Patent office governance and patent examination quality

Journal of Public Economics 104 (2013) 14–25

Contents lists available at SciVerse ScienceDirect

Journal of Public Economics journal homepage: www.elsevier.com/locate/jpube

Patent office governance and patent examination quality☆ Pierre M. Picard a, b, Bruno van Pottelsberghe de la Potterie c,⁎ a b c

CREA, University of Luxembourg, Luxembourg CORE, Université catholique de Louvain, Belgium Université Libre de Bruxelles (ULB), Solvay Brussels School of Economics and Management, iCite, ECARES and CEB, Belgium

a r t i c l e

i n f o

Article history: Received 22 July 2011 Received in revised form 14 April 2013 Accepted 23 April 2013 Available online 1 May 2013 JEL classifications: L30 O30 O31 O34 O38

a b s t r a c t The present paper discusses the role of quality in patent examination process from the perspective of patent offices' behavior and organization. After documenting original stylized facts, the paper presents a model in which patent offices set patent fees and the quality level of their examination process. A higher effort in the examination process enhances the patent holders' protection in the judicial system and strengthens the screening of innovations with small inventive steps. We compare the quality of the examination process for various objectives of patent offices. Patent examination quality is the highest in an office maximizing incentives to innovate and the lowest in that maximizing the number of granted patents. A rent-seeking patent office can provide good incentives to innovate if it does not set too high markups on fees. © 2013 Elsevier B.V. All rights reserved.

Keywords: Patent system Quality Intellectual property Public firm organization

Over the past century patent offices have evolved to very large and heavy administrations. To put some perspective, when the Venice republic passed the first-known law on intellectual rights in 1474, it organized a ‘general welfare board’ gathering some government officials and wise citizens to examine the usefulness of the inventions for the city. Similarly, in the US, after the Congress passed the first US patent law in 1790, the patent applications were initially reviewed by Thomas Jefferson who, at the time, was the Secretary of State and a prolific inventor. Jefferson simply passed the documents of accepted innovations to the Secretary of War for review and then gathered the signatures from the Attorney General and the President George Washington. Since then the rate of innovation and the stock of knowledge have drastically increased worldwide. As a consequence, the patent examination process requires large organizations, patent offices, with thousands of highly specialized professionals who search and examine the inventions

submitted by universities and independent inventors, and an ever increasing number of firms engaged into industrial research and development. 1 For most economists, the central debate about patenting focuses on the protection of intellectual property, where a balance must be found between dynamic efficiency effect (worth to the public) and the static inefficiency effect (embarrassment) caused by the market power that patent systems grant to innovating firms.2 However, the constant growth in patent filings and the consequent backlogs (longer pendency during the examination process) have drawn the attention of many observers on the nature and the cost of the examination process. In particular, it is questioned whether examiners should spend less time on the examination of patent files and whether the resulting lower quality and cost of the patent process is socially valuable. The question is of particular interest as two of the largest patent offices in the world, namely the US Patent and Trademark Office (USPTO) and the European Patent Office (EPO), have adopted opposite strategies. In the former the examination process

☆ The authors are grateful to Paul Belleflamme, Yassine Lefouili, Patrick Legros and Gaïetan de Rassenfosse for their useful comments and remarks. This project has been supported by the grant F2R-CRE-PUL-10EGQH at the University of Luxembourg. We thank the editors and two referees for their improvement suggestions. ⁎ Corresponding author. E-mail addresses: [email protected] (P.M. Picard), [email protected] (B. van Pottelsberghe de la Potterie).

1 To fix ideas, the USPTO receives more than 500,000 patent applications each year, of which a large percentage will be granted and potentially enforced. About 2 million patents are currently in force in the US. 2 Thomas Jefferson (1794) penned perhaps one of the best-known maxims: “Patents should draw a line between the things which are worth to the public the embarrassment of an exclusive patent, and those which are not. Patents are, after all, governmentenforced monopolies and so there should be some ‘embarrassment’ (and hesitation) in granting them.”

1. Introduction

0047-2727/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jpubeco.2013.04.009

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25

is cheap and fast, with a low-to-medium relative quality of the examination process, whereas in the latter the examination is slower and more expensive, with a medium-to-high quality. The Japan Patent Office (JPO) is in an intermediate position (van Pottelsberghe de la Potterie, 2011). According to Lemley (2001) the USPTO performs ‘rationally’ a low quality examination, with many patents per examiners. The author argues that only limited resources should be allocated to the examination process, because only a small share of patents are worth it economically. Indeed, the USPTO tackles about three times more applications per year than the EPO, with approximately the same number of examiners. However, more patents have been erroneously granted and fostered the perception of a bad quality in the US patent system, as convincingly demonstrated by Jaffe and Lerner (2004), amongst others. The objective of this paper is to investigate the relationship between the governance of patent offices (in terms of organization and resources) and the quality of patent systems. In the present paper this latter concept is defined as the stringency of the examination process. As a point of departure one has to assume that the degree of quality in a patent system is important for patent holders. As emphasized by Ayres and Klemperer (1999), a patent is best viewed as a ‘probabilistic’ property right that gives the patent holder the right to sue potential infringers and a fair chance either to win the litigation in court or to reach a favorable agreement. So, it can fairly be assumed that the quality of the patent selection process reduces the uncertainty associated with the effective exploitation of granted intellectual property rights. King (2003) indeed shows that the examination hours spent on patent examination are statistically correlated with lower patent litigation activity. As a result, more examination time potentially lowers the transaction costs associated with enforcement of intellectual property rights. The present paper relies on this assumed positive relationship between the degree of quality in a patent system and the level of certainty associated with the granted patent. The paper puts forward a model where each invention is supposed to be associated with an inherent and idiosyncratic uncertainty associated with the testing of the novelty with respect to the prior art, and a sufficient degree of inventiveness, called the inventive step at the EPO and non-obviousness at the USPTO. Some inventions are obvious, others are not. So, even with a comprehensive knowledge of prior art, effort and time, an examiner faces the risk of wrongly granting a patent. A high quality patent system can therefore be defined as the one that discriminates between inventions keeping the same inherent and idiosyncratic error risk. A medium quality patent system makes more errors and grants more than necessarily monopolistic rights to applicants. A zero quality patent system would always grant a patent and would simply consist in stamping and recording the patent application files. Following King (2003), it can be assumed that a better quality patent system is likely to be taken more seriously at the litigation stage by the Court. With a zero quality examination system, the court will give no advantage to a patent holder and the patent will have no or little legal value. By contrast, if the patent office offers an excellent quality, the Court will more easily uphold the patent in case of a validity challenge by infringers, and will more likely rule the litigation in favor of the patent holder. In practice the link between the quality of patent examination process and patent values is confirmed by the fact that US patent owners frequently ask for re-examination of previously granted patents by the USPTO, in order to strengthen the certification offered by the initial patent and therefore increase the bargaining power during settlement or licensing negotiations. To sum up, the quality of a patent examination process would contribute to the credibility of the certification process and therefore to the private value of patents. This quality/credibility relationship is also developed by Bessen and Meurer (2008), Jaffe and Lerner (2004), and Quillen (2008). Finally, the production function of patent offices, and hence examiners' behavior, also need to be formalized. In particular, it is assumed

