CMR17

Transcription

1 Discrimination against foreigners in the patent system : evidence from standard-essential patents Citation for published version (APA): de Rassenfosse, G., Raiteri, E., & Bekkers, R. N. A. (2017). Discrimination against foreigners in the patent system : evidence from standard-essential patents Paper presented at DRUID 2017, June 12-14, 2017, New York City, New York City, United States. Document status and date: Published: 01/06/2017 Document Version: Publisher s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication: A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. The final author version and the galley proof are versions of the publication after peer review. The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. Users may download and print one copy of any publication from the public portal for the purpose of private study or research. You may not further distribute the material or use it for any profit-making activity or commercial gain You may freely distribute the URL identifying the publication in the public portal. If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the Taverne license above, please follow below link for the End User Agreement: Take down policy If you believe that this document breaches copyright please contact us at: openaccess@tue.nl providing details and we will investigate your claim. Download date: 29. Jul. 2019

2 Powered by TCPDF ( p ー エッ ーイ ウ nyu@s エ オウゥョ ウウ L@n ッイォ L@j d ゥウ イゥュゥョ ァ ヲッイ ゥァョ ー エ ウケウエ ュ Z@e カゥ ヲイッュウエ ョ イ M ウウ ョエゥ ー エ ョエウ e ゥエ イゥ e ーッャケエ ヲ オウ c ッャャ ョ ァ ュ ィョッャッァケ ュゥャゥッ N イ ゥエ イゥ ` ーヲャ N ィ g ウウ ョヲッウウ epfl c ッャャ ョ ァ ュ ィョッャッァケァ エ ョ N イ ウウ ョヲッウウ ` ーヲャ N ィ r オ ォォ イウ e ゥョ ィッカ ョゥカ ィョッャッァケ s ョョッカ ゥ ョ ウ rnnnanb ォォ イウ ` エオ a ウエイ エ t ー ー エ エイ ゥウ イゥュゥョ ァ ヲッイ ー エ ーイッウ ーイッ ウウ N@i ヲッ オウ ウー ゥヲゥ ー エ ーーャゥ ャ ウウ ョエゥ エッ エ ィョッャッァゥ ャウエ ョ イ N@t ウエ ョ イ M ウウ ョエゥ ー エ ウ I@ ー イエゥ オャ ウエイ エ ァゥ ゥューッイエ ウオ ー エ ィゥョ ゥョ ゥァ ゥョョッカ ーイッァイ ュ ゥョヲッイュ ッヲ ゥウ ヲエ ー エ ーーャゥ ョエ ク ュゥョ エゥッョ I@ ゥョヲ ャゥォ ーイ ウ ッヲ ゥウ イゥュゥョ エゥッョ エィ ー エ ーーャゥ ヲッイ エイ オョヲ カッイ キィ ク ュゥョ ォョッキエィ エィ ャ ウエ ョ ウウ ョエゥ ャ ゥョエ イーイ イ ウ エ ーイッエ エゥッョゥウュ N j ャ ッ ウ ZoSTLM

3 Discrimination against foreigners in the patent system: Evidence from standard-essential patents Draft February 15th, Abstract This paper tests for traces of discrimination against foreign firms in the patent prosecution process. It focuses on the case of China and looks specifically at patent applications declared as essential to a technological standard. The choice of standard-essential patents (SEPs) is particularly suited because of the strategic importance of such patents for China s indigenous innovation program. We exploit information on the timing of disclosure as SEP (before or after the patent application enters examination) to infer the likely presence of discrimination. We find that patent applications by foreign firms are treated unfavorably when examiners know that they are declared as standard essential. We interpret this result as a case of technology protectionism. JEL Codes: F53, F68, K39, L52, L63, O25, O34 Keywords: discrimination, indigenous innovation, national treatment principle, standard essential patent, technology protectionism 1

4 1. Introduction The global rise of Chinese corporations is undeniable. They initially prospered by relying on cheap labour and exploiting economies of scale that the sheer size of the internal market offers. They are now becoming more sophisticated, buying firms in technologically-advanced countries and challenging established innovation champions. This situation is the result of a long transformation process that accelerated in the 2000s, when the country embraced at set of policies aimed at promoting indigenous innovation. The overarching objective was to become ascientificallyandtechnologicallyadvancedcountrybyyear2020withuniqueintellectual property (IP) assets. No industry illustrates better the rise of Chinese champions than the telecommunications industry. It is one of the selected strategic industry that the government has actively sought to nurture (Ernst, 2011). This industry heavily relies on technical standards to ensure devices and networks can interoperate. For companies active in this industry, owning patents on technologies used in those standards, known as standard-essential patents (SEPs), is a business imperative. By being indispensable to any party wanting to implement a technical standard, SEPs offer opportunitiesforcollectinglicensingfees,provideastrongbargainingpositionforcross-licensing deals, and represent a strategic asset to counter patent infringement accusations by competitors (Kang and Bekkers, 2015). Ownership of SEPs that have legal validity in China is particularly important as the country is not only one of the largest consumer markets for high-tech products, but also a manufacturing hotspot for many products relying on these standards. A well-known example is the iphone city of Zhengzhou, in east-central China, where the manufacturing firm Foxconn has 350,000 workers producing up to 500,000 iphone devices per day (Barboza, 2016). Failure for Chinese companies to own essential patents for such standards would force them to remain low-margin players. Initially, the Chinese government was keen to promote home-made standards and opposed to opening itself to global standards. Examples of such unique Chinese standards are the 3G mobile communications standard TD-SCDMA and the wireless encryption standard WAPI (Lee and Oh, 2008). In the end, this strategy was not very successful. These unique Chinese standards have generally been a failure: none of these standards have gained significant market support outside of China, and their adoption within China itself remained limited, despite strong government backing. Moreover, while the goal was that such indigenous standards 2

