Essays on the empirical analysis of patent systems

2. The evolution: Chapter 2 presents a detailed descriptive analysis of the evolution in the size of patent applications filed to the European Patent Office (EPO). In this chapter, we propose two measures of patent voluminosity and identify the main patterns in their evolution. Based on a dataset with about 2 million documents filed at the EPO, the results show that the average voluminosity of patent applications – measured in terms of the number of pages and claims contained in each document – has doubled over the past 25 years. Nevertheless, this evolution varies widely across countries, technologies and filing procedures chosen by the applicant. This increasing voluminosity of filings has a strong impact on the workload of the EPO, which justifies the need for regulatory and policy actions.

3. The drivers: The evolution in patent voluminosity observed in chapter 2 calls for a multivariate analysis of its determinants. Chapter 3 therefore proposes and tests 4 different hypotheses that may contribute to explaining the observed inflation in size: the influence of national laws and practices and their diffusion to other countries with the progressive globalization of patenting procedures, the complexification of research activities and inventions, the emergence of new sectors with less established norms and vocabularies, and the construction of patent portfolios. The econometric results first reveal that the four hypotheses are significantly associated with longer documents and are therefore empirically supported. It appears however that the first hypothesis – the diffusion of national drafting practices through international patenting procedures – is the strongest contributor of all, resulting in a progressive harmonization of drafting styles toward American standards, which are longer by nature. The portfolio construction hypothesis seems a less important driver but nevertheless highlights substantial changes in patenting practices. These results raise two questions: Do these evolving patenting practices indicate more valuable patents? Do they induce any embarrassment for the patent system?

4. Measuring patent value: If the former of these two questions is to be addressed, measures are needed to identify higher value patents. Chapter 4 therefore proposes a review of the state of the art on patent value indicators and analyses several issues in their measurement and interpretation. Five classes of indicators proposed in the literature may be obtained directly from patent databases: the number of countries in which each patent is enforced, the number of years during which each patent has been renewed, the grant decision taken, the number of citations received from subsequent patents, and whether it has been opposed by a third party before the EPO. Because the former two measures are closely connected (the geographical scope of protection and length of maintenance can hardly be observed independently), they have been subjected to closer scrutiny in the first section of chapter 4, which shows that these two dimensions have experienced opposite evolutions. A composite measure – the Scope-Year Index – reveals that the overall trend is oriented downwards, which may suggest a substantial decline in the average value of patents. The second section of chapter 4 returns to the five initial classes of measures and underlines their main patterns. It appears that most of them witness the well-known properties of patent value: a severe skewness and large country and technology variations. A closer look at their relationships, however, reveals a high degree of orthogonality between them and opposite trends in their evolution, suggesting that they actually capture different dimensions of a patent’s value and therefore do not always pinpoint the same patents as being the most valuable. This result strongly discourages the reliance on one of the available indicators only and opens some avenue for the creation of one potential composite index of value based upon the five indicators to maximize the chances of capturing all potentially valuable patents in a large database. The proposed index reflects the intensity of the signal provided by all 5 constituting indicators on the potential value of each patent. Its declining trend reflects a rarefaction of this signal on average, leading to different plausible interpretations.

5. The links with patent value: Based upon the six indicators of value proposed in chapter 4 (the five classical ones plus the composite), the question of the association between filing strategies and the value of patents may be analysed. This question is empirically addressed in chapter 5, which focuses on all EPO patents filed between 1990 and 1995. The first section presents a comprehensive review of the existing evidence on the determinants of patent value. The numerous contributions in the field differ widely along three dimensions (the indicator of value chosen as dependent variable, the sampling methodology, and the set of variables tested as determinants), which have translated into many ambiguities across the literature. Section 2 proposes measures to identify different dimensions of filing strategies, which are essentially twofold: they relate to the routes followed by patent filings toward the EPO (PCT, accelerated processing), and to their form (excess claims, share of claims lost in examination), and construction (by assembly or disassembly, divisional). These measures are then included into an econometric model based upon the framework provided by the literature. The proposed model, which integrates the set of filing strategy variables along with some of the classical determinants, is regressed on the six available indicators separately over the full sample. In addition, the sensitivity of the available results to the indicator and the sampling methodology is assessed through 18 geographic and 14 industrial clustered regressions and about 30 regressions over random samples for each indicator. The estimates are then compared across countries, industries and indicators. These results first reveal that filing strategies are indicative of more valuable patents and provide the most stable determinants of all. And third, the results do confirm some classical determinants in their positive association with patent value, but highlight a high degree of sensitivity of most of them to the indicator or the sample chosen for the analysis, requiring much care in generalizing such empirical results.

