Effect Of Advertising On Pharmaceutical Innovation

Rev Ind Organ (2007) 31:221–236 DOI 10.1007/s11151-007-9146-8

The Effect of Advertising on Pharmaceutical Innovation Winghan Jacqueline Kwong · Edward C. Norton

Published online: 11 October 2007 © Springer Science+Business Media, LLC. 2007

Abstract Although there is much controversy in the economic literature about how advertising affects market competition, little is known about the effect of advertising on product innovation. We examined the relationship between advertising expenditures and the research and development activities of pharmaceutical firms using empirical data from eight therapy areas. This study finds that detailing advertising may have a significant positive effect on the number of new products entering into clinical development. Markets of chronic disease with high levels of detailing advertising were more attractive to pharmaceutical firms. However, the effect of advertising on new product novelty remains inconclusive. Keywords Advertising · Entry · Innovation · Pharmaceuticals · Research and development

W. J. Kwong Department of Clinical and Administrative Pharmacy, College of Pharmacy, University of Georgia, Athens, GA, USA W. J. Kwong (B) Johnson and Johnson Pharmaceutical Services, L.L.C., Worldwide Health Economics & Pricing, 700 Route 202 South, Raritan, NJ 08869, USA e-mail: [email protected] E. C. Norton Department of Health Policy and Administration, School of Public Health, University of North Carolina at Chapel Hill, 1101C McGavran-Greenberg Hall, CB #7411, Chapel Hill, NC 27599-7411, USA e-mail: [email protected] E. C. Norton Department of Economics, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA

123

222

W. J. Kwong, E. C. Norton

1 Introduction The proliferation of direct-to-consumer pharmaceutical advertisements in recent years has fueled many studies that examine the effects of advertising on prescribing decisions and prescription drug demand. An important question that has not been studied adequately is how pharmaceutical advertising may affect product innovation. The relationship between advertising and product innovation has several implications for health care quality and cost containment issues. It is well established that advertising increases product demand and drug expenditures. If advertising encourages development of medicines that have better efficacy, fewer side effects, greater patient compliance, and reduced utilization of other health care resources, then the benefits of advertising would offset the increased drug expenditures. If advertising merely encourages the entry of generics and close substitutes with similar efficacy and safety profile to existing products, then it would have minimal effects on advancing medical science despite the consumer surplus that may arise from the competition of these products. In this study, pharmaceutical advertising data from the 1990s were analyzed to examine the relationship between advertising and products that pharmaceutical firms have invested in developing. By examining the effects of advertising expenditures on the types of products entering into clinical investigations, we can conjecture on the effect of advertising on product innovation.

2 Policy History The United States Congress has repeatedly passed laws to spur innovation in the pharmaceutical industry. For example, the 1983 Orphan Drug Act encourages development of pharmaceutical treatments for rare diseases and medical conditions by providing pharmaceutical firms with market exclusivity for seven years. The 1984 Drug Price Competition and Patent Term Restoration Act (frequently described as Hatch-Waxman Act) restored some of the patent life that is lost during the lengthy drug development and regulatory review process. A longer patent life increases the returns on investment and is expected to encourage innovation (Grabowski and Vernon 1986). The 1992 Prescription Drug User Fee Act allows pharmaceutical firms to submit a fee with its new drug application so that the Food and Drug Administration (FDA) can hire more resources to expedite the regulatory review process. By shortening the review time and restoring patent life, the number of new drugs entering clinical testing, which had been declining in the late 1980s and early 1990s, increased by 10% from 1992 to 1994 (DiMasi 2001a). In 1997, the FDA clarified its interpretation of direct-to-consumer advertising laws, effectively making it easier for pharmaceutical firms to advertise directly to consumers through the mainstream media. Although this ruling was not directly aimed at innovation, anything that affects demand, costs, and strategic behavior necessarily affects product entry decisions.

