An Empirical Study on the Effects of the America Invents Act on Patent Applications Owned by Small Businesses
AAn Empirical Study on the Effects of the AmericaInvents Act on Patent Applications Owned by SmallBusinesses
Yoo Jeong HanDepartment of StatisticsUniversity of California, Los AngelesFebruary 2021
Abstract
This paper evaluates the heterogenous impacts of the America Invents Act of 2011 (AIA)on patent applications for small and large businesses. Using data collected from the UnitedStates Patent and Trademark Office and Google Patents, I compare how the probability of suc-cessfully overcoming an initial rejection is affected by the AIA for small- and large-businessapplicants, respectively. This comparison is achieved by analyzing the data using a difference-in-differences approach. Results suggest that after the enactment of the AIA, small-businessapplicants were relatively favored when compared against large-business applicants. This ef-fect is statistically significant and also practically large. a r X i v : . [ s t a t . A P ] F e b Introduction
It has been argued that small businesses have been unfavorably affected by the major reforms(Braun, 2012) of the Leahy-Smith America Invents Act of 2011 (AIA). Vandenburg (2014) andSutton (2014) propose that the transition from the First-To-Invent system to the First-To-File sys-tem (Section 3, AIA) favors large businesses that have the financial resources to accommodatefaster filing, thus placing small businesses at a relative disadvantage. Case (2013) notes the AIA’selimination of the post-disclosure grace period as another source of disadvantage. The lack of apost-disclosure grace period forces inventors to either file before fully developing their ideas or tofully develop their ideas while at risk of premature disclosure by others. This is advantageous tolarger businesses that have relatively abundant financial resources for faster development.While Vandenburg (2014), Sutton (2014), and Case (2013) focus on small businesses that havenot yet filed their patent applications, Lerner, Speen, and Leamon (2015) attend to the possi-ble disadvantages to small-business patent owners by measuring market reaction to the AIA forpublicly-traded patent-owning small businesses. Their study, however, does not find clear evidenceof unfavorable market reactions, and reports only minimal differential impact by the AIA.In this paper, I focus on another group of small businesses that may be affected by the AIAenactment – those that have applications that are undergoing pre-grant patent prosecution. Specif-ically, I compare how the probability of successfully overcoming an initial rejection is affected bythe AIA for small- and large-business applicants, respectively. This comparison pertains to a phaseof patent prosecution called the ‘Search and Examination phase,’ summarized as follows.A patent application is examined for its patentability by a patent examiner at the United StatesPatent and Trademark Office (USPTO). If the application is deemed unpatentable as is, the appli-cant would most likely receive a Non-Final Rejection. This Non-Final Rejection details the argu-ments against patent issuance of the current application. The applicant’s response to the Non-FinalRejection attempts to overcome the examiner’s listed arguments by arguing against the examiner orby amending the current application. The successful response places the application in a positionready for patent issuance. The unsuccessful response fails to convince the examiner that the appli-cation is ready for patenting. This can be because the response inadequately addresses the issuesset forth in the Non-Final Rejection or because the response creates new reasons for rejection inits amendments.To identify the AIA’s effect on applications in the early Search and Examination phase, I ex-amine the success rate, i.e., the proportion of patent applications that successfully respond to their2nitial Non-Final Rejections. I compare the difference in success rate of small-business applicantsto that of large-business applicants before and after the AIA. If the AIA does not favor or disfavorsmall businesses in this phase, the difference in success rate for small-business applicants shouldroughly equal that of large-business applicants. In other words, the two differences would be nodifferent.For this comparison, I take the difference-in-differences (DID) approach. In applying the DIDmethod, I assume that trends in the success rate due to uncontrolled characteristics are independentof business size. This is called the parallel trends assumption and is key to the validity of the DIDanalysis. (See Section 3 for details; see also Wing, Simon, and Bello-Gomez, 2018, and Schiozer,Mourad, and Martins, 2021.)It is notable that when a new law is enacted, the behavioral adjustment of its related entities,including applicants, patent lawyers and agents, and USPTO examiners, may not be immediatelyclear. This is especially relevant to the research in this paper in that it involves variables that aredifficult to measure with precision. Some of these variables are the quality of a small-businessapplicant’s attempt to overcome a rejection and the examiner’s judgement on that small-businessapplicant’s attempt. These variables make it difficult to deduce the AIA’s effects on small-businessapplicants using a priori arguments, which in turn, makes the current research topic more inter-esting. Furthermore, to the best knowledge of the author, this paper is the first attempt in currentliterature to explain the AIA’s effects on patent applications owned by small and large businessesthat are undergoing pre-grant prosecution.The results of this paper show that there is a significant difference in differences. Small businessapplicants experienced more success in their Non-Final Rejection responses than large-businessapplicants, at least for those which were ultimately issued as patents.The rest of this paper is organized as follows. Section 2 describes data sources and the mea-surement of variables and provides summary statistics. A brief summary of the DID methodologyand estimation results are presented in Section 3. Section 4 contains concluding remarks.
