Patent System, Firm Patenting Strategy and Technology Progress. by Siwei Cao

Transcription

1 Patent System, Firm Patenting Strategy and Tehnology Progress by Siwei Cao A dissertation submitted in partial satisfation of the requirements for the degree of Dotor of Philosophy in Agriultural and Resoure Eonomis in the Graduate Division of the University of California, Berkeley Committee in harge: Professor Brian Wright, Chair Professor Jeffrey M. Perloff Professor Carl B. Shapiro Spring 2015

2 Patent System, Firm Patenting Strategy and Tehnology Progress Copyright 2015 by Siwei Cao

3 Abstrat Patent System, Firm Patenting Strategy and Tehnology Progress by Siwei Cao Dotor of Philosophy in Agriultural and Resoure Eonomis University of California, Berkeley Professor Brian Wright, Chair The patent system of the People s Republi of China has only a history of about 30 years, yet it has undergone three major poliy hanges to keep up with the tremendous pae of eonomi and tehnologial progress in China. One a ountry in whih tehnologial advane relies primarily on opying and imitating foreign innovations, China is gradually shifting towards a regime that emphasizes on strong intelletual property and indigenous innovation. It is fasinating, therefore, to witness this transition. It is important to enhane my understanding of the patent regime as a mehanism to promote innovation and the dissemination of knowledge, as is the purpose of the thesis. Both of the hapters in this dissertation fous on patent appliants preferene for how quikly they want their patent appliations to be proessed. This is nontrivial primarily due to heterogeneity in firm harateristis, nature of invention, market struture, tehnology bakgrounds and many unobservable fators. I propose two fators that partially determine an inventor s preferene for speed of patenting: the inter-temporal value patents of their invention and the speed of tehnologial progress. The first hapter investigates the heterogeneous inter-temporal patterns of inventions value flow and the assoiated appliants patenting strategies for a small set of United States patents. To measure the harateristis of different inventions inter-temporal value flow, we exploit a poliy lever provided by the Chinese government that allows patent appliants to freely hoose between one patent protetion of short examination delay, short protetion period and another patent protetion of long examination delay, long protetion period. We find that the majority of appliants that favor the former protetion (short&shot) have Eletronis and Mehanis inventions, whereas appliants that favor the latter (long&long) are mostly within Pharmaeutials and Drugs. We then exploit this variation in patent hoies under the Chinese regime with appliants patenting strategies in the U.S., for idential inventions. The empirial results suggest the short&short Chinese patent holders have a strong tendeny of pursuing early U.S. patent issuane whereas the long&long patent holders have a strong tendeny of maintaining their U.S. patents for long periods. The results are robust with or without tehnology field fixed effets. The seond hapter analyze the effet of faster tehnology development on firms patenting strategy. Using a dataset with information about patent appliations in both China and the US, I find that firms are willing to seure early patent grants when tehnology moves ahead faster. The onventional 1

4 wisdom that a patent seures a flow of monopoly profits that depreiates at a onstant speed over time is not onsistent with my empirial findings. Faster tehnology progress shifts the profits towards the early periods, making early grants more important. The empirial results suggest that a more flexible patent regime whih offers options for speed is more effiient. 2

5 Contents 1 Flow of Invention Value and Speed versus Length of Patent Protetion INTRODUCTION Coneptual Issues Inventions with front-loaded vs. bak-loaded value flows invention patents vs. utility models in China Hypotheses Empirial Strategy and Data Results Comparison of Patent Quality: invention patents vs. utility models Speed and Length of Patent Protetion: utility models vs. invention patents Conlusion Speed of Patent Protetion, Rate of Tehnial Knowledge Obsolesene and Optimal Patent Strategy: Evidene from Innovations Patented in the US, China and several other ountries INTRODUCTION Chinese Invention Patent vs. Utility Model Delay in Patent Pendeny Appliation, Attorney and Maintenane Costs Patent Sope and Validity Issues The Strategi Impliations of Patent Pendeny under Tehnology Progress: Explanation and Historial Example Patent Pendeny as Firm Strategy A historial example in the Eletri Lamp Industry Theory Measuring Rate of Tehnial Knowledge Obsolesene Data Desription and Summary Statistis Eonometri model and Identifiation Estimating Equations Identifiation Main Results Rate of Tehnial Knowledge Obsolesene and Patenting Choie Heterogeneity: SIPO Administration Dynamis Strategi Patenting: Appliant s Patent Portfolio Size Subsample estimations i

6 2.8 Robustness heks with SIPO-EPO patent dyads Conlusion Appendix ii

7 Chapter 1 Flow of Invention Value and Speed versus Length of Patent Protetion 1.1 INTRODUCTION The United States Patent and Trademark Offie (USPTO) reeives more than 500,000 patent appliations annually sine This ever-inreasing trend has made any new oming patent appliation an exerise of delay traditional total pendeny is about 1.5 years in 1991 and quikly inreases to more than 3 years in A reent report of USPTO onludes that this baklog ould ost the U.S. eonomy billions of dollars annually in "foregone innovation" business opportunities that fail to get off the ground due to the late arrival of patents (Rai, Graham, and Doms 2010). Responding to suh onerns, Congress has relaxed restritions on fee diversion and examiner hiring limits for the patent offie (Hedlund 2007); USPTO has undergone a series of ations in 2011 inluding the introdution of a "Fast Trak" patent examination 2 and a "First Ation Interview Pilot Program" 3 whih failitates ommuniations between examiners and appliants at no additional harge. So far, these pioneer ations have not reeived signifiant attentions from patent appliants. Drawing on the idea that the inter-temporal value patterns of different inventions are heterogeneous, we highlight one mehanism that explains the magnitude of attentions of patent appliants towards fast patent grant. We distinguish inventions with front-loaded value flows from those with bak-loaded value flows (Figure 1). An invention with a front-loaded value flow brings ommerial value to the inventor shortly after its disovery but the value is likely to maintain only for a short period of time. Examples of suh inventions inlude those in fields that feature fast tehnology and produt turnovers. In ontrast, an invention with bak-loaded value flow brings the inventor ommerial values that are 1 Traditionally, patent examination pendeny is measured as the average number of months from the patent appliation date to the date the appliation has reahed its final disposition (e.g., issued as a patent or abandoned). The statistis represent traditional patent pendeny inluding patent appliations that have requested for ontinued examinations. 2 The "Fast Trak" examination program allows an appliant, who is willing to pay an additional fee ($4,950 for large entities and $2,550 for small entities) to request for prioritized examination that guarantees a final deision within twelve months of the filing date. 3 The "First Ation Interview" pilot program permits an appliant to ondut an interview with the examiner after reviewing a pre-interview ommuniation providing the results of a prior art searh ondued by the examiner to at least partially failitate possible early allowane. 1

8 heavily weighted towards the later years of the invention. New drug inventions, for instane, whose ommerialization often requires a lengthy period of preparation for FDA approval, fall into this ategory. For inventions that differ in the inter-temporal patterns of value flows, one would expet that inventors have different preferene for speed and length of patent protetion. We hypothesize that inventors with expeted front-loaded inventions will prefer fast patent grant but do not require long patent protetion in order to appropriate the invention s high ommerial value in early periods. In ontrast, inventors with expeted bak-loaded inventions might not onsider speed of patent grant to be utterly important but find the length of protetion to be ruial. We study a small set of USPTO utility patents for whih appliants behaviors reveal both their inventions inter-temporal patterns of value flows as well as their preferenes for grant speed and length of protetion. The study involves USPTO patents with Chinese priorities, i.e. U.S. patents that are first sought for patent protetion in China then in the United States. The inter-temporal patterns of value flow are roughly predited through different inventors hoies between two types of Chinese patents, namely, invention patent and utility model. The differenes between these patent protetions involve a differene in the examination proedure 4 and a differene in protetion length (see more details in the next setion). In short, the grant of a Chinese utility model requires a short examination proess and provides a short protetion period (short&short). In ontrast, the pursuane of a Chinese invention patent entails a muh longer examination proess and in reward, yields a longer protetion period (long&long). Beside the tradeoff between grant speed and protetion length, approval of a Chinese utility model issuane is also signifiantly easier due to the simpliity of the examination proess. Therefore, the tehnial quality of the inventions proteted by the Chinese utility models is more questionable. However, we are able to ontrol the tehnial quality of inventions beause the idential materials are also granted USPTO utility patents. Sine the appliant has the freedom to hoose between these two types of patent protetion, we are able to infer, based upon their unique hoies, the rough shape of their invention s expeted inter-temporal patterns. Patentees preferenes for speed of grant are measured by their observed behaviors while proessing and managing their own patent appliations using the Patent Appliation Information Retrieval (PAIR) database provided by USPTO. Patentees preferenes for length of protetion are refleted by their renewal behaviors after their patents are granted. Our empirial results support the two hypotheses: USPTO patents with Chinese utility model priority are more likely to be filed and granted faster ompared to USPTO patents with Chinese invention patent, suggesting appliants preferene for fast patent grant. USPTO patents with Chinese utility model priority are also less likely to be renewed after grant, indiating a shorter value horizon of the inventions. 4 The State Intelletual Property Offie of China provides two examination proedures sequentially for eah new invention patent and utility model appliation. The preliminary examination, whih heks only formality issues, is requred for both types of appliation. After this proedure is finished, a substantial examination, whih heks for novelty and inventiveness (similar to the USPTO non-obviousness), is required only for invention patent. This last step often takes years to finish 2

9 The insights that the heterogeneity of inter-temporal patterns of invention value flows ontribute to the broad literature on the optimal design of patent system (see, e.g., O donoghue, Sothmer, and Thisse 1998; Gallini and Sothmer 2002; Hall and Harhoff 2012). In partiular, these results lend empirial support for the trend of major patent offies towards offering "aelerated" and "deferred" examination options. Our paper also makes a methodologial ontribution to the literature on patent evaluation. The existing literature, often assuming that a patent s private value deays over time, has often used patent renewals as indiators of private patent value. In our sample, however, the USPTO patents with the Chinese utility model priority are less likely to be renewed, not neessarily beause they are less valuable (they are suffiiently valuable to justify the substantial expense of filing overseas at USPTO); rather, they are abandoned sooner beause they have front-loaded value flows of whih the inventions ommerial value are high in the early years, but delines quikly to a level at whih marginal expeted returns are insuffiient to justify further renewal expenditure. The paper also ontributes to a growing researh lien that explores appliant behavior regarding equivalent inventions in different patent systems. This line of researh, as does our paper, aims to gain better understanding of firm patenting strategies as well as institutional differenes aross patent systems. The remaining of this artile is organized as follows. Setion 2 presents a brief omparison of the poliy designs for Chinese invention patent and utility model. Setion 3 summarizes hypotheses about the determinants of patenting strategies. Setion 4 desribes the onstrution of the dataset and our empirial strategy. In Setion 5 we desribe the results of our empirial estimations. Setion 6 onludes. 1.2 Coneptual Issues Inventions with front-loaded vs. bak-loaded value flows Figure 1 illustrates the value flow over time, disounted bak to the time of disovery, for a hypothetial invention that is front-loaded in the value flow and a hypothetial invention with a bak-loaded value flow, respetively. The two inventions are equally valuable in the sense that the present value of their value flows (i.e. the areas underneath the two urves) is same. However, the inter-temporal patterns of their value flows are different. Invention A is front-loaded in its value flow. An invention with a front-loaded value flow brings ommerial value to the inventor shortly after its disovery but the value is likely to maintain only for a short period due to fast tehnology upgrading. Examples of suh inventions inlude new Eletronis and Mehanis produts. Within these fields, firms seek patents for new inventions that represent improvement or add-ons of existing produts. These inventions, whih normally an be ommerialized fair quikly, an yield great benefit if patent an be effetive in time and shield the patentees from market ompetition. Their ommerial values, however, ould depreiate quikly one they are superseded by the immediate future generation of produts. In addition, new inventions that suits the purposes suh as attrating investment, meeting venture apital fundõs milestone date or inreasing firm reputation, are also likely to be filed and pursued for fast patent grant. 5 In the above examples, 5 See Cohen, Nelson, and Walsh (2000) and Graham et al. (2009). Also, Gans, Hsu, and Stern (2008) find that patent allowane signifiantly inreases the hazard rate for liensing of inventions for start-ups. 3

10 not only do patent appliants pay espeial attentions to how quikly they are able to get their patents granted, but they also are less about the length of patent protetion. 6 In ontrast, invention B is bak-loaded in value-flow. The ommerial value is small in early years and beomes high in later periods. The ommerialization of new drugs, whih often requires many years of linial trial after the disovery of the invention, falls into this ategory. For these inventions, firms plae little value on the speed of patent grant but high value on the length of protetion invention patents vs. utility models in China In China, the State Intelletual Property Offie (SIPO) offers patent appliants the freedom to hoose between two types of property right protetions for their produt innovations. The allowane of a Chinese invention patent requires substantial examination of novelty, inventiveness (similar to the non-obviousness standard required for granting a utility patent at USPTO) and pratiability; grant of an invention patent onfers a statutory patent life of 20 years. In ontrast, grant of a Chinese utility model offers a statutory life of 10 years; it is granted quikly with almost no unertainty of rejetion, as long as the appliant pays a (lower) filing fee and omplies with the filing requirements. For the same invention, an appliant an simultaneously file a "dual appliation" of both an invention patent and a utility model at SIPO. If both appliations are granted (usually at different dates), the appliant must hoose one property protetion and abandon the other immediately. 8 The Chinese utility model, the grant of whih requires no substantial examination of patentability, are often viewed as "petty inventions" that would not be qualified for the patentability standard of the invention patent. However, the utility model offers an advantage to patent appliants: speedy patent allowane in about 14 months, as ompared to an average of 54 months for invention patent appliations. If the patentee wishes to establish the validity of a utility model, she an ex-post obtain a substantial examination report from the SIPO. Table 1 presents detailed omparison of provisional differenes between the Chinese invention patent and the Chinese utility model. Figure 2 illustrates the average grant delays of Chinese invention patent and Chinese utility model that were granted during the period The invention patents were granted on average 1621 days or 54 months after appliation (with a minimum of 732 days and a maximum of 2148 days). The utility models, on the other hand, have an average grant lag of 434 days or 14 months (with a minimum of only 199 days and a maximum of 599 days). The average grant lags of invention patent inreased signifiantly during the period and dereased afterwards; whereas the average grant lags of utility models have been stable. The utility model offers a onsistently faster time line for obtaining patent protetion. 6 Jeff Bezos, the founder of Amazon.om and owner of the "one-lik" patent, even proposes that "business and software patents should have a muh shorter lifespan... (of) 3 to 5 years of patent protetion." 7 The Hath-Waxman At offers ertain drug patents an additional 30 months of patent protetion beyond the original 20 years. 8 See e.g. Patent Law of the People s Republi of China