123

Advertising and Innovation

223

3 Literature on the Effect of Advertising on Product Entry and Quality It is difficult to determine how advertising will affect product entry in the pharmaceutical market because the role of pharmaceutical advertising is not known a priori. The Food and Drug Administration (FDA) mandates that product-specific advertisements disclose information on adverse effects to satisfy the “fair-balance” requirement. However, critics contend that pharmaceutical advertisements often emphasize the benefits of a product while minimizing the information on side effects so that the product is perceived to be safe and effective. The agency relationship between physicians and consumers also complicates prescribing decisions, especially when the physician and patient perceive the product’s value differently (Steinke et al. 1999). Although informative advertising may drive patients to physicians to request prescriptions, if these physicians are unwilling to prescribe the requested product, then advertising would have no effect on use. A competing product’s advertising may lead physicians to prefer it to other products if the competing product is perceived to have better efficacy and fewer side effects. Therefore, when studying the effect of advertising on market competition in the pharmaceutical market, advertising activities that are targeted to physicians and consumers should be considered simultaneously. Only a few published papers have directly examined the effect of advertising on product entry. All of these papers utilized sales and advertising audit data published by IMS America, Ltd. but with different study samples, timeframes, and methods. Telser et al. (1975) analyzed the sales and promotional data of products in 17 therapeutic categories from 1963 to 1972, and found that in 13 of the 17 therapeutic categories, changes in promotional intensity were positively correlated with the changes in entry. Using data from 51 new products introduced from 1968 to 1977, Leffler (1981) found that pharmaceutical advertising encouraged the entry of superior new products and may retard the entry of low priced cost substitutes. Two other papers focused on how advertising activities of brand products affected generic entry but did not evaluate the effect on entry of other brand products in the same therapeutic market. Hurwitz and Caves (1988) compared the market activities of oral products for 29 therapeutic markets between 1978 and 1983 and found that the number of generic entrants decreased with the pioneer brand’s promotional outlays. A similar research question was pursued again by Scott Morton (2000) using data from products that lost patent protection from 1986 to 1991. The author used instrumental variables to correct the possible endogeneity bias between advertising and entry, and found that advertising had no significant effect on the number of generic entrants. Based on these findings, it appears that advertising either encouraged or had no significant effect on new product entry. In addition to the papers that examined direct effect of advertising on product entry, several other studies suggested that favorable market conditions encourage product entry. Using IMS data, Ellison and Ellison (2000) found sales revenue had a significant and positive effect on generic entry. Finkelstein (2003) studied the vaccine market and found that policies that support reimbursement and use of immunizations were associated with an increase in the number of vaccine clinical trials. Acemoglu and Linn (2003) studied the new drugs approved by the FDA and found that the number of new approvals increased with potential market size. While these papers did not

123

224

W. J. Kwong, E. C. Norton

directly test the effect of advertising on innovation, they suggested that if advertising increases market size, then it encourages entry and innovation. With the exception of Leffler’s work, previous studies did not consider the effect of advertising on the quality of new entries. With the exception of Finkelstein’s (2003) study previous papers used the number of marketed products as the dependent variable, which does not adequately represent pharmaceutical firms’ research and development activities. Drug development is a lengthy process and a risky business. On average, it took 3.8 years from synthesis to human testing, 8.6 years from human testing to regulatory submission, and another 1.3 years for regulatory review before a product can be marketed in the U.S. (DiMasi 2001a,b). During this development process, many products fail to reach market. Before committing to each stage of drug development, pharmaceutical firms estimate the expected cost of subsequent research, the likelihood of achieving market approval, expected marketing expenses, and expected sales revenue after the product is launched (Wiggins 1981; Frank 2003). This information allows profit-maximizing pharmaceutical firms to estimate the net present value of making additional investments and to decide if development should continue (Frank 2003). In fact, the success rates of products progressing from development to market approval range from only 19% to 30% (DiMasi 2001b). Termination happens for reasons of clinical efficacy, safety, economics, or regulation. Because few products eventually gain market approval, we analyze data on investigational products instead of marketed products. The former better reflect firms’ product entry decisions.

4 Model and Hypotheses Advertisements could heighten physician and consumer awareness of product quality (Findlay 2001) and provide additional incentives for firms to invest in product innovation to enhance differentiation (Calfee 2002). Advertising may provide higher returns to pharmaceutical firms, freeing resources to develop more innovative products (Scherer 2001; Calfee 2002). However, if advertising is a necessary condition for market entry and its cost is too high relative to total return, then advertising may discourage entry and innovation. This latter point of view was supported by a study showing that product innovation and advertising are strategic substitutes in the German service sector (Kaiser 2005). Pharmaceutical firms often choose between various clinical development programs based on return-on-investment analyses. During each stage of development, firms update the assessment of the net present value of making an additional investment in the development of a specific product (Frank 2003). We propose that current advertising expenditures in a therapeutic market provide information to prospective entrants. Higher current advertising expenditures indicate that greater advertising expenditures will be required in the future. Because more difficult markets require higher marketing budgets with lower returns on investments, we hypothesize that when advertising expenditures are high, pharmaceutical firms prefer to develop close substitute products or new formulations to novel or creative products. Under this conceptual framework, the following hypotheses were developed for empirical testing:

123

Advertising and Innovation

225

Hypothesis 1: New product entries are less likely in markets with higher advertising expenditures. Hypothesis 2: The deterrence effect of advertising is greater for creative entry than other types of entry.