15

that the examination process is intrinsically subject to uncertainty in the sense that the inventiveness and breadth of some innovations cannot be assessed with certainty. Given the properties explained above, the paper explores the behavior of patent offices that can set the fee and the quality of the examination process under budget neutrality. This paper sequentially analyzes offices run by benevolent social planners and compares them with the offices managed by directors who follow politician objectives such as increasing the number of patent applications or the number of granted patents. 3 The paper also considers the hypothetical case of offices with rent seeking objectives. The results suggest that the quality levels are too low under politician structures but may be appropriate in the social planner and monopoly settings. 1.1. Related literature This paper focuses on the role of patent offices' governance and organizational structure in the setting up of quality in patent systems. In this respect, it departs substantially from the mainstream ‘patent’ literature of the past 50 years. Early theoretical investigations into the role of patent systems originated with Barzel (1968), Nordhaus (1969) and Scherer (1972), who argued that stronger patent systems would induce more investment in research and development. Following these early theoretical investigations, most landmark papers have essentially focused on three major aspects of policy making: the optimal length, the optimal breadth (or the optimal combination of these two dimensions), and the optimal geographical scope of protection. For instance, Gallini (1992) analyzes the optimal length of a patent as a function of imitation costs. Klemperer (1990) examines the optimal scope of protection, whereas Gilbert and Shapiro (1990) identify the optimal mix between length and breadth of patents. Scotchmer (1991) explores how patent scope may affect the speed of generation and diffusion of new knowledge in a context of cumulative innovation processes. A patent protection that is too strong could lead to socially inefficient monopoly pricing and might stifle second-stage R&D. On the other hand, a too small inventive step could lead to ‘hold-up’ problems, whereby a patent granted for a small increment would actually provide more power to large resourceful imitators. So far, however, it seems that the status-quo has prevailed in the “examination field”, with little noticeable changes since the landmark contribution of Penrose (1951) and Machlup (1958): “If we did not have a patent system, it would be irresponsible, on the basis of our present knowledge of its economic consequences, to recommend instituting one. But since we have had a patent system for a long time, it would be irresponsible, on the basis of our present knowledge, to recommend abolishing it”.4 Recent studies on the effectiveness of patent systems seem to flourish, apparently in correlation with the recent boom in patent applications and criticisms to it (Jaffe and Lerner, 2004; Guellec and van Pottelsberghe de la Potterie, 2007; Bessen and Meurer, 2008). The present paper takes an alternative – and novel – approach to analyze the effectiveness of patent systems and the degree of quality in the examination process. Few authors have discussed the optimal quality in patent systems. Lemley (2001) advocates for low standards of patent examination on the ground of the ‘rational ignorance principle’. As the vast majority of patented inventions do not lead to successful products in the market place, it may be preferable to ignore them and wait for the small number of contentions in the product market and organize their resolution in 3 For instance, the recently published intellectual property strategy of China has amongst its goals for 2015 a target of 2 million yearly applications of patents, utility models and design rights (cf. the document published in November 2010 by the State Intellectual Property Office of China, quoted by Steve Lohr, New York Times, January 1st, 2011). 4 This quotation is often attributed to Machlup (1958, p. 80). It however appears that Machlup was in fact quoting Penrose (1951), according to Bronwyn Hall and Josh Lerner.

16

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25

Court. This argument is contested by King (2003), who suggests that a low quality examination would increase the uncertainty associated with the granted patents in case of litigation. One of the few authors who explicitly consider patentability requirement is O'Donoghue (1998). His theoretical model suggests that more stringent selection criteria would create longer incumbency and, thereby, would raise innovation incentives. Dewatripont and Legros (2008) show that litigation threats contribute to reducing the propensity to file low-quality applications, while they also hinder the production of strong patents. One method of reducing this negative side effect would be to sharpen the filtering process. Farrell and Shapiro (2008) also emphasize the importance of filtering, as they find that determining patent validity prior to licensing is socially beneficial. The present paper models the trade-off between the cost of patenting and the uncertain benefit of the patented invention after the grant date. Caillaud and Duchêne (2011) and Langinier and Marcoul (2003) consider the complexity and cost of information processing during the examination process, which creates a social cost of granting bad patents. Their main concerns relate to the efficiency of ‘equal treatment’ of patent applications and to the possibility of bad equilibria in which patent offices are attacked by firms that file inefficiently high numbers of bad patent applications. Gans et al. (2004) model the self-funded offices and compare their outcome to the social optimal one. None of these papers discusses the impact of the objectives of patent office's governors and examiners, or the role of governance, in the setting-up of the quality level in patent systems. The paper is organized as follows. Section 2 presents new and original stylized facts about patent offices that motivate our analysis. Section 3 presents the model and Section 4 discusses the demand for patents. Section 5 studies the production of patents under various contexts and objectives. Section 6 concludes. 2. Stylized facts about patent offices This section presents stylized facts on differences between the U.S. Patent and Trademark Office (USPTO), the European Patent Office (EPO), and the Japan Patent Office (JPO). The five following tables subsequently describe the demand for patents, the examination resources, the workload of examiners, their incentives and the governance structure of patent offices. A first stylized fact is that there is a much higher propensity to file patents at the USPTO than at the two other offices. Table 1 displays the demand for monopolistic rights in the three offices, in both absolute and relative terms. The demand can be measured either with the total number of patents filed or with the number of claims included in these patents. The latter indicator takes stock of the patent size in terms of claims, which significantly differs across patent offices. With 448,000 applications in 2008 (nearly 11 million claims), the USPTO faces the largest number of filings. The demand for patents at the EPO is 51% of the demand at the USPTO (38% in terms of claims). Although the JPO and USPTO file

Table 1 Demand for monopolistic rights: patents and claims, 2008.

Patents filed (000s) Aver. number of claims per patent Total claims filed (000s) Patents filed per 1000 capita Claims filed per 1000 capita Claims in force per 1000 capita

EPO

JPO

USPTO

EU/US

JP/US

227 18 4083 0.4 9 13

391 9 3519 3.1 30 91

448 24 10,752 1.5 35 150

0.51 0.75 0.38 0.27 0.26 0.09

0.87 0.38 0.33 2.10 0.86 0.61

The number of claims in force per 1000 capita in Europe is based on the total number of claims in force in Germany (were more than 90% of the patents granted by the EPO are validated), relative to the EU population. Source: national patent offices, trilateral statistical report and information provided by patent offices for the number of claims (cf. de Rassenfosse and van Pottelsberghe de la Potterie, 2012). The number of capita comes from the OECD/MSTI database. The number of patent in force is from WIPO statistical series (2009).

a similar number of patents (391 and 448), the number of claims filed in Japan falls to 33% of the U.S. level because of a smaller coverage of patents in Japan. In other words, the demand for patent rights, as measured by the number of claims filed, is much higher at the USPTO than at the EPO or JPO, which both receive less than 40% of the USPTO yearly workload. As the EPO covers a much larger market than the USPTO or JPO, the demand per capita is probably a better measure. It is highest in the US and smallest in Europe, Japan being in an intermediate position (see the claims filed per capita in Table 1). The consequence of the heterogeneous propensities to rely on patent systems can also be gauged with the number of “claims in force” per capita in a geographical area, where then “claims in force” are measured as the number of previously granted patents that are maintained in the country. This measure is presented in the last row of Table 1. The number of claims in force per capita is much higher in the USA (about 150 claims per 1000 capita) than in Japan and Europe, which are respectively at 60% and 10% of the US level. This very low ratio for Europe is partly explained by a stricter selection process at the EPO (van Pottelsberghe de la Potterie, 2011) and by the higher costs that follow the decision to grant a patent. In particular, once a patent is granted it must be enforced in each desired country, implying translation costs, validation fees and renewal fees in each desired country (van Pottelsberghe de la Potterie and Mejer, 2010). A second stylized fact is that examination fees are larger in Europe (cf. Table 2). The cumulated fees from the filing to the decision to grant a patent are the most expensive in the E.U. and the cheapest in the USA and Japan. For instance, the examination of a patent filed at the EPO would cost about USD 9348 (more than USD 500 per claim), against USD 2325 at the JPO (USD 258 per claim) and USD 2426 at the USPTO (USD 101 per claim). Such differences have persisted for a significant period of time (de Rassenfosse and van Pottelsberghe de la Potterie, 2012). Because patent offices are “officially” non-profit organizations, the firms' cost of filing patents can also be assessed with the patent offices' revenues or budgets. Patent offices receive various payments like filing fees, examination fees, renewal fees and ancillary procedures. The offices' budgets per filed patent or claim are higher than the cumulated examination fees because the former include renewal fees and ancillary procedures. This is however less true for the EPO, which must process many applications under the Patent Cooperation Treaty (PCT), and which receives half the renewal fees collected by national patent offices in Europe (Danguy and van Pottelsberghe de la Potterie, 2011). However, both the fees per claim and budget per claim yield the conclusion that the EPO and the JPO allocate two to three times more examination resources per patent or per claim than the USPTO. A third stylized fact is that examiners at the USPTO process more patent applications than at the EPO or JPO. This is shown in Table 3, which provides alternative measures of the workload per examiner. The number of claims ‘searched’ (search for prior art) per examiner, and the number of claims ‘examined’ (examination for inventiveness) per examiner can be used as indicators of examiners' workload and as very rough approximation of their productivity. The figures must take into account the fact that in Japan the examiners do not perform the Table 2 Cumulated patent fees and office budget per patent, 2008. Source: Own calculation from annual reports of patent offices, de Rassenfosse and van Pottelsberghe de la Potterie (forthcoming) and van Pottelsberghe de la Potterie and Mejer (2010). 2008, USD

EPO

JPO

USPTO

E.U./US

JP/US

Cumulated fees per filed patent Cumulated fees per filed claim Budget per filed patent Budget per filed claim

9348 519 7527 418

2325 258 3042 338

2426 101 3627 151

3.9 5.1 2.1 2.8

1.0 2.6 0.8 2.2

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25 Table 3 Workload per examiner, 2008. Source: Own calculation from annual reports of patent offices and de Rassenfosse and van Pottelsberghe de la Potterie (2012) for claim numbers. Workload per examiner

