.”
The negative effects of mergers on R&D in the pharmaceutical industry were confirmed empirically by Omaghi (2009[111]). Using a dataset covering the universe of large pharmaceutical firms between 1988 and 2004, Omaghi (2009[111]) found that merged firms spend less on R&D and file fewer patent applications than a control group of comparable firms.15 Omaghi (2009[111]) also attempted to estimate the role of product-market relatedness and technological similarity, although the dataset did not allow for a proper distinction between substitutes and complements.
The seed and biotech industry
There is relatively little literature on the specific impact of market concentration on innovation in the seed industry. An exception is Schimmelpfennig et al. (2004[112]), who focus on maize, cotton and soybeans in the United States and report an inverse relationship between firm concentration and R&D intensity. While an important contribution to the literature, there are several methodological aspects of the study which raise doubts around the validity of its findings.
The structural approach used by the authors to estimate the effect of concentration on R&D intensity assumes all firms are symmetric and hence have equal market shares. The authors’ measure of market concentration is therefore defined as (1/n) where n is the number of firms conducting field trials, regardless of their actual market shares. The R&D intensity is measured as the number of registered biotech field trials divided by a measure of market size. While the authors find a negative correlation, the interpretation of this result is problematic. During most of the period under study (1989-1998), the number of firms conducting field trials increased strongly. At the same time, the number of field trials increased as well. Hence, the negative correlation between market concentration and R&D intensity is most likely picking up the simultaneous increase in the number of firms conducting trials and the total number of trials in the early 1990s. This study therefore does not lend strong support to the view that market concentration in the US seed and biotech industry negatively affects innovation.
A related paper by Brennan et al. (2005[113]) looks at market concentration in US field trials for biotechnology as an indicator of competition in the “innovation market.” They calculate a Hirschman-Herfindahl Index (HHI) of the “market shares” of different firms in the total number of field trials, which can be seen as a proxy for crop-specific R&D effort.16 Between 1988 and 1996, there was a downward trend in the HHI. However, the concentration in field trials grows strongly after that date, increasing from less than 1 500 in 1996 to about 5 000 in 2002. The increase is driven in large part by Monsanto. In 1996, it was responsible for less than 30% of field trials, but by 2002, Monsanto conducted 70% of all field trials. Brennan et al. (2005[113]) show that a similar pattern is found when looking at ownership of agricultural biotechnology patents. Furthermore, “market shares” in patents appear robust over time. The authors conclude there is a risk of high concentration negatively affecting innovation.
An updated version of the “field trial” analysis by Brennan et al. (2005[113]) is found in Figure 4.1, showing the HHI and the four-firm concentration ratio for approved field trials in US biotech. There was indeed a strong increase in concentration in the late 1990s. However, it appears that this trend reversed around 2002-2004. The HHI, which had increased to a high level of 5 000 points in 2002, fell to levels comparable to those seen in the early 1990s. Similarly, the four-firm concentration ratio fell from 80% back to around 62% in recent years. Based on these data, it could be argued that Brennan et al. (2005[113]) captured what seems to have been a temporary event.
Figure 4.1. Concentration in US field trials for biotech, 1985-2017
Note: Data up-to-date to 4 December 2017. Scope includes all applications for permits and all notifications regarding “Release” (including e.g. combinations of Release & Import, Release & Interstate) which have been approved or accepted (for permits) or acknowledged (for notifications), and which were submitted by a private firm (defined as any entity which is not a university or a US government agency). Year is defined by the date the application was received.
Source: United States Department of Agriculture, APHIS BRS data file.
Underlying these numbers is a major structural change (Figure 4.2). Between the early 1990s and the early 2000s, Monsanto strongly increased its field trials until it accounted for more than half of all field trials. After 2002, however, Monsanto’s annual number of field trials fell significantly, accounting for nearly all of the decrease in field trials during this period. Excluding Monsanto, there is no clear downward trend in the number of field trials in the post-2002 period. The results are nearly identical if applications rather than approvals are used.17 In other words, both the increase and decrease in concentration of field trials (whether measured with the HHI or the four-firm concentration ratio) reflect the impact of Monsanto. Figure 4.3 shows the same data expressed per crop. The rise and fall of field trials is largely explained by variations in field trials for maize. Over the entire period, field trials for maize accounted for about 45% of all field trials, followed by soybeans (13%). The peak in soybean field trials (around 2008) appears to occur a few years later than the peak for maize (around 2002).