11 1.3 Hypotheses A granted patent provides a temporary period of exlusivity. However, inventors do not have the legal authority to stop infringement behaviors immediately after their patent appliation. In the United States, the onditions for obtaining royalty damage for infringement behaviors during the pre-grant period 9 are hallenging modifying patent laims raises legal unertainty in enforing patent right even if the appliant has onfidene in getting the patent issued eventually. 10 An invention with a "frontloaded" inter-temporal value flow rewards the inventor with high ommerial value in early periods an interval during whih pending patent right is unable to onvey the omplete legal authority that the inventor requires to establish exlusivity by shielding off ompetition. On the other hand, if an invention has a "bak-loaded" inter-temporal value pattern, aquiring early patent protetion is of less importane beause the ommerial value the invention brings to the inventor is heavily weighted on the later periods of the invention. In this ase, the inventor an exploit the advantages of market exlusivity through lengthening the assoiated patent period. Sine the Chinese utility model yields a fast effetive patent right and a short protetion, we roughly ategorize eah patented invention into a "front-loaded" group and a "bak-loaded" group by identifying the assoiate patent appliants hoie between Chinese invention patent and Chinese utility model, i.e. an invention that is proteted under a Chinese utility model is interpreted as having a "front-loaded" inter-temporal value flow and vie versa. As appliants of eah Chinese patent appliation file a U.S. patent appliation for the idential invention, we are able to exploit the variation of the inter-temporal value patterns and the U.S. patent appliantsõ patenting strategies. More speifially, we propose two hypotheses to test our theory: Hypothesis 1: Appliants that have hosen Chinese utility model should are more about the speed of patent grant in filing their U.S. ounterparts, relative to those that have hosen Chinese invention patent. Hypothesis 2: Appliations that have hosen Chinese invention patent should are more about the length of patent protetion in managing their U.S. ounterparts, relative to those that have hosen Chinese invention patent. 1.4 Empirial Strategy and Data To test the above two hypotheses, we employ a small set of USPTO patents with Chinese priority. Besides the advantage of measuring the inter-temporal patterns of eah invention through the hoie 9 In both China and the United States, effetive patent right starts from the patent publiation date. Sine the publiation date is normally 18 months (or less) from the appliation date and the patent grant (or patent allowane) date is muh later, there is a pre-grant pending period that an appliant will be unable to fully exploit the legal authority of her patent right. 10 Aording to 35 U.S.C. d 154, the right to obtain reasonable royalty damages for ertain infringing ativities that our before the patent s date of issuane requires (1) the infringing ativities our after the publiation of the patent appliation, (2) the patented laims are substantially idential to the laims in the published appliation, and (3) the infringer had "atual notie" of the published patent appliation. 5

12 of either a Chinese invention patent or a Chinese utility model, two other key advantages are assoiated with this data seletion. First, we are able to observe USPTO appliants behaviors during their patent proseution proesses and infer, based on the hoies made, their attitudes towards fast patent grant. The inferred preferene for speedy patent grant is in relative, instead of absolute, terms. For example, for two USPTO patent appliations that are filed in the same year and the same tehnology field, if the inventors for the first appliation spend less time negotiating with examiners or file the USPTO appliation earlier during the priority year 11 ompared to the inventors of the seond appliation, we interpret the first set of inventors to have a stronger tendeny for early patent grant. Seond, as mentioned in the previous setion, there is a substantial differene between the examination proesses assoiated with the Chinese utility model appliations and the Chinese invention patent appliations, resulting in a wedge of average tehnial quality of the inventions proteted under the two types of property rights. More speifially, besides the advantage of a quik grant, appliants might be inentivized to file Chinese utility model simply beause they expet an inability to aquire a Chinese invention patent for their low tehnial quality inventions. If the pursuit for Chinese utility model is primarily driven by the onern of approval ertainty, then the hoie of utility model is not a valid proxy variable for an invention s inter-temporal value pattern. In our dataset, this onern is mitigated beause for eah invention that is proteted with a Chinese utility model, the idential invention is also proteted with a USPTO patent. Sine USPTO offers a uniform patent examination proedure that is presumably at least as rigorous as the Chinese invention patent examination standard, the hoie between a Chinese utility model and Chinese invention patent should mainly reflet the differene in the inter-temporal patterns of the invention. For eah USPTO patent that is granted between Jan 1 st 1993 and De 31 st 2010, we use the "patent priority" information to retrieve eah inventionõs foreign patenting history and selet a subsample that have been applied for patent protetion in China. Together, we have obtained 4101 USPTO patents with Chinese priority (heneforth USPTO-SIPO patent). We then math this dataset with the Chinese patent dataset published by the State Intelletual Patent Offie of China (SIPO) to extrat relevant information on the Chinese patents. Both proess and produt inventions are eligible to be proteted under Chinese invention patent while only produt inventions might be proteted with Chinese utility model. We manually identify proess inventions by reading the title and abstrat of eah patent in the sample, resulting in a drop of 441 USPTO-SIPO proess patents. Finally, there are 76 USPTO-SIPO patents that are Òdual appliedó for both an invention patent and a utility model in China. 12 We drop these observations due to their inonsisteny with our definition of inter-temporal patterns. Our final sample inludes 3584 USPTO-SIPO patents, 2920 of whih are proteted under invention patents in China while the remaining has utility model property right. Table 2 presents the perentages of USPTO-SIPO patents with Chinese utility model priority aross six broadly defined tehnology areas as in Hall el al These statistis show differentiated usefulness of Chinese utility modelñin teh-field of Eletrial&Eletroni, Mehanials and Others, where we expet inventions tend 11 At the USPTO, the priority year refers to the 12 months grae period prior to a patent s filing date that allows the appliant to laim priority of the urrent appliation to a previous appliation of the same invention. Claiming priority has the advantage of starting the effetive patent as of the filing date of the previous patent, onditioning on grant. 12 We identify "dual appliations" by searhing the entire Chinese patent dataset and looking for other appliations that have similar abstrat and idential assignees and inventors. The titles fo a Chinese invention patent and a Chinese utility model in a "dual appliation" are often not idential 6

13 to be "front-loaded" in value flow, one third to one half (33%, 37.5% and 69.1%, respetively) of appliations hoose Chinese utility models. 13 In sharp ontrast, few Chinese utility models are seen in teh-fields of Chemials or Drugs&Medials (3.1% and 3.6%) where we tend to believe the inventions have "bak-loaded" value flows. However, one puzzling result appears in the teh-field Computer&Communiation, in whih we expet a high perentage of utility models but we only see very little (6%). To measure an USPTO appliant s attitude towards speed of patent grant, we utilize detailed data on time lags between different events that happen during and up to one year prior to the USPTO appliation proess. Sine the appliants have laimed priority of their USPTO patent to a prior Chinese patent, the USPTO patent has an effetive starting date from the Chinese filing date upon approval of grant (footnote 14). So USPTO appliants an shorten their patent grant delay through filing their USPTO patent appliations earlier during the priority year and/or reduing the atual USPTO examination delay. Consistent with this idea, we build metris that represent deompositions of the total USPTO grant lag. More speifially, we first use the lags between the filing date at SIPO and the filing date(s) at USPTO as measure of how quikly an appliant files a USPTO appliation. We ompute three filing lags: (1) Total filing lag, the time lag between the Chinese filing date and the last filing date at the USPTO; (2) Filing lag, the lag between the Chinese filing date and the first USPTO filing date or the date of entering the U.S. national stage for a PCT filing; (3) Continuation lag, the lag between the first USPTO filing date and the last USPTO filing date if the U.S. patent appliation has some prior history of ontinuation (and 0 otherwise). Seond, to measure USPTO appliantsõ tendeny of filing patent appliations early during the priority year, we define a dummy variable, Filing lastminute, whih takes the value of 1 if the USPTO filing is within the last 10 days of the grae period. 14 Third, we use a dummy variable, Continuation, to indiate whether the USPTO appliant files ontinuing patent appliations subsequent to his urrent patent appliation. Finally, we look at the Grant lag of the USPTO patent, defined as the time delay between the last USPTO filing date to the final allowane date. The measures Filing lag, ontinuation lag and Grant lag are three disjoint time intervals the union of whih spams the total patent approval delay from the starting date of the priority year to the date of the USPTO-SIPO patentõs USPTO allowane date. If the appliant ares about the speed of patent protetion, he should file at USPTO sooner, be less likely to file ontinuations and if he files ontinuations, file them sooner. When the expeted ommerial value of an invention beomes low enough, the assoiated patent(s) is also likely to lose its private value to the patentee. We therefore use the USPTO patent maintenane status information to measure how long a patent owner maintains his patent right. In partiular, we use dummy variables Maintain 4, Maintain 8 and Maintain 12 to indiate whether a USPTO patent has been maintained after 4 th year, 8 th year (onditional on a 4 th year renewal) and 12 th year (onditional on a 8 th year renewal) after grant, respetively. We expet that the probability of maintaining patents for 4 years, 8 years and 12 years should differ between USPTO-SIPO patents with "front-loaded" value flow and "bak-loaded" value flow. 13 The teh-field Others inludes Agriultural, Amusement Devies, Apparel&Textile, Furniture, Reeptale et. whih we also expet the invention to have "front-loaded" value patterns. 14 We test various ut-offs inluding last month, last ten days and last day, the results are qualitatively the same. 7

14 1.5 Results Comparison of Patent Quality: invention patents vs. utility models First, we hek whether there is any differene in patent quality between U.S. patents with Chinese invention patent and utility model priorities. We look at five indiators for patent quality: number of laims (Lanjouw and Shankerman 2004), number of patent lassifiations (Lerner 1994) 15, number of bakward and forward itations (Hall, Jaffe, and Trajtenberg 2001; Harhoff, Sherer, and Vopel 2003) 16 and number of inventors. Speifially, for a U.S. patent i in tehnology field j with USPTO filing year t, we estimate the following equation: Y ijt = α 0 + α 1 UM i + u j + v t + ɛ ijt (1.1) where Y ijt represents one of the five patent quality indiators above. The key regressor is UM i, a binary variable that is equal to one if the patent has a Chinese utility model priority. We also inludes tehnology and U.S. filing year fixed effets, u j and v t. Results from OLS regressions are presented in Table 3. Controlling for tehnology fields and USPTO filing year fixed effets, there seems to be no signifiant differenes between these two groups, exept that patents with Chinese invention patent priority have more inventors (at the 5% level). The results lend further support to the laim that Chinese utility models in the sample are likely to quality for invention patent at SIPO Speed and Length of Patent Protetion: utility models vs. invention patents We test whether speedy patent protetion is more important, and length of protetion is less important, for U.S. patents with Chinese utility model priorities. As shown in Table 4, U.S. patents with Chinese utility model priorities have signifiantly shorter filing lags between SIPO filings and USPTO filings, are less likely to delay filing abroad until the end of the grae period, are less likely to file ontinuations, and have a shorter grant lag at USPTO 17. Table 4 also shows that U.S. patents with Chinese utility model priorities have a short value horizon: they are less likely to be maintained by patentees by the end of the 4 th, 8 th and 12 th years, respetively, after grant. To formally examine the two hypotheses, we run the following eonometri speifiation: Y ijt = α 0 + α 1 UM i + α 2 X i + u j + v t + ɛ ijt (1.2) where Y ijt represents the indiators for demand for speed and length patenting strategies desribed above for U.S. patent i in tehnology field j with USPTO filing year t. We use OLS if the outome 15 Lerner 1994 uses number of international patent lassifiaionts (IPC). We use the number of U.S. patent lassifiations (USPC). 16 An average U.S. patent in our sample has 13 laims, 2.78 inventors, 2.26 U.S. patent lassifiations, 12 ited referenes and reeives 2.83 itations. 17 Given that all these metris are at best noisy measures of appliants eagerness to seure patent protetion, it is likely that the observe signifiant differenes underestimate the atual differene in the fous on speed protetion between the two groups of patent appliations 8

15 variable is ontinuous and Logit if the outome is a binary variable. The key variable of interest is UM t. X i inlude patent quality ontrol variables. Tehnology and U.S filing year fixed effets, u j and v t, are also inluded. Table 5 presents the results. The lag between Chinese filing dates and the last U.S. filing dates, Total Filing Lag, is 81 days shorter for U.S. patents with Chinese utility model priorities. This differene reflet variations both aross and within tehnology fields. If we ontrol for tehnology, u j, the differene drops to 30 days, but is still signifiant at the 5% level. 18 We then deompose Total Filing Lag into two omponents: Filing Lag and Continuation Lag. The differene in Filing Lag, the delay between the SIPO filing date and the fist USPTo filing date or the date of the PCT filing, is signifiant, averaging 35 days earlier for utility models. The differene beomes smaller and insignifiant, though with the same sign after ontrolling for tehnology fields. Under our hypotheses, appliants want to file at USPTO sooner for inventions with Chinese utility model priorities. However, it is oneivable that it takes more time to onfer a Chinese utility model appliation to a U.S appellation than Convert a Chinese invention patent appliation. This is one possible reason why the differene in Filing Lag is less signifiant when we add in u j. Continuation Lag, the time lag between the first and the last filing date at USPTO, is also signifiantly shorter for utility models: 48 days and 27 days with or without ontrolling for u j, respetively. 19 Panel B of Table 5 shows that U.S. patents with Chinese utility models priorities are 12 perentage points less likely to be filed at USPTO in the last 10 yeas of the grae period, and3 perentage points less likely to be have ontinuations filed at USPTO. Their grant lags are 157 days shorter as well. Altogether, these results support our first hypothesis that appliane who desire a fast patent protetion tend to file utility models in China. In Panel C of Table 5, we test the seond hypothesis that inventions filed for utility models at SIPO have a shorter value horizon. U.S. patents with Chinese utility model priorities are signifiantly less likely to be renewed at the 4 th year after grant. 20 U.S. patents with Chinese utility model priorities that are renewed at the 4 th (4 th and 8 th ) year are also less likely to be maintained at the 8 th (12 th ) year. The point estimates for Maintain 8 and Maintain 12 are negative and larger in absolute terms than those for Maintain 4. They are less signifiant when ontrolling for tehnologies, partly beause of the redution in sample size. Overall, these results suggest that appliants tend to protet these inventions with relatively short value horizons for relatively short periods. 18 The drop is expeted, as now the differene is only due to variations within eah tehnology. 19 If we, instead, fous on the subs maple of U.S. patents that filed ontinuations at USPTO, the results are muh larger (not shown here): patents with Chinese utility model priorities have ontinuation lags 180 days less than those with invention patent priorities in the same tehnology fields (signifiant at 1% level). 20 For the USPTO-SIPO patents in our sample, the 4 th year renewal deision at USPTO tends to ome towards the late stage of the 10-year life of a Chinese utility model. This is beause in our sample, the average lag between SIPO filing dates and USPTO grant dates is about 3.8 years. So a U.S. patent s 4 th year renewal is about years after SIPO filing, towards the late stage of a utility model s 10 year patent term. 9

