Difference between revisions of "Galasso Schankerman (2010) - Patent Thickets Courts And The Market For Innovation"

From edegan.com
Jump to navigation Jump to search
imported>Ed
imported>John
Line 80: Line 80:
  
 
The expected length of a dispute is then <math>\mathbb{E}(t^*) = p^*\;</math>.
 
The expected length of a dispute is then <math>\mathbb{E}(t^*) = p^*\;</math>.
 +
 +
==Review==
 +
 +
===Measures of Patent Thicket===
 +
 +
Two measures of the fragmentation of patent rights faced by the infringer related to the litigated technology class are defined.
 +
*(1) Fragmentation of patent ownership among the top 4 firms in a technology class associated with citations of the infringer's patents is calculated as follows:
 +
**The share of citations by the infringer's patents in the litigated technology class to patents in each of the 417 technology classes over a 10-year window (5 years before and after the observation date) is calculated;
 +
**The share of patents held by the top 4 patentees in each of these technology classes over the same 10-year window is also calculated;
 +
**Fragmentation is then: one minus the product of these two shares summed over all technology classes.
 +
**A separate indicator variable is also defined to identify infringers that have no patent in the technology class of the litigated patent (25% of infringers).
 +
*(2) Fragmentation in assignees for the patents cited by the infringer's patents in the litigated USPTO technology class is calculated as follows:
 +
**First, the share of citations by the infringer's patents in the litigated technology class to distinct assignees is calculated over a 10-year window (5 years before and after the observation date);
 +
**Then, following Ziedonis (2004), framentation is defined as: 1 minus the sum of squared assignee citation shares, (also multiplied by a bias correction to correct for sample size, N/(N-1)).
 +
**Fragmentation measure 2 is used as a robustness check, since many infringers have very small patent portfolios in the litigated technology class, which makes Fragmentation measure 1 that relies on larger portfolios of top 4 patentees more reliable.
 +
 +
===Sample===
 +
 +
In 5,131 U.S. federal court cases where infringers could be matched to a patent assignee are analyzed from 1976-2000, based on data from:
 +
*9,219 patent infringement cases filed during 1975-2000 and terminated before 2001 were identified in Lit-Alert database;
 +
*Patent litigation progress and resolution obtained from U.S. federal district courts;
 +
*NBER patent dataset that matches US patents to Lit-Alert database.
 +
 +
===Results===
 +
*When infringers require access to more fragmented patent rights, disputes are settled faster, (a one standard deviation increase in the main fragmentation measure -fragmentation of litigated patent's citations among the top 4 firms in the technology class- leads a case to be settled 22 days sooner).
 +
**The effect is 30% larger when indicators of technology and field are included.
 +
**The effect is reduced with establishment of CAFC that reduced uncertainty of disputes substantially (from a marginal effect of -55 months to -7 months) - similarly higher variance courts exacerbated impact of fragmentation.
 +
**The second fragmentation measure (assignee fragmentation) measures an effect about half as large as the main fragmentation measure (top 4 firms).
 +
*Stronger complementarity among patents increases duraction of disputes, (a one standard deviation increase in the complementarity measure leads case to be settled 23 days later)
 +
*Establishment of CAFC reduced average settlement by 6 months.
 +
*A one standard deviation increase in patent value (as measured by citations) extends disputes by about 0.8 months, (=marginal effect(3.067)*absolute coefficient value(0.165)/st.dev(0.047)).
 +
*Results are robust to the inclusion of other controls such as accounting for serial infringers, serial patentees, different sized portfolios.
 +
 +
===Social Welfare Consequences===
 +
:''"There are two main empirical findings. First, patent disputes in U.S. district courts are settled more quickly when infringers require access to fragmented external rights, but this effect is much weaker after the introduction of the CAFC. Second, the introduction of the CAFC is associated with a direct and large reduction on the duration of disputes, which the model attributes to less uncertainty about the outcome if the dispute goes to trial. This finding has implications for current policy debates over the establishment of a European Patent Court with exclusive jurisdiction in cases dealing with infringement and revocation of European patents. Our result suggests that such a centralized court is likely to facilitate transactions in the European market for innovation. In addition, our calculations suggest that fragmentation may have reduced total negotiation delay, and thus sped up rather than retarded technology diffusion, in some technology areas during the period before the CAFC."''
 +
 +
===Dependent Variable and Model===
 +
*The dependent variable is legal dispute duration, measured by the number of months between original case filing and case termination, (averaging 18 months and 18 days).
 +
*A proportional hazard model with an exponential specification is used for analysis.
 +
*The model accounts for two measures of fragmentation:
 +
**Fragmentation of patent ownership among the top 4 firms in each technology class that is associated with citations of the infringer's patents;
 +
**Fragmentation in distinct assignees of patents cited by the infringer's patents.
 +
**Various interactions between the fragmentation measures and court regimes (post CAFC, high-variance courts).
 +
*The model also accounts for:
 +
**Complementarity of the patent to the litigated technology field, measured by the ratio of (non-self) citations to the litigated  patent up to 2002 in the litigated technology class compared to all (non-self) citations of patents in a technology class over a 10 year window; and normalized so it captures infringers share of citations in the litigated technology field.
 +
**Patent value, measured by the total citations to the litigated patent;
 +
**An indicator for patent suits filed after CAFC, (=1 after 1982);
 +
**An indicator for the top 3-4 courts with highest variance in outcomes;
 +
**An indicator for duplicative cases (same infringer and patent);
 +
**Indicators for technology field;
 +
**Indicators for district courts;
 +
**Indicators of small or large portfolios;
 +
**Indicators of serial patentees or serial infringers.

