Direct-to-Consumer Advertising of Pharmaceuticals (Appendix)

Appendix Supplemental Appendix For Online Publication A A.1 Model Simulation Results We simulate a Logit formulation of the above setting to explore estimation bias. Our formulation has the following utility functions in each simulated market m ui jm = α j + β1 ln(1 + a jm ) + β2 ln(1 + a− jm ) + ξ jm + εi jm ui0m = εi0m , where ui0 denotes the utility of the outside good. Assuming εi jm is i.i.d. type I extreme value, market shares D jm can be computed given parameters and advertising levels using the standard Logit formula. The per-unit cost of advertising is c, and profit per unit sold is ρ. Firm profits in this model are given by π jm = ρD jm − ca jm , where ρ is the margin on an individual unit. We draw values of ξ jm and solve for advertising levels in each market such that both firms’ firstorder conditions are satisfied and create a dataset containing demand and advertising data. We then estimate equation (1), and compare the estimated elasticity with respect to own and rival advertising with analytic values (full details are in Appendix A). We simulate 200 markets and optimal advertising decisions for both firms at a range of parameter values for β1 and β2 . The plots below show the difference between estimated and analytic elasticities. The level of the surface indicates the bias in different areas of the parameter space: it is apparent that there can be upward (greater than zero) or downward (less than zero) bias in both own and rival advertising elasticities. In no simulation were own and rival elasticities both estimated with less than 5% bias.40 40 Table 12 shows estimates and standard errors for a particular set of parameter values. Figure 8: Simulations of OLS Estimate Bias A.2 Simulation Details Parameters were set to the following values: α1 = α2 = −0.3, c = 1, ρ = 1000. Matlab’s FSOLVE function was used to set a system of first-order conditions to zero. We use 200 markets and we draw values of ξ for each firm in each market where ξ ∼ N(0, 0.25). Analytic values of own and rival advertising elasticities are calculated as the mean over all observations of
aj β1 (1 − s j ) − β2 s− j 1+aj
a− j β2 (1 − s j ) − β1 s− j = 1 + a− j ηown = ηrival We drop any simulations where Matlab’s FSOLVE function failed to converge to a solution for firm first-order conditions for advertising levels. The full space of simulations covered β1 ∈ [0.01, 0.2] and β2 ∈ [−0.2, 0.1], both in increments of 0.005. We drop cases where β2 > β1 as firms would choose negative advertising. The share of simulations where the bias in estimating own advertising elasticity was less than 5%, was only 0.3% of simulations, and 1.5% for rival advertising elasticity. Table 12 shows for one particular set of parameter values the OLS bias in estimating elasticities of own and rival ads. For completeness, we also performed the same analysis where only a single firm chooses advertising. Figure 9 shows the bias from OLS estimation of the elasticity of revenue with respect to own advertising. As is clear, the bias can be positive or negative. Figure 9: OLS Bias with One Firm Table 12: Sample Model Simulation Results Dependent Variable: Specification: Log(1+Own Ads) Log(1+Rival Ads) Control: ξ N R2 Naive (1) 0.0290∗∗∗ (0.0065) −0.0770∗∗∗ (0.0059) 200 0.687 Log(Revenue) With ξ Analytic Values (2) (3) ∗∗∗ 0.0796 0.0763 (0.0006) −0.1305∗∗∗ −0.1266 (0.0006) X 200 0.998 Notes: Parameter values for these results were β1 = 0.06 and β2 = −0.15. Firm optimal advertising levels were solved for using Matlab’s FSOLVE routine and first-order conditions for profit maximization. Estimates are OLS results for equation 1, with ξ j and ξ− j as additional controls in the second column. Analytic values are computed as the means of the expressions for ηown and ηrival shown above. The controlled