16 1.6 Conlusion The United States Patent and Trademark Offie onfronts a dilemma on the one hand, substantial attention has been direted towards shortening the patent examination proess, inluding adding a "Patent Proseution Highway" and a "First Ation Interview Pilot Program." On the other hand, only a few patent appliants are using the "Patent Proseution Highway" to expedite their patent appliations. 21 Sine the launh of the "Fast trak" option, a lot of debate has been foused on explaining the apparent low take-up rate. Many suggestions are offered suh as the inertia to adaption and ost onsideration. Given this senario, we are interested in understanding what kind of patent appliants would be interested in getting early patent grant. An obvious explanation is tehnologyñaverage patent appliants in short-yle tehnologies would naturally want earlier patent grant ompared to appliants in long-yle tehnologies. However, sine inventions are extremely heterogeneous, intra-tehnology omparison is useful, but annot explain within-tehnology heterogeneity in preferenes for speed of patenting, as shown in the empirial results. This is our purpose of writing the paper. Among all other plausible explanations, we resort to a lassi patenting deision model (to maximize the inventionõs monopoly profit within the effetive term of the patent right) to explain the heterogeneity in terms of appliantsõ preferene towards fast patent grantñthe variation in the inter-temporal value flow of the inventions. We use a small sample of USPTO patents that were first filed for Chinese patent protetion. We investigate the linkage between the hoie of Chinese utility models versus invention patents at SIPO (whih indiates whether inventions are Òbak-loadedÓ or Òfront-loadedÓ in value flows) and the subsequent filings and renewal deisions at USPTO. The results suggest that inventions differ in the inter-temporal patterns of value flows, and onsequently, appliants differ in their preferenes over speed versus length of patent protetion and opt for the appropriate type of protetion that best suit their needs and invention harateristis. Chinese inventors are willing to opt for shorter but quiker protetion afforded by utility models even for inventions of substantial value on the world market that is Òfront-loadedÓ in value flows. The results are robust after ontrolling for tehnology fixed effets. The notion that there exist heterogeneous temporal patterns in value flows among inventions ould provide an interesting and useful oneptual framework for the disussions on the optimal design of patent systems. In partiular, sholars have argued that a uniform patent system, suh as the one in the U.S., might not aommodate the heterogeneous appliant needs and reate potential distortion in firm inentives in innovation (Cornelli and Shankerman, 1998; O Donoghue, 2004). In a reent study, Budish, Roin, and Williams (2014) shows that the lags between the disovery and ommerialization of drug innovations vary aross aner types and render, among drug inventions that are generally "bak-loaded" in value flows, some to be more "bak-loaded" than others. Thus, a fixed patent term ould distort firm R&D inentives in aner drugs. 21 There is less than 1% of USPTO patent appliants that use the "Patent Proseution Highway" to expedite their patent appliation during

17 Figure 1.1: front-loaded Value Flow vs. Bak-loaded Value flow for different inventions note: the above graph shows the inter-temporal patterns of value flows for two hypotheti inventions. The blue line represents an invention with "front-loaded" value flow and the red line represents an invention with "bak-loaded" value flow. 11

18 Figure 1.2: invention patent vs. utility models: delay in patent grant (days), note: Mean grant lags for invention patents and utility models are estimated, using SIPO patent dataset whih ontains all Chinese patents with filing dates between The ost of patent appliation and renewal are obtained from SIPO website at 12

19 Figure 1.3: Perentage of U.S. patents with Chinese utility model Priority: U.S. appliation date note: perentages of U.S. patents with Chinese utility model patents are estimated, using SIPO-USPTO patent dataset whih ontains all Chinese patents with filing dates between and the orresponding U.S. patent appliations with filing dates between

20 Table 1.1: invention patents vs. utility models in China invention patent utility model Any new tehnial solution Any new tehnial solution Definition or improvement relating to or improvement relating to a produt or a proess. the shape, struture or their ombination of a produt. Subjet Matter proess and produt innovations produt innovations Patentability substantial examination of No substantial examination novelty, non-obviousness and utility. of novelty and non-obviousness. Average Grant Lag 54 months 14 months Term 20 years 10 years Appliation (YMB): Examination (YMB): 2500 N/A Attorney fee (YMB): , 1 st 3 rd years; 1200, 4 th 6 th years; st 3 rd years Maintenane Fees (YMB) 2000, 7 th 9 th years; th 5 th years (annual renewal shedule) 4000, 10 th 12 th years; th 8 rth years 6000, 13 th 15 th years; th 10 th years 8000, 16 th 20 th years; Note: Mean grant lags for invention patents and utility models are estimated, using SIPO patent dataset whih ontains all Chinese patents with filing dates between The ost of patent appliation and renewal are obtained from SIPO website at For attorney fees, we interviewed several lawyers from different law firms loated in Beijing, China. We asked for the attorney fees they harge for invention patents and utility models, respetively. In general, law firms harge the same rate regardless of the loations of their lients. 14

21 Table 1.2: Distribution of Chinese invention patent and utility model Patent by Tehnology fields Tehnology Fields invention patent utility model Dual % of U Chemials Computer & Communiation Drugs & Medial Eletris & Eletronis Mehanials Others Total note: The six tehnology fields are defined as in Hall, Jaffe, and Trajtenberg 2001, based on U.S. patent lassifiations. Sample inludes SIPO-USPTO patent dyads with Chinese priorities between

22 Table 1.3: Attributes of U.S. patents: Chinese invention patent priorities vs. Chinese utility model priorities No. of Claims No. of Inventors No. of USPC No. of referenes made No. of referenes reeived (1) (2) (3) (4) (5) UM i (0.50) (0.11)*** (0.25) (0.57) (0.26) U.S. appliation year yes yes yes yes yes U.S. tehnology field yes yes yes yes yes N R Notes: Patent-Level Observation. All estimates are from OLS models. Sample inludes all USPTO (The United States Patent and Trademark Offie) patent (appliations) with Chinese Priority from ohort. Heterogenous Robust standard errors are shown in parentheses. *: p<0.10; **: p<0.05; ***: p<0.01. Dependent variables are number of laims, inventors, U.S. patent lassifiation, referenes made and referenes reeived. UM i is a dummy variable indiating whether the U.S. patent has a Chinese utility model priority. Appliation Year Dummies: 0/1 indiator variable for appliation year. Tehnology Field Dummies: 0/1 indiator variables for tehnology fields (total of 6 tehnology lassifiations following Hall, Jaffe, and Trajtenberg 2001). 16

23 Table 1.4: Desriptive Statisti of U.S. Patents with Chinese Priorities: invention patent vs. utility model utility model invention patent differene (mean) (mean) (UM-IPat) (A) Indiators for eagerness to seure patent protetion Total Filing Lag *** Continuation Lag *** Continuation *** Filing Last Minute *** Grant Lag *** (B) Indiators for demand for length of patent protetion Maintain *** Maintain *** Maintain *** Note: Unites of Total Filing Lag Continuation Lag, Grant Lag are days; Continuation, Filing Last Minute and Maintain 4 Maintain 8 Maintain 12 are binary indiators. * Signifiant at 10%; ** signifiant at 5%; *** signifiant at %. 17

24 Table 1.5: Regression results: U.S. patent with Chinese invention patent priority vs. Chinese utility model priority (1) (2) (3) (4) (5) (6) Panel A Total Filing Lag Total Filing Lag Filing Lag Filing Lag Continuation Lag Continuation Lag UM i (33.372)** (15.275)** (16.592)** (14.778) (19.644)** (10.573)** N R Panel B Filing Last Minute Filing Last Minute Continuation Continuation Grant Lag Grant Lag UM i (0.042)*** (0.028)* (0.014)** (0.039)*** (49.623)** (32.092)** N P seudo R Panel C Maintain 4 Maintain 4 Maintain 8 Maintain 8 Maintain 12 Maintain 12 UM i (0.03)*** (0.029)** (0.08)** (0.061) (0.083)** (0.08) N P seudo R U.S. appliation year yes yes yes yes yes yes U.S. tehnology field no yes no yes no yes Notes: Patent-Level Observation. Panel A estimates are from OLS models; Panel B and C estimates are from Logit models. Sample inludes all USPTO (The United States Patent and Trademark Offie) patent (appliations) with Chinese Priority from ohort. Heterogenous Robust standard errors lustered at firm level are shown in parentheses. *: p<0.10; **: p<0.05; ***: p<0.01. UM i is a dummy variable indiating whether the U.S. patent has a Chinese utility model priority. In olumns (1) and (2) of panel C, the dependent variable is a binary indiator for whether a U.S. patent is minted at the 4 th year after grant, and the sample inludes patents whose U.S. issue years are no later than In olumns (3) and (4) of panel C, the dependent variables is a binary indiator for whether a U.S. patent is maintained at the 8 th year after grant, onditioning on the patent has been renewal after 4 years of grant. The sample inludes patents whose U.S. issue years are no later than In olumns (5) and (6) of panel C, the dependent variables is a binary indiator for whether a U.S. patent is maintained at the 12 th year after grant, onditioning on the patent has been renewal after 8 years of grant. The sample inludes patents whose U.S. issue years are no later than Appliation Year Dummies: 0/1 indiator variable for appliation year. Tehnology Field Dummies: 0/1 indiator variables for tehnology fields (total of 6 tehnology lassifiations following Hall, Jaffe, and Trajtenberg 2001). The results from speifiations with sub-tehnology fields fixed effets (36 total), not shown here, are similar. 18

25 Chapter 2 Speed of Patent Protetion, Rate of Tehnial Knowledge Obsolesene and Optimal Patent Strategy: Evidene from Innovations Patented in the US, China and several other ountries 2.1 INTRODUCTION A rih set of eonomi literature investigates the strategi impliations of the patent regime for firm-level patenting behaviors (Levin et al. 1987, Cohen, Nelson, and Walsh 2000, Hall and Ziedonis 2001). Strengthening appropriability onditions 1 (suh as inreasing duration and sope of patent protetion) is likely to have both positive and negative influenes on the private returns to firms patents. 2 An interesting yet somehow underexplored strand of literature is to onsider how firms address the benefits and osts of given appropriability onditions (Ahuja, Lampert, and Tandon 2008). In theory, given that firms are strategi players, one would expet that firm strategies would try to utilize the advantage of appropriability onditions to maximize the private returns to their patents. These private returns to patent do not, however, depend only on the institutional design of the patent regime. New disoveries and improvements in tehnology might alter the value of a firm s existing and future innovations whih, in turn, imply adjustments of its patenting behaviors. This paper provides empirial evidene that a firm s patenting strategy is determined by the rate of tehnial knowledge obsolesene embedded in patents. I quantify the effet of this knowledge obsolesene on Chinese firms patenting deisions aross more than 400 distint tehnology fields during , a period during 1 Appropriability onditions refer to the environmental fators, apart from firm and market struture, that enable an innovator to apture the rents of innovation (Teee 1986). The patent system is the most widespread and ommonly studied appropriability mehanism in the ontext of innovation. Other mehanisms inlude serey, lead-time advantage, trademark, opyright et. In this paper, I only fous on the appropriability impliations of a patent regime. 2 A strong patent regime alleviates the onern of expropriation by rivals and affords the innovative firms with an opportunity to reover investments in researh and development (R&D) (Arrow 1962, Anton and Yao 1994); it might also disourage researh beause strong patent rights make it more diffiult to "invent around" a prior innovation (Gallini 1992), disouraging follow-on researh. O donoghue, Sothmer, and Thisse (1998) introdued the onept of leading breadth and lagging breadth and analyze the optimal design of a patent regime for sequential innovations. 19

26 whih no major shifts in the patent regime took plae in China. My results indiate that, when new tehnial knowledge supersedes existing tehnial knowledge at a faster pae, firms are more willing to seure early patent grants for their innovations. I show that these patterns from the data are onsistent with a firm-level return-topatent maximization problem used ommonly in the innovation literature (Nordhaus 1969; Gilbert and Shapiro 1990; Gallini 1992), with one additional assumption: the return to a patent depends on how quikly the urrent tehnologial knowledge beomes obsolesent, whih is treated as exogenous to the firm and haraterized by a onstant depreiation fator over time. A firm forms expetations about the impat of rate of knowledge obsolesene on the private returns to its own patents based on its urrent stok of tehnial knowledge and adjusts its patenting strategies aording to the following rule: given that the firm s stok of tehnial knowledge is fixed, higher rates of tehnologial knowledge obsolesene shift the reward to a patent towards the early periods of patent life, induing the firm to seure early patent grants in order to maximize the benefits of the patent right. Delay in patent protetion has been a universal problem for big patent offies suh as the United States Patent and Trademark Offie (USPTO), the European Patent offie (EPO) and the State Intelletual Patent Offie of China (SIPO). A reent report of USPTO onludes that patent pendeny ould ost the US eonomy billions of dollars annually in "forgone innovation" business opportunities that fail to get off the ground due to the late arrival of patents. 3 There has been a long history of major patent offies offering a uniform patent appliation proess. 4 The institutional design of the patent regime in the US and Europe, therefore, prevents us from understanding the determinants of firms demand for speed of patent protetion. To address this issue, I utilize a poliy design provided by the State Intelletual Patent Offie of China (SIPO), whih provides two types of patent protetion for produt innovation, namely, the invention patent and the utility model. Utility model protetion is granted muh faster than invention patent protetion (an average of one year as ompared to 4.5 years) beause an appliation for utility model protetion does not require a omplete and substantial patent examination. A firm s hoie of utility model might, however, also suggest that the invention represents low tehnial quality. In order to distinguish between firms onerns for speed and quality, I arefully selet the sample to inlude inventions for whih patents are sought in both China and the United States. The United States Patent and Trademark Offie (USPTO) employs a uniform and rigorous patent examination standard, whih is presumably at least omparable to the patentability standard for the invention patent in China. I find that although firms hoose the utility model available in China for a signifiant potion of their innovations, the overall US grant rates for US patents with Chinese utility model and invention patent priorities 5 are almost idential. In other words, some firms that have innovations that qualify for full patent protetion are hoosing weaker and faster protetion. I also ondut the analysis using inventions for whih patents are sought in both China and Europe and find the results are onsistent with the results found in the China-US patent sample. 3 In the past few years (2009-), USPTO has repeatedly mentioned its grave onerns about the impat of patent pendeny on future innovation and eonomy. These oasions inlude speehes by David Kappos (former Diretor of the USPTO) during the Innovation Alliane Conferene (Jan. 2011), World IP Day (Apr. 2012), Center for Amerian Progress (Jun. 2010), a US Dept. of Commene Report on the Role of Patent Reform in Supporting Innovation and Job Creation (Apr 2010) and many others. 4 It was not until 2011 that the United States Patent Offie (USPTO) deided to undergo a shift in the patent appliation proess that introdued "fast trak" patent examination. It allows appliants, subjet to a hefty fee, to obtain a final disposal within 12 months from the filing date. The EPO, on the other hand, still offers a single route for patent appliations. 5 In both the US and Chinese patent law, a priority right is a time-limited right, triggered by the first filing of an appliation for a patent. The priority right allows for laimant to file a subsequent appliation in another ountry for the same invention effetive as of the date of filing the first appliation. When filing the subsequent appliation, the appliant must laim the priority of the first appliation in order to make use of the right of priority. 20

