Int. J. of the Economics of Business,
`Vol. 13, No. 1, February 2006, pp. 15–38
`
`Generic Competition in the US Pharmaceutical Industry
`
`ATANU SAHA, HENRY GRABOWSKI, HOWARD BIRNBAUM,
`PAUL GREENBERG and ODED BIZAN
`
`
`AtanuSahaasaha@analysisgroup.comTaylor and Francis LtdCIJB_A_151973.sgm10.1080/13571510500519905International Journal of the Economics of Business1357-1516 (print)/1466-1829 (online)Original Article2006International Journal of the Economics of Business1310000002006
`
`
`
`
`
`
`
`
`
`
`
`ABSTRACT We develop a simultaneous equations estimation framework to understand
`the interactions among generic entry, prices, and market shares. We base our estimates
`on a panel data sample of 40 brand-name drugs that first experienced generic competi-
`tion during the period July 1992–January 1998. We find that generic share and price
`are simultaneously determined, while the number of generic entrants is a key determi-
`nant of generic market share and the generic-to-brand price ratio. In addition, we find
`generic competition to be particularly intense for blockbuster drugs, which experience
`significantly more generic entrants, price erosion, and generic penetration than other
`drugs.
`
`Key Words: Generic Entry; Competition.
`
`JEL Classifications: I11, L11.
`
`1. Introduction
`
`Generic competition has intensified in the US prescription drug industry and
`become a major source of health care cost savings since the mid-1980s. The
`Congressional Budget Office (CBO) estimated that purchasers saved between $8–
`10 billion in 1994 by substituting generics for brand name drugs (CBO, 1998).
`Recently, several leading brand name drugs have experienced generic competition
`
`The authors wish to thank Ernst Berndt and Arthur Havenner for their excellent comments on an
`earlier version of the paper. Research support from Mark Spelber and Sarah Whitney and
`editing support from Amelia Greenberg are gratefully acknowledged. Any errors that remain are
`ours.
`Atanu Saha, Analysis Group, 10 Rockefeller Plaza, 15th Floor, New York, NY 10020, USA; e-mail:
`asaha@analysisgroup.com, Henry Grabowski, Duke University, Department of Economics, Box 90097,
`Durham, NC 27708, USA; e-mail: grabow@econ.duke.edu, Howard Birnbaum, Analysis Group, 111
`Huntington Ave., 10th Floor, Boston, MA 02199, USA; e-mail: hbirnbaum@analysisgroup.com, Paul
`Greenberg, Analysis Group, 111 Huntington Ave., 10th Floor, Boston, MA 02199, USA, e-mail: pgreen-
`berg@analysisgroup.com, Oded Bizan, Microeconomic Analysis, 21 Orchard Street, Cambridge, MA 02140,
`USA.
`Exhibit 1078
`ARGENTUM
`IPR2017-01053
`
`1357-1516 Print/1466-1829 Online/06/010015–24
`© 2006 International Journal of the Economics of Business
`DOI: 10.1080/13571510500519905
`
`000001
`
`

