Global competition and the United States pharmaceutical industry.

Ramrattan, Lall ; Szenberg, Michael

1. Introduction

Most studies embed pharmaceutical companies within firms in different industries in studies of innovation, prices, and returns to make up for sample size and to infer aggregate industry performance from market structure. But the industry structure appears fragmented at best, with wave of mergers occurring to confront globalization and intellectual property rights (The Economist (US), June 23, 2001). Information about the structural direction that the industry will take resides within the brain cells of the CEOs of major companies.

The endogenous growth model predicts nonrivalrous behavior for R&D behavior in a national and global setting. This information translates into process and product innovation at the level of the firm, where R&D, advertising, and productivity are the driving forces for success. We have collected time series from 1980-1999 for 7 firms: Abbot Laboratories, American Home Products, Bristol-Myers Squibb, Eli Lilly, Merck, Johnson and Johnson, and Pfizer to investigate rivalry among them. We statistically fitted four equations corresponding to four hypotheses and found that smaller firms tend to set their R&D and advertising budgets taking Merck's previous outlays as given. However, when Total Factor Productivity is investigated for the same period, large firms tended to react to small firms, reaffirming concerns in the literature regarding size versus innovation.

2. Background

The U.S. Pharmaceutical Industry has enjoyed economies from the aging baby boomer population, aggressive R&D, advertising and productivity efforts; and now from the opportunities available in the global economy. The potential opportunities and challenges for pharmaceutical innovation are tremendous. Groundbreaking advances in technology have led to unprecedented pharmaceutical discoveries. Yet a major concern is that regulations by the FDA will generate low returns to investments in R&D. For instance, the rate of return in the late 1970s fell by a third to its 1960 levels, and the cost of discovering and developing new drugs increased 18-fold (Business Week, February 21, 1977).

During the 1980s, the pharmaceutical industry received a boost from the Reagan administration that lengthened the patents on prescription drugs and hastened the pace of approving generic drugs to substitute for drugs with expired patents. The immediate result in the 1980s was that R&D expenditure in drugs was about 10% of the industry's sales, versus 3% for all manufacturing industries (S&P Industry Survey, January 1985, H16). But the FDA's Center for Drug Evaluation and Research (CDER) still regulates the industry brand name, generic prescriptions and OTC drugs, placing a heavy time delay on production. The time it takes to develop a new drug has almost doubled from its 1960 levels. The actual trend is 8.1 years in the 1960s, 11.6 years in the 1970s, 14.2 years in the 1980, and a stable 14.9 years during 1990-1996 (Pharmaceutical Industry Profile, 2000, VI). CDER claimed that with the user-fee approach in the mid-1990s, where the applicant pays the government for its review, they have doubled the number of new drugs approved and halved the review time (FDA Consumer, September-October 1997, 21). Other policies such as the streamlining of the IND and the International Conference on Harmonization also reduced review time. However, the review time continues to generate concern. The industry's strategy is:

1. To have an ample supply of R&D projects and patents in the pipeline,

2. To lobby Congress and get extension of time on their patents in order to recoup their investment costs,

3. To allow the speedy approval of generic drugs in order to substitute for drugs whose patent has expired, and

4. To make drugs available before approval possible in special cases such as in the HIV cases in South Africa.

3. Globalization Effects

At the firm level, big changes such as NAFFA have not noticeably affected firms in the pharmaceutical industry relative to firms in other industries such as the textile, shoes, autos, and steel industries. An early concern in the international scene was the debate between Canada and the U.S. over compulsory licensing of firms' pharmaceutical patents (Hufbauer et al., 1992, 173, 179). This issue seems to be resolved under The Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS) under WTO that grants 20 years patent protection for pharmaceutical and other product. Today, the international scene is populated with demonstrations, such as in Seattle in December 1999 against the WTO and others against the G8. These demonstrations are not so much related to free trade as to issues of fairness, transparency, and environment (JAMA, June 12, 2001, V285, i22, 2844). Other specific concerns hover around generic pricing as it relates to South Africa. In that situation 39 drug companies tried to stop the importation of cheaper medicines and the substitution of generics, as was permitted under TRIPS. However, the lawsuit was dropped due to global political pressure on the companies.

On the employment impact side, competition from abroad has made little indent on the market of domestic firms. This might be because the American firms dominate the global market, where 6 of the top 10 firms are based in the United States (S&P: Industry Survey, June 28, 2001, 9).The statistics show that over the 1980-1999 period, pharmaceutical companies have filed only 28 petitions with the International Trade Commission (ITC), covering 4,535 employees, of which most were denied (20 with 3,906 workers) and few were approved (8 with 629 workers). Perhaps the reason for the small import impact had been that pharmaceutical companies have always been on guard against foreign competition. They use strategies such as drug licensing, joint ventures and mergers to counter foreign competition. The push for intellectual property rights through the WTO's TRIPS rules illustrates the former.

Beginning in the mid-1980s, the pharmaceutical industry has been characterized by larger and more frequent merger and acquisition activity. The threat of patent expirations has influenced the increased merger activity within the industry. Many pharmaceuticals with high sales histories fear losing their patent protection and face competition from generic copies. There has been evidence that sales can decrease by as much as 75 percent in the year preceding patent expiration. Through merging with other industry players, pharmaceutical companies are able to pool their advertising, R&D, and productivity efforts while simultaneously cutting costs.

While the literature has extensively assessed prices, profits, and R&D efforts, advertising and productivity efforts were generally given a lower profile. These issues were considered by the Kefauver Committee in 1962 and again by the Subcommittee on Monopoly in 1976, which notes that "The drug industry has vast resources at its disposal. Its expenditure for advertising and promotion of drugs is now well over $1 billion per year or about $5,000 per physician per year" (Subcommittee 1976, 1395). Considering that advertising outlay is substantially greater than R&D outlay, and that long-term productivity is a major source for its economic growth, it is surprising that the study of advertising and productivity have taken a back-seat in the literature.

In this paper we develop four hypotheses and associated corollaries in order to perform an integrated analysis of seven major firms in the pharmaceutical industry. The model section develops these hypotheses around the concepts of scale economies, R&D, advertising, and Total Factor Productivity. The rest of the paper is divided into sections on the statistical results of at least four equations that were developed to evaluate the hypotheses with data.

4. Model

Traditional models focus on how prices and quantity in a market are determined. The pharmaceutical industry is useful for investigating price and non-price competition. Commenting on Caves et al., 1991 article, Pakes (1991) wrote that it is "cleaner" than most related industrial organization problems for several reasons. First, there is a legal monopoly for the first T years of the product's existence, and then free entry occurs at a fixed sunk cost thereafter (the cost of approval by the Federal Drug Administration), giving us a well-defined set of rules to determine possible market interactions. Second, it is reasonable to argue that there are common and fairly constant costs of production for the drugs being sold. Third, after the introduction of the branded drug, there seems to be only one major type of investment (advertising), and we have reasonably detailed data on it. There is, however, a difficult set of economic problems in modeling demand and in defining precisely what we mean by "brand loyalty" (Pakes, 1991). We model the pharmaceutical industry from its price and non-price aspects, and bring out rivalry among the major firms within this framework. No one model is broad enough to account for all of these activities, hence we begin with some simple abstraction, namely that the firms arrive at a pricing strategy through non-market means, and that the firms rely heavily on non-price competition for their survival.

Fudenberg et al., (1993) provide the essence for a model in the form of a time game where a firm's strategy set includes a time to either "stop" or "not stop" their efforts, which can include pricing, R&D, advertising, and productivity efforts. A firm can, for instance, lower its R&D cost to C, from a higher level, C'. Assuming it rivals do not react, the firm will expect a stream of benefits V(t) from such research efforts. Such a formulation allows one to estimate social gains if a social gain function can be specified. Scherer (1967) was one of the earliest to advance such a model within a profit maximization framework to predict a firm's market structure. He gave it an exponential form, which was further expanded by other authors. Reinganum (1984) has summarized some of those models to derive stylized facts about firm's size, excess capacity, strategies for increasing or decreasing efforts in equilibrium, leadership role, pre-emption strategies, and licensing. Some of the modern features of the exponential model are summarized in Tirole's book (1997, Ch. 10). Rather than developing the model here in symbolic form for each of the variables, we next discuss the form it takes in the estimation of each of the price and non-price categories.

4.a. Price Competition

On the demand side, the consumer is not the one that usually makes the choice of using a particular drug. Mostly, drugs are prescribed by physicians, who sometimes lack the necessary information about relative prices (Ellison et al., 1997, 437). Consumers, in attempts to gather or aggregate decentralized information, may want to free-fide on information from another patient that has already gone through that experience, the so called "herd" behavior effect (Choi 1997, 409-410).