Revision as of 19:14, 29 March 2013

Reference

  • Galasso, A. and Schankerman, M. (2010), "Patent thickets, courts, and the market for innovation", The RAND journal of economics, Vol.41, No.3, pp.472--503
@article{galasso2010patent,
  title={Patent thickets, courts, and the market for innovation},
  author={Galasso, A. and Schankerman, M.},
  journal={The RAND journal of economics},
  volume={41},
  number={3},
  pages={472--503},
  year={2010},
  abstract={We study how fragmentation of patent rights (?patent thickets?) and the formation of the Court of Appeal for the Federal Circuit (CAFC) affected the duration of patent disputes, and thus the speed of technology diffusion through licensing. We develop a model of patent litigation which predicts faster settlement agreements when patent rights are fragmented and when there is less uncertainty about court outcomes, as was associated with the ?pro-patent shift? of the CAFC. The model also predicts that the impact of fragmentation on settlement duration should be smaller under the CAFC. We confirm these predictions empirically using a dataset that covers nearly all patent suits in U.S. federal district courts during the period 1975-2000. Finally, we analyze how fragmentation affects total settlement delay, taking into account both reduction in duration per dispute and the increase in the number of required patent negotiations associated with patent thickets.},
  discipline={Econ},
  research_type={Empirical},
  industry={},
  thicket_stance={},
  thicket_stance_extract={},
  thicket_def={},
  thicket_def_extract={},  
  tags={},
  filename={Galasso Schankerman (2010) - Patent Thickets Courts And The Market For Innovation.pdf}
}

File(s)

Abstract

We study how fragmentation of patent rights (?patent thickets?) and the formation of the Court of Appeal for the Federal Circuit (CAFC) affected the duration of patent disputes, and thus the speed of technology diffusion through licensing. We develop a model of patent litigation which predicts faster settlement agreements when patent rights are fragmented and when there is less uncertainty about court outcomes, as was associated with the ?pro-patent shift? of the CAFC. The model also predicts that the impact of fragmentation on settlement duration should be smaller under the CAFC. We confirm these predictions empirically using a dataset that covers nearly all patent suits in U.S. federal district courts during the period 1975-2000. Finally, we analyze how fragmentation affects total settlement delay, taking into account both reduction in duration per dispute and the increase in the number of required patent negotiations associated with patent thickets.

Model

Variables:

  • [math]V\;[/math] is the revenue of the licensee if he uses all [math]n\;[/math] patents
  • [math]\theta\;[/math] is a measures of complementarity:
[math]\theta = \begin{cases} 0 &\mbox{if perfect complements} \\ 1 &\mbox{if equally contributing} \\ \frac{n}{m} &\mbox{if perfect substitutes} \end{cases} \;[/math]
  • [math]\frac{m}{n}\theta V\;[/math] is the revenue of the licensee if he uses [math]m \lt n\;[/math] patents
  • The negotiation value when the firm already has [math]m=n-1\;[/math] of the patents is:
[math]z(n,\theta,V) = V - V\frac{n-1}{n}\theta\;[/math]

The set-up:

  • Both the patentee and the infringer are risk neutral
  • The infringer has private information and estimates that the patentee will prevail at trial with probability [math]p\;[/math]
  • The patenteee doesn't know [math]p\;[/math] but knows [math]p \sim U[0,1]\;[/math]
  • At [math]T=0\;[/math] the patentee makes a TIOLI settlement offer to the infringer. If the infringer accepts the game ends.
  • If the offer is rejected, at [math]T=1\;[/math] there is a trial. The parties incur costs [math]L_p\;[/math] and [math]L_i\;[/math]. If the infringer is found liable the court awards damages of [math]z\;[/math]. The paper says that this is consistent with the Unjust Enrichment doctrine.

By backwards induction:

An infringer of type [math]p\;[/math] would accept a settlement of [math]S\;[/math] only if:

[math]S \le p z + L_i\;[/math]
[math]\therefore p \ge \frac{S-L}{z}\;[/math]

The patentee's optimization problem is to maximize his expected profit by choosing a cutoff [math]p^*\;[/math] (such that above this cut-off the infringer accepts):

[math]\max_p \pi = \int_p^1 (pz + L_i) dy + \int_0^p yz - L_p dy\;[/math]

Letting [math]L = L_i + L_p\;[/math]:

[math]p^* = 1 - \frac{L}{z}\;[/math]

The expected length of a dispute is then [math]\mathbb{E}(t^*) = p^*\;[/math].

