China Economic Review

The value of invention patents in China: Country origin and technology field differences Zhang GUPENG⁎, Chen XIANGDONG School of Economics and Management, Beihang University, China a r t i c l e i n f o a b s t r a c t Article history: Received 12 June 2011 Received in revised form 6 February 2012 Accepted 6 February 2012 Available online 15 February 2012 This paper improves renewal payment based patent value model by introducing unexpired patents into the model. We estimate value of the invention patents based on SIPO records in China and compare values of patents between local owners and owners from the U.S., Japan and European countries. The study reveals that patent value from Chinese owners is much lower than that of overseas owners. This larger value gap implies important difference in motive of the patenting and R&D quality between China and those technology intensive sources usually from economically advanced countries and regions. The model developed in this study is also applied to patent data in different technical fields, successfully differing tech- nical sectors with higher value (e.g., machinery) and those with lower value (e.g., pharmaceu- ticals), in terms of China market based patent records. Both applications prove that the newly developed model can be an important analytical tool for providing classified patent value on different purposes, especially in case of China. Crown Copyright © 2012 Published by Elsevier Inc. All rights reserved. JEL classifications: O31 O33 O34 Keywords: Patent value Renewal period Renewal fee Forward value 1. Literature review It is commonly believed that patented technology plays an important role in economic development and social welfare by balancing patent owners' exclusive benefits from technology within a limited time frame and knowledge spillover in society in the long run. Patents have their own effective lifecycles in terms of legal protection, which can be divided into full live terminations accord- ing to the patent system and patents becoming invalid when owners refuse to pay renewal fees. If the publicity of the invention information or the lifecycle in terms of knowledge spillover are considered, the patent lifecycle can be divided into two further categories: the provisional life (the time between the filing and the request for examination) and the active life (the time between the granting and the lapse of the patent) (Van Zeebroeck, 2007). Many international scholars focus more on the active life, or the legal protection period, based on the patent renewal information, as an important implication for patent value (Pakes, 1986; Schankerman and Pakes, 1986; Maurseth, 2005; and Bessen, 2008). Patent value is an interesting economic quantity for several reasons: it informs policy because it is a measure of the reward that the patent system provides inventors, and it helps to account for the value of intangibles and to measure the productivity and quality of R&D (Bessen, 2008). The basic argument is that the value of a patent can be revealed based on its owner's beliefs, by examining the corresponding renewal statue of the specific pat- ent, and apparently, the effective patent should be considered as worth more than the renewal fee required to keep it in force. Thus, it is reasonable to measure the patent value over its renewal payment during a certain period. Such research has been in- creasing in recent years, including studies by Pakes (1986), Pakes and Schankerman (1984), Schankerman and Pakes (1986), Schankerman (1991, 1998), Putnam (1996), Lanjouw (1998), Barney (2002), Serrano (2005), Baudry and Dumont (2006), Yi (2007) and Bessen (2008). China Economic Review 23 (2012) 357–370 ⁎ Corresponding author at: Xueyuan Road No. 37, P.S. 100091, Beijing, China. E-mail addresses: zhanggupeng@163.com (Z. Gupeng), chenxdng@gmail.com (C. Xiangdong). 1043-951X/$ – see front matter. Crown Copyright © 2012 Published by Elsevier Inc. All rights reserved. doi:10.1016/j.chieco.2012.02.002 Contents lists available at SciVerse ScienceDirect China Economic Review The original model developed by Pakes and Schankerman (1984), Schankerman and Pakes (1986) provides measurement methodologies concerning the probability distributions of patent renewal payments by certain groups. The measurement is based on the requirement that patent holders pay renewal fees to keep their patents in force. The researchers assume that renew- al decisions are based on purely economic criteria and that patent owners only renew their patents if the expected value of hold- ing those patents for an additional year is sufficiently higher than the cost of renewal (Pakes, 1986). Many researchers estimate corresponding models from patent renewal information that recover the distribution of returns from holding patents at each stage of the patents’ lifespan. This estimation principle enables patent evaluation to be directly connected