Patent Trolls and Small Business Employment Ian Appel† , Joan Farre-Mensa‡ and Elena Simintzi∗ November 2016 Abstract We analyze how frivolous patent-infringement claims made by “patent trolls” affect small firms’ ability to create jobs, raise capital, and survive. Our identification strategy exploits the staggered passage of anti-patent-troll laws at the state level. We find that the passage of this legislation leads to a 1.8% increase in employment at small firms in high-tech industries, which are a frequent target of patent trolls. By contrast, the laws have no significant impact on employment at larger or non-high-tech firms. Anti-troll legislation is also associated with fewer business bankruptcies. Financing appears to be a key channel driving our findings: in states with an already established VC presence, the passage of anti-troll laws leads to a 19% increase in the number of firms receiving VC funding. Consistent with this, we find that the effect of patent laws on employment is driven by states with above-median VC presence. Our findings suggest that measures aimed at curbing the recent explosion in patent litigation may play an important role in reducing both real and financing frictions faced by small businesses.

Keywords: employment, patent litigation, venture capital. Affiliations: † Carroll School of Management, Boston College; ‡ Harvard Business School, Harvard University; ∗ Sauder School of Business, University of British Columbia. e-mails: [email protected], [email protected], [email protected].

I

Introduction

Small businesses play a critical role in creating jobs and promoting economic growth. Yet, such firms face a number of frictions that limit their ability to grow, create jobs, and innovate.1 One friction that has received considerable attention in recent years is patent litigation, giving rise to concerns that “patent litigation now imposes substantial costs, particularly on small and innovative firms, and these costs have tended overall to reduce R&D, venture capital investment, and firm startups” (Asay et al. 2015). Indeed, the number of patent lawsuits has increased tenfold since 2000 (Cohen, Gurun, and Kominers 2015). Perhaps even more worryingly, in recent years much of this litigation has been brought by non-practicing entities or “patent trolls”, organizations that own patents but do not make or use the patented technology directly: The share of patent cases brought by patent trolls increased from 30% in 2009 to over 60% in 2014 (CEA 2016). Examples abound of the negative consequences of patent trolls on job dynamics. A case in point is Kunin, a Vermont technology company, whose CFO was cited in the Washington Post in 2013 as saying: “We had two projects that we were close to doing. Then our clients contacted us saying, ‘We’re being threatened by patent trolls’.” As a result, the Post reports, “both projects were cancelled, nixing the firm’s plans to hire several new employees.”2 Motivated in part by these concerns, several bills have been introduced in Congress since 2012.3 However, as of today, none of them have become law. As Cohen, Gurun, and Kominers (2015) note, those opposing these bills argue that patent trolls “serve a key financial intermediary role, policing infringement by well-funded firms that could 1

According to the Small Business Administration, small businesses have created over 7 million of the 11 million jobs created since the end of the Great Recession. 2

“How Vermont could save the nation from patent trolls” (The Washington Post, Aug. 1, 2013).

3

In the 114th Congress (2015-2016) proposed bills include the The Protecting American Talent and Entrepreneurship (PATENT) Act (S. 1137), Innovation Act (H.R. 9), Venue Equity and NonUniformity Elimination Act (VENUE) Act (S. 2733), Trade Protection Not Troll Protection Act (H.R. 4829), STRONG Patents Act (S. 632), Targeting Rogue and Opaque Letters (TROL) Act (H.R. 2045), Demand Letter Transparency Act (H.R. 1896), and Innovation Protection Act (H.R. 3349).

–1–

otherwise infringe upon small inventors’ intellectual property at will” (p. 1). In response to Congress’s lack of action to curb the activities of patent trolls, 31 state legislatures, beginning with Vermont in 2013, have taken action at the state level to limit patent trolls’ ability to bring frivolous legal threats against local businesses. In this paper, we use a difference-in-differences approach to provide the first analysis of the effect that these state-level anti-patent-troll laws have had on small firms’ ability to create jobs, raise capital, and survive. We find that the passage of anti-patent-troll legislation in a state leads, on average, to a 1.8% increase in employment at high-tech small firms (those with fewer than 20 employees) located in the state. By contrast, we find no significant effect on the employment at larger high-tech firms or at non-high-tech firms regardless of their size. The increase in employment at small high-tech firms is most significant for college-educated workers; that said, we also find somewhat weaker evidence of an employment increase for workers without a college degree, thus highlighting the complementary nature of their jobs. While the fact that many of the anti-troll laws have been passed in the last three years limits our ability to capture their long-term effects, we find no evidence that their effect on employment is short-lived–if anything, it appears to increase over time. Our results indicate that the net effect of anti-troll legislation is to increase employment at small high-tech firms, which are key for economic growth and are precisely the firms that tend to be increasingly targeted by patent trolls (Chien 2015). Thus, anti-troll laws’ positive effect in protecting small high-tech firms from frivolous legal threats appears to on average dominate any negative effect that the laws may have in limiting patent trolls’ ability to act as financial intermediaries for these firms. Most importantly, patent trolls seem to hurt those sectors of the economy that are deemed to be job creation drivers, namely small high-tech businesses. Empirically identifying the effects of policy changes such as the passage of anti-patenttroll legislation is plagued with endogeneity concerns. At the heart of these concerns is the possibility that there could be confounding variables, such as macroeconomic shocks, that

–2–

could affect both a state’s decision to pass anti-troll legislation and the hiring decisions of firms in the state. Our difference-in-differences identification strategy helps us overcome these endogeneity challenges by exploiting the fact that not all U.S. states have adopted anti-patent troll laws (as of today, 19 have yet to do so, including some states with a large concentration of high-tech firms like California); in addition, among those states that have adopted them, not all of them did so at the same time. Thus, the staggered passage of the anti-patent troll laws allows us to construct a time-varying control group that provides a plausible counterfactual for how employment would have evolved in the treated states had they not adopted anti-patent-troll legislation. For our control group to be a valid counterfactual, the evolution of employment at treated and control states needs to share parallel trends. Consistent with this assumption, we find that employment trends at treated and control states were indistinguishable pre-treatment, both economically and statistically. While this identifying assumption is ultimately untestable, the following facts are also consistent with it. First, our results are robust to controlling for standard state-level macroeconomic variables: state aggregate and per-capita income, unemployment, and patenting volume. These macroeconomic variables are in addition to the state and year-quarter fixed effects that we include in all specifications, which account for time-invariant employment differences across states and aggregate shocks affecting employment in all states, respectively. Second, the fact that the effect of anti-troll laws is concentrated among the small high-tech firms that tend to be the focus of patent trolls is consistent with our capturing the effect of the laws and not some other state-level shock. Third, small firms are unlikely to be able to spend much on lobbying, thus alleviating the concern that our estimates may be driven by reverse causality or by some unobservable shock affecting both the growth opportunities and lobbying efforts of high-tech firms. What are the likely mechanisms behind the positive effect that anti-troll laws have on small high-tech firms’ ability to create jobs and grow? By reducing the instances of frivolous intellectual property (IP) lawsuits, anti-troll laws allow small firms to avoid having to devote their often limited resources to defending themselves in court or to settling out-of-court.

