Moser and Voena (2012)

Question 1

Findings of Moser and Voena (2012)?

Answer 1

While our analysis suggests that compulsory licensing encourages domestic invention in the licensing country, the policy’s long-run effects include potentially important incentive effects on invention in the country whose inventions are licensed. Ex ante these effects are unclear because, for example, increased competition may either encourage or discourage innovation. In the case of the TWEA, the quick re-entry of German patentees suggest that negative incentive effects may be limited if compulsory licensing is a one-shot response to an emergency situation. Systematic analyses with additional data, however, are required to evaluate these effects. The response of US pharmaceuticals to compulsory licensing provisions in India and more recently under TRIPS offers a promising contemporary setting.

Question 2

How does compulsory licensing affect welfare domestically and abroad? Which channel do Moser and Voena provide evidence on?

Answer 2

1. Erosion of foreign monopoly power:

• Lower consumer prices (good domestically, possibly bad globally)
• Lower innovation incentives (bad, globally)

2. Facilitates diffusion of knowledge (MV)
3. Lower domestic innovation incentives (MV)
4. Reduced access, investment in CL countries (MV)

Question 3

How do MV measure innovation? Discuss advantages and disadvantages. Suggest alternative measures.

Answer 3

• Patents.
• Pro: Easy to measure, easy to value, easy to trace follow-up (citations)
• Cons: Not all innovation is patent (maybe in chemical industry they are); not all patents have the same effect
• Alternative measures: R&D expenditure, employment in research, productivity, value, patent sales, . .

Question 4

Table 2 shows that the DiD estimator of the effect of compulsory licensing on domestic innovation is positive. What are possible sources of bias in this estimate? How do MV attempt to rule out these sources of bias?

Answer 4

1. Attenuation bias due to measurement error in assignment of inventor nationality. (Checking assignment manually)
2. Spillovers?
3. Preexisting trends in licensing

• CL more common in classes with low US patenting (Figure 5)
• Reversion to mean or catching up in these subclasses
• MV run statistical tests on that. No evidence

4. Endogeneity of treatment (CL of promising industries) → MV run IV, no evidence
5. OVB (tripple difference estimation, US patenting rose relative to non-US, non-German inventors)

Question 5

Table 9 shows how compulsory licensing affected patenting by DuPont. What do these results imply about spillovers?

Answer 5

• Evidence for spillovers but:
• Spillovers slow (so are these the true effects?)
• Spillovers are asymmetric
• Endogeneity in patent selection (DP)

Question 6

Empirical strategy of Moser and Voena (2012)?

Answer 6

Diff-in-Diff regression – treatment group: subclasses where at least one german patent was licensed to U.S. firms.

Control group: Subclasses where no german patent was licensed to U.S. firms.

Question 7

how did they test for parallel trends assumption? (i.e. whether licensing decisions were endogenous?)

Answer 7

How did the authors check whether the licensing decisions are endogenous?

1. Intention to treat regression – found that ITT estimates were slightly smaller than OLS, suggesting that there were selection bias -> OLS underestimate the effect of compulsory licensing.
2. IV strategy: the number of enemy patents as instrument for licensed patents.
   
   1. Exclusion restriction: variation in enemy patents should not by itself increase domestic invention.
   2. Relevance: Enemy patents is highly correlated with the number of licenses that were granted to US firms

Question 8

How did the authors account for the potential of differential growth paths across treated and untreated subclasses?

Answer 8

Include the interaction term between year dummies and each of the broader 19 USPTO main classes. -> still showed significant effect from subclasses with at least 1 license to patents by US inventors per USPTO subclass and year.

Question 9

Why did the U.S. domestic firms need a learning period to fully adopt the German patents?

Answer 9

Because many German patents did not include full information regarding their processes, so U.S. firms needed time to fully understand how to utilize those patents.

Question 10

limitation?

Answer 10

When sample size is restricted to the primary subclasses, despite positive and significant coefficient on number of licenses’ and remaining lifetime of licensed patents’ effect on patents by US inventors per subclass per year, the subclasses having at least one license coefficient becomes statistically insignificant.