The data is collected through a series of maneuvers within the USPTO website and Google Patents.I first used the USPTO’s PatFT search engine to retrieve a comprehensive list of 1,866 patents3hat have an application date ranging from the 2nd to the 6th of January 2010. I obtained thepatent number and application number for each of the patents on this list. The patent number’sformat contains information regarding its invention subject category, i.e., whether the patent isa utility, design, or plant patent. Using this information, I exclusively retained the numbers forutility patents. I then used each of the retained application numbers to locate and scrape the cor-responding utility patent’s public data on the USPTO’s electronic filing and patent applicationmanagement tool, Patent Center. Patent Center provides detailed information for each individualpatent such as the history of documents and transactions between the applicant and the USPTOfrom application filing to patent issuance. It also provides information regarding the patent’s ap-plication meta data, continuity, patent term adjustment, and attorney information and more. Forexample, Patent Center provides that Patent No. 7,992,975 (Non-Conductive Fluid Droplet Form-ing Apparatus and Method) falls under Class 347 and Group Art Unit 2861, has business entitystatus LARGE/UNDISCOUNTED, has a particular correspondence address in the United States,has four attorneys, claims priority to two parent applications (1 DIV type, 1 PRO type), has a Pri-ority Date of 10/04/2004, has two inventors (all US inventors), and has 44 incoming and outgoingdocuments from application filing to patent issuance. After filing the application on 01/05/2010,the first USPTO Office Action is a Non-Final Rejection (CTNF), which took place on 02/18/2011.This is followed by the applicant’s response to the Non-Final Rejection (REM) on 04/07/2011,and a Notice of Allowance (NOA) on 04/15/2011. Using the data scraped from Patent Center,I narrowed the list of utility patents down to 1,467, which excludes national-stage applicationsoriginating from foreign or international patent applications.An Office Action is defined in this paper as a Notice of Allowance (NOA), Ex Parte QuayleAction (CTEQ), Requirement for Restriction/Election (CTRS), Non-Final Rejection (CTNF), orFinal Rejection (CTFR). Using the scraped data from Patent Center, I identify the first OfficeAction (OA1) the applicant receives for each utility patent. If OA1 is a Non-Final Rejection, Imanually downloaded the OA1 document in XML or PDF form. This part of the data collectionprocess could not be automated because Patent Center does not allow access to the documents inprogrammable ways that I know of. For each OA1 that is a Non-Final Rejection, I extracted thenumber of claims that are rejected under each of 35 USC § 112, 101, 102(a), 102(b), 102(e), 103(a)and 103(e). This process was automated for the vast majority of these files in XML format, while Imanually counted the number of claims rejected under each section for the remaining few in PDFformat.Finally, I obtained the number of independent claims for each utility patent. This information4as gathered by scraping the text of each claim in each utility patent from Google Patents. Thetext of each claim was processed to determine whether that claim is independent or dependent.