–3–

Consistent with this direct effect of the laws on improving the viability of constrained firms by allowing them to conserve their cash, we find somewhat noisy evidence that the passage of anti-troll legislation in a state leads to a decrease in the number of business bankruptcies in the state.4 Perhaps even more importantly, by simply reducing the threat of frivolous IP lawsuits, anti-troll laws can increase the net present value (NPV) of investing in the high-tech small firms that tend to be the target of patent trolls. This higher NPV may not only affect the investment decisions of the firms’ founders and managers, but also of professional investors such as venture capitalists (VCs). In line with this hypothesis, we find that, in states with above-median VC presence at the beginning of our sample period, anti-troll laws lead to a 19% increase in the number of firms raising VC funding and to a 29% increase in the capital invested by VCs in the state. By contrast, we find that the laws have no significant effect on VC investments in those states with low VC activity to begin with, thus suggesting that the passage of anti-troll laws is, on its own, not sufficient to attract VCs to states with historically scarce VC presence. Further reinforcing the importance of this financing channel, we find that the positive effect of anti-patent-troll laws on the employment of small high-tech firms is driven by states with above-median VC presence. In these states, we also find a positive effect of the laws on the number of high-tech establishments, thus suggesting that the growth in employment we observe is driven both by the growth of existing establishments and by the creation of new ones. Our study joins a small but growing literature that aims to analyze the economic consequences of the explosion in patent litigation, much of which has been brought by patent trolls. Chien (2015) shows that patent trolls tend to target small firms: 55% of unique defendants in lawsuits brought by patent trolls make less than $10 million in annual revenue (by contrast, only 16% of firms sued by operating companies such as IBM make 4

By contrast, as expected, we find no effect of anti-troll laws in the number of non-business bankruptcies in the state. This placebo test further alleviates concerns that our findings are driven by unobservable shocks that coincide, both in time and geography, with the passage of the laws.

–4–

less than $10 million in revenue). Cohen et al. (2015) show that, when suing large public firms, patent trolls tend to focus on those that are cash rich, leading to a decrease in their innovative activities (see also Chien 2013; Tucker 2014; and Smeets 2015). On the financing side, Kiebzak, Rafert, and Tucker (2016) show that litigation by patent trolls is associated with lower levels of VC investment. In contrast to these studies that focus on specific firms targeted by patent lawsuits, our empirical setting has the advantage of capturing both the direct and indirect consequences of frivolous patent infringement claims by capturing the effect of a reduction in the risk that any firm faces of being targeted by patent trolls. Our paper’s key contribution is to provide the first analysis of the effects of state antipatent-troll laws. The adoption of these laws has been surrounded by an intense debate on whether limiting the activities of patent trolls may do more harm than good by hindering trolls’ ability to help small innovative firms monetize their inventions. Also, a number of commentators have expressed doubts on whether states should legislate in an area, IP protection, that has long been seen as a federal matter. Our results indicate that state antitroll laws have had a positive net effect for small firms in high-tech industries, helping them create jobs and making them more attractive to VC investors, and no significant effect for larger or non-high-tech firms. These findings suggest that anti-patent-troll laws can have a multiplier effect, by not only decreasing the resources that small firms’ need to spend on litigation expenses, but by also facilitating their access to the funding, monitoring, advice, and networks access provided by VCs (Hellmann and Puri 2002; Bernstein, Giroud, and Townsend 2015; Hochberg, Ljungqvist, and Lu 2007). Our paper also contributes to the literature in finance studying the effects of constraints on employment. Several papers have shown that financial constraints impede firm employment growth (e.g. Benmelech, Bergman, and Seru 2011; Chodorow-Reich 2014; Giroud and Mueller 2015, 2016; Benmelech, Frydman, and Papanikolaou 2016). This literature argues that disruptions in financial markets matter for real economic outcomes. This study argues instead that constraints imposed on small innovative firms by frivolous patent-infringement claims can be disruptive for employment and firm growth.

–5–

The remainder of the paper is organized as follows. Section II discusses the institutional background of the anti-troll regulation. Section III summarizes the data and describes the methodology. Section IV presents the baseline results. Section V discusses the mechanisms that explain our findings. Section VI concludes.

II

Institutional Background

In the US, patent law generally falls under the purview of the federal government. The Constitution authorizes Congress to enact legislation that protects inventors so as to promote scientific progress, and federal courts have jurisdiction over related civil actions. Recent years have seen considerable debate within the legislator regarding reforms to the patent system, culminating with the passage of the America Invents Act (AIA) in 2011. This effort to modernize US patent law featured a number of significant changes, including the implementation of a first-inventor-to-file system, post-grant review of patents, and new joinder rules making it more difficult to sue multiple defendants simultaneously. The changes to the joinder rules were specifically intended to curb abusive patent litigation (Bryant, 2012). However, evidence suggests the AIA has failed to significantly reduce the number of lawsuits brought by NPEs.5 In response to concerns that the AIA does not sufficiently protect firms from bad-faith infringement claims, a number of states have taken action (Desisto, 2015). While the “substance” of patent law is federal, state law plays an important role in some instances (Gugliuzza, 2015).6 Beginning with Vermont in 2013, state legislators adopted patent reforms that protect businesses accused of patent infringement under consumer protection laws. The goal of the laws, according to legislators, was to reduce the effect of abusive patent litigation on the local economy. For example, the Vermont statute (ACT 44) notes 5 Specifically, the number of patent lawsuits was close to a record high in 2015 (Cohen, Gurun, and Kominers, 2015). Cotropia, Kesan, and Schwartz (2014) note, however, that the number of targeted firms has remained approximately constant before and after AIA. 6

For example, licensing agreements are subject to state contract law, and ownership of a patent after the inventor’s death is determined by state probate law (Gugliuzza, 2015).

–6–

that such litigation can “impose a significant burden on individual Vermont businesses” and limit their ability to “invest, produce new products, expand, or hire new workers”. In Vermont, the push for patent reform was led by a tech firm with 100 employees that eventually started a grassroots collation with other local firms (Washington Post, 8/1/2013). The Vermont law has served as a model for other states, and a total of 31 states currently have adopted patent reform based on consumer protection laws.7 The statutes limit bad faith assertions of patent infringement. While the definition of bad faith is “amorphous”, it includes instances of allegations that lack factual allegations, have no merit, or are deceptive (Gugliuzza, 2015). The laws allow defendants to recover costs and/or punitive damages from bad faith assertions. There are, however, some differences between the laws passed by different states. One such difference is the behaviors that can trigger the law. In most states, the statutes only apply to demand letters. However, in some (e.g., Georgia, North Carolina, and Maryland), the law also applies to litigation if the complaint violates the statute (Patent Progress, 2016).8 Another difference is that the bulk of states permit for private right of action (allowing alleged infringers to bring suit), though in some only the attorney general can enforce the statute (Gugliuzza, 2015). In addition, approximately half of the statutes include a bond requirement, requiring plaintiffs to post up to $500k for the defendant’s legal fees. Finally, while the laws of all states apply to firm located within it, some laws also apply to demand letters sent from patent holders from the state (Patent Progress, 2016). 7

Table A1 provides a list of the states along with the corresponding date legislation was signed.