27 The previous literature offers no onsensus measure of the rate of tehnial knowledge obsolesene. 6 I extend Bosworth s approah to use patent renewal data to reate a proxy variable for the rate of obsolesene (Bosworth 1978). The index is based on the idea that the duration of a patent reflets the lifetime of the tehnial knowledge embedded in the patent. More speifially, when an inumbent tehnology s ompetitive advantage diminishes due to emergene of some superior tehnial knowledge, the private value of the assoiated patent will derease, ontributing to an early mortality of the patent right. Individual patent renewal deisions have been used extensively in the literature as a means to estimating private returns to patent (Pakes 1986; Shankerman and Pakes 1987; Lanjouw and Shankerman 2004 et.). Instead of fousing on the renewal deisions of the entire lifespan of a given patent ohort, I look at the perentages of ineffetive patent rights (based on renewal deisions) for a given age of patents over different ohorts and treat them as a ontinuous measure of tehnial knowledge obsolesene. My identifiation exploits within-tehnology variation of the above indies over time. A key advantage of this approah is that it mitigates onerns of omparing patenting behaviors aross different tehnology fields. 7 The empirial analysis indiates that, when development of tehnial knowledge is arried out at a faster pae, firms propensity for hoosing the utility model inreases. One standard deviation inrease in the rate of tehnial knowledge obsolesene inreases firms propensity to file for the utility model by 6%-8%. I find the sensitivity in the hoie of utility model is non-uniform aross tehnology obsolesenes: the largest effet is in fields with a tehnology index above the 67 th perentile of the distribution. I also find evidene that the sensitivity to hanges in tehnologial progress an be deomposed into a firm "entry and exit" effet and a shift in the patenting strategy of existing firms, with the magnitude of the seond effet muh larger than that of the first. I find that these within-tehnology variations in the rate of tehnologial progress reflet distint patenting strategies depending on the size of the individual firm s patent portfolio: firms with almost no stok of patents are most sensitive to the hanges in tehnology development whereas firms with a sizable stok of patent follow a more stable patenting strategy. This result is onsistent with previous literature showing that startup firms pursue patent rights for finaning and liensing onsiderations (Gans, Hsu, and Stern 2008). To onsolidate my results, I exploit exogenous variations in SIPO s administrative effiieny in examining invention patents. I find that when SIPO examines invention patents more effiiently (as represented by a shorter aggregate grant lag), there is a derease in the sensitivity in the hoie of the utility model to the speed of tehnologial hange, a result onsistent with my hypothesis that patent pendeny is an important onern for firms while filing patents. My paper is related to several different strands of literature. The analysis ontributes to a growing body of work that evaluates the distortion of innovation under a uniform patent system. Using exogenous variation in a linial trial period, Budish et al find that a fixed patent term shifts private R&D resoures towards pharmaeutial drugs with shorter ommerialization delay. O Donoghue and Zweimüller 2004 inorporate the fixed attributes of the patent system into a dynami endogenous growth model and onlude that a uniform patent system auses misalloation of resoures aross industries. From a different perspetive, this paper empirially demonstrates firms heterogenous demand for speed of patent protetion and suggests that urrent patent regimes that employ a uniform patent appliation standard might be no longer as effetive as before when tehnology is progressing rapidly. Previous eonomi and legal literature has foused on other aspets of patent poliy (e.g. patent sope, 6 Bosworth uses patent renewal to haraterize the rate of tehnial knowledge obsolesene (Bosworth 1978). Comin, Diego and Hobijn use the hanges in market share of existing tehnologies vs. new tehnologies as a measure of how quikly tehnology beomes obsolete (Comin and Hobijn 2004). Bilir uses average patent itation lag as a measure of tehnology yle (Bilir 2013). 7 As disussed extensively in the literature, the effetiveness of patents varies signifiantly aross different tehnology areas (Merges and Nelson 1990; Levin et al. 1987; Cohen, Nelson, and Walsh 2000; Burk and Lemley 2003) 21

28 length, validity et). My paper, whih examines the importane of patent pendeny in addition to patent length and sope as a publi poliy instrument in addition to length and sope, adds insight into the optimal design of patent systems. This paper also makes a methodologial ontribution to the literature on patent evaluation, by providing a novel perspetive on the use of patent renewal as an indiators of private patent value. Given that it is onsiderably more expensive for Chinese firms to file overseas at USPTO, my results are not onsistent with the previous understanding that patents are not renewed beause they are no longer valuable inventions. Rather, they are abandoned sooner beause the flow of value they bring to the patentees is more heavily weighted toward the early years of patent life. The value flow is likely to be high in the early years for these patents, but to deline quikly to a level at whih marginal expeted returns are insuffiient to justify further renewal expenditure. The paper shed lights on a growing researh line that explores patentee behavior regarding equivalent inventions in different patent system, offering insight into firm patenting strategies as well as institutional differenes aross patent systems. The paper proeeds as follow: setion 2 outlines some of the key provisional differenes of Chinese invention patent and utility model; setion 3 lays out a simple theory and a historial example in the eletrial lightning industry to elaborate our eonomi intuition; setion 4 disusses our definition of the rate of tehnial knowledge obsolesene; setion 5 desribes our data and summary statistis of all regression variables; setion 6 presents our eonometri model and identifiation strategy; setion 7 and 8 report our empriia findings and robustness heks; setion 9 onludes. 2.2 Chinese Invention Patent vs. Utility Model This setion briefly ompares some of the important provisional differenes between the Chinese invention patent and the Chinese utility model. It serves as the foundation upon whih I develop my researh design and sample onstrution Delay in Patent Pendeny The Chinese patent law was enated in 1984 and put into pratie in Two types of patent protetion for industrial produt innovation are available in China, namely the invention patent and the utility model. The invention patent is the onventional patent: the appliation will go through a substantial examination for novelty, inventiveness and pratiability. The Chinese utility model is designed following the German and Japanese utility model. This lesser-known form of IP protetion was initially designed to protet property rights in a way that is less expensive, quik and easy to obtain. Faster protetion under the utility model is ahieved, as no examination is required. As a result, the delay in patent pendeny for a utility model typially ranges from six months to a one year and a half (average 14 months), as against four to five years (average 4.5 years) for an invention patent. Figure 2 illustrates the mean grant lags of Chinese invention patent and utility models by appliation year for the period Invention patents are granted, on average, 1621 days (or 54 months) after, appliation with a minimum average of 732 days (per year) and a maximum average of 2148 days (per year). In ontrast, utility models are granted, on average, 434 days (14 months), with a minimum annual of only 199 days (per year) and a maximum annual of 599 days (per year). In addition, there is a large variation in the grant lag of the invention patents over time (std. dev: 337 days), whereas average grant lag for the utility models remains relatively stable (std. dev: 109 days). 22

29 2.2.2 Appliation, Attorney and Maintenane Costs Table 1 presents the legal provisions, and the proess of granting and maintenane of the Chinese invention patent and the Chinese utility model. Besides the advantage of a fast patent grant, the utility model is also more attrative beause it is signifiantly heaper than the invention patent. For instane, preparation of a utility model appliation through a Chinese attorney typially osts an appliant $500 (3000 rmb), whereas that for an invention patent is around $1,300 (8000 rmb), 8 an amount more than double. The differenes in maintenane osts over the first ten years after patent issue are also signifiant: the aggregate ost of renewing a utility model is about 60% the ost of renewing an invention patent for the same effetive periods. Sine the appliation for a utility model does not require examination, the appliation fee for a utility model is signifiantly lower. Overall, the total ost of applying and maintaining a utility model is around 30% of the ost of applying and maintaining an invention patent Patent Sope and Validity Issues The Chinese Patent Law does not set different patent sope standards for the invention patent and the utility model. 9 In addition, the bases for laiming damages aused by patent infringement on an invention patent and a utility model are the same. 10 However, when infringement litigation is filed for a utility model, the plaintiff is required to present an evaluation report prepared by SIPO during the proeeding as evidene supporting its validity. In China, the validity of an invention patent or a utility model is determined by the Patent Reexamination Board rather than by a ourt. An assertion of patent validity, therefore, often results in a delay at the ourt while the validity is pending. In addition, the redibility of a utility model s evaluation report is subjet to many onerns. Before 2009, SIPO personnel in harge of preparing the report is seleted from a pool of examiners who did not represent the most qualified examiners in eah tehnology field. The resoures available for prior art searh are limited to prior Chinese invention patents and utility models, whih limits the examiners ability to get aess to other soures suh as aademi journals, other online publiations and issued patents in foreign ountries. Moreover, the evaluation report is not an "ironlad" proof of the utility model s validity. That is to say, even if the report fails to find the utility model invalid, other evidene might overturn the utility model s validity. Based on the above evidene, I treat the Chinese utility model as a faster and heaper but weaker IP protetion ompared to the Chinese invention patent. 2.3 The Strategi Impliations of Patent Pendeny under Tehnology Progress: Explanation and Historial Example This setion is divided into three parts. The first part outlines our eonomi interpretation that a firm s propensity to seure early patent rights (or shorten patent pendeny) is determined by hanges in the rate of tehnial knowledge obsolesene. The seond part desribes a historial example in the eletri lamp industry that 8 Cost of attorney is based on interviews with one senior patent attorney at Tee & Howe, an IP law firm liensed by the Chinese government to represent domesti and foreign lients. 9 Aording to Chinese Patent Law (2008) Artile 11: After the patent right is granted for an invention or a utility model...no...individual may... manufature, use, sell or import the patented produts without permission of the patentee. 10 The 2000 Amendment of the Chinese Patent Law sets forth a standard for alulating infringement damages based on four alternative methods: lost profit to the patentee, unjust enrihment to the infringer, exploitation fee for the patent under ontratual liense and a statutory amount between 10, 000 rmb to 1000, 000 rmb, depending on various fators related to the harateristis of patent right and infringement. 23

30 follows the above intuition losely. The third part applies this intuition to a hoie model in whih firms an selet between the Chinese invention patent and utility model to maximize the private returns to their patents. Based on this model, we derive several theoretial impliations that will be empirially tested Patent Pendeny as Firm Strategy Under both US and Chinese patent laws, a patent right beomes effetive after the patent offie issues the patent. Conditioning on the grant, a patent owner an obtain reasonable royalty damages for infringement ativities that our after the patent publiation, an event that happens no later than 18 months from appliation (and, in most ases, muh earlier than the patent grant). 11 Assuming firms are risk neutral and aware of the probability of the patent grant, 12 firms might want to seure the patent grant early, rather than late, beause early resolution of unertainty ontributes to long-term planning (Epstein, Farhi, and Strzaleki 2013). Researh-oriented firms treat patents as their strategi assets and use them in many different ways. More speifially, firms utilize their patents (or patent portfolios) for ross-liensing negotiations, patent-pool onstrution, standard-setting organizations, preemptions, defense against litigation, finaning purposes and branding (Lemley 2000). In the above ases, an early patent grant ontributes to the gains of speifi tehnologial or market advantage. For example, shorter pendeny aelerates a firm s buildup of its patent portfolio, improving its bargaining position in ross-liensing negotiations and its defense against litigation (Cohen, Nelson, and Walsh 2000). For individual start-up firms, patent grants improve the effiieny of forming liensing ontrats with their downstream manufaturers (Gans, Hsu, and Stern 2008). These firms might also seure patent grants for the purpose of meeting venture apitalists milestones (Kortum and Lerner 2000). These studies suggest that there is an option value assoiated with having patents granted early. On the other hand, firms might strategially lengthen the patent proess beause they expet that a patent issued later will be more valuable than one issued earlier. This value might ome from athing rivals "off-guard" the ability to fore ompetitors into liensing ontrats through the pratie of "submarine patenting" (Graham and Mowrey 2004). In addition, firms might file patents without a thorough advane understanding of their ommerial values. In this ase, firms an take advantage of the patent pendeny delay to refine their laims as they figure out how the market unravels overtime (Lemley and Moore 2004). These studies suggest that firm strategies should take advantage of patent pendeny to maximize the benefits of patent rights while minimizing their profit-reduing effets. In the lassi inentive theory of patent systems, owning a patent right is treated as seuring a stream of monopoly profits, the size of whih depends on the institutional design of the patent (length, breadth, validity), the harateristis of the firm, market and tehnology. 13 In this senario, an inventor faes a maximization problem (either profit or private return to patent) with respet 11 In the US, there are several onditions that need to be satisfied before the patentee an laim retrospetive damages for infringement before a patent grant. One of the key onditions is that the patented laims need to be substantially idential to the laims in the published appliation (Title 35 of the United States Code 154). The Chinese patent law, on the other hand, also permits the patentee to laim retrospetive royalty damages (Chinese Patent Law 2008 Artile 13). Legal praties in China, however, suggest that the infringement behaviors will be based on either the laims desribed in the patent publiation or the finalized patent, depending on whih is narrower in sope in relation to the infringed laim(s). See, e.g., ourt ase on Chinese Patent No In summary, repeated negotiations between a patent appliant and the patent examiner are likely to subjet the sope of the patent right to signifiant unertainty (Merrill, Cohen, et al. 2003), making unauthorized imitation behavior eonomially attrative despite the litigation risk. 12 The US patent grant rate is around 90% when aounting for "ontinuing patent appliations" (Quillen and Webster 2001). The Chinese invention patent grant rate is around 33%. The Chinese utility model grant rate is almost one. 13 See e.g. Gallini and Sothmer for a literature review on optimal design of patent poliy. 24

