As explained in previous SA articles (17 Healthcare Companies To Consider Based On Patent..., 40% Return In ~4 Months With Our IP Selected Health..., In Which Company Should You Invest In, Taking Into ..., In Which Field/Sub-Field Should You Invest In, Taki..., Patent Dynamics And Stock Performance), Innovalpha has:
- developed proprietary IP and quantitative investment models based exclusively on patents. Patent applications are processed on a weekly basis and result in a proprietary IP index for each company that files patents. Patent patterns/clusters are identified from the IP index, grading all companies that are filtered through.
- shown that more alpha is generated amongst innovative companies (backtesting and real track performance), as measured by weekly fluctuations in specific patent filing activity. From a weekly intellectual patent index (PI), it is possible to calculate innovation/patent scores for any company worldwide that files patents.
These IP models provide buy and sell signals for stocks essentially saying that now is the time to consider those stocks based on patent dynamics/patterns. In other words, companies that have seen quite recent notable increase in patent activity (see examples below) corresponding to identified patent patterns/clusters by the IP models provides support to integrate these stocks in your watch-list.
As mentioned in a previous SA article:
Such patent patterns reflect in principle interesting internal innovation that may lead to products on the market. At the minimum, such recent patent patterns reflect management belief that such innovation deserves consideration with consequently substantial financial and human resources allocated through increased patent-filing activity over a defined period of time (and the smaller the company is, the greater likelihood this is correct).
In summary, identification of patent patterns or profiles is an indication of growth within a company. The IP models do not assess the value of a patent per se, which is already a very difficult and likely impossible task for a patent expert, but takes into account the type of patent and fluctuations of patent activity over time (patent patterns/clusters). The SA article "Patent Dynamics And Stock Performance - Part II On IP/Patent/Innovation Indexes" provides further information, explanations and examples of the IP/patent models/Indexes.
Construction of IP models and IP/Patent Index
Currently, nearly six thousand (5796) companies are taken into account by the IP algorithms corresponding to Nasdaq, NYSE and Amex constituents.
The models/algorithms are built as follows:
i. Patent Profile: Each company's patent activity over time is unique (= history of patent activity). Algorithms have been designed to capture each company's patent profile on a weekly basis = IP [Patent] index for each company. The IP index fluctuates over time.
The IP index takes the following factors into account:
- Type of patent data - this is a quality measure as patents are not equivalent per se (different types of patents coexist, e.g. utility patents, design patents, innovation patents, patent applications, granted patents…).
- History of patent activity of a given company (distinct period of time for each company).
The IP index is therefore calculated in the same manner for all companies worldwide, whether private or public, for all sectors and regardless of the size of the company.
ii. Patent Dynamics: Among the thousands of different patent profiles (IP Index over time), specific patterns or fluctuations of patent activity over time have been detected to be correlated with market outperformance (alpha) via backtesting (patterns may be related to intense innovation, product launch, good management, etc.) = Patent Dynamics
iii. Ranking: The degree of outperformance is dependent on the type of patent pattern (e.g. a pattern A results in a better outperformance than a pattern B). This enabled the construction of a ranking system...
The IP models therefore enable the determination of buy and sell signals but also recommendations. The actual implementation of the buy/sell signal, i.e. the buy and sell recommendation (the moment shares of the company are bought/sold,) is not necessarily immediate following the buy/sell signal and may depend on the sector of activity or determined exclusively by each patent index . The buy/sell signal may be implemented after a few weeks or months. The time interval between the buy signal and its implementation (buy recommendation) may represent the average time needed for a company in a given sector to actually develop and market the products derived from in-house innovation, which will in turn materialize into equity return.
Criterion for success? Which are the factors and patterns that lead to outperformance? The main factors consist of a certain level of patent history and evidently the type of data used. Patterns that lead to outperformance include i) constant and progressive filing activity over a certain length of time and ii) increased or intensive filing activity over a shorter period of time. A lot of variations have been derived from these basics concepts. In general, a steady (even slow) increase over a certain period of time (from approximately 3 years) is in principle better than a rapid increase in a short period of time (e.g. 6 months). This makes sense for the rationale as it underlies a sustainable innovation/business.
In summary, Patent Dynamics is the design of unique models to identify specific patterns of innovation in any given Patent Profile. The use of patent data with patent dynamics represent a synergistic combination, which provides an entirely new way of identifying and selecting innovative growth-driven companies.
Evidence of alpha generation has already been shown in previous Innovalpha SA's articles/blogs. From the Patent Index, other scores can be computed like a patent value score, a patent risk score and a patent disruptive score. This article focuses on the intrinsic quality value of PCT (Patent Cooperation Treaty) patent applications administered by the World Intellectual Property Office (WIPO) from patent indexes and models/algorithms based thereof.