8

Patent Progress (2016) notes that this can give rise to a situation where a complaint meets federal pleading standards but is illegal under state law. Federal patent law may preempt the state law in this case under the Supremacy Clause of the Constitution, though this has not been tested in court.

–7–

III

Data and Methodology

III.1

Data

We obtain state-level employment data from the Census’ Quarterly Workforce Indicators (QWI) database. Employment is measured at the end of each quarter (EmpEnd). Hires (HirAEnd) include workers that begin a new job in the quarter which continues into the next quarter. Separations (Sep) include workers whose job with a given employer ended in the specified quarter. Payroll (Payroll) is the total payroll for jobs in a given quarter. We use the QWI aggregation of employment by firm size and 4-digit NAICS industry in a given state-quarter. QWI reports five size categories (in employees): 0-19; 20-46; 50-249; 250-499; 500+. In the interest of space, we collapse those in three groups keeping the first two categories the same and combining the last three categories which include the largest firms. Thus, we measure employment for small (<20 employees), medium (20-50), and large (50-500+) firms. Our results remain qualitatively the same if we instead use the original five categories. We also classify high-tech employment using 4-digit NAICS codes following Kile and Phillips (2009).9 In addition, we use demographic information from QWI to analyze employment based on education level: QWI reports five education categories, which we again collapse in two groups. “Completed college” workers include QWI groups “Some college or Associate degree” and “Bachelor’s degree or advanced degree”, and “Not completed college” includes QWI categories “Less than High school” and “High school or equivalent, no college”. Our results are similar if we use instead the original education categories provided by QWI. We start our sample in 2012Q1, six quarters before the introduction of the first anti-patent troll legislation in Vermont, and end in 2015Q3, the last quarter of data available. We obtain establishment data from the Quarterly Census of Employment and Wages 9

Specifically, the following industries are classified as high-tech: 3254, 3341, 3342, 3344, 3345, 3346, 3353, 3391, 5112, 5133, 5141, 5172, 5173, 5179, 5181, 5182, 5413, 5415, 5416, 5417, 5161, 5171, 5179, 5191.

–8–

(QCEW) database at the (4-digit) NAICS industry-state-year-quarter level. Similarly to employment data, we aggregate establishment data at the 4-digit NAICS industry level, which we subsequently collapse in two groups, high-tech and non high-tech establishments, following the classification provided by Kile and Phillips (2009). We collect this data from 2012Q1, six quarters before the introduction of the first anti-patent troll legislation in Vermont, and end in 2016Q1, the last quarter of data available.10 Venture capital data is from Thomson’s VentureXpert. We isolate early stage venture capital rounds raised by high-tech firms founded in 2005 or later.11 In addition, we exclude firms that are not in high-tech industries. Statewide bankruptcy data are from the Administrative Office of the United States Courts. Bankruptcies are classified as business (e.g., corporations or partnerships) and non-business (e.g., consumer debt). Finally, we also include several macroeconomic controls in our regression specification. We control for state-level quarterly GDP (log-transformed), state-level quarterly income per capita (log-transformed) from Bureau of Economic Analysis (BEA), state level quarterly unemployment rates from Bureau of Labor Statistics (BLS), entrepreneurial activity proxied by the number of patents granted in a given state and year from the United States Patent and Trademark Office (USPTO). Summary statistics are reported in Table 1. Panel A reports summary statistics for the entire sample. The average state-level employment is approximately 2.1 million. Of this, about 20% (399 thousand) is for establishments with fewer than 20 employees and 8% is in industries that we classify as high-tech. On average, there are 256 thousand new hires each quarter and 350 thousand separations. The mean total payroll is $28 billion. The number of business bankruptcies at the state level averages 150 per quarter, while the number of non-business bankruptcies averages over 14 thousand. On average, there are 15 unique firms that receive VC funding in each state for an aggregate dollar amount of $138 10

Note establishment data are not available by size (in employees) at the quarterly frequency.

11

Specifically, we focus on companies classified as “startup/seed”, “early stage, or “expansion.” The inclusion of firms founded before 2005 does not have a material effect on the results.

–9–

million. As expected, there is significant variability for both of these variables; the standard deviation is over three times the mean value for both of the VC outcomes of interest. This variation is driven by the concentration of startups and VC funding in a few states (e.g., California). For the control variables, the mean state-level quarterly GDP is $309 billion, income per capita averages $44,600, the number of granted patents is nearly 3 thousand, and the average unemployment rate during our sample period is 6.3%. Panel B compares the average values for states with and without patent troll laws for the four quarters prior to the first law. Specifically, the first column reports the mean for states that adopt patent reform at some point during the sample, and the second column reports the mean for those that do not. The difference between these values and corresponding p-values (accounting for clustering at the state-level) are reported in Columns 3 and 4, respectively. This table indicates the observable characteristics of the treated and control groups are similar prior to the adoption of patent laws. Specifically, while states without patent laws tend to be larger than those without (in terms of employment, payroll, etc.), this difference is not statistically significant for the variables of interest in this paper. While not statistically different at conventional levels, the mean value for the number of unique firms funded by VC and the dollar amount of VC funding are considerably larger for states that do not have patent troll laws relative to those that eventually adopt them. This difference is primarily driven by the presence of California in the “never treated” group. To address this, we exclude California from the analysis in Table A3 and find that this does not have a material effect on the main findings.

III.2

Methodology

To identify the effect of the passage of the laws on employment, we estimate the following difference-in-differences regression specification at the state-year-quarter level:

yst =αt + λs + δ · P atentLawst + β · Xst−1 + st ,

(1)

where s denotes a state and t denotes a year-quarter. The dependent variable yst is our

– 10 –

measure of employment; P atentLaw takes a value of one the year-quarter the legislation is signed at a given state and is 0 otherwise; Xst−1 is the vector of control variables; λs is a state fixed effect and αt is a year-quarter fixed effect. Xst−1 includes the lagged effect of macroeconomic controls (GDP, income per capita, unemployment rate) and of patent activity. Robust standard errors are clustered at the state level. An important feature of this empirical setting is that it allows the same state to be part of the treated and control groups at different points in time. It also allows some states not to be treated at all or to some states being treated since the start date of the sample. In fact, 23 states are never treated in our sample (19 have not passed a law and 3 have passed a law in 2016 – after our sample ends), while the remainder are treated at some point during our sample period. Overall, 18% of state-year-quarter observations are treated observations. The staggered nature of the events therefore helps to reduce potential noise and bias that may result from analyzing a single event (Roberts and Whited, 2012).