This section summarizes information in the data set relevant to the research question. The typesof first Office Actions (OA1) are Notice of Allowance (NOA), Ex Parte Quayle Action (CTEQ),Requirement for Restriction/Election (CTRS), Non-Final Rejection (CTNF), and Final Rejection(CTFR). The distribution of OA1 types is summarized as follows.Table 2.1: Distribution of OA1 typesType NOA CTEQ CTRS CTNF CTFR TotalNumber 176 21 276 991 3 1,467Share (%) 12.0 1.4 18.8 67.6 0.2 100.0Out of 1,467 patents, 176 applications (12.0%) were allowed in OA1. OA1 is a CTNF type in991 applications (67.6%), and the other 300 applications (20.4%) had the remaining types.For the 991 applications that received a CTNF-type OA1, the types for the second Office Action(OA2) are distributed as in Table 2.2. Among the 991 applications which received CTNF-typeOA1s, 363 of them have a CTFR-type (36.6%), 78 of them have a CTNF-type (7.9%), and 539 ofthem have a NOA-type (54.4%) OA2.Table 2.2: Distribution of OA2 types after CTNF-type OA1Type NOA CTEQ CTRS CTNF CTFR TotalNumber 539 2 9 78 363 991Share (%) 54.4 0.2 0.9 7.9 36.6 100.0The composition of the business entity status is presented in Table 2.3 for the total 1,467 appli-cations. 5able 2.3: Distribution of business entity status category (all patents)Type Large Small Micro TotalNumber 1,138 305 24 1,467Share (%) 77.6 20.8 1.6 100.0A vast majority of the patent applications have applicants with business entity status ‘LARGE/UNDISCOUNTED’, while the proportion of those with ‘SMALL’ or ‘MICRO’ status is 22.4%.Of those with a CTNF-type OA1, the distribution is as given in Table 2.4. The proportion of smallbusinesses (including SMALL and MICRO) is slightly smaller (21.3%) when conditioned on theevent of a CTNF-type OA1.Table 2.4: Distribution of business entity status for applications with CTNF-type OA1Type Large Small Micro TotalNumber 780 194 17 991Share (%) 78.7 19.6 1.7 100.0A full description of the OA1 types for each business entity status is given in Table 2.5.Table 2.5: Cross table for OA1 typesType NOA CTEQ CTRS CTNF CTFR SumLarge 147 17 191 780 3 1,138(12.9) (1.5) (16.8) (68.5) (0.3) (100.0)Small 27 4 80 194 0 305(8.9) (1.3) (26.2) (63.6) (0.0) (100.0)Micro 2 0 5 17 0 24(8.3) (0.0) (20.8) (70.8) (0.0) (100.0)Sum 176 21 276 991 3 1,467(12.0) (1.4) (18.8) (67.6) (0.2) (100.0)
Note:
Row-wise percentage shares are in parentheses.
Note:
Row-wise percentage shares are in parentheses.
The quality of the application as filed and the quality of the applicant’s response to the firstNon-Final Rejection are important determinants in the resulting type of OA2. To control for them,I note that the quality of the application as filed can be correlated with the complexity of theinitial Non-Final Rejection and also with the difficulty to overcome that rejection. That difficultycan influence the strength of the applicant’s response, which directly affects the type of OA2 thatis received. Based on this observation, instead of directly processing the applications as filed, Iconsider the delay between the application filing date and the date of receipt of CTNF-type OA1as a possible measure of the quality of the application as filed. Delayed receipt of a CTNF-typeOA1 may indicate that the application as filed contains more flaws, although I recognize that the7elay may also be influenced by the flux of new patent applications at the USPTO. The delayuntil receipt of a CTNF-type OA1 since filing is distributed as in Figure 2.1. On average, receiptof a CTNF-type OA1 took 1.84 years (approximately 1 year and 10 months) since filing, and thestandard deviation is 0.76 years (approximately 9 months).Figure 2.1: Delay until receipt of CTNF-type OA1
Years to first PTO action D e n s i t y . . . . . . . mean = 1.84sd = 0.76 The delay of the applicant’s response since receipt of a CTNF-type OA1 would be related withthe difficulty in addressing the issues raised by USPTO. Except one application that took 2.6 yearsfor a response, the distribution of this delay is given in Figure 2.2. The mean length of delay isapproximately 3 months with a standard deviation of approximately 1 month.8igure 2.2: Delay until applicant response to CTNF-type OA1
Years to applicant response D e n s i t y mean = 0.25sd = 0.09 Similarly, the delay until the receipt of an OA2 from the applicant response date may be corre-lated with the quality of the applicant’s response. This delay is distributed as in Figure 2.3.Figure 2.3: Delay until receipt of OA2 since applicant response to CTNF-type OA1
