The coefficient on the paragraph IV indicator variable is estimated to be significantly positive

These surveys are conducted by the National Center for Health Statistics to assess the use of ambulatory medical care in the US, through questionnaires sent to randomly selected hospitals and physicians’ offices. One part of the survey asks for information on “drug visits” during a fixed reference period. A drug visit occurs when a patient visits a health care facility and a drug is prescribed. I estimate the total number of drug visits in the US for each drug in the sample for every year between 1992 and 2004, based on the number of drug visits recorded by the surveys. Then, the total number of drug visits during the one- to five-year period before generic market opening is used to represent the size of the user population for each drug market. Because the focus of NAMCS/NHAMCS is on outpatient services, drugs that are primarily used in inpatient settings are not captured by the surveys. Such drugs are therefore excluded from the sample. The average user population for the drugs in the sample is 2.57 million people. “Per-User Expenditure” is a measure of patients’ and insurers’ willingness to pay for a drug product. Willingness-to-pay varies greatly across drug products because medical conditions vary in terms of morbidity and mortality for the patient, plant growing rack while pharmaceuticals vary in their effectiveness at preventing or treating those conditions as well as in the number of available substitutes .

Such variation may influence generic companies’ incentive to enter a market because it is likely to affect the number of firms that can profitably enter. As a proxy for the willingness to pay for a drug, I use the per-user average annual expenditure on the drug, including out-of-pocket expenses as well as payments made by insurers and other payers, during the year immediately prior to generic entry. This is estimated in two steps. In the first, the average consumed quantity per user is estimated for each drug using data from the Medical Expenditure Panel Survey . Co-sponsored by the Agency for Healthcare Research Quality and the NCHS, MEPS is a nationwide survey that collects data on households’ use of medical goods and services, supplemented with information from the respondents’ health care providers and pharmacies. Using MEPS data for the period 1996-2005, I calculate the average quantity of each drug consumed by a user in one year. Instead of producing separate values for each year, ten years’ worth of observations are pooled together to generate one figure for each drug to cover the entire observation period. In the second step, the average wholesale price of each drug in the year immediately before generic market opening is obtained from different editions of the Red Book. The per-user consumed quantity is then multiplied by the average wholesale price to generate the average per-user annual expenditure. The mean of this variable for the drugs in the sample is 979 US dollars.

The drugs in the sample are grouped into six broad therapeutic classes: anti-infectives, cardiovascular agents, central nervous system agents, gastrointestinal and endocrine-metabolic agents , oncology drugs, and others. The first three categories each make up between one-fifth and one-quarter of the markets in the sample. The drugs are also classified into three distinct dose form groups: oral solids, injectables, and topicals. Oral solids, which make up 82.4 percent of the in-sample drugs, consist of tablets and capsules including extended-release and other enhanced versions. Injectables are liquids that are usually contained in vials and ampoules. Topicals include creams, lotions, and gels. There are two reasons for including indicators for therapeutic classes and dose form groups as covariates. First, they are expected to capture unobserved factors that are related to the revenue potential and product development costs for each market, and that may affect generic entry behavior. For instance, patients may be more willing to switch from originator products to generics in certain therapeutic classes than in others. Second, technological economies due to vertical integration may be stronger for certain drug types than for others. For instance, the production of injectables is subject to quality and manufacturing standards that are generally more stringent than the ones for oral solids . Thus, the returns to vertical integration, which enables tighter control over manufacturing processes, may be higher for injectables.Before turning to the vertical integration equation which is of primary interest, let us consider the other two equations.