`

`16 A. Saha et al.
`
`– e.g., Prozac, Vasotec and Taxol – and many more commercially significant brand
`name drugs will face generic competition in the next five years.
`In this paper, we seek to understand the process of generic competition better
`by developing a model that captures the interactions among generic entry, prices,
`and market shares using a simultaneous equations framework. The model is esti-
`mated on a panel data sample of 40 drugs first exposed to generic competition
`over the period July 1992–January 1998.
`The next section of this paper considers the historical and institutional factors
`encouraging the growth of the generic industry and summarizes prior findings
`reported in economic literature. Section 3 discusses the structure of the model and
`our estimation methodology. Section 4 describes the characteristics of the dataset
`and our sample. Section 5 discusses the estimation results. Section 6 contains a
`preliminary analysis of the impact of generic entry on brand price. The final
`section provides a brief summary and conclusions.
`
`2. Background
`
`2.1. Important Industry Developments
`
`The growth of the generic drug industry over the past two decades has been
`affected by important changes on both the demand and supply sides. One key
`event was the passage of the Drug Price Competition and Patent Term Restoration
`Act of 1984, better known as the Hatch-Waxman Act. This act significantly reduced
`the costs and time of entry for generic drugs by establishing an Abbreviated New
`Drug Application (ANDA) procedure. With an ANDA, generic firms need only
`show that their products are bioequivalent to the branded product in order to gain
`Federal Drug Administration (FDA) approval.1 In addition, the law established a
`research exemption so that generic firms could perform their bioequivalence test-
`ing and receive conditional FDA approval prior to the expiration of the brand
`product’s patents. The 1984 law also tried to strike a balance between generic price
`competition and drug innovation by providing brand name firms with the oppor-
`tunity for patent term extension to compensate for time lost during the clinical test-
`ing and regulatory approval stages.2
`On the demand side, the development of the generic industry has been aided
`by the growth of managed care and the more intensive coverage of prescription
`drugs by health insurers.3 Pharmacy benefit management firms (PBMs) have
`evolved as managers of pharmaceutical reimbursement programs for both HMOs
`and employers and have actively promoted the use of generic drugs as a cost-
`saving measure (Berndt, 2002). Generic competition has also been encouraged
`through various benefit designs, including a tiered formulary in which generics
`are placed in the least costly co-payment tier.4 PBMs also provide incentives to
`pharmacists in the form of higher fees for generics, compared to branded
`products.5 In addition, PBMs often monitor and attempt to alter physicians’
`prescribing habits among those who disproportionately prohibit generic substitu-
`tion. Grabowski and Mullins (1997) found that these various incentive measures
`can save payers 10% or more of their total drug budget.
`Thus, there have been powerful institutional forces at work accelerating the
`degree of generic competition since the mid-1980s. This is reflected in the fact that
`47% of prescription drug units consumed in the United States in 1999 were
`generic products, compared to only 19% in 1984 (PhRMA, 2001: 61). With several
`
`000002
`
`

`

`Generic Competition in the US Pharmaceutical Industry 17
`
`widely prescribed branded products scheduled to go off patent in the next five
`years, the percentage of generic utilization is likely to increase in the years ahead.
`
`2.2. Prior Economic Studies of Generic Competition
`
`Several economic studies have examined the characteristics and determinants of
`generic competition after the passage of the 1984 Hatch-Waxman Act. Caves et al.
`(1991) conducted an early exploratory analysis of generic competition using a
`sample of 30 drugs that went off patent between 1976 and 1987. Their analyses
`spanned the period prior to Hatch-Waxman and a few years after its passage.
`They found that the initial generic drug entered the market at a significant
`discount to the branded product (40% on average) and this discount grew larger
`as the number of generic competitors expanded over time. However, even with a
`significant number of generic competitors in the market, the average market
`shares6 of the generic products were relatively small in this period. In this regard,
`their analysis was consistent with a number of studies of the pre-1984 period that
`found the impact of generic competition on branded sales to be very limited.7
`Several papers have focused on generic price. In two related studies,
`Grabowski and Vernon (1992, 1996) examined a sample of 40 branded products
`that faced generic competition between 1984 and 1993, when the intensity of
`generic competition increased significantly.8 Using a regression model in which
`the number of generic competitors was driven by the expected profitability of
`entry, they found that the price of a generic product tended toward marginal cost
`over a multi-year time frame.9 In a recent paper, Reiffen and Ward (2002) esti-
`mated a structural model of the relation between generic drug prices and the
`number of ANDA approvals, and concluded that eight or more ANDAs were
`generally sufficient to cause generic prices to converge to long run marginal cost.
`Another subject of prior studies has been generic market share. Grabowski
`and Vernon (1996) found that the speed at which generics captured market share
`was positively related to the size of the brand product’s pre-entry sales, the thera-
`peutic class of the product, and the calendar date of generic entry. Greater rates of
`generic utilization were observed for more recent time cohorts of brand products.
`In particular, by the early 1990s, generic shares averaged about two-thirds of a
`molecule’s unit sales one year after the initiation of generic competition.
`Fiona Scott Morton examined generic entry decisions in two recent papers. In
`the first paper, Scott Morton (1999) showed that firms are more likely to venture
`into markets in which they have some experience, e.g., in form, therapy or ingre-
`dient. In addition, firms have a tendency to enter large markets and markets
`where the drug treats a chronic condition. In a second paper, Scott Morton (2000)
`looked at factors that might thwart generic entry, including switching costs, FDA
`regulations, and brand firm advertising. Using a sample of 98 drugs with patent
`expirations from 1986 to 1992, she found that generic entry was positively related
`to brand revenue and price elasticity, and negatively affected by FDA
`regulations.10 She also found no evidence that brand advertising has deterred
`generic entry.
`A number of investigators dating back to Caves et al. (1991) have considered
`the response of brand firms to generic entry and whether branded firms have
`pursued entry-deterring strategies. There is little evidence to support the
`hypothesis of entry deterrence. First, with respect to promotional activities,
`branded firms typically curtail most of their expenditures, usually beginning in
`
`000003
`
`

