U.S. Patent and Trademark Office Economics Lecture Series: Understanding the Innovation Economy

1. Global Diffusion of New Drugs: The Role of Patent Policy, Price Controls and InstitutionsIain Cockburn, Jean Lanjouw and Mark Schankerman U.S. Patent and Trademark Office Economics Lecture Series: Understanding the Innovation Economy 8 December 2011

2. Introduction • Welfare depends on “effective access” to new drugs • Effective access turns on: • Drug launches in country • Local prices of new drugs • Mechanisms for distribution Government policy/institutions affect all three elements. Trade-off between incentives for launch and accessible prices after launch.

3. We focus on three leading policies: Patent Policy: new drug manufacturing processes and compounds Price Regulation Government Drug Procurement/Regulatory Institutions These policies interact in shaping private incentives to launch new drugs.

4. Key Research, and Policy, Questions: Do patent and price regulation policy choices affect whether/when new drugs are launched in a country? Does the duration of protection matter? Do health policy institutions have any effect on drug launches, and if so, why? Do these factors matter for low and medium income countries as much as for upper income countries?

5. Related Economics Literature Empirical work on impact of price controls Kyle (2004) 21 OECD countries, 1577 molecules during 1980-2002 Negative impact of price controls on drug launches Empirical work on impact of patent policy Qian (2007) Weak effect of patent policy on domestic drug patenting Danzon, Wang and Wang (2005) 25 major markets, 85 new chemical entities 1994-1998 Launches more likely/faster for larger markets, higher prices

6. Economics of Drug Launches Launch = first appearance of a molecule in a country, whether proprietary or generic Price controls Reduce price-cost margins, and thus returns to market entry With fixed costs of entry → fewer/delayed launches Process patents Short process patents (“India strategy”) facilitate generic competition based on process innovation. Two effects: weakened incentives for product innovators → fewer/delayed launches greater local capacity to produce copies → more/faster launches

7. Economics of Drug Launches, cont. Long process patents can be used by drug companies to block this kind of generic, process-based entry, with two competing effects: weakened incentives for process innovators → fewer/delayed launches greater incentives for product innovators (less generic competition) → more/faster launches Product patents Longer, or stronger, property rights increase private returns to market entry and thus incentives to enter With fixed costs of entry → more/faster launches

8. The Launch Data IMS Health “LifeCycle: Drug Launches” database All product launches in 80+ countries - most cases for 1982-2002 Trade names of drugs Anatomical Therapeutic Classification (ATC) Composition (chemical constituents) Month and year of first sales in a given country Indian market data from ORG-MARG Only subset of therapeutic classes, incl. antibiotics, ulcer, cancer Includes products launched 1986-98 We match products on the basis of active ingredient (NCE)

9. “Names Game” • 130k records of product launches • ~ 2500 molecules, identity obscured by spelling variations, language, use of brand vs. generic name etc. • For the 2500 molecules, we identify >250k synonyms from multiple reference sources to allow standardization on INN (International Nonproprietary Name) or equivalent • Sample selection: remove pre-1981 and drugs launched only in subset Japan/Taiwan/Korea

10. Background on Drug Patent Protection Process patents: cover new methods of manufacture of pharmaceuticals Product patents: cover new compounds Statutory patent term: 1-20 years (application v. grant date) 1994 TRIPS Agreement harmonized term to 20 years, but with transition periods for some countries Measure of “Strength” of Property Rights Legal rules affecting content of patent protection Enforcement (institutional quality measures)

11. Patent Policy Variables Process and Product Patent Duration Dummies: None = no process/product patents Short = duration ≤ 10 years Medium = duration >10 and <16 Long = duration > 15 years Total of 6 dummies (3 for process, 3 for product patents)

12. Pro-patent Index • Count of various statutory provisions, including: • Same patent term for domestic/foreign applicants • Term extension available • Liability for contributory infringement • Preliminary injunctions available • Patents non-revocable if not worked • No patent working requirement • Compulsory licenses not usually issued • Based on Ginarte/Park + other country sources

13. Price Regulation Policy • None = no formal drug price regulation • Weak = formal but not extensive or restrictive • Strong = formal and extensive/restrictive • In regressions, we aggregate None + Weak (no significant difference in their effects, itself a noteworthy finding)

