Most of the empirical analysis on the determinants of vertical integration has focused on efficiency motives

A common approach is to investigate the relationship between certain market characteristics – such as those associated with the importance of non-contractible relationship-specific investments – on the one hand, and the incidence or prevalence of vertical integration on the other. Numerous studies have found a significant relationship between non-contractible investment requirements and vertical integration.1 This has provided support to the transaction cost and property rights theories of vertical integration represented by Williamson , Klein et al. , and Grossman and Hart . In principle, the vertical foreclosure motive can also be explored through similar methodology – for instance, by examining whether markets that are more susceptible to foreclosure are characterized by higher rates of vertical integration. However, market characteristics associated with vulnerability to foreclosure, such as the level of market concentration, also tend to be related to the degree of relationship specificity in investments. Thus, studies that find a positive relationship between market concentration and vertical integration attribute those findings to efficiency rather than foreclosure motives . 2 To date, the significance of bandwagon effects as a cause of vertical integration has received little attention from empirical researchers. This is despite suspected cases of bandwagon behavior often being discussed and documented in business and legal circles.

For example, industry executives in the cement and ready-made concrete industries, which experienced a vertical merger wave during the 1960s, weed growing systems justified their vertical integration decisions as an inevitable response to increasingly integrated rivals . A more recent example is the acquisition of Kinko’s by Fedex in 2004. The shipping company’s acquisition of the office services provider, which enabled the former to access small-business owners and other customers more directly, was seen by commentators as a response to rival shipper UPS’s acquisition of Mail Boxes Etc., another office services provider .3 This chapter looks at the causes of vertical integration in the US generic pharmaceutical industry. This industry consists of a number of markets, each identified by a particular drug product. Each market starts off as a patent-protected monopoly served by an originator pharmaceutical company – also called an innovator or brand-name firm. New markets open up to competition by generic manufacturers at different points in time, following the expiration of patents and other exclusivities held by the originators. This competition has a significant impact on the market price of drugs. Berndt and Aitken find, in a sample of nine drug markets that went generic during 2006-2008, that the daily cost of drug treatment fell by 50.1 percent on average in the first two years after generic entry. The same study finds that since 2007, the average volume-based share of generic products has been higher than 90 percent in markets where they exist. By generating large cost savings for consumers and insurers, generic competition has successfully reversed an earlier trend – observed up to the early 2000s – where pharmaceutical expenditure growth outstripped growth in the quantity of drugs being prescribed .

The generic drug industry is a suitable setting for investigating the motives for vertical integration because each market exhibits a clear demarcation between the upstream and downstream segments, and each entrant decides whether or not to vertically integrate. Upstream plants produce active pharmaceutical ingredients , which are chemical compounds with therapeutic prop- erties, using raw materials such as basic and intermediate chemicals, solvents, and catalysts. The downstream segment manufactures finished formulations by combining APIs with inactive ingredients and processing them into dose forms such as tablets and injectables. There is a significant degree of vertical integration in generic drug markets and it has been rising over time. Since the late 1990s, markets opening up in later years have tended to exhibit a greater prevalence of vertical integration. Using a sample of 128 markets, I calculate the average proportion of vertically integrated entrants among all downstream entrants as 8.1 percent in markets that went generic during 1993-2000. The corresponding figure for markets that opened up during 2001-2005 is 24.2 percent. Using firm-level data from generic drug markets, I estimate the determinants of a firm’s decision to vertically integrate. The first finding is that a firm has a higher probability of vertically integrating, conditional on its decision to enter the downstream segment, if it has greater past entry experience in the upstream API segment. This suggests that a firm’s upstream experience lowers its cost of vertical integration. In addition, a firm is more likely to vertically integrate when the average upstream experience level among its rivals is higher.