On the supply side, pricing strategies are complicated by the fact that a firm can transfer or license encoded experience to other firms (Levitt and March 1990, 24). The tendency has been for firms in developed countries to press their government for strengthened patents regulation in foreign, particularly less developed, countries. Domestic manufactures claim that they can sell abroad at higher prices if patent laws are strengthened. Even developed countries that trade with the U.S. were reluctant to agree. Canada signed a law in 1987, allowing 7-10 years exclusivity to new drugs from abroad. Several other developed nations, including Japan and the EC, have pledged to adopt more uniform patent laws recommended by the Uruguay round of negotiations under GATT. The Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS), under the WTO seeks minimum 20 years patent protection for pharmaceutical products, which should be in full effect by 2006.

Models on the global scene have demonstrated that price discrimination is evident in the industry performance. Schut and VanBergeijk (1986) have argued that where patents are allowed, higher prices are expected to prevail. On the other hand, lower prices are expected if competition is encouraged, if the market is large, and if price control is practiced. Levy (1999, 74) offers several "competing explanations for observed price differences. Differential pricing may be the result of increased opportunities for price discrimination or may reflect the presence of quality or cost variations in different segments of prescription drug markets." It seems that price discrimination and differentiation are driven by cost, product amenities, and not just by the vulnerability of a certain segment of the population which can afford it.

Price competition is further complicated by the presence of generics. After considering the pros and cons for increase or decrease in brand versus generic pricing, Scherer concluded that the most likely scenario is ... "for the incumbent to maintain or increase its price, while ceding a substantial share of the market to much lower-priced generic rivals" (1993, 101). Such complications are also present in the pricing of "me too" drugs, which are variations of drugs already in the market place, "orphan" drugs that do not have a parent company, and "OTC" drugs.

It is clear that the pricing of pharmaceutical products is not the result of spontaneous or induced market forces. Forces internal to the cost structure of the firms and external to the firms' organizational structure contribute to price formation. The idea of price discrimination is high on the hierarchy of pricing strategies related to brand versus generics, and across international trading partners that adopt varying degrees of foreign patent protection.

4.b. Non-Price Competition

Concerns about non-price competition focuses mostly on R&D and patents, and somewhat less on advertising, sales promotion, and productivity. On the R&D side, the central thesis of the "exogenous growth" point of view under non-price competition for the pharmaceutical industry is Schumpeter's argument that innovation is a public good that must be encouraged by a patent system. However, from the endogenous growth point of view, the R&D sector is primary. To produce a new chemical entity (NCE) or a new molecular entity (NME), the pharmaceutical manufacturer must lay out a substantial portion of R&D expenditures, estimated at over $500 million, seek patents for its invention, and advertising to promote knowledge of its product.

The pharmaceutical industry holds the special dishonor in that the cost of patent protection for this industry is about three times the commercialization of innovation relative to petroleum, machinery, and fabricated metal products, and far more relative to autos or textiles (Mansfield 1986). Firms, however, are not dissuaded by the cost barrier. We enumerate some of the reasons of this in the literature, with a view to be able to specify a statistical model for non-price competition in the pharmaceutical industry, as follows:

1. Patents are arranged long enough to allow returns to cover R&D outlay. This has been the foundation of the Waxman-Hatch Act of 1984,

2. Competition seems to be driven by the "Patent First" initiative. This has been modeled in a gaming situation between two firms--one leading, the other following. Fudenberg and Tirole (1993, 123) argue that "it is optimal for each player in such games to stay in until discovery once his opponent has quit,"

3. It is not easy to imitate a good discovery. Successful imitation requires substantial learning (Mansfield 1968). Specialized imitation models were examined for the drug and auto industries. Grabowski (1968) has investigated imitation in the pharmaceutical industry along with the chemical and petroleum industries, and

4. Models of R&D follow a memoryless, Poisson state, implying that if there is any reaction at all, it is immediate (Tirole 1997, 384). It implies that lagged R&D may not be a good specification for a rivalry situation such as that advocated by Grabowski and Baxter (1973). We will examine this lagged aspect empirically around the central thesis of this paper, which is with regards to the imitation feature of firms, whether they react to a leader or just to each other.

Although "endogenous growth" models classify R&D as a nonrivalrous good, we find tremendous rivalry among firms. However, we do find that this industry is a good example of willingness to cooperate with the government by paying higher patent review fees in order to speed up the patent approval process. In close concert with this collaborative effort detailing and sales promotion efforts are also important for the industry after R&D. According to Caves et al. (1991), "The pharmaceutical innovators have two principal instruments, price and sales-promotion outlays, for maximizing the value of their innovations, both during the period of exclusive marketing and in the post-entry game" (Caves et al., 1991, 5). We observe that the large companies spend up to three times on marketing as they do on R&D. Detailing requires the firms to maintain a large staff, and hence high fixed cost, in order to inform the medical profession about the firm's product. It accounts for about three quarters of the firm's promotion outlays.

In an integrated model, we also examine rivalry from the Total Factor Productivity (TFP) perspective of the firm. This aspect is posed in the form of whether "basic research, as contrasted with applied research and development, (does) make a significant contribution to an industry's or firm's rate of technological innovation and productivity" (Mansfield 1980, 863). As presented, this model requires a distinction between basic and applied R&D, which is not generally available to firms. Mansfield estimates that the percent of R&D expenditure for basic research declined from 20.7 to 15.3 percent between 1967-1980 in the drugs industry, and points out that the latter forecast are not data consistent with the earlier ones (Ibid., Table 2). A more detailed view will distinguish between "product" versus "process" knowledge, implying that "firms are even willing to reduce the quality of their products to increase productivity," in cases where they would standardize their product. (Thompson and Waldo 2000, 158-159) Following Mansfield (1988, 223), we will use R&D expenditure as a measure of the firm's R&D capital, without the basic and applied distinction. We then concentrate on "process" knowledge measured by TFP in the sense of Thompson and Waldo.

In sum, we have formulated the following hypotheses to enable our statistical investigation.

Hypothesis I (Scherer and Ross 1990, 657-658): Economies of scale and timing games are important for pharmaceutical firm choices, behavior, or survival.

While economies of scale in production are not important for the pharmaceutical industry, based on the small scales nature of fermentation processes (Caves et al., 1991, 8), they are important from the R&D and promotion side. The literature, to our knowledge, has neglected this line of research. Following Scherer (1999), we provide this analysis for firms R&D outlays. It requires fitting the firm's R&D expenditures on the firm's sales, either through a linear or polynomial specification. The mere fact that R&D expenditures may escalate as rivals compete to get a patent first on a new product makes such a study necessary. We note that, subject to diminishing returns, the expected completion date of a project can be shortened through a firm's increasing R&D expenditures, which also underscore economies of scale. Also, because patents can be licensed to rival firms, R&D outlays can spillover since the research findings of a firm is available to others free of charge, a sort of external economy. Scherer also underscored the view "that advertising, by making known the availability of new products, enables innovators to tap larger markets more rapidly, enhancing the profits from innovation and hence strengthening incentives for investment in innovation" (Scherer 1980, 378).

The Grabowski and Vernon (1977, 361) three industry study and Scherer's (2001, 657) 196 industries study advocate regression analysis to establish reaction patterns between innovational output, such as R&D or patent and sales, in linear and polynomial form. Grabowski and Vernon have tried innovative output linearly on sales alone, including a polynomial up to the third degree. They were only after a good fit, reporting an [R.sup.2] of 65 percent. Scherer advocated only a second degree polynomial, which is consistent with Grabowski and Vernon's specification. We have decided to adopt the latter, which implicitly involves fitting the following equation:

Output = f(Sales, [Sales.sup.2]) (1)

Hypothesis II (Griliches 1990, 1702): Patents are good indicators of differences in inventive activity across firms.

Corollary I to Hypothesis II: Patents are an output and R&D an Input to the firms rivalry process.

Corollary II to Hypothesis II (Tirole 1997, 394): Efforts to obtain a Patent via R&D expenditure proceeds in a "memoryless" or "Poisson" state.

The hypothesis requires a test for association between a firm's R&D expenditure and the number of patents. This relationship sheds light on the innovation process and technical change. Cross-sectional statistics explain differences in a firm's inventive activities. However, rather than correlating just R&D with patents, we will expand on this model in the direction of rivalry, i.e., pitting one firm's R&D outlay against another, taking into consideration the number of patent the firms received in the previous period. In symbolic form:

[R&D.sub.it] = f([R&D.sub.j(t-1)], [Patent.sub.i(t-1)]). (2)

where "i" is the ith firm, "j" is the rival firm, and "t" is time. The above specification resembles Grabowski and Baxter's (1973) model, except that we substitute patent as an independent variable for their previous period R&D expenditure, and have dropped some collinear variables.