Years to second PTO action D e n s i t y mean = 0.19sd = 0.16 The number of rejected claims in OA1 is a more direct indicator of the difficulty of address-ing the issues raised by the USPTO. Non-Final Rejections provide information on the sections in95 U.S.C. under which claims are rejected. These sections are §101, §102(a), §102(b), §102(e),§103(a), §103(e), and §112 for the collected data. Table 2.7 summarizes the distributions of theclaim numbers rejected under these 35 U.S.C. sections, which is used in the analyses in Section 3.Table 2.7: Distributions of claims rejected under 35 U.S.C. sectionsNumber of 35 USC §rejected claims 101 102(a) 102(b) 102(e) 103(a) 103(e) 1120 858 962 579 857 331 990 6721–5 51 8 130 33 189 0 1536–10 35 11 102 36 155 0 6111–15 24 4 78 24 100 0 3216–20 16 2 50 20 127 0 3621–25 4 3 25 8 39 0 1826–30 2 0 10 5 23 0 631– 1 1 17 8 27 1 13Sum 991 991 991 991 991 991 991One of the key variables in the present research is whether OA2 was received before or afterthe enactment of the AIA. Limited to the applications that have a CTNF-type OA1, the estimatedcumulative distribution function (CDF) of the date of receiving OA2 is given in Figure 2.4. A pointon the estimated CDF represents the proportion of considered applications that receive an OA2 bythat date. For example, 22.9% of the applications have received an OA2 by September 16, 2011(depicted by the vertical dashed line). 10igure 2.4: Cumulative distribution of OA2 receipt date . . . . . . Date of second PTO action C D F ← September 16, 20112010 2011 2012 2013 2014 2015 The number of independent claims, described in Figure 2.5, can also be indicative of the com-plexity of the patent application. Most applications (97.2%) have six or less independent claims.Figure 2.5: Distribution of the number of independent claims for applications with CTNF-typeOA1 Number of independent claims F r e q u e n c y ( ) Finally, I control for the total number of parent applications, the number of patented parentapplications, the number of parent applications for each continuity type and the earliest priority11ate. These data points attempt to capture examination on matter related to the current applicationprior to its filing.The dependent variable for the following analyses is the indicator for a successful response toa CTNF-type OA1, i.e., the dependent variable ‘success’ takes on the value 1 if the OA2 type isnot CTNF or CTFR. This dependent variable and the independent variables to be used in this workare listed in Table 2.8 below. Table 2.8: List of control variablesVariable Name Meaningsuccess =1 if OA2 type is not CTFR or CTNFSmallMicro =1 if small or micro businessAIA =1 if OA2 is received in the post-intervention periodnumIndClaims number of independent claimsdblPat =1 if CTNF-type OA1 contains double patentingwait1 delay from filing to receipt of OA1 (in years)adelay delay from receipt of OA1 to applicant response (in years)wait2 delay from applicant response to receipt of OA2 (in years)US =1 if the corresponding address is in the United StatesnumParents number of parent applicationsnoparent =1 if number of parent applications is zeropryears length of period (in years) between priority date and filing date;0 if there are no parent applications.numParentTypeCON number of parent applications with CON type continuitynumParentTypeCIP number of parent applications with CIP type continuitynumParentTypeDIV number of parent applications with DIV type continuitynumParentTypeNST number of parent applications with NST type continuitynumParentTypePRO number of parent applications with PRO type continuityUSC101 number of claims rejected under 35 U.S.C. § 101USC102a number of claims rejected under 35 U.S.C. § 102(a)USC102b number of claims rejected under 35 U.S.C. § 102(b)USC102e number of claims rejected under 35 U.S.C. § 102(e)USC103a number of claims rejected under 35 U.S.C. § 103(a)(Continued on next page)12ariable Name MeaningUSC103e number of claims rejected under 35 U.S.C. § 103(e)USC112 number of claims rejected under 35 U.S.C. § 112hasUSC101 =1 if USC101 > hasUSC102a =1 if USC102a > hasUSC102b =1 if USC102b > hasUSC102e =1 if USC102e > hasUSC103a =1 if USC103a > hasUSC103e =1 if USC103e > hasUSC112 =1 if USC112 > Summary statistics for the dependent variable and the variables in Table 2.8 are provided inTable 2.9. Table 2.9: Summary statisticsVariable Obs Mean SD Min Median Max