5 would mainly incorporate IP of Chinese owners, in reality they also incorporated significant amounts of foreign IP (Breznitz and Murphree, 2013). Yet, despite the above policy, some strong Chinese companies such as Huawei and ZTE developed into very capable and successful adopters of, and contributors to, global standards such as leading standards for 3G and 4G mobilecommunicationsdevelopedby3gpp(knownaswcdmaandlte,respectively). These companies are now among the most innovative in the world, according to the number of patents they file at the U.S. Patent and Trademark Office (USPTO) and the European Patent Office (EPO). The fact that China now finally embraces global standards, has not come without tension at the international level. Western observers expressed concern that Chinese competition authorities may target for investigation foreign firms that hold [patents] that may be essential to the implementation of certain standard technologies (USITC, 2014, 35). In fact, in April 2014, the Guangdong High Court of China published its judgements in a case between Chinese firm Huawei Technologies and U.S. firm InterDigital. The latter was found guilty of abusing its dominant market position regarding essential patents (Orrick, 2014). In 2015, China s National Development and Reform Commission (NDRC) found that the patent licensing schemes used by U.S. firm Qualcomm violated China s Anti-Monopoly Law. The firm was ordered by the NDRC to rectify its patent licensing schemes in a way that is compliant with China s laws and to pay a fine equivalent to $1billion(Lexology,2015). Another concern is that foreign companies may simply be denied patent protection for inventions that are standard essential. de Rassenfosse and Raiteri (2016) provide tentative evidence that foreign firms have a particularly low probability to receive a patent grant in China for inventions in selected strategic fields. In this study we aim to shed more light on the cause(s) for potential anti-foreign bias in the prosecution of SEP applications at the SIPO. We focus on patent applications declared as standard essential to two of the world s economically most important standards: the 3G WCDMA and 4G LTE standards for mobile communications, as created by 3GPP. We investigate two possible explanations: the availability of earlier search reports at other patent offices, and whether the patent application is known to be essential to a standard at the time of examination. We consider three dimensions of the prosecution process, namely the examination outcome, the duration of the examination process and the possible reduction of scope between 3

6 the patent application and the patent that was eventually granted. 2. Hypothesis development An invention deserves patent protection in a jurisdiction if it meets the patentability criteria in that jurisdiction. Generally, these criteria include novelty, inventive step/non-obviousness, and industrialapplicability,althoughtheirprecisedefinitionsdifferbetweenjurisdictions(derassenfosse et al., 2016). In terms of patents related to technical standards, for instance, the EPO has adopted a broader definition of what comprises the prior art for novelty searches compared to other patent offices (Bekkers et al., 2016). However, differences in patentability criteria across jurisdictions do not represent an anti-foreign bias, since these criteria apply to all applicants regardless of their country of origin. We are concerned with systematic differences in the outcome of the Chinese prosecution process between foreigners and locals. This section discusses two possible explanations that may account for such differences. Both explanations are related to the information that is available to SIPO examiners. The first potential explanation concerns the availability of prior art searches by other patent offices at the time a SIPO examiner scrutinizes the application. Inventions can be patented in multiple countries and, hence, be examined multiple times. Once the first application describing an invention is filed, the applicant has 12 months to seek protection in additional jurisdictions by filing so-called second filings. Thus, an examiner at an office of second filing may be able to consult search reports already written by colleagues at other offices. If patent examiners have access to earlier prior art searches, they may have additional information on the basis of which apatentcouldberejected informationtheymightnothavefoundthemselves. Thismight result in a less favourable examination outcome. Hence, we hypothesize: H1: Applications for which an earlier search report is available will have a less favourable application outcome, ceteris paribus. Note that validation of Hypothesis H1 would not be evidence of discrimination against foreign firms. The less favourable outcome of foreigners would simply be a consequence of the fact that foreign applicants are more likely to have filed their invention at other offices before filing at the SIPO compared to Chinese firms. The second explanation is a direct test of discrimination. They key element here is whether 4