Figure 4.2. Number of approved US biotech field trials per firm, 1985-2017
Note: Data up-to-date to 4 December 2017. Scope includes all applications for permits and all notifications regarding “Release” (including e.g. combinations of Release & Import, Release & Interstate) which have been approved or accepted (for permits) or acknowledged (for notifications). “Public” includes universities and US government agencies. Year is defined by the date the application was received.
Source: United States Department of Agriculture, APHIS BRS data file.
Figure 4.3. Number of approved US biotech field trials per crop, 1985-2017
Note: Data up-to-date to 4 December 2017. Scope includes all applications for permits and all notifications regarding “Release” (including, for example, combinations of Release & Import, Release & Interstate) which have been approved or accepted (for permits) or acknowledged (for notifications). Year is defined by the date the application was received.
Source: United States Department of Agriculture, APHIS BRS data file.
It is not clear what explains the large variation over time in field trials, although it has been suggested that the overall decrease in field trial applications may reflect maturity of the “first generation” of GM and a slow introduction of subsequent generations of GM crops (Anderson and Sheldon, 2017[55]). At any rate, it is clear that the increase and decrease in measured concentration in field trials directly correspond to the increase and decrease in field trials by Monsanto. Looking at overall market averages, this “composition effect” would give the misleading impression that high concentration is associated with a higher rate of innovation. More generally, these findings mean that caution is needed when calculating and interpreting concentration measures based on field trial data.
Oehmke and Naseem (2016[69]) studied mergers, market structure and innovation in the US seeds and biotech industry using the same field trials data described here. They look at field trials for tomatoes, cotton, maize, soybeans, and potatoes over the period 1987-2012. The results show that the total number of organisations conducting field trials has a positive and statistically significant effect on the number of field trials conducted per year. Moreover, the number of public sector organisations conducting trials appears to have a positive impact. However, the same is true for the four-firm concentration ratio of field trials. Hence, both the number of firms and the concentration of field trials among a small number of firms appear to contribute to a larger number of field trials overall.
The four-firm concentration ratios used by Oehmke and Naseem (2016[69]) are not based on market shares but on the number of field trials per firm. At the same time, Oehmke and Naseem (2016[69]) use the total number of field trials as their dependent variable. As the previous analysis shows, this can lead to misleading conclusions. If one firm ramps up its R&D efforts faster than others (as Monsanto did in the late 1990s) this increases the total number of field trials, but may also lead to a higher concentration ratio. Conversely, when a single firm with a large number of field trials winds down its programmes faster than others, this gives the impression of a reduction in concentration coinciding with a decline in innovation. An ideal setup would measure how a high degree of concentration in actual market shares (defined by sales) affects subsequent innovation. Hence, it is difficult to assign a clear interpretation to this finding.18
As noted by Chan (2011[114]), studies relying exclusively on field trials ignore innovation taking place in conventional breeding. Chan (2011[114]) looks at the number of new varieties (as measured by patents or plant variety protection certificates) introduced by a sample of 115 firms in the United States between 1976 and 1999. Firms with larger patent portfolios create greater numbers of new varieties, but there is no evidence of economies of scale; in fact, firms appear to experience diminishing returns. Firms that have merged do not appear to produce a significantly greater number of new varieties, nor do they experience greater economies of scale in comparison with firms that did not merge. Chan (2011[114]) concludes that these results do not support the argument that firms should be allowed to merge to increase efficiencies in innovation. On the other hand, neither do the results imply that mergers would negatively affect innovation. In fact, there appears to be evidence that a more monopolistic structure (as measured by concentration of patents) has a positive effect on rival firms’ innovation.
In summary, the empirical literature broadly confirms the principles identified by Shapiro (2012[106]), including the “parallel paths” principle. In contrast, there appears to be little evidence so far of a negative link between market concentration and innovation in the seeds and biotechnology industry, based on historical data.19 This does not imply that the current merger wave poses no risks to innovation, however.
Implications for merger policy
A merger may simultaneously impact all the principles in contradictory directions. For instance, a horizontal merger between rivals may reduce the contestability (by removing a competitor), increase the appropriability (by reducing spillovers), and potentially lead to innovation synergies – while also closing off parallel paths. Shapiro (2012[106]) suggests that a horizontal merger with possible innovation effects should be evaluated according to the following questions:
Will a merger between two rivals significantly reduce their incentive to innovate (by reducing contestability and/or appropriability)?
