Abstract:

The relationship between scientific knowledge—a critical input to technological systems—and invention value remains a central concern in science policy and innovation management. While existing studies have primarily focused on correlational analyses, yielding inconsistent findings, the integration of scientific knowledge into technological systems is not random but intimately linked to knowledge recombination patterns and invention team characteristics. The inconsistency in research findings likely stems from endogeneity issues caused by selection bias, which traditional regression analyses fail to address. Causal inference methods are thus necessary to accurately estimate the net causal effect of scientific knowledge on invention value, thereby providing more reliable evidence for science and technology policymaking.

Using a counterfactual causal framework, we employed Propensity Score Matching (PSM) and Generalized Propensity Score Matching (GPSM) to estimate the causal effects of both the presence and intensity of scientific knowledge inputs on patents' technological and commercial value. Existing measures of scientific knowledge input intensity focus solely on citation counts, overlooking variations in scientific literature impact. We address this limitation by developing the Patent Scientific Citation h-index (PSH), inspired by the h-index methodology, which integrates both quantitative and qualitative dimensions of scientific citations to more accurately assess patents' scientific knowledge absorption. We measured patents' technological value using two indicators: forward citation counts and technological generality. Commercial value was quantified through the frequency of patent transfers. Analyzing 1,685,970 utility patents granted by the USPTO (2001-2010) and their 6,473,214-citation links to 1,407,439 scientific papers, we established the following findings:

(1) Our findings demonstrate positive causal effects of scientific knowledge input on both the technological and commercial value of patents. Patents incorporating scientific knowledge receive significantly more forward citations than those without such inputs. This indicates that scientific knowledge integration enhances a patent's capacity to serve as a foundation for subsequent innovations. Moreover, scientific knowledge input significantly enhances patents' technological generality. Specifically, we observe that scientific knowledge incorporation enables the development of more influential general-purpose technologies. The theoretical mechanism underlying this relationship lies in how comprehensive scientific understanding enhances researchers' cognitive capabilities to effectively identify, access, and recombine knowledge from distant technological domains. Finally, patents incorporating scientific knowledge demonstrate significantly higher transfer frequencies compared to those without such knowledge inputs. This pattern suggests that science-based inventions exhibit enhanced potential for private value realization, primarily because established scientific frameworks provide a shared cognitive platform and common language among market participants.

(2) A nonlinear relationship exists between scientific knowledge input intensity and inventive value. Empirical analysis reveals a fluctuating "rise-decline-rise" pattern in patents' forward citations and transfer frequencies as the PSH index increases. Patents with either low or high PSH intensities demonstrate superior technological impact and market transaction potential. In low-PSH domains, moderate scientific knowledge integration maintains technological advancement while aligning with market cognition. Mid-range PSH shows diminishing returns due to potential divergence from user expectations. High-PSH patents leverage rich scientific knowledge to overcome technical barriers, yielding enhanced technological impact and market value. Regarding technological generality, our findings reveal that patents characterized by higher PSH indices embody more fundamental scientific principles. The universal applicability of these underlying principles facilitates technology extension and diffusion across diverse application contexts, resulting in enhanced levels of technological generality.

This study makes several significant contributions to the innovation policy literature. First, we employ a causal inference framework to establish the causal relationship between scientific knowledge input and inventive value. Second, we advance methodological development by introducing a novel approach that incorporates the H-index to quantify scientific knowledge input intensity. These empirically grounded findings provide robust evidence for optimizing science and technology policies, bridging the gap between theoretical frameworks and practical policy implementation.

Key words: scientific knowledge, invention's value, propensity score matching (PSM), generalized propensity score matching (GPSM), counterfactual framework.

摘要：

科学知识被广泛视为创新过程中的“发明地图”,其与发明价值之间的关系一直是科技政策与创新管理领域的核心议题。然而,已有研究主要聚焦于二者的相关性分析,研究结论不尽一致,难以为科技决策提供确切的参考依据。本文基于反事实框架,运用倾向得分匹配(PSM)与广义倾向得分匹配(GPSM)方法,以美国专利商标局USPTO 2001-2010年授权的1,685,970件发明专利及其与科学论文构成的6,473,214个专利-论文引用对为研究样本,构建专利科学引用H指数(PSH)作为量化科学知识输入强度的代理指标,系统考察了科学知识输入及其强度对发明价值的因果效应。研究结果表明:(1)科学知识输入对专利的前向引用、技术通用性和专利转让均具有显著的正向因果效应;(2)科学知识输入强度与专利前向引用、专利转让次数均存在非线性关系,随着PSH指数的递增,专利的前向引用和转让次数表现出“先上升-后下降-再上升”的波动趋势,而低强度与高强度区间的发明专利则都具有更高的技术影响力和市场交易潜力;(3)高强度的科学知识输入是形成通用性技术的重要驱动力。本研究从因果推断视角验证了科学知识对发明价值的影响效应,并结合H指数提出了评估科学知识输入强度的新方法,为科技政策优化提供了重要的实证依据。

关键词: 科学知识, 发明价值, PSM匹配, GPSM匹配, 反事实框架