EPO

JPO

USPTO

E.U./US

JP/US

Claims filed (inc. PCT) Claims ‘searched’ Claims examination request Claims granted

1055 1055 642 278

2095 Outsourced 1406 857

1776 1776 1776 613

0.59 0.36 0.45

– 0.79 1.40

Equivalent workload (claims searched + examined)

1698

1406

3551

0.48

0.40

search report, which takes at least 50% of the time of the average examiner at the USPTO and EPO. The ‘equivalent workload’ indicator, which adds the claims searched and the claims examined per examiner are similar in Japan and Europe (between 1400 and 1700 claims per examiner each year), but is less than 50% the workload of an average US examiner. In terms of claims granted, the average EPO (JPO) examiner ‘grants’ each year 45% (79%) of the number of claims granted by the average USPTO examiner. 5 In a nutshell, be it in terms of incoming workload per examiner, or outgoing (granted claims per examiner), the average USPTO examiner processes at least two times more patents or claims each year than the average EPO or JPO examiner. The above measures of workload or productivity do not seem to be positively correlated with incentives. Another fact is indeed that E.U. and Japanese patent offices offer more attractive compensations to their examiners. In 2008, a typical USPTO examiner earns a civil servant GS12 (step 5) gross salary of about 87,000 USD with some 24% locality adjustments that are subject to federal and state taxation. 6 By contrast, the EPO offers one of the most comprehensive and family-friendly benefits packages in Europe. The offer comprises an internationally competitive basic salary that is exempt from national income tax. The 2009 gross salary for an A4 step 2 examiner (there are 7 grades under the President, and 13 steps per grade) is about 120,000 EUR (or about 160,000 USD) and is not subject to state taxation. In addition, EPO employees receive excellent social security coverage, medical insurance, and pension and saving plans and benefit from several other attractive allowances. 7 In Japan a 35 year old examiner being married and having one child would earn more than 7 million Yen (or 81,772 USD) per year, subject to taxation. He/she also receives insurance packages and support for lodging expenses (through dormitories facilities or financial allowance for apartment rental, up to 334,000 Yen each year). The variety in compensation packages makes comparisons difficult. An alternative approach is presented in Table 4, which compares each office's compensation budget and reports it in terms of the numbers of examiners and total staff (the staff/examiners ratio lies at about 1.65 in all offices). This provides us with a higher and a lower bound for the examiners' gross compensation packages. To ease comparison, those bounds are presented in US PPPs. At the USPTO, the total compensation per staff of 106,272 US PPPs roughly corresponds to the gross salary mentioned above. Given that the tax wedge is

5 The JPO has a higher ratio than the EPO for granted patents partly because the grant rate of examination requests is higher in Japan than in Europe. This is due to the fact that applicants in Japan have three years to request an examination, against 18 months at the EPO. This longer period leads to more stringent self-selection process in Japan and hence to higher quality patents submitted for examination (see Yamauchi and Nagaoka, 2009). Since it takes less time to grant a patent than to refuse it, examiner in Japan turn out to grant more patent each year. 6 See http://www.popa.org/txt/salary2008.txt. 7 The EPO offers allowances for expatriation (16–20%), installation (a month salary), household (up to 6%), dependent children, child education (international schools fees), learning languages (fluency in English, French, German and other languages is frequent).

17

Table 4 Budget/incentive per examiner. Source: Own calculations from the annual reports of patent offices. US PPPs, 2008

EPO

JPO

USPTO

E.U./US

JP/US

Tot. compensation/examiner Tot. compensation/staff

286,514 165,779

342,914 198,654

167,052 106,272

1.72 1.56

2.05 1.87

much smaller for the EPO examiners, the figures in Table 4 confirm the fact that the EPO and the JPO offer much more attractive compensation packages than the USPTO. We may now attempt to draw some conclusions about the quality of the patent process. Combining the above information, it is more convenient to measure patent quality through the inputs in the patent office: work time and on-the-job experience. In a perfect labor market, the fact that EPO examiners process a smaller amount of patent applications and are better paid implies that EPO gives more time and more incentives to their examiners and that it allows for a higher quality in the patent examination process. This conclusion is corroborated by the staff turnovers and job experience in each office (van Pottelsberghe de la Potterie, 2011). Indeed, whereas the EPO and the JPO have a very small staff turnover, the USPTO has an annual staff turnover of about 30% in 2008. Although this difference can be partly explained by more attractive compensation packages, it also gives evidence of a much longer job experience of EPO or JPO examiners. This discussion may give ground to the idea that the patents granted by the EPO are of a higher quality in the sense that their assessment is made in a more rigorous way. The governance structure has also implications on the formulation of the institutions' objectives. In particular, the EPO is an international organization directed by a board (called European Patent Organization) that gathers representatives of national patent offices (or contracting States). As an intergovernmental organization, the EPO is required to balance its budget because contracting States are obliged (but reluctant) to finance any deficit. The governance structure is prone to conflict of interest on the one hand because some national patent offices are somewhat competitors for patent filing and on the other hand because national patent offices receive the proceeds of renewals of the patents granted by the EPO. The board could therefore be criticized to favor quantity over quality of patents because this strategy is expected to raise national patent offices revenues and maintain their national demand for patents. Recently, the board has attempted to reduce the backlog and has promoted productivity measurement systems, by putting more pressure on the shoulders of examiners. The USPTO has a quite different governance structure. The USPTO is a federal agency of the U.S. Department of Commerce that is fully funded by the fees since 1991. It is led by the Director of the USPTO (who directly becomes Under Secretary of Commerce for Intellectual Property). This person is appointed by the President (heard and confirmed by the US Senate) and directly reports to the Secretary of Commerce. He/she is responsible for providing policy direction and management supervision for the USPTO and for the issuance of patents and registration of trademarks. Since 1991, USPTO directors were outsiders of the USPTO and have changed according to the changes in federal administrations. As a result, one may infer that USPTO directors are linked to current federal administrations and their political agendas. Similarly, the JPO is an agency of the Japanese Ministry of Economy, Trade and Industry and is led by a commissioner. To conclude, the above stylized facts establish that the EPO processes fewer patents and claims per capita and per examiner, pays higher compensations and provides a better patent quality. 8 The purpose of the following sections is to frame the relationships between 8 This relationship was empirically tested and validated by de Saint-Georges and van Pottelsberghe (2013), with 32 patent systems.

18

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25

those features. Towards this aim, we discuss a model of patent office where examiners produce patents with a quality level that affects the agents' confidence about the protection warranted by patents. 3. Model In this model we assume a single patent office and a set of firms (or individuals) that have an invention to patent and to potentially submit to the market place. Those agents enter into a two stage game. The first stage concerns those firms' choice to apply for a patent. The second stage concerns the life of the patent after its validation by the patent office. In the first stage, the patent office receives patent applications and asks its examiners to investigate whether the inventions described in the applications are patentable (i.e., not obvious and unknown in the existing body of knowledge, or prior art). The patent office sets a (uniform) fee f ≥ 0 and the examination effort (time) e ≥ 0 to process an (average) patent application. We assume that it costs nothing to deliver a patent without examination; that is, stamping, copying and archiving the application forms involve no examination cost. It however costs more in term of examiners' attention and training to check the novelty, the usefulness and the inventive step of patent applications. For the sake of simplicity, we assume that the unit cost of a patent application does not depend on the number of applications that are scrutinized by the patent office. In other words, there are constant returns to scale in the patent examination process. Each application therefore costs we where w is the examiner's competitive wage. Inventions differ with respect to their specific inventiveness probability b ∈ [0, 1], which captures the exogenous, inherent and idiosyncratic uncertainty that technical knowledge embodied in the invention is patentable, i.e. novel with respect to the existing knowledge (or prior art) and non obvious. In particular, an invention with b = 1 would be granted with probability one by an examiner who has very good knowledge, expertise and time to process the patent file. An invention with lower b would be granted with lower probability by the same examiner. For the sake of simplicity, we assume that a unit mass of firms is endowed with an invention and that the inventiveness b is distributed across inventions according
 to the cumulative and density distribution function ˜ b b and g(b) ≡ G′(b) > 0. For the sake of conciseness GðbÞ≡proba b and with some abuse of language we will shortly call b the ‘inventiveness’. Each firm is uniquely determined by its inventiveness and can simply be indexed by b. Each firm chooses whether to apply for a patent in the patent office. If it does, the firm pays the patent application fee f. In practice, judging an invention is a complex process and can be subject to assessment errors. 9 Examiners are more likely to grant patents to inventions with higher inventive step. Their examination and search for prior art are however imperfect so that some significant inventions may fail whereas some inventions with a weak degree of inventiveness can successfully pass the patent examination process. We assume that a firm with inventiveness b will have its patent granted with the probability p(e, b) where p(0, b) = p(e, 1) = 1. Under those assumptions, all applications are patented if the patent office implements no effort when it delivers the patents (p(0, b) = 1). The patent office is just stamping and archiving the application. Also, undoubtedly significant innovations get a patent with certainty (p(e, 1) = 1). In addition, we assume that stronger inventiveness increases the probability of getting a patent (pb ≡ ∂ p/∂ b > 0) and that higher examination efforts decrease the probability of success of patent applications (pe ≡ ∂ p/∂ e b 0) but decreases this probability less