In Corollary I, we intend to use a firm's patent award in the previous period in order to cut through the controversy of whether patents are an input or output. According to Kamien and Schwartz (1975, 4), R&D is an input and patent is an output in the innovation process. We may think of an R&D input as generating an R&D output as diagrammed by Hay and Morris (1979, p. 444). Many authors have used it as a measure of a firm's capital for innovative input. Mansfield separates it into basic research and applied research capital and uses it as inputs in a firm's production function to explain total factor productivity (Mansfield 1980, 861). Grabowski and Baxter (op. cit., 1973) have advocated the use of a firm's R&D expenditure in the past period. However, because its influence may be dominated by previous patent awards, we use patent instead. Our substitution of patent as an independent variable in place of R&D in this context has other purposes as well. We use the patent variable in a conditional probability sense. A firm might, in the most naive sense, want to forecast its current level of R&D outlay based on current information. However, if some information is gathered on the firm's past patent's award and effectiveness, the firm may want to incorporate that information into its decision making on R&D as well, thus making a conditional probability forecast.

In Corollary II, our use of patents is also instrumental in throwing light on the "memoryless" or "Poisson" game of the firm's race for a patent. According to Tirole (1997, 394), in such a model, "a firm's probability of making a discovery and obtaining a patent at a point in time depends only on this firm's current R&D expenditure and not on its past R&D experience." As stated above, the literature for the specification of equation 2 above would require that we put a lagged [R&D.sub.it] variable in the place of patent. A significant statistic on such a variable would act as a potential falsifier of this "Poisson" model. In fact, we find that the patent variable does perform better.

On the empirical side of the literature on R&D rivalry, Grabowski and Baxter (1973) have demonstrated for eight firms in the U.S. chemical industry that rivals do react to the R&D outlays of other firms, when other determinants such as cyclical movements are controlled. They argued that firms do not match R&D expenditures as precisely as they do advertising. In an earlier work for the chemical, petroleum and drug industries, Grabowski (1968) discovered that such reactions proceed with a one period lag. However, the pharmaceutical firms made up only a subset of that sample. This study, on the other hand, exclusively considers pharmaceutical firms.

Hypothesis III [Grabowski and Mueller (1969, 1970, 1971)] Advertising competition among pharmaceutical firms leads to imitation of their advertising expenditures.

The pharmaceutical industry advertising budget has received attention since the Kefauver-Harris Act of 1962 noted that the industry is spending more on advertising than on R&D. Today, the pharmaceutical firms have about six major channels to target advertising expenditures: detailing, sampling, direct mailing, journal advertising, general media advertising, and ads directed at physicians for continuing medical education (Schweitzer 1997, p. 48). According to Measday (1977), "No other products on the market are promoted as intensively as ethical drugs," and ethical drugs had been the faster growing segment of the industry relative to proprietary drugs.

The traditional--perhaps antitrust--view of advertising is that innovative firms engage in large amount of advertising and promotional expenditures, which act as a barrier to entry for new and small firms. Another view is that advertising sells invention such as a new discovery, or the molecular imitation of a rival's new product. Although the way that the literature treats advertising hypothesis are not well-formed enough for statistical investigation, we find that this aspect of the industry has received a fair share of research (Comanor 1986, 1196). For instance, in his interpretation of the advertising effect of generic drugs on brand, Frank et al. (1992, 173-174), notes that "The econometric model of advertising and market share yielded estimates of the impact of number of sellers on the advertising effort of the leading firm." Another study by Caves et al. was primarily concerned with the relationship between advertising and the rate of return to R&D. We see the basis in these early studies for an imitative work in advertising among the rival firms. Accordingly we have formulated the following specifications for the investigation of Hypothesis III.

[Advertising.sub.it] = f([Advertising.sub.j(t-1)], [Cashflow.sub.i(t-1)]). (3)

where "i" is the ith firm, "j" is the rival firm, and "t" is time.

Hypothesis IV: (Comanor, 1986): Firm size may have an influence on technical advance.

This hypothesis involves the use of R&D activities as a measure of a firm's new products, and sales of new products as a measure of a firm's marketing activity. One issue here is how to separate the effect of advertising and sales promotion (Comanor 1986, 1191). We first estimate total factor productivity for the seven firms in our model using the following specification for each firm over time:

Q = [Ae.sup.[lambda]t][R.sup.[alpha]][L.sup.[omega]][K.sup.1-[omega]] (4)

where Q is the firm's value added, A is a measure of TFP and its growth rate, R is R&D capital, L is labor, and K is capital. We then proceed to measure total factor productivity (TFP) via the equation Log(Q/L) - vLog(K/L) - (1 - v)R/L for each firm. We integrate advertising into the sales and R&D hypothesis by using to calculate TFP instead of R&D in Eq. 3, and jointly estimate them with the advertising specification in Eq. 4.

Equations 2, 3, and 4 above take the Cobb-Douglas-type production function form, relating R&D and advertising, and indirectly TFP, to the firm's previous rival's outlay, conditioned on the firms current states of patent awards and cashflow, respectively. They are dynamic in that they involve lags and purport to show the influences of the firms changing states of patents and cashflow over time. Ordinarily, in a production like specification, the constant term would capture the firm's process technology, and advertising would make known to the public the product quality of the firms. However, such interpretation would be placed in the background until now where we introduce the traditional input of capital and labor in our estimation of the TFP model. The main issue we are poised to estimate with these specifications is that each firm would generate in a simultaneous manner, R&D, advertising, and "process" knowledge in order to enhance the quality and growth of their product, with an eye on their rival's previous level of performance.

5. Results

In this section, we first investigate the relationship between economies of scale to R&D and Patents for the seven firms. We then consider the rivalry question among firms. We also study R&D, advertising and productivity in an integrated manner.

5.a. Scale Economies

Table 1 presents our results of Eq. 1 above. It underscores that the inclusion of the constant term is necessary because they turned out mostly significant. In the fit of Grabowski and Vernon, the t-values were insignificant for both the linear and non-linear forms. This is the case for our results for only American Home Product where the coefficients of -157.79 in the patent, and 169.07 in the R&D equation, were insignificant. In the case of Pfizer, only the constant for patents, viz., 57.82 was insignificant. The other constant terms are significant.

The linear sales coefficients are significant for 8 of 14, implying that 6 of 14 are negative. Of the nonlinear coefficients, only 8 are significant, and half are positive indicating increasing returns. Adding up significant intercept and nonlinear coefficient, 5 of 14 are significant indicating constant returns. Four nonlinear coefficients are positive indicating increasing returns, which means that five other coefficients of which some are insignificant are indicating decreasing returns.

Compared with the Grabowski and Vernon results, whose sales coefficient in the linear model was 0.74, and 0.94 in the polynomial equation, our results are small, ranging between -0.006 for Pfizer, to 0.27 for Lilly. The sales coefficient was -0.88 x [10.sup.-3] in Grabowski and Vernon, which is in the vicinity of what we find. Their [R.sup.2] of 0.64 compares with ours, which are mostly in the 80-90 percent range.

5.b. Non-price Rivalry Results

The results in this section explain competitive reaction among the seven firms for market share or dominance. The contribution here is that pharmaceutical firms do compete with each other from the R&D and advertising perspectives. We first examine the rivalry from the possibility of the leader and followers hypothesis. This requires us to identify one firm as the leader among the seven. The literature is unanimous on this. We therefore examine all the pairs of firms, namely 42 ((7!/(7-2)!) = 42 in our model. We perform the examination for the two main rivalry weapons, advertising and R&D, in a SUR model in order to take care of inter-correlation about the residuals as a consequence of the firms being in the same industry.

We begin our estimation by looking at the literature for some guidance as to whether we can identify a leader, unambiguously. Scherer (1993), for instance, talks of Merck as a leader. Further, in his book, Scherer expresses deep concern with the relationship of R&D and market outlays within the context of uncertainties about market acceptance, Scherer mentions that firms are involved with both consumer and rivals' reaction to their product. "Although marketing research can provide some clarification, most of the uncertainties remain until well after the majority of R&D investment, and indeed, appreciable equipment and marketing rollout investment, has been shouldered" (1999, 65).

Traditional selection criteria would suggest that the examination of winners versus losers among R&D outlays; the size of the firm's portfolio of R&D projects; the number of patents awarded to stop imitators will give the innovating firm an advantage in recovering R&D expenditures. Indeed, some studies that look at the return to R&D suggest that the returns are highly skewed to a few top innovators. In their 99 new drugs study, for instance, Grabowski and Vernon (1990) found that the top ten deciles of 99 new drugs received about 55% of the profits, measured in terms of sales less cost of production, including marketing and distribution costs. This result also implies that the size of the portfolio of R&D projects is not a good measure of leadership because success will be skewed to only a few projects. We therefore fall back on the historical role of the firm, complemented with a few rule-of-thumb descriptive statistics to suggest a leader. Afterwards, a few cross-checks, such as bootstrapping and alternative discrete choice models, are performed to examine the robustness of the choice.