`

`18 A. Saha et al.
`
`the pre-entry period (Caves et al., 1991).11 Ellison and Ellison (2000) find that the
`trends in advertising and product proliferation are non-monotonically related to
`the probability of generic entry: advertising is reduced and presentation prolifera-
`tion increased in the period preceding patent expiration among drugs that face an
`intermediate probability of entry. Second, there is scant evidence that brand firms
`take pre-emptive actions or match generic prices, except when they offer selective
`discounts to their large institutional customers (CBO, 1998). Rather, most studies
`have found that branded drug firms continue to raise prices after generic entry,
`although there is some disagreement about whether generic entry has positively
`or negatively affected the rate of increase in these prices.12 Grabowski and Vernon
`observe specific cases where brand name firms have pursued a two-tier strategy,
`entering the generic market either through a subsidiary firm or in partnership
`with a generic firm. Even in these latter situations, however, entry has not been
`effectively deterred and generic price competition has remained intense.
`
`2.3. Objectives of Our Analysis
`
`Our study builds on the studies discussed above, but contributes to the literature
`in several dimensions. First, we explicitly account for the interaction between three
`key variables: generic entry, generic share, and generic-to-brand price ratio. We
`posit that these variables are part of a simultaneously determined system; specifi-
`cally, generic entry affects the share of generic suppliers and the price of generics.
`These two variables are then endogenously determined. That is, generic share
`depends on, and is influenced by, generic price. While a few papers in the existing
`literature have acknowledged the endogeneity of generic entry,13 prices, or shares,
`to our knowledge ours is the first paper to adopt a simultaneous estimation
`procedure to address the issue of the endogeneity of all of the key variables. Our
`empirical results clearly show that generic share influences and is influenced by
`prices, corroborating our model’s econometric estimation framework.
`Second, our study examines a relatively large sample of drugs that experi-
`enced generic competition between July 1992 and January 1998. The analysis of
`more recent data is particularly relevant in light of the marked growth of generic
`drug sales fuelled by the dominant role of managed care and PBMs in the 1990s.
`Finally, we adopt an estimation framework that is appropriate for panel data.
`Our estimation approach allows and corrects for heteroskedasticity and serial
`correlation of errors. That is, we allow for idiosyncratic differences across drugs
`(cross sectional units) through the heterogeneity of error variances. Additionally,
`since we have time series observations on each drug, we allow for drug-specific
`serial correlation of errors. The ordinary least squares (OLS) model is a special
`case of this more general estimation approach. Our results demonstrate that the
`OLS estimation framework yields, in many cases, seriously erroneous inferences
`about the determinants of generic competition.
`Our empirical results provide valuable insights into the determinants of
`generic entry, prices, and generics’ market share. We find generic competition to
`be particularly intense for ‘blockbuster’ drugs, which we define as drugs having
`pre-generic annual sales of $500 million or more. Specifically, we find that block-
`buster drugs average two more generic entrants annually compared with non-
`blockbusters. We further find that the number of generic entrants, in turn, directly
`affects the level of generics’ share and price. Blockbuster drugs thus experience
`not only significantly more generic entrants, but also more price erosion and
`
`000004
`
`