31 to different approriability onditions (longer vs. shorter patent pendeny, among others). Based on the intuition from Bosworth (1978), I treat the speed of tehnologial progress as a onstant depreiation fator on the private value of the patent over time. Faster disovery of new and superior tehnial knowledge is represented as a higher degree of depreiation and shorter value horizon of the patent. This is beause, when an inumbent tehnology s ompetitive advantage diminishes due to emergene of a superior tehnology, the private value of the assoiated patent will derease and the effetive value horizon of the patent will be shortened. In this ase, it is straightforward to see that firms will have a stronger inentive to seure early patent grants if new tehnologies are emerging at a faster rate A historial example in the Eletri Lamp Industry To be more onrete, onsider for example the dramati hanges in the R&D (researh and development) efforts and patenting behaviors of General Eletri s major ompetitor, Westinghouse, before and after Thomas Edison s disovery of the inandesent lamp. Edison was the first to disover and patent an inandesent lamp with fibrous material illuminants. 14 The patent was filed in November 1879 and quikly issued in less than 3 months in the United States. 15 Later on, Edison produed inandesent lamps using an illuminant of arbonized paper, whih proved to be of huge ommerial value. The development of the tehnology, like many other eletri goods, was a umulative nature. Westinghouse, on the other hand, also had patent rights over many related tehnologial progresses subsequent to Edison s ore patent. However, Westinghouse was produing inandesent lamps that were substantially similar to the Edison lamp, a potential infringement behavior (Bright 1972). In 1891, Edison s ore patent was held valid in ourt. General Eletri quikly obtained a series of injuntions that shut down many ompetitors, inluding Westinghouse. 16 Getting aught off-guard, Westinghouse responded by speeding up its R&D and patenting effort of an older and non-infringing tehnology. 17 In addition, Westinghouse also employed a "defensive patenting" strategy by trying to obtain patent rights related to the inandesent lamp, as fast as possible and as many as possible. For instane, one of its subsidiary firms, the Consolidated Eletri Light Company, undertook to assert its right over another important inandesent lamp patent the paper illuminant patent by Sawyer and Man 18 immediately after the patent was granted. 19 Although Edison s patent helped his ompany gain a market share of around 75%, the non-infringing lamp of Westinghouse was nonetheless produed at a ommerial sale large enough to help the ompany survive until Edison s ore patent expired about 6 years later. Thereafter, Westinghouse immediately resumed prodution of Edison s lamp. Patent appliations in the late 19 entury US would be granted only after a few months. Compared to the 17 years of statutory protetion, patent pendeny was not even an issue. Nevertheless, Westinghouse s R&D and patenting strategies refleted its intention of seuring early, rather than late, patent rights, when its urrent and future profits (and private returns to patent as well) were seriously threatened by the introdution of a superior tehnology. 14 U.S. patent No. 223, Edison also filed international patent appliations for the same innovation in Britain, Canada and Frane around the same date as when he filed the U.S. patent appliation. In Britain and Frane, the patents were granted within 2 weeks. 16 General Eletri did not allow ompetitors to stay in the industry even as liensees (Bright 1972). 17 The lamp produed by Westinghouse employed the older stoppered base instead of a hermetially sealed glass globe whih maintained the vauum more steadily (Bright 1972). 18 William E. Sawyer and Albon Man, like Edison, ontributed signifiantly to the tehnial improvement and ommerialization of the inandesent eletri lighting industry. 19 The Saywer and Man paper illuminant patent was granted in 1885, 6 years before Edison s patent was held valid in ourt, but later than when Edison started the litigation. 25

32 2.3.3 Theory In this setion, I apply the above intuition to a onrete setting under the Chinese patent regime that offers both the invention patent and the utility model. I develop a model that aptures the essential trade-off between getting a patent quikly, but with weaker protetion (in terms of length and breadth), and getting a patent through a slower proess that provides stronger protetion for the same innovation. In partiular, I examine whether the relative effiieny hanges with exogenous variations in the rate of tehnial knowledge obsolesene. I will all the slow, strong patent the Tier 1 patent and the fast, weak patent the Tier 2 patent. The ost differenes between these two types of patents are exluded from the model beause they do not provide further insight into the questions at issue. A patent seures a flow of monopoly profits that depends on the strength of the patent (see Nordhaus (1969), Gilbert and Shapiro (1990), Gallini (1992) et). To highlight the importane of speed, I add in another variable that depits patent pendeny. The optimal hoie between a Tier 1 and Tier 2 patent is based on an ex-ante pre-filing profit flow omparison: beause a Tier 2 patent is granted early, it seures profits primarily in the early periods of patent life, while a Tier 1 patent seures profits in later periods. Time is ontinuous in the model. My approah formalizes the intuition that the relative effetiveness of these two types of patents is affeted by hanges in the rate of tehnial knowledge obsolesene. The model is also relevant in the presene of strategi patenting, as I will disuss in the last part of this setion. This model permits examination of the relationship of researh and innovation to the profits flowing from patented tehnologies. Speifially, I assume that firms ompete in R&D in a number of tehnology areas indexed by j, j = 1, J. Tehnology areas are haraterized by the rate of tehnial knowledge obsolesene ST D j, whih I assume is exogenous to individual firms. Suess in researh labs an later be developed into ommerially viable produt innovations that represent the highest quality among all existing horizontally differentiated produts. When researh ahieves tehnologial breakthroughs, the state of the art is pushed forward. A new innovation thus has value to its owner until the tehnology it is utilizing beomes obsolesent. When tehnology beomes obsolesent, I assume that the assoiated produt will lose its value for ustomers and that the intelletual property (invention patent or utility model), will lose its value for the owner. For simpliity, I assume the firm has already deided to seek a patent instead of using informal mehanisms to protet the innovation. 20 The firm s problem is to selet between the invention patent and the utility model to maximize the ex-ante flow of profit. I assume that imitation an redue the patentee s per-period monopoly profit, depending on the sope of the patent. 21 The rate of tehnology obsolesene determines the maximum periods of monopoly the patentee an enjoy. Higher ST D j orresponds to shorter periods of monopoly and thus shorter periods of effetive patent life. When the appliation is still pending, the appliant does not have the legal patent right, whih often results in delayed business ooperation and early infringement. The ex-post profit is therefore also affeted by how early the patent an be issued. 22 In this model, I assume appliants an 20 Some of the most frequently used tait mehanisms inlude serey, lead-time advantage, omplementary assets et (Cohen, Nelson, and Walsh 2000). A fully-saturated model should ompare the relative effiieny in terms of reouping returns to R&D between eah pair of the IP mehanisms. However, what is important in my empirial analysis is whether the relative effiieny between fast-weak patent and slow-strong patent hanges with exogenous variations in the rate of tehnology obsolesene. 21 Gallini (1992) disussed the extent of patent breadth as measured by imitation ost. Alternatively, Klemperer (1990) defined the patent breadth as the spatial produt differentiation. Gilbert and Shapiro (1990) defined it instead as the patentee s ability to raise prie. 22 Although examination is partly affeted by the appliants response to referee reports, most of the time delay an be attributed to PTO administration and baklog fritions (Popp, Juhl, and Johnson 2003). 26

33 expedite the patent examination proess only through filing for the fast patent. Suppose a patent appliation is filed with PTO at t = 0. Then patent protetion an be desribed by three fators: (t, b, T ), where t is the starting period of the effetive patent right (the date of patent allowane); b is the breadth of patent, whih an take a value in the interval [0, 1], with b = 0 orresponding to a zero-effetiveness patent that allows free imitation from ompetitors, and b = 1 orresponding to perfet patent protetion that bloks imitation until the the end of patent life; and T is the ending period of a statutory patent. The rate of tehnial knowledge obsolesene ST D j is defined as a patent value depreiation fator: ρ j. A higher rate of obsolesene orresponds to a higher ρ j, whih makes the monopoly profits depreiate at a faster pae. In addition, I assume there is a non-zero patent maintenane ost for eah effetive period until the patent expires. In this model, the renewal ost and the rate of tehnial knowledge obsolesene together will determine the patent "shut-down" period. 23 Immediately following this setup, (0, 1, + ) represents the strongest possible patent protetion (immediately granted, largest breadth and infinite periods). With suh a patent, I assume a patentee s innovation will reward him with a per-period monopoly profit of π until the tehnology beomes obsolesent and is replaed. In the last period during whih the patent is renewed, the marginal profit must equal the marginal ost: π e ρ T = (2.1) T = 1 ρ log π (2.2) Notie that, under the strongest patent protetion, the effetive patent life is not infinite. The length of the patent depends on the rate of tehnial knowledge obsolesene, the per-period monopoly profit and the patent renewal ost. Patentees would prefer longer patents when tehnology moves slowly, when per-period profits are larger, or when the patent maintenane ost is lower. Beause faster, broader and longer patents always seure higher profits, it suffies to ompare the differenes in profits under a faster but "weaker" patent with those under a slower but "stronger" patent. Define two distint types of patent protetions: Tier 1 (t 1, b 1, T 1 ) and Tier 2 (t 2, b 2, T 2 ) with the following relations: t 2 < t 1, b 2 < b 1, T 2 < T 1 and t 1 < T 2. The first three onditions indiate that Tier 2 patent is granted earlier with a narrower breadth and shorter protetion length than Tier 1. The fourth ondition shows that the protetion horizons of the two patents have ertain overlaps. This ondition, although not essential to the model, is onsistent with the urrent two-tier patent poliy designs in most ountries. With a Tier 1 patent, the disounted sum of profits is: Π T1 = min(t1, 1 ρ log b 1 π ) t 1 e rs (b 1 π e ρs )ds (2.3) where r is the disount fator. Similarly, with a Tier 2 patent, the disounted sum of profits is: Π T2 = min(t2, 1 ρ log b 2 π ) t 2 e rs (b 2 π e ρs )ds (2.4) 23 At major PTOs in the world, patent renewal fees have been inreasing over time. However, inreasing renewal fees in this model will not add further insight. 27

34 Using (3) and (4), the profit differene under alternative patent protetions is: Π 2 Π 1 = t1 t 2 e rs (b 2 π e ρs )ds (2.5) min(t2, 1 ρ log b 2 π ) t 1 min(t1, 1 ρ log b 1 π ) e rs ((b 1 b 2 )π e ρs )ds min(t 2, 1 ρ log b 2 π ) e rs (b 1 π e ρs )ds Tier 2 patent offers more profits in the earlier periods beause it is granted earlier, t 2 < t 1. However, a Tier 1 patent offers more per-period profits (b 1 > b 2 ) as well as longer periods of protetion min(t 1, 1 ρ log b 1π ) > min(t 2, 1 ρ log b 2π ). It is therefore straightforward to see that a Tier 2 patent will outperform a Tier 1 patent if and only if the differenes in profit during the early periods outweigh the differenes in profit during the later periods. Notie that (5) is weakly inreasing in T 1. That is, if the Tier 1 patent is granted faster, the advantage of the Tier 2 patent will beome smaller. For ρ smaller than a threshold value and T 1 larger than a threshold value, Π 2 Π 1 < 0 and the patent appliant will prefer a Tier 1 patent to a Tier 2 patent. Lemma 1. If (1) the statutory patent life for Tier 2 patent, T 2 is short enough and (2) the delay in Tier 1 patent, t 1 is quik enough, suh that the following regularity ondition holds: then there exist ρ and T 1 suh that Π 2 Π 1 < 0. Proof: please see appendix for details of proof. e rt 2 e rt 1 + e rt 2 (2.6) In other words, a patent appliant will prefer the Tier 1 patent beause most of the profit will arue during the later periods of the patent life (i.e., there is a small ρ); this later period would not be overed by a Tier 2 patent. This roughly orresponds to the ase of the pharmaeutial industry, as pharmaeutial firms generally renew their patents to full term, sine most of the profits is seured during the later periods of patent life (Budish, Roin, and Williams 2013). When the rate of tehnology obsolesene beomes greater, the per-period profits depreiate at a faster speed. In that ase, a Tier 2 patent is more favorable, beause a Tier 2 patent seures early periods of profit. Simultaneously, a Tier 1 patent beomes less attrative beause profits in the later periods might even fall short of the patent renewal osts. Proposition 1. If Lemma 1 s regularization and the following regularity onditions hold: ( b 2 π 1 log b 2π t 1 rt 1 logb 1 π log )+1 b 1π > 1 T 1 log b 1π ( rt 1 logb 1 π )+1 + e (rt 1+logb 1 π log) e (rt 2+ t 2 T1 (logb 1 π log)) > 0 (2.7) Π 2 Π 1 is inreasing in ρ when 1 t 1 log b 2π > ρ > 1 T 1 log b 1π ; Π 2 Π 1 is inreasing in ρ when Π 2 Π 1 < 0. In addition, there exists ρ suh that Π 2 Π 1 ρ = 0 and ρ > ρ, Π 2 Π 1 ρ > 0, making Tier 2 patent more favorable. (2.8) 28