9 In this paper we abstract from the modeling of information processing (as is done in Caillaud and Duchêne, 2011; and in Langinier and Marcoul, 2003) and rather focus on the governance structure of patent offices.

for more inventive ones (peb ≡ ∂ 2p/∂ e ∂ b > 0). 10 Note that, for the sake of our discussion, we abstract from trade secret discussions in the sense that firms benefit from the rent of their innovations if they become protected by a patent (e.g. Coca-cola secret formula). Innovations are perfectly replicable by competitors. Such a configuration of an invention without patent protection could still allow the inventor to appropriate rents, as illustrated by Anton and Yao (1994), but this case is not tackled in the present paper. In the second stage, the patent is granted by the patent office and the invention follows a standard patent life. For the sake of clarity, we retain the most important aspects of the patent life for the purpose of our discussion. Typically, we assume that the innovation creates a market value v and has a value for society s (consumer and producer surplus) and that its owner faces a potential competitor who challenges the firm's ownership over the invention, say with the exogenous probability α. There are many ways to challenge existing patents. The challenger may hold a patent that includes technical knowledge that precedes or is close to patented invention b; it may hold a patent that pertains to an application field unrelated to patent b; or it may hold no patent and claim that the patent is simply not valid. In any case, the main reason of the contention lies in the patent office's difficulty of ascertaining the exact inventiveness of the invention. The only way to solve this challenge is to go to court (or negotiate). Importantly we assume that the judge (or jury) is sensitive to the examination efforts e and is more likely to settle the case in favor of the firm holding the patent when the patent system is reputed to secure high quality examination services. For instance, Fisher (2010) observes that the validity of a patent is less often challenged in German Courts when it is delivered by the EPO, as opposed to those granted by other patent offices. Hence, the examination efforts by the patent office translate into patent strength, as in Farrell and Shapiro (2008); stronger patents are more likely to be validated in Court. We assume that the judge (jury) grants a winning case to the patent holder with a probability equal to P(e) that is an increasing function of the quality of the patent system (P′ > 0). We assume that defending the innovation in court has a lawyer's cost C.11 If the firm holds no patent, the judge (or jury) grants a winning case to the firm with a probability equal to P(0). Hence, the firm receives less protection when it does not receive or apply for a patent. In this model, the examination effort is clearly a synonym of examination quality and patent system quality because a higher examination effort gives a better credibility to the certification granted by the patent office. We may use those terms interchangeably. In the following sections, we first determine the demand for patent examination and then discuss the production of patents by patent offices according to various governance structures. 4. Demand for patents The demand for patents is obtained by solving the above game backward. The last stage can readily be solved as follows. Before the possibility of a challenge over invention ownership and before developing further its product, the firm has net benefit of holding a patent given by V ðeÞ≡ð1−α Þv þ α ½vP ðeÞ−C : In this expression, the first and second terms represents the value of the invention when it is respectively unchallenged and challenged. The function V(e) increases in e because P′ > 0. The net benefit of holding a patent augments with patent examination effort because 10 Under the highest examination effort (e = 1), examiners take decisions that are congruent with the inherent probability of a true positive of each innovation. Under low effort (e = 0), examiners take decisions that are independent from the inherent probability of a true positive of each invention. 11 This is the cost of hiring lawyers. In this text, this cost is supposed to be constant for the sake of clarity.

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25

judges care about the certification level given by the patent office when they make their judgment. This would remain true if the two firms negotiated an outcome before going to court. This net benefit increases with higher values of innovation (higher v), smaller likelihood of being challenged (lower α), and smaller lawyer's cost (smaller C). When the firm holds no patent for its invention, it gets a net benefit equal to the latter expression where e is set to zero. The first stage of the game allows us to establish the demand for patents. A firm with inventiveness b receives a patent with the probability p(e, b) and has an expected profit equal to p(e, b)V (e) + (1 − p(e, b)) V(0) when it files an application. By contrast, it gets V(0) when it does not file an application. Hence, the firm b applies for a patent if the former profit is larger than the latter. That is, if the private benefit of submitting a patent F ðe; bÞ ≡ pðe; bÞ½V ðeÞ−V ð0Þ

ð1Þ

is larger than the patent fee f. In this expression, we have Fb ≡ ∂ F/∂ b > 0 and Feb ≡ ∂2F/∂ e ∂ b > 0 as pb, peb and V′ are all positive. For simplicity, we further assume concavity in examination effort, Fee b 0. This assumption ensures an interior solution in the choice of the office's examination efforts. ˜ Condition
(1)  determines a unique inventiveness threshold b ˜ ¼ f ) such that a firm with inventiveness b applies if (solving F e;b ˜ The demand for patent is then equal to and only if
b≥ b. ˜ It therefore ˜ . The demand falls with the threshold b. Dðf ; eÞ≡1−G b ˜ =df ¼ 1=F b > 0. It may however fall or falls with the fee f because d b rise with patent examination effort as the sign of ′

˜ db F p p V ðeÞ ¼− e ¼− e− de Fb pb pb V ðeÞ−V ð0Þ is ambiguous. On the one hand, an increase in patent examination effort accentuates the screening of submitted applications, which de˜ (first positive ters low innovative patent applicants and increases b term). On the other hand, it offers a better protection in the judicial system and increases the net benefit of holding a patent, which re˜ (second negative term). Hence patent quality creates two opduces b posite forces of patent screening and patent protection. Note that only the second force depends on the value of innovation v. Finally, the expected number of patents granted by the patent office is given by 1

Bðf ; eÞ≡∫ pðe; bÞdGðbÞ: b ˜ The number of patents granted is then a decreasing function of the fees f. It is smaller than the number of applications (B(f, e) b D(f, e)) because p(e, b) ≤ 1. 5. Production of patents We compare four benchmark organizations of patent offices. The first organization is controlled by a social planner, the second by a profit maximizing manager and the two last ones by politicians who maximize either the number of patent applications or the number of patents issued. 5.1. Social planner Patent systems are designed to preserve the firms' incentives to innovate. Innovations lead to novel products as well as higher quality and less costly goods, which ultimately benefit consumers. Incentives to innovate are driven by the profits that firms expect from their research investments. A key point however is that firms make their research investments before they realize the innovativeness and market potential of their expected innovations. In this section we first focus on the benchmark case of a social planner who maximizes those

19

incentives to innovate. We then discuss the case where the planner considers the consumer benefit and the externalities in the judicial system. 12 Consider first the incentives to innovate of the firms that make research investments before knowing the innovativeness parameter b of their inventions. They get this piece of information only after completing their research but before applying to the patent office. At that moment, their expected profit is simply equal to the net surplus F(e, b) + V (0) − f when they get a patent and to V(0) if they do not. We assume that the social planner shares the same information about each firm's inventiveness parameter b and takes the judicial system as given.13 She chooses the patent examination effort e and selects the profile of the innovations to examine θ(b) ∈ {1, 0} that maximize the firms' expected net surplus: 1

∫0 fθðbÞ½F ðe; bÞ þ V ð0Þ−we þ ð1−θðbÞÞV ð0ÞgdGðbÞ

ð2Þ

where the social cost of an application is equal to its labor cost we. The optimal examination profile θ o(·) and patent examination effort e o is determined as follows. On the one hand, using pointwise maximization of the above objective function, one readily finds that the planner chooses to examine all applicants (θ o(b) = 1) that satisfy the condition F(e, b) ≥ we. Since Fb > 0, the condition  o o o F e ; b ¼ we

ð3Þ

yields a unique threshold bo above which inventions are examined. In other words, the social benefit of examining a patent should be above its cost. On the other hand, the optimal examination effort e o is given by the first order condition 1

∫bo

  o   F e e ; b −w dGðbÞ ¼ 0:

ð4Þ

The marginal benefit of submitting a patent should be equal on average to the office's marginal examination cost. Because Fee b 0, this condition is necessary and sufficient. The optimal allocation (e, b) is denoted by the superscript o. Fig. 1 depicts the planner's allocation in the space (e, b). The optimal allocation (e o, b o) is determined by the intersection of the loci of conditions (3) and (4). We show in the Appendix A that the locus of Eq. (3) (bold curve) is a function that accepts a minimum when it crosses the locus of Fe − w = 0, which is itself a monotonically increasing function (dashed curve). The U-shape of locus (3) reflects the trade-off between patent screening and protection. When the patent office's budget is balanced (f = we), an increase in patent examination effort e implies a rise in the level of fees. However, for low efforts e, it also raises judicial protections so that less innovative firms apply and push the innovativeness threshold b down. By contrast, for high examination efforts, it strengthens patent screening and therefore lowers patent attractiveness, which ultimately pushes the innovativeness threshold b up. Finally one can show that the locus of Eq. (4) is also an increasing function (thin curve) and lies to the right of Fe − w = 0. To sum up, the planner sets the innovativeness threshold and the patent examination effort such that the labor cost of examining a patent is equal to the social benefit of submitting a patent and to its average marginal private benefit. Yet, the planner may also consider the 12

We thank a referee and the editor for this suggestion. We assume informational symmetry between the social planner and the innovators for the sake of clarity. In fact, the social planner can implement the social optimal allocation under asymmetric information when it has no information on each patent applicant but has information on the innovativeness distribution G(·). This is because the planner is shown to be able to decentralize its choice by setting the patent fee f, which depends only on the distribution G(·). 13

20

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25

Fe ( e, b) = w (5)

b

1

∫ [F (e, b) − w] dG(b) = 0 e

(4)

bo

F ( e, b) = we (3)

o * *’

e Fig. 1. Socially optimal production of patents.

potential externalities, such as those affecting consumers and the judicial system. We can readily extend the above analysis for a planner who considers the social value of innovation, which includes the benefit to the consumers and the externalities in the judicial system. Indeed, each innovation brings an economic surplus above the private surplus while each challenge in court includes a cost to the judicial system and the challenging firm. The social net benefit of holding a patent is then given by V ðeÞ≡ð1−α Þs þ α ½sP ðeÞ−C−K ðeÞ where s is the sum of producer and consumer surplus of the innovation (s ≥ v) and K(e) is the cost to the judicial system and the challenging firm. This is the expression of V(e) where the private surplus v is replaced by s and the private cost of a trial challenged is increased by K(e). This cost K(e) is assumed to fall with e as longer examination processes provide more confidence in the true innovativeness of the patent and should reduce the length of a trial and the lawyers' costs. As a result, V ðeÞ is an increasing function of e. The social net benefit V ðeÞ is larger than the private net benefit V(e) if and only if s is sufficiently larger than v and/or if K(e) is sufficiently small. The social benefit of examining a patent can then be defined as F ðe; bÞ≡pðe; bÞ½V ðeÞ−V ð0Þ which increases in e and b. For simplicity, we assume that F is endowed with the same properties as F. The planner's objective is then the same as expression (2) where F is substituted for F . The optimality conditions (3) and (4) therefore apply with F substituted for F . We denote the resulting allocation by the superscript ⁎. By Eq. (1), the social planner can decentralize this allocation by setting the fee f ∗ = F(e ∗,b ∗). However, since F differs from F , this fee does not generally match the new first order condition (3):  F ðe ; ; b Þ ¼ we . The patent office then makes a loss and needs to   be subsidized if and only if f ∗ ≤ we ∗. That is, if F ðe ; b Þ ≤ F ðe ; b Þ   and thus V ðe Þ≤V ðe Þ. This will occur if the social surplus of inventions is sufficiently higher than their private surplus and/or if the cost in the judicial system is low. As a result, one must infer that self-funded patent offices cannot generally follow the objective to maximize the social incentives to innovate when those include externalities.   This occurs only in the very special case where F ðe ; b Þ ¼ F ðe ; b Þ. The discrepancy between social and private return to innovative activities, and its implication for patent policy, has been investigated since the seminal work of Arrow (1962). Compared to this literature, our setting includes an additional discrepancy caused by judicial costs. In the absence of empirical estimates of those effects, we present the following discussion. In Fig. 1, the allocation (e⁎, b⁎) is represented by the point indexed

by a star *. This allocation is determined by the intersection of the loci of conditions (3) and (4) in which F is substituted for F and which are not displayed for the sake of clarity. It obviously depends on how the externalities shift those two loci. We show in Appendix A that a rise in s or K increases V and F . A rise in social surplus s indeed entices the planner to examine innovation with lower innovativeness. A rise in judicial cost K entices the planner to raise the patent examination quality in order to reduce further socially costly litigation processes. So, this is equivalent to a fall in the wage w in condition (3) and implies a downward shift of the respective locus in Fig. 1. However, a rise in s or K has ambiguous effects on the marginal social benefit of examining a patent F e . A rise in s increases F e if and only if ðd=dsÞlnV ′ ðeÞ > ðd=dsÞln½V ðeÞ−V ð0Þ. The same condition applies with K. When the latter condition holds, a rise in s or K is equivalent to a fall in the wage w in condition (4) and implies a downward shift of the respective locus in Fig. 1. The new optimal allocation (point *) lies in the south-east corner of the optimal allocation without considering externalities (point o). By contrast, when the last condition does not hold, the same rise in s or K implies an upward shift of the locus (4) and the allocation (point *′) lies in the south-west corner of the optimal allocation without externalities (point o). In this case, the examination quality falls while the inventiveness threshold decreases and raises the number of patent applications. To sum up, this section has discussed the optimal allocation of patent examination quality of a benchmark office operated by a planner who promotes innovation. The planner sets the innovativeness threshold and the examination quality such that the labor cost of examining a patent is equal to the social benefit of submitting a patent and to its average marginal private benefit. When the planner considers the potential externalities to consumers and in the judicial system, the office may need to be subsidized. The impact on the examination quality and demand depends on specific functional forms. We now turn to the analysis of the monopoly patent office. 5.2. Monopoly patent office The patent office can be delegated to rent seeking agents. This would be the case if the patent office were delegated to a private firm (or manager) that maximizes the office's profit. Although there exists no example of such a pure profit oriented patent office, it is instructive to discuss the optimal quality chosen by such an organization. 14 As shown below, this setup is also interesting because it is equivalent to a patent office managed by its insiders. As discussed in Section 2, patent offices hire examiners with strong scientific and law backgrounds and require intense on-the-job training for technical expertise. As a special case, the EPO is an international organization that hires examiners with additional language competencies and therefore attracts workers with diverse cultural backgrounds in the districts of Munich and The Hague. Examiners can therefore be suspected to exert some power on the institution. We first show that delegations to private managers or to examiners yield the same outcome. On the one hand, private managers choose the patent fee f and examination effort e that maximize their profit where Π(f,e) = (f − we)D(f,e). On the other hand examiners choose their wage ω and examination effort e that maximize their surplus over their outside option. Assuming the competitive wage w as the outside option, we have that examiners maximize their total wage bill surplus (ω − w) eD(ωe, e). Since examiners have no incentives to leave a rent to their institution, they set the patent fee f equal to the labor cost ωe of examining a patent. The examiners' objective can then be written as (f − we) D(f, e), which is the profit maximized by the private managers. The patent fee and examination effort are therefore the same when examiners' outside option is their competitive wage. We can therefore focus on the analysis of a monopoly patent office. 14 The Swiss patent office could potentially be considered as such a monopoly patent office, but it does not have full control over its fees.

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25


 ˜ and the patent demand Using Eq. (1), the fee f is equal to F e;b
 ˜ D(f, e) is given by 1−G b . The profit can thus be re-written as
i h
 ih ˜ . Profit is maximized when the following ˜ −we 1−G b F e; b

21


 ˜ , office therefore maximizes its demand for patent Dðwe; eÞ ¼ 1−G b which amounts to minimize the inventiveness threshold bD that solves
 D D D F e ; b ¼ we :

ð7Þ

˜ hold first order conditions with respect to e and b  m m F e e ; b −w ¼ 0 

m

ð5Þ

m 

Fb e ; b

 m    m m  m  m 1−G b − F e ; b −we g b ¼ 0

ð6Þ

where the optimal values are denoted by a superscript m. On the one hand, the monopoly patent office increases its examination effort until the marginal cost of a more detailed examination (w) equals the increase in private benefit (Fe) of the last firm b m that submits a patent (first condition). On the other hand, the inventiveness threshold b m is increased as long as a stronger screening of inventions brings an additional private benefit to the examined inventions (Fb) that is larger than the firm's markup on the last invention examined (F(e m, b m) − we m) (second condition). The last condition can be reshuffled in the more traditional price–cost margin formula:  m f m −wem F b 1−G b ¼ fm F g ðbm Þ