`

`Generic Competition in the US Pharmaceutical Industry 19
`
`generic penetration than non-blockbuster drugs. We also find that the extent of
`HMO coverage has a positive impact on the market share garnered by the gener-
`ics. Additionally, generic prices are significantly and positively related to the
`costs of drug production.
`Our results also include a preliminary analysis of brand prices. In contrast to
`prior studies’ findings based on data from earlier time periods, our results indi-
`cate that brand prices do respond to generic competition: each additional entrant
`on average is associated with a 0.2% decline in brand price. Nevertheless, unless
`the number of generic competitors is large, brand prices continue to rise in abso-
`lute terms. Consistent with prior studies, we do not find any evidence of entry-
`deterrent pricing by brand manufacturers.
`
`3. Econometric Model Specification
`
`The objective of the econometric model is to explain the determinants of three key
`variables for each drug: P, S, and N, where P denotes the average generic-to-
`brand price ratio, S represents the share of the group of generic substitutes of a
`branded drug, and N is the number of generic manufacturers of that compound.
`The number of generic manufacturers for the ith drug at time t is defined as
`follows:
`
`N
`
`it
`
`≡
`
`N
`
`it
`
`+−1
`
`E
`it
`
`,
`
`
`
`1( )
`
`where Eit is the number of generic entrants for ith drug at time t. The number of
`entrants, in turn, is determined by:
`= (
`)
`
`ε
`E
`it
`
`,
`
`iE
`
`f N
`
`,
`
`X
`
`−1
`
`it
`
`E
`it
`
`
`
`2( )
`
`where XE is a set of exogenous variables related to the conditions of entry faced by
`generics, and εE denotes random errors. All variables on the right hand side of (2)
`are thus pre-determined, that is, (2) is not a simultaneous equation and therefore
`can be estimated directly.
`By contrast, for each drug the generic-to-brand price ratio (P) and the share
`for all generics (S) are jointly-determined (i.e., endogenous) variables. The simul-
`taneous equation framework determining these variables is:
`= (
`)
`,
`,
`g S N X
`it
`it
`= (
`)
`,
`,
`h P N X
`it
`it
`
`P
`it
`
`S
`it
`
`P
`it
`
`,
`
`ε
`P
`it
`
`S
`it
`
`,
`
`ε
`S
`it
`
`
`
`( )3
`
`
`
`( )4
`
`where XP and XS are sets of exogenous variables that affect the price ratio and
`generic share, and εP and εS are random errors. Note that while the number
`of generics (N) affects the generic-to-brand price ratio and generic share, it is a pre-
`determined variable since it is fully determined by information available at time t−1.
`Equation (3) includes exogenous variables affecting the intensity of price
`competition on the supply side of the market. Equation (4) includes exogenous
`variables affecting the intensity of demand by managed care and other purchasers.
`Equation (3) is identified by excluded demand side shifts (i.e., HMO coverage),
`while equation (4) is identified by excluded cost shifters (i.e., manufacturing costs).
`
`000005
`
`

`

`20 A. Saha et al.
`
`Eit = f(Nit–1, Xit )E
`
`Eit Nit–1 + Eit
`
`Pit = g(Sit, Nit , Xit )P
`
`Sit = h(Pit, Nit , Xit )S
`
`Figure 1. The interaction between entry (E), number of generics (N), generic
`share (S), and generic-to-brand price ratio (P)
`
`Figure 1. The interaction between entry (E), number of generics (N), generic share (S), and generic-to-brand price ratio (P)
`
`Figure 1 summarizes the relationships between the key variables, N, P, and S,
`in the econometric model. The direction of the arrows in the figure indicates the
`causality relationship. Thus, a one-directional arrow going from, say, N to P
`indicates that N is exogenous to P. The two-directional arrow connecting P and S
`indicates endogeneity of these two variables.
`We estimate equation (2) separately from (3) and (4). The latter two form a
`system of simultaneous equations, which is estimated using the instrumental
`variables (IV) regression approach. Each of these equations is estimated using
`both OLS and pooled time series cross sectional estimation methods, with appro-
`priate correction for heteroskedasticity and serial correlation. In particular, we
`assume that in each equation, (2)–(4): εit = ρiεit−1 + uit, where uit is identically and
`σi
`independently distributed random error, and V(εit) =
`2
`. Thus, OLS is a special
`2 σj
`σi
` = σ2, ∀i, j,
`case wherein: ρi = 0, ∀i, implying no autocorrelation, and
`2
` =
`implying homoskedasticity of errors.
`
`4. Data and Descriptive Analysis
`
`4.1. Source Data
`
`The data for our analysis are primarily derived from an IMS Health information
`product entitled ‘Generic Spectra.’ IMS Health is a leading provider of informa-
`tion products to the pharmaceutical and healthcare industries.14 The Generic
`Spectra dataset has product shipment information for purchases made by phar-
`macies and hospitals, including units and revenues.
`This dataset contains dollar and gram sales data for brand and generic drugs
`for, at most, three years before and three years after first generic entry.15 The brand
`drugs in the dataset first faced generic competition between July 1992 and January
`1998. Our sample is restricted to oral drugs utilized primarily on an outpatient
`basis. While the original dataset included 41 drugs, we omitted one – Micronase
`(glyburide) – because it was launched in a co-marketing agreement with Diabeta
`
`000006
`
`