IV

Results: Patent Troll Laws and Employment

We begin by examining employment changes for subsets of firms based on their size and on whether they operate in high-technology industries. Because patent trolls tend to target smaller firms and those operating in high-tech industries, we predict the effect of state reforms will be concentrated among such firms. To conduct this analysis, we use QWI classifications to study the effect of legislation on firms of three sizes: 0-19, 20-549, and 50-500+. Table 2 presents the results of this analysis. Column 1 tests the effect of patent troll laws on employment of small (0-19 employees), high-tech firms. The specification controls for state fixed effects (to control for time-invariant state characteristics) and year-quarter fixed effects (to control for time-varying heterogeneity). Column 2 repeats the specification in Column 1 adding time-varying macroeconomic controls at the state level. While the inclusion of such controls may be problematic if they are affected by the legislation, in practice we find the estimates are largely unchanged. Specifically, for both specifications

– 11 –

the coefficient for the patent law indicator is positive and statistically significant at the 5% level. The estimated effects are economically significant as well: the passage of the law is associated with an approximately 2% increase in employment for small, high-tech firms relative to the control group. The point estimates for the macroeconomic controls have the expected sign, with more positive economic outcomes predicting higher employment. Columns 3 and 4 of Table 2 report the effect of patent troll laws on medium (20-49 employees) and larger (50-500+ employees) firms in high-tech industries. In contrast to the previous results, we find no evidence that the laws are associated with changes to employment for these firms. Columns 5-7 examine the effect for firms that are not in high-tech industries. Again, the results are statistically indistinguishable from zero. Overall, this table indicates that the effects of state patent troll laws are primarily confined to small businesses in high-tech industries. This finding is consistent with evidence suggesting patent trolls target high technology firms and, over time, have shifted their focus to smaller companies as they are more likely to settle once faced with the threat of litigation. The fact we only observe an increase in employment in one out of the seven employment subsets mitigates concerns regarding confounding factors that could drive the results, i.e. an omitted variable that is correlated with the adoption of patent troll laws. Such an omitted variable would need to have differential effects on small and large firms and high-tech and non-high-tech industries in order to explain our findings. Figure 1 shows the within-state variation in employment as a function of changes in patent troll legislation for small firms in high-tech industries. Specifically, the figure plots the average quarterly employment change between treated and control states in quarters t=-6 to t=+6, while controlling for state and year-quarter fixed effects and macroeconomic controls.12 Prior the passage of the laws, there is no difference in the employment of treated and control states. However, employment increases in treated states by about 2% following 12

We focus on six quarters around the adoption of the laws as employment data for most states that adopt patent trolls laws is not available past this. We exclude observations outside of this window as well as the year of adoption so the point estimates in the figure can be interpreted relative to the signing date (i.e., quarter t in the figure).

– 12 –

the laws as compared to control states, and this increase persists for six quarters. While the individual point estimates are not all statistically significant at conventional levels, the regression analysis in Table 2 indicates they are jointly significant at the 5% level. In Table A2, we further examine whether the adoption of patent troll laws is related to state-level economic trends. Such trends may pose a problem if they drive the adoption of the laws as well as subsequent increases in employment in small, innovative firms. However, the linear probability model presented in this table suggests the signing of the laws is unrelated to state-level GDP, income per capita, unemployment rate, and patenting activity. In sum, these findings are consistent with the parallel trends assumption necessary for empirical identification in this setting. Next, we examine whether the increase in employment is driven by firms hiring more new employees or firing fewer existing employees. The results of this analysis are presented in Table 3. The sample is confined to small firms (0-19 employees) in high-tech industries. The dependent variable in Columns 1 and 2 is the logarithm of the number of new hires, defined as the number of workers who started a new job in the specified quarter which continued into the next quarter. The dependent variable in Columns 3 and 4 is the logarithm of the number of workers whose job with a given employer ended in the specified quarter. Column 1 indicates an economically significant increase in new hires of 3.3% in treated states relative to control states following the adoption of legislation. This effect is statistically significant at the 10% level, and the estimate is similar in terms of magnitude and statistical significance when macroeconomic controls are included in the specification (Column 2). In contrast, we do not find a significant effect on employee separations. Specifically, the point estimates reported in Columns 3 and 4 are positive, though not statistically different from zero.13 In Columns 5-6, we examine the effect of patent troll laws on total firm payroll. Consistent with the fact that employment increases, we find that overall wage cost to firms also goes up. Total wages are higher by about 3%, a coefficient statistically 13

Our proxy for firing not only captures involuntary separations but also voluntary separations of workers who change firms. A positive coefficient may therefore reflect the fact that the increase in the supply of jobs due to the legislation (and, subsequently, an increase in workers’ outside options) results in a higher propensity of workers to move and to a larger number of voluntary separations.

– 13 –

significant at the 5% level and robust to the inclusion of macroeconomic controls. This increase in total wages suggests the change in employment is not driven by hiring lower paid workers. Table 4 sheds further light on what type of employment increases in states with patent troll laws. Again, we confine the sample to small firms (0-19 employees) in high-tech industries. The outcomes in this table characterize employment according to workers’ levels of education (i.e., college and non-college). Consistent with the high concentration of college educated workers in high-tech industries, we find much of the increase in overall employment is driven by such workers. Specifically, patent troll laws are associated with a 2% increase in employment of college educated workers, which is statistically significant at 5% (1%) level without (with) controls in Column 1 (Column 2). Interestingly, we also find evidence of an increase in employment for workers without college degrees; the point estimates in columns 3 and 4 are similar to those for college educated workers, though they are statistically noisier and only the specification with macroeconomic controls is significant at conventional levels. This result may reflect the need to hire more employees for lowskill jobs (e.g., janitors) when there is an increase in employees with technical jobs (e.g., engineers). This suggests, in turn, that the benefits of the legislation are not only captured by the right tale of the wage distribution, namely the higher skilled workers, but may also benefit workers in the left tale of wage distribution, as proxied by workers with no college education, thereby creating jobs for a variety of socio-economic groups.

V

Mechanisms: Patent Troll Laws, Firm Survival, and Firm Financing

Section IV shows that the passage of state-level anti-patent-troll legislation leads to an increase in employment at small high-tech firms. In this section, we aim to shed light on the potential mechanisms driving this effect. We begin by examining whether patent troll laws are associated with a decrease in business bankruptcies. We next study the laws’ effect on the financing environment faced by small innovative firms by analyzing how anti-troll

– 14 –

legislation impacts venture capital investment.