to the value placed on particular patents or grouped patents by the patent holders. Lanjouw (1998) and Baudry and Dumont (2006) further expand- ed the original model with more detailed conditions. For example, Lanjouw (1998) included infringement during the patent life in his version of the model. Baudry and Dumont (2006) rebuilt the patent value model by introducing more random components. Additionally, Sullivan (1994), Bessen (2008) and Grönqvist (2009) produced excellent work when they simplified the original setting of Schankerman and Pakes (1986) by introducing less random components and controlling factors to reflect the patents’ nature. Because patents are heterogeneous in nature, e.g., with different characteristics or competing behaviors by their owners, es- pecially in the fields of technology or with technology-based controlling effects, patent values can differ significantly among owners and technical fields. International studies of patent values based on renewal payments also tend to focus on the correla- tions between patent characteristic variables and patent value measurements, such as the firm market value (Chen and Chang, 2010; Hall et al., 2005), litigation statuses (Allison et al., 2004; Harhoff and Wagner, 2003; Lanjouw and Schankerman, 2004; Marco, 2005), more subjective survey measures of patent quality (Harhoff and Wagner, 2003; Harhoff et al., 1999, 2003), or the nature of the patent family (Lanjouw and Schankerman, 2004; Putnam, 1996). Apparently, these typical correlated variables of patent value can be used as important references or benchmarks to establish the final value more efficiently. Furthermore, among other important empirical studies, with a simplified version of the patent value model combined with the correlated factors, Bessen (2008) and Grönqvist (2009) estimate patent value over renewal data and controlling variables on typ- ical patent nature. These representative empirical studies of patent value estimation are listed in Table 1. However, current studies on patent value focus only on the information of expired patents. Although there are related case studies of both expired and unexpired patents, such as the work of Van Zeebroeck (2007), Maurseth (2005), Svensson (2007), Nakata and Zhang (2009), and Xie and Giles (2007), these studies do not focus on patent value but possible determinant factors of the patent length, or correlated factors over effective patent lifecycles (e.g., patent citations, by Van Zeebroeck, 2007; Maurseth, 2005; Nakata and Zhang, 2009; the nature of the commercialization of patents, by Svensson, 2007; the size and type of the appli- cant, by Svensson, 2007; Van Zeebroeck, 2007; Nakata and Zhang, 2009; Xie and Giles, 2007; the patenting source countries, by Svensson, 2007). These studies are important references for patent value studies when corresponding influencing factors are taken into account. However, the survival analysis technique in these studies can also be a unique reference for further patent value studies in terms of the equal consideration of unexpired patents. Although almost all the literature on patent value in international journals focuses on expired patents, based on the idea that information from expired patents can be used as the best indicator for the values of patents with varied term lengths, these stud- ies overlook patents that are currently in effect, and these patents are more relevant to updated technology markets. Given that that the history of the Chinese patent system is short, with only 25-year term lengths since 1985, it would be necessary for a pat- ent value estimation to include both expired and unexpired patent records. In fact, according to our brief calculations, only 1/5 of invention patents in China are expired in total. In this case, if only expired patents are included in the patent value studies, a shortage of correlated variables and a limited number of patenting sources may cause bias in the estimation (Davidson and MacKinnon, 2004; Cameron and Trivedi, 2005). For example, smaller business owners may occupy a larger share within these limited samples of expired records. Table 1 Main empirical researches on patent value with renewal period. Literature Patent data Barney (2002) U.S. (1986, 1996) Bessen (2008) U.S. (1985–1991) Yi (2007) E.U. (1978–1996) Serrano (2005) U.S. (1983–2002) Putnam (1996) U.S. (1974) Japan (1974) Germany (1974) France (1974) England (1974) Pakes (1986); Pakes and Schankerman (1984), Schankerman and Pakes (1986); Schankerman (1991, 1998) France (1951–1979, 1969–1987) England (1950–1974) Germany (1952–1972) Lanjouw (1998) Germany (1953–1998) Baudry and Dumont (2006) France (1970–2002) Grönqvist (2009) Finland (1971–1990) Sullivan (1994) Britain and Ireland (1852–1876) 358 Z. Gupeng, C. Xiangdong / China Economic Review 23 (2012) 357–370 The current patent value studies based on renewal information of the patent records are summarized