7 the examiner (or, in fact, any other party), can obtain information on whether a particular patent is essential to a technical standard or not. Such information is publicly available via disclosure processes at Standard Setting Organizations (SSOs). Virtually all large SSOs have policies in place that require members to timely disclose patents or patent applications that are essential to a standard or, if a standard is still under development, essential to technical proposalsforthatstandard(seelemley,2002;bekkersandupdegrove,2012). 1 Suchdisclosures are then made publicly available via the websites of these SSOs. In theory, such information on essentiality should not affect the prosecution process because it does not alter the nature of the invention. But if foreign discrimination occurs, then the fact that the examiner is aware of the application s SEP status before the application enters the substantive examination phase could affect the outcome of the examination. 2 On the contrary, if the declaration to the SSO was made after the patent application entered the substantive examination phase, the examiner cannot identify that application as SEP when he (she) begins the examination process, and no such bias should occur. On the basis of the above, we hypothesize the following: H2: Foreign patent applications that are known to be standard essential at the time they enter the substantive examination phase at SIPO have a less favourable examination outcome, ceteris paribus. Clearly, validation of Hypothesis H2 would be evidence of discrimination against foreign firms. 3. Econometric approach 3.1. Regression models Our analysis covers three facets of the prosecution process: the likelihood of a grant, the duration of examination, and the reduction in scope of the application. The first outcome variable, grant i,capturesthegrantoutcomeofpatentapplicationfor invention i. Ittakesthevalue1ifthepatentapplicationwasgrantedand0ifitwasrejected 1 Because such policies also require potentially essential patents to be disclosed, not all declared patents will eventually be factually essential to the final standard. But this fact does not have any implications for the way this disclosure information is used in our paper. 2 Chinese patent law requires the applicant to file a request for examination within three years after the application date. 5

8 or withdrawn after the filing of a request for substantive examination. In order to test the two hypotheses, we estimate the following model: grant i =β 1 EPO sra i +β 2 USPTO sra i +β 3 PCT sra i +β 4 foreign i +β 5 known SEP i +β 6 (foreign known SEP) i (1) +β 7 PFE i +X i γ +ε i ThevariablesEPO sra i,uspto sra i andpct sra i (where sra standardsfor SearchReport Available ) are used to test Hypothesis H1. We focus on these three search reports as they account for the bulk of search reports in our sample. The variable EPO sra i indicates whether an EPO search report was available at the time the substantive examination at SIPO took place. It takes the value 1 if a patent application at SIPO has a direct equivalent at the EPO for which the initial search report performed by an EPO examiner is already publicly available before the applicant files the request for examination at SIPO. The variables USPTO sra i and PCT sra i are similarly defined. (See section 4 for more details.) The variables foreign i, known SEP i and interaction term (foreign known SEP) i are used to test Hypothesis H2. The dummy variable foreign i takes the value 1 if application for invention i is filed by a foreign applicant and 0 otherwise. The dummy variable known SEP i takes the value 1 if the date of the public disclosure of a patent application being a SEP for invention i pre-dates the request for examination, and thus the substantive examination phase, and 0 otherwise. Since our study focuses on the 3GPP WCDMA and LTE standards, we consider the date of disclosures at European Telecommunications Standards Institute (ETSI), the European SSO that is the partnering organization within 3GPP where the lion s share of patent disclosures for these standards are made. 3 The interaction term (foreign known SEP) i is our variable of interest for Hypothesis H2. It takes value 1 when the applicant is foreign and the patent is publicly known to be a SEP. Failure to control for invention quality would lead to biased estimates. In particular, we 3 3GPP is a partnership of regional SSOs and does not have its own IP policy or disclosure rules. Instead, companies participating in 3GPP must also be member of one or more of the partnering SSOs, and must use these SSO to disclose their IP. In practice, the bulk of disclosures for 3GPP standards takes place at ETSI. Disclosures at other partnering organizations are few, and usually overlap with those already present at ETSI. Baron et al. (2015) provides a detailed discussion of the 3GPP standards 6

9 may observe a discriminatory treatment against foreigners if patent applications by foreign firms are of systematically lower quality than applications by Chinese firms. To account for this possibility, we build on the recent twin patent approach (Webster et al., 2014; Sampat and Shadlen, 2015; de Rassenfosse et al., 2016). We track twin applications of invention i in other jurisdictions and we measure the variable PFE i as the average grant rate of these twin applications (de Rassenfosse and Raiteri, 2016). We interpret the variable PFE i as an invention pseudo fixed-effect that captures other patent offices assessment of the patentability of invention i. Finally, the vector variable X i includes a range of control variables that may affect the outcome of the examination process. The control variables are presented at the end of this section. The second outcome variable, grant lag i, reports the duration (in months) between the request for examination and the grant decision (for those patents that do get granted). We estimate the following regression model: grant lag i =β 1 EPO sra i +β 2 USPTO sra i +β 3 PCT sra i +β 4 foreign i +β 5 known SEP i+β 6 (foreign known SEP) i (2) +β 7 fast i +β 8 slow i +X i γ +ε i Most variables are similar to the earlier model. But instead of the pseudo fixed effect variable, we now include the variables fast i and slow i,whicharesimilarinspirittothepseudofixed effect. We define a patent applicant as fast (slow) if the average deviation from the mean of the grant lag at the equivalents patent office is in the top (bottom) decile. Thus this dummy reports whether the twin applications at other patent authorities were granted particularly fast (slow) compared to the average grant-lag for SEP patents at each authority. (Note that this control variable is used exclusively for the grant-lag analysis. 4 The third outcome variable relates to changes in the scope of the invention described in the patent document. We estimate the following model: 4 We lose some observations because it was not possible to retrieve the information on the grant-lag for some twin applications. 7