V.1

Patent Troll Laws and Business Bankruptcies

Small, privately-held firms often face important financial constraints (e.g., Farre-Mensa and Ljungqvist 2016). Those that are targeted by patent trolls may have to devote their limited resources to defending themselves in court or to settling out-of-court, which may not only limit the firms’ ability to grow and create jobs, but may even force some of them into bankruptcy. Table 5 examines the effect that the passage of anti-troll legislation in a state has on the number of business bankruptcies in that state. Column 1 shows that patent troll laws are associated with a 3.8% decrease in the number of business bankruptcies in a state, although this estimate is not statistically significant (p=0.199). When we introduce our usual state-level macroeconomic controls in Column 2, the estimated reduction in the number of bankruptcies increases to 4.9% and becomes significant at the 10% level. Thus, the evidence in Table 5 is consistent with the notion that part of the employment effect of anti-troll laws captured in Section IV reflects the fact that by protecting small firms from frivolous lawsuits, the laws have a positive effect on small firms’ survival probability. As expected, Columns 3 and 4 in Table 5 show that anti-troll laws have no economically or statistically significant effect on the number of non-business bankruptcies in a state. This placebo test helps alleviate concerns that our findings may be driven by unobservable shocks that coincide, both in time and geography, with the passage of the laws.

V.2

Patent Troll Laws and Venture Capital Investment

In addition to reducing the instances of frivolous patent lawsuits and thus the negative consequences that these lawsuits have on the targeted firms, patent troll laws also reduce the litigation threat faced by all those firms in a state that could at some point be targeted by trolls. By reducing the firms’ litigation risk, anti-troll laws can increase the net present value (NPV) of investing in the high-tech small firms that patent trolls often target, thereby

– 15 –

making the firms more attractive to professional investors such as venture capitalists. Table 6 investigates this hypothesis by examining how the passage of anti-troll legislation in a state affects venture capital investment in the state.14 In Columns 1 through 4, we include all 50 states in our analysis, while Columns 5 through 8 focus only on those states with a level of VC activity in 2012 (the year before the first anti-troll law was passed) at or above the median of all 50 states. In Column 1, we find that the passage of anti-troll legislation in a state is associated with a 6.8% increase in the number of firms raising venture capital in the state, though this effect is not statistically significant (p=0.256). We continue to find positive but insignificant estimates in Column 2 (where we include state-level macroeconomic controls) and in Columns 3 and 4, where the dependent variable is the dollar amount of capital invested by VCs in the state (log-transformed). By including all states in our analysis, Columns 1-4 include many states that have historically had very little VC activity: Recall that Table 1 shows that in the median state-quarter, only three firms raise venture capital. When we exclude those states with below-median VC activity in Columns 5 through 8, we find that the effect of anti-troll laws on VC investment becomes stronger, both economically and statistically. Indeed, Column 5 shows that the number of firms raising venture capital increases by 18.7% with the passage of anti-troll legislation (significant at the 5% level). This growth in the number of firms raising venture capital translates into a 28.9% increase in the amount of capital invested by VCs in the state which is significant at the 10% level (Column 7). These effects are only slightly smaller when we include our macroeconomic controls in Columns 6 and 8, respectively. The results in Table 6 are thus consistent with the notion that by reducing the litigation 14

Venture capital is not the only source of funding for startups, but prior work has shown that it is particularly helpful to the kinds of innovative startups that tend to be targeted by patent trolls (Hellmann and Puri 2000; Kortum and Lerner 2000; Gompers and Lerner 2001). In addition to offering funding, VCs provide monitoring and advice (Hellmann and Puri 2002; Bernstein, Giroud, and Townsend 2015), access to networks of potential customers, suppliers, and strategic partners (Hochberg, Ljungqvist, and Lu 2007), and help recruiting talented individuals (Gorman and Sahlman 1989).

– 16 –

threat faced by small innovative firms in a state, anti-troll laws make firms in that state more attractive to venture capitalists–although this effect does not appear to be strong enough to draw VCs to states with historically scarce VC presence. Further reinforcing the importance of this financing channel, Table 7 shows that the positive effect of anti-patent-troll laws on the employment of small high-tech firms is driven by states with a level of VC activity in 2012 at or above the median of all 50 states. Indeed, Column 1 shows that in high-VC-presence states, anti-patent-troll laws lead to a 2.2% increase in the employment of small firms in high-tech industries, which is significant at the 5% level, and a 2.6% increase when we include macroeconomic controls in Column 2, significant at the 1% level . By contrast, the employment increase in states with belowmedian VC presence is 1.1% and insignificant (Columns 3 and 4). Table 8 shows that in high-VC-presence states, anti-troll laws also lead to an increase in the number of high-tech establishments: In Column 1, the estimated increase is 2.1% (significant at the 5% level), and it increases to 2.7% (significant at the 1% level) in Column 2. By contrast, Columns 3 and 4 again show no significant effect in states with belowmedian VC presence. The results in Table 8 thus suggest that the growth in employment we observe after the passage of anti-patent-troll legislation is driven both by the growth of existing establishments and by the creation of new ones.

VI

Conclusion

We explore the impact of the passage of staggered state patent troll legislation on employment. Patent reform targets bad faith assertions of patent infringement, an increasingly common practice by patent trolls, namely entities that own patents but do not make or use the patented technology directly. We show that, following the change in regulation, state employment of high-tech small businesses increases by 1.8%. At the same time, we find no significant effects on employment of non high-tech or large businesses. A direct explanation of these findings is that patent litigation threat imposes costs on high-tech firms, namely those frequently targeted by trolls, and these costs have real effects on job

– 17 –

creation for exactly those firms for which these costs are more likely to be binding. The following example illustrates this point: Shipping & Transit [a patent troll] has turned its sights on scores of small online retailers and logistics startups. It typically demands licensing fees of $25,000 to $45,000, amounts just small enough to discourage a legal battle, yet painful for businesses with only a few employees. “I am literally losing sleep over this,” said Pat Nastri, chief operating officer of CD Universe, an online seller of music, movies and games. The $25,000 sought by Shipping & Transit, he said, “is one of our employees’ salaries” (WSJ, 10/27/2016). We also find support of a financing channel driving the growth in employment. We show that VC funding increases following the passage of troll laws and this increase seems to feed employment growth, firm survival and creation as evidenced by an increase in employment and number of firms and a decrease in bankruptcies in states with abovemedian VC presence at the beginning of our sample period. The economic intuition behind these findings is that by simply reducing the threat of frivolous IP lawsuits, anti-troll laws can increase the net present value (NPV) of investing and such higher NPV may not only affect the investment decisions of the firms’ themselves, but also of investors in small, hightech firms, such as venture capitalists (VCs). Our findings suggest that measures aimed at curbing the recent explosion in patent litigation may play an important role in reducing both real and financing frictions faced by small businesses. Overall, these findings are important in light of a heated debate how to address the trends that lead to increasingly polarized U.S. labor markets in recent decades (Autor and Dorn, 2013; Acemoglu et al., 2015). Policymakers understand the increasing need to promote sectors of the economy that can foster economic growth and create jobs, such as high-tech, small, innovative firms.15

15

The view that small businesses create the most jobs has been more contentious in the literature. Despite early findings by Birch(1971, 1981, 1987) that small businesses create more jobs, Haltiwanger, Jarmin, and Miranda (2010) find a more nuanced pattern. More recently, Newmark, Wall, and Zhang (2011) provide further support for the view that small businesses create more jobs, albeit their results indicate smaller differences compared to prior studies.