in the following chart. This study focuses on Chinese market-based invention patent records, and by developing and modifying a corresponding pat- ent value model, the study includes both expired and unexpired patent records in China to estimate patent value among different owner groups and different technical fields. The corresponding contents of the paper are as follows: Section 2 presents a modified model of patent value; Section 3 provides a basic structure of the patent system in China in terms of the effective life cycle of typ- ical patents, including the renewal fee payment structure; independent variables in a related model will be introduced subse- quently; research findings are discussed in Section 4, and the conclusion appears in Section 5. 2. The research model 2.1. A model of patent renewal The framework of patent values based on renewal information was initially proposed by Pakes and Schankerman (1984). While patent owners could achieve exclusive benefits from their own patents, as long as patent rights exist, it is frequently not easy to adopt patents for real production or to license patent rights for successful returns. Generally, if the expected benefit is less than the annual renewal fee, the patent owner will stop paying the annual renewal fee, and the patent rights will naturally expire. This process implies that the renewal decisions or renewal statuses of patent owners reflect information about their expected benefits from their patents in the near future. Let Ri(t) be the flow of rents for patent i at time t. Because the knowledge content of patented technology is fixed, the value of patented technology is usually considered to diminish because new technology will continually force the old patented technology Determinant of patent value (Harhoff et al., 2003; Lanjouw and Schankerman, 2004 etc) Original model of patent value (Pakes and Schankerman, 1984; Pakes, 1986) Introduction of su rvival analysis Sim plified m odel Expanded m odel Determinant of patent life with both expired and unexpired patents (Maurseth, 2005; Svensson, 2007; Van Zeebroeck, 2007) Improved model with patent characteristics (Bessen, 2008; Grönqvist, 2009) Improved model with litigation and more random settings (Lanjouw, 1998; Baudry and Dumont, 2006) Improved model of patent value of both expired and unexpired patents(This research) Fig. 1. Researches on patent value. Fig. 2. Comparison of Kaplan-Meier survival curve by owners among different countries — patent records in China. 359Z. Gupeng, C. Xiangdong / China Economic Review 23 (2012) 357–370 to be depreciated in technical value or discounted with a fixed rate, as demonstrated in studies by Bessen (2008) and Maurseth (2005), where Ri(t) is discounted with a fixed rate of d . Thus, Ri(t)=Ri(0)e−dt. Furthermore, let initial rent Ri(0) be a function of Xi, which is the vector denoting the heterogeneous factors of the patent. Thus, lnRi 0ð Þ ¼ Xiβ þ εi where εi~N(0, σ2), hence lnRi(0)~N(Xiβ, σ2). Xi is set to include typical factors influencing patent value, such as fields of technology, the inventors’ or owners’ nationality, country differences in patent systems, etc. (Bessen, 2008; Grönqvist, 2009; Schankerman, 1998; Schankerman and Pakes, 1986; Yi, 2007). However, in many cases, Xi is set as a constant, such as in the studies of Putnam (1996) and Harhoff et al. (2003). In this case, only d, σ and β need to be estimated. The value between t and t+1 with the payment time span t is ∫ tþ1 t Ri τð Þe−s τ−tð Þdτ ¼ Ri 0ð Þzt ;wherezt ¼ e−dt 1−e− dþsð Þ dþ s s is the discount rate in terms of the financial market (which is different from the depreciation of technical value), and based on normal treatments in current international studies, we set s=0.1. Then, the necessary and sufficient condition that the patentee pays the renewal fee at time t is lnRi 0ð Þ≥ ln ct=ztð Þ where ct is the annual renewal fee the patentee should pay at time t. To patent i, there should be three possibilities at time t: (1) The patent is renewed at t because the renewal fee is paid; (2) the patent lapses at t because the renewal fee is not paid; and (3) the patent expires at t because it is renewed to full term. Then, the probability that the patentee pays the renewal fee at time t is Pr Ti > tð Þ ¼ 1−Φ ln ct=ztð Þ−Xiβ σ � � The probability that the patentee stops paying the renewal fee at t is Pr Ti ¼ tð Þ ¼ Φ ln ct=ztð Þ−Xiβ σ � � −Φ ln ct−1=zt−1ð Þ−Xiβ σ � � Fig. 3. Comparison in Kaplan-Meier survival curve by different technology field — patent records in China. Table 2 Annual renewal fee for invention patents in China (RMB). Time Span 1985–1993 1994–2000 2001– Version Version Version 1–3 Years 200 600 900 4–6 Years 300 900 1200 7–9 Years 600 1200 2000 10–12 Years 1200 2000 4000 13–15 Years 2400 4000 6000 16–20 Years –* 8000 8000 * Renewal period for invention patents before 1993 is 15 years at most. 360 Z. Gupeng, C. Xiangdong / China