6. In Search of the Historically Dominant Firm

New medicines and vaccines derived from R&D efforts play an essential role in the progress of treatment of various forms of disease. Since the turn of the century, many causes of disease have been eliminated, and Americans of all ages have experienced progressive increases in life expectancy and improved overall health. Although drug trade and the use of plants and minerals for medicinal purposes date back several centuries, the industry began to have a noticeable presence beginning in the early 1930s.

Introduction of the first sulfa drug in 1935 led to an increased interest in pharmaceutical research and opened up the market for the launch of penicillin. 1938-1953 was coined "The Age of Antibiotics" as many new drugs were introduced to the market. In World War II, there was a great need for safe and easily administered drugs to protect wounds from infection. Although Alexander Fleming discovered penicillin in 1928, he was unable to produce the antibiotic in sufficient quantity to be of any clinical value. Twenty years later, Howard Florey and Ernst B. Chain produced solid penicillin and searched for vendors to produce the drug commercially. With the assistance of the United States government and several United States pharmaceutical firms, penicillin became the first commercially produced drug. After penicillin became a commercial product, it became evident that investment in the industry could lead to substantial profits and resources were allocated towards intensive research discovery.

Over the years, antibiotics and vaccines were important in the abolition of polio, measles, and other diseases. According to a PHRMA Study (Profile 99) of 152 major drugs developed between 1975 and 1994, 45 percent were developed in the United States. The data in Chart 1 illustrates how staggering death rates for various diseases in the 1920s have decreased over the years as the industry has developed.

Historically, the most important discovery has been penicillin, which was discovered in 1928 by Alexander Fleming. Because it was not patented, it was licensed freely by governments. Merck and Pfizer were early producers that enjoyed a sizeable mark-up over cost advantage. But the original price of $6000 per billion units in 1945 fell to about $100 and reached $15 by 1962 when entrants were able to penetrate the market (Measely, 267). Following penicillin, the next most important discovery was antibiotic, in particular tetracyclines that combat a wider range of organisms than penicillin. According to Scherer (1990), Pfizer held the markup advantage of about $60 per 100 tablets bottle. From this historical sketch, therefore, both Merck and Pfizer were leaders in the penicillin revolution, with Pfizer dominating in the post-penicillin antibiotic revolution. We now should take a look at the more current positions of these two firms to find out if other leading role reversals are present.

7. Some Descriptive Statistical Measures of Lead Role Reversals

Advances in biomedical science have helped pharmaceutical researchers develop new drugs and cures to attack various diseases. A PHRMA Industry Study (1) of R&D expenditures and sales have provided the following results:

1. R&D expenditures by research-based pharmaceutical companies have increased by 14.1 percent between 1998 to 1999, reaching about $24.0 billion.

2. Over our sample period of 20 years, the percentage of United States sales allocated to R&D has increased from 11 percent to 20.8 percent. Meanwhile, the average R&D-to-sales ratio for all United States industries is less than 4 percent.

3. Total drug development time has increased from an average of 8.1 years in the 1960s, 11.6 years in the 1970s, 14.2 years in the 1980s, to 14.9 years for drugs approved during 1990-1996.

4. About 350 fifty new biotechnology medicines (produced by 140 companies) are in the pipeline for development, and

5. Effective R&D requires an exorbitant amount of capital in order to succeed. A Boston Consulting Group estimate states that the pre-tax cost of developing a new drug in 1990 was $500 million, which includes the cost of research failures in addition to interest costs over the life of the investment. As the length of time needed to develop the drug increases, so do the costs. These costs of capital increase as pharmaceutical companies are exposed to economic risk and uncertainty over a longer period. In addition, since the cost of developing new drugs is so expensive, commercial success is usually possible only for a few products, therefore, companies concentrate their R&D efforts on fewer products.

To help in the assessment of leadership, we have tabulated some descriptive statistics on the seven companies in Table 2. While Merck and Bristol-Myers are not far apart in R&D, Merck is clearly the leader in terms of sales force, median number of patents registered over the sample period, number of products, and market share. It is interesting to note that the company's percentage change in R&D is typically less than the percentage change in the industry's R&D during the sample period. It implies that smaller firms in the industry are making larger percentage expenditures on R&D in order to keep the industry percentage change above those of the firms in our sample. The large firms' first preference is not to escalate their R&D expenditure above the industry percentage, perhaps with a view to avoid R&D wars. Their second preference is to keep the change in their R&D outlay in line with the industry level of change, which demonstrates a Cournot type rivalry. Further, Merck chooses not to beat the industry's percentage as indicated by the fact that it has done so only one time in 19 years.

Pfizer, which shows aggressive historical dominance as discussed above, is low on the time series of R&D and patent trend, with a median of $585.5M and 90, respectively. It has the dominant effort in domestic sales force, 7,600, which may mean that it wants to use sales promotion strategies in order to bring up its market share, which is now below Merck's. Because some of the data in market share and list of products are missing, we can only conclude that it appears that Merck and Pfizer dominate. However, to leap to a conclusion that they lead the industry would be a hasty conclusion. We therefore will consider all the rivalry combination between any two firms in their competition in R&D and advertising expenditure in order to shine more light on the leadership hypothesis, which underscores that the firms engage in non-price over price competition.

8. Statistical Results on R&D

Although Merck leads in both the historical and size measure points of views, we plan to see how each firm responds to each other's lagged level of expenditure. Table 3 gives the rivalry results for the 42 different combinations of the firms in an R&D rivalry situation. The model that generates this result is from Hypothesis II above, effectively a combination of specifications from Grabowski and Baxter (1973) and Griliches (1990). A novelty of our approach is that firms make R&D outlays in a rivalrous domain in order to be the first to get a patent.

One notable feature of the results is that the rivalry does not proceed in a memoryless state via the Poisson probability distribution for success. In fact the firms do remember their lagged period's award of patents, which they put side-by-side their rival's previous period R&D outlay in order to make a decision on their current period outlay of R&D. Although the advertising results are discussed separately in the section below, the R&D and advertising rivalry decision are estimated jointly. The model has worked well in that most of the R^2 are in the 90 percent range.

The results of Table 3 can be reviewed in seven clusters depending on which firm is positioned as the leader. The first cluster indicates that firms respond positively to Merck's previous period outlay of R&D outlay. The responses vary from 0.96 for Johnson & Johnson to 1.38 for American Home Products, and they are all significant at the 99 percent confidence level. The average reaction for the six firms on Merck's outlay is 1.23, which indicates an elastic response overall, indicating that the firms will always tend to do more R&D than Merck's previous period outlay. The influence of Merck's previous period patent award indicates an overwhelming negative influence on other firm's R&D behavior. The reacting firms exhibit strong level of complacent behavior in that when their patent awards are up, they tend to ease up on their current R&D outlay, even though their rival's previous period R&D might be up. This behavior underscores that getting a patent is as important as the literature underscores.

The second cluster indicates firms' reactions on Pfizer's previous period outlay of R&D, which are all significant at the 99 percent level as well. However, we note some differences of reaction to Pfizer from what we saw for Merck as the leader. Only American Home Products reacts fully with a coefficient of 1.09. The other firms indicate partial reaction ranging from 0.54 to 0.84. American Home Product is also the loner in this cluster in complacency relating to patents awarded. The other firms react positively.

The other five clusters indicate reactions to Abbott, American Homes Product, Bristol-Meyers, Johnson and Johnson, and Lilly, respectively. That firms react positively in these clusters is not in question. However, full reaction is limited to only American Homes Products and Pfizer versus Abbott; Bristol-Meyers versus American Home Products; no firms versus Bristol-Meyers; all but Lilly and Merck versus Johnson and Johnson, and all but Johnson and Merck versus Lilly. Whether the reacting firms are complacent based on their previous period patent awards is less significant in these five clusters. We have 18 of 30 significant patent coefficients, of which 11 are with negative signs, indicating a dominance of the complacency hypothesis.

Overall, we find that firms are reacting to each other's previous period R&D outlay, indicating strong non-price competition in that area. However, we are able to underscore that they react to the apparent leader, Merck, in an elastic way. The dictum seems to be that what the apparent leader does, the generality of firms tends to imitate strongly. We have overwhelming confirmation that firms are happy with relative high past period award of patents. It allows them to be complacent about non- price competition by allowing them to ease up on their current period outlay of R&D.

9. Statistical Results on Advertising

Table 4 contains the results for advertising. Again the results are grouped in seven clusters, based on each finn being given the opportunity to lead. Of the 42 advertising coefficients, 28 are significant, of which only the reaction to Lilly are of the incorrect (negative) signs. The size of the significant advertising coefficients on Merck is 0.23 to 0.78, Pfizer is 0.32 to 0.39, Abbott is 0.27 to 106, Bristol-Meyers 0.66 to 2.10, Johnson and Johnson 0.69 to 1.45, and Lilly is -1.1 to 0.39.