14. Control Variables Demographics GDP per capita (in PPP) Population size and age composition Gini coefficient of income inequality Health expenditures per GDP Health Policy Institutions Dummy for formal National Drug Policy Dummy for National Formulary Dummy for Essential Drug List Other Index of bureaucratic (regulatory) quality Index of rule of law

15. Income groups: Low, Middle and Upper (World Bank categories) Drug categories: All FDA Approved FDA Approved and Priority Reviewed Observations at level of molecule/country/year: 980 distinct molecules, 76 countries, 1982-2002

16. First Look: Univariate Analysis • Extent of drug launches • How many countries gain access to new drugs? • How does this vary with the quality of drugs and per capita income level? • Launch lags: Non-parametric Kaplan-Meier survival curves for different • Patent regimes • Price control regimes • Income levels

19. Summary • Spatial diffusion of drugs varies widely, with majority of drugs limited to very few countries • This geographic disparity is less sharp for more “important” drugs (FDA approved, Priority Reviewed) • Likelihood of drug launches rises with income per capita. High and low income countries differ by a factor more than two, even for “important” drugs

23. In summary • Lags are long, and many drugs are not launched at all in most countries • Patent protection reduces launch lags: both moving from None to Short/Medium (‘Other’) and from Short/Medium to Long • Strong price controls increase launch lags • Higher per capita income reduces launch lags

24. Hazard Model Analysis Weibull model with time-varying covariates Monotone hazard rate: prob. of first launch falls with time Adjusts for censoring (drugs may not have launched yet) Robust to other, more flexible functional forms (e.g. log-logistic) Therapeutic category dummies in all regressions Standard errors clustered at drug level (partial control for unobserved random drug effects) Interpretation: Negative of coefficients represent percentage impacts on the median (and mean) launch lag

27. Summary of Hazard Model Estimates Incentives matter: Effective market size Population and income per capita reduce launch lags Health expenditures per capita more than income per capita Inequality raises effective demand (given mean income) Older population raises demand and reduces launch lags Price-cost margins: price regulation raises launch lags Patent regime Duration: Both short and long patents make a difference Details of protection matter too: Pro-patent Index

28. Summary continued Health Institutions matter: Essential Drug List and National Formulary strongly reduce launch lags, but not National Drug Policy. These two institutions may be “gatekeepers” and thus have informal effect of reducing prices, but they strengthen entry hurdles for generic and proprietary drug competition. More effective (regulatory) bureaucracy screens out “low quality” drugs (non-FDA approved), increasing their launch lags. This screen is, sensibly, weaker on “high quality” drugs.

29. Robustness 1 FDA-approved drugs (high quality drugs) only Low/middle income countries only Introducing random (unobserved) drug effects Random (unobserved) country effects Control for domestic technical capability for generic competition (counts of U.S. chemical patents)

30. Robustness 2: FIML with Instruments Endogeneity: Are we confounding influence of policy regimes with unobserved country (fixed) effects? Do ‘high quality’ countries adopt stronger patent regimes (or less price controls), and that is why these regimes are correlated with faster drug launches? Statistical tests reject hypothesis that policy regimes are exogenous, so we estimate full model treating price control regime, process and product patent regimes as endogenous We need variables (instruments) that affect the choice of policy regimes, but do not affect launches except through these regimes.

31. Specification includes Hazard model for drug launches Equation for price controls (Probit) Equation for process patent regime (Ordered Probit) Equation for product patent regime (Ordered Probit) Political/legal structure variables: index of severity of political constraints political orientation of executive (L-C-R) ethno-linguistic fractionalization legal origins number of bilateral trade agreements

33. What Changes with FIML? Impacts of price controls and patent regimes are larger with instruments than in the baseline, nearly double the size Estimates are very stable across different sets of instruments LPROC now matters too (in baseline, it did not) SPROD is the only regime that has no effect consistently Health institutions still matter: National Drug Policy and National Formulary are most important

34. Key Lessons and Next Steps Patent policy matters Both process and product patents speed up launches, but short patents have no effect Strong price regulation delays launches Regulatory/procurement institutions also matter Need to consider how these policies interact to ensure maximum incentive effects, but weighed against ability of population to purchase the “available” drugs

35. Next Steps Update drug launches through 2010 to capture more of post-TRIPS experience Get more detailed information about health policy institutions, and new policies such as data exclusivity, and on reference pricing policies Study impact of these policies not just on drug launches but also on extent of use