Economic Review 23 (2012) 357–370 When the patent is renewed to full term, the value of the patent can be written as Pr Ti ¼ tfull � � ¼ 1−Φ ln ctfull=ztfull � � −Xiβ σ 0 @ 1 A where Φ(∙) represents a standard normal cumulative distribution, with tfull=18 years in a China patent study case. The more traditional or previously used models in international studies deal with only the latter two possibilities, i.e., the patent is either renewed to full term or expires at time t (tb18 in China's case). Thus, the likelihood function of the traditional model is Li ¼ Pr Ti ¼ tfull � �h iδi Pr Ti ¼ tð Þ½ �1−δi ð1Þ Table 3 Regression by traditional model and improved model of patent value: country differences. Impact Factors Traditional Model Improved Model According to Model (1) According to model (2) Number of Inventors 0.3419*** 0.2545*** (0.0071) (0.0068) Co-Application # 0.2658*** 0.1696*** (0.0015) (0.0159) Size of Applicant 0.0007*** 0.0005*** (0.0002) (0.0001) References: China U.S. # 0.0983** 0.2286*** (0.0455) (0.0329) Japan # 0.4806*** 0.7317*** (0.0251) (0.0344) E.U. # 0.3054*** 0.2453*** (0.0148) (0.0359) Other Countries # 0.0545*** 0.0776*** (0.0129) (0.0220) Technology Field (References: Consumer Appliances) Electrical Engineering # 0.3369*** 0.0597* (0.0084) (0.0304) Instruments # 0.0890 0.0838*** (0.0858) (0.0305) Chemistry, Pharmaceuticals # −0.1234*** −0.3654*** (0.0167) (0.0424) Process Engineering, Special Equipment # −0.1191*** −0.1137 (0.0381) (0.1027) Mechanical Engineering, Machinery # 0.0935*** −0.0684*** (0.0173) (0.0113) Grant Year (Reference: 1997–2000) 1985–1988 # −0.1537*** −1.3139*** (0.0051) (0.0297) 1989–1992 # −0.2259*** −0.0976*** (0.0283) (0.0261) 1993–1996 # 0.3790*** 0.0933*** (0.0297) (0.0244) 2001–2004 # 0.1234*** −0.0551*** (0.0061) (0.0111) 2005–2009 # 1.1778*** 0.4732*** (0.0120) (0.0020) Constant 11.3183*** 10.3217*** (0.0010) (0.0039) d 0.2428*** 0.3432*** (0.0001) (0.0087) σ 3.1382*** 4.5685*** (0.0698) (0.0391) No. of Obs. (No. of Censored) 119,823 568,596 (448,773) Robust standard error in the bracket. # Dummy variable. *** significant at 1% level; ** significant at 5% level; * significant at 10% level. 361Z. Gupeng, C. Xiangdong / China Economic Review 23 (2012) 357–370 However, the improved model in this study will cover all three possibilities; thus, the likelihood function of the improved model can be written as Li ¼ Pr Ti ¼ tfull � �h iδi Pr Ti ¼ tð Þ½ �1−δi n oφi Pr Ti > tð Þ½ �1−φi ð2Þ where δi and φi are indicator functions. δi ¼ 1 if patent i is renewed full term0 otherwise φi ¼ 1 if patent i is expired 0 otherwise �� Maximizing the likelihood function-containing n observations of the patent records will achieve the estimated values of d, σ and β. In this way, the corresponding estimated values can be used further to measure the currency-based value of the patents examined in this study. 2.2. A patent value model based on expired patent data Based on the model developed by Bessen (2008), for patent i expired at time t, the initial rent should satisfy the following equation: ln ct−1=zt−1ð Þ≤ lnRi 0ð Þ≤ ln ct=ztð Þ Given that Xiβ is fixed, the random component εi should make Ri(0) satisfy the above condition. That is, ln ct−1=zt−1ð Þ−Xiβ≤ εi≤ ln ct=ztð Þ−Xiβ The conditional expectation of εi is E εij ln ct−1=zt−1ð Þ−Xiβ≤εi≤ ln ct=ztð Þ−Xiβ½ � ¼ ρi ∫ ln ct=ztð Þ−Xiβ ln ct−1=zt−1ð Þ−Xiβ ε σ � � ϕ ε σ � � dε ð3Þ where ρi ¼ 1= Φ ln ct=ztð Þ−Xiβσ � � −Φ ln ct−1=zt−1ð Þ−Xiβσ � �h i and ϕ(.) are standard normal density functions. The estimate of initial return Ri(0) that satisfies condition (3) is Ri 0ð Þ ¼ exp Xiβ þ E εij ln ct−1=zt−1ð Þ−Xiβ≤εi≤ ln ct=ztð Þ−Xiβ½ �f g Here, the estimate of initial return Ri(0) and, thus, Ri(t) can be achieved; then, the present value of patent i expired at time ti is Vi ¼ ∫ ti 0 Ri τð Þe−sτdτ− Xti−1 t¼0 cte −st ð4Þ However, the value measured here is derived from information on expired patent records. For the unexpired patent records, the measuring model is discussed as follows. 2.3. A patent value model based on unexpired patent data for forward values Based on the previous discussion, unexpired patents should be considered in patent value studies, particularly for patent data in China. With similar principles used in studies of expired patent data, for an unexpired patent i with age t, the present value of profits from t to t+1 should satisfy R0zt > ct Thus, the value of εi should satisfy εi > ln ct=ztð Þ−Xiβ The estimate of εi will have its distribution as a normally distributed value. Thus, the conditional distribution function of εi is F εð Þ ¼ Pr εi≤ε εi > ln ct=ztð Þ−Xiβj � ¼ Pr ln ct=ztð Þ−Xiβbεi≤ε½ � Pr εi > ln ct=ztð Þ−Xiβ½ � ¼ Φ εð Þ−Φ ln ct=ztð Þ−Xiβð Þ½ � 1−Φ lnct=zt−Xiβð Þ½ � � 362 Z. Gupeng, C. Xiangdong / China Economic Review 23 (2012) 357–370 Let f(ε) be the conditional density function of εi, i.e., f(ε)=∂F(ε)/∂ε . The sufficient condition that patent i is renewed at period t+1 is