4.1 Empirical Model Specification We studied the effect of advertising on investigational product entry for eight therapy markets (asthma, depression, dyslipidemia, gastric/duodenal ulcer, migraine, obesity, Parkinson’s disease and epilepsy) from 1995 to 2001. The analysis was performed at the market level where products targeted for the same therapeutic indications compete with each other. Because the number of new products entering into clinical testing (investigational products) was small in each quarter, the number of entries was modeled using negative binomial regression with the following specification: Entryit = f(Market advertising expenditureit−k , Total Market Sales Revenueit−k , Number of existing marketed products for the same therapeutic indicationit−k , Remaining patent life of market leaderit−k , Market share of generic productsit−k , Chronic diseasei , Specialist prescriberi , Yeart−k ) + εit The dependent variable was the total number of investigational products entering into clinical development in market i in period t. The time-frame of t was 3 months. The variable of primary interest was market advertising expenditures in period t − k, where k = 1, 2, 3, or 4. In addition, the number of creative product entries, close substitute product entries, and new formulation entries (as defined in the next section) were estimated separately using negative binomial estimation. There are two reasons to model lagged advertising expenditures. First, advertising expenditures were lagged because it was expected that there was a time lag between the collection of advertising data by market research firms and the delivery of such data to pharmaceutical executives for research and development decisions. Therefore, we lagged all explanatory variables. Because there was no information from the literature on how long a time lag is appropriate, we estimated models with several different lags. Second, it is possible that market advertising expenditures and new product entry are endogenous. Firms that produce existing products may want to take preemptive action to increase or decrease advertising expenditures if they expect a new product entry would take place in later periods. While the launch date of a new product can be predicted based on the typical time required for regulatory review, the exact date of regulatory approval is often uncertain and it depends on whether firms are required to provide additional data after initial regulatory review. One way to correct for potential endogeneity bias is to estimate the model using instrumental variables. Unfortunately, good instruments that are correlated with advertising but not with new product entry could not be found in the data. In the absence of good instruments, we used lagged values of advertising expenditures farther back in time (e.g., t − 3 and t − 4), hoping that the lagged variables would not be correlated with the dependent variable at time t.

123

226

W. J. Kwong, E. C. Norton

5 Variable Definitions In this study, a market is made up of products with the same therapeutic indications. The primary variable of interest is the number of new investigational product entries. A new product is a new chemical entity or new formulation that entered into clinical development at time t, but that did not exist at time t − 1. This definition excludes any generic products. Investigational products were classified into three categories: creative products, close substitute products and new formulations. Creative products are products with a new pharmacological mechanism of action that differs from that of other existing marketed or investigational products for the same therapeutic indication. For example, Imitrex
was considered to be a creative entry for migraine treatment when it was launched in 1991 because at that time no other existing product approved for migraine treatment worked by selectively targeting the serotonin receptor. Close substitute products are products with a pharmacological mechanism of action that is similar to that of other existing marketed or investigational products for the same therapeutic indication. For example, Zantac
is a close substitute of Tagamet
. Zantac was approved and marketed after Tagamet in the United States. Although the products are chemically different, both drugs inhibit gastric secretions by blocking the histamine-2 receptors in the gastric parietal cells. New formulations are new dosage forms of existing products. An example of a new formulation is Prozac Weekly™, which is a delayed-release formulation of Prozac
that allows patients to take the medication once a week instead of once a day. The categorization of creative, close substitute, and new formulation was made independent of the novelty ratings provided in the data source. Timing of entry is defined as the date when a new product first enters into clinical development (i.e., human testing) for a therapeutic indication. Although it is preferable to define investigational product entry using the start date of clinical development, the information was not always revealed to the public for proprietary reasons. In cases where a new product was reported to have started clinical development but the start date of clinical development was unknown, the start date of pre-clinical testing was used instead. If a product was not known to undergo development until results of clinical trials were reported, then the date when results are reported was used as the start date of clinical development. These substitutions for the exact clinical development start date were used for less than 10% of observations. Using the beginning of clinical testing to define the time of product entry is reasonable because clinical research and development activities constitute a large proportion of research and development expenses to bring a pharmaceutical product to the market. Before clinical testing can begin, manufacturers are required to file an investigational drug application (IND) with the FDA that describes the clinical research plan and all known information about the product. Of the total cost of $802 million to bring a product to market, $467 million is for clinical development (DiMasi et al. 2003). Pharmaceutical firms incur large expenses for research and development risks. Therefore, defining product entry as the time a new product enters into human testing instead of FDA approval is a more realistic approach to study the effect of advertising on pharmaceutical research and development investments.