35 Proof: Please see appendix for details of proof. The onditions desribed in Lemma and P roposition1 regulate the exogenous variables (t 1, b 1, T 1 ) and (t 2, b 2, T 2 ) suh that either a Tier 1 patent or a Tier 2 patent will be preferable depending on the rate of tehnial knowledge obsolesene. Based on this simple setup, however, the model predits that a faster rate of tehnologial obsolesene tends to make a Tier 2 patent more attrative than a Tier 1 patent only when 1 t 1 log b 2π > ρ > 1 T 1 log b 1π. This set of inequalities has an interesting eonomi interpretation. ρ > 1 T 1 log b 1π orresponds to the range of tehnial knowledge obsolesene in whih appliants will not renew their patents to Tier 1 s maximum statutory life. 1 t 1 log b 2π > ρ, on the other hand, refers to the range of obsolesene that will reward appliants with positive net per-period profits for some periods after the Tier 1 patent is issued. Only within this range of obsolesene will Π 2 Π 1 be stritly inreasing in ρ. Empirially, in major PTOs, on average less than 10% of patents will be renewed to the maximum term and less than 2% of patents will be abandoned before or immediately after patent issue (itation for U.S, SIPO, EPO statistis for patent life). Thus, our model is able to predit the behaviors of patent strategy in a wide range of tehnologies. The result shown in P roposition1 is the main hypothesis of this empirial exerise. Beause the Tier 2 patent (the utility model) is a registration model, the assumption that t 2 is fixed seems to be plausible. On the other hand, in major PTOs, baklogs are reating a signifiant variation in terms of Tier 1 patent examination; these baklogs an be attributed to ommuniation fritions, the extent of the examiners diligene, the need for a thorough hek of novelty and non-obviousness based upon existing prior art. The assumption that t 1 is fixed is likely to fail. It is thus important to understand whether hanges in Tier 1 patent examination effiieny are likely to influene the effet of tehnial knowledge obsolesene on the propensity to hoose Tier 2 relative to Tier 1. Given the rate of tehnial knowledge obsolesene ρ fixed within [ 1 T 1 log b 1π, 1 t 1 log b 2π ], a slower examination of Tier 1 patents will make Tier 2 patent more favorable. Beause the differene in profits is a ontinuous funtion of ρ, there exists a group of marginal "ρ" appliants who are willing to shift from filing for Tier 1 to Tier 2 patent when t 1 inreases. A greater t 1, therefore, orresponds to a smaller range of ρ appliants that find a Tier 1 patent more favorable. Proposition 2. Suppose the regularity onditions in Lemma 1 and Proposition 1 hold. Let ρ 1 and ρ 1 denote the rates of tehnologial obsolesene that make appliants indifferent between hoosing Tier 1 and Tier 2 patent when Tier 1 patent s examination delay is t 1 and t 1, respetively. If ρ 1 and ρ 1 both lie in [ 1 T 1 log b 1π, 1 t 1 log b 2π ], t 1 > t 1 if and only if ρ 1 < ρ 1. Thus, inreasing t 1 will make Tier 2 patent more favorable for a larger range of ρ appliants. Proof: please see appendix for more details. This model an also shed light on deisions about when to file for a patent, under the assumption that firms treat patent portfolios as their strategi assets. A rih set of literature has disussed the issue of strategi patenting, espeially after the "pro-patent" shift with the establishment of the Court of Appeals of the Federal Ciruit (CAFC) by the US Congress in Among the formal and tait mehanisms to protet intelletual property, patents are relatively ineffiient in terms of appropriating returns to R&D (Cohen, Nelson, and Walsh 2000). The deision to patent differs by firm harateristis. Firms with large patent portfolios exploit patent rights for preemptive purposes (Gilbert and Newbery 1982), strengthening ross-liensing bargains (Hall and Ziedonis 2001), defense against potential litigation(lanjouw and Shankerman 2001). Firms with small or no stok of 29

36 patents (e.g. researh-oriented start-up firms) might emphasize fast patent grants, "iron-lad" patents for purposes suh as seuring VC funding, and liensing agreements, signaling strong R&D abilities and enhaning ompetition potential. The ability to protet IP is, thus, influened by patent portfolio harateristis. Speifially, a firm an enhane protetion on a partiular patent by threatening rivals and imitators with his other patents. A firm an also preempt a rival s entry by filing "sleeping patents" or reating "thikets" of patents. The ability to file a patent for strategi purposes is largely influened by the stok of patents owned by the firm. Tehnology obsolesene is thus likely to have differential effets on patenting strategy aross firms with heterogeneous patent portfolio size 24. The relative differenes between Tier 1 and Tier 2 patents should be smaller for firms with larger patent portfolios. Proposition 3. Tehnial knowledge obsolesene has a differential impat on patenting strategies. The effet is less notable for firms with a larger portfolio. 2.4 Measuring Rate of Tehnial Knowledge Obsolesene The rate of tehnial knowledge obsolesene is one determinant of how fast and slow patent approval affet profits. This is the rate at whih new tehnologies emerge on the market and displae the urrent tehnology. I use patent renewal deisions to reate a proxy variable for this onept. This is beause when new and superior tehnology is introdued, the produt using the urrent tehnial knowledge will lose its value to the produers; the assoiated patent will lose its value to the inventor. So, higher rate of tehnial knowledge obsolesene will orrespond to shorter effetive patent life, expediting the mortality of patents. More speifially, we aggregate the renewal deisions for patents that are in the same tehnology fields to proxy for the rate of tehnial knowledge obsolesene. For patent i that is filed in year t, we an denote the tehnology fields as (S1 it, Sit 2, Sit 3,...Sit n ), where Sj it = 1 if patent i is loated in tehnology field j and 0 if not. Define Dm it = 1 if patent i is abandoned within m years after the grant or 0 otherwise. Suppose there is a total of Q patents that are filed in year t. With these notations, we define the tehnology level rate of obsolesene for tehnology j and ohort t as: ST D jt (m) = Q i=1 Sit j Q Dit m That is, we ategorize eah patent by its appliation year and tehnology. For eah ohort-tehnology ategory, we use the perentage of the patents that are given up within m years as the proxy variable of the rate of tehnial knowledge obsolesene. The above definition treats eah patent as a separate patent in eah of its tehnology fields. In my empirial analysis, I alulate the rates of tehnial knowledge obsolesene (m=4) in the United States, Germany, Frane and Great Britain during the period For United States patents, the tehnology definition follows the United States Patent Classifiation (USPC-3 digit, 435 distint lasses); for the three European ountries, the tehnology definition follows the International Patent Classifiation (IPC- 4digit, 639 distint lasses). Figure 1 illustrates the trends of development of the above indies in the 4 ountries under a more aggregate definition of tehnology fields defined in Hall, Jaffe and Trajtenberg (2001) (HJT). The six large tehnology fields are Chemial, Computer&Communiation, Drugs&Medial, Eletris&Eletronis, Mehanis and Others (whih inludes misellaneous tehnial areas suh as Amusement Devies, Apparel and Textile, Furniture, Heating, et.). Beause the HJT definition assoiates eah USPC lassifiation into the six tehnology fields, the ategorization of the European patents are made using a USPC-IPC onordane. 24 Due to the limitation of the data, our empirial findings are not able to inlude other harateristis at the firm level. 30

37 The first observation is the onsiderable variation in the rate of obsolesene aross tehnology fields. In the US, for example, the tehnology field "Others" has the highest rate of tehnology obsolesene (18.3%), whih is almost twie the measure for Computer and Communiation (9.5%). 25 Seond, the rankings of the tehnology obsolesene measures hange over time in all four ountries. In Frane during , the tehnology field Computer&Communiations had the lowest ST D jt but it surpassed Chemial and Drugs&Medial in the following five years. Third, ST D jt in different tehnology fields have similar trends over time. Table 5 illustrates the hanges of ST D jt over the period for United States patents. The tehnology fields that have the largest dereases in the rate of tehnial knowledge obsolesene are Leather Manufaturing and Musial Instruments. The tehnologies whih have the highest inrease in developments are Beds, Books and Amusement Devie. The fields that remain relatively stable inlude X-ray, Drug and Organi ompounds. It is interesting to notie that the most volatile hanges in ST D jt our in tehnologies that fous on the development of shape, struture or funtion of produts, while the ST D jt remains unhanged for basi and applied researh. Similar tables for the European ountries are inluded in the appendix. Beause the model relates tehnial knowledge obsolesene with patent hoie, and eah patent an belong to more than one tehnology field, we further define the mean rate of tehnial knowledge obsolesene for eah patent i filed in year t as: MeanST D it (m) = n j=1 S it j ST D jt (m) that is, the mean rate of tehnology obsolesene for patent i filed in year t is a summation of the tehnial knowledge measurement, weighted by the patent s tehnology fields. In the empirial analysis, I take into aount that, when a patent appliation is filed, the appliant an only observe the past rate of tehnial knowledge obsolesene. So, for eah patent, I further demean the measure by a three-year average of MeanST D it in the previous periods of the patent s appliation date (shown speifially in the setion desribing my eonometri model). Bilir (2013) has proposed to use mean forward itation lag to measure the length of produt (tehnology) lifeyle. 26 While the "itation lag" measure has the advantage that it exploits relative information throughout a patent s lifetime, it also has the disadvantage of a trunation problem, as a signifiant portion of itations appear 5 years after the patent grant (Hall, Jaffe, and Trajtenberg 2001). In addition, iting a previous patent is more onsistent with the understanding that the previous patent is "narrowing" the sope of the urrent patent rather than the idea that innovations overed by the previous patent are superseded by the innovations overed in a patent. Coneptually, therefore, the "renewal" measure is more losely related to the rate of tehnial knowledge obsolesene. Another advantage of the "renewal" measure is that I ategorize the tehnology fields at a muh smaller luster and allow the measure to vary over ohorts. Nevertheless, I also ompare estimates of my regression based on Bilir s measure using mean itation lag and find onsistent results In the United States, software patents are extremely valuable although produt-yle for software rarely goes beyond five years (Graham and Mowrey 2004). This does not onern me beause I am proxying the hanges in tehnial knowledge obsolesene over time and within the same tehnology field, based on the hanges in the effetive patent rights for a given age of patents. 26 She mainly uses the term produt yle, with oasional use of tehnology yle. Her measures, however, are based on the idea of tehnial obsolesene. 27 The itation lag measure I use is also defined at the USPC 3-digit level and is allowed to vary over ohorts. Results are not shown in this paper. 31

38 2.5 Data Desription and Summary Statistis To empirially evaluate the propositions presented in the theoretial model, I need measures for the rate of tehnial knowledge obsolesene and patent information for innovations filed in the U.S, China and Europe. I test the same set of hypotheses using two different datasets: inventions that sought patent protetion in both China and the US (SIPO-USPTO patent dyads) and inventions that sought patent protetion in both China and Europe (SIPO-EPO patent dyads). In either of these datasets, beause protetion is applied for in both China and another system (the US or EPO), I an observe whether the firm is applying for a Chinese invention patent or a Chinese utility model. I then link this hoie with a measure of tehnial knowledge obsolesene. I desribe the method of reating measures and data seletion riteria below. I ombine information from several datasets: patent data published by SIPO , patent information from the USPTO website, Harvard Patent Dataverse and EPO Worldwide Patent Statistial Database (PATSTAT). Both SIPO-USPTO and SIPO-EPO patent dyads an be identified using the "priority number" information from eah US and EPO patent and mathing it to the priority appliation number of the Chinese patent in the dyad. 28 Using the "legal status" information in PATSTAT, I further identify all EPO patents that are designated to Germany, Frane and Great Britain. Harvard Patent Dataverse provides information on US patents. In addition, I use the USPTO website to extrat all the USPC lassifiations for eah US patent. Beause Chinese utility model patents an only protet industrial produt innovation while invention patents an protet both proess and produt innovation, it is important to distinguish proess and produt invention patents. Fortunately, the title of eah Chinese patent appliation needs to follow a strit format. For proess innovation, the title needs to ontain key words suh as "proess," or "method." For produt innovation, the key words are "produt," "struture," or "devie". I also identify the types of patentee (inluding firms, publi researh institutes and individual) through a keyword mathing method, and exlude non-firm patentees from the data. Table 6 provides summary statistis of the SIPO-USPTO patent dyad dataset for all variables used in regressions. Every observation in the dataset onsists of a patent appliation in China and a patent appliation in the US for the same invention during the period 2001 to The sample is deomposed into two groups: inventions for whih appliation is made for both a Chinese invention patent and a US patent (77% of the sample) and inventions for whih appliation is made for both a Chinese utility model and a US patent (23% of the sample). As of 2014, 28% of US patent appliations were granted while the rest are either rejeted or still under examination % of Chinese invention patents in this sample were granted US patents; 27.75% of Chinese utility models were granted US patents. The differene in grant rates is statistially insignifiant. There are a total of 4,652 U.S. patent appliations with Chinese invention patent priority of whih 1,424 (30.61%) are produt innovations, 901 (19.37%) are proess innovations and 2,327 (50.02%) are both produt and proess innovations. There are 1,556 U.S. patent appliations with utility model priority and all of them are produt innovations. There are a total of 1155 distint assignees. There are 368 assignees that have never filed any other invention patent prior to the one in our sample. On the other side, only 21 assignees have an invention patent portfolio larger than 50 prior to the patent they filed in both China and the US, with the largest assignee having 1681 invention patents. In Table 2, I further deompose the two groups of patents into 6 HJT tehnology lassifiations defined by their main USPC. As shown, the perentage of Chinese utility model patents varies signifiantly aross different tehnologies. In Eletris & Eletronis, Mehanial and Others, the perentage of U.S. patents with Chinese utility model priority ranges from 40 % to 60 %. In ontrast, in Chemials, Computer & Communiation and Drugs & Medials, less than 15 % inventions were filed under the utility model. Clearly shown, the hoie of IP 28 Foreign appliants seeking patent protetion in China seleted the utility model patent in only 88 ases during the period So I only inlude Chinese firms filing patents in both China and the the US (EPO). 32