Further, this article provides the patent index based on PCT applications with related scores for the following companies: Micron Technology, Inc. (MU), NTT DOCOMO, Inc (DCM), Lam Research Corporation (LRCX), International Business Machines Corporation (IBM), Alibaba Group Holding Limited (BABA), Nippon Telegraph and Telephone Corporation (NTT), POSCO (PKX), Edwards Lifesciences Corporation (EW), Novartis AG (NVS) and General Electric Company (GE).
To test the hypothesis, portfolio performance (backtesting) has been computed from Nasdaq, NYSE and Amex listed companies since August 30, 2012 till early 2019.
Without even building any model from the patent indexes, there’s already alpha generation for companies having a Patent index based on PCT applications (around 1000 companies), see Figure 1. Hence, companies filing PCT patents fare better than the ones which don’t.
This is the first time that evidence is provided of the strong intrinsic value of patent applications, in particular of PCT applications. It is therefore highly recommended to select companies having filed recently at the least one PCT application. Further, it is a strong incentive for companies to file PCT applications (administered by WIPO), which will provide a competitive hedge.
Multitude of models have thereafter been built from the PCT patent indexes. Some results are presented below. They all show the important outperformance of basic and more advanced models.
This can be explained by the novel approach adopted focused on forward looking innovation, taking into account the recent patent activity of companies in contrast to other methods relying on granted patents (with innovation that occurred several years ago). Indeed, alpha lies in expectations not in past innovation already part of companies’ market valuation.
Figure 2 shows the outperformance of a basic model with companies having a patent index >= 15 (but better performance achieved with different settings, e.g. with IP<100). Innovalpha has also shown outperformance of models with a basic increase in the patent index (results not shown).
Figure 3 shows the outperformance of another model based on a specific pattern in the patent index, buying 12 weeks following pattern detection.
Figure 4 shows the outperformance of another model summing points awarded to companies depending on the way their patent index fluctuates on a weekly basis.
Figure 5 shows the outperformance that can be obtained by combining different models.
Even though we have shown that companies filing a certain number of PCT applications fare better than those filing very few PCT applications (Figure 2), this is certainly not the criteria to take into account when selecting companies for your portfolio. Indeed, all the other models show better market outperformance when specific fluctuations/patterns are detected over time in the weekly PI of companies independently of the PI level.
Figure 6 displays some examples of companies having a PI greater than 101 (heavy PCT filers) and which shows the best companies according to their weekly overall patent score that considers patent value, risk and disruptive capacity computed from their patent index. As these companies file intensively PCT applications, they shall already grasp the attention of the investor. As mentioned, the mere filing of PCT applications in an indication of success, but certainly not the best way to select companies. What is much more interesting are the specific patterns or fluctuations in the patent index detected and translated in the various patent scores.
As can be seen in Figure 6 (displaying 30 companies), among all the heavy PCT filers comprised in the NASDAQ, NYSE and AMEX, the top 10 companies are MU, DCM, LCRX, IBM, BABA, NTT, PKX, EW, NVS and GE.
These companies display a high overall patent score taking into account patent value, patent risk and patent disruptive scores. Such high scores are evidence of the intense innovation activity within these companies with identification of clear patterns/clusters. Investors shall definitively consider these companies for further analysis.
A more detailed view of these companies is presented in Figure 7 displaying the patent index, grade, patent value, patent risk and patent disruptive scores.
Patent value score is related to the identification of the patterns/clusters which translate into alpha generation. The higher the value, the more patterns/clusters have been identified. The investor should note that it is better to select companies at the start of their innovation cycle (not indicated in the Figure) as stock market price might have already increased. Further, financial analysis is therefore warranted before including any of these companies in your portfolio.
Patent risk score is related to patent activity. As these companies are filing a lot of PCT applications annually, patent risk is considered very low in comparison with companies filing much less. Hence, all these companies display very good patent risk scores. Nonetheless, the investor can choose companies according to their patent risk score in order to adapt to the risk profile of the investor. However, as is usually the case, the investor should note that the less risk taken, the less probability that alpha will be generated, and, on the other hand, the more risk taken, the more probability that alpha will be generated (backtesting evidence).
Patent disruptive score is also related to patent activity and in particular to specific patterns of intense patent activity over a particular period of time. In general, it is more difficult for heavy PCT filers to display high values in such scores. Disruptors are more easily found in new emerging companies. Nonetheless, Figure 7 shows the disruptive capacity of these companies which are good taking into account their size.
