Information, Marketing, and Pricing
`
`in the U.S. Antiulcer Drug Market
`
`
`Introduced into the United States in 1977,
`Tagamet was the pioneer product in the
`class of antiulcer drugs known as H,-
`antagonists. By promoting ulcer healing
`through inhibiting acid secretion, Tagamet
`was able to heal ulcers and treat pre-ulcer
`conditions pharmacologically on an outpa-
`tient basis, thereby substituting for more
`costly hospital admissions and surgeries. In
`1983 another H2-antagonist called Zantac
`entered, and by early 1987 U.S. Zantac sales
`surpassed those of the pioneering Tagamet.
`Today there are four H2-antagonists sold in
`the United States: Tagamet, Zantac, Pep-
`cid, and Axid. Zantac is now the world's
`largest selling prescription drug, having esti-
`mated worldwide sales in 1994 of about $4
`billion. Each of the four H2-antagonists is
`among the top 100 in world drug sales,
`although Tagamet lost U.S. patent protec-
`tion on May 17, 1994.
`In this paper we examine empirically the
`role of information in facilitating and ex-
`plaining growth of the overall antiulcer drug
`market, as well as in shaping the changing
`market shares of the four patented prod-
`ucts. The dissemination of information is
`due largely to the use of marketing chan-
`nels, such as visits by manufacturers' repre-
`sentatives to physicians (called "detailing"),
`
`* ~ e r n d t : Sloan School of Management, Mas-
`sachusetts Institute of Technology, Cambridge, MA
`02142; Bui: Department of Economics, Boston Univer-
`sity, 270 Bay State Road, Boston, MA 02215; Reiley:
`Department of Economics, Massachusetts Institute of
`Technology; Urban: Sloan School of Management,
`Massachusetts Institute of Technology. Financial sup-
`port from the Alfred P. Sloan Foundation is gratefully
`acknowledged, as is data support from Stephen C.
`Chappell of IMS International, J. Stanley Hull of Glaxo
`Pharmaceuticals, Ditas Riad of Merck & Co., and
`William Moore of Lowe & Partners/SMS. Any views
`and opinions expressed here are attributable only to
`the authors.
`
`advertising in medical journals, and most
`recently, by direct-to-consumer advertising.
`We examine these and also explore pricing
`policies, product differentiation, and order-
`of-entry effects.
`
`I. Background
`
`There are two cost conditions that have
`considerable bearing on the structure and
`behavior of the pharmaceutical industry.
`First, sunk costs are very large. In particu-
`lar, the costs of bringing a product to mar-
`ket (doing basic research, winning patent
`approval, engaging in development, per-
`forming clinical trials, and obtaining final
`approval from the Food and Drug Adminis-
`tration [FDA]) are currently estimated at
`about $360 million per drug. Second, for
`most traditional pharmaceutical products,
`the marginal costs of manufacturing are very
`small. Although appropriate cost data are
`not publicly available, it is not uncommon
`for generic drugs to sell at 25-30 percent of
`the pre-patent-expiration price. Informal
`discussions with industry officials suggest
`that for the H2-antagonists, production costs
`are about 10-25 percent of the price.
`These cost conditions have implications
`for pricing. Patent protection gives firms the
`ability to influence price, and to the extent
`one is willing to use the Lerner markup
`relation as a pricing rule of thumb, one
`would expect price and marginal-cost condi-
`tions to approximate ( P -MC)/P = - I/&,,
`is the demand price elasticity.
`where a,
`With manufacturing costs at 10-25 percent
`of price (markups 75-90 percent), the im-
`plied demand price elasticity would range
`from - 1.1 to - 1.3. However, elasticities of
`that size contrast with the common percep-
`tion that demand for prescription drugs is
`extremely price inelastic. Peter Temin (1980
`Ch. 5), for example, notes that physicians
`
`

`
`VOL. 85 NO. 2
`
`INFORMATION, EDUCATING, AND MARKETING IN HEALTH CARE
`
`101
`
`traditionally have been relatively unaware
`of drug prices. Other observers have sug-
`gested that moral hazard in the form of
`third-party (insurance) payment practices
`also contributes to low price responsiveness.
`Very little econometric evidence on demand
`elasticities for drugs is available, in part
`because the traditional consumer demand
`paradigm (utility maximization, marginal
`rates of substitution equal
`to relative
`marginal prices, etc.) cannot be expected to
`describe behavior adequately in a market in
`which principal-agent problems (stemming
`from relationships among physicians, pa-
`tients, and insurers) are widespread.' In this
`paper we report elasticity estimates viewed
`from the vantage of the firm, not the "con-
`sumern-whoever
`that may be.