123

Advertising and Innovation

227

Market advertising expenditures (including promotion expenditures) were included as two independent variables in the model: 1) expenditures for physician detailing activities (i.e., costs associated with pharmaceutical sales representatives’ visits to physician office to discuss product information), and 2) expenditures for other advertising activities (i.e., the sum of direct-to-consumer advertising activities and physician meetings and events expenditures). Physician detailing by pharmaceutical representatives is a traditional form of advertising that has been used by pharmaceutical firms for decades. Even with the increases in direct-to-consumer advertising activities over the past decade, professional detailing remains the dominant form of advertising activity and it accounts for about 65% of total advertising expenditures in our data. Because of the importance of professional detailing to the pharmaceutical industry, expenditures on detailing are modeled separately from other forms of advertising to allow for comparison. To control for potential multicollinearity problems between advertising expenditures and sales revenue, we modeled the ratios of advertising expenditures to sales revenue. If advertising has an entry deterrence effect, the coefficient of these ratios would be negative. We also controlled for demand factors that may affect a firm’s investment decision for a therapeutic market. Firms may be more interested in developing products for chronic diseases that require long-term treatment (i.e., asthma, depression, dyslipidemia, Parkinson’s disease and seizure disorder). Diseases that are predominantly managed by specialists may be less attractive to firms because they often have lower disease prevalence than illnesses that are managed by primary care providers. In addition, we included three variables to control for other market characteristics that may affect competitiveness. Markets are expected to be more attractive for product entry when they have fewer existing products, when generic products have a smaller market share, and when the market leader has a shorter remaining patent life.

6 Data Sources The data used in this study are from multiple sources. The data source for investigational product entry is Pharmaprojects, a database that contains details on about 34,000 developmental products investigated since 1980. Pharmaprojects contains detailed information on products that are under active development, their development history, and progress to date. From this information we can determine the timing of investigational product entry. The primary explanatory variable of interest, pharmaceutical advertising expenditure, was based on the Scott-Levin Market Research Audit data that monitor product promotion and market performance for U.S. and international pharmaceutical firms. Specifically, we aggregated market-level advertising expenditure data from three different market research audits: 1) the office-based and hospital-based personal selling audit from physicians, 2) physician meetings and events audit, and 3) the direct-toconsumer audit. Audit data from the physicians were obtained through a monthly survey of 11,330 physicians that practiced in hospital or office-settings. These health care professionals

123

228

W. J. Kwong, E. C. Norton

reported sales representative detailing activities for one week per month in the survey. The survey data were then projected to provide a national estimate. A fixed cost was assigned to each detailing visit, and the cost estimate was adjusted annually. Data from the direct-to-consumer audit covers commercial activities across 11 media types, including television, magazines, radio, newspapers and outdoor advertising. Both non-product specific and product-specific advertising are captured. In this study, we did not utilize Scott-Levin data on journal advertising and detailing activities to nurse practitioners and physician assistants because these data were not available for the entire study period. Market-level sales revenue data were obtained by aggregating individual product sales from the Scott-Levin Source Prescription Audit, which includes products dispensed at more than 36,900 retail pharmacies. Audited pharmacies include independent and chain pharmacies, mass merchandisers, deep discounters, and food stores. Seventy percent of these panel pharmacies are chain stores. These stores represent about 65% of the pharmacy stores in the U.S. The distribution of prescriptions was calculated for each of the 1,300 stratified regions, and then aggregated to the national level. Prescription data were reported at product form and strength level. Sales revenue is based on the full price that the pharmacy charges to the consumer, regardless of the co-pay. Information about existing products was extracted from the electronic Physician Desk Reference,US Pharmacopeia Dispensing Information (USPDI), Drug Facts and Comparisons (F&C), and the electronic Orange Book (http://www.fda.gov/cder/ob/). Product patent information was obtained from the publication Drugs Under Patent 2002 and the Orange Book. We analyzed data from 1995 to 2001 for products in eight therapy areas (asthma, depression, dyslipidemia, gastric/duodenal ulcer, migraine, obesity, epilepsy and Parkinson’s disease) on a quarterly basis. The therapy areas were limited to medical conditions for which Scott-Levin market research audits data were available to the authors from a pharmaceutical firm. These therapy areas represent a wide range of disease areas, ranging from those that are commonly known to those that are often under-diagnosed and under-treated, diseases that affect mostly the young and the elderly, and diseases that are typically managed by primary care physicians to those that are managed mostly by specialists. According to the Novartis Pharmacy Benefits Report, the U.S. retail pharmacy market was $204 billion in 2002. Annual total sales revenue for the eight therapy markets included in this analysis averaged $30.2 billion over the study period. Therefore, the therapy markets analyzed in this study represents about 15% of the national prescription drug market.