ð> 0Þ

where the right hand side is simply the inverse elasticity of demand with respect to the fee. The margin falls when this elasticity increases; that is, when innovativeness has less impact on the private benefit of submitting a patent (smaller Fb/F) and when the mass of marginal applicants increases compared to the mass of inframarginal applicants (i.e. for smaller hazard rate [1 − G(b)]/g(b)). We can compare this outcome with the welfare benchmark case. The monopoly allocation (em, bm) is determined by the intersection of the loci of conditions (5) and (6). The locus of Eq. (5) has been shown in Fig. 1 (dashed line). The locus of Eq. (6) is such that F(em, bm) − wem > 0 so that it lies above the locus of Eq. (3), F(e, b) − we = 0, that we depicted in Fig. 1. As a result, the monopoly allocation (point m) lies in the North-West cone determined by the welfare benchmark allocation (point o). Note that the locus of Eq. (6) gets closer to the locus of Eq. (3) when the monopoly reduces its price–cost margin (fm − wem/fm). If the latter is not too high, the point m lies to the West of the point o and the monopoly patent office sets a lower quality than the planner. For even smaller margins, the point m will lie to the South-West of the point o so that the monopoly office sets a lower patent quality and innovativeness threshold (em, bm) compared to the social planner. This discussion is in accordance with Spence's (1975) analysis: a monopoly sets the quality of a facility according to the marginal benefit of its marginal user whereas a planner sets it according to the average of the users' marginal benefits. As a result, the monopoly patent office sets a too low patent quality. 5.3. Application oriented patent office

Differentiating totally expression (7) we get
 V ′ eD dbD F e −w pe p 1 w     ¼− ¼− − þ ¼0 Fb de pb pb V eD −V ð0Þ pb V eD −V ð0Þ

ð8Þ

The office raises its examination quality as long the firm's marginal cost from better patent screening (first term) is smaller than the benefit from better patent protection in the judicial system (second term) plus the cost of a more detailed examination (last term). Since Fee b 0, the condition expresses a minimum for b D and therefore a maximal patent demand D. The above discussion allows us to draw two conclusions. First, even though a politician oriented management does not formally value the patent examination quality per se, it implements a positive quality if V′(0) > w. Indeed, at e D → 0, V(e D) − V(0) tends to zero and p tends to one so that the sum of the last two terms in Eq. (8) tend to negative infinity under this condition. So, db D/de b 0 so that the office has an incentive to decrease the innovativeness threshold b through a rise in patent examination quality. It will do so if the first examination hour brings a net benefit of holding a patent, V′(0) higher than the examiner's hourly rate, w. A better examination quality increases the firms' value of a patent examination through a lower uncertainty associated with their intellectual property rights protection. The office indeed needs to set a low enough examination quality to help applicants to pass through the examination procedure but at the same time it needs to set a high enough examination quality to insure a good reputation for the patent office and a better patent judicial protection. Second, higher personnel costs w diminishes the demand for patent whereas it has ambiguous effects on the patent examination quality. Indeed, totally differentiating Eq. (7) and applying Eq. (8) yield db D/dw = e/Fb > 0 and de D/dw = (Fb − eFeb)/(FbFee) > 0 ⇔ dlnFb/ d ln e ≡ eFeb/Fb > 1, which occurs for high enough degree of concavity of the net benefit of holding a patent V(e). Higher examination costs entice the office to raise its fees and to lose some demand; to compensate for this loss, the office increases its examination quality when the last condition is satisfied. Personnel costs may thus be positively correlated with the examination quality even though the office's management puts no formal weight on this quality. We can finally compare this outcome with the welfare benchmark and monopoly case in Fig. 2. The allocation (e D, b D) of the application

b

(11)

Fe ( e, b) = w (5) 1

Most patent offices are self-financed. An unresolved issue is what objectives patent offices follow after they have balanced their budget. As discussed in Section 2, patent offices are generally (the EPO is one exception) government agencies run by politically linked managers. One may therefore assume that patent offices could follow politician objectives. Politicians often assess and praise the effectiveness of patent systems by the numbers of patent applications or/and by the number of granted patents. These two numbers are obvious, measurable outcomes of each patent office for which politicians are accountable to the public. They are also frequently used as measures (despite strong biases) of the country's industrial innovativeness. In this section we discuss the patent examination quality chosen by patent office governors who seek to maximize the number of patent applications. In such a patent office, the budget balance condition implies that the fees should be set to cover the examination cost so that f = we. The

∫ [F (e, b) − w] dG(b) = 0 e

bo

F ( e, b) > we (6) F ( e, b) = we (3)

m

B D

o

e Fig. 2. Production of patents.

(4)

22

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25

oriented governor is determined by the intersection of the loci of conditions (7) and (8). The locus of Eq. (7) is the same as the locus of Eq. (3) (bold curve). The locus of Eq. (8) is determined by the condition Fe − w = 0, which appear as the numerator to the right of the first equality in condition (8) and which has been established earlier in Figs. 1 and 2 (dashed line). As in the monopoly case, this locus corresponds to the quality that maximizes the private benefit of the marginal applicant. Hence, this allocation (e D, b D) (point D) lies at the minimum of locus of Eq. (3) and is located to the south-west of the welfare benchmark (point o) and to the south-west of the monopoly allocation (point m). Therefore, the patent examination quality e D is smaller than e o and e m while the innovativeness threshold b D is smaller than b m and b o. Quite naturally, the demand for patent is the largest. 5.4. Patent issuance oriented office Politicians also assess and praise the effectiveness of patent systems by the number of granted patents. In particular, the rate of invention has recently sharply increased in some sectors. Policy makers have therefore recommended to promote a certain form of ‘quality of service’ in order to fasten examination processes and thus increase the number of patents granted. For instance, in Section 2, we noted that EPO's governance structure is biased towards rewarding national patent offices. The EPO board includes representatives of the national patent offices, which collect the proceeds from renewals of patents granted by the EPO. The EPO board has thus some incentives to raise the number of granted patents. 15 Suppose here that the patent offices follow an objective measured by the number patents.
of granted    Under this objective, the office B maximizes B weB ; b ¼ ∫1bB p eB ; b dGðbÞ where b B solves
 B B B F e ; b ¼ we :

ð9Þ

The optimal quality is given by the first order condition

 dbB
 B B B 1 B þ ∫bB pe e ; b dGðbÞ ¼ 0: −p e ; b g b de From Eq. (9), we have db/de = − (Fe − w)/Fb so that the above condition can be re-written as −


 F e −w B B −Φ e ; b ¼ 0 Fb

ð10Þ

or equivalently

 V ′ eD pe p 1 w B B  D  D þ −Φ e ; b ¼ 0 − − pb pb V e −V ð0Þ pb V e −V ð0Þ

ð11Þ

where 1

Φðe; bÞ ¼ −

∫b pe ðe; zÞdGðzÞ pðe; bÞg ðbÞ

>0

represents the (absolute value of the) elasticity of granted patents with respect to the patent examination quality. As this quality falls, more patents are granted. This effect of patent head-count combines with the three other effects discussed above, namely patent screening, patent protection and the examination cost (the three first terms in Eq. (11)). Its existence gives incentives to the office to reduce its examination effort. Yet, for the same reason as above, such a politician oriented 15 See Danguy and van Pottelsberghe de la Potterie (2011) for evidence on how patent offices share their renewal fees income with the EPO in the European system.

management may still ask the examiners to implement a positive effort in their examination process. We can readily compare this equilibrium with the previous ones using Fig. 2. The allocation (e B, bB) of the patent oriented governor is determined by the intersection (point B) of the loci of conditions (9) and (11). As above, the locus of Eq. (9) is the same as the one of (3) (bold curve). The locus of Eq. (11) (dotted-dashed curve) lies to the left of the locus of Fe = w (dashed curve). Indeed, on the locus of Fe = w, the first term in Eq. (10) is nil so that this office has incentives to reduce patent examination quality. From this figure, it obvious that e B is smaller than eD, em and e B. This office diminishes its examination quality to reduce patent screening with the objective to raise the number of granted patents. Unsurprisingly, it offers the smallest examination effort amongst the above office structures. The innovativeness threshold bB is higher than bD but may be above or below bo and bm depending on the functional form of F. Compared to the application oriented office, this patent office implements a larger innovativeness threshold and has thus a smaller demand for patent. This is because the office reduces the patent protection in the judicial system so much that it deters the applicants with low innovativeness.