If so, is there an offsetting effect through an enhanced ability to innovate (through synergies)?
A similar approach has been proposed by Gilbert and Sunshine (1995[115]) and is known as an “innovation markets” approach.
As Shapiro (2012[106]) notes, if a horizontal merger leads to a highly concentrated market, antitrust law uses a “rebuttable presumption of harm” – that is, the merger is assumed to be harmful unless demonstrated otherwise. Although Shapiro (2012[106]) does not mention explicitly whether this principle should also apply to the specific question of innovation, the parallel paths principle could be interpreted as justifying such an approach.
4.3. Potential effects on product choice
With a horizontal merger, two firms are potentially offering products to the same set of customers. The newly merged firm may decide to reduce its product offering. This section briefly reviews the literature on such possible effects.20 In the case of the seed industry, some observers fear that reducing product choice may have negative ecological effects by diminishing genetic diversity (Mammana, 2014[66]). This topic is also briefly discussed.
Economic effects
Only a handful of papers have studied the question of how mergers affect product choice.21 The conclusion of both theoretical and empirical work is that the effects are ambiguous and context-dependent.
Berry and Waldfogel (2001[116]) studied the effect of mergers on product choice in the US radio broadcasting sector. In theory, this effect is ambiguous. If post-merger, two radio stations with the same owner are similar, the owner could shut down one of the stations to save costs and avoid competition between the two stations (“cannibalisation”). Alternatively, the owner could reposition one of the two radio stations to increase the differentiation and attract a different audience. A third possibility is that the owner differentiates his two radio stations, but not too much in order to avoid opening up a niche for competitors to enter.22 Empirically, Berry and Waldfogel (2001[116]) find that consolidation reduced the number of stations, but the stations themselves became more diverse. The combined effect was for consolidation of ownership to increase product choice.
Theoretical work by Gandhi et al. (2008[117]) confirms that firms may indeed have an incentive to change the positioning of products post-merger to reduce the problem of cannibalisation. However, as Gandhi et al. (2008[117]) point out, product repositioning may in reality be expensive and time-consuming, thus preventing or reducing the repositioning effect. Given the long time lags involved in developing new varieties and/or new traits, any “repositioning” in the seed industry is likely to take at least several years. It therefore seems possible that post-merger, a firm with two similar breeding programmes will decide to discontinue one of the breeding programmes. Over time, this would lead to fewer varieties for farmers to choose from, as well as a slower rate of innovation.
Ultimately, whether firms decide to shut down a “duplicate” offering will depend on the specific circumstances. Such an outcome is more likely if the initial products are similar, if the cost of “differentiating” the two products is high, if there is limited scope to differentiate in a meaningful way, and if there is little reason to believe that competitors will introduce a new offering in the segment.
Empirical evidence on seed markets
Only a handful of studies have investigated the issue of product choice in seed markets, and all of these focus on the US market. Schenkelaars et al. (2011[29]) point out that between 1997 and 2008 the number of maize hybrids available in the US market increased from around 3 000 to 4 300, while the number of soybean varieties almost doubled from 650 to 1 130. Over the same period, the total acreage devoted to these crops remained roughly the same, nor did the number of firms active in the seed market increase. Hence, the greater number of varieties seemed to be driven by more rapid product introductions by seed firms.
Data on the average number of varieties available per crop reporting district (CRD) in the United States are provided by Ciliberto et al. (2017[24]) (Figure 4.4). These data confirm the argument of Schenkelaars et al. (2011[29]). For both maize and soybeans, the number of conventional varieties started to decline after 1996 as more farmers adopted genetically modified varieties. Data shows a strong growth in the number of GM varieties. The total number of varieties available temporarily increased in both markets due to the co-existence of conventional and GM varieties, before reaching a peak and then stabilising at a level higher than in 1996. For maize, there were 16.7 varieties available in 2011 compared with 6.5 in 1996 (i.e. 2.6 times as many), while for soybeans there were 6.2 varieties available in 2011 versus 4.3 in 1996 (i.e. 1.4 times as many).