6. The links with patent length: Chapter 6 focuses on one particular dimension of patent value: the length of patents. To do so, the censored nature of the dependent variable (the time elapsed between the filing of a patent application and its ultimate fall into the public domain) dictates the recourse to a survival time model as proposed by Cox (1972). The analysis is original in three main respects. First of all, despite the fact that renewal data have been exploited for about two decades to obtain estimates of patent value (Pakes and Schankerman, 1984), this chapter provides – to the best of our knowledge – the first comprehensive analysis of the determinants of patent length. Second, whereas most of the empirical literature in the field focuses on granted patents and investigates their maintenance, the analysis reported here includes all patent applications. This comprehensive approach is dictated by the provisional rights provided by pending applications to their holders and by the legal uncertainty these represent for competitors. And third, the model integrates a wide set of explanatory variables, starting with the filing strategy variables proposed in chapter 5. The main results are threefold: first, they clearly show that patent rights have significantly increased in length over the past decades despite a small apparent decline in the average grant rate, but largely due to the expansion of the examination process. Second, they indicate that most filing strategies induce considerable delays in the examination process, possibly to the benefit of the patentee, but most certainly to the expense of legal uncertainty on the markets. And third, they confirm that more valuable patents (more cited or covering a larger geographical scope) take more time to process, and live longer, whereas more complex applications are associated with longer decision lags, but also with lower grant and renewal rates.

7. Conclusions: The potential economic consequences and some policy implications of the findings from the dissertation are discussed in chapter 7. The evolution of patenting practices analysed in these works has some direct consequences for the stakeholders of the patent system. For the EPO, they generate a considerable increase in workload, resulting in growing backlogs and processing lags. For innovative firms, this phenomenon translates into an undesired increase in legal uncertainty, for it complicates the assessment of the limits to each party’s rights and hence of the freedom to operate on a market, which is precisely what the so-called ‘patent trolls’ and ‘submariners’ may be looking for. Although empirical evidence is lacking, some fear that this may result in underinvestment in research, development or commercialization activities (e.g. Hall and Harhoff, 2004). In addition, legal uncertainty is synonymous with an increased risk of litigation, which may hamper the development of SMEs and reduce the level of entrepreneurship. Finally, for society, we are left with a contrasted picture, which is hard to interpret. The European patent system wishes to maintain high quality standards to reduce business uncertainty around granted patents, but it is overloaded with the volume of applications filed, resulting in growing backglogs which translate into legal uncertainty surrounding pending applications. The filing strategies that contribute to this situation might reflect a legitimate need for more time and flexibility in filing more valuable patents, but they could also easily turn into real abuses of the system, allowing some patentees to obtain and artificially maintain provisional rights conferred by pending applications on inventions that might not meet the patentability requirements. Distinguishing between these two cases goes beyond the scope of the present dissertation, but should they be found abusive, they should be fought for they consume resources and generate uncertainty. And if legitimate, then they should be understood and the system adapted accordingly (e.g. by adjusting fees to discourage some strategies, raising the inventive step, fine-tuning the statutory term in certain technologies, providing more legal tools for patent examiners to reject unpatentable applications, etc.) so as to better serve the need of inventors for legal protection in a more efficient way, and to adapt the patent system to the challenges it is or will be facing.