version does not perfectly match the analytic values as the Logit model creates a non-linear error term when estimating equation1. B Additional Summary Statistics and Robustness Checks Figure 10: Political Ad Levels, July-November 2008 Notes: The above maps show a circle for each DMA in the USA. The diameter of each circle is proportional to the number of political ads aired in that market, in that month, for all races (Presidential, Senatorial, House, Gubernatorial). The first row are July and August; second row are September and October, and third row is November. Figure 11: Time Trends Notes: The above graphic plots the share of Lipitor and Crestor from the Truven data as a percentage of total category sales over the period of January 2007-November 2008. Note different axes. Figure 12: Instrument Effect Heterogeneity Lipitor Crestor, Vytorin, Zetia All Months Except Regulatory Action Regulatory Action Months (April-August 2008) Notes: The above graphic plots binned scatterplots to show that political ads displace drug ads, as Figure 4, for different sub-samples of the data. “Regulatory Action” months refer to the months when Lipitor was banned from advertising by congress. Figure 13: Effect of Placebo Primaries on Shares of Non-Advertised Sales Note: The above plots estimated coefficients for timing dummies relative to a market’s primary month, with the “timing” of the primary shifted 12 months forward. The dependent variable is the (one-month) change in market share, defined as the percentage of the population taking a non-advertised statin, Crestor, or Lipitor, respectively. Table 13: Robustness: No Substitution to Earlier/Later Months Model: Political Ads (1000s) One Month Lag One Month Lead Controls: Market FEs Year-Month FEs Drug FEs Drug National Ads N R2 Dependent Variable: Local Drug Ads, Product-Market-Year-Month Level OLS OLS OLS OLS ∗∗∗ ∗∗ −0.0819 −0.0632 (0.0263) (0.0304) 0.0265 0.0012 (0.0284) (0.0299) −0.0239 -0.0405 (0.0301) (0.0294) X X X X 8, 925 0.225 X X X X 8, 925 0.225 X X X X 8, 120 0.219 X X X X 8, 120 0.218 Notes: Regressions combine the Wisconsin and Kantar data sets in 2008. OLS standard errors clustered at the market-year-month level. Results differ from Table 3 as this is at the individual drug level. Statistical significance at the 10%, 5% and 1% levels are denoted by ∗ , ∗∗ , and ∗∗∗ . Table 14: Robustness: No Substitution to Other Media Model: Political Ads (1000s) Local TV Drug Ads National TV Drug Ads Controls: Market FEs Year-Month FEs Drug FEs N R2 Dependent Variable: Local Non-TV Advertising Spending OLS OLS OLS ∗ −0.3000 −0.1731 −0.1962 (0.1770) (0.1802) (0.1804) 0.8664∗∗∗ 0.9515∗∗∗ (0.1366) (0.1433) −0.0724∗∗∗ (0.0138) X X X 14, 867 0.074 X X X 14, 867 0.086 X X X 14, 867 0.087 Notes: Regressions combine the Wisconsin and Kantar data sets for the months of July 2007-November 2008. OLS standard errors clustered at the market-year-month level. Statistical significance at the 10%, 5% and 1% levels are denoted by ∗ , ∗∗ , and ∗∗∗ . Table 15: Political Ads Displace Drug Ads Model: Political Ads (1000s) Controls: Market FEs Year-Month FEs Drug FEs Drug-Year-Month FEs N R2 OLS −0.2668∗∗∗ (0.0245) X Dependent Variable: Statin Ads OLS OLS Tobit Tobit ∗∗∗ ∗∗∗ ∗∗∗ −0.2102 −0.2102 −1.2548 −0.6824∗∗∗ (0.0263) (0.0264) (0.1198) (0.0992) X X X X 14, 280 0.193 14, 280 0.244 X X X X 14, 280 0.366 X X X X X 14, 280 0.140 14, 280 0.219 Tobit −0.7536∗∗∗ (0.0228) X X X X 14, 280 0.312 Notes: Unit of observation is the market-month-product combination. Data (source: Wisconsin and Kantar) span 17 months from July 2007 to November 2008 and 210 markets and include Crestor, Lipitor, Vytorin, and Zetia. OLS and Tobit standard errors clustered at the market-year-month level. The results with Drug-Year-Month fixed effects are identical if local plus national statin ads are used as the dependent variable. Statistical significance at the 10%, 5% and 1% levels are denoted by ∗ , ∗∗ , and ∗∗∗ . Reported R2 is adjusted for OLS, pseudo for Tobit. Table 16: Robustness Checks (IV Results) Own Ads (2-Month Trailing) Rival Ads (2-Month Trailing) Jan.