5.5. Summary and discussion The patent examination effort and the patent demand have been compared for the above benchmark organizations. Noting that fees f i are equal to the labor cost we i in self-funded offices, i ∈ {B, D, o}, we can similarly rank the patent fees. We summarize the main results in the following proposition. Proposition 1. The effort in the patent examination process is such that e B ≤ e D ≤ e o while the demand for patents is such that D D ≥ min {D o, D B, D m}. The fees are ranked such that f B ≤ f D ≤ f o. If the monopoly patent office does not set too large markups, the patent examination effort and the patent demand are such that e m ≤ e o and D o ≤ D m. The quality of patent examination process and the patent fees can be ranked under the general conditions of our model. The examination effort is set to its highest level by social planner and to its lowest by the patent office oriented towards patent issuance. The fees are obviously the highest in the office that maximizes incentives to innovate and the smallest in the office with a governor that maximizes the number of granted patents. Our analysis shows that the monopoly's examination effort is lower than the social planner's level but it does not allow us to determine the exact position of the monopoly's effort and fee in the above ranking. The latter depends on the realized markup of the monopoly patent office and therefore on the elasticity of patent demand. In particular the monopoly office has incentives to increase its fee fm over marginal cost wem to make more profit but at the same time it is enticed to reduce its patent examination quality compared to the welfare level eo because it cares only about the benefit of marginal patent applicants. A lower examination quality yields lower costs and incentive to reduce the fees. Hence the fee and examination quality cannot be ranked without additional model specifications. The same is true for the demand for patents that is more difficult to rank without more specific assumptions.16

16 For instance, Picard and van Pottelsberghe de la Potterie (2011) provides a specific model where patent quality, demands and numbers of issued grants and fees can be ranked. Under their specification, they confirm that the quality of a patent system is maximal for the social planner office, is smaller for the self-funded patent office maximizing the number of patent applications and even smaller for the self-funded patent office maximizing the number of granted patents. Under their specification, the monopoly firm sets so high markups that the patent quality is as high as the social planer's and the patent demand is lower in the monopoly patent office than in the social planner's office. Our present analysis highlights that the latter result depend on functional forms.

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25

In general, one expects that the objective of a patent office is a combination of the three objectives studied above. Hence, the patent examination effort and the innovativeness threshold can be represented by a point within the triangle BDm in Fig. 2. Its precise position depends on the weight placed on each objective. The quality of the patent examination process in office structures lying in this triangle BDm is therefore shown to be smaller than the socially optimal level that promotes innovation. A larger weight on profits or examiners' claims raises the patent examination effort and diminishes the number of applications. A larger weight on the counting of patent applications decreases the examination effort and augments the number of those applications. A higher weight on the number of granted patents decreases further the examination effort but is likely to also reduce the number of patent applications. We ultimately come back to the incentives to innovate. The economic literature on patent systems emphasizes the role of incentives to innovate in dynamic settings where firms invest in research before getting an innovative idea and asking for patent protection. In this paper we show that the incentives to innovate are optimal neither in the monopoly patent office nor in the offices oriented towards application or issuance. To see how distorted are those incentives, we can observe the points D and B that lie on the locus (3) in Fig. 2. Since incentives to innovate decreases on the latter locus as we move from o to D and then to B, the incentives to innovate are smaller with a patent office that promotes patent applications and even smaller in the one that favors patent issuance. Finally, the incentives to innovate under monopoly are also too low. How do they compare to the incentives with an office oriented toward patent demand? In Fig. 2, the incentives to innovate are larger under monopoly than those with the patent application oriented office if the point m lies within the iso-welfare locus that includes point D. Iso-welfare loci (thick gray ellipses) are curves that revolve around the social optimal point o and that have a vertical tangent at any point included on the locus of Eq. (3). This is because the latter locus maps the planner's first order condition with respect to b: vertical moves from any point on this curve bring no first order change in welfare. As a consequence, the iso-welfare crossing point D is orthogonal to the locus (3). By contrast, the monopoly allocation lies on the locus of Eq. (6) (i.e. Dm) that has a finite positive slope at D. It readily comes that the allocation m lies within the iso-welfare locus that includes point D as long as m is not too distant from D. This will happen when the monopoly office does not set too high markups on fees as the locus of Eq. (6) will remain close to the one of Eq. (3). As a result, the incentives to innovate are stronger with the monopoly patent office if the latter does not set too high markups. We summarize the last two results in the following proposition. Proposition 2. Incentives to innovate are too low in the patent office that seeks patent demands and even lower in the one that seeks patent issuance. A monopoly or examiner-managed patent office that does not set too high markups on fees provides better incentives to innovate than those two offices. 6. Conclusion The objective of this paper is to study the relationship between the governance of patent offices and the quality of the patent systems — defined here as the stringency of patent examination processes. The paper firstly establishes some stylized facts about the major patent offices in the U.S., E.U. and Japan. As a case in point, the E.U. patent office processes fewer patents and claims – in per capita and per examiner – pays higher compensations to examiners and devotes more resources per staff that potentially contribute to a better patent quality. The stylized facts suggest that compensation packages and examiners' power are discriminating elements of patent office behaviors. Also, while patent offices can influence their own targets, they are

23

managed by politically driven governors who must fund the offices with their own resources. Although patent offices are managed under different contexts and constraints, their objectives – be they explicit or implicit – most probably include the maximization of a combination of the numbers of patent applications, the number of granted patents and the examiners' interests. The paper then presents a model that describes the potential behaviors of patent offices with respect to the setting of their fees and the quality of their examination processes. These fee/quality settings depend on the four alternative objectives that patent offices' governors may target, namely, incentives to innovate, patent office's profit, number of patent applications and number of patents granted. As a point of departure, we assume that the patent system quality is important for patent holders. A patent is viewed as a ‘probabilistic’ property right that gives the patent holder the right to sue potential infringers and a fair chance either to win the litigation in court or to reach a favorable agreement. Therefore, a higher quality in patent examination lowers the transaction costs associated with enforcement of intellectual property rights and increases the demand for patents. The quality of the patent system contributes to the credibility of the patent certification process and hence to the private value of and the demand for patents by firms. The results of the theoretical model firstly show that the demand for patent examination increases with smaller patent fees but is a non-monotone function of the patent system quality. Indeed, a higher quality patent system implies a more rigorous selection of the inventions submitted by firms and a better enforcement of granted patents in the judicial system. Whereas the latter effect is good for all innovators, the former effect is bad for the innovators with ‘small’ inventions and reduces their incentives to apply for a patent. Hence, whatever its objective, a self-funded patent office faces a trade-off between softer examination and better credibility of the granted patents. The question is whether the governors and examiners have incentives to soften the examination process and add noise on the quality of the patent system. We review four types of governors' objective. On the one hand, we study the two hypothetical but interesting cases of a patent office that is run either by a planner who maximizes firms' incentives to innovate or by stakeholders who maximize their rent, be they shareholders, management or examiners. We show that the planner chooses the highest patent quality level and therefore does not add any noise on the examination process. By contrast, the rent maximizing office has an incentive to set a higher patent fee and decrease the patent quality to compensate for the fall in patent demand resulting for its fee. Indeed, this office targets the marginal applicant's benefit of patent quality rather than the firms' aggregate benefit and therefore prefers to cut on quality. On the other hand, we study the more realistic case of a self-funded office where the governor maximizes either the number of patent applications or the number of issued patents. We show that, even though such a governor does not formally value the quality of the patent system, it implements a positive quality level because a more rigorous patent examination process reduces the degree of uncertainty associated with the effective enforcement of the intellectual property rights. The office indeed decides to set a low enough quality to help applicants pass through the examination procedure but at the same time needs to set a high enough quality to insure a good reputation for the patent office and a better patent judicial protection. In such offices, higher examiners' wages (or resources devoted to the examination process) can increase with the quality of patent systems, even though the office's management puts no formal weight on quality. Comparing those four cases we show that the quality of the patent system is maximal in the patent office maximizing incentives to innovate, is smaller for the self-funded patent office maximizing the number of patent applications and even smaller for the self-funded patent office maximizing the number of granted patents. The patent quality under the rent seeking patent office is lower than the one that maximizes incentives to innovate as

24

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25

long as the rent seeking patent office does not set too large markups. As a result, incentives to innovate are too low in the patent office that seeks patent demands and even lower in the one that seeks patent issuance. A rent seeking office that does not set too high markups on fees can nevertheless provide better incentives to innovate than those two offices. The present paper addresses the issue of patent quality from the perspective of the patent offices' behaviors and organizations. It deliberately abstracts from the other – no less relevant – issues covered in the existing patent literature such as optimal length, scope and breadth. However, to the best of our knowledge, the literature has not properly covered the relationship between the offices' organization and the effective quality of the patent systems. Our main result is that offices offering high compensation packages like the EPO are likely to keep a high quality whereas offices offering lower compensation packages like the USPTO have incentives to soften examination process and introduce noise in the patent system, with the consequence of having more patent filed and granted. The discussion and results presented in this paper have their limitations as many additional features of patent systems have not been taken into account. The reality is indeed somewhat different and much more complex than suggested in our stylized model. Compensation packages, fee settings, budget components and even quality are composed of many facets that cannot be tackled in one model. 17 However, examiners must be managed, and their wage must be set. Last but not least, patent offices' governors often set priorities that implicitly or explicitly relate to the number of patent filings or patents granted. In the light of the stylized facts presented in this paper, it seems that the EPO gets closer to the model where the governor maximizes the number of patent filings and insiders' rents. The USPTO seems closer to the model of a patent office that maximizes the number of patents granted. This paper does not only offer a grid of analysis for existing patent offices but it also provides evidence about the relationship between, on the one hand, the governance of patent offices and, on the other hand, the quality of patent systems, its cost and the demand for patents. Our analysis ultimately sets the stage for further research on the quality of patent systems. In particular, it will be interesting to study the issue of patent quality in the context of competition between patent offices. The US and Japan patent offices indeed set lower examination stringency and attract a more than proportional share of patent applications and search requests from international firms. Similarly, it will be important to understand the introduction of a single European patent system in the context of countries with different patent system qualities and judicial procedures.