– 18 –

References [1] Acemoglu, D., D. Autor, D. Dorn, G. H. Hanson, and B. Price, 2015, “Import Competition and the Great US Employment Sag of the 2000s,”Journal of Labor Economics 34, S141-S198. [2] Asay, C. D., et al., 2015, Letter to Congress (March 3). [3] Autor, D., and D. Dorn, 2013, “The Growth of Low Skill Service Jobs and the Polarization of the U.S. Labor Market,”American Economic Review 103, 1553-1597. [4] Benmelech E., N. Bergman, and A. Seru, 2011, “Financing Labor,”Working Paper. [5] Benmelech E., C. Frydman, and D. Papanikolaou, 2016, “Credit Market Disruptions and Employment during the Great Depression: Evidence from Firm-level Data,”Working Paper. [6] Bernstein, S., X. Giroud, and R. R. Townsend, 2015, “The Impact of Venture Capital Monitoring,”Journal of Finance 71, 1591-1622. [7] Birch, D., L., 1979, “The Job Generation Process,”Washington DC: MIT Program on Neighborhood and Regional Change for the Economic Development Administration, U.S. Department of Commerce. [8] Birch, D., L., 1981, “Who Creates Jobs?,”Public Interest 65, 3-14. [9] Birch, D., L., 1987, “Job Creation in America: How Our Smallest Companies Put the Most People to Work,”New York: Free Press. [10] Bryant, T., L., 2012, “The America Invents Act: Slaying Trolls, Limited Joinder,”Harvard Journal of Law & Technology 25, 674-695. [11] Chien, C. V., 2015, “Startups and Patent Trolls,” Stanford Technology Law Review, forthcoming. [12] Chodorow-Reich, G., 2014, “The Employment Effects of Credit Market Disruptions: Firm-level Evidence from the 2008-09 Financial Crisis,”Quarterly Journal of Economics 129, 1-59. [13] Cohen, L., U. G. Gurun, and S. D. Kominers, 2016, “Patent Trolls: Evidence from Targeted Firms,”Working Paper. [14] Cotropia, C. A., J. P. Kesan, and D. L. Schwartz, 2014, “Unpacking Patent Assertion Entities (PAEs),”Minnesota Law Review 99, 649-700. [15] Council of Economic Advisers, 2016, “The Patent Litigation Landscape: Recent Research and Developments.” [16] DeSisto, R., 2015, “Vermont vs. the Patent Troll: Is State Action a Bridge Too Far?”Suffolk University Law Review 109, 109-130.

– 19 –

[17] Farre-Menda, J., and A. Ljungqvist, 2016, “Do Measures of Financial Constraints Measure Financial Constraints?”Review of Financial Studies 29, 271-308. [18] Gompers, P., and J. Lerner 2001, “The Venture Capital Revolution,”Journal of Economic Perspectives 15, 145-168. [19] Gorman, M., and W. A. Sahlman 1989, “What Do Venture Capitalists Do?”Journal of Business Venturing 4, 231-248. [20] Giroud, X., and H. Mueller, 2015, “Firm Leverage, Consumer Demand, and Employment Losses during the Great Recession”Quarterly Journal of Economics, forthcoming. [21] Giroud, X., and H. Mueller, 2016, “Redistribution of Local Labor Market Shocks through Firms’ Internal Networks”, Working paper. [22] Gugliuzza, P., 2015, “Patent Trolls and Preemption,”Virginia Law Review 101, 15791647. [23] Haltiwanger, J., R. S. Jarmin, and J. Miranda, 2013, “Who Creates Jobs? Small versus Large versus Young,”The Review of Economics and Statistics 95, 347-361. [24] Hellman, T., and M. Puri, 2000, “The Interaction Between Product Market and Financing Strategy: The Role of Venture Capital,”Review of Financial Studies 134, 959-984. [25] Hellman, T., and M. Puri, 2002, “Venture Capital and the Professionalization of StartUp Firms: Empirical Evidence,”Journal of Finance 57, 169-197. [26] Hochberg, Y. V., A. Ljungqvist, and Y. Lu, 2007, “Whom You Know Matters: Venture Capital Networks and Investment Performance,”Journal of Finance 62, 251-301. [27] Kiebzak, S., G. Rafert, and C. E. Tucker, 2016, “The Effect of Patent Litigation and Patent Assertion Entities on Entrepreneurial Activity ”Research Policy 45, 218-231. [28] Kile, C. O., and M. E. Phillips, 2009, “Using Industry Classification Codes to Sample High-Technology Firms: Analysis and Recommendations,”Journal of Accounting, Auditing & Finance 24, 35-58. [29] Kortum, S., and J. Lerner, 2000, “Assessing the Contribution of Venture Capital to Innovation,”RAND Journal of Economics 31, 674-692. [30] Neumark, D., B. Wall, and J. Zhang, 2011, “Do Small Businesses Create More Jobs? New Evidence for the United States from the National Establishment Time Series,”The Review of Economics and Statistics 93, 16-29. [31] Patent Progress, 2016, Guide to State Patent Legislation. [32] Roberts, M. R., and T. M. Whited, 2012, “Endogeneity in Empirical Corporate Finance, ”in George Constantinides, Milton Harris, and Rene Stulz, eds. Handbook of the Economics of Finance 2, 493-572.

– 20 –

[33] Smeets, R., 2015, “Does Patent Litigation Reduce Corporate R&D? An Analysis of US Public Firms,”Working Paper. [34] Tucker, C. E., 2014, “Patent Trolls and Technology Diffusion: The Case of Medical Imaging,”Working Paper.

– 21 –

Figure 1: Map of Patent Troll Laws This figure shows states that have adopted a patent troll law. Signing dates are provided in Table A1.

– 22 –

Figure 2: Small Business Employment Coefficient Trend – High-Tech This figure shows the timing of the effect of patent laws on small business employment in high-tech industries. Point estimates are obtained from the baseline specification including state and yearquarter fixed effects, except Patent Law is replaced with indicators for each quarter before/after signing of the law (time 0 in the figure). The specification also includes controls for state-level GDP, income per capita, unemployment, and granted patents. The signing quarter is omitted from the specification, so the magnitude of each point estimate is relative to t+0 in the figure. Standard errors are clustered at the state-level, and dashed lines indicate the 90% confidence interval.