Focusing on the signs for both the advertising and cashflow coefficients helps us to narrow down the leader in the case of advertising rivalry. This is suggested by the fact that while all of the advertising coefficients are significant with Johnson and Johnson as the leader, the results of the cashflow variable for that case are not supportive. They are significant only for American Home Products and Lilly, but the signs are negative rather than being positive. The joint cases where both of the coefficients are significant and are of the correct a priori signs is one for Lilly, two each for Pfizer, Abbott, and Bristol, and three each for Merck and American Home Products as leader.

There are criteria that would make advertising leadership swing between Merck and American Home products. Using size of the advertising coefficient in the joint cases, the American Home Products will come out ahead with the vector of advertising coefficient of [0.78, 0.74, 0.60 ] versus the vector [0.23, 0.37, 0.37] for Merck. However, using R2 for the joint cases, Merck's average will be 0.93 (0.92+0.95 + 0.92)/3), and American Home Products' average will be 0.83 (0.83 + 0.91 + 0.93)/3), which is lower, thus giving the leadership to Merck. We also note that while American Home Products is reacting to Merck, Merck does not react to American Home Products. This asymmetry further tilts the leadership towards Merck.

Both the R&D and advertising rivalry results indicate a leader versus follower type of non-price competition. The analysis points to Merck as the industry leader. This underscores a remark by Scherer that Merck was the industry leader in 1992 (Scherer 1993, 102). In the R&D battle ground, firms show the highest reaction to Merck and smaller reactions to other firms. In the advertising arena, the most significant reactions again point to Merck as the leader.

10. Firm Size, and Innovation through a Total Factor Productivity Model

Hypothesis III addresses rivalry through an environment in which the size of a firm's influences its innovative activities. The model we identified above was intensively measured by Mansfield (1980, 1983, 1988) at the firm, industry, and country levels. We have adopted this model for our firm size versus innovation hypothesis. It requires us to use some different measures of the data. We have used the S&P definition for total capital and employment, sales as a proxy for Q, and, following Mansfield, R&D as a measure of innovation. The traditional use of the model is to regress TFP against time to obtain an estimate of its growth rate, [lambda], or against R&D to obtain the influence of R&D on TFP. Instead, we use it in our rivalry equations above to ascertain rivalry. In the table, we indicate with an "x" only regression equations with significant reaction coefficients of the proper signs. To fit the equations for each firm, we experimented with a variety of techniques such as SUR and 3SLQ, and several specifications such as Cobb-Douglas and CES. The best fits were associated with 3-stage least squares, using the log values of the firms' patents as instruments. The results are as follows:

The results for the first phase of the calculations in Table 5 indicate that the R&D variables are significant for Abbott, and Bristol. All the capital coefficients are significant, and only one intercept term, viz., for Bristol, is insignificant. We use these results to calculate the TFP, insert these values into EQ1 for R&D, and then proceed to estimate the second phase of the analysis. Because we estimate only nine of 42 TFP reaction coefficients, we qualitatively discuss the results via a grid.

 Reacting Firms to Leaders TFP
Leading in the Previous Period

Firms: Pfizer Abbott Bristol Merck

Abbott X x x
American X x
Britol X x
Johnson X
Lily x

We note that Pfizer reacts in its TFP rivalry to all the firms excepting Merck and Lilly, Abbott reacts to Bristol and Lilly, Bristol reacts to Abbott and American; and Merck reacts only to Abbott. The notable feature of the grid is that Pfizer and Merck are absent from the leading role in TFP rivalry, and Pfizer is a strong reactor to small firms TFP. Along with the one instance of Merck reaction to Abbott, we lean towards the conclusion that a large firm shows a significant reaction to a small firm in TFP.

From the above observations we lean towards the conclusion that small firms are leaders in total factor productivity rivalry; at the very least, large firms do not have a TFP advantage over small firms. This is reminiscent of the Kefauver Committee argument that innovation is not necessarily occurring within the firms, but is instead occurring outside of the industry (Comanor 1986, 1189). An alternative explanation as to why the large firms react to small firms is because small firms are most likely to seize the opportunity to develop a new molecular rather than a new chemical entity. According to a USDC study, "As soon as a new chemical compound with useful physiological activity is discovered and patented by a pharmaceutical company, numerous competitors try to improve it by finding a new patentable chemical variant (through 'molecular manipulation')" (USDC 1984, 11). This conclusion is also at the heart of the Kamien and Schwartz conclusion that non-price competition is "primarily a tool of small firms seeking profit improvement by introducing new substitutes for the existing product" (Kamien and Schwartz 1975, 15). Therefore, it would be in the interest of large firms to monitor the small firms TFP activities closely.

In sum, while we found that Merck is a viable leader in R&D and advertising separately, we find that large firms have a tendency to follow small firm's activities when TFP is the rivalry weapon. The fact that a firm's capital and labor are behind the calculation of TFP may account for the differences in the two divergent conclusions. However, the drive to get a NME rather than a NCE is more in the province of the smaller firms that are more likely to be strapped for cashflow and behind in their patent awards. We recognize the need for better measures of R&D capital, NCE instead of sales, and perhaps a separation of basic from applied research. Improved data is required to further investigate these conclusions. However, at this point we observe that the size versus innovation hypothesis holds out a contrary possibility for the large versus small firm rivalry hypothesis that exists under the R&D and advertising hypothesis.

11. Data

The firm financial data is from the S&P 500 Stock Market Encyclopedia and Compustat. Industry R&D is taken from Pharmaceutical Research and Manufacturers of America (PhRMA Annual Survey, 2001, 117). Patent data is taken from the United States Patent & Trademark Office, Patent Full Text and Image Database (http://www.uspto. gov/patft/index.html). Advertising data represents the cost of advertising media (radio, television, newspapers, and periodicals) and promotional expenses, and excludes selling and marketing expenses. Compustat has another series entitled selling, general and administrative expenses, which we did not use because it included R&D expenditure whose influence we are separating from advertising. For the TFP hypothesis, we used S&P's employment and total capital data.

12. Conclusion

We investigated a broad array of hypotheses related to the pharmaceutical industry. For the globalization period, we find that R&D is a focal variable that is supplemented with detailing and sales promotion activities. Although R&D is considered a nonrivalrous good from the "endogenous growth" model viewpoint, firms do compete with respect to outlay as they are showing willingness to cooperate with the government on the other. We found that the seven firms we investigated over the 1980-1999 period react to each other in their R&D and advertising outlays. We found leadership-follower patterns, where the other six firms react to Merck's previous period outlay of R&D and advertising, given the state of their patent awards and cash flow, respectively. We arrived at these results using traditional models in the literature with careful econometric specification and specialization to the pharmaceutical industry. To our knowledge, the mere finding of rivalry reaction patterns is here investigated for the first time in the literature. The focal concern of firms on R&D in this industry justifies making it the primary sector from the "endogenous growth" model viewpoint. It confirms the belief that the pharmaceutical industry may have traditional structural pattern, despite its reputation of being concentrated in R&D and patent activities.

Our results also found that rivalry in regards to TFP does not honor large firm dominance over small firms. As noted above, this touches on an old concern of the Kefauver Committee investigation that most new discovery can take their source from outside the industry. Because of their disadvantage in size, we have suggested that an explanation for this behavioral pattern is anchored in small firms' likelihood to look for NMEs rather than for NCEs. Small firms are most apt to copy an industry technology, for instance by being aggressive to create a new molecular entity. This is their way to make inroads into the larger firm share. Such activities over time are captured in the TFP measure, of which we find that large firms are mindful.

References

Agrawal, M., Global Competitiveness in the Pharmaceutical Industry, (Binghamton, NY: The Haworth Press, 1999).

Balance, R., J. Pogany, and H. Forstner, The World's Pharmaceutical Industries: An International Perspective on Innovation, Competition, and Policy, (Aldershot, England: Edward Elgar Publishing Ltd., 1992).

Berndt, E.R., R.S. Pindyck and P. Azoulay, "Consumption Externalities and Diffusion in Pharmaceutical Markets: Antiulcer Drugs," NBER Working Paper 7772, June 2000.

Berndt, E.R., I.M. Cockburn, and Z. Griliches, "Pharmaceutical Innovations and Market Dynamics: Tracking Effects on Price Indexes for Antidepressant Drugs," Brookings Papers on Economic Activity, 1997, pp. 133-199.

Caves, R.E., M.D. Whinston, and M.A. Hurwitz, "Patent Expiration, Entry, and Competition in the U.S. Pharmaceutical Industry," Brookings Papers on Economic Activity, 1991, pp. 1-45. Choi, J.P., "Herd Behavior, The "Penguin Effect," and the Suppression of Informational Diffusion: An Analysis of Informational Externalities and Payoff Interdependency," The RAND Journal of Economics, Volume 28, Number 3, Autumn 1997, pp. 403-425.