`Since marginal production costs are small,
`enhancing revenues is essentially the same
`as increasing profits, and thus drug firms
`face strong incentives to shift out the de-
`mand curves. Thus it is not surprising that
`marketing-sales ratios are quite high in the
`pharmaceutical industry. The largest com-
`ponent (70-80 percent) of marketing has
`traditionally involved detailing to physi-
`cians; it consists of a company representa-
`tive providing as much product information
`as possible to physicians, given the typical
`short time of the visit (3-10 minutes) and
`the content regulation enforced by the FDA.
`Medical journal advertising is also carried
`out but is less extensive than detailing. Fi-
`nally, in the last few years, pharmaceutical
`companies have increasingly employed di-
`rect-to-consumer advertising
`in various
`media.
`The information content of marketing ef-
`forts deals primarily with product differen-
`tiation and nonprice aspects. In the H2-
`antagonists market, five quality attributes
`are of particular importance.2 First, the var-
`ious H,-antagonists are viewed as being
`roughly similar in efficacy (the four- to six-
`
`'see, however, Michael Baye et al. (1994).
`or more extensive discussion, see Berndt et al.
`(1994).
`
`week treatment healing rate is about 70-80
`percent for duodenal ulcer patients), al-
`though there is some evidence suggesting
`that Zantac has a significantly lower relapse
`rate than does Tagamet for patients on duo-
`denal maintenance treated at recommended
`dosages (see K. R. Gough et al., 1984).
`Second, less frequent dosages are thought
`to enhance patient compliance. When Zan-
`tac entered the U.S. market in 1983, its
`twice-daily dosing frequency was considered
`more favorable than the regimen of four
`times a day recommended for Tagamet.
`Tagamet responded with a twice-a-day ver-
`sion in late 1984, after which considerable
`rivalry ensued; today all four H2-antagonists
`have a once-a-day version. A third quality
`attribute involves adverse interactions with
`other drugs. Here Tagamet has been on the
`defensive, for early on it was discovered
`that Tagamet interacted with the liver and
`kidney system in a way that could affect the
`metabolism of other drugs. As of 1994,
`Tagamet had reported to the FDA signifi-
`cant drug interactions with ten other drugs,
`whereas Zantac and Axid had only one re-
`ported drug interaction, and Pepcid had
`none. A fourth quality characteristic in-
`volves side effects. Here again Tagamet has
`been somewhat on the defensive, for condi-
`tions such as mental confusion in the el-
`derly and gynecomastia (breast swelling) for
`males are apparently not as prevalent with
`Zantac, Pepcid, and Axid. Finally, the four
`products compete in terms of medical con-
`ditions (indications) for which the FDA has
`granted
`treatment approval. Although
`Tagamet was the first to win approval for
`the treatment of duodenal ulcers, duodenal
`ulcer maintenance, and gastric ulcers, in
`1986 Zantac was the first to obtain approval
`for gastroesophageal
`reflux disease
`(GERD), a rather common condition that
`ranges from modest heartburn and acid in-
`digestion to being a very serious condition.
`The FDA permits marketing only for ap-
`proved indications. Although Tagamet ob-
`tained FDA approval for GERD in 1991,
`and even though Tagamet had very similar
`effects to Zantac, suggesting that it would
`likely also be effective in treating GERD,
`not having FDA approval for GERD
`
`

`
`102
`
`AEA PAPERS AND PROCEEDINGS
`
`MAY 1995
`
`(whereas Zantac did) may have constituted
`a significant marketplace disadvantage for
`Tagamet.
`In terms of pricing, at entry Zantac was
`priced at an 80-percent premium over Taga-
`met, but by May 1994 this premium had
`gradually declined to 19 percent. In May
`1994, the price per day's treatment (to drug
`stores) was $2.61 for Zantac, $2.56 for Axid,
`$2.30 for Tagamet, and $2.17 for Pepcid;
`quantity shares for the four products were
`49 percent, 12 percent, 22 percent, and 17
`percent, respectively.
`To understand the roles of marketing,
`pricing, and quality attributes in explaining
`the growth and changing composition of the
`H2-antagonist market, we now outline an
`econometric model first
`for
`the H,-
`antagonist industry as a whole, and then for
`the market shares garnered by the four H,-
`antagonist drugs.
