1/ 🧵 New paper! “Patent Obfuscation and Reliance on External Financing” We use LLMs on 2.6m patents to decompose their linguistic complexity into a fundamental and strategic component. We show a link btw innovation disclosure strategies & finance across 4 empirical designs.

@iandgow @ProfAnalytics @nikir1 Oh - @nikir1 : We are updating the draft to mention your paper prominently! Not discussing yours was a big oversight in the rush to get this out for fall conference deadlines.

18/ Whew. You're still here. Thanks! This paper is brand new and has lots of room to grow. We are still collecting feedback and would love to hear what you think.

17/ This is joint work with my awesome coauthors: Kathleen Hanley and Qianqian Yu @LehighU Sungjoung Kwon @IlitchBusiness @waynestate William Mann @EmoryUniversity #AcademicTwitter #EconTwitter #FinTech #Innovation #Patents #LLM #TextAsData #CorporateFinance #IPfinance

16/ Paper & details: - SSRN: https://t.co/JWoSRpcmJK - Direct link + AI reading help: https://t.co/ZcS9QCCKFM - Slides: https://t.co/CJoK3sYlp8 I hope you'll read if you care about innovation policy, IP finance, or how new AI tools can shed light on disclosure strategies.

15/💡Takeaway 3: The patent system is built on a basic exchange: inventors receive legal protection in return for disclosing knowledge that supports further innovation. This paper shows that the nature of that exchange is shaped by firms' financing needs.

14/💡Takeaway 2: We show that patent obfuscation may hinder capital formation. This suggests that financing needs can temper the incentives to obfuscate. ➡️ Shows a positive role for financial constraints and for the growing use of IP assets!

13/ 💡 Takeaway 1: Financing considerations can influence the strategic choices made during the patent filing process. Because obfuscation reduces knowledge spillovers, our results identify an additional channel through which financing can shape the trajectory of innovation.

12/ Our decomposition approach updates the approach of Brian Bushee, @iandgow, @ProfAnalytics (JAR, 2018: https://t.co/j1cxq39ADY) and is related to great "Human-AI" difference paper by Bai, @nikir1, Cao, Liu, and Wan (R&R at MS: https://t.co/WhqLGUDdJg)

11/ 🛠️ Methodologically: Our LLM-based “text as data” approach offers a generalizable way to decompose **fundamental complexity** vs **strategic opacity**. This could be applied beyond patents: e.g., in corporate disclosures, financial reports, regulatory filings.

10/ 🌐 Importance to policy & innovation: More obfuscation means lower knowledge spillovers, fewer potential acquirers, and weaker tech-market liquidity. But external financing needs may push firms toward clarity — which could increase transparency and innovation dissemination.

9/ ➡️ Obfuscation isn’t costless. While it may deter competitors or reduce litigation risk, we show it reduces a patent’s value for external financing. This is probably because opaque descriptions are less useful to investors/lenders.

8/ ⚖️ Implication: There is a trade-off — strategic value (concealment/deterrence) vs financing. Firms don’t always maximize obfuscation. Instead, they balance strategic deterrence against raising capital.

7/ 📊 Firm-level shock to financing constraints: When firms’ financing constraints ease, we observe a *rise* in obfuscation — consistent with lower need for transparent collateral.

6/ We find: • IV (left fig): Patents with higher obfuscation are less likely to be cited, pledged as collateral, litigated, or sold. • Firms that pledge patents (right fig) or get hit by credit-rating shocks reduce obfuscation. → Suggests financing needs constrain obfuscation.

5/ 🧑‍💼 Evidence comes from four ID designs: • IV with examiner strictness regarding *clarity* specifically (lowers obfuscation) • Nautilus v. Biosig clarity ruling • American Inventors Protection Act (AIPA) early-publication rule • Change to Moody’s credit rating algo

4/ 📈 Stylized fact in the xsec: Obfuscation is negatively related to patent value and citations (it’s not costless!) and occurs in tech areas where strategic uses of patents are more salient (validation).

3/ 📈 Stylized fact in the time series: Over the past decade, obfuscation has surged across nearly all technology fields. Firms increasingly choose to cloak their innovation in complexity.

2/ 🔍 What we do: We use a leading LLM to rewrite patent abstracts in a “simplified, clear” form. Then we compare the readability of the original vs the LLM-simplified abstract. The gap = a novel measure of *strategic obfuscation*. 2 examples here:

It's incredible how powerful LLM's already are. As the tool ecosystem improves, they will have ever larger impacts. I have research using 100,000+ LLM q&as. It was done 15 months ago in the pre @ExpectedParrot era. The pain I could have avoided! https://t.co/6RgXihBuFz