6.1 Summary Statistics During the study period of January 1995 to December 2001, sales revenue data were available for a total of 737 unique products among the eight disease areas. Of these products, 204 (28.1%) were generic products. Asthma had the largest number of products on the market during this period (n = 317), followed by migraine (n = 100), obesity (n = 77), Parkinson’s disease (n = 67), seizure disorders (n = 63),

123

Advertising and Innovation

229

depression (n = 59), lipid disorder (n = 29), and gastric and duodenal ulcer (n = 25). Among the 737 products with sales revenue data, 326 (44.2%) products were advertised. The remaining 411 products did not have any advertising activities from 1995 to 2001 according to the Scott-Levin Market Research Audits. Asthma has the largest number of products with recorded advertising activities (n = 116), followed by depression (n = 39), Parkinson’s disease (n = 35), seizure disorder (n = 36), migraine (n = 32), obesity (n = 25), lipid disorder (n = 24), and gastric/duodenal ulcer (n = 22). Three products indicated for the treatment of seizure disorder (Depakote, Depakote ER and Divalproex sodium) were also indicated for the treatment of migraine. There were a total of 328 investigational product entries in the eight disease areas from 1995 to 2001, of which 130 are creative products, 130 are close substitute products, and 68 are new formulations. Among the eight disease areas, asthma had the largest number of investigational product entries per quarter on average (mean = 3.39, SD = 2.87), followed by depression (mean = 1.93, SD = 1.82), Parkinson’s disease (mean = 1.64, SD = 1.44), lipid disorder (mean = 1.25, SD = 1.04), migraine (mean = 1.00, SD = 1.28), seizure disorder (mean = 0.96, SD = 1.20), obesity (mean = 0.89, SD = 0.99) and gastric/duodenal ulcer (mean = 0.64, SD = 0.68). Across the eight therapy areas over the study period, detailing advertising expenditures averaged about 4.4% of sales revenue, and other advertising averaged about 3.1% of sales revenue on a quarterly basis (see Table 1). There was a general trend of increasing advertising expenditures over time for all disease markets, with the exception that advertising expenditures for Parkinson’s disease appeared to be stable over time. On average, physician detailing accounts for about 65% of total advertising expenditure. Depression products were most heavily detailed, followed by lipid disorder, asthma, gastric/duodenal ulcer, migraine, and seizure disorder; obesity and Parkinson’s disease had the lowest detailing advertising expenditures per quarter. On the other hand, lipid disorder products had the highest other advertising expenditure, followed by depression and migraine. Seizure disorder and Parkinson’s disease had the lowest other advertising expenditures. Based on quarterly sales revenue, depression was the largest market, followed by gastric/duodenal ulcer, lipid disorder, asthma, seizure disorder, migraine, obesity, and Parkinson’s disease. Similar to advertising expenditure, sales revenue also appeared to increase over time for each disease area, with the exception of Parkinson’s disease and obesity. 7 Results 7.1 Econometric Issues We tested for the presence of heteroskedasticity and autocorrelation in the panel data. A likelihood ratio test comparing the log likelihood of iterated generalized least squares models with and without heteroskedastic errors suggested the presence of heteroskedasticity. Therefore, robust errors are presented in the results. Using the approach suggested by Wooldridge (2002), first-order autocorrelation in the panel data was ruled out.

123

230

W. J. Kwong, E. C. Norton

Table 1 Summary statistics of key variables during the period 1995-2001 on a quarterly basis Variables Number of product entries Investigational products entries across disease areas Creative entries Close substitute entries New formulation entries

Mean

Standard Deviation Minimum Maximum

1.5 0.60 0.60 0.31

1.68 0.90 0.87 0.75

0 0 0 0

13 5 5 5

Detailing advertising ($ mil.) Asthma Depression Gastric and duodenal ulcer Lipid disorder Migraine Obesity Parkinson’s disease Seizure disorder

41.52 59.81 103.81 51.74 62.38 28.53 9.25 4.37 12.33

34.75 14.62 25.59 12.09 18.49 7.85 6.83 1.75 2.52

0.27 33.63 66.24 35.82 27.12 17.11 2.71 1.97 7.91

143.62 89.09 143.62 79.87 84.44 46.59 22.45 7.73 16.23

Other advertising ($ mil.) Asthma Depression Gastric and duodenal ulcer Lipid disorder Migraine Obesity Parkinson’s disease Seizure disorder

26.07 35.21 44.73 26.79 47.64 36.09 14.54 0.38 3.14

27.86 26.50 35.97 20.20 25.21 23.22 20.67 0.48 1.79

0 5.60 10.14 5.66 5.98 3.58 0 0 0.61

115.02 105.66 115.02 90.82 103.14 88.37 83.95 1.76 7.78

Total product sales ($ mil.) Asthma Depression Gastric and duodenal ulcer Lipid disorder Migraine Obesity Parkinson’s disease Seizure disorder

942.44 953.91 1876.56 1788.73 1420.10 538.80 166.57 155.97 638.85

758.08 284.66 707.89 477.83 640.19 163.84 81.12 15.16 278.97

37.61 626.92 911.29 1205.30 577.12 258.17 37.61 127.78 286.27

3234.53 1653.80 3234.53 2628.33 2608.74 787.28 303.87 183.88 1188.50

Number of existing products Remaining patent life for market leader (months)

83.97

89.72

9

329

68.97

44.06

0

140.98

Note: There were a total of 224 observations, with 28 observations in each disease area