This is equivalent to saying that a firm’s vertical integration probability is decreasing in its rivals’ cost of vertical integration. Employing a simple duopoly model, I show how such a finding would arise if the payoff function of an individual firm has the following characteristic: the firm gains more from vertical integration when more of its rivals are vertically integrated – which is equivalent to saying that firms’ vertical integration decisions are strategic complements. Intuitively, when a firm is faced with rivals who have low vertical integration costs, it expects a higher degree of vertical integration in the equilibrium market structure. Given that the firm’s gain from vertical integration is greater when more of its rivals are integrated, we should observe a higher probability of vertical integration by the firm itself. In sum, firms in the generic drug industry are responding to the expected prevalence of vertical integration among rivals by becoming vertically integrated themselves. This can be classified as a type of bandwagon effect. Put another way, firms in the generics industry have payoff functions that are conducive to bandwagon behavior. A second set of findings pertains to the relationship specificity of investments as a determinant of vertical integration. I find that generic drug companies are more likely to be vertically integrated in markets where they try to enter with a “paragraph IV certification” – a certification that one or more patents held by the originator pharmaceutical firm are either invalid or not infringed. Generic entrants have an incentive to engage in such patent challenges, because the first one to enter with a paragraph IV certification may be awarded a 180-day exclusivity in the generic market. I employ a simple model to argue that in markets characterized by paragraph IV patent challenges, upstream investment into API development tends to be relationship-specific. This is because in such markets, indoor farming system the upstream product has a much higher value if it is used by the first-to-file paragraph IV entrant than when it is used by some other firm. Such relationship specificity does not exist in other generic drug markets. Therefore, it is likely that the higher relationship specificity of upstream investments in paragraph IV markets explains the higher incidence of vertical integration in such markets. The remainder of the chapter is structured as follows. In Section 2.2, I describe the process of entry and vertical market structure formation in the generic drug industry. The section also examines how vertical integration patterns have evolved over time. Section 2.3 employs simple theoretical models to derive testable predictions. The first model shows that when a firm’s payoff gain due to vertical integration is increasing in the vertical integration status of its rival, the firm’s probability of vertical integration rises as its rival’s cost of integration falls. The second model demonstrates that in a market where generic companies engage in a race to be the first-to-file, investment into API development is characterized by relationship specificity. It also demonstrates the advantage of being vertically integrated in such a market. In Section 2.4, I present the econometric specification used to analyze the determinants of vertical integration by individual firms. Section 2.5 describes the data for the US generics industry and Section 2.6 presents the empirical results.

Section 2.7 concludes. A pharmaceutical product market is born when an originator company receives approval from the Food and Drug Administration to market a new drug. The approval process involves the submission of a New Drug Application by the originator, and the FDA’s review of the NDA based on the criteria of safety and efficacy. Included under the definition of new drugs are formulations containing entirely novel active pharmaceutical ingredients , formulations containing new combinations of existing APIs, new dosage forms of existing APIs, and existing drugs for use in previously unapproved indications. Most newly approved drugs are awarded a period of marketing exclusivity by the federal government. For example, a drug containing a new chemical entity is usually protected by a patent on the API as well as by a five-year period of data exclusivity. The term “data” in data exclusivity refers to the clinical trials information generated by the originator and submitted to the FDA as part of its NDA. The data are protected in the sense that the FDA is not authorized to use it for the purpose of reviewing marketing approval applications submitted by generic manufacturers. In fact, the FDA is not even allowed to accept applications from generic companies until one year before the expiration of the originator’s data exclusivity period if, as is normally the case, those applications rely on the originator’s clinical trials data. New drugs that do not contain new chemical entities are also subject to data exclusivity: new combinations, new formulations, and new uses are all eligible for three years of data protection . In many cases, a new drug is protected by multiple patents. Each patent basically has a minimum term of twenty years from the date of filing so that patent protection usually outlasts the data exclusivity period. The one covering the API is often called a basic product patent. In addition, there are patents that protect new formulations and new uses for existing drugs. Originators also employ additional patents relating to the API, such as those covering new processes of manufacture and those protecting new chemical forms of the same compound . Such additional patents, sometimes called secondary patents, are especially valuable when a new drug is not protected by a basic product patent. This was the case for the antiviral drug zidovudine, whose basic product patent had already expired when it was developed as a pioneering treatment for HIV infection . Even in cases where a basic product patent exists, secondary patents are often used to extend the exclusivity of a new drug beyond the life of the basic patent . This is done by filing the secondary patents ´ during or after the drug development stage, when the life of the basic patent has already been eroded by several years . From the viewpoint of originators, a limitation of secondary patents as an entry barrier is that, unlike data exclusivities and basic product patents, they tend to provide incomplete protection against generic entry. It is sometimes possible for generic companies to produce and sell a drug without infringing any of its secondary patents. For example, if a drug is protected only by a process patent, a generic firm can avoid infringement by employing an alternative process. Moreover, the patentability of innovations that underlie secondary patents is often open to question even after the patent is granted. For instance, combining an anti-hypertension compound and a cholesterollowering agent into the same pill creates significant benefits for some consumers, given that physicians often prescribe such combinations. However, it is a challenge to argue that the combination satisfies the non-obviousness requirement of patentability. Thus, the validity of Pfizer’s patent on Caduet, a combination of amlodipine besylate and atorvastatin calcium, has been challenged by several generic firms . In this way, many secondary pharmaceutical patents belong to the category of what Lemley and Shapiro call “probabilistic patents”. Lei and Wright shed light on the question of why such patents are allowed to exist in the first place. Their empirical analysis indicates that while patent examiners at the US Patent and Trademark Office generally have the ability to correctly judge the patentability of an application, the pro-applicant rules and procedures within the organization drive them to issue more patents than they should. The proliferation of secondary patents creates a potential “patent minefield” where generic firms face the risk of being sued by the originator for infringing a patent that they did not even know existed. Such litigation risks are harmful not only for the generic firms but also for consumers, because they may lead to the abrupt removal of approved generic products from the market.