10 scope i =β 1 EPO sra i +β 2 USPTO sra i +β 3 PCT sra i +β 4 foreign i +β 5 known SEP i+β 6 (foreign known SEP) i (3) +X i γ +ε i The outcome variable scope is computed as the difference in the number of words per independent claim included in the granted patent and in the patent application. As suggested by Malackowski and Barney (2008) and Yoshimi et al. (2016), an increase in the number of words per independent claim between the patent application and the granted document is a proxy for the reduction in the scope of the patent due to examination. The rationale behind this interpretation is that each word added in a claim introduces a further legal limitation upon its scope. We provide a simplified example: suppose that the first independent claim of an application reads A bike brake using a round disk, while the first claim of the granted patent reads A bike brake using a round disc made of metal. Apparently, during the patent prosecution process, the examiner believed the first claim was too broad. The resulting, granted patent is reduced in scope, as it no longer covers breaks using non-metal discs, for instance carbon ceramic discs Covariates In all the above equations, the vector X i includes control variables that may affect the probability of grant at the SIPO. We consider the following patent-level control variables: Patent family size (family size) is the number of distinct countries covered by the IN- PADOC family. Number of IPC classes (tot IPC)is the number of IPC classes mentioned in the SIPO application. Number of inventors (nb inv)reportsthetotalnumberofinventorslistedinthepatent document. Examination-request lag (exam request lag)reportsthetime-laginmonthsbetweenthe date of application at the SIPO and the date of the request for examination. 8

11 Priority-declarationlag(dec prior lag)reportsthetime-laginmonthsbetweenthepriority date of the invention and its declaration date at ETSI. This variable allows to control for the age of the invention when is declared as essential to the standard implementation. Number of independent claims (nb indep claims) reportsthenumberofindependent claims listed in the patent application. Number of words per claim (words claim)reportstheaveragenumberofwordsperclaim included in a patent application. Difference in independent claims (diff ic)isavariablethatcollectsthedifferenceinthe number of independent claims between the patent application and the granted patent (Note that this control variable is used exclusively for the scope reduction analysis). We also control for four fixed effects: a invention pseudo fixed effect (discussed above); a firm fixed effect; an application year fixed effect; and an attorney agency fixed effect. Regarding the latter, China patent law stipulates that a foreign applicant that has no residence in China must appoint a licensed patent attorney agency to act as its agent to handle the patent application. Chinese applicants may instead appoint any patent attorney agency. The quality of the agency may affect the grant outcome and the grant lag, especially if there are differences in the quality of attorneys between foreigners and locals. We then include a binary variable for each of the patent agency in our sample. Although previous studies on SEPs suggest that disclosure timing may be correlated with patent characteristics that affect both the grant outcome and the duration of the examination process(bekkersetal.,2012;bergeretal.,2012),thereisnoreasontobelievethattheincentives to disclose an application at a given point in time should be systematically different for Chinese and foreign applicants. 4. Data 4.1. Data sources and sample construction We combine data from different sources. The EPO Worldwide Patent Statistical Database (PATSTAT, April 2015 edition) is the main source of patent information. We identify SEP applications by collecting disclosure data from ETSI and focus on disclosures related to the 3G WCDMA and 4G LTE standards developed by 3GPP; Appendix A expands on this data 9

12 collection. The INPADOC legal status table (a PATSTAT add-on) provides the information on the grant outcome at the SIPO and on the date of the grant. We also crawled the Google Patent website and the SIPO website to recover the number of independent and dependent claims at the SIPO, the number of words per claim, and information on the attorney agency that was used for the patent in question. 5 In order to put locals and foreigners on the same level, we impose that all applications in the sample have a direct equivalent at selected patent authorities. The selection ensures that we compare foreign applications with Chinese applications of international stature. A direct equivalent is a patent protecting exactly the same invention in a different jurisdiction (Martinez, 2010). We identify direct equivalents by identifying for each INPADOC family Chinese applications that claim only one priority filing and/or that are claimed by only one priority filing in a jurisdiction (that is, we exclude split equivalents and merged equivalents). One can think of this requirement as similar in spirit to the common support requirement in econometric matching models. The sample is then composed of SEP applications filed at the SIPO by foreign and domestic firms between 2001 and 2009, that have at least one unique direct equivalent in one of the following patent offices: CIPO, EPO, JPO, KIPO, RFSIP, TIPO, and USPTO. All these applications are disclosed at ETSI and related to the 3G WCDMA and 4G LTE standards. Due to data availability, we have produced four samples: Sample 1: 1653 SEP applications used for estimating regression model 1. Of these, 421 are filed by Chinese applicants and 1,232 filed by foreign applicants. A total of 457 applications (349 foreign and 108 Chinese) was declared as SEP before entering the examination phase at SIPO. Sample 2: the analysis for model 2 and model 3 is based on patents that are actually granted, which reduces our sample to 1,477 applications. The sample further reduces to 1,426 because for some patents the grant date of the twin application at other patent office is not available or the full text of the patent application is not available on the Google Patents website. The construction of the variables capturing the availability of search reports deserves further 5 See and 10