-Jun. ’07 Ads ...Controls: Market FEs Year FEs Drug FEs Drug FE*Time Trend Drug-Market FEs Clustering N R2 Dependent Variable: Log(Revenue per Insured) 0.0753∗∗∗ 0.0669∗∗∗ (0.0220) (0.0213) ∗∗∗ −0.0536 −0.0466∗∗∗ (0.0182) (0.0177) ∗∗∗ 0.0013 (0.0004) X X X X DMA 11, 465 0.831 X X X X X DMA 11, 465 0.945 Dependent Variable: Log(Days Supply per Insured) 0.0696∗∗∗ 0.0601∗∗∗ (0.0213) (0.0209) ∗∗∗ −0.0522 −0.0438∗∗∗ (0.0178) (0.0175) X X X X DMA 11, 465 0.822 X X X X X DMA 11, 465 0.946 Notes: Unit of observation is the market-month-product combination. Data (source: Truven) span 17 months from July 2007 to November 2008 and include Crestor, Lipitor, Vytorin, and Zetia. “Own Ads” and “Rival Ads” are constructed as log(1+X). All specifications are utilize a “Two Month Trailing Average,” which indicates that the independent variables are constructed as the average of advertising during the revenue month and the month before. First stage excluded instruments are political advertising, its square and cube, a dummy that takes on a one during April 2008-August 2008 (regulatory action dummy), and the interactions of the political variables and the regulatory action dummy. Statistical significant at the 10%, 5%, and 1% levels are denoted by *, **, and *** respectively. Table 17: Effect of Business Stealing (IV Results) Dependent Variable: Log(Revenue per Insured) Log Own Ads Log Rival Ads ...Controls: Market, Year, Drug FEs Drug FEs*Time Trend N R2 This Month 0.0136∗∗∗ 0.0961∗∗∗ (0.0018) (0.0331) −0.0608∗∗∗ (0.0243) Two-Month Trailing Average 0.0117∗∗∗ 0.0761∗∗∗ (0.0015) (0.0216) −0.0547∗∗∗ (0.0180) X X X X X X X X 11, 466 0.849 11, 466 0.810 11, 465 0.850 11, 465 0.827 Notes: Unit of observation is the market-month-product combination. Data (source: Truven) span 17 months from July 2007 to November 2008 and include Crestor, Lipitor, Vytorin, and Zetia. “Own Ads” and “Rival Ads” are constructed as log(1+X). “This Month” indicates contemporaneous advertising, while “Two Month Trailing Average” indicates that the independent variables are constructed as the average of advertising during the revenue month and the month before. First stage excluded instruments are political advertising, its square and cube, a dummy that takes on a one during April 2008-August 2008 (regulatory action dummy), and the interactions of the political variables and the regulatory action dummy. Statistical significant at the 10%, 5%, and 1% levels are denoted by *, **, and *** respectively. Standard errors are clustered at the market level. Table 18: Crestor and Lipitor Specifications, First Stage Robustness Own Ads Rival Ads Market Fixed Effects Year Fixed Effects Product Fixed Effects Prod-Time Trend Prod-Regulatory Action Fixed Effects Order of Polynomials, First Stage Observations Adjusted R-Squared (1) (2) 0.0320** -0.00481*** (0.0147) (0.0178) -0.0280** -0.0409*** (0.00121) (0.0146) X X X X X X X X Third 5,914 0.831 Second 5,914 0.819 (3) 0.0575** (0.0280) -0.0483** (0.0229) X X X X First 5,914 0.810 (4) 0.0626** (0.0287) -0.1154** (0.0578) X X X X X First 5,914 0.834 Notes: Unit of observation is the market-month-product combination. Data (source: Truven) span 17 months from July 2007 to November 2008 and include Crestor and Lipitor. “Own Ads” and “Rival Ads” are constructed as log(1+X). All specifications are utilize a “Two Month Trailing Average,” which indicates that the independent variables are constructed as the average of advertising during the revenue month and the month before. Statistical significant at the 10%, 5%, and 1% levels are denoted by *, **, and *** respectively.