Magnier et al. (2010[118]) provide an empirical analysis of product life cycles in the US maize seed market. In 1998, a new variety introduced on the market remained available for an average of 4-5 years. This expected lifetime declined gradually between 1998 and 2003, and then more strongly between 2004 and 2007 (a period which corresponds to the strong increase in the number of varieties in Figure 4.4). Varieties introduced in 2007, for example, could only expect to remain on the market for 2.5–3.5 years on average. This decline affected both conventional and GM varieties. These trends suggest intensifying competition between 2004 and 2007, although the decline in the number of maize varieties since then may imply at least a partial reversal. Building on these results, Ma and Shi (2013[119]) incorporated information on market conditions into the analysis. They found that between 2000 and 2007, products from a firm with a larger market share had a longer expected lifetime. Similarly, integrated firms (i.e. firms combining germplasm and traits) performed better, potentially because they had access to higher-quality germplasm.
As few studies have looked into the matter of product choice in seed markets, it is difficult to generalise from these findings to other geographies, other crops, or other time periods. However, the ongoing consolidation of the US maize seed industry in the 1990s and 2000s does not appear to have reduced the number of product varieties or the speed of new product introductions. However, the studies reviewed here only include data until 2007 (for the product life cycle studies) or 2011 (for the number of varieties in Figure 4.4). It is possible that product choice and dynamism may have deteriorated in recent years, or could do so after a merger.
Figure 4.4. Number of maize and soybean varieties in the United States, 1996-2011
Note: Charts show the average number of seed products per crop reporting district (CRD). A CRD is a region identified by the United States Department of Agriculture’s National Agricultural Statistical Service, with relatively homogeneous agronomic and climatic conditions. On average, the dataset used here contains 242 CRDs per year (Ciliberto, Moschini and Perry, 2017[24]).
Source: Ciliberto et al. (2017[24]), using GfK Kynetec data.
Effects on genetic diversity
Maintaining genetic diversity is important both as an input into plant breeding (to ensure sufficient variation to select from) and as an output (to ensure a resilient ecology). Genetic homogeneity in agriculture may create a vulnerability to certain pests and diseases, for instance. Some observers (e.g. Mammana (2014[66])) have expressed concern that mergers could reduce genetic diversity. However, as the literature reviewed in this section shows, this is by no means obvious. Before discussing this question, it is necessary to review the more general literature on genetic diversity and plant breeding.
Conceptual issues
Plant breeding relies on the creation of variation and subsequent selection. Hence, while plant breeding is in one sense a process of reducing genetic diversity through selection, breeding can also deliberately increase diversity through “base broadening”, the practice of incorporating diversity from wild relatives or landraces (van Etten et al., 2017[120]); as well as through the creation of new variation (e.g. through mutagenesis, genetic modification or genome editing).
To analyse the impact of plant breeding on genetic diversity for agricultural crops, two conceptual distinctions must be made. First, there is a question of the proper geographical scale at which to define genetic diversity. Regions which were formerly distinct and homogeneous could gradually become more similar. This would increase diversity within each region, but could be interpreted as reducing diversity across regions as these become less distinct, with unclear implications for global diversity. It appears that such a process has occurred regarding the global diversity of food crop species between 1961 and 2009 (Khoury et al., 2014[121]). Food crop supplies have become more diversified at the national level, as the number of crops contributing to national food supplies has increased, the relative contribution of these crops has become more even, and the dominance of the most important commodities has decreased. At the same time, countries have become more similar in terms of their food crop supplies, so that the total diversity of food crops globally has narrowed.
Second, even within a specific region, genetic diversity operates at different levels: the diversity of alleles within a variety, of varieties within a crop, and of crops grown in a region (van de Wouw et al., 2010[122]). All three levels are relevant for agricultural resilience and as a store of genetic material which can be useful for future plant breeding efforts, but different mechanisms or trends can affect these three levels yielding a complicated overall picture. For instance, Bonneuil et al. (2012[123]), studying genetic diversity of bread wheat in the French département of Eure-et-Loir, found that overall diversity decreased compared to the 19th century even though the number of distinct wheat varieties grown had increased. A greater diversity in terms of the number of varieties grown can lead to less genetic diversity when new varieties are genetically more similar and/or when they displace more heterogeneous varieties, and/or when a small number of varieties becomes dominant in the landscape.