13 explanation. We construct the variable EPO sra i by exploiting the different publication kind codes at the EPO. A1 and A3 publication kinds refer to publications that include a search report. The variable takes value 1 if A1 or A3 publications pre-date the request for examination at SIPO. For the same reason, the variable takes value 1 of the EP patent was granted before the request for examination (publication code B1 or B2). In a similar way, USPTO sra i indicates whether an USPTO search report was available at the time the substantive examination at SIPO took place. It takes the value 1 if a patent application at SIPO has a direct equivalent at the USPTO for which the initial search realized by USPTO examiners is already available before the applicant files the request for examination at SIPO or if the USPTO patent was granted. To construct this variable we exploit the data from the USPTO - PUBLIC Patent Application Information Retrieval (Public PAIR) Database. 6 We determined the date of the PTO-892 form ( Notice of References Cited ) for the patent in question, which lists the first set of citations the examiner made to prior art. If the release of that document pre-dates the request for examination the variable USPTO sra i takes the value 1. Finally, PCT sra i is a binary variable that takes value 1 if the application reached SIPO through the PCT route and 0otherwise.SinceforpatentsthatfollowthePCTroute,thePCTsearchreportwillalwaysbe available before the substantive examination at SIPO takes place, we do not have to consider a specific date Descriptive statistics Table 1 displays descriptive statistics by country of residence for Sample 1. The last column of the table reports the difference in the averages of the two groups and its statistical significance. As the table shows Chinese applications on average have a higher issuance rate at SIPO than foreign applications (variable grant), although their grant rate at other patent offices is on average significantly lower (variable PFE). Chinese applications are also granted significantly faster than foreign applications (grant lag). There is no statistical difference between the two groups in the share of patents that are publicly disclosed as SEP when they enter into the examination phase at SIPO. In 60 per cent of the cases foreign applications reached the Chinese patent office through the PCT route. In half of the cases the search report by the EPO is already available for foreign applications, whereas the one by the USPTO is available only 6 Available at 11

14 in 17 per cent of the cases. Given that for Chinese applications the application at SIPO is very often the priority application, only a minor share reached SIPO through the PCT route and have USPTO or EPO search report available when the examination process began. Table 1: Descriptive statistics by country of residence Chinese applicants Foreign applicants t-test min mean max sd min mean max sd Diff. granted grant lag Scope known sep EPO sra USPTO sra PCT sra exam request lag nb inv dec prior lag log fam size log tot IPC nb indep claims log words claim PFE fast slow N The column t-test reports the difference between the averages of the two groups and the statistical significance of that difference. p< Results Table 2 displays the regression coefficients for the first model (Grant Outcome), estimated using as a linear probability model and a probit regression. 7 Concerning Hypothesis 1, we observe that the availability of USPTO and EPO search reports has a negative impact, although only the USPTO search report has a statistically significant effect. The lack of statistical significance for EPO search reports is surprising, as it is a common believe that the EPO is more rigorous than the USPTO in its searches, leading to lower 7 Note that the sample reduces to 1,425 applications for the probit model, as some of the agency and firms fixed-effects perfectly predicts the outcome. 12

15 Table 2: Results for Grant Outcome OLS Probit (1) (2) (3) (4) (5) (1) (2) (3) (4) (5) EPO sra (0.017) (0.017) (0.017) (0.014) (0.012) (0.011) USPTO sra (0.032) (0.031) (0.031) (0.020) (0.018) (0.017) PCT sra (0.027) (0.029) (0.030) (0.021) (0.023) (0.023) foreign (0.067) (0.068) (0.071) (0.048) (0.045) (0.031) known sep (0.035) (0.036) (0.036) (0.045) (0.044) (0.029) foreign known sep (0.041) (0.041) (0.040) (0.048) (0.046) (0.030) exam request lag (0.001) (0.001) (0.002) (0.001) (0.001) (0.001) log fam size (0.019) (0.021) (0.019) (0.014) (0.020) (0.013) log tot IPC (0.018) (0.019) (0.018) (0.012) (0.018) (0.011) nb inv (0.005) (0.006) (0.005) (0.004) (0.006) (0.004) dec prior lag (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) log NB indep claims (0.014) (0.015) (0.014) (0.009) (0.014) (0.008) log words claim (0.021) (0.023) (0.021) (0.015) (0.021) (0.014) Fixed effects: PFE (0.026) (0.027) (0.028) (0.030) (0.027) (0.019) (0.019) (0.023) (0.024) (0.018) Firm Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Agency Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes App Year Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes cons (0.045) (0.113) (0.082) (0.134) (0.133) N R Standard errors in parentheses p<0.1, p<0.05, p<0.01 Right-hand side columns report the marginal effect for the probit model allowance rates (Bekkers et al., 2016). Furthermore, the EPO has a more-encompassing definition of prior art which is especially relevant for the area this study focuses on: in contrast to other patent offices, also technical proposals that are shared in the context of standards setting are considered to be part of the state of the art (Ibid.). Even though a SIPO examiner should strictly adhere to the SIPO definitions of prior art, this still may lead to documents that the SIPO examiner could find a basis to reject a patent. Despite both factors, nevertheless, our data suggest no impact of the availability of EPO search reports. Concerning the effect of the availability of a PCT search report we see that, contrary to what we hypothesized, there is a positive impact: the availability of an PCT search report strongly increases the likelihood 13