`
`11. An Econometric Model
`
`of the H,-Antagonist Market
`
`
`At the industry level, we expect the quan-
`tity demanded (number of patient days of
`duodenal-ulcer therapy) to depend on price
`per treatment day, various marketing ef-
`forts, and quality attributes. Since market-
`ing efforts provide long-lived information, it
`is important that cumulative information
`stocks be distinguished from current-period
`new information flows. Define the cumula-
`tive marketing information stock S, at end
`of month t as
`
`advertising (DCA)., It is worth noting that
`the DCA efforts for H2-antagonists did not
`mention any drug by name, but only encour-
`aged viewers to seek advice from their
`physician if they experience heartburn and
`acid indigestion.
`Although such DCA advertising is plausi-
`bly intended to augment overall industry
`demand, when two or more products exist,
`marketing efforts are often only focused on
`a particular brand. During its monopoly era,
`Tagamet recouped all the benefits of its
`marketing efforts (it had 100 percent market
`hare).^ However, once Zantac entered,
`even though rivalry between Tagamet and
`Zantac was intense, some of Tagamet's
`marketing efforts might have spilled over to
`the benefit of Zantac, and vice versa. Simi-
`larly, once Pepcid and Axid entered, while
`marketing efforts were typically focused on
`specific brands, spillovers to Zantac and
`Tagamet might have occurred. To allow for
`marketing spillovers affecting
`industry
`(rather than just product-specific) demand,
`we define the effective industry marketing
`stock ST as a weighted sum of the market-
`ing information stocks originally formed in
`various market structures:
`
`where S,, is the surviving marketing infor-
`mation stock at end of month t that origi-
`nally accumulated in the Tagamet monopoly
`era, S,, is the similar stock formed during
`the Tagamet-Zantac duopoly, S,,
`is that
`from the Tagamet-Zantac-Pepcid triopoly,
`and S4, is that from the Tagamet-Zantac-
`Pepcid-Axid rivalry. Since in a monopoly all
`marketing efforts affect industry demand,
`
`where F, is the flow of new marketing infor-
`mation efforts during month t, and 6 is the
`monthly depreciation rate. Since 6 is un-
`known, we estimate it econometrically. In
`terms of marketing efforts, we distinguish
`three channels: the minutes of detailing to
`physicians (DET), the number of pages of
`medical-journal advertising (PJL), and the
`target rating points of direct-to-consumer
`
`3 ~ a r g e trating points are defined as the target reach
`(the percentage of the over-age-35 population who
`view the message over the course of the ad campaign)
`times the frequency, where frequency is the number of
`times the average target individual views the message.
`For further discussion, see Philip Kotler (1991 pp.
`606-8). The proprietary DCA data were kindly pro-
`vided us by Lowe & Partners/SMS in cooperation with
`Glaxo, Inc.
`4 ~ h ediscussion that follows is based in large part
`on Berndt et al. (1994).
`
`

`
`VOL. 85 M . 2
`
`I N F O W T I O N , EDUCATING, AND W R K E T I N G I N HEALTH CARE
`
`103
`
`we normalize the p's by setting p, = 1.
`Several interesting hypotheses involve the
`p's. First, if the effectiveness of firms' mar-
`keting on industry sales is independent of
`market structure, then p2= p3= p4.= 1.
`Second, if in the presence of competition
`marketing efforts only affect market shares
`and have a zero-sum impact on industry
`demand, then p2= p3= p4= 0. Finally, if
`the industry sales-augmenting effects of
`firms' marketing decline as the number of
`products in the industry increases, then 1>
`p 2 > ~ 3 > ~ 4 > O .
`For our industry demand equation, we
`specify a log-log model, where Q, is quan-
`tity, PI is CPI-deflated price, DETT, PJLT,
`and DCA*, are the effective industry stocks
`defined in (1) and (2), and DGERD is a
`dummy variable taking on the value of 1
`following FDA approval for GERD:
`
`(3) LNQ, = Po+ PILNPI+ P2LNDETT
`
`Since the effective industry marketing stocks
`depend nonlinearly on the p's and S's, and
`since marketing efforts, pricing, and quan-
`tity demanded are likely to be jointly deter-
`mined (see Richard Schmalensee 1972), we
`estimate parameters in equation (3) by non-
`linear two-stage least squares (NL-2SLS).'
