The Impact of the Leahy-Smith America Invents Act on Firms’ R&D Disclosure

Abstract

The Leahy-Smith America Invents Act of 2011 (AIA) fundamentally changed the U.S. public patent system. This study examines whether and how the enactment of the AIA affects innovation-intensive firms’ narrative R&D disclosure. Adopting a difference-in-difference design, we find that the enactment of the AIA is associated with a greater decline in narrative R&D disclosure of innovation-intensive firms relative to non-innovation-intensive firms. This effect is more pronounced for firms that are more concerned about the proprietary costs of disclosure and firms with lower financial constraints. Moreover, we show that the enactment of the AIA dampens analysts’ information environment and aggravates information asymmetry. Overall, our results suggest that the AIA has a negative impact on corporate disclosure and information asymmetry. The documented unintended effect of the AIA could potentially work against lawmakers’ efforts to bolster innovation and the dissemination of knowledge.

Keywords:

• The America Invents Act
• Patents
• Intellectual property
• Narrative R&D disclosure
• Analysts’ forecast
• Information asymmetry

Acknowledgements

We appreciate the helpful comments of Ane Tamayo (the editor), two anonymous reviewers, Neil Fargher, Gary Monroe, Rencheng Wang, Mark Wilson, Yangxin Yu, seminar participants at the Australian National University, participants at the 2016 American Accounting Association Annual Meeting and participants at the 2017 European Accounting Association Annual Congress.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 This was determined at the discretion of the U.S. Patent and Trademark Office after taking all relevant evidence into consideration.

2 The 1952 Patent Act required the application to ‘contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention’ (35 U.S.C. § 112(a), 2010).

3 We thank a reviewer for the suggestion to require at least one patent in the post-AIA period to define our innovation-intensive firms. We use three patents as the cut-off because this is the median of the total number of patents filed in our sample period. As part of our robustness checks, we re-estimate all regressions after redefining our treatment firms using the following alternative criteria: (1) firms that have at least one patent during our sample period; and (2) firms that have at least three patents since 2005 (10 years prior to the AIA). We obtain consistent results. For each of these tests, we do not require each firm to have all patent data across our sample period.

4 We did not impose the requirement that the three patents necessary to classify a firm as treated be spread over three years. As a robustness check, we also require that the 3 patents to classify a firm as treated be spread over 3 years and find that after we impose this requirement, while our observations is reduced by 1,026, we obtain consistent results (p < 0.05) for all three R&D narrative disclosure variables. We thank a reviewer for providing this suggestion.

5 Apart from using different patent number cut-offs to identify treatment samples, we use the following two ways to measure firms’ innovativeness. First, we use an input measure, R&D intensity, to measure the firms’ innovativeness. Second, we thank a reviewer for the suggestion to consider other patent characteristics and thus use patent citation numbers to classify treatment/control firms.

6 Lower disclosure level increases information asymmetry between managers and investors, amplifying adverse selection (e.g., Grossman, 1981; Grossman & Hart, 1980). Investors are less likely to provide capital, or demand greater compensation, which constrains firms’ ability to invest in risky R&D projects (Myers & Majluf, 1984).

7 For example, an applicant could file a new compound (medicine) with the USPTO by disclosing elements including molecular composition, temperature, and reactive agents. This information might allow a person with ordinary skills in the art to replicate a comparable compound. However, without disclosing the precise methodology (best mode) to implement the compound, later inventors might require further development or experimentation to produce a compound with similar efficacy, which is both time consuming and expensive (Sohi, 2013).

8 ‘(3) Invalidity of the patent or any claim in suit for failure to comply with—(A) any requirements of section 112, except that the failure to disclose the best mode shall not be a basis on which any claim of a patent may be canceled or held invalid or otherwise unenforceable’ (AIA, § 15(3)(A), 2011).

9 1952 Patent Act, 35 U.S.C §102(f)–(g) (2010).

10 This refers to both filing the invention application with the USPTO and the subsequent detailed disclosure of the invention made by the USPTO to the public.

11 First, the innovator can pre-empt rivals' R&D investment and use defensive disclosure to diminish the probability that the rival has of receiving a patent of similar technology (Ponce, 2011); Second, voluntary disclosure of proprietary innovative information has a strategic benefit by allowing the rival to infer the innovator’s cost efficiency and deter new entrants (Hughes & Pae, 2015); Third, voluntarily revealing such information encourages other rivals to adopt the invention, ultimately making the innovation the informal standard for future development (Harhoff et al., 2003); Fourth, voluntarily disclosing proprietary innovative information allows innovators to search for its potential partners who can help with the innovation together by reducing information asymmetry between the innovator and outsiders. (Ettredge et al., 2017; Hughes & Pae, 2015).

