Changes
Jump to navigation
Jump to search
Siebert VonGraevenitz (2010) - Licensing In The Patent Thicket Timing And Benefits (view source)
Revision as of 18:46, 2 April 2013
, 18:46, 2 April 2013no edit summary
*This page is referenced in the [[Patent Thicket Literature Review]]
*This page is listed on the [[PTLR Core Papers]] page
Complex high technology industries are increasingly affected by patent thickets in which firms patents mutually block the use of important technologies. Firms facing patent thickets patent intensively to acquire bargaining chips and use licensing to ensure freedom to operate. Such licensing allows rivals to either avoid or resolve hold-up from blocking patents. R&D incentives depend on whether licensing takes place ex ante or ex post. We model the choice between ex ante licensing and entry into patent portfolio races leading to ex post licensing. It is shown that higher degrees of blocking lead firms to license ex post, while stronger product market competition leads firms to license ex ante. Empirical results support these theoretical predictions.
==Review==
===Measures of Patent Thicket===
Two measures related to patent thickets are considered:
*Realized blocking ability between firms as a measure of hold-up potential and is the product of:
**The sum of citation intensity for firms paired for analysis, each of which is the share of citations on the patents of a firm (i) that point to patents belonging to the paired firm (j) given the total citations by the firm (i) to other firms' patents.
**Technological proximity, measured by the uncentered correlation coefficient of paired firms' patent applications in a year across 9 patent classes (semiconductors, memory and microcomponents).
**Expected blocking is also considered as a factor in ex ante licensing decision using the lagged value of the realized blocking measure.
*Fragmentation of the patent citation stock between the two firms at a point in time following the measure constructed by Ziedonis (2004), with a sample size correction following Hall (2005).
===Sample===
30,905 firm pairs observed annually between 1989-1999, where at least one firm's principal line of business is in the semiconductor industry.
*Data analyzed includes 212 ex ante and 261 ex post licensing events after sample selections and exclusions due to missing data.
*Firm sample restrictions include:
**Only firms producing output in a given year (positive market share);
**Only firms that engage in horizontal (as opposed to vertically integrated) technology licensing - at least 50% of sales of firms are from semiconductors;
**Excluding licenses related to production, marketing or litigation;
**Only firms with non-missing data.
*847 licenses were gathered from Thomson Financial, supplemented by public information including business reports and filings published in the National Cooperative Research Act, and classified into 549 ex ante contracts and 298 ex post contracts prior to sample restrictions.
*Firm level financial data is from Gartner Group, and 60% were headquartered in the U.S.
*U.S. patents in 9 technology classes (related to memory chips, microcomponents and other semiconductor devices) are drawn from the NBER patent database, which covers 1963-1999, with citations covering 1975-1999.
===Results===
In aggregate:
*Licensing increased from 1988, then declined after 1994, driven primarily by a decline in (predominant) ex ante contracts.
*Blocking ability in firm pairs similarly increased then decreased, while fragmentation of patent rights in firm pairs generally rose.
Ex ante licensing:
*Chances of ex ante licensing increase with expected blocking ability.
:''"A one standard deviation increase in the expectation of blocking raises the probability of observing ex ante licensing by 0.0009 at the mean. This is an increase of 17% in the probability of ex ante licensing."''
*Chances of ex ante licensing fall with fragmentation.
:''"Turning to the fragmentation of firms patent citations we observe that this variable increased by over 0.4 over the sample period. This corresponds to a smaller probability of observing ex ante licensing of 10%."''
Ex post licensing:
*Chances of ex post licensing increase with actual blocking ability.
:''"The probability of observing ex post licensing rises by 0.0038 (54%) if blocking increases by one standard deviation at the mean."
*Chances of ex post licensing fall with fragmentation.
Measures of relative bargaining strength have qualitatively similar impacts on the mean chance of ex ante and ex post licensing:
*Chances of licensing rise as paired firms have larger market shares;
*Chances of licensing rise as paired firms have more symmetric market shares;
*Chances of licensing rise as paired firms have a larger patent stock.
The level of patenting by firm pairs:
*Falls by 12.5 patents for firms entering ex ante licenses;
*Rises by 4.36 patents with a 1 standard deviation increase in blocking;
*Increases by 6.5 patents for a 0.4 increase in fragmentation (the average increase in fragmentation over the sample).
===Social Welfare Consequences===
:''"These results demonstrate that expected blocking has important effects on firms’ propensity to license ex ante. Additionally, we find that ex ante licensing matters especially for large firm pairs in which partners are symmetrical. If firms have large patent portfolios they are less inclined to engage in ex ante licensing. More importantly, perhaps, we find that the trend towards greater fragmentation of patent citations undermines ex ante licensing significantly. This shows that as the potential for hold up grows due to greater fragmentation of patent ownership, ex ante licensing is less useful in preventing hold-up."''
:''"Ex post licensing is an important mechanism for firm pairs that have not licensed ex ante but find that they block each others patents to a significant degree. Especially larger and more symmetrical firm pairs resolve this problem by licensing ex post. Increasing fragmentation of patent citations also has a significant marginal effect on the ex post licensing decision. The sign of the coefficient indicates that ex post licensing is less likely as fragmentation increases."''
:''"Our results underline that licensing in the patent thicket is primarily important for large, symmetric firm pairs and is used to resolve potential hold-up. Overall the empirical results confirm semiconductor firms behave as if they are competing in patent portfolio races."
===Social Welfare Consequences===
:''"we find that the fragmentation of patent rights reduces firms’ propensity to license ex ante and ex post. Thus, a deepening of patent thickets resulting from more complex blocking relationships seems to undermine the usefulness of licensing to resolve blocking."''
===Dependent Variable and Model===
There are 2 independent variables:
*Whether a firm enters into an ex ante, and ex post or no contract modeled using a bivariate probit selection model.
*The joint level of patent applications in a firm pair for a given year.
Models control for:
*Realized blocking ability between firms as a measure of hold-up potential and is the product of:
**The sum of citation intensity for firms paired for analysis, each of which is the share of citations on the patents of a firm (i) that point to patents belowing to the paired firm (j) given the total citations by the firm (i) to other firms' patents;
**Technological proximity, measured by the uncentered correlation coefficient of paired firms' patent applications in a year across 9 patent classes (semiconductors, memory and microcomponents).
*Expected blocking ability, is the value of realized blocking described above lagged one year prior to the decision to engage in ex ante licensing.
**Expected blocking is insignificant when included as an explanatory variable for the ex post licensing decision, and therefore the preferred model excludes this variable when modelling the ex post licenting decision.
*Average patent stock;
*Difference in patent stocks of paired firms;
*Fragmentation of the patent citation stock between the two firms at a point in time following the measure constructed by Ziedonis (2004), with a sample size correction following Hall (2005);
*Market shares, as a measure of product competition;
*Difference in market shares of paired firms, as a measure of product competition;
*multimarket participation, measured by the number of different product markets (3 markets microcomponents, memory chips, other devices) within the semiconductor industry in which the firm has positive market shareand is intended as a measure of product competition;
*Previous ex post contracts;
*Previous ex ante contracts;
*Indicators for each year from 1990-1998.