The impact of plant breeding on genetic diversity within a crop
From a historical perspective, it seems plausible that the diversity within a crop has been reduced over time by a number of “genetic bottlenecks” as well as through slower-moving trends (van de Wouw et al., 2010[122]). A first bottleneck likely occurred through domestication, as only a subset of the full range of wild ancestors is selected and reproduced. A second bottleneck is related to geographical dispersal of crops, as typically only a subset of the broader population is transferred to new regions (e.g. the introduction of coffee to South America was based on a single tree). Following such bottlenecks, selection of desirable characteristics could further reduce genetic diversity over time. However, for some species a significant “gene flow” is possible between domesticated and wild varieties. In addition to spontaneous (de novo) genetic variation, this would tend to increase the genetic diversity over time.
Van de Wouw et al. (2010[122]) hypothesise the existence of a third bottleneck related to modernisation as improved cultivars replace land races, but point out that scientific evidence on this hypothesis is hard to find. According to Van de Wouw et al. (2010[122]), the most likely scenario seems to be that the introduction of modern cultivars led to an initial increase in diversity at the regional level as new cultivars supplemented existing genetic diversity in landraces, followed by a decrease as new cultivars substituted for landraces (a pattern similar to that shown in Figure 4.4 for the introduction of GM varieties in maize and soybean). Once this transition was completed, the evidence seems to suggest no further reduction in genetic diversity.
A meta-analysis by the same authors of 44 studies (Van de Wouw et al. (2010[124])) indeed suggests that in the long run there has been no significant decrease in the genetic diversity of crop varieties released by plant breeders, although there was some variation over time (with a notable decline in the genetic diversity of crops released in the 1960s compared to the preceding decade). Importantly, these analyses look at varieties released by plant breeders, and hence do not take into account the popularity of different crops and varieties. Van de Wouw et al. (2010[122]) emphasise that farmers’ choices among crops and varieties play an important role in determining the overall diversity in the field.
Regarding the impact of modern plant breeding, Van de Wouw et al. (2010[124]) concluded that it was not clear whether an active breeding programme contributes to maintaining a high level of diversity or on the other hand leads to genetic erosion, while Rauf et al. (2010[125]) argued that the introduction of hybrids led to a reduction of genetic diversity. Fu (2015[126]) has argued that a reduction in genetic diversity is plausible, but that the impact of modern plant breeding on genetic diversity is still poorly understood from both the empirical and the theoretical point of view.
Orphan crops
While the link between plant breeding and genetic diversity within a crop is not clear, it does seem plausible that the threat of “genetic erosion” is highest for those crops in which breeders are not interested, as these crops might gradually disappear altogether (van de Wouw et al., 2010[124]).
Some crops with minimum commercial value may not receive much plant-breeding effort despite their importance in terms of genetic diversity, food security, or livelihoods in some parts of the world. Such crops have been labelled “underutilised,” “neglected,” “minor” or “orphan” crops. Many of these crops may have a high potential commercial value which goes unrealised because of a lack of scientific knowledge and recognition of the benefits of the crop downstream (i.e. the lack of a market for the final product), or because of market imperfections and market failures (Gruère, Giuliani and Smale, 2006[127]).23
A similar situation is found in the market for pharmaceuticals, where commercial incentives would not normally favour R&D on treatments for rare diseases. For this reason, government policies in many countries stimulate research on such “orphan drugs” (Franco, 2013[128]). Serious policy challenges exist to stimulate research at the global level on diseases such as malaria or tuberculosis and other diseases that affect the developing world (Kremer and Glennerster, 2004[129]).
Empirical evidence confirms that the size of the market is an important determinant of the degree of innovation in the pharmaceutical industry (Acemoglu and Linn (2004[130]); Dubois et al. (2015[131])). Similarly, evidence for the French seed industry suggests that the number of new varieties introduced each year is positively correlated with market size, although this correlation disappears for hybrid crops (Charlot et al., 2015[132]).24 The link between market size and innovation suggests that commercially less important crops indeed receive less attention from plant breeders, which in the long run could make these crops even less attractive to farmers.
Implications for consolidation
As the preceding discussion shows, the question of maintaining or improving genetic diversity in modern plant breeding is complex and depends in large part on the crop mix and on technological developments. To assess the potential effects of consolidation, a further conceptual distinction should be made. While there is clearly some link between the number of varieties introduced into the market and the degree of genetic diversity, the link is not clear. For instance, a large number of firms might be selling similar varieties, as could occur in systems where public R&D is responsible for the development of new varieties which are then commercialised by independent private firms. Conversely, a small number of suppliers could provide genetic diversity over time by ensuring rapid varietal turnover (thus avoiding the build-up of resistance among pests).