16 that the SIPO examiner grants the patent. All the above findings are robust when we allow for alternative explanations via control variables. Altogether we conclude that the availability of a USPTO search report has a negative effect on the grant rate, one of EPO has no effect, and one of PCT has an unexpected positive effect, and thus reject the grant outcome part of Hypothesis 1. Concerning Hypothesis 2, we find a strong negative and significant effect (between 8.8 and 9.3 percentage points) on the grant outcome for foreign application disclosed as SEP before entering the substantive examination at SIPO. This effect is stable after control variables are added. For the estimation of the second model (Grant Lag) we both use OLS and Poisson regression models. Results are shown in Table 3. Because we can only consider patents that are actually granted, the overall sample is somewhat smaller than that for the first model. Concerning Hypothesis 1, we find that the availability of USPTO or EPO search reports does not affect the time lag. Yet, we observe that the availability of PCT search report before the SIPO examiner starts the substantive examination has a strong, positive effect on the time it takes to examine a SIPO patent. Note that the time lag we investigate is defined as the period between the applicant requests examination, and the time of patent grant, so the effect we observe cannot be simply attributed to the lengthy PCT procedure as such, which allows that national phases start as late as 30 months after the priority date of the original patent(wipo, 2015). Altogether, we only find partial support for Hypothesis 1. Concerning Hypothesis 2, we find a substantial positive and significant effect of our interaction term on the time to grant, that is on average between 8.5 and 12.6 months longer for foreign application disclosed as SEP before the examination process. This effect is always strongly present, and its magnitude is independent of the inclusion of controls. Altogether, this provides strong support for Hypothesis 2. The third model (Reduced Scope) is estimated through an OLS regression. Results are shown in Table 4. Note that a positive result means that the scope of a granted patent is reduced. Concerning Hypothesis 1, we observe that the prior availability of search reports (USPTO, or EPO) generally does not have much impact on the scope of the granted patent, compared to the scope of the patent application. While the availability of a PCT search report does impact the Scope variable. 14

17 Table 3: Results for Grant Lag OLS Poisson (1) (2) (3) (4) (5) (1) (2) (3) (4) (5) EPO sra (0.857) (0.857) (0.845) (0.751) (0.745) (0.730) USPTO sra (1.269) (1.254) (1.243) (1.087) (1.078) (1.058) PCT sra (1.421) (1.503) (1.499) (1.227) (1.297) (1.292) foreign (2.429) (2.461) (2.444) (2.510) (2.505) (2.458) known SEP (1.651) (1.713) (1.775) (2.780) (2.765) (2.826) foreign known SEP (1.943) (1.917) (1.959) (2.904) (2.865) (2.913) exam request lag (0.073) (0.069) (0.075) (0.066) (0.064) (0.069) log fam size (0.947) (0.983) (0.951) (0.858) (0.904) (0.860) log tot IPC (0.915) (0.919) (0.902) (0.862) (0.876) (0.847) nb inv (0.278) (0.280) (0.277) (0.238) (0.240) (0.238) dec prior lag (0.014) (0.019) (0.018) (0.013) (0.017) (0.016) log nb indep claims (0.740) (0.744) (0.724) (0.699) (0.703) (0.683) log words claim (1.082) (1.102) (1.082) (1.041) (1.058) (1.037) Fixed effects: fast (1.268) (1.225) (1.239) (1.207) (1.192) (1.377) (1.320) (1.354) (1.313) (1.280) slow (1.450) (1.449) (1.477) (1.491) (1.432) (1.167) (1.162) (1.194) (1.206) (1.146) Firm Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Agency Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes App Year Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes cons (2.709) (5.857) (3.666) (6.064) (6.226) N R Standard errors in parentheses p<0.1, p<0.05, p<0.01 Concerning Hypothesis 2, we see that foreign application disclosed as SEP (our interaction term) also of larger reduction in scope, with an average of 13.4 additional words per independent claim included during the examination process. Summing up our findings for Hypothesis H1 (related to the availability of search reports), we conclude that the availability of PCT reports lead to higher grant rates (favourable for the applicant), but also to longer grant lags and reduced patent scope (both unfavourable for the applicant). While availability of USPTO reports leads to lower grant rates, the availability of neither USPTO nor EPO search report has any other significant impact otherwise. Altogether, we see mixed and inconclusive evidence, and reject Hypothesis 1. 15