`

`Generic Competition in the US Pharmaceutical Industry 21
`
`(also glyburide). The list of drugs in our analysis, their therapeutic classification,
`and the manufacturer of the brand are reported in Table 1. We used these data to
`construct the key variables in our analysis, including brand price, generic price,
`generic share, market size, and generic-to-brand price ratio.
`We measure generic share for a drug by dividing the sum of all generic
`manufacturers’ sales (in grams) by the total sales (i.e., generic plus brand) of the
`compound. We calculate price as dollars per gram. We recognize that this differs
`from the definition of price used in previous studies. For example, Grabowski
`and Vernon (1992) have used the average cost per unit paid by drugstores and
`hospitals for the most frequently consumed dosage size of each compound. Frank
`and Salkever (1997) define price as the average revenue per extended unit, while
`Caves et al. (1991) identify the most popular dosage of each drug and then divide
`sales revenue by quantities sold in wholesale transactions (i.e., transactions
`involving pharmacies and hospitals) to compute an average price. Although our
`dollars-per-gram price does not account for the variation in prices arising from
`different dosage forms, it is the most relevant price for our analysis given that it
`captures the entire market for the drugs rather than a specific segment.
`Data on the number of generic manufacturers of each drug come from IMS
`America’s Product Directory in the Market Research database. These data include
`drugs’ manufacturers and entry dates.16 We verified these dates using the Federal
`Food and Drug Administration’s Orange Books.17
`Data on HMO coverage come from PhRMA’s Pharmaceutical Industry Profile
`(2000). It is an annual series based on IMS audits that reports third-party pharma-
`ceutical reimbursement. We calculate the manufacturing cost variable from the
`Bureau of Labor Statistics (BLS) intermediate goods price index for basic inorganic
`chemicals, which we divide by the BLS pharmaceutical producer price index to
`express cost in real dollars.18
`We created indicator (or dummy) variables for the various therapeutic classes
`to which the drugs belonged to capture the class-specific differences across drugs.
`We also created an indicator variable for the drugs Clozaril, Mexitil, Toradol, and
`Zarontin, because their usage is restricted due to the possibility of serious side
`effects, a disincentive to generic entry. Specifically, Clozaril has a high incidence
`of agranulocytosis, which makes patients susceptible to life-threatening infection.
`The drug is administered only through a process that mandates a weekly moni-
`toring of the patients’ white blood cell count. Mexitil is indicated only for patients
`with life threatening arrhythmia and its use can be initiated only in hospitals.
`Similarly, Toradol must be started in a hospital in IV form. Use of the oral form is
`limited to five days due to its potential for severe side effects, including bleeding
`ulcers. Zarontin can reduce the body’s ability to manufacture certain blood cells
`that are important to fight infections and prevent bleeding. As a result, patients
`on Zarontin must have their blood levels monitored periodically. For each of
`these four drugs, Clozaril, Mexitil, Toradol, and Zarontin, the ‘restricted usage’
`indicator variable takes a value of one; it is zero for all other drugs.
`
`4.2. Descriptive Statistics
`
`Table 1 lists all compounds in our dataset, the therapeutic classes to which they
`belong, and the brand manufacturers. The 40 brands in the sample belong to nine
`therapeutic classes and are manufactured by 20 pharmaceutical firms. The largest
`therapeutic class in the sample, Cardiovascular, has 14 compounds.
`
`000007
`
`