Note: “
I use the difference-in-differences (DID) method to measure the effect of the AIA on the successrate, the rate of successful response to initial Non-Final Rejection, of small and large businesses.The DID method is widely used in social sciences to measure causal effects of policy interventions.(See Card and Krueger, 1994, for example.) In general terms, the DID method is used in thefollowing manner.There exists a sample consisting of two groups of observational units. One group is the treat-ment group and the other the control group. Given a chosen intervention date, the periods aredichotomized into a pre-intervention and a post-intervention period. The DID method begins withmeasuring the average values of the dependent variable for each group in each period. Table 3.1exhibits these averages as A00, A01, A10, and A11, respectively, where A00 and A01 are the av-erages for the control group in the pre-intervention and post-intervention periods, respectively, andA10 and A11 repeat the same practice for the treatment group.14able 3.1: Diagram for difference-in-differencesControl group Treatment group DifferencePre-intervention A00 A10 A10 – A00Post-intervention A01 A11 A11 – A01Change A01 – A00 A11 – A10 (A11 – A10) – (A01 – A00)The change in the average dependent variable for the treatment group (A11 – A10) consists ofthe treatment effect and the associated uncontrolled trend between the dichotomized periods:A11 − A10 = (
Treatment effect ) + (
Trend for treatment group ) , (1)where random errors are ignored due to the Law of Large Numbers. It is notable that the changeA11 − A10 contains not only the treatment effect of interest, but also the changes due to factorsunrelated to the policy intervention. To find the treatment effect using (1), it is necessary to knowthe value of A11 − A10 and the trend effect for the treatment group. The value of A11 – A10 can beestimated by the sample. To find the trend effect for the treatment group, we turn to the over-timechange for the control group.The over-time change for the control group consists of only the trend effectA01 − A00 = (
Trend for control group ) (2)because the control group is not treated. The random errors are again ignored due to the Law ofLarge Numbers. Here the key assumption for the DID method is implemented, that is, the paralleltrends assumption which states that the treatment and control groups both share the same trendeffect (Wing et al., 2018, Schiozer et al., 2021). Under this assumption, the treatment effect is canbe found by taking the difference between (1) and (2) – the difference-in-differences (DID):DID = ( A11 − A10 ) − ( A01 − A00 ) . (3)The DID method is alternatively called a comparative interrupted time series design or a nonequiv-alent control group pretest design (Wing et al., 2018).The DID method can be implemented by the standard ordinary least squares (OLS) regression.To estimate the DID effect, one can regress the dependent variable (Y) on the dummy variable(TR) for treatment group, another dummy variable (POST) for the post-treatment period, and the15nteraction of the two, that is, TR*POST. Then the DID estimate is simply the coefficient on theinteraction term (Wing et al., 2018, p. 456). The standard errors and associated confidence intervalsto use for inferences are reported by standard statistical software such as R.The parallel trends assumption is vital for the regression of Y on TR, POST, and TR*POSTto give a valid policy effect estimator. If this assumption is violated, the trend heterogeneity con-founds the causal effect. In that case, one can control for the factors that are responsible for thetrend heterogeneity. For example, if the distribution of observations over time for the treatmentgroup differs from that of the control group, and there are other interventions within the vicinity ofintervention date in question, year dummies or year-quarter dummies can be included to mitigatethe effect of those other unwanted interventions. Also, quality measures can be controlled for ifquality difference drives group-wise heterogeneity in trends.In these analyses, the treatment group is composed of the patent applications owned by small-business applicants (including micro-business applicants), and the control group is composed ofthose owned by large-business applicants. The date of the intervention in question is the date theAIA has been signed