12 For some examples of defensive disclosure, see ‘Protecting Intellectual Property’ by The New York Times on 02/18/2002; ‘Suddenly, “Idea Wars” Take on a New Global Urgency’ by The New York Times on 11/11/2002, and ‘On the Defensive about Invention’ by The Financial Times on 09/19/2001. For examples of voluntary disclosure of proprietary information, see Ford Motor Company’s disclosure of information about its assembly line system (Anton & Yao, 2004) and IBM’s disclosure of information of the process to manufacture semiconductors (Hughes & Pae, 2015); See also Gu and Li (2003); Jones (2007); Merkley (2014) for more examples.

13 See footnote 2 for a detailed description of the ‘best mode requirement’.

14 See footnote 7 for an example of best mode disclosure.

15 We also examine alternative event windows as robustness tests in Section 6.6.3.

16 Only for firms whose year-ends are February 28 or March 31, the pre-AIA period includes fiscal years 2009, 2010, and 2011, and the post-AIA period includes fiscal years 2013,2014, and 2015. Our results are robust if we exclude those firms.

17 The forward-looking word list includes ‘will’, ‘could’, ‘should’, ‘expect’, ‘anticipate’, ‘plan’, ‘hope’, ‘believe’, ‘can’, ‘may’, ‘might’, ‘intend’, ‘project’, ‘forecast’, ‘objective’, and ‘goal’.

18 As robustness check, we also exclude patents that have been filed before the enactment of AIA and granted in the post-AIA period when defining treatment sample.

19 We acknowledge that this limited data availability may lead to misclassifications of some innovation-intensive firms as non-innovation-intensive firms in our tests (Type II error); however, such bias may only work against finding the results.

20 As robustness checks, we also re-estimate all regressions after redefining our treatment firms using the following alternative criteria: (1) firms that have at least one patent during our sample period, and (2) firms that have at least three patents from 2005 to 2014. We obtain consistent results.

21 We also test the robustness of the results to an alternate method of classifying innovation-intensive and non-innovation-intensive firms using R&D intensity.

22 The patent data are available at: http://iu.app.box.com/patents.

23 Available at https://www.uspto.gov/learning-and-resources/electronic-data-products/historical-patent-data-files and https://www.uspto.gov/learning-and-resources/electronic-data-products/patent-assignment-dataset.

24 The name-matching algorithm is developed based on code written by Jim Bessen, available at http://goo.gl/m4AdZ (Kogan et al., 2017).

25 We follow Merkley (2014) and include Age² and (R&D/OPX)² to control for nonlinearity.

26 Because of the inclusion of both firm and year fixed effects in this specification, both dummy variables, PostAIA and Treated, are suppressed.

27 We thank Bill McDonald for publicly sharing these data, which are available at https://sraf.nd.edu/data/stage-one-10-x-parse-data/.

28 As a robustness check, we exclude chemicals industry (SIC 28); and our results remain unchanged.

29 Our main results are also robust if we truncate all continuous variables at the top and bottom one-percentiles.

30 The mean for Analyst, the natural logarithm of the number of analysts, is 1.744.

31 Given the dependent variable is in the form ln(1 + y), the first derivative should be Δy ÷ (1 + y). Therefore, economic magnitude is estimated by β₁ × (1 + y) ÷ y, which is [(−0.078 × (1 + 31.38) ÷ 31.38)=] 8%.

32 DeFond et al. (2017) suggest that matching treatment firms with control firms with replacement reduces bias in treatment effects. This is because it allows each treatment group to be matched with the closest control group.

33 We set the caliper to value 0.01, which is less than 25% of the standard deviation of the predicted value from Model (2); using a caliper larger than 0.01 results in finding no close match and using a caliper smaller than 0.01 results in a loss of observations.

34 Our results continue to hold if we use quintiles rather than the median to classify innovative (non-innovative) firms.

35 We thank a reviewer for the suggestion to consider the impact of AIA on patent characteristics. In addition to patent quality, AIA may affect other patenting characteristics including patent generality, patent originality etc.

36 Our classification approach is consistent with prior studies in a DID setting (e.g., Byard et al., 2011; Irani & Oesch, 2013).

37 Consistent with Byard et al. (2011), Horizon and Noest are excluded in the regressions with COV as the dependent variable.