Review

Measures of Patent Thicket

Two measures of the fragmentation of patent rights faced by the infringer related to the litigated technology class are defined.

  • (1) Fragmentation of patent ownership among the top 4 firms in a technology class associated with citations of the infringer's patents is calculated as follows:
    • The share of citations by the infringer's patents in the litigated technology class to patents in each of the 417 technology classes over a 10-year window (5 years before and after the observation date) is calculated;
    • The share of patents held by the top 4 patentees in each of these technology classes over the same 10-year window is also calculated;
    • Fragmentation is then: one minus the product of these two shares summed over all technology classes.
    • A separate indicator variable is also defined to identify infringers that have no patent in the technology class of the litigated patent (25% of infringers).
  • (2) Fragmentation in assignees for the patents cited by the infringer's patents in the litigated USPTO technology class is calculated as follows:
    • First, the share of citations by the infringer's patents in the litigated technology class to distinct assignees is calculated over a 10-year window (5 years before and after the observation date);
    • Then, following Ziedonis (2004), framentation is defined as: 1 minus the sum of squared assignee citation shares, (also multiplied by a bias correction to correct for sample size, N/(N-1)).
    • Fragmentation measure 2 is used as a robustness check, since many infringers have very small patent portfolios in the litigated technology class, which makes Fragmentation measure 1 that relies on larger portfolios of top 4 patentees more reliable.

Sample

In 5,131 U.S. federal court cases where infringers could be matched to a patent assignee are analyzed from 1976-2000, based on data from:

  • 9,219 patent infringement cases filed during 1975-2000 and terminated before 2001 were identified in Lit-Alert database;
  • Patent litigation progress and resolution obtained from U.S. federal district courts;
  • NBER patent dataset that matches US patents to Lit-Alert database.

Results

  • When infringers require access to more fragmented patent rights, disputes are settled faster, (a one standard deviation increase in the main fragmentation measure -fragmentation of litigated patent's citations among the top 4 firms in the technology class- leads a case to be settled 22 days sooner).
    • The effect is 30% larger when indicators of technology and field are included.
    • The effect is reduced with establishment of CAFC that reduced uncertainty of disputes substantially (from a marginal effect of -55 months to -7 months) - similarly higher variance courts exacerbated impact of fragmentation.
    • The second fragmentation measure (assignee fragmentation) measures an effect about half as large as the main fragmentation measure (top 4 firms).
  • Stronger complementarity among patents increases duraction of disputes, (a one standard deviation increase in the complementarity measure leads case to be settled 23 days later)
  • Establishment of CAFC reduced average settlement by 6 months.
  • A one standard deviation increase in patent value (as measured by citations) extends disputes by about 0.8 months, (=marginal effect(3.067)*absolute coefficient value(0.165)/st.dev(0.047)).
  • Results are robust to the inclusion of other controls such as accounting for serial infringers, serial patentees, different sized portfolios.

Social Welfare Consequences

"There are two main empirical findings. First, patent disputes in U.S. district courts are settled more quickly when infringers require access to fragmented external rights, but this effect is much weaker after the introduction of the CAFC. Second, the introduction of the CAFC is associated with a direct and large reduction on the duration of disputes, which the model attributes to less uncertainty about the outcome if the dispute goes to trial. This finding has implications for current policy debates over the establishment of a European Patent Court with exclusive jurisdiction in cases dealing with infringement and revocation of European patents. Our result suggests that such a centralized court is likely to facilitate transactions in the European market for innovation. In addition, our calculations suggest that fragmentation may have reduced total negotiation delay, and thus sped up rather than retarded technology diffusion, in some technology areas during the period before the CAFC."

Dependent Variable and Model

  • The dependent variable is legal dispute duration, measured by the number of months between original case filing and case termination, (averaging 18 months and 18 days).
  • A proportional hazard model with an exponential specification is used for analysis.
  • The model accounts for two measures of fragmentation:
    • Fragmentation of patent ownership among the top 4 firms in each technology class that is associated with citations of the infringer's patents;
    • Fragmentation in distinct assignees of patents cited by the infringer's patents.
    • Various interactions between the fragmentation measures and court regimes (post CAFC, high-variance courts).
  • The model also accounts for:
    • Complementarity of the patent to the litigated technology field, measured by the ratio of (non-self) citations to the litigated patent up to 2002 in the litigated technology class compared to all (non-self) citations of patents in a technology class over a 10 year window; and normalized so it captures infringers share of citations in the litigated technology field.
    • Patent value, measured by the total citations to the litigated patent;
    • An indicator for patent suits filed after CAFC, (=1 after 1982);
    • An indicator for the top 3-4 courts with highest variance in outcomes;
    • An indicator for duplicative cases (same infringer and patent);
    • Indicators for technology field;
    • Indicators for district courts;
    • Indicators of small or large portfolios;
    • Indicators of serial patentees or serial infringers.