into law, September 16, 2011.A complication arises with regard to Equation (2). The standard DID method states that thecontrol group is free from policy intervention. However, in the case of this analysis, patent applica-tions owned by large businesses may be affected by the AIA just as those of small businesses are.Because of this, I look at the ‘relative policy effect’, defined as the difference in the AIA’s effectson small business patent applications and its effects on large business patent applications. It canbe observed that if the parallel tends assumption is satisfied, the DID approach will consistentlyestimate the ‘relative policy effect’. To demonstrate this, see that Equation (1) can be expressed asA11 − A10 = (
Policy effect for treatment group ) + (
Trend for treatment group ) , (4)while Equation (2) can be generalized toA01 − A00 = (
Policy effect for control group ) + (
Trend for control group ) . (5)The ‘relative policy effect’ is defined as the difference of the two policy effects in (4) and (5). Underthe assumption that the trends are identical for the treatment group and the control group, the DIDstatistic estimates the ‘relative policy effect’, which is the difference between the policy effectsbetween the two groups. In other words, the relative policy effect is the heterogeneity in the policyeffects between the two groups. A positive DID statistic indicates that small-businesses applicantsenjoy a relatively higher success rate than large-business applicants after the AIA became effective.16his could suggest that small businesses received a more favorable treatment within the USPTOpost-AIA in comparison to large businesses. A negative DID statistic indicates otherwise. Prior to conducting formal regression analysis, I present trends in success probability for patentapplications owned by large-business applicants and those owned by small-businesses applicants.Figure 3.1 exhibits centered three-month moving averages of successful responses for each of largeand small business groups in the sample of 991 applications, where ‘success’ is defined in Table2.8, and the centered moving average at a certain date is defined as the ratio of successes in the91-day period centered at that date. For example, the moving average for December 1, 2011 is theratio of successes from October 17, 2011 (45 days prior) to January 15, 2012 (45 days posterior).Figure 3.1: Success ratios during centered three-month periods . . . . . . Week of second PTO action S u cc e ss P r o b a b ili t y Small & Micro Large2011 2012 2013
Note:
The vertical dashed line is for September 16, 2011.
Prior to the AIA, the patent applications owned by large-business applicants and those ownedby small-business applicants do not seem to display a systematic difference between their successrates. However, for more than a year starting approximately one month after the introduction ofAIA, small-business-owned patent applications exhibit systematically higher success rates in com-parison to large-business-owned patent applications. Loosely speaking, regression analyses in thissection measure the average change of the success-rate differentials over time between small and17arge businesses.Results from the formal DID regressions are presented in Table 3.2.Table 3.2: Difference-in-differences estimation resultssuccess (1) (2) (3) (4)SmallMicro*AIA 0.2347*** 0.2037** 0.1850** 0.1655**SmallMicro -0.0502 -0.0310 -0.0100 -0.0035AIA -0.1550*** -0.2385 -0.1813 -0.1508log(numIndClaims) -0.0806*** -0.0621**dblPat 0.0244 -0.0227wait1 -0.0210 0.0092adelay -0.6348** -0.5031*wait2 -0.6606*** -0.5864**US 0.0697 0.0967numParents 0.0263 0.0413**noparent 0.0310 0.0041pryears -0.0218* -0.0252**numParentTypeCON 0.0471 0.0090numParentTypeCIP -0.0628 -0.0818*numParentTypeDIV 0.0799 0.0281numParentTypeNST -0.0198 -0.0197USC101 0.0031USC102a -0.0240**USC102b -0.0038*USC102e -0.0067*USC103a -0.0049***USC103e -0.0309**USC112 0.0010hasUSC101 -0.0564hasUSC102a 0.1005hasUSC102b -0.0577hasUSC102e -0.0030(Continued on next page)18uccess (1) (2) (3) (4)hasUSC103a -0.1290***hasUSC112 -0.0654Year-quarter dummies No Yes Yes YesIntercept 0.6440*** 0.0310 0.3507 0.4278 n
991 991 991 991R-squared 0.0255 0.0431 0.1080 0.1584Adj. R-squared 0.0225 0.0234 0.0773 0.1174
Note:
Robust standard errors are in parentheses. The DID estimates are on the first row. ***, ** and * stand forstatistical significance at the 1%, 5% and 10% levels, respectively.