Comanor, W.S., "The Political Economy of the Pharmaceutical Industry," Journal of Economic Literature, Volume 24, Issue 3, September 1986, pp. 1178-1217.

Danzon, P. M., Pharmaceutical Price Regulation, (Washington, DC: The AEI Press, 1997).

Dunning, J.H., "International Direct Investment in Innovation: The Pharmaceutical Industry," In Multinationals, Technology, and Competitiveness, (London: Unwin Hyman, 1988).

Ellison, S.F., I. Cockburn, Z. Griliches, and J. Hausman, "Characteristics of Demand for Pharmaceutical Products: An Examination of Four Cephalosporins," The RAND Journal of Economics, Volume 28, Number 3, Autumn 1997, pp. 426-446.

Frank, R.G., and D. Salkever, "Pricing, Patent Loss and the Market for Pharmaceuticals," Southern Economic Journal, Volume 59, Number 2, October 1992, pp. 165-179.

Fudenberg, D. and J. Tirole, Game Theory, (Cambridge, MA: The MIT Press, 1993).

Grabowski, H., "The Determinants of R and D in Three Industries," Journal of Political Economy, Volume 76, 1968.

Grabowski, Henry G., and Nevins D. Baxter, "Rivalry in Industrial Research and Development," The Journal of Industrial Economics, Volume 21, Number 3, 1973, pp. 209-235.

Grabowski, H., and Mueller, D.C., "Non-Price Competition in the Cigarette Industry," The Antitrust Bulletin, Volume 14, 1969, pp. 607-628.

Grabowski, H., and Mueller, D.C., "Non-Price Competition in the Cigarette Industry: A Comment, The Antitrust Bulletin, Volume 14, 1970, pp. 679-686.

Grabowski, H.G., and Mueller, D.C., "Imitative Advertising in the Cigarette Industry," The Antitrust Bulletin, Volume 16, Summer 1971, pp. 257-292.

Grabowski, H.G., and J.M. Vernon, "Innovation and Invention: Consumer Protection Regulation in Ethical Drugs," AER Proceedings, Volume 67, Number 1, 1977, pp 359-364.

Grabowski, H.G., and J.M. Vernon, "A New Look at the Returns and Risks of Pharmaceutical R&D," Management Science, Volume 36, July 1990, pp. 804-821.

Grabowski, H.G., and N.D. Baxter, "Rivalry in Industrial Research and Development," The Journal of Industrial Economics, Volume 21, Number 3, 1973. pp. 209-235.

Griliches, Z., Patent Statistics as Economic Indicators: A Survey", Journal of Economic Literature, Volume 28, December 1990, pp. 1661-1707.

Hay, D.A., and D.J. Morris, Industrial Economics: Theory and Evidence, (New York: Oxford University Press, 1979).

Levitt, B. and J.G. March, Chester I. Barnard and the Intelligence of Learning, in Oliver E. Williamson ed., Organization Theory, (New York: Oxford University Press, 1990), pp. 11-37.

Kamien, M. I., and N. L. Schwartz, "Market Structure, Elasticity of Demand, and the Incentive to Invest," Journal of Law and Economics, Volume 13 Number 1, 1970.

Kamien, M. I., and N. L. Schwartz, "Market Structure and Innovation: A Survey," Journal of Economic Literature, Volume 13, Issue 1, March 1975, pp. 1-37.

Kolassa, E. M., Elements of Pharmaceutical Pricing, (Binghamton, NY: The Haworth Press, 1997).

Levy, R., The Pharmaceutical Industry: A Discussion of Competitive and Antitrust Issues in an Environment of Change, (Washington DC: Bureau of Economics Staff Report, Federal Trade Commission, March 1999).

Mansfield, E., The Economics of Technology Change, (New York: Norton, 1968).

Mansfield, E., Basic Research and Productivity Increases in Manufacturing," The American Economic Review, Volume 70, Issue 5, December 1980, pp. 863-873.

Mansfield, E., "Technological Change and Market Structure: An Empirical Study," American Economic Review, Volume 73, Issue 2, May 1983, pp. 205-209.

Mansfield, E., "Patents and Innovation: An Empirical Study," Management Science, Volume 32, Number 2, February 1986, pp. 173-181.

Mansfield, E., "Industrial R&D in Japan and the United States: A Comparative Study," American Economic Review, Volume 78, Issue 2, May 1988, pp. 223-228.

Measday, W., The Pharmaceutical Industry, in Walter Adams, ed., The Structure of American Industry, Fifth Edition, (New York: Macmillan Publishing Co., Inc, 1977), pp. 250-284.

NAE, The Competitive Status of the U.S. Pharmaceutical Industry: The Influences of Technology in Determining Industrial Competitive Advantage, (Washington, DC: National Academy Press, 1983).

Pakes, A., "Comments and Discussion on Richard E. Caves, Michael D Whinston, and Mark A. Hurwitz, "Patent Expiration, Entry, and Competition in the U.S. Pharmaceutical Industry," Brookings Papers on Economic Activity, 1991, pp. 1-45.

Reinganum, J., "Practical Implications of Game Theoretic Models of R&D," American Economic Review, Volume 74, Issue 2, May 1984, pp. 61-66.

Phillips, A., "Concentration, Scale and Technological Change in Selected Manufacturing Industries, 1899-1939," Journal of Industrial Economics, June 1956.

Scherer, EM., New Perspectives on Economic Growth and Technological Innovation, (Washington, DC: Brookings Institution Press, 1999).

Scherer, EM., "Pricing, Profits, and Technological Progress in the Pharmaceutical Industry," Journal of Economic Perspectives, Volume 7, Number 3, Summer 1993, pp 97-115.

Scherer, EM., and David Ross, Industrial Market Structure and Economic Performance, Third Edition, (Boston: Houghton Mifflin Company, 1990).

Scherer, F.M., Industrial Market Structure and Economic Performance, Second Edition, (Chicago: Rand McNally College Publishing Company, 1980).

Scherer, F.M., "Research and Development Resource Allocation Under Rivalry," The Quarterly Journal of Economics, Volume 81, Number 3, August 1967, pp. 359-394.

Schwartzman, D., Innovation in the Pharmaceutical Industry, (Baltimore, MD: Johns Hopkins University Press, 1976).

Schweitzer, S.O., Pharmaceutical Economics and Policy, (New York: Oxford University Press, 1997).

Shut, ET., and EA.G. Van Bergeijk, "International Price Discrimination: The Pharmaceutical Industry," World Development, Volume 14, Number 9, 1986, pp. 1141-1150.

Taggart, J., The Worm Pharmaceutical Industry, (London: Routledge, 1993).

Thomas, L.G., "Implicit Industrial Policy: The Triumph of Britain and the Failure of France in Global Pharmaceuticals," Industrial and Corporate Change, Volume 3, Number 2, 1994, pp. 451-489.

Thomas, L.G., "Industrial Policy and International Competitiveness," In Competitive Strategies in the Pharmaceutical Industry, edited by R.B. Helms (Washington, DC: AEI Press, 1996), pp. 107-129.

Thompson, P. and D. Waldo, "Process versus Product Innovation: Do Consumption Data Contain Any Information?" Southern Economic Journal, Volume 67 Number 1, 2000, pp. 155-170.

Tirole, J., The Theory of Industrial Organization, (Cambridge, MA: The MIT Press, 1997).

USDC: A Competitive Assessment of the U.S. Pharmaceutical Industry (United States Department of Commerce, Westview Press, 1986).

Note

(1.) www.searchforcures.com/publications/ industry.

Lall Ramrattan, Lecturer, Department of Economics, California State University, EastBay, lall_b._ramrattan@hud.gov

Michael Szenberg, Corresponding Author, Lubin School of Business, Pace University, NY, NY, 10038, mszenberg@pace.edu. This article will appear in Distressed U.S. Industries in the Era of Globalization (Ashgate Press, 2007) coauthored by the above.

TABLE 1
Sales and Innovation

R&D and Sales data is from Standard and Poors Industry Surveys, various
years, and also available from the Compustat Tapes. Patent data are
extracted from the U.S. Patent and Trade Mark: Full Text and Image
data base (http://www.uspto.gov). The * below the dependent variable
indicates which of the two equations performed better from the
statistical point of view.