`Our econometric model of market shares
`follows Urban et al. (1986) in specifying
`variables relative to the incumbent (Taga-
`met). In particular, using a log-log frame-
`work, we specify that in month t, demand
`quantities of product j relative to the in-
`cumbent [In(Qj/Q, ) = LNQJ1, j = Zantac,
`Pepcid, h i d ] depend on: relative prices,
`LNPRJ1; relative detailing and journal-
`
`pages marketing stocks, LNDTJl and
`LNJPJ1; the number of adverse drug inter-
`actions for product j relative to Tagamet,
`LNINTJ1;6 a discrete variable, DSGERD,
`indicating whether product j has a GERD
`indication advantage relative to Tagamet (1,
`advantage; 0, no advantage; - 1, disadvan-
`tage); an order-of-entry variable, ENTRY,
`taking on the value of 2 for all Zantac
`observations, 3 for Pepcid, and 4 for h i d ;
`and an AGE variable indicating the number
`of months product j has been in the mar-
`ketplace. Again, an instrumental-variable
`procedure is employed to allow for simul-
`taneity.
`Our data sources are described more fully
`in Berndt et al. (19941.' The direct-to-con-
`sumer marketing data are for a campaign
`begun by Glaxo (the manufacturer of Zan-
`tac) in June 1992, and they extend through
`May 1994.
`
`111. Econometric Results
`
`Based on 201 monthly observations from
`September 1977 through May 1994, we esti-
`mated parameters of equation (3) for the
`industry using NL-2SLS. To be parsimo-
`nious in parameters, we constrained the p's
`and 6's to be the same for the DET and
`PJL marketing stocks, but allowed 6 to dif-
`fer for DCA. The preferred model was cho-
`sen based on the lowest value of the tradi-
`tional NL-2SLS residual criterion function.
`Our estimated H2-antagonist industry
`price elasticity is -0.689 (t = 3.80), while
`elasticity estimates for the DET, PJL, and
`DCA surviving stocks are 0.553 (t = 7.52),
`0.198 (t = 2.79) and 0.008 (t = 2.67).' Hence,
`industry demand is positively affected by all
`three of the firms' marketing channels, but
`DET is most effective; the sum of the three
`marketing elasticities is 0.759, suggesting
`decreasing returns to scale. In terms of
`
`5 ~ sinstruments, we employ the producer price in-
`dex for intermediate goods, production worker wages
`in the pharmaceutical industry, cumulative marketing
`efforts by the four companies on non-Hz-antagonist
`products for each of the three instruments, and time.
`
`6 ~ oaccommodate zeros, 1.0 is added to both the
`DCA and the INT variables.
`7 ~ e r ewe extend the Berndt et al. (1994) data base
`to May 1994. Data on prices, quantities, detailing, and
`journal pages are from IMS International.
`equation R'
`"he
`is 0.995, and the Durbin-Watson
`statistic is 1.912.
`
`

`
`104
`
`AEA PAPERS AND PROCEEDINGS
`
`MAY 1995
`
`spillovers, the estimates of p 2 , p3, and p4
`are 0.601 (t = 6.591, 0.924 (t = 5.301, and
`0.410 (t = 4.00); these p's are jointly signif-
`icantly different from 1, and from zero, indi-
`cating that marketing spillovers occur and
`that the effectiveness of firms' marketing
`efforts on industry sales depends on market
`structure. The extent to which spillovers
`occur, however, does not decline monotoni-
`cally with the number of products in the
`market. The DGERD dummy variable co-
`efficient is 0.104 (t = 3.201, indicating that
`FDA approval for GERD increased the size
`of the Hz-antagonist market by about 10
`percent. Finally,
`the NL-2SLS criterion
`function is optimized at the point where 6
`for the DET and PJL stocks is 0.00, while
`that for the DCA stock is 0.15 (t = 0.20),
`implying an annual DCA deterioration rate
`of about 80 percent. Although we are some-
`what surprised that the information stocks
`of DCA and PJL marketing do not depreci-
`ate at all, we note that a similar 6 = 0
`finding in the context of R&D knowledge
`stocks has been reported by Zvi Griliches
`and Frank Lichtenberg (1984).
