Changes

Jump to navigation Jump to search
no edit summary
 
*This page is referenced in the [[Patent Thicket Literature Review]]
*This page is listed on the [[PTLR Core Papers]] page
Licensing can be a solution for hold-up in patent thickets. In this paper we study whether licensing is an effective solution for hold-up. To do this we distinguish between ex ante and ex post licensing. A theoretical model shows that firms’ expectations of blocking in a patent thicket determine whether they license ex ante while ex post licensing arises if expected blocking was low but realized blocking turns out to be high. It can also be shown that ex ante licensing will allow firms to reduce their patenting efforts. A sample selection model of licensing is derived from the theoretical model. Applying this to data from the semiconductor industry we show that licensing does help firms to resolve blocking. However, the probability of observing licensing decreases as fragmentation of property rights increases and arises mainly between large firms with similar market shares. Licensing experience is also an important determinant of licensing. As expected ex ante licensing allows firms to reduce the level of patenting.
 
==Review==
 
===Theoretical Model===
 
Overview:
*The article models the decision to enter into ex ante or ex post licenses as a 3 stage game between two firms competing in product and technology markets:
**Stage 1: Choosing ex ante contracting,
**Stage 2: Choosing hazard rate of becoming technology leader (via R&D investment), and
**Stage 3: Entering an ex post license to divide surplus if no ex ante license was entered into.
 
Model predictions:
*Proposition 1: The profit incentive becomes stronger relative to the competitive threat if (i) blocking becomes greater and/or (ii) the degree of product market substitution becomes smaller for 2>r.
:''"An intuitive way of thinking about these results is that higher blocking reduces competitive pressure in the patent portfolio race, while higher degrees of product market substitution increase competitive pressure in the patent portfolio race."''
*Proposition 2: Ex ante and ex post licensing are of equal value when new patents are completely blocked.
:''"The intuition for identical values of ex ante and ex post licensing under full blocking is that this removes any threat that firms may emerge with asymmetric payoffs ex post and reduces the competitive threat to zero."''
*Proposition 3: Two firms will always prefer ex post licensing over ex ante licensing if the profit incentive is stronger than the competitive threat and blocking is not complete. If the competitive threat is stronger than the profit incentive ex ante licensing can be more valuable than ex post licensing.
*Proposition 4. If blocking is close to complete (B > B(E=0)) any increase in the degree of blocking reduces the difference between the expected value of ex post and ex ante licensing.
 
===Measures of thicket===
 
The article defines a measure of the extent that two firms are in a patent thicket with each other:
*The blocking intensity of a paired firms technology, as measured by technologically more similar innovations and higher cross-citations between patents held by a pair of firms. It is the interaction of two measures:
**Following Jaffee (1986), the uncentered correlation coefficient between the distribution in the number of each firms' patents across the 9 semiconductor industries; and
**Citation intenstity, defined as number of firm i's citations to firm j's patents relative to the total number of firm i citations.
 
===Sample===
 
476 observations on pairs of U.S. semiconductor firms that entered into a licensing agreement from 1989 to 1999:
*847 licensing contracts between firms is drawn from Thomson Financial (including date of agreement and firms involved) and excludes agreements resulting from litigation.
**Ex ante licensing rose, peaked in 1994 at 103 licenses, then fell, while ex post licensing is relatively flat.
*Market shares and revenues for 268 firms is drawn from Gartner Group data.
*firm-level semiconductor patent (and citations) information is drawn from NBER utility patent database restricted to 9 technology classes related to the semiconductor industry, (semiconductors: 257, 326, 438, 505; memory: 360, 365, 369, 711; and microcomponents 714).
**semiconductor patents rose from 4,063 in 1989 to 12,624 in 1999.
 
===Results===
 
Key results related to patent thickets:
*The elasticity of ex ante (vs ex post) licensing to blocking is -0.183 and significant at the 1% level, indicating an effect from overlapping patent rights.
*Although the blocking variable is endogenous (based on Wald tests of whether it is correlated with errors of the equation), lagged blocking provides a strong instrument, and instrumental variables specifications using the lagged blocking variable confirm the above results are robust.
:''[F]irms are more likely to enter into ex ante (ex post) licensing agreements if their patents block each other to a lower (greater) extent. This finding confirms Hypothesis 1 that lower levels of blocking will induce firms to license ex ante.''
*The elasticity of ex ante (vs ex post) licensing to the product market competition measure is 0.180 and significant at the 5% level.
:''"[Confirming] Hypothesis 2 that higher levels of product market substitution will induce firms to license ex ante. A one standard deviation increase in the degree of product market substitution raises the probability of ex ante licensing by 7.2%."''
 
Other results:
*Prior experience in ex ante licensing is significantly positively related to ex ante licensing, reflecting transaction costs.
*Ex ante licensing is less likely in larger (revenue) markets.
*Ex ante licensing is less likely for smaller firms.
 
===Social Welfare Consequences===
:''"The model we provide in this paper shows that licensing allows firms to overcome some serious problems arising from the patent thicket. Thus, ex ante licensing helps firms to reduce somewhat the very strong incentives to patent that can arise in patent thickets. This paper has focused on the determinants of licensing ex ante and ex post and we have found that ex ante licensing is preferred, as predicted, when competition in the patent thicket is most intense."''
 
===Dependent Variable and Model===
*The dependent variables indicates when a firm-pair enters into an ex ante license agreement (=1), as opposed to engaging in an ex post licensing agreement (=0)
*A probit model is estimated that includes features of the theoretical model:
**Degree of blocking ability (see above detail);
**Degree of product market competition, as measured by the uncentered correlation between firm pair's market shares in the 3 semiconductor technology areas (following Bloom 2005);
**firms' cost of licensing ex ante and ex post, as proxied by a count of previous ex ante or ex post licensing agreements signed with other firms.
*The model also includes market controls:
**semiconductor industry revenues;
**firms' average market shares in the semiconductor industry;
**firms' difference in market shares;
**firms' average semiconductor patent stocks;
**time trend.
Anonymous user

Navigation menu