`

`22 A. Saha et al.
`
`2
`
`6
`
`6
`
`9
`
`3
`
`12
`
`1
`
`7
`
`9
`
`4
`
`2
`
`5
`
`10
`
`6
`
`12
`13
`
`8
`
`3
`
`13
`20
`21
`
`8
`
`18
`20
`19
`13
`
`64%
`52%
`66%
`39%
`66%
`40%
`60%
`62%
`76%
`63%
`60%
`67%
`65%
`58%
`52%
`33%
`71%
`62%
`21%
`44%
`5%
`70%
`17%
`10%
`73%
`30%
`
`33%
`58%
`39%
`41%
`42%
`62%
`72%
`66%
`45%
`28%
`77%
`10%
`58%
`81%
`59%
`58%
`72%
`63%
`64%
`86%
`79%
`83%
`85%
`72%
`66%
`75%
`
`82.91
`98.29
`101.50
`118.84
`143.41
`145.02
`154.97
`159.72
`185.32
`205.89
`263.43
`281.73
`288.60
`331.17
`343.83
`371.26
`395.95
`539.03
`557.67
`573.37
`639.75
`656.71
`697.05
`805.49
`875.40
`1,856.94
`
`after generic entry
`generics 12 months
`
`after generic entry
`
`price ratio 12 months
`
`Number of
`
`Generic to brand
`
`after generic entry
`share 12 months
`Generic market
`
`$ millions)
`
`entry (in 2000
`Market size at
`
`entry
`generic
`
`Date of first
`
`launch
`drug
`Date of
`
`Manufacturer
`
`Therapeutic class
`
`Drug
`
`Table 1.Summary statistics by drug
`
`Ciba-Geigy (Novartis)
`Jan-76
`Jan-98
`Merck
`Apr-82
`Oct-92
`Zeneca PharmaceuticalsAug-84
`Jul-92
`Somerset Pharmaceutical
`Sep-89
`Aug-96
`Rhone-Poulenc Rorer
`Jul-83
`Jul-93
`Wyeth-Ayerst
`Jan-86
`Dec-92
`Upjohn
`Dec-88
`Jun-94
`Hoechst Marion RousselNov-81
`Nov-96
`Bristol Laboratories
`Mar-85
`Aug-93
`DuPont PharmaceuticalsMay-75
`Jan-93
`Hoechst Marion RousselMay-84
`Apr-94
`Apr-91
`Dec-97
`Pfizer
`May-84May-94
`Ciba-Geigy (Novartis)
`Jul-88
`Aug-95
`Wyeth-Ayerst
`Apr-91
`Feb-97
`Ciba-Geigy (Novartis)
`Jul-78
`Oct-93
`Roche Laboratories
`Aug-75
`Sep-96
`Parke-Davis
`Jan-82
`Jan-93
`Hoechst Marion RousselNov-82
`Oct-92
`Feb-85
`Apr-97
`Glaxo
`Apr-81
`Dec-95
`Bristol-Myers Squibb
`Aug-79
`Oct-94
`Eli Lilly
`Mar-76
`Sep-93
`Syntex
`Nov-81
`Sep-93
`Psychotherapeutics/sedativesUpjohn
`Aug-77May-94
`Jul-97
`Jun-83
`
`Antivirals
`Cardiovascular
`Antibiotics/anti-infectives
`
`SmithKline Beecham
`Glaxo
`
`Analgesics
`
`Cardiovascular
`Antiarthritic
`Antiarthritic
`
`Diabetes
`Psychotherapeutics/sedativesCiba-Geigy (Novartis)
`
`ParlodelNeurological disorders
`Dolobid
`TenoreticCardiovascular
`EldeprylNeurological disorders
`Lozol
`Orudis
`Ansaid
`CarafateGastrointestinal
`Cardiovascular
`Corgard
`SinemetNeurological disorders
`Diabeta
`Clozaril
`GlucotrolDiabetes
`VoltarenAntiarthritic
`Antiarthritic
`Lodine
`LopressorCardiovascular
`KlonopinNeurological disorders
`Lopid
`Cardiovascular
`CardizemCardiovascular
`Zovirax
`Capoten
`Ceclor
`NaprosynAntiarthritic
`Xanax
`TagametGastrointestinal
`Gastrointestinal
`Zantac
`
`000008
`
`