Hence, it is a mistake to automatically equate a smaller number of suppliers with a reduction in genetic diversity. As Louwaars et al. (2009[133]) have pointed out, the results of Van de Wouw et al. (2010[124]) show that genetic diversity increased between 1970 and 2000, a time of strong consolidation in the seed industry. Louwaars et al. (2009[133]) credit technological developments for this observed increase. However, they also warn that other technological developments, in particular the emergence of more precise breeding techniques, could reduce genetic diversity in the future.25
Louwaars et al. (2009[133]) also point out that a consolidated breeding programme which spans different regions could increase diversity in each region while reducing diversity at the global level. Less competition could reduce the breeding of specifically adapted varieties. However, it is not clear that a firm post-merger would find it profitable to close down specific breeding programmes. A larger firm could benefit from economies of scale such as investments in costly technology and could more easily focus on breeding for a crop with a smaller market, or for a specific geography or soil type.
These arguments apply not only to genetic diversity within a crop (e.g. plant breeders’ incentives to tailor a new variety to specific agro-ecological conditions), but also to research efforts into crops with smaller markets. It is not clear whether a merged firm would find it profitable to shut down “niche” plant-breeding programmes.
In summary, the connection between consolidation and genetic diversity is ambiguous. While various mechanisms in theory could connect the number of firms to the overall level of genetic diversity, there is little empirical evidence to draw any strong conclusions in this regard.
4.4. Conclusion
This chapter reviewed literature on the potential effects of mergers on prices, innovation, and product choice. While information on product choice is limited, clear conclusions emerge concerning the potential effects on prices and innovation.
Similar “risk factors” determine potential harm in both cases. As a result, analysing when a merger has a negative impact on innovation corresponds closely to analysing when a merger may lead to higher prices. A reduction in contestability which allows a firm to charge higher prices would also allow it to reduce R&D efforts. However, if products are complementary, appropriability may increase as the firm now reaps the full benefits of innovations and has a greater incentive to innovate. Finally, synergies are most likely to occur when there are complementarities in technology.
Horizontal mergers in the seed industry are unlikely to lead to large appropriability or synergy effects on innovation, while they would eliminate parallel paths to R&D and may affect contestability. By contrast, non-horizontal mergers are more likely to lead to positive innovation outcomes (or to avoid negative outcomes) when there are complementarity effects. The central questions for the analysis of the current mergers are whether and in which markets the mergers are horizontal or non-horizontal; to what extent they eliminate a potential entrant; and whether there is sufficient evidence to expect efficiency gains through complementarities.
At a high level of aggregation, the recent mergers appear complementary. The merging firms have different profiles both in terms of their focus on seeds versus agrochemicals, and in terms of geographic footprint. At the same time, aggregate numbers may hide important horizontal effects in specific markets. The correct way of evaluating mergers is therefore to analyse outcomes market by market, using an appropriate market definition. Analysis at this level of granularity is what competition authorities do in the course of a merger review. During such a review, competition authorities can access confidential materials held by the merging firms, such as sales data, pricing information or estimates of market shares. This makes it possible to conduct a thorough review. For instance, in reviewing the Bayer-Monsanto merger, the European Commission evaluated more than 2 000 product markets, and analysed 2.7 million internal documents of the companies (European Commission, 2018[134]). Such a detailed analysis is clearly beyond the scope of the present report, but it is possible to go beyond highly aggregated numbers cited in the public debate by using market data from private providers, as is done in the next chapter.
Notes
← 1. See, for example, the European Commission’s guidelines on horizontal mergers (Official Journal C 31, 05.02.2004) and non-horizontal mergers (Official Journal C 265, 18.10.2008). Likewise, the Department of Justice and the Federal Trade Commission have published horizontal merger guidelines (the latest version dating from 2010) as well as guidelines on non-horizontal mergers (dating from 1984). The discussion of price effects here will broadly follow the treatment in the European and American merger guidelines. Merger guidelines set out in broad terms the reasoning and main analytical techniques and types of evidence used by competition authorities and offers a good introduction to the economic issues regarding mergers. However, analyses of actual mergers are always based on the specifics of the situation, and hence highly fact-intensive and context-dependent.