In the paragraph IV equation, the coefficient on the downstream originator patents variable is significantly positive. Thus, the observation by Grabowski and Hemphill and Sampat that patents on new formulations and new uses are more vulnerable to challenge by generic entrants is supported. This finding has interesting implications regarding the effectiveness of such patents as entry barriers. To the extent that formulation patents and new use patents induce more aggressive entry behavior by generic firms – in the form of paragraph IV ANDA filings – they may be ineffective at delaying generic entry. In fact, the existence of vulnerable secondary patents might make a drug market more attractive in the eyes of potential generic entrants because it creates an opportunity for 180-day exclusivity, and may induce more of them to enter. In the downstream entry equation, the user population variable has a significantly positive coefficient, which agrees with the intuition that larger downstream markets attract more entrants. On the other hand, the coefficient on per-user expenditure is not significantly different from zero. This suggests that downstream generic entrants are attracted more by market size than by the willingness-to-pay of patients and other payers. Two therapeutic classes – central nervous system agents and oncology drugs – have a significantly positive coefficient, which implies that drugs in these classes tend to attract more generic entry than those in the “Other Therapeutic Class” category. The coefficient on the firm’s own downstream experience is positive and highly significant, confirming earlier results by Scott Morton and Gallant et al. that past downstream entry experience reduces firms’ entry costs in current markets. On the other hand, the coefficient on the own upstream experience variable is not significantly different from zero, which suggests that the effect of upstream experience on downstream entry costs is small. The number of potential upstream-only entrants has a significantly positive coefficient in the downstream entry equation. This implies that in markets where the number of potential unintegrated API suppliers is large, downstream entrants expect to earn higher payoffs. It may not be obvious why the number of potential entrants in the unintegrated upstream category, as opposed to the number of actual entrants, affects the expected payoffs of potential downstream entrants. A likely explanation is that when there are many potential unintegrated upstream entrants, downstream firms expect the equilibrium market structure to be characterized by a greater presence of unintegrated upstream suppliers – in other words, a lower degree of vertical integration. The payoffs of downstream entrants would be higher in markets with less vertical integration if such markets have lower API prices – in other words, if foreclosure effects exist. Meanwhile, indoor vertical garden system the coefficient on the number of potential downstream rivals is not significantly different from zero. Keeping the size and other characteristics of the market fixed, one would expect an individual firm’s entry probability to fall with the number of rivals vying to enter the same market, because the equilibrium number of entrants is not expected to change. Therefore, it comes as somewhat of a surprise that this coefficient is not significantly negative. In the vertical integration equation, the user population variable has a significantly positive coefficient and its marginal effect on the probability of vertical integration is also positive and significant. An increase in the number of users by one million raises a potential downstream entrant’s marginal probability of vertical integration by 16.4 percent. Conditional on the firm entering the downstream segment and on the market not being subject to paragraph IV certification, the same increase in user population raises the probability of vertical integration by 8.2 percent. The finding of a positive relationship between market size and vertical integration, which runs counter to Stigler’s hypothesis that vertical integration is less prevalent in larger markets, is somewhat puzzling. One possible explanation is that unintegrated upstream firms are more efficient in the manufacture of APIs than vertically integrated firms.

If the equilibrium selection process for the entry game is such that the more efficient API manufacturers are given higher priority in entry, then we are likely to see a higher share of the upstream market being taken up by unintegrated entrants in smaller markets. Of the therapeutic class dummy variables, the one for gastrointestinal and endocrine-metabolic agents has a significantly negative marginal effect on the probability of vertical integration. This may be because for some drugs belonging to this class , tighter control over the upstream manufacturing process through vertical integration is not as important as it is for cardiovascular drugs, the baseline category. The dummy variable for injectable formulations has a significantly positive coefficient, and its marginal effect on the marginal probability of vertical integration is also positive and significant. This is consistent with the notion that control over manufacturing processes is more important for injectables than for oral solids, and that vertical integration enables firms to have better control. This lends some support to the hypothesis that vertical integration is motivated by the need for early API development when pursuing a paragraph IV patent challenge. However, the marginal effect of the paragraph IV variable on the marginal probability of vertical integration is not significantly different from zero. The inability to detect a significant marginal effect may be due to the following: while paragraph IV patent challenges change the format of the entry game from a simultaneous-move one to a race to be first, not all firms want to participate in the race. The change in game format raises the benefit of being vertically integrated only for those firms who participate in the race – i.e., those who intend to file paragraph IV certifications. Thus, in order to more accurately capture the impact of the paragraph IV indicator on vertical integration probabilities, one should examine the effect among firms that actually participate in the race. Unfortunately, the data required for such an analysis – firm-level observations on paragraph IV certification – are not available to me at this time. A firm’s past experience at entering the upstream segment of a market has a significantly positive impact on its probability of vertical integration. One additional upstream entry event during the previous year raises the marginal probability of vertical integration by 9.6 percent and increases the conditional probability of vertical integration by 7.7 percent. This finding indicates that the past upstream experience of a potential downstream entrant lowers its cost of vertical integration – that is, past entry experience has a cost-lowering effect in the upstream API segment just as it does in the downstream finished formulation segment. Meanwhile, the downstream entry experience of a firm has a significantly negative marginal effect on the conditional probability of vertical integration. Since the downstream experience variable appears only in the downstream entry equation , this is entirely attributable to an indirect effect . As the downstream experience variable rises, firms having a low value of ε1 become more likely to enter the downstream segment . Because such firms tend to have low values of ε2 due to the positive correlation between the two error terms, their inclusion into the selected group lowers the conditional probability of vertical integration. The mean upstream experience among potential downstream entrants has a significantly positive coefficient in the vertical integration equation, and its marginal effect on the probability of vertical integration is also positive and significant. When the mean upstream experience of rivals increases by one unit , the representative firm’s marginal probability of vertical integration rises by 36.4 percent and its conditional probability of integration increases by 26.7 percent. Combined with the earlier result that a downstream entrant’s own upstream experience increases its probability of vertical integration by lowering its cost of vertical integration, this finding implies the following: lower vertical integration costs among rivals raises a firm’s incentive to vertically integrate. According to the model presented in Section 2.3.1, this implies that firms’ vertical integration decisions are strategic complements, which, in the context of a simultaneous move vertical integration game, is equivalent to the existence of bandwagon effects.