Appendix A We here characterize the properties of the loci displayed in Figs. 1 and 2. First, totally differentiating Eq. (3) we get db/de = −(Fe − w)/ Fb. Because Fee b 0 b Feb, the locus of Eq. (3) has a minimum at Fe − w = 0 since one computes d2b/(de) 2 = −Fee/Fb > 0 for Fe − w = 0. Second, since Feb > 0, we have that ∫ b1 [Fe(e,z) − w]dG(z) > [Fe(e, b) − w] ∫ b1 dG(z). So, by Fee b 0, the two conditions ∫ b1[Fe(e′,z) − w] dG(z) = 0 and Fe(e, b) − w = 0 imply that e′ > e. So, the locus of Eq. (4) lies to the right of Fe − w = 0. Third, the locus of Eq. (4) is an increasing function. Indeed, totally differentiating Eq. (4) gives db/ de = − ∫ b1Fee(e,z)dG(z)/[∫b1Feb(e,z)dG(z) − [Fe(e,b) − w]g(b)], which is positive because Fee b 0 b Feb and because the locus of Eq. (4) lies to the right of Fe − w = 0 so that Fe(e, b) − w is negative on this locus. 17 For instance the renewal fees paid for previously granted patents may be an important component of a patent office's budget. At the EPO the renewal fees correspond to approximately 25% of the total budget. An interesting extension would be to assess the extent to which the maintenance rate of patents depend on the quality of the examination process.

We now show how the social benefit of examining a patent, F , increases with a rise in s and K(Q) whereas F e has ambiguous slope. We first replace K(Q) by ηK(Q) and make variation analysis on η. We have V ðeÞ ¼ ð1−α Þs þ α ½sP ðeÞ−C−ηK ðeÞ and V ′ ðeÞ ¼ h i α sP ′ ðeÞ−ηK ′ ðeÞ > 0. Also, using F ðe; bÞ ¼ pðe; bÞ½V ðeÞ−V ð0Þ, we get dF =ds ¼ pðe; bÞ ½P ðeÞ−P ð0Þ > 0 and dF =dη ¼ pðe; bÞ½K ð0Þ−K ðeÞ > 0. Second we give the condition that F e increases if s or η rises. We write dlnF =de ¼ pe ðe; bÞ=pðe; bÞ þ V ′ ðeÞ=½V ðeÞ−V ð0Þ. Since only V de 2 pends on s, we get d lnF =ðdedsÞ > 0 if and only if ðd=dsÞ V ′ ðeÞ= ½V ðeÞ−V ð0Þg > 0; that is, if ðd=dsÞlnV ′ ðeÞ > ðd=dsÞln½V ðeÞ−V ð0Þ. The same argument applies for variations of η.

References Anton, J., Yao, D.A., 1994. Expropriation and inventions: appropriable rents in the absence of property rights. American Economic Review 84 (1), 190–209. Arrow, K., 1962. Economic welfare and the allocation of resources for invention. NBER: The Rate and Direction of Inventive Activity: Economic and Social Factors, pp. 609–626. Ayres, I., Klemperer, P., 1999. Limiting patentees' market power without reducing innovation incentives: the perverse benefits of uncertainty and non-injunctive remedies. Michigan Law Review 97, 985. Barzel, Y., 1968. Optimal timing of innovation. The Review of Economics and Statistics 50, 348–355. Bessen, J., Meurer, M.J., 2008. Patent Failure: How Judges, Lawyers and Bureaucrats Put Innovators at Risk. Princeton University Press, Princeton, NJ. Caillaud, B., Duchêne, A., 2011. Patent office in innovation policy: nobody's perfect. PSE Working Papers 2009–39, PSE (Ecole normale supérieure). International Journal of Industrial Organization 29 (2), 242–252. Danguy, J., van Pottelsberghe de la Potterie, B., 2011. Cost–benefit analysis of the Community patent. Journal of Benefit–Cost Analysis 2 (2), 3. de Rassenfosse, G., van Pottelsberghe de la Potterie, B., 2012. On the price elasticity of demand for patents. Oxford Bulletin of Economics and Statistics 74 (1), 58–77. de Rassenfosse, G., van Pottelsberghe de la Potterie, B., 2013. The role of fees in patent systems: theory and evidence. Journal of Economic Surveys. http://dx.doi.org/ 10.1111/j.1467-6419.2011.00712.x (forthcoming). de Saint-Georges, M., van Pottelsberghe de la Potterie, B., 2013. A quality index for patent systems. Research Policy 42 (3), 704–719. Dewatripont, M., Legros, P., 2008. Essential patents, FRAND royalties and technology standards. ECARES Working Paper 2008_010. Farrell, J., Shapiro, C., 2008. How strong are weak patents? American Economic Review 98 (4), 1347–1369. Fisher, T., 2010. Defending Patents' Validity — An Empirical Analysis of German Patent Invalidation Proceedings. Technische Universiteit Munchen. Gallini, N., 1992. Patent policy and costly imitation. The RAND Journal of Economics 23 (1), 52–63. Gans, G., King, S., Lampe, R., 2004. Patent renewal fees and self-funding patent offices. The B.E. Journal of Theoretical Economics, vol. 0(1). Berkeley Electronic Press. Gilbert, R., Shapiro, C., 1990. Optimal patent length and breadth. The RAND Journal of Economics 21 (1), 106–112. Guellec, D., van Pottelsberghe de la Potterie, B., 2007. The Economics of the European Patent System. Oxford University Press, Oxford (250 pp.). Jaffe, A., Lerner, J., 2004. Innovation and Its Discontents: How Our Broken Patent System is Endangering Innovation and Progress, and What to Do About It. Princeton University Press, New Jersey (236 pp.). King, J.L., 2003. Patent examination procedures and patent quality. In: The National Research Council (Ed.), Patents in the knowledge-based economy. The National Academic Press. Klemperer, P., 1990. How broad should the scope of patent protection be? The RAND Journal of Economics 29 (1), 113–130. Langinier, C., Marcoul, P., 2003. Patents, search of prior art, and revelation of information. Iowa State University, Department of Economics Staff General Research Papers No 10489. Lemley, M.A., 2001. Rational ignorance at the patent office. Northwestern University Law Review 95 (4). Machlup, F., 1958. An Economic Review of the Patent System. Report to the US government. Nordhaus, W., 1969. Invention, growth and welfare: a theoretical treatment of technological change. M.I.T. Monographs in Economics, 10. M.I.T. Press, Cambridge, MA (168 pp.). O'Donoghue, T., 1998. A patentability requirement for sequential innovation. The Rand Journal of Economics 29 (4), 654–679. Penrose, E.T., 1951. The Economics of the International Patent System. Johns Hopkins Press. Picard, P.M., van Pottelsberghe de la Potterie, B., 2011. Patent office governance and patent system quality. Center of Economic Policy Research DP8338. Quillen, C., 2008. Commentary on Bessen and Meurer's patent failure: an industry perspective. Journal of Intellectual Property Law 16 (1), 57–82. Scherer, F.M., 1972. Nordhaus' theory of optimal patent life: a geometric reinterpretation. American Economic Review 62, 422–427.

P.M. Picard, B. van Pottelsberghe / Journal of Public Economics 104 (2013) 14–25 Scotchmer, S., 1991. Standing on the shoulder of giants: cumulative research and the patent law. Journal of Economic Perspectives 5 (1), 29–41. Spence, M., 1975. Monopoly, quality, and regulation. Bell Journal of Economics 6 (2), 417–429. van Pottelsberghe de la Potterie, B., 2011. The quality factor in patent systems. Industrial and Corporate Change 20 (6), 1755–1793.

25

van Pottelsberghe de la Potterie, B., Mejer, M., 2010. The London agreement and the cost of patenting in Europe. European Journal of Law and Economics 29 (2), 211–237. Yamauchi, I., Nagaoka, S., 2009. Reforms of patent examination request system in Japan: some lessons. IIR Working Paper.