Dependent Constant Sales Sale^A

Abbott: Pat.: 184 -0.07 0.000001
* (10.95) *** (-9.06) *** (10.80) ***
R&D: -157.79 0.12 0.000002
 (-7.03) *** (12.15) *** (0.17)
Am.: Pat: -0.55 0.04 -0.000001
* (-1.31) (2.35) ** (2.01) **
R&D: 169.07 -0.06 0.00001
 (1.26) (-1.36) (3.69) ***
Bristol: Pat.: 17.94 0.005 -0.0000001
 (1.81) * (1.64) * (-0.54)
R&D: -468.76 0.20 -0.000006
* (-2.91) *** (3.96) *** (-1.86) *
Johnson: Pat.: 46.71 -0.001 -0.0000001
 ((2.09) ** (-0.10) (-0.31)
R&D: -175.29 0.10 -0.0000004
* (-2.19) ** (5.33) *** (-0.41)
Lilly: Pat.: 203.43 -0.045 0.000004
* (4.71) *** (-2.27) ** (1.90) **
R&D: -429.24 0.27 -0.00001
 (-3.12) *** (4.18) *** (-1.37)
Merck: Pat.: 211.35 -0.03 0.000004
 (6.61) *** (-2.38) *** (4.01) ***
R&D: -117.69 0.15 -0.000003
 (-3.36) *** (11.90) *** (-2.72) ***
Pfizer: Pat.: 57.82 0.006 -0.0000001
 -1.36 -0.35 (-0.06)
R&D: 150.49 -0.06 -0.00002
* (2.18) *** (-2.01) ** (-8.08) ***

Dependent R^2 DW

Abbott: Pat.:
* 0.89 1.42
R&D:
 0.99 1.94
Am.: Pat:
* 0.11 1.07
R&D:
 0.97 0.70
Bristol: Pat.:
 0.84 1.88
R&D:
* 0.96 1.36
Johnson: Pat.:
 0.38 2.44
R&D:
* 0.98 0.67
Lilly: Pat.:
* 0.49 2.32
R&D:
 0.98 2.57
Merck: Pat.:
 0.81 1.80
R&D:
 0.99 0.97
Pfizer: Pat.:
 0.34 2.13
R&D:
* 0.99 0.94

TABLE 2.
Descriptive Statistics of a Firm's Dominant Position in the Industry
The median is over the sample period 1980-2000. Top 20 list from
Schweitzer (1997, 24). Market Share is from: Chemical Market Reporter,
10/04/99, pg 1. Sales Force: Med Ad News: Engel Publishing Partners,
West Trenton NJ [Various years, September issue].

 U.S. Top 20
 R&D Sales Patent list of
Company Median Force Median Products
Names ($M) 1999 (#) 1995

Abbott 534.5 2759 88 7
American 357.3 3600 51
Bristol 835.0 3912 56 9
Johnson 776.5 5000 38
Lilly 654.0 2300 95 6
Merck 802.5 5000 219 17
Pfizer 585.5 7600 90 6

 Firms' vs. Ind. R&D
 Market Share
 Less Equal Greater
Company Than to Than
Names Ind. Ind. Ind. 1996 1998

Abbott 10 5 4
American 13 3 3
Bristol 12 4 3 4.8% 5.3%
Johnson 11 4 4 4.0%
Lilly 11 2 6 4.3%
Merck 12 6 1 7% 7.6%
Pfizer 6 7 6 4.6% 5.9%

TABLE 3.
SUR Regressions of Firms R&D Outlays on Rivals

All variables are in logarithmic form. Numbers in brackets are
t-values. Three asterisks represent significance at the 99% level;
two, significance at the 95% level, and one, significance at the
90% level.

Dep./Ind. Constant Rivals R&D (-1)

Abbott/Merck -1.74 (-7.73) *** 1.26 (34.19) ***
American/Merck -2.85 (-6.13) *** 1.38 (20.47) ***
Bristol/Merck -1.74 (-4.15) *** 1.37 (16.32) ***
Lilly/Merck 0.37 (0.79) 1.03 (15.33) ***
Johnson/Merck -0.5 (-0.25) 0.96 (32.61) ***
Pfizer/Merck -1.49 (-4.71) *** 1.36 (34.87) ***
Abbott/Pfizer 0.69 (2.30) *** 0.84 (18.22) ***
American/Pfizer -0.67 (-3.05) *** 1.09 (34.48) ***
Bristol/Pfizer 0.16 (0.34) 0.54 (7.38) ***
Lilly/Pfizer 1.45 (4.57) *** 0.75 (17.65) ***
Johnson/Pfizer 1.78 (9.63) *** 0.72 (25.29) ***
Merck/Pfizer 2.01 (16.62) *** 0.71 (40.90) ***
American/Abbott -0.68 (-1.71) * 1.06 (17.74) ***
Bristol/Abbott 0.50 (1.77) * 0.92 (17.00) ***
Lilly/Abbott 1.08 (4.97) *** 0.76 (15.73) ***
Johnsn/Abbott 1.89 (9.43) *** 0.73 (23.97) ***
Merck/Abbott 2.26 (16.09) *** 0.75 (37.13) **
Pfizer/Abbott 0.16 (0.52) 1.00 (23.72) **
Abbott/American 1.45 (4.28) *** 0.82 (14.81) ***
Bristol/American 2.81 (5.14) *** 1.09 (11.77) ***
Lilly/American 2.19 (6.48) *** 0.69 (15.68) ***
Johnson/American 2.20 (7.67) ** 0.62 (13.25) **
Merck/American 2.86 (12.51) *** 0.67 (19.66) ***
Pfizer/American 2.23 (8.55) *** 0.96 (26.32) **
Abbott/Bristol 0.94 (3.65) *** 0.94 (23.73) ***
American/Bristol -0.50 (-0.90) 0.92 (11.01) ***
Lilly/Bristol 0.87 (2.47) *** 0.64 (14.01) ***
Johnson/Bristol 2.26 (7.08) *** 0.69 (14.55) ***
Merck/Bristol 2.41 (11.90) *** 0.72 (23.89) ***
Pfizer/Bristol 0.37 (0.92) 0.92 (16.21) ***
Abbott/Johnson -1.56 (-3.79) *** 1.23 (19.28) ***
American/Johnson -2.62 (-5.41) *** 1.34 (19.29) ***
Bristol/Johnson -1.44 (-2.69) *** 1.27 (13.27) ***
Lilly/Johnson 0.26 (0.60) 0.98 (16.12) ***
Merck/Johnson 0.43 (1.79) * 0.94 (26.22) ***
Pfizer/Johnson -1.79 (-5.71) *** 1.30 (30.02) ***
Abbott/Lilly -0.475 (-1.03) 1.08 (15.79) ***
American/Lilly -2.76 (-5.26) *** 1.33 (16.93) ***
Bristol/Lilly 0.08 (0.13) 1.06 (10.34) ***
Johnson/Lilly 0.86 (2.16) ** 0.81 (13.04) ***
Merck/Lilly 1.24 (4.01) *** 0.87 (18.39) ***
Pfizer/Lilly -0.87 (-1.92) ** 1.24 (18.74) ***

Dep./Ind. Own Patent (-1) [R.sup.2] D.W.

Abbott/Merck -0.07 (-3.44) *** 0.98 0.60
American/Merck -0.03 (0.87) 0.95 0.40
Bristol/Merck -0.21 (-2.17) ** 0.94 0.46
Lilly/Merck -0.14 (-2.00) ** 0.92 1.41
Johnson/Merck 0.14 (4.28) *** 0.98 2.83
Pfizer/Merck -0.22 (-4.18) *** 0.98 0.59
Abbott/Pfizer 0.05 (2.71) *** 0.94 0.22
American/Pfizer -0.04 (-2.04) ** 0.97 0.79
Bristol/Pfizer 0.73 (9.12) *** 0.86 0.27
Lilly/Pfizer 0.08 (1.39) 0.95 2.23
Johnson/Pfizer 0.11 (4.50) *** 0.97 2.00
Merck/Pfizer 0.03 (2.05) ** 0.98 0.24
American/Abbott 0.09 (2.54) *** 0.93 0.35
Bristol/Abbott 0.09 (1.03) 0.96 0.53
Lilly/Abbott 0.01 (0.14) 0.92 1.54
Johnsn/Abbott 0.11 (3.77) *** 0.97 1.69
Merck/Abbott -0.03 (-1.27) 0.98 0.74
Pfizer/Abbott 0.04 (0.97) 0.97 0.39
Abbott/American -0.03 (-1.43) 0.91 0.24
Bristol/American -0.72 (-8.17) *** 0.77 0.39
Lilly/American 0.05 (0.61) 0.94 1.79
Johnson/American 0.22 (9.50) *** 0.91 0.98
Merck/American -0.03 (-1.31) 0.94 0.35
Pfizer/American -0.04 (-9.32) *** 0.97 0.76
Abbott/Bristol -0.16 (-8.02) *** 0.95 0.50
American/Bristol 0.20 (9.12) *** 0.83 0.37
Lilly/Bristol 0.34 (8.61) *** 0.92 1.76
Johnson/Bristol 0.03 (1.28) 0.91 0.80
Merck/Bristol -0.05 (-2.58) *** 0.96 0.36
Pfizer/Bristol 0.06 (1.97) ** 0.91 0.24
Abbott/Johnson -0.08 (-3.17) *** 0.94 1.18
American/Johnson -0.05 (-1.70) * 0.95 1.35
Bristol/Johnson -0.12 (-1.22) 0.90 0.67
Lilly/Johnson -0.05 (-0.65) 0.93 1.66
Merck/Johnson -0.01 (-0.19) 0.97 1.88
Pfizer/Johnson -0.08 (-2.22) 0.98 1.89
Abbott/Lilly -0.06 (-2.26) *** 0.92 0.94
American/Lilly 0.08 (2.11) ** 0.93 2.20
Bristol/Lilly -0.11 (-1.24) 0.83 0.57
Johnson/Lilly 0.18 (6.26) *** 0.92 1.82
Merck/Lilly -0.03 (-1.66) * 0.94 1.42
Pfizer/Lilly -0.14 (-3.79) *** 0.95 2.00

TABLE 4
SUR Regressions of Firms Advertising Outlays on Rivals

All variables are in logarithmic form. Numbers in brackets are
t-values. Three asterisks represent significance at the 99%
level; two, significance at the 95% level, and one, significance
at the 90% level.