← 2. Examples include situations where a merged firm would have control over patents or brands allowing it to make expansion or entry by rivals difficult. Likewise, in markets where interoperability between platforms is important (such as energy or telecommunications), the merged firm may be large enough to unilaterally set standards in the market in such a way as to raise the costs of its rivals. The merged firm may even be large enough to refuse to interconnect with rivals altogether.
← 3. To understand this, assume a firm with market power facing an inverse demand curve given by , constant marginal costs given by and fixed costs given by . Total profits are then given by where is the output level of the firm. Maximising profits with respect to leads to the optimal pricing rule , where denotes the elasticity of demand. The optimal price thus depends on the marginal cost and the elasticity of demand , but not on the fixed cost . A similar result holds for the standard Cournot-Nash oligopoly model, which takes into account strategic interactions between competitors. Reductions in fixed costs are hence less likely to be passed on to consumers in the form of lower prices, in contrast with a reduction in marginal costs.
← 4. Official Journal of the European Union 2008 C 265/07, point 49.
← 5. The discussion here focuses on effects on prices and efficiency, ignoring other effects such as those on employment studied by McGuckin and Nguyen (2001[223]), Lehto and Böckerman (2008[224]) and Siegel and Simons (2010[225]). An older literature explored the price and efficiency effects of mergers and takeovers. Notable contributions include Ravenscraft and Scherer (1987[226]), (1989[229]); Scherer (1988[230]), Lichtenberg (1992[227]), and Kaplan (2000[228]).
← 6. Winston et al. (2011[217]) show that mergers of railroads in the western United States have not had a negative effect on the cost of transporting grain. These results suggest that efficiency gains offset the increased market power of the merging firms.
← 7. It is possible (although not analysed explicitly by Blonigen and Pierce (2016[81])) that these market power effects explain the lower prices found by Sheen (2014[80]). Mark-ups depend on the ratio between output price and marginal production cost. After a merger, a firm may use its buyer power to negotiate lower prices with its suppliers. Subsequently, it may pass on part (but not all) of this cost saving to consumers. This mechanism would imply lower prices for consumers and higher profitability (as found by Sheen (2014[80])) combined with greater mark-ups and no evidence of efficiency gains (as found by Blonigen and Pierce (2016[81])).
← 8. Ashenfelter and Hosken (2010[78]) point out that these are short-run price effects and hence may not capture more long-term efficiency effects, such as those found by Sheen (2014[80]).
← 9. A study on the effect of an individual merger typically compares the prices of products of the merging firms before and after the merger with those of competitors. A grouped-merger study, by contrast, would construct a larger dataset with observations of many firms, then compare price changes between groups of merging firms and all non-merging firms. Grouped-merger studies thus look at an “average” impact across different mergers.
← 10. Weinberg (2011[216]) studied whether commonly used merger simulation techniques would have led to good predictions of the price effects of one of the mergers assessed by Ashenfelter and Hosken (2010[78]). He shows that two commonly used techniques considerably underestimate the actual price effects. These results are in line with earlier work by Peters (2006[218]), who showed that observed price changes in five airline mergers sometimes deviated considerably from what would be predicted ex ante (sometimes overstating and sometimes understating the actual price changes). Hence, while merger simulation is an important tool, caution is required in interpreting the results.
← 11. In addition, the analysis by Bryant et al. (2016[97]) uses the seed market share as its measure, which does not fully reflect the competitive situation. In 2014, Bayer did not have any soybean sales according to this measure, but Bayer was active in soybean markets through its LibertyLink (glufosinate-tolerant) trait. LibertyLink has grown strongly in popularity, from 6% market share in 2015 to a projected 20% in 2018 (Bayer, 2017[221]), which includes traits in non-Bayer seed. Bayer soybean seed and LibertyLink traits are part of the business sold to BASF.
← 12. This idea has been popularised in the management literature by Clayton Christensen as “the innovator’s dilemma” (Christensen, 1997[222]). A large firm may have little incentive to innovate given the risk of cannibalisation, but competitors with a smaller market share have little risk of cannibalisation and are therefore more likely to introduce disruptive innovations. The incumbent firm therefore faces the dilemma to either innovate and cannibalise its own sales, or lose market share to competitors who innovate.
← 13. In fact, the endogenous sunk cost argument discussed earlier implies that a high degree of contestability can lead to high endogenous fixed costs for R&D and hence a high degree of market concentration – together with a high sustained rate of innovation.