18 Table 4: Results for Reduced Scope scope (1) (2) (3) (4) (5) EPO sra (2.717) (2.729) (2.750) USPTO sra (4.419) (4.362) (4.388) PCT sra (4.003) (4.289) (4.290) foreign (7.285) (7.322) (7.293) known SEP (6.468) (6.607) (6.622) foreign known SEP (7.267) (7.345) (7.358) exam requestlag (0.219) (0.210) (0.226) log fam size (2.965) (2.920) (2.990) log tot IPC (3.097) (3.092) (3.087) nb inv (1.194) (1.219) (1.202) dec prior lag (0.042) (0.052) (0.052) log nb indep claims (2.719) (2.779) (2.750) diff ic (0.701) (0.733) (0.710) Fixed effects: Firm Effects Yes Yes Yes Yes Yes Agency Effects Yes Yes Yes Yes Yes App Year Effects Yes Yes Yes Yes Yes cons (13.129) (19.724) (15.468) (19.931) (21.305) N R Standard errors in parentheses p<0.1, p<0.05, p<0.01 Summing up our findings for Hypothesis H2 (related to the timing of disclosure as SEP), we conclude that in all aspects (grant rate, grant lag, and reduced scope), the outcomes for foreign applications of essential patents are less favourable than those for domestic, Chinese applicants. Thus, we find strong support for Hypothesis H2. 6. Robustness We performed several robustness checks to confirm the validity of our results. 16

19 6.1. Time window We ran the above regression models on a sample that excludes applications for which the (absolute) time lag between the declaration date and the request for examination date is shorter than three months. The rationale for conducting this test is to reduce potential confounding effects. It is indeed less likely that an application for which the examination process at SIPO begins immediately after it has been declared as essential at ETSI, could be already publicly knownasasepapplication. Inthesameway,itismorelikelythatanexaminercouldidentifyas SEP an application that is declared as essential soon after the start of the examination process at SIPO. Excluding applications in this three months time-window should then mitigate the potential bias introduced by this confounding factor. Table 5 reports the results of the analysis conducted on this reduced sample. As the table shows, the negative effect on the grant outcome for foreign SEP application disclosed before examination becomes larger in magnitude, reaching between 11 and 13 percentage points. Also the effect on the grant lag increases in size and is now on average between 9.9 and 15.1 months longer for foreign application disclosed as SEP before the examination process. The results on the reduction in scope is similar in magnitude but loses statistical significance Larger sample We also ran the analysis on a larger sample that is not restricted to applications that have adirectequivalentatotherspecificpatentauthorities. Thenewsampleiscomposedofapplications that belong to an international family (i.e., we do not consider families composed exclusively of the one application filed at SIPO) that have a unique application in China (i.e., we do not consider continuations and divisional applications). In this way our larger sample counts to 2,764 patent applications filed at SIPO, of which 2,207 are filed by foreigners and 557 by Chinese firms. 872 applications belong to families declared as SEP before the start of the examination process at SIPO and are hence flagged as known SEP. Table 6 reports the results of the analysis conducted on this enlarged sample. As the table shows, this robustness check confirm the direction and the significance of the effects that we find in the main analysis Measuring change in scope Although it is clear that an increase in the number of words per claim implies as reduction in scope, it would be wrong to interpret a decrease in the number of words per claim as an increase 17

20 Table 5: Results with a time window Grant Grant lag scope OLS Probit OLS Poisson OLS EPO sra (0.018) (0.011) (0.887) (0.750) (2.868) USPTO sra (0.032) (0.017) (1.170) (0.954) (4.205) PCT sra (0.031) (0.023) (1.789) (1.486) (4.505) foreign (0.073) (0.029) (2.553) (2.533) (7.589) known SEP (0.038) (0.031) (2.001) (3.060) (6.790) foreign known SEP (0.042) (0.033) (2.146) (3.115) (7.588) exam request lag (0.002) (0.001) (0.083) (0.074) (0.234) log fam size (0.020) (0.013) (1.012) (0.895) (3.138) log tot IPC (0.019) (0.012) (0.920) (0.850) (3.209) nb inv (0.006) (0.004) (0.295) (0.243) (1.255) dec prior lag (0.000) (0.000) (0.020) (0.017) (0.056) log nb indep claims (0.014) (0.008) (0.730) (0.668) (2.840) log words claim (0.022) (0.014) (1.074) (1.010) diff ic (0.707) Fixed effects: PFE (0.028) (0.019) fast (1.246) (1.366) slow (1.410) (1.103) Firm Effects Yes Yes Yes Yes Yes Agency Effects Yes Yes Yes Yes Yes App Year Effects Yes Yes Yes Yes Yes cons (0.139) (6.521) (22.085) N R Standard errors in parentheses p<0.1, p<0.05, p<0.01 in scope. Indeed, an increase in scope between the application and the grant is by definition not possible in patent law. Looking manually through several cases of patent applications having experienced a decrease in the number of words per claim, we almost invariable came to the conclusion that the changes were also associated with a reduction in scope. Again, using a 18