Two econometric tests were performed to determine the appropriateness of modeling the equation separately for each type of product entry. The Breusch-Pagan test of independence found that the error terms of the three separate equations are not correlated. The seemingly unrelated estimation version of the Hausman test confirmed coefficients of the single-equation estimation were significantly different from the three separate equation estimations at the 0.01 level. 7.2 Effect of Pharmaceutical Advertising Expenditures on New Product Entry When no differentiation was made between the different types of entries, the coefficients for detailing advertising were uniformly positive, and the coefficients for other

123

Advertising and Innovation

231

advertising activities were uniformly negative across all four lag models (Table 2). The coefficient for detailing advertising was significant in the three- and four-period lag models, but none of the coefficients for other advertising were statistically significant. These results indicate that more detailing advertising activities were associated with more new products entering into clinical development. As expected, markets of chronic treatments were significantly more attractive to new product entry in all four lag models. Markets for treatments that were predominantly prescribed by specialists were significantly less attractive to new product entry in the one-period and two-period lag models only. Over time, there was a significant increase in the number of new product entries. To our surprise, neither the length of remaining patent life nor the market share of generic products was significant in any of the four lag models. The sign of the number of existing products was ambiguous—it was positive and significant in the one- and two-period lag models but was negative and significant in the four-period lag model. These results suggest that although a firm’s decision to begin clinical development in a therapeutic market was not affected by the market leader’s exclusivity status, markets with a large number of existing products might be less attractive to pharmaceutical firms. On the other hand, a fragmented market with many (but no dominant) players may be attractive to a new entrant because it may be easier to establish dominant status. This relationship deserves further examination in future studies. Because the proper lag structure is unknown, we ran several lag models. The pseudo R 2 increased with longer lag periods up to three lags (nine months). The possibility of increased R 2 due to shrinking sample size was ruled out after running the one-period lag model on the subset of observations used in the four-period lag model. Although detailing advertising expenditures was significant in three- and four-period lag models, the three-period lag models had higher z-statistics for advertising expenditures (z statistics = 1.92) as compared to the four-period lag models (z statistics = 1.78). Because the three-period lag model used more data than the four-period lag model, three-period lags were used in determining whether the effect of advertising varied for different types of product entry. 7.3 Effect of Advertising on Different Types of Entry The coefficient for detailing was consistently positive for all types of entry, but the sign of other advertising activities was positive for close substitute and negative for creative and new formulation entries (Table 3). However, neither detailing advertising expenditures nor other advertising expenditures were significant in predicting the number of creative, close substitute, or new formulation product entries. Similar to the results of single equation estimation, the remaining patent life of the market leader did not significantly affect any types of new product entries. The coefficients of other covariates offer some interesting findings. While chronic treatment markets were significantly more attractive to all three types of product entries than were acute treatment markets, markets dominated by specialist prescribers were significantly less desirable for close substitute product entries. It is possible that specialists may have stronger brand loyalty and risk aversion than do primary care physicians so

123

123 2-period lag (N = 208) −0.935(0.716) 0.225(0.185) −0.021(0.021) 0.175∗ (0.101) 1.32(0.91) −0.0012(0.0052) 1.18 × 10−5 (3.91×10−5 ) 0.713∗∗∗ (0.172) −0.511∗ (0.284) 0.136∗∗∗ (0.040) −310.41 71.41∗∗∗ 0.093

1-period lag (N = 216)

−1.50∗∗ (0.68) 0.129(0.173) −0.026(0.025) 0.222∗∗ (0.094) 1.34(0.84) 0.00540(0.00473) −1.57 × 10−5 (3.77 × 10−5 ) 0.691∗∗∗ (0.173) −0.596∗∗ (0.250) 0.118∗∗∗ (0.038) −320.32 73.39∗∗∗ 0.092 −0.462(0.803) 0.317∗ (0.165) −0.021(0.024) 0.149(0.101) 1.23(0.89) −0.00136(0.00534) 2.60×10−5 (4.12×10−5 ) 0.740∗∗∗ (0.174) −0.437(0.281) 0.146∗∗∗ (0.043) −300.45 72.53∗∗∗ 0.097

3-period lag (N = 200)

∗ Statistically significant at 0.10 level; ∗∗ Statistically significant at 0.05 level; ∗∗∗ Statistically significant at 0.01 level

Constant Log detailing to sales ratio Log other advertising to sales ratio Log number of existing products in market Total generics market share Remaining patent life of market leader Square of remaining patent life Chronic disease Specialist prescriber Year Log likelihood Likelihood Ratio χ 2 (9) Pseudo R 2

Variable

Table 2 Negative binomial estimates of total number of new investigational product entries (robust standard errors are in parentheses)

−0.767(0.820) 0.292∗ (0.164) −0.0274(0.0229) −0.202∗ (0.119) 0.99(0.89) 0.00342(0.00596) −1.27×10−5 (4.45×10−5 ) 0.738∗∗∗ (0.179) −0.456(0.287) 0.143∗∗∗ (0.07) −292.07 64.28∗∗∗ 0.093