← 14. In the context of seed markets, the practice of farm-saved seed limits the appropriability of the benefits of innovation. The introduction of hybrid maize in the 1930s dramatically changed appropriability, as the seed produced by F1 hybrids gives maize of poorer quality. With hybrid maize, farmers for the first time had a clear incentive to purchase new maize seed every year, thus greatly increasing the share of the benefits flowing to plant breeders even in the absence of intellectual property rights.
← 15. A working paper by Haucap and Stiebale (2016[214]), focusing on the pharmaceutical industry in Europe, obtains similar results and finds that R&D expenditures of non-merging rivals decrease.
← 16. The Hirschman-Herfindahl Index is a standard measure of market concentration, calculated as the sum of the squares of the market shares of all firms. For a monopolist, the HHI therefore has a value of 10 000. In a market with 100 firms, each of which has 1% of the market, the HHI would have a value of 100. The HHI is often used by competition authorities. For instance, the Horizontal Merger Guidelines of the US Department of Justice and the Federal Trade Commission indicate that markets are generally considered ‘highly concentrated’ if the HHI is above 2 500, while markets are regarded as “unconcentrated” if the HHI is below 1 500.
← 17. “Approval” here means the approval to start field testing and not the final approval for commercial use. See Fernandez-Cornejo (2004[72]) for a discussion of the regulatory procedures and the distinction between notifications and permits.
← 18. Oehmke and Naseem (2016[65]) also look at the impact of mergers. When both firms are conducting field trials for the same crop, mergers do not impact the industry-wide number of field trials. When both firms are conducting field trials but in different crops, mergers appear to reduce the industry-wide number of field trials. Finally, when one of the firms is not conducting any field trials, mergers increase the industry-wide number of field trials. In contrast with the studies on mergers and R&D mentioned earlier, the analysis in their study is done by regressing industry-level outcomes on the cumulative number of mergers over time. This approach ignores the relative size of merging firms. Moreover, it is difficult to identify a causal relationship, as many other factors might explain the observed correlations.
← 19. A possible explanation is that innovation in plant breeding involves relatively long lags. In conventional plant breeding, for instance, it typically takes around ten years from the first crossing to the first commercial sales (KWS, 2017[35]). With such long lags, firms may prefer to keep R&D programs in place, since it would be difficult to quickly ramp up innovation in response to unexpected moves by competitors.
← 20. In the economic literature, the term “variety” is typically used in this context. To avoid confusion with plant varieties, this study will use the term “product choice” instead.
← 21. See in particular empirical research by Berry and Waldfogel (2001[116]) and Sweeting (2010[211]) on US radio stations and by Argentesi et al. (2016[212]) on the merger of two Dutch supermarket chains; and theoretical contributions by Gandhi et al. (2008[117]) and Mazzeo et al. (2012[213]).
← 22. A spatial example may help illustrate these possibilities. If two grocery chains merge, they may re-evaluate the location of their stores. In some neighbourhoods, two stores may be located close together and it may be optimal to close one store. Another possibility is to relocate one of the stores to a different neighbourhood to attract new customers. However, if the two stores are then located too far apart, this may create an opportunity for a competitor to open a store halfway between the two, stealing business from both. Hence, it may be optimal to relocate one store, but not too far away. Such spatial analogies are common in the literature on horizontal differentiation.
← 23. In the early 2000s, research on underutilised species was stimulated by the Global Facilitation Unit for Underutilized Species (www.underutilized-species.org), a joint initiative by FAO, IFAD, Bioversity International, the International Center for Underutilised Crops (ICUC), and the German Federal Ministry for Economic Cooperation and Development (BMZ). Since 2009, this initiative has been undertaken by Crops for the Future (www.cffresearch.org), based in Malaysia.
← 24. Unpublished research by Sébastien Parenty (INRA) using data on French seed markets similarly finds a positive correlation between market size and the total number of varieties available for sale.
← 25. Genetic modification and New Plant Breeding Techniques allow the precise introduction of a desired trait, and that trait alone, in an existing variety. In contrast, traditional plant breeding techniques rely on, for example, crossing with wild relatives or landraces, or on mutagenesis, both of which introduce a much wider “genetic load” than just the desired trait. While the new techniques improve precision, they also reduce the unintended increase in genetic diversity caused by the plant breeding process.