21 Table 6: Results with larger sample Grant Grant lag scope Probit Poisson OLS foreign (0.098) (5.680) (27.375) known SEP (0.031) (2.062) (6.157) foreign known SEP (0.033) (2.123) (6.561) log fam size (0.011) (0.620) (2.820) log ipc (0.008) (0.427) (1.858) decprior lag (0.000) (0.012) (0.046) PCT sra (0.016) (1.032) (3.314) log nb indep claims (0.008) (0.443) (2.170) log words pc (0.014) (0.753) nb inv (0.003) (0.158) (0.761) diff ic (0.568) Fixed effects: Firm Effects Yes Yes Yes Agency Effects Yes Yes Yes App Year Effects Yes Yes Yes cons (24.841) N R Standard errors in parentheses p<0.1, p<0.05, p<0.01 simplified example, think of an application with the claim A bike seat covered with leather, microfibre, or hemp canvass and the granted patent with the claim A bike seat covered with microfibre, or hemp canvass. Therefore, we also propose an alternative variable, namely the absolutenumberinthechangeofthenumberofwords. Therationalehereisthatanysignificant change, be it adding or removing parts of the claim text, leads to a reduction in scope. We call this variable Absolute Scope. Table 7 reports the results of the analysis when the variable Absolute Scope is used to measure reduction in scope. As the table shows, as in the case of the Scopeused in the main analysis, prior availability 19

22 Table 7: Results for Absolute reduction in scope Absolute Scope (1) (2) (3) (4) (5) EPO sra (2.503) (2.534) (2.554) USPTO sra (4.072) (4.005) (4.021) PCT sra (3.662) (3.862) (3.854) foreign (6.679) (6.808) (6.857) known SEP (6.047) (6.156) (6.176) foreign X known SEP (6.718) (6.737) (6.755) exam requestlag (0.205) (0.198) (0.213) log fam size (2.818) (2.760) (2.838) log tot IPC (2.868) (2.847) (2.847) nb inv (1.144) (1.164) (1.153) dec prior lag (0.038) (0.048) (0.048) log nb indep claims (2.472) (2.483) (2.480) diff ic (0.580) (0.589) (0.582) Fixed effects: Firm Effects Yes Yes Yes Yes Yes Agency Effects Yes Yes Yes Yes Yes App Year Effects Yes Yes Yes Yes Yes cons (12.110) (18.348) (14.191) (18.617) (19.949) N R Standard errors in parentheses p<0.1, p<0.05, p<0.01 of search reports does not affect the change in scope of the granted patent. A foreign application disclosed as SEP (our interaction term) experience a larger reduction in scope, with a change of up to 21 words per independent claim between the application and the granted document. 7. Conclusion and discussion This paper examines anti-foreign bias in the prosecution of patent applications. It focuses on patent applications filed at the SIPO and declared as standard essential to two of the world s economically most important standards: the 3G WCDMA and 4G LTE standards for mobile 20

23 communications. The choice of standard-essential patents is particularly suited because of the strategic importance of such patents for China s indigenous innovation program. Furthermore, this choice allows us to exploit information on the timing of disclosure as SEP to infer the likely presence of discrimination. Our findings can be summarised as follows. First, the availability of search reports before substantive examination at the SIPO has a mixed and inconclusive effect on the outcome of the prosecution process. The largest effect come from the PCT search report, although we find both favourable and unfavourable effects of PCT search report on the prosecution process. Second, standard-essential applications disclosed as SEP before the entrance into substantive examination phase at the SIPO are significantly less likely to receive a grant when the patent owner is foreign. Domestic patent owners do not experience such a drop in the likelihood to obtain a patent. Besides, if such foreign-owned patents do receive a grant, the grant decision arrives substantially later and the scope of the application is significantly reduced. We come to these findings after extensively controlling for a number of alternative explanations, including year effects, firm fixed effects, and patent attorney agency fixed effects, as well as for the examination request lag, patent family size, number of technology classes, number or inventors, time lag for essentiality declaration, number of independent claims, and difference in the number of independent claims between the patent application and the granted patent. Our findings have considerable implications, as China is not only one of the world s largest markets for products based on technical standards, but also a country where many of such products are manufactured for other markets. A weaker patent position of non-chinese firms is detrimental to innovative companies that have developed the underlying technology. An immediate application of our results is that foreign firms should disclose their SEPs after the patent has entered the substantive examination phase in order to increase the chance of a faire prosecution process. Yet, doing that would be acting in conflict with the ETSI disclosure rules, that stipulates such disclosures be made in a timely fashion. More generally, our findings suggest that China breaches the national treatment principle, one of the pillar of the international patent system that states that the prosecution process may not depend on the nationality of the applicant. An interesting question is whether the observed effect should be seen as a part of China s policy concerning technical standards, and its goal to strengthen its position in patents for such standards. Should this be the case, this finding may 21