Figure 8 shows that among all the companies having a PI>101, some are hence more disruptive than others (bubble chart, with the overall patent score on the Y axis and disruptive capacity on the X axis, the bigger the bubble the higher the PI index).
Hence, it can be deduced from this Figure that the 10 companies mentioned above display good patent disruptive scores in comparison with others as a lot of companies with PI>101 have patent disruptive scores below 0 (which indicates that such companies have not been able, have difficulties to put in place innovation management/strategy that will translate into disruptive innovation).
Figure 9 shows the map location of the companies.
The map is useful to immediately show if there's some diversification in the geographical location of companies selected. It is prudent for the investor not to select companies only located within one geographical region.
Figure 10 shows the weekly scores over time of Micron Technology, Inc.
It shows the constant increase of the scores over time since 2017 with some acceleration in 2019. It is a good indication of the innovation capacity of this company. The investor should definitively select this company for further analysis. As shown below for the correlations of patent scores with stock price, the investor should check the historical patent profile of companies which enables to immediately identify in a convenient way any modifications in the patent scores over time.
Figure 11 shows MU’s scorecard with grade, various scores, category, phase and alerts.
This view provides some further information and indicates for example the position of MU in comparison with other companies within the Technology sector. Hence, according to this analysis, it is the best company in this sector for the patent overall score, for the patent value score and for the patent disruptive score. It is in position eight for the patent index and for the patent risk score. Alerts indicate that MU is in a second cycle of innovation (taking into account the available historical data) and is a medium company according to patent filing activity.
Figures below show some nice correlations of the patent overall score with the stock price of various companies. They provide evidence that an increase in the overall patent score can translate into an increase in the stock price (and in some instances that a decrease in the overall patent score can translate in to a decrease in the stock price). After all, innovation represents one of the core values of a company, with strong R&D/innovation and good innovation strategy/management translating into the filing of patent applications and hopefully in the launch of new products on the market. As such, all these Figures indicate that the investor should have selected these companies already some time ago, but that their actual scores still warrant further analysis for portfolio incorporation.
Figure 12 shows the correlation of MU’s patent overall score with its price.
Figure 12 shows that the investor should have considered MU already in 2016.
Figure 13 shows the correlation of IBM’s patent overall score with its price.
Figure 13 shows that the investor should have considered IBM already in late 2018.
Figure 14 shows the correlation of LRCX’s patent overall score with its price.
Figure 14 shows that the investor should have considered LRCX already in 2016.
Figure 15 shows the correlation of BABA’s patent overall score with its price.
Figure 15 shows that the investor should have considered BABA already in mid 2015 and a stagnant/decrease since 2018 should have warranted attention of the investor.
Figure 16 shows the correlation of PKX’s patent overall score with its price.
Figure 16 shows that the investor should have considered PKX already in early 2016 and a stagnant/decrease since early 2018 should have warranted attention of the investor (same is true for the period mid 2014-2015).
Figure 17 shows the correlation of EW’s patent overall score with its price.
Figure 17 shows that the investor should have considered EW already at the least in 2015.
This article has shown once again the strength of the novel approach taken by Innovalpha relying on patent dynamics and cycles of innovation, in contrast to other methods that usually automatically provide a score to an individual patent, which is something already difficult to do for an expert. Obviously, the two approaches can be combined (patent applications and granted patents), but Innovalpha further demonstrates here that methods based on patent applications provide strong outperformance and are therefore reliable in the detection and selection of companies.
Companies and investors should take into account the following points:
- Companies filing PCT patents fare better than the ones which don’t.
- Models presented rely on PCT patent applications which translate into forward looking innovation and strong market outperformance versus current methods relying on granted patents (past innovation). This is evidence of the strength of patent dynamics with the identification of patent patterns/clusters and cycles of innovation.
- PI (Patent Index) level is interesting to know, but other factors are more important, i.e. specific fluctuations/patterns in the PI. Patent value, risk and disruptive scores with historical profiles/data provide further useful information for company selection.
- Based on their patent scores, investors should consider for further analysis MU, DCM, LRCX, IBM, BABA, NTT, PKX, EW, NVS and GE.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: The information provided in this document is given for indicative purposes only. Innovalpha Sàrl provides no assurance as to its completeness or accuracy nor the reasonableness of the conclusions based upon such information. This document is not intended to constitute an offer or solicitation for the purchase or sale of an investment program or of any one or more of the models mentioned in the document. There is no assurance that the models investment objectives will be achieved and investment results may vary significantly over time. Past performance should not be construed as a guide to future performance. The content of this document is subject to change without prior notification. All information and data presented in this document is for informational purpose only and should not be reproduced, distributed nor used without prior written authorization of Innovalpha Sàrl.