39 protetion varies aross tehnology fields. Two limitations of the datasets need to be addressed. One disadvantage of the SIPO patent dataset is that it does not show the length of eah patent appliation doument, whih affets the ost of applying for a patent. As mentioned above, one of advantage of the Chinese utility model is the relatively low ost of appliation and maintenane. The appliation ost largely onsists of an attorney to prepare patent appliations. Interviews with law firms reveal that the number of words in eah appliation is a proxy variable for the legal ost. For example, one lawyer from Tee&Howe told us that they harge 220 rmb ($34) per 100 Chinese haraters, as of Google transforms the original Chinese patent appliation pdf file into an online html format that is able to be extrated. Thus we are able to aquire the total number of (both independent and dependent) laims for eah Chinese patent appliation. Unfortunately, the html webpages also ontain muh other ontent of patents, so a simple word ount of the entire file does not give us the exat number of words in the patent appliation. The seond limitation is that we are unable to find a good dataset that provides firm finanial data. Similar to Compustat, the National Bureau of Statistis (NBS) in China also provides finanial data for Chinese ompanies that have annual sales above five million rmb. Mathing the NBS data with our patent dyads would eliminate almost 70% of our patent observations beause most patent dyads are not filed by large firms. For this reason, we deided to inorporate only one firm level variable, the size of patent portfolio prior to the patent appliation. Beause eonomy and tehnology growth might indue hanges in patent law and, in turn, influene firms patenting strategies, we speifially selet our sample in the period during whih there were no hanges to odified Chinese patent law. 30 Failure to ontrol for institutional hanges in patent regime ould lead us to severely biased estimates. 2.6 Eonometri model and Identifiation Estimating Equations The invention patent orresponds to the Tier 1 patent while the utility model orresponds to the Tier 2 patent. Proposition 1 states that when the rate of tehnology obsolesene inreases, patent appliants will find the utility model more favorable. This result motivates an estimating equation of the following form: D(UM iklt ) = β + β 1 t 1 z=t 3 MeanST D iz(4) + Γ X i + α l θ t + 3 n j=1 S it j + λ k + ɛ iklt (2.9) where D(UM iklt ) is a dummy variable that equals one if the appliant has hosen the utility model in China for patent i of firm k loated in provine l during year t. Beause eah patent i belongs to multiple tehnology fields defined by (S1 it, Sit 2, Sit 3,...Sit n ), the tehnology fixed effets, n j=1 Sit j ontrol for eah tehnology field t 1 z=t 3 in whih the patent is defined. MeanST D iz(4) 3 is defined as the average perentage of patents that are given up within four years from issue date for all patents (either U.S. or European) that are filed in the past three years and in the same tehnology fields (USPC 3-digit or IPC 4-digit) as the observed patent. X i is the ontrol variables at patent and firm level. As explained above, the inventions filed under the utility model might 29 The SIPO patent dataset in the CD-ROM only provides one independent laim per patent. The full ontents of eah patent an be retrieved in a PDF file on the SIPO website. 30 See setion 2.1 for a brief introdution to the three Chinese patent law amendments whih happened in 1993, 2000 and 2008, respetively. 33

40 have signifiantly lower tehnial quality ompared to those under invention patents beause no examination is required. To ontrol for this issue, I add in a variable Grant whih measures whether the US or EPO patent is granted. I also add in patent level variables at the appliation date to further ontrol for patent quality variations. Consistent with the previous literature, these ontrols inlude the number of patent laims, number of inventors (Chinese and foreign inventors are treated equally), and number of assignees (Lanjouw and Shankerman 2004). Previous literature has pointed out that the number of ountries in whih an appliation is filed is positively orrelated with the quality of patent (Putnam 1997). Beause our data already inludes patents that are filed in at least two ountries (and mostly two to three patent offies), we do not expliitly ontrol for this. In addition, international patent filing an be proessed either diretly, through filing in the designated ountry or ountries, or through the Patent Cooperation Treaty (PCT) route. Beause different routes might reflet appliants heterogeneous motives for patenting, it is important to ontrol for this variable. Beause firms treat patents as strategi assets, one would expet their patent strategies to be orrelated with other firm-level harateristis. Strategi patenting is found to be most ommon in the group of "experiened patent filers" (Kortum and Lerner 1998) and R&D intensive firms (Hall and Ziedonis 2001). However, our data prevent us from adding any variable that is related to the firm s finanial performane (suh as R&D investment). Nevertheless, we ontrol for the size of the patent portfolio prior to the appliation date of the observation and for firm fixed effets to aount for unobservable harateristis that are fixed over time. Starting in 1995, many provines in China implemented patent subsidy and tax dedution poliies as a response the entral government s theme of "indigenous innovation." 31 Sine these poliies have a diret effet on patenting behaviors, I add in interations of provine and year, α l θ t to fully ontrol for any onurrent poliy hanges that are likely to bias our estimates. The main oeffiient of interest, β 1, aptures the influene of tehnial knowledge obsolesene on patenting behavior. Aording to the model, where β 1 > 0 indiates that the firm is in a fast-developing field, appliants should have a higher propensity of seleting utility model patents. Proposition 2 states that the impat of tehnology obsolesene on patent hoie is also affeted by how fast SIPO proesses invention patents. If appliants expet the invention patent to be proessed faster, the advantage of filing for the utility model will be diminished. Empirially, we examine this hypothesis as follows: D(UM iklt ) = β + β 1 +β 3 t 1 z=t 3 MeanST D iz(4) t 1 z=t 3 MeanST D t 1 iz(4) z=t 3 3 MeanGrantlag iz 3 (2.10) 3 + β 2 t 1 z=t 3 MeanGrantlag iz 3 + Γ X i + α l θ t + n j=1 Sit j + λ k + ɛ iklt t 1 z=t 3 MeanGrantlag iz 3 is the ohort-tehnology level average grant lag of invention where the new variable patent examination delay in the three years before the patent appliation for the observation. Eah omponent under the summation, MeanGrantlag it, is a sum of average invention patent grant lags in tehnology fields, n ). Smaller MeanGrantLag it indiates a weighted by the tehnology presene of patent i, (S it 1, Sit 2, Sit 3,...Sit higher speed of invention patent examination and a redued advantage assoiated with utility model patents. Aording to the model, appliants should be less sensitive to tehnology obsolesene hanges for smaller MeanGrantLag it ompared to larger MeanGrantLag it. This orresponds to β 2 > 0. Proposition 3 states that appliants heterogeneity of patent portfolio size affets their patent hoie responses to hanges in tehnial knowledge obsolesene. An appliant with a larger patent portfolio is less sensitive 31 Aording to our own olletion of information, we find at least nine provines in China have adopted some form of patent subsidy and/or tax dedution poliies by These provines inlude Guangdong, Liaoneng, Jilin, Hubei, Shanghai, Beijing, Anhui, Jiangxi and Tianjin. 34

41 to hanges in tehnology obsolesene beause he an utilize the advantage of his patent stok to partially overome the relative ineffiienies due to the slow speed of the invention patent or the low enforement strength and short protetion term of the utility model. Empirially, we estimate the following speifiation: D(UM iklt ) = β + β 1 t 1 z=t 3 MeanST D iz(4) t 1 z=t 3 MeanST D iz(4) 3 + β 2 +β 3 P atentstok kt 1 + Γ X i + α l θ t + n j=1 Sit j 3 P atentstok kt 1 (2.11) + λ k + ɛ iklt where P ortfolio kt 1 is the size of patent portfolio (e.g. total number of invention patents previous to the appliation year of the observation) in firm k in period t 1. Proposition 3 orresponds to β 3 < Identifiation Identifiation of β 1 is based on within-tehnology variation of the rates of tehnology obsolesene. A key advantage of this approah is that it mitigates onerns of omparing patenting behaviors in different tehnology fields (Mansfield 1986). Patents have been pereived to be most effetive in pharmaeutials; their effetiveness varies in other tehnology areas, notieably Eletris and Eletronis. Beause suh fators an affet appliant s patenting strategies, the interpretation of β 1 in a regression without ontrolling for tehnology fixed effets will be unlear. As pointed out by the literature, hanges in patent behavior an be deomposed into two effets: hanges in the omposition of firms in the tehnology over time (entry and exit) and hanges in the eonomi behavior of existing firms (Hall and Ziedonis 2001). To address this onern, we ompare estimates of β 1 obtained from regressions that inlude both tehnology and firm fixed effets with regressions only inluding tehnology fixed effets. Patent strategies are likely to be influened by other eonomi variables, suh as the market struture. Therefore, when a new or superior tehnology is introdued, it is likely to influene firms patenting deisions, either diretly, as desribed in our theory, or indiretly, through its influene on market struture. To mitigate this onern, we speifially hoose our measure of tehnial knowledge obsolesene to be alulated based on patents in foreign ountries. This is beause tehnologial progress in foreign ountries is unlikely to diretly influene on the market struture in the domesti ountry. 2.7 Main Results Rate of Tehnial Knowledge Obsolesene and Patenting Choie Beause the dependent variable is a dummy, a probit model is used. 32 I luster the standard errors either at the main tehnology field or at the firm level, depending on whether the eonometri speifiation inludes a firm fixed effet. Estimates of (9) appear in Table 7. The results are strongly onsistent with Proposition 1 in the theoretial model. In olumn 1, I find evidene that, when tehnial knowledge obsolesene beomes larger, firms have a higher propensity to hoose the utility model to protet their inventions. In addition, the variable Grant is not signifiantly orrelated with an appliant s patenting hoie. This result supports the validity of our sample seletion riteria, i.e., that the ex ante tehnial quality of the inventions is not systematially differentiated between the Chinese invention patent group and the utility model group. In olumn 2, I add in patent level variables to further ontrol for potential patent quality differenes and find that the influene of the rate 32 I also ompare my results with estimates using OLS and logit, and the results are similar. 35

42 of tehnial knowledge obsolesene remains nearly idential. The point estimates of β 1 ( in olumn 1 and in olumn 2) orrespond to a marginal effet of a 133% inrease in the propensity of filing under the utility model (alulated using results from olumn 2). More speifially, a one standard deviation inrease in the rate of tehnial knowledge obsolesene will result in a 7.6% inrease in the propensity to file utility models in the oming year. In olumn 3, firm fixed effets are added and patent appliants that have filed only one during the period are dropped. In this ase, the point estimate an be interpreted as the effet of tehnial knowledge obsolesene on patent hoies at the firm level. Comparing point estimates in olumn 2 and 3, I find that the one in olumn 3 is 80% larger. The inrease in the point estimate of β 1 after ontrolling for firm dummies shows that the influene of tehnology obsolesene is not mainly driven by entry of new firms but rather by hanges in the eonomi behavior of existing firms Heterogeneity: SIPO Administration Dynamis Table 8 provides estimates orresponding to Proposition 2. Regression results are onsistent with the theoretial model. The signifiant drop in the number of observations (e.g. in olumn 1, the original sample inludes 4712 observations, now only 2582) after ontrolling from MeanGrantlag it is due to the imperfet math between USPC and IPC. 33 Nevertheless, MeanST D it are still signifiant in olumns 1 and 3 and are not very different from results in the previous table. The key estimate of interest, β 2 in (10), is always positive and signifiant, whether inluding patent level ontrol or assignee fixed effets. This indiates that patent appliants are more sensitive to hanges in the rate of tehnial knowledge obsolesene when SIPO s overall effiieny in examining invention patent dereases. In olumn (4), the marginal effets of β 1 and β 2 are and , respetively. At the 25 th perentile and 75 th perentile of MeanGrantlag it (1578 days and 1731 days), a one standard deviation inrease in the rate of tehnology obsolesene will inrease patent appliants propensity for filing the utility model by 3.0% and 9.4%, respetively. At the mean rate of tehnology obsolesene (0.1457), a one standard deviation inrease in the examination delay at SIPO (147 days) will inrease the propensity to file for the utility model by 16.49%. At SIPO s mean examination delay (1672 days), this estimate orresponds to a grant lag elastiity of demand of 2.82 for the utility model, whih means that, if there is a 6 months inrease in grant lag for invention patents, about 10% more appliations are willing to swith from filing for an invention patent to filing for a utility model, a signifiant hange. Given that an examiner s effort for eah patent appliation is relatively limited (Farrell and Shapiro 2008), a derease in PTO s examination effiieny has a signifiant impat on the effetiveness of the invention patent Strategi Patenting: Appliant s Patent Portfolio Size To evaluate Proposition 3, we investigate the effet of tehnial knowledge obsolesene on patenting hoie aross appliants with different patent portfolio sizes. The variable Large Patent Portfolio Dummy equals 1 for all appliants whose Chinese patent portfolio size exeeds the mean patent portfolio size of the sample (14 patents) prior to the urrent patent appliation, and 0 otherwise. We estimate (11) for all appliants and also estimate variants of (11) separately for large portfolio size appliants and small portfolio size appliants. Firm fixed effets are inluded to ontrol for appliant level unobservables that do not hange over time. The results (Table 9) show that small portfolio holders are more sensitive than large portfolio holders to hanges in the rate of tehnial knowledge obsolesene, whih is onsistent with the theoretial model that a large portfolio holder is more apable of irumventing the ineffetiveness of invention patent and utility models by utilizing other patents in his pool. In fat, results in olumns 1-2 show that large portfolio holders do not respond to hanges in tehnial knowledge obsolesene. 33 E.g. in the US there is a lass named G9B after 1980, but there is no orresponding IPC lass. 36

43 2.7.4 Subsample estimations To further evaluate the effet of tehnial knowledge obsolesene on optimal patent hoie, we estimate speifiations that allow the oeffiient to vary aross different sizes of tehnology obsolesene. We ategorize MeanST D it into Small, Medium and Large ST D it by the sample s 33 rd and 67 th perentile, defined in dummy variables LargeST D, MediumST D and LowST D. We estimate the following speifiation: D(UM iklt ) = β + β 1 +β 3 t 1 z=t 3 MeanST D iz(4) t 1 z=t 3 MeanST D iz(4) 3 M ediumst D (2.12) 3 + β 2 t 1 z=t 3 MeanST D iz(4) 3 LowST D + Γ X i + α l θ t + n j=1 Sit j + ɛ iklt Table 10 provides regression results for (11). Comparing olumns (1)-(3), the appliant s patent hoie is signifiantly influened by hanges in MeanST D ijt only in the group with the highest rates of tehnial knowledge obsolesene. As the rate of tehnology obsolesene dereases, not only do the estimated oeffiients beome smaller but they also beome statistially insignifiant. These results offer further support to our theoretial model: the advantage of fast protetion is more salient when the rate of tehnial knowledge obsolesene is relatively high. 2.8 Robustness heks with SIPO-EPO patent dyads While our theoretial results relate individual appliant s patenting strategies to a measure of how quikly new tehnologies are introdued, we test our hypotheses based on SIPO-USPTO patent dyads and alulate tehnial knowledge obsolesene measures using only United States patent data. One of the unique features about the United States patent system is that it allows patent appliants to file ontinuations. Even if the patent examiner onludes that an invention is patentable, the appliant an still abandon the right an infinite amount of times and start the proess over again. This institutional design makes it more flexible for patent appliants to manipulate patent pendeny to their advantage and has the result that 90% of US patent appliations are granted eventually (Quillen and Webster 2001). Based on the high grant rate of US patent, one would expet that many weak patents, inventions that would not be granted patent rights under stringent examination, are allowed by the USPTO (Farrell and Shapiro 2008). If there is a large variation in the tehnial quality of US patents, then the renewal measures I am using ould both apture the hanges in tehnology as well as hanges in patent quality. I address these onerns in two steps. First, I hange the ountry for reating measures of tehnial knowledge obsolesene to three European ountries, namely, Germany, Frane and Great Britain, in whih patent appliants are muh more restrited (ompared to the US) in their flexibility to delay patent issue. Seond, instead of using the entire patent pool within one ountry, I use only the EPO granted patents that are designated to the ountry. This areful seletion has two advantages: first, the European Patent Offie sets a uniform and stringent patentability standard. 34 Patents issued by EPO are presumably high in terms of tehnial quality. Seond, previous studies have shown that the patent quality is positively orrelated with the number of ountries applied to. Filing a EPO patent reflets the appliant s intention to seure patent protetions in multiple European ountries. So the EPO patent should be onsidered as the "top tier" patent within one ountry. In other words, the hanges in the renewal behaviors over time for EPO patents should primarily reflet appliants onern for tehnology upgrading and turnovers. Following the above strategy, we ondut our analysis using SIPO-EPO patent dyads and alulate three alternative measures of tehnial knowledge obsolesene based on EPO patent data designated to the following three 34 A omparative study of the inventive step standards in the European, Japanese and United States patent offies show that grant rates at EPO, JPO and USPTO (without ounting ontinuation) are 55%, 49% and 54%, respetively. See for more information. 37