`Turning now to econometric results based
`on the market-share model, we obtained the
`following NL-2SLS results, based on 291
`monthly observations:
`
`+ 0.052 DSGERD, - 0.252LNINTJ1
`(2.17)
`(9 .00)
`
`with an R~ of 0.983. Order-of-entry effects
`are negative and strong, implying significant
`first-mover advantages, consistent with evi-
`dence from other markets (see Urban et al.,
`1986). The within-Hz-antagonist price-elas-
`ticity estimate is -0.67 and significant, while
`coefficients on relative stocks of detailing
`(0.649) and journal pages of advertising
`(0.167) are positive and significant. The esti-
`
`mated monthly depreciation rate for the
`DET and PJL stocks is 0.030 (t = 13.771,
`implying that relative information market-
`ing stocks deteriorate at about 30 percent
`per year. With respect to quality variables,
`the DSGERD coefficient is 0.05, while that
`on relative adverse drug interactions is
`-0.25, suggesting that Tagamet's market
`share was significantly negatively affected by
`its disadvantages in terms of GERD and
`adverse drug
`interactions
`in
`the H,-
`antagonist market. Finally, the age coeffi-
`cient is positive and significant, implying
`that, ceteris paribus, longevity in the mar-
`ketplace positively affects market shares.'
`
`N.Concluding Remarks
`
`We have reported results on factors af-
`fecting the growth and composition of the
`H2-antagonist drug market. With an H2-
`antagonist industry own-price elasticity of
`-0.69 and between-drug price elasticities
`of -0.66, the implicit brand-specific own-
`price elasticities in May 1994 are -0.80 for
`Tagamet (SE = 0.08), - 1.03 (SE = 0.12) for
`Zantac, -0.76 (SE = 0.08) for Pepcid, and
`(SE = 0.08) for Axid. Except for
`-0.74
`Zantac, these elasticity estimates are still
`slightly smaller than the - 1.1 to - 1.3 val-
`ues one might expect based on the Lerner
`markup rule of thumb; nevertheless they
`are not far from 1, and clearly differ from 0.
`It is worth noting that when marketing vari-
`ables are omitted from the relative-demand
`equations, price-elasticity estimates fall to
`about half these values.
`We find that marketing information stocks
`positively affect sales, that the sales elastic-
`ity is largest for detailing, followed by jour-
`nal pages of advertising, and is smallest for
`direct-to-consumer advertising. Marketing
`information appears to display overall de-
`creasing returns to scale. We also find that
`
`' ~ l t h o u ~ hDCA is arguably intended to affect indus-
`try demand rather than market shares, when the DCA
`information variable is added, shares of Tagamet and
`Axid were positively affected relative to those of Zan-
`tac and Pepcid.
`
`

`
`VOL. 85 NO. 2
`
`INFORMA TION, EDUCATING, AND MARKETING IN HEALTH CARE
`
`105
`
`order-of-entry effects are significant, as are
`quality attributes.
`
`REFERENCES
`
`Baye, Michael R.; Maness, Robert and Wiggins,
`Steven N. "Demand Systems and the 'True'
`Cost of Living for Pharmaceuticals."
`Working paper, Department of Eco-
`nomics, Texas A&M University, May
`1994.
`Berndt, Ernst R.; Bui, Linda; Reiley, David and
`Urban, Glen. "The Roles of Marketing,
`Product Quality and Price Competition in
`the Growth and Composition of the U.S.
`Anti-Ulcer Drug Industry." National Bu-
`reau of Economic Research (Cambridge,
`MA) Working Paper No. 4904, October
`1994.
`Gough, K. R.; Korman, M. G.; Bardhan, K. D.;
`Lee, F. I.; Crowe, J. P.; Reed, P. I. and Smith,
`R. N. "Ranitidine and Cimetidine in Pre-
`vention of Duodenal Ulcer Relapse." The
`
`Lancet, 22 September 1984, 2(8404), pp.
`659-62.
`Griliches, Zvi and Lichtenberg, Frank. "R&D
`and Productivity Growth at the Industry
`Level: Is There Still a Relationship?" in
`Zvi Griliches, ed., R&D, patents, and pro-
`ductiuity. Chicago: University of Chicago
`Press, 1984, pp. 465-502.
`Kotler, Philip. Marketing Management, 7th
`Ed. Englewood Cliffs, NJ: Prentice-Hall,
`1991.
`Schmalensee, Richard L. The economics of ad-
`uertising. Amsterdam: North-Holland,
`1972.
`Temin, Peter. Taking your medicine: Drug reg-
`ulation in the United States. Cambridge,
`MA: Harvard University Press, 1980.
`Urban, Glen L.; Carter, Theresa; Gaskin, Steve
`and Mucha, Zofia. "Market Share Rewards
`to Pioneering Brands: An Empirical
`Analysis and Strategic Implications."
`Management Science, June 1986, 32(6),
`pp. 645-59.