Changes
Jump to navigation
Jump to search
Harhoff VonGraevenitz Wagner (2012) - Conflict Resolution Public Goods And Patent Thickets (view source)
Revision as of 14:35, 1 April 2013
, 14:35, 1 April 2013no edit summary
*This page is referenced in the [[Patent Thicket Literature Review]]
*This page is listed on the [[PTLR Core Papers]] page
Litigation and post-grant validity challenges at patent offices provide an important mechanism for correcting erroneous patent grants. However, such challenges will only be initiated if the (expected) private gains from challenging a granted patent right exceed the respective costs. Two important aspects may influence the likelihood of challenges. First, there is a public goods problem: firms may refrain from challenges if they anticipate that others will also benefit from the revocation of a weak patent. Second, as more firms are caught up in patent thickets, challenges to weak patents will become too costly as they invite counter-challenges. We use data on opposition proceedings initiated against patents granted at the European Patent Office (EPO) to study the importance of these mechanisms. This paper identifies a significant increase in the incidence of opposition in technical fields characterized by high concentration of patent ownership. Additionally, in fields with a large number of mutually blocking patents, the incidence of opposition is sharply reduced, particularly amongst those firms that are caught up in and driving the growth of patent thickets. Thus, while post-grant reviews may help to resolve problems in some areas, they are less suited to deal with patent thickets and contexts with dispersed patent ownership. We discuss the implications of these results for efforts to deal with patent thickets and weak patents.
==Review==
===Measures of Patent Thicket===
Patent thicket measures include:
*"Density of a patent thicket", following Graevenitz (2011), is measured by "triples" that count the number times critical patent references ("X or Y references") found in search reports, over the preceeding two years in a technology area, that are found among 3 patents held by the top 10 most frequently referenced firms in a given year. To test an "insider-outsider" hypothesis (whether firm is "inside" or "outside" the patent thicket), the measure is computed in two ways:
**A firm-level count of the number of triples in which the firm itself is involved (inside the thicket);
**A count of triples present in the technology area of the patent without involvement of the firm that holds the patent (outside the thicket).
*The paper also measures the overall concentration of patent ownership in the technology area among a firm's rivals using a Herfindahl index, to capture a "public goods" effect whereby fewer rivals increases benefits from successfully opposing a patent.
**While not directly viewed as a measure of patent thicket density, this feature is characteristically lower in complex technology area that tend to have patent thickets.
*The authors also consider Ziedonis's (2011) measure for fragmentation of ownership in a technology area in a year as an alternative measure of patent thickets and ownership concentration in a technology area.
===Sample===
1,044,069 patents granted by the European Patent Office (EPO) between 1980 and 2007 that were potentially subject to the 64,946 opposition proceedings recorded by the end of March 2011.
*PATSTAT database provides biliographic and legal information on patents.
*Patent applicant names are drawn from ECOOM-EUROSTAT-EPO PATSTAT Person Augmented Table (EEE-PPAT).
===Results===
Patent Thicket Results:
*Aggregate trends from 1980 to 2007:
**Opposition rates have declined in major technology groups.
**In complex technologies (Audiovisual Technology, Telecommunications, IT and Semiconductors all within the broader Electrical Engineering area), triple counts also rose while opposition declined.
*The marginal effect of another triple on the likelihood of oppposition to a firm's patent is -0.0010.
:''"Ten additional triple relationships (slightly more than half a standard deviation) reduce the incidence of opposition by one percentage point, that is a 16% reduction relative to the average unconditional probability of opposition [16%=1%reduction/mean opposition rate of 6%]."''
*When separately counting triples in which the firm is "inside" or "outside" the triple, the count of an "insider's" triples measures a larger effect than the count of "outsider" triples, although both effects remain sizeable, (the marginal effect of another own triple is -0.0013 vs -0.0009 from an outsider triple).
:''"[W]e confirm theoretical expectations that in patent thickets, incentives for adversarial actions are considerably reduced when agents are themselves directly affected by thickets. However, there is also an area-wide effect on outsiders who are not involved in the thicket, but are active in the technology in which a high density of thickets exists. This effect is only a third weaker than the effect for directly affected firms."''
*The concentration of patent ownership in a technology area by a firm's rivals increases the likelihood of opposition (with a marginal effect of 0.9039).
:''"The concentration variable should have a positive sign if there is a public goods effect. Indeed, we find that greater concentration of patents among rivals of the focal patent holder leads to an increase of oppositions: if concentration is increased by one standard deviation (0.0106), this leads to an increase in opposition by almost one percentage point (0.0160*0.9039=0.0096).
*Ziedonis' measure of the fragmentation of patent ownership is only weakly related to opposition (especially given large sample size) when the patent thicket measures and the rival ownership concentration measures are included.
**The authors interpret this as indicating that Ziedonis' fragementation measure is inferior to their measures for the density of patent thickets.
Other Results:
*Firms with larger patent portfolios face a lower incidence of opposition (with a marginal effect on likelihood of opposition of about -0.004 depending on controls included, or for a 100 additional patents a 0.4% lower chance of opposition).
*Authors find higher value patents less likely to face opposition.
**Higher generality, originality, backward X-type references increase likelihood of opposition.
**PCT filings, Japanese filings are less likely to face opposition.
===Social Welfare Consequences===
:''"The dense interaction between holders of mutually blocking patents may render litigation and post-grant review rather ineffective in particular technological areas as mechanisms that ensure that strong patents survive on the register while weak patents that are asserted get eliminated...the study shows that once firms have adopted the logic of patent portfolio races, the incentives to patent more and litigate less against holders of large patent portfolios spreads even to firms initially not caught up in the thicket but sharing the same technology. This tendency of patent thickets to self perpetuate and grow is worrying."''
===Dependent Variable and Model===
The dependent variable is an indicator of whether legal opposition is filed against a granted EPO patent within the statutory 9 month limit, (a mean 6% faced opposition).
*Probit models of the likelihood of opposition are estimated.
*The model controls for Patent Thicket measures including:
**A firm-level count of the number of triples in which the firm itself is involved;
**A count of triples present in the technology area without involvement of the firm;
**Concentation of patent ownership, measured by a Herfindahl index of rivals' granted patents in 30 (OST-INPI/FhG-ISI) technology classifications in a given year;
**The fragmentation of patent ownership in a technology area and year (following Ziedonis' measure) is also considered as an alternative measure.
*The model controls for patent applicant characteristics including:
**Cumulative number of patent applications filed by each applicant;
**Applicant's country of origin (U.S., Japan, Europe, other);
**Whether applicant is an individual, government institution, university or other (private enterprises).
*The model controls for patent value using:
**The total number of references (log of 1+total);
**The number of citations a patent receives over a 5 year period (log of 1+total);
**The number of jurisdictions in which equivalent patents have been filed;
**Number of claims;
**Share of backward references in each patent's search report associated with A-type, X-type or Y-type references;
**An indicator of international patent filing via the Patent Cooperation Treaty (PCT) application path;
**Number of equivalents, applicant's ROW, patent generality, patent originality.
*The model also includes grant year and 30 technology area fixed effects.