In Column (1) of Table 3.2, the intercept 0.6440 measures the estimated success rate for thecontrol group (large businesses) during the pre-AIA period. The coefficient on the ‘SmallMicro’variable is the average differential of the success rate between the treatment group (small busi-nesses) and the control group (large businesses) in the pre-treatment period (pre-AIA). The esti-mated value -0.0502 means that the success rate of the treatment group is estimated to be lowerthan that of the control group in the pre-treatment period.The coefficient on the AIA variable measures the average ‘before/after’ change in the successrate for the control group. Patent applications filed by large businesses experienced a decreasedsuccess rate (-0.1550). Though not shown in the table, the treatment group, on the contrary, ex-perienced an increase in the success rate by 7.97 percentage points. In comparison to the controlgroup’s experience of a decrease by 16.01 percentage points, the treatment group experienced arelative increase in success rate by 23.47 percentage points, which is the estimated coefficient(DID) on the interaction term. This effect is practically large in magnitude amounting to 36% oflarge businesses’ average success rate in the pre-AIA period, and is statistically significant. Theseresults are summarized in Table 3.3.Table 3.3: Simple DID tableLarge Small & Micro DifferencePre-invention 0.6440 0.5938 -0.0502Post-invention 0.4890 0.6735 0.1845Change -0.1550 0.0797 0.234719 graphical illustration of Figure 3.2 plots success rates for both groups during the pre- andpost-AIA periods. In the pre-AIA period, applications owned by large businesses showed a slightlyhigher average success rate in comparison to those owned by small businesses by 5.02 percentagepoints according to Table 3.3. This is in contrast with the post-AIA period. While the averagesuccess rate for applications owned by large businesses experienced a fall of success rate by 15.5percentage points, the average success rate of those owned by small businesses rose by approxi-mately 8 percentage points. The difference of 23.47 percentage points of the two changes is theestimated DID effect, i.e., the ‘relative policy effect’.Figure 3.2: Difference-in-differences graphically illustrated . . . . . . period S u cc e ss P r o b a b ili t y pre-AIA post-AIALarge LargeSmall & Micro Small & Micro Column (2) controls for the year-quarter effects by including dummy variables (where ‘year-quarter’ means the year-quarter of the OA2 receipt date). These year-quarter dummy variablesaccount for the changes in possible differences in the outflow of Office Actions in each quarterbetween the treatment and the control groups. As a result, the estimated effect slightly decreasesin magnitude in comparison to Column (1) but remains statistically significant at the 5% level.Column (3) controls for various characteristics of the patent applications together with theyear-quarter effects. These extra control variables are the log number of independent claims (‘nu-mIndClaims’), whether the CTNF-type OA1 contains remarks on double patenting (‘dblPat’), thedelay between filing and the receipt date of the OA1 (‘wait1’), the delay between the OA1 receiptdate and the applicant’s response date (‘adelay’), the delay between the applicant’s response dateand the receipt date of the OA2 (‘wait2’), the country of corresponding address (‘US’), the number20f parent applications (‘numParents’), the dummy variable indicator of no parent applications (‘no-parent’), the interval between the priority date and the filing date in years (‘pryears’), the numberof parent applications with continuity types CON, CIP, DIV and NST. The remaining continuitytype PRO is omitted in order to avoid collinearity with ‘numParents’. The estimated DID effectdecreases by 1.5 percentage points from Column (2), but the estimated value 0.1850 again seemsconsiderable in magnitude and is statistically significant at the 5% level.Shifting focus to the other variables in Column (3), the log number of independent claimsis negatively correlated with the success rate and its effect is statistically significant at the 1%level. Double patenting does not have significant effects. The delay periods after the first Non-Final Rejection are estimated to have adverse effects on success rate. The ‘pryears’ variable is alsonegatively correlated with the success rate. It is notable that the inclusion of the ‘noparent’ dummyvariable limits the comparison of the effects of ‘pryears’ to among patent