0.12 0.1 0.08 0.06 0.04 0.02 0 -0.02 -0.04 t-6

t-5

t-4

t-3

t-2

t-1

t

– 23 –

t+1

t+2

t+3

t+4

t+5

t+6

Table 1: Summary Statistics Panel A reports summary statistics for the full sample. Panel B compares the mean values for states that adopt patent laws at some point during the sample (column (1)) and those that do not have patent laws at any point during the sample (column (2)) for the four quarters prior to the first law (i.e., 2012Q2 - 2013Q1). The p-value for the difference between these values is reported in column (4) and accounts for clustering at the state level.

Panel A N

Mean

Median

SD

Total Employment, thousands

745

2066.9

1331.4

2258.7

Employment (<20), thousands

745

399.0

255.8

466.9

Employment (High-Tech), thousands

745

164.7

90.3

219.3

Hires, thousands

745

256.2

174.8

284.5

Separation, thousands

745

350.8

234.4

397.7

Total Payroll, billions

745

28.0

15.7

34.6

Bankruptcies (Business)

900

150.1

92.0

196.9

Bankruptcies (Non-Business)

900

4772.0

3247.5

5562.6

Establishments (High-Tech)

850

14679.6

8719.5

16518.3

VC (# Unique firms)

900

15.1

3.0

47.7

VC ($ Amount), millions

900

137.8

9.0

649.2

GDP, billions

745

308.9

188.7

377.3

Income per Capita

745

44599.1

43556.0

7217.2

Patents

745

2980.5

1100.0

5860.5

Unemployment Rate

745

6.3

6.3

1.7

– 24 –

Panel B Eventually Treated

Never Treated

Difference

P-value

Total Employment

1768.0

2405.2

637.2

0.373

Employment (<20)

338.2

483.0

144.8

0.35

Employment (Tech)

136.6

196.6

60.0

0.402

Hires

219.2

278.3

59.1

0.489

Separation

301.1

390.7

89.5

0.465

Total Payroll

22.2

34.0

11.9

0.278

Bankruptcies (Business)

168.7

210.0

41.3

0.599

Bankruptcies (Non-Business)

5297.3

6117.1

819.8

0.723

Establishments (High-Tech)

13270.4

15153.3

2630.2

0.706

VC (# Unique firms)

5.2

27.8

22.6

0.166

VC ($ Amount), millions

22.1

181.2

159.0

0.199

GDP

255.4

375.6

120.2

0.324

Income per Capita

42430.8

45427.9

2997.1

0.175

Patents

1851.4

4209.1

2357.7

0.211

7.5

7.1

0.4

0.396

Unemployment Rate

– 25 –

Table 2: Employment by Firm Size This table reports the effect of state patent laws on the employment by firm size. The dependent variable in columns (1)-(4) is employment in high-tech industries (see text for definition), and the dependent variable in columns (5)-(7) is employment in non-tech industries. Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

Tech Firms Firm Size=

Patent Law

Non-Tech Firms

0-20

0-20

20-50

50-500+

0-20

20-50

50-500+

(1)

(2)

(3)

(4)

(5)

(6)

(7)

0.0183** 0.0195** -0.00875 -0.00194

-0.00355 -0.00305

(0.0089)

(0.0053) (0.0055) (0.0043)

Ln(GDP)

(0.0077)

(0.0187) (0.0112)

0.0000

0.552*** (0.165)

Ln(Income per Capita)

0.0420 (0.223)

Unemployment Rate

-0.00692* (0.0040)

Ln(Patents)

0.0117 (0.0383)

State FE

yes

yes

yes

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

yes

yes

yes

Observations

745

745

745

745

745

745

745

0.024

0.142

0.002

0.000

0.001

0.001

0.000

Within R-squared

– 26 –

Table 3: Small Business Hires, Separations, and Payroll – High-Tech This table reports the effect of state patent laws on aspects of employment for small firms (<20 employees) in high-tech industries. The dependent variables are the natural logarithm of new hires (columns (1) and (2)), separations (columns (3) and (4)), and total payroll (columns (5) and (6)). Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

Ln(Hires)

Patent Law

Ln(Separations)

(1)

(2)

(3)

(4)

0.0327*

0.0321*

0.0103

0.0121

(0.0179)

(0.0176)

(0.0148)

(0.0148)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

Ln(Payroll) (5)

(6)

0.0281** 0.0292** (0.0124)

(0.0109)

0.587

1.168***

0.851***

(0.506)

(0.362)

(0.225)

0.273

-0.0959

-0.0276

(0.453)

(0.469)

(0.329)

0.00910

-0.00303

-0.00774

(0.00963)

(0.00882)

(0.00543)

0.0634

-0.0440

0.0413

(0.0921)

(0.114)

(0.0573)

State FE

yes

yes

yes

yes

yes

yes

Year-quarter FE

yes

yes

yes

yes

yes

yes

Observations

745

745

745

745

745

745

0.008

0.024

0.001

0.043

0.023

0.123

Within R-squared

– 27 –

Table 4: Small Business Employment by Education Level – High-Tech This table reports the effect of state patent laws on employment by education level for small firms (<20 employees) in high-tech industries. The dependent variables are the natural logarithm of employment of college graduates (columns (1) and (2)) and employment of non-college graduates (columns (3) and (4)). Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

Ln(College Emp.)

Patent Law

Ln(Non-College Emp.)

(1)

(2)

(3)

(4)

0.0182**

0.0195***

0.0165

0.0181*

(0.00837)

(0.00723)

(0.0113)

(0.00992)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

0.433**

0.683***

(0.162)

(0.164)

0.0380

0.0904

(0.230)

(0.281)

-0.00774*

-0.00975*

(0.00392)

(0.00519)

0.00832

0.0321

(0.0373)

(0.0435)

State FE

yes

yes

yes

yes

Year-quarter FE

yes

yes

yes

yes

Observations

745

745

745

745

0.030

0.135

0.014

0.152

Within R-squared

– 28 –

Table 5: Bankruptcies This table reports the effect of state patent laws on the number of bankruptcies within a state. The dependent variables are the natural logarithm of the number of business bankruptcies (columns (1) and (2)) and non-business bankruptcies (columns (3) and (4)). Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

Business bankruptcies

Patent Law

Non-business bankruptcies

(1)

(2)

(3)

(4)

-0.0383

-0.0486*

0.0254

0.0237

(0.0294)

(0.0267)

(0.0239)

(0.0236)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

-0.416

-0.324

(0.496)

(0.695)

-1.976**

-0.355

(0.769)

(0.447)

0.0350**

0.0221

(0.0165)

(0.0135)

-0.0945

-0.104

(0.165)

(0.108)

State FE

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

Observations

900

900

900

900

0.003

0.047

0.006

0.054

Within R-squared

– 29 –

Table 6: VC Financing This table reports the effect of state patent laws on VC funding at the state level. The dependent variables are the natural logarithm of the number of VC-funded firms and the dollar amount of VC financing. The sample includes all states in columns (1)-(4) and states with greater than the median level of 2012 VC financing in columns (5)-(8). Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