Dep./Ind. Constant Rival AD (T-1)

Abbott/Merck 0.92 (2.93) *** 0.55 (3.29) ***
American/Merck 3.76 (11.98) *** 0.23 (3.72) ***
Bristol/Merck 3.55 (19.25) *** 0.37 (5.75) ***
Lilly/Merck 7.54 (14.01) *** 0.03 (0.22)
Johnson/Merck 1.47 (4.84) *** 0.37 (4.15) ***
Pfizer/Merck 3.24 (9.10) *** 0.78 (9.11) ***
Abbott/Pfizer 1.50 (2.29) ** 0.30 (1.47)
American/Pfizer 2.64 (10.09) *** -0.13 (-0.77)
Bristol/Pfizer 3.39 (21.03) *** 0.39 (11.27) ***
Lilly/Pfizer 6.15 (9.43) *** -0.40 (-1.24)
Johnson/Pfizer 1.48 (6.45) *** 0.32 (5.66) ***
Merck/Pfizer 3.23 (3.00) *** -0.22 (-0.66)
American/Abbott 4.13 (9.48) *** 0.40 (3.65) ***
Bristol/Abbott 3.77 (16.75) *** 0.27 (3.16) ***
Lilly/Abbott 6.96 (13.06) *** 0.18 (1.16)
Johnsn/Abbott 1.61 (5.21) *** 0.41 (3.97) ***
Merck/Abbott 0.60 (1.97) ** 1.06 (7.83) ***
Pfizer/Abbott 3.01 (7.81) *** 0.50 (4.71) ***
Abbott/American -1.58 (-1.25) 0.78 (2.81) ***
Bristol/American 0.98 (1.25) 0.74 (4.41) ***
Lilly/American 6.88 (5.05) *** -0.002 (-0.01)
Johnson/American -0.93 (-1.28) 0.60 (3.48) ***
Merck/American -0.52 (-1.28) 0.53 (1.25)
Pfizer/American -0.76 (-0.95) 0.95 (4.80) ***
Abbott/Bristol -0.10 (-0.09) 0.15 (0.47)
American/Bristol 2.97 (11.57) *** -0.17 (-0.76)
Lilly/Bristol 6.74 (7.50) *** -0.01 (-0.02)
Johnson/Bristol -0.38 (-1.02) 0.66 (6.30) ***
Merck/Bristol -2.63 (-2.05) ** 0.76 (2.15) **
Pfizer/Bristol -4.95 (-4.44) *** 2.10 (6.53) ***
Abbott/Johnson -0.53 (-0.62) 0.87 (3.24) ***
Amer./Johnson 3.38 (14.03) *** 0.77 (7.06) ***
Bristol/Johnson 2.09 (7.89) *** 0.70 (7.10) ***
Lilly/Johnson 6.58 (8.67) *** 0.69 (2.59) ***
Merck/Johnson -0.98 (-0.93) 0.98 (2.78) ***
Pfizer/Johnson -2.20 (-3.58) *** 1.45 (6.31) ***
Abbott/Lilly 6.17 (4.54) *** -0.30 (-1.09)
American/Lilly 5.89 (9.95) *** -0.18 (-1.98) **
Bristol/Lilly 2.49 (3.93) *** 0.39 (3.40) ***
Johnson/Lilly 3.12 (3.19) *** -0.18 (-1.28)
Merck/Lilly 12.94 (5.51) *** -1.11 (-2.40) ***
Pfizer/Lilly 7.57 (6.29) *** -0.21 (-0.87)

Dep./Ind. Own CF (T-1) [R.sup.2] D.W.

Abbott/Merck 0.15 (1.05) 0.98 0.59
American/Merck 0.17 (2.08) ** 0.92 0.06
Bristol/Merck 0.21 (5.41) *** 0.95 1.19
Lilly/Merck -0.62 (-4.44) *** 0.69 1.79
Johnson/Merck 0.40 (4.98) *** 0.92 1.55
Pfizer/Merck -0.28 (-3.00) *** 0.87 1.35
Abbott/Pfizer 0.24 (2.42) *** 0.77 0.39
American/Pfizer 0.61 (4.23) *** 0.92 2.00
Bristol/Pfizer 0.21 (6.52) *** 0.96 0.57
Lilly/Pfizer -0.07 (00.29) 0.58 1.22
Johnson/Pfizer 0.42 (7.00) *** 0.96 1.34
Merck/Pfizer 0.42 (2.74) *** 0.54 0.21
American/Abbott 0.03 (0.24) 0.88 1.48
Bristol/Abbott 0.27 (6.05) *** 0.93 1.14
Lilly/Abbott -0.63 (-4.57) *** 0.68 1.67
Johnsn/Abbott 0.38 (4.96) **** 0.92 1.84
Merck/Abbott -0.06 (-0.60) 0.94 1.47
Pfizer/Abbott -0.01 (-0.08) 0.84 1.24
Abbott/American 0.23 (2.70) *** 0.83 0.63
Bristol/American 0.19 (4.14) **** 0.91 0.56
Lilly/American -0.49 (-3.65) *** 0.69 1.74
Johnson/American 0.49 (7.69) *** 0.93 1.89
Merck/American 0.34 (2.66) *** 0.76 0.26
Pfizer/American 0.04 (0.44) 0.84 0.59
Abbott/Bristol 0.57 (3.21) *** 0.89 0.92
American/Bristol 0.64 (2.99) *** 0.93 2.00
Lilly/Bristol -0.46 (-1.74) * 0.69 1.69
Johnson/Bristol 0.30 (4.45) *** 0.96 1.15
Merck/Bristol 0.37 (1.95) ** 0.84 0.58
Pfizer/Bristol -0.57 (-2.58) *** 0.77 0.25
Abbott/Johnson -0.01 (-0.05) 0.77 0.55
Amer./Johnson -0.27 (-2.75) *** 0.89 1.69
Bristol/Johnson 0.08 (1.17) 0.95 0.87
Lilly/Johnson -1.07 (-5.48) *** 0.60 1.88
Merck/Johnson 0.02 (0.11) 0.79 0.44
Pfizer/Johnson -0.17 (-0.88) 0.89 0.48
Abbott/Lilly -0.06 (-0.72) 0.03 0.19
American/Lilly 0.13 (2.57) *** 0.65 0.64
Bristol/Lilly 0.44 (10.71) *** 0.83 0.33
Johnson/Lilly 0.52 (6.83) *** 0.81 1.58
Merck/Lilly -0.55 (-3.94) *** -0.62 0.19
Pfizer/Lilly -0.23 (-2.57) ** -0.43 0.22

TABLE 5
Regression Estimates of Production Functions for Firms
Sample (1980-1999). 3SLQ Estimates.

Dependent
Variable:
Sales/Labor Constant Capital/Labor R&D/Labor [R.sup.2]

 2.67 0.23 0.51
Abbott (12.78) *** (3.74) *** (15.45) 0.98
 5.12 -0.43 0.78
American (19.50) *** (-4.50) *** (10.41) *** 0.97
 -0.15 1.23 -0.12
Bristol (-0.17) (4.63) *** (-0.76) 0.94
 1.24 0.84 0.02
Johnson (4.93) *** (7.69) *** (0.24) 0.98
 2.14 0.39 0.33
Lilly (3.49) *** (1.87) * (2.32) *** 0.94
 0.81 0.61 0.47
Merck (3.00) *** (6.12) *** (3.64) *** 0.95
 3.28 0.09 0.49
Pfizer (18.86) *** (1.76) * (21.65) *** 0.99