4-period lag (N = 192)

232 W. J. Kwong, E. C. Norton

Close Substitute −0.810 (1.125) 0.276 (0.291) 0.00225(0.0424) 0.0375 (0.162) 2.35 (1.54) 0.00107 (0.00805) −0.0000173 (0.000060) 0.786 (0.261)∗∗∗ −0.999 (0.545)∗ 0.0972 (0.0657) −191.27 43.34 0.0898 200

Creative −0.871 (1.254) 0.438 (0.317) −0.0375 (0.0299) 0.116 (0.157) 0.158 (1.17) −0.00926 (0.00724) 0.0000911 (0.0000564) 0.639 (0.263)∗∗ 0.268 (0.391) 0.185 (0.062)∗∗∗ −194.95 37.99 0.0722 200

∗ Statistically significant at 0.10 level; ∗∗ Statistically significant at 0.05 level; ∗∗∗ Statistically significant at 0.01 level

Constant Log detailing to sales ratio Log other advertising to sales ratio Log number of existing products in market Total generics market share Remaining patent life of market leader Square of remaining patent life Chronic disease Specialist prescriber Year Log Likelihood Likelihood Ratio Pseudo R2 Number of observations

Variable

−4.11 (1.50)∗∗ 0.146 (0.343) −0.0212(0.0527) 0.305 (0.261) 1.95 (2.21) 0.0142 (0.0136) −0.0000529 (0.0000897) 0.872 (0.458)∗∗ −1.10 (0.764) 0.201 (0.098)∗∗ −129.96 23.45 0.0907 200

New formulation

Table 3 Negative binomial estimates of the number of creative, close substitute and new formulation product entries (robust standard errors are in parentheses)

Advertising and Innovation 233

123

234

W. J. Kwong, E. C. Norton

that they prefer products with which they are familiar, making penetration for close substitute products more difficult.

8 Discussion In contrast to previous research that focused on the effect of advertising on product entry based on marketed products, this study is the first that examined the effect of advertising on products entering into clinical development. Because the success rate of research and development is low, investigational product entry data better reflect a firm’s investment decisions than do marketed product data. This study shows that it is feasible to use an alternative approach to study the entry deterrence effect of advertising. The results of this study suggest that detailing advertising may encourage new product entry by increasing the number of new products entering into clinical testing. When the three types of entry were pooled together, higher detailing advertising expenditure was significantly associated with more new product entries. However, other types of advertising had no significant effect on product entry. These results can be explained by the unique role that detailing advertising plays in the pharmaceutical market. In contrast to direct-to-consumer advertising that relies on patient inquiry of a specific medication to drive sales, professional detailing provides opportunities for sales representatives to defend the advantages of a product over its competitors. Previous studies also showed that professional detailing advertising has a greater effect on brand switching than does direct-to-consumer advertising (Narayanan et al. 2004; Wosinska 2002). As a result, markets with more detailing activities are more attractive to new product entries because they offer more opportunities for product differentiation. However, when the three types of entry were modeled separately, statistical significance disappeared. The lack of significance in the three equation estimations may be due to the small sample size and insufficient variation in the dependent variable. Over the study period, the average number of new product entries per quarter was only 1.5 across disease areas. When the number of entries was further divided by entry types, the number was even smaller (range = 0.31 to 0.60). Insufficient variation in the dependent variable may lead to higher standard errors, making it more difficult to detect statistical significance. The effect of advertising on novelty of product entries warrants further examination in future studies. The findings of this study should be viewed in light of their limitations. First, the data used in this study were from a convenience sample of eight disease areas from the period of 1995 to 2001. Although the disease areas cover a broad range of medical conditions, products that were included in this analysis represented only about 15% of the retail pharmaceutical markets. Therefore, the generalizability of results to other disease areas could be limited. Second, like other studies, our results may be subject to endogeneity and omitted variable bias. It is widely recognized that sales and advertising are endogenous. Scott Morton (2000) suggested entry and advertising may be endogenous as well. However, the search for good instrumental variables was unsuccessful, and the use of instrument variables to address endogeneity is not feasible with a small sample size.