`

`Generic Competition in the US Pharmaceutical Industry 23
`
`Note: Drugs are sorted in descending order by the variable Market Size at Entry.
`
`1
`
`5
`
`5
`
`4
`
`6
`
`3
`
`1
`
`1
`
`3
`
`8
`
`10
`
`9
`
`4
`
`5
`
`2
`
`54%
`63%
`69%
`63%
`
`56%
`52%
`62%
`36%
`64%
`66%
`60%
`63%
`46%
`76%
`60%
`
`55%
`26%
`1%
`97%
`
`44%
`43%
`40%
`48%
`40%
`46%
`34%
`70%
`67%
`34%
`50%
`
`284.94
`6.55
`11.92
`13.60
`
`14.97
`15.30
`27.47
`39.80
`42.42
`47.71
`50.06
`50.89
`59.43
`69.64
`74.78
`
`Mar-68May-96
`Aug-94
`Nov-87
`Jul-92
`Jan-76
`
`Jan-91
`Jul-97
`Oct-82
`Sep-94
`Feb-89
`Jul-96
`Jan-86
`Dec-97
`Mar-86
`Jun-95
`Mar-85May-95
`Oct-82
`Oct-92
`Apr-90May-97
`Dec-96
`Feb-90
`Jan-95
`Apr-83
`Sep-93
`Dec-82
`
`Parke-Davis
`
`Psychotherapeutics/sedativesEli Lilly
`
`Vivactil
`ZarontinNeurological disorders
`Aventyl
`
`Average
`Psychotherapeutics/sedativesMerck
`
`Prosom
`WytensinCardiovascular
`Cardene
`Cardiovascular
`CapozideCardiovascular
`Cardiovascular
`Mexitil
`Cardiovascular
`Sectral
`Cardiovascular
`Visken
`Analgesics
`Toradol
`AnafranilPsychotherapeutics/sedativesCiba-Geigy (Novartis)
`Bumex
`Halcion
`
`Pharmaceuticals
`Psychotherapeutics/sedativesTakeda-Abbott
`
`Wyeth-Ayerst
`Syntex
`Bristol-Myers Squibb
`Boehringer Ingelheim
`Wyeth-Ayerst
`Ciba-Geigy (Novartis)
`Syntex
`
`Roche Laboratories
`
`Cardiovascular
`Psychotherapeutics/sedativesUpjohn
`
`after generic entry
`generics 12 months
`
`after generic entry
`
`price ratio 12 months
`
`Number of
`
`Generic to brand
`
`after generic entry
`share 12 months
`Generic market
`
`$ millions)
`
`entry (in 2000
`Market size at
`
`entry
`generic
`
`Date of first
`
`launch
`drug
`Date of
`
`Manufacturer
`
`Therapeutic class
`
`Drug
`
`(Continued)
`
`Table 1.
`
`000009
`
`

`

`24 A. Saha et al.
`
`The branded drugs in the sample were launched between March 1968
`(Vivactil, in the class of Psychotherapeutics/sedatives, launched by Merck) and
`April 1991 (Clozaril, launched by Novartis and Lodine in the class of drugs for
`Psychotherapeutics/sedatives, and by Wyeth-Ayerst in the class of Antiarthrit-
`ics). The brands in the sample faced generic competition between July 1992 and
`January 1998.
`Table 1 reveals considerable variability in entry rates across drugs. While
`some compounds are supplied by at least 20 generic firms within a year after
`generic entry occurs, for others there is only one. Table 1 also shows an associa-
`tion between entry rate and market size.19 The average annual sales by the brand
`prior to generic entry for drugs with at least 20 generic entrants are $673 million.20
`This figure drops to $101 million21 for drugs facing competition from only one or
`two generic entrants.
`The statistics in Table 1 also suggest a strong association between entry rate
`and the generic-to-brand price ratio. The average generic-to-brand price ratio a
`year after the first generic entry for drugs with at least 20 generic suppliers is 20%;
`by contrast, this ratio is 65% for drugs with two or fewer generics by the end of
`the first year.
`Table 1 data reveal considerable variability in the degree of generic penetra-
`tion across drugs. The average market share a year after the first generic entry for
`compounds with at least 20 suppliers is 79%, while this share is only 47% for
`compounds with two or fewer generics. Even within the group of compounds
`that experienced very little entry, there are marked differences. During the first
`year after generic entry, both Aventyl and Zarontin experienced competition
`from only one generic manufacturer. But, the shares of these generics were 97%
`and 1%, respectively, the highest and lowest one-year generic shares in the
`sample. Thus, although the statistics in Table 1 indicate a strong association
`between the number of generic entrants and both generic share and generic price,
`clearly there are other important determinants of these variables. We examine
`these through regression analysis in the next section.
`Table 2 defines variables used in the regression analysis and contains post-
`entry summary statistics.22 The minimum and maximum values of the key vari-
`ables suggest a high degree of heterogeneity across drugs and over time. For
`example, for the 40 drugs in the sample, the number of generic competitors
`ranges between 1 and 27, the generics’ share between 0.02% and 99%, and the
`generic-to-brand price ratio between 0.05 and 1.13.23 These figures suggest a
`considerable degree of variability in the nature and extent of generic competition
`across drugs. The focus of the next section is an examination of the factors that
`explain this variability.
`
`5. Estimation Results
`
`This section first considers the determinants of generic entry and then discusses
`the impact of generic entry and other factors on the generic-to-brand price ratio
`and generic market share.
`
`5.1. Determinants of Generic Entry
`
`For most drugs in our sample, significant entry occurs after the first generic
`manufacturer has entered the market. Figure 2a depicts the average number of
`
`000010
`
`