44 European ountries: Germany, Frane and Great Britain, respetively. Table report results for model (9) and provide onfirmation for proposition 1. Estimates in olumns 1-3 reveal a qualitatively similar pattern of sensitivity to tehnology hange ompared to olumns 1-3 in Table 7. For the SIPO-EPO patent sample, a one standard deviation inrease in rate of tehnial knowledge obsolesene inreases the propensity to file for the utility model by 8.28%, 6.76% and 10.04%, using the Germany, Frane and Great Britain EPO patents as obsolesene measures, respetively (using the point estimates from olumn (3) of Table 11-13). Given that the SIPO-USPTO ounterpart is 7.6%, we find the sensitivities of patent hoie with respet to rate of tehnial knowledge obsolesene, estimated using different MeanST D it, measured based on four ountries, US, Germany, Frane and Great Britain are losely omparable with eah other, with similar signifiane. Tables reports results for Proposition 2. Consistent with the results shown in Table 8, we also find that an inrease in the examination effiieny of SIPO s invention patents results in a derease in sensitivity of patent hoies with respet to tehnial knowledge obsolesene, measured using EPO patent dataset. Table 17 reports results for Proposition 3. In olumns (1) - (6), although the point estimates of the key variable MeanST D it LargeP atentp ortfolio it 1 are always negative, they are not statistially signifiant (exept in olumn (5)). We, therefore, are not able to find support for Proposition 3 in the EPO dataset. 2.9 Conlusion In this paper, we provide theoretial and empirial analysis of how firms optimal patent hoie is influened by hanges in the rate of tehnial knowledge obsolesene. In a simple model where inventors rely on patents to seure monopoly profits due to invention, we highlight the importane of speed of patent grant and develop results with regard to tradeoffs between fast and slow patent protetions. The model indiates that the tradeoff is influened by hanges in the rate of tehnial knowledge obsolesene, a measure we reate to proxy for the speed of tehnology development. Firms propensity of hoosing a fast patent inreases as tehnology develops faster; this response beomes very salient in tehnology fields that have espeially fast tehnology upgrading. Our empirial results are strongly onsistent with our theoretial model. In tehnology fields where the rates of tehnial knowledge obsolesene inrease over time, we find subsequent patent appliants propensity of hoosing the utility model also inreases signifiantly. Although the utility model is inferior to the invention patent due to shorter protetion periods and narrower patent sope, interations between rate of tehnology obsolesene and SIPO s average grant lag explain that the hoie between the invention patent and the utility model is mainly due to whether appliants need fast protetion. The results provide evidene that the rate of tehnology obsolesene is a strong determinant of appliants patenting strategy both at the tehnology and assignee level, establishing the ausal effet of hanges in tehnology development on patenting hoie. Our results find their usefulness in the literature of optimal design of patent poliy. Previous literature has argued that a uniform patent system might be unable to satisfy the heterogeneous demands for patent protetion. With the patent attributes fixed, the system tends to over-reward some inventors but under-reward others. We suggest there is an additional poliy lever that is worth analyzing: patent appliants might differ in their preferenes for how fast the patent an be granted. Our findings suggest that speed of patent grant is an important onsideration; patent appliants might even willing to seure a fast patent right at the expense of protetion length and enforement strength. These omparisons imply a potential welfare enhanement: beause weaker patents reate lesser per-period distortion and total periods of distortion and faster protetion makes them more effetive 38

45 to appliants, it is welfare enhaning (ompared to the urrent uniform system) to provide fast but weak patent protetions in fast-moving tehnology fields. Based on this result, we propose two diretions for future researh. First, will a hybrid patent system that offers flexibility in protetion speed, width and term be welfare enhaning? To answer this question, one needs to ompare the welfare of a hybrid system to not only the urrent patent system with a uniform patent poliy but also to a ounterfatual ase where there is a uniform alternative patent poliy with different attributes. Seond, future researh for patent poliy should also fous on how a uniform patent system affets R&D inentives in different tehnology fields. Further quantitative investigation of these possibilities is important to promote the understanding of patent poliy and tehnology development. 39

46 2.10 Appendix Proof of lemma 1: when ρ = 0, + min(t 2, 1 ρ log b 2 π ) e rs (b 1 π e ρs )ds = Using the ondition from the Lemma, we see that: + T 2 e rs (b 1 π )ds t1 t 2 + T 2 e rs (b 1 π )ds e rs (b 1 π )ds t1 t 2 e rs (b 2 π )ds Define K(ρ, T 1 ) = t 1 t 2 e rs (b 2 π e ρs )ds T 1 T 2 e rs (b 1 π e ρs )ds. We have just shown that K(0, + ) < 0. Moreover, K(0, T 1 ) is stritly inreasing in ρ at (0, + ): dk(ρ, T 1 ) dρ (0,+ ) = > t2 t 1 se rs (b 2 π )ds + t2 t 1 t 2 e rs (b 1 π )ds + + T 2 + se rs (b 1 π )ds T 2 T 2 e rs (b 1 π )ds > 0 where the last inequality follows from the regularity ondition. Sine K(0, + ) < 0, there exists suffiiently small ρ suh that K(ρ, + ) < 0 (although K(ρ, ) > K(0, )). K(ρ, T 1 ) is dereasing in T 1. Sine K(ρ, + ) < 0, there exist suffiiently large T 1 suh that K(ρ, T 1 ) < 0. We an always hoose T 1 and ρ suh that T 1 1 ρ log b 1π hold. Therefore, T 1 = min(t 1, 1 ρ log b 1π ). Sine T 2 is fixed, for suffiiently large ρ, it must be true that T 2 = min(t 2, 1 ρ log b 2π ). So K(ρ, T 1 ) < 0 t 1 t 2 t1 t 2 e rs (b 2 π e ρs )ds T 1 T 2 e rs (b 1 π e ρs )ds < 0 e rs (b 2 π e ρs )ds min(t 1, 1 ρ log b 1 π ) min(t 2, 1 ρ log b 2 π ) e rs (b 1 π e ρs )ds < 0 It is straightforward to see that Π 2 Π 1 is smaller than the last inequality so the onditions on t 1, t 2, T 2 are suffiient but not neessary. Proof of Proposition 1: Sine the integral interval points are determined by the minimum of two variables, the best way to illustrate the first order derivatives is to disuss under separate ases. Suppose ρ and T 1 and T 2 satisfies 1 ρ log b 1π = min(t 1, 1 ρ log b 1π ) and 1 ρ log b 2π = min(t 1, 1 ρ log b 2π ). In addition, assume t 1 < 1 ρ log b 2π so that the 2nd integral exists. In this ase, taking the first order derivative w.r.t. Π 2 = Π 1 d(π 2 Π 1 ) dρ = t1 + 1 ρ 2 log b 2π (b 2π r b ) 1 b ρ 2 b 2 t1 > t 1 e (r+ρ)s b 1 πds + t 2 = b 1 πt 1 (e r+ρ ρ log b 2 π > b 1 πt 1 ( b 2π ( t 2 se (r+ρ)s b 2 πds + rt 1 1 ρ log b 2 π + 1 ρ 2 log b 1π (b 1π 1 ρ log b 2 π t 1 se (r+ρ)s (b 1 b 2 )πds + r ) t 1 t 1 e (r+ρ)s (b 1 b 2 )πds + + e (r+ρ)t 1 e (r+ρ)t 2 e r+ρ ρ log b 1 π ) logb 1 π log )+1 b 1π ( rt 1 ρ ( b 1 ) 1 b 1 ρ 2 log b 2π (b 2π 1 ρ log b 1 π 1 ρ log b 2 π 1 ρ log b 1 π 1 ρ log b 2 π r ) se (r+ρ)s b 1 πds ρ ( b 1 b 2 ) t 1 e (r+ρ)s b 1 πds logb 1 π )+1 + e (rt 1+logb 1 π log) e (rt 2+ t 2 T1 (logb 1 π log)) ) > 0 > 0 40

47 where the first inequality utilize the fats that b 1 > b 2, t 1 < 1 ρ log b 2π and the seond line in the first order ondition anels out ompletely. Calulating the integrals in the third line gives us the fourth line. Sine 1 t 1 log b 1π > ρ > 1 T 1 log b 1π, shrinking the positive terms in the above parenthesis to the lower bound and enlarging the negative terms to the upper bound we have the seond inequality. Using this most ompliated ase as the benh mark, suppose ρ dereases suh that T 2 < 1 ρ log b 2π but T 1 > 1, there is only one integral bound that involves ρ, the first order ondition thus beomes: ρ log b 1π d(π 2 Π 1 ) dρ + 1 ρ 2 log b 1π t1 T2 1 = se (r+ρ)s b 2 πds + se (r+ρ)s ρ log b 1 π (b 1 b 2 )πds + se (r+ρ)s b 1 πds t 2 t 1 T 2 (b 1π r ) ρ ( b 1 ) b 1 t1 > t 1 e (r+ρ)s b 1 πds + t 2 T2 t 1 t 1 e (r+ρ)s (b 1 b 2 )πds + = b 1 πt 1 (e (r+ρ)t 2 + e (r+ρ)t 1 e (r+ρ)t 2 e r+ρ ρ log b 1 π ) > b 1 πt 1 (e r+ρ ρ log b 2 π > b 1 πt 1 ( b 2π ( rt 1 + e (r+ρ)t 1 e (r+ρ)t 2 e r+ρ ρ log b 1 π ) logb 1 π log )+1 b 1π where the seond inequality follows sine T 2 < 1 ρ log b 2π ( rt 1 1 ρ log b 1 π T 2 t 1 e (r+ρ)s b 1 πds logb 1 π )+1 + e (rt 1+logb 1 π log) e (rt 2+ t 2 T1 (logb 1 π log)) ) > 0 > 0 and the rest of the derivation is idential to the above. Suppose ρ ontinue to derease and T 1 < 1 ρ log b 1π, no integral bound involves ρ so the first order ondition will be: d(π 2 Π 1 ) t1 T2 T1 = se (r+ρ)s b 2 πds + se (r+ρ)s (b 1 b 2 )πds + se (r+ρ)s b 1 πds dρ t 2 t 1 > t1 t 2 t 1 e (r+ρ)s b 2 πds + T2 t 1 t 1 e (r+ρ)s (b 1 b 2 )πds + = t 1 ((Π 1 Π 2 ) + 1 r (e rt 2 e rt 1 e rt 2 + e rt 1 ) = t 1 ((Π 1 Π 2 ) 1 r K(0, T 1)) > 0 T1 T 2 T 2 t 1 e (r+ρ)s b 1 πds the sat step uses the assumption that Π 1 Π 2 > 0 and the regularity ondition from Lemma 1. It ρ is big enough suh that t 1 > 1 ρ log b 2π and t 2 < 1 ρ log b 2π, Π 2 Π 1 = 1 ρ log b 2 π t 2 e rs (b 2 π e ρs )ds > 0 So Tier 2 patent is more favorable. The Lemma demonstrates a ase where Π 2 Π 1 < 0 Sine Π 2 Π 1 is a ontinuous funtion of ρ, there exists a ρ suh that Π 2 Π 1 ρ = 0. If ρ satisfies 1 ρ log b 1π < T 1 and 1 ρ log b 2π > t 1, sine Π 2 Π 1 is inreasing in ρ for all ρ in interval ( 1 T 1 log b 1π, 1 t 1 log b 2π ), for all ρ > ρ, Π 2 Π 1 > 0 holds. If ρ satisfies 1 ρ log b 1π > T 1, pik ρ to be the largest ρ that satisfies this ondition. For all ρ > ρ, Π 2 Π 1 > 0 holds. 41

48 Figure 2.1: Heterogeneity in Patent Renewal Behaviors by Tehnology Fields: US, Germany, Frane and Great Britain The remain trivial ase ours when ρ is so big that t 2 > 1 ρ log b 2π, then the appliant will not even file for patent, a situation ruled out by the assumption of model. Proof of Proposition 2: take the first-order derivative of Π 2 Π 1 w.r.t. t 1 : d(π 2 Π 1 ) dt 1 = e rt 1 (e ρt 1 b 2 π ) + e rt 1 (e ρt 1 (b 1 b 2 )π) = e rt 1 (e ρt 1 b 1 π ) > 0 so Π 2 Π 1 is an inreasing funtion of t 1. For a given t 1, sine Π 2 Π 1 ρ1,t 1 = 0, we have Π 2 Π 1 ρ1,t > 0 for t 1 > t 1. Aording to P roposition 1, Π 2 Π 1 is an inreasing funtion of ρ when ρ is in [ 1 T 1 log b 1π, 1 t 1 log b 2π ]. This means if Π 2 Π 1 ρ1,t 1 = 0, then it must be true that ρ 1 < ρ 1. So when t 1 inreases to t 1, all the ρ that lies in (ρ 1, 1 t 1 log b 2π ] will make Π 2 Π 1 > 0, enlarging the range of ρ appliants that will make Tier 2 patent more favorable. Notes: These 4 figures plot the time trend of the perentages of ineffetive patent rights (within 4 years of patent issue) for all granted US (EPO) patents filed in the same year and the same tehnology. Data soures inlude USPTO patent dataset and EPO Worldwide Patent Statistial Database April The tehnology 42