123

Advertising and Innovation

235

Third, a potentially important omitted variable in these analyses is a measure of the scientific opportunities arising from basic research, drug discovery, and pre-clinical development for novel approaches in specific disease areas over time. This information is difficult to capture and quantify, and it was not included in the model estimations. Finally, measurement error in product entry and advertising data could also have biased the results. Pharmaceutical firms may not disclose all their research and development activities to the public in a timely manner. The data source of investigational product entry relied on tracking at scientific conferences and may not be all inclusive. Nevertheless, the consistency of current results with those reported previously in the empirical and theoretical literature increases our confidence in their validity. In summary, this study finds that detailing advertising may have a significant positive effect on product entry. It remains unclear whether advertising would affect product innovation in the pharmaceutical market. Future research investigating the effect of advertising on product novelty, and the social benefits of different types of new product entry is needed to allow for a more comprehensive debate on the controversy of pharmaceutical advertising to inform advertising regulations. Acknowledgements This study was supported by the American Foundation for Pharmaceutical Education Pre-doctoral Fellowship. The authors thank Verispan and GlaxoSmithKline for providing ScottLevin Market Research Audits data for this study. The authors also thank Gary Biglaiser, Marisa Domino, Betsy Sleath, David W. Miller, participants at the 5th World Congress of the International Health Economics Association, Barcelona, Spain, and two journal reviewers for their helpful comments.

References Acemoglu, D., & Linn, J. (2003). Market size in innovation: Theory and evidence from the pharmaceutical industry. Working paper no. 10038, National Bureau of Economic Research, Cambridge, MA. Calfee, J. E. (2002). The role of marketing in pharmaceutical research and development. Pharmacoeconomics, 20(Suppl 3), 77–85. DiMasi, J. A. (2001a). New drug development in the United States from 1963 to 1999. Clinical Pharmacology and Therapeutics, 69, 286–296. DiMasi, J. A. (2001b). Risks in new drug development: Approval success rates for investigational drugs. Clinical Pharmacology and Therapeutics, 69, 297–307. DiMasi, J. A., Hansen, R. W., & Grabowski, H. G. (2003). The price of innovation: New estimates of drug development costs. Journal of Health Economics, 22, 151–185. Drug Facts and Comparisons, 2001 ed. St. Louis, Missouri: Facts and Comparisons, A Wolters Kluwer company. Drugs Under Patent. (2002). Gaithersburg, MD: Freedom of Information Services, Inc. Ellison, G., & Ellison, S. F. (2000). Strategic entry deterrence and the behavior of pharmaceutical incumbents prior to patent expiration. Retrieved April 4, 2007, from Massachusetts Institute of Technology, Department of Economics. website: http://econ-www.mit.edu/faculty/download_pdf.php?id=225 Findlay, S. D. (2001). Direct-to-consumer promotion of prescription drugs. Economic Implications for patients, payers and providers. Pharmacoeconomics, 19(2), 109–119. Finkelstein, A. (2003). Health policy and technological change: Evidence from the vaccine industry. Working paper no. 9460, National Bureau of Economic Research, Cambridge, MA. Frank, R. G. (2003). New estimates of drug development cost. Journal of Health Economics, 22, 325–330. Grabowski, H., & Vernon, J. (1986). Longer patents for lower imitation barriers: The 1984 Drug Act. American Economic Review, 76(2), 195–198. Hurwitz, M. A., & Caves, R. E. (1988). Persuasion or information? Promotion and the shares of brand name and generic pharmaceuticals. Journal of Law & Economics, 31(2), 299–320. Kaiser, U. (2005). A Microeconometric note on product innovation and product innovation advertising. Economics of Innovation and New Technology, 14(7), 573–582.

123

236

W. J. Kwong, E. C. Norton

Leffler, K. B. (1981). Persuasion or information? The economics of prescription drug advertising. Journal of Law and Economics, 24(1), 45–74. Narayanan, S., Desiraju, R., & Chintagunta, P. K. (2004). Return on investment implications for pharmaceutical promotional expenditures: The role of marketing-mix interactions. Journal of Marketing, 68(October), 90–105. Scherer, F. M. (2001). The link between gross profitability and pharmaceutical R&D spending. Health Affairs, 20(5), 216–220. Scott Morton, F. M. (2000). Barriers to entry, brand advertising, and generic entry in the U.S. pharmaceutical industry. International Journal of Industrial Organization, 18(7), 1085–1104. Steinke, D. T., MacDonald, T. M., & Davey, P. G. (1999). Doctor-patient relationship and prescribing patterns. View from primary care. Pharmacoeconomics, 16(6), 599–603. Telser, L. G., Best, W., Egan, J. W., & Higinbotham, H. N. (1975). The theory of supply with application to the ethical pharmaceutical industry. Journal of Law and Economics, 18(2), 449–478. United States Pharmacopeia Dispensing Information Volume I—Drug Information for the Health Care Professional, 20th ed. Englewood, CO: Micromedex, Inc. Wiggins, S. N. (1981). Product quality regulation and new drug introductions: Some new evidence from the 1970s. Review of Economics and Statistics, 63(4), 615–619. Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. Cambridge: MIT Press. Wosinska, M. (2002). Just what the patient ordered? Direct-to-consumer advertising and the demand for pharmaceutical products. HBS Marketing Research Paper Series, No. 02–04, 2002. doi: 10.2139/ssrn.347005

123