applications that have anonzero number of parent applications. The negative estimated coefficient (-0.0218), significant atthe 10% level, suggests that the longer the time difference between the priority date and the filingdate, the lower the success rate among comparable patent applications.Column (4) further controls for the numbers of claims rejected under each section within 35U.S.C. The numbers of rejected claims under each section are negatively correlated with the suc-cess rate. This aligns with the intuition that there are more rejected claims in a Non-Final Rejectionthat is more difficult to overcome. The estimated DID effect again falls by approximately 2 per-centage points, but is still large in magnitude (0.1655) and statistically significant.I have defined the post-AIA period as September 16, 2011 until present day. It is interesting toobserve the change or lack of change of the magnitude and statistical significance of the DID effectwhen the policy intervention date is changed. Specifically, I explored changing the interventiondate to October, November, and December of 2011, respectively. These dates respectively markthe one month, two month, and three month date since the AIA was signed into law. For theseintervention date alterations, the period between September 16, 2011 and the new interventiondate was disregarded as a ‘gray period.’ The DID effects change to Columns (2)–(4) of Table 3.4for model (4) in Table 3.2. The results remain robust.21able 3.4: Various intervention periodssuccess (1) (2) (3) (4)SmallMicro*AIA 0.1655** 0.1769** 0.1739** 0.1670**SmallMicro -0.0035 -0.0062 -0.0080 -0.0040AIA -0.1508 2.1323 2.1955 2.2020Other variables omitted n
991 979 958 935R-squared 0.1584 0.1576 0.1558 0.1586Adj. R-squared 0.1174 0.1170 0.1142 0.1160
Note: (1)=Model (4) in Table 3.2; (2)=comparison of pre 09/16/2011 and post 10/16/2011; (3) comparison of pre09/16/2011 and post 11/16/2011; (4)=comparison of pre 09/16/2011 and post 12/16/2011. Robust standard errors arein parentheses. The DID estimates are on the first row. ***, ** and * stand for statistical significance at the 1%, 5%and 10% levels, respectively.
This paper examines the effects of the AIA on small businesses that have pending patent appli-cations in the Search and Examination phase of patent prosecution. Data was collected from theUSPTO’s PatFT search engine and Patent Center tool, and Google Patents. The results of this em-pirical study show that after the enactment of the AIA on September 16, 2011, small-business ap-plicants experienced a statistically significant higher success rate of 16.5 to 23.5 percentage points,measured by difference-in-differences (DID), than that of large-business applicants. The estimatedeffect on small businesses is robust in that the inclusion or removal of control variables in the DIDregressions does not considerably affect the DID coefficient and its statistical significance.While this study uses a large enough sample to show statistically significant DID effects, col-lection of further data would allow for an assessment of long-term effects on small-business appli-cants. This may be desirable since the AIA was fully implemented over the course of 18 months(USPTO, 2011). Also, this study only uses patent applications that were ultimately issued as apatent. Investigation of full data including abandoned patent applications is left for future researchdespite the data for such applications having limited public accessibility.22 eferences
Braun, R. G. (2012). America Invents Act: First-to-file and a race to the patent office, 8
Ohio St.Entrepren. Bus. L. J. , 47.Card, D., and A. B. Krueger (1994). Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania,
American Economic Review , 84, 772–793.Case, J. (2013). How the America Invents Act hurts American inventors and weakens incentives toinnovate,
University of Missouri-Kansas City Law Review , 82(1), 29–77.Lerner, J., A. Speen, and A. Leamon (2015). The Leahy-Smith America Invents Act: A Prelim-inary Examination of Its Impact on Small Businesses,
SBA Office of Advocacy
Revista de Administrac¸ ˜ao Contem-porˆanea , 25(1), e200067. Available at: https://doi.org/10.1590/1982-7849rac2021200067.Sutton, P. J. (2014). AIA’s impact upon small businesses,
World Intellectual Property Review
Idaho Law Re-view .201. Available at: https://digitalcommons.law.uidaho.edu/idaho-law-review/vol50/iss1/8.Wing, C., K. Simon, and R. A. Bello-Gomez (2018). Designing difference in difference studies:Best practicies for public health policy research,