`

`Generic Competition in the US Pharmaceutical Industry 25
`
`0.57
`
`82%
`
`1.00
`1.00
`27.00
`1.13
`2715.16
`1783.16
`
`99%
`
`1856.94
`
`0.42
`
`44%
`
`0.00
`0.00
`1.00
`0.05
`0.61
`0.24
`0%
`
`6.55
`
`0.03
`
`9%
`
`0.42
`0.28
`6.32
`0.20
`451.30
`257.55
`24%
`
`328.09
`
`0.53
`
`65%
`
`0.23
`0.09
`8.74
`0.50
`184.58
`93.32
`58%
`
`280.24
`
`1338
`
`1338
`
`1338
`1338
`1338
`1338
`1338
`1338
`1338
`
`1338
`
`pharmaceutical preparations PPI
`PPI for intermediate pharmaceutical goods didvided by
`Pharmaceutical Industry Profile
`Third party reimbursement share as reported in PhRMA’s
`
`$500M (in 2000 $)
`Indicator for whether market size at entry is greater than
`Indicator for whether product usage was restricted
`Number of generic drugs
`Ratio of generic-to-brand price
`Brand price per gram (in $)
`Average price per gram of all generics (in $)
`Quantity share of all generics
`millions)
`Annual sales of the brand before generic entry (in 2000 $
`
`Manufacturing cost
`
`HMO coverage
`Other:
`
`Blockbuster dummy
`Restricted usage dummy
`Number of generics
`Generic-to-brand price ratio
`Brand price
`Generic price
`Generic share
`
`Market size at entry
`For each branded drug:
`
`Max
`
`Min
`
`Std. dev.
`
`Mean
`
`Obs
`
`Description
`
`Variable
`
`Table 2.Definition of variables and summary statistics
`
`000011
`
`

`

`26 A. Saha et al.
`
`1
`
`5
`
`9
`
`25
`21
`17
`13
`Months after first generic entry
`
`29
`
`33
`
`16
`
`12
`
`8 4 0
`
`Number of generics
`
`Figure 2a. Average number of generic drug manufacturers.
`
`Figure 2a. Average number of generic drug manufacturers.
`
`generics in the months following generic entry. As the figure shows, by the end of
`the first month, on average two generic manufacturers compete with the brand.
`Within four months the average number of generics exceeds five, and within a
`year it is close to eight. There is essentially no exit of generic firms observed in our
`panel data sample.
`We base our entry variable on several factors. We hypothesize that entry
`depends in part on the number of independent firms marketing the drug through
`wholesalers and chain pharmacies. We expect early entry in particular to be
`strongly influenced by the size and profitability of the market prior to patent
`expiration, as well as other market pull factors. Additionally, FDA regulations
`affect entry. As previously stated, current entry is regulated by the Hatch-
`Waxman Act, which sanctioned bioequivalence testing and application for condi-
`tional approval prior to patent expiration. The law also allowed a generic supplier
`to enter the market not only through its own approved ANDA but also through a
`licensing arrangement with a firm that had an approved ANDA. Most of the 40
`products in our sample have in fact considerably more generic suppliers than
`approved ANDAs, which suggests that many generic firms have entered cross-
`licensing deals in order to extend their product portfolios. In other cases, firms
`with approved ANDAs do not actually market the