All States

Patent Law

States w/ Above Median VC

Ln(# Firms)

Ln($ Amont)

Ln(# Firms)

(1)

(2)

(3)

(4)

(5)

0.0678

0.0562

0.0757

0.0775

0.187** 0.144** 0.289* 0.229*

(0.0590) (0.0579) (0.116) (0.112)

(0.0795) (0.0698) (0.148) (0.124)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

(6)

Ln($ Amont) (7)

(8)

0.808

0.401

-0.107

0.595

(0.849)

(1.763)

(2.174)

(4.262)

-0.291

1.525

-3.219

-3.995

(1.483)

(2.498)

(3.044)

(4.520)

-0.00247

-0.0700

-0.00641

-0.0397

(0.0316)

(0.0870)

(0.0523)

(0.136)

0.503*

0.684

1.716**

3.101**

(0.263)

(0.673)

(0.690)

(1.447)

State FE

yes

yes

yes

yes

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

yes

yes

yes

yes

Observations

900

900

900

900

486

486

486

486

0.002

0.010

0.001

0.008

0.019

0.042

0.011

0.030

Within R-squared

– 30 –

Table 7: Small Business Employment by VC Activity – High-Tech This table reports the effect of state patent laws on employment for small firms (<20 employees) in high-tech industries. The sample includes states with above median VC financing in 2012 in columns (1)-(2) and states with below median 2012 VC financing in columns (3)-(4). Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

High VC

Patent Law

Low VC

(1)

(2)

(3)

(4)

0.0221**

0.0264***

0.0112

0.0109

(0.00999)

(0.00818)

(0.0151)

(0.0132)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

0.533***

0.564**

(0.155)

(0.216)

0.406**

-0.310

(0.196)

(0.307)

0.000504

-0.00722

(0.00369)

(0.00652)

0.00810

-0.0158

(0.0373)

(0.0468)

State FE

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

Observations

404

404

341

341

0.065

0.228

0.007

0.101

Within R-squared

– 31 –

Table 8: Number of Establishments by VC Activity This table reports the effect of state patent laws on the number of business establishments. The sample includes states with above median VC financing in 2012 in columns (1)-(2) and states with below median 2012 VC financing in columns (3)-(4). Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

High VC

Patent Law

Low VC

(1)

(2)

(3)

(4)

0.0212**

0.0266***

-0.0105

-0.0126

(0.00947)

(0.00918)

(0.0125)

(0.00857)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

0.324

0.692***

(0.192)

(0.146)

0.740**

-0.502***

(0.339)

(0.140)

0.00226

-0.0114**

(0.00565)

(0.00539)

-0.0694

0.00505

(0.0452)

(0.0253)

State FE

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

Observations

459

459

391

391

0.041

0.140

0.012

0.317

Within R-squared

– 32 –

Table A1: Patent Troll Laws

State

Date Signed

Alabama

3/18/2014

Arizona

3/24/2016

Colorado

6/5/2015

Florida

6/2/2015

Georgia

4/15/2014

Idaho

3/26/2014

Illinois

8/26/2014

Indiana

5/5/2015

Kansas

5/20/2015

Louisiana

5/28/2014

Maine

4/14/2014

Maryland

5/5/2014

Mississippi

3/28/2015

Missouri

7/8/2014

Montana

4/2/2015

New Hampshire

7/11/2014

North Carolina

8/6/2014

North Dakota

3/26/2015

Oklahoma

5/16/2014

Oregon

3/3/2014

Rhode Island

6/4/2016

South Carolina

6/9/2016

South Dakota

3/26/2014

Tennessee

5/1/2014

Texas

6/17/2015

Utah

4/1/2014

Vermont

5/22/2013

Virginia

5/23/2014

Washington

4/25/2015

Wisconsin

4/24/2014

Wyoming

3/11/2016

– 33 –

Table A2: Predictive Regressions This table reports the effect of state economic conditions on the timing of the adoption of patent laws. Observations for states that adopt a patent law are excluded from the sample after the law is signed. The explanatory variables are lagged by 1 quarter in columns (1)-(4) and by 4 quarters in column (5). Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

Patent Law Signed

Ln(GDP )t-1

(1)

(2)

(3)

(4)

0.563

0.897

0.844

0.738

(0.554)

(0.579)

(0.632)

(0.636)

-1.532

-1.601

-1.795

(1.236)

(1.204)

(1.205)

-0.00726

-0.00521

(0.0182)

(0.0177)

Ln(Income per Capita)t-1

Unemployment Ratet-1

Ln(Patents)t-1

(5)

0.272 (0.194)

Ln(GDP)t-4

0.275 (0.764)

Ln(Income per Capita)t-4

0.805 (1.351)

Unemployment Ratet-4

0.0150 (0.0258)

Ln(Patents)t-4

0.0654 (0.244)

State FE

yes

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

yes

Observations

662

662

662

662

512

0.200

0.204

0.204

0.208

0.235

R-squared

– 34 –

Table A3: Robustness Test: Excluding California This table excludes California from the main analysis. Panel A repeats the analysis in Table 2, and Panel B repeats the analysis in Table 5. Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

Panel A Tech Firms Firm Size=

Patent Law

Non-Tech Firms

0-20

0-20

20-50

50-500+

0-20

20-50

50-500+

(1)

(2)

(3)

(4)

(5)

(6)

(7)

0.0192** 0.0195**

-0.0074

-0.0010

-0.0028

-0.0026

0.0006

(0.0090)

(0.0190) (0.0113)

Ln(GDP)

(0.0078)

(0.0053) (0.0055) (0.0043)

0.550*** (0.168)

Ln(Income per Capita)

0.0476 (0.225)

Unemployment Rate

-0.00675 (0.00408)

Ln(Patents)

0.0113 (0.0389)

State FE

yes

yes

yes

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

yes

yes

yes

Observations

730

730

730

730

730

730

730

0.027

0.138

0.001

0.000

0.001

0.001

0.000

Within R-squared

– 35 –

Panel B

All states

Patent Law

States w/ above median VC in 2012

Ln(# Firms)

Ln($ Amont)

Ln(# Firms)

(1)

(3)

(5)

0.0688

(2)

(4)

(6)

Ln($ Amont) (7)

(8)

0.0539 0.0910 0.0856

0.191** 0.139* 0.317**

0.237*

(0.0594) (0.0582) (0.117) (0.114)

(0.0803) (0.0711) (0.150)

(0.126)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

0.852

0.257

0.0145

0.397

(0.873)

(1.799)

(2.348)

(4.563)

-0.256

1.487

-3.369

-4.185

(1.491)

(2.502)

(3.089)

(4.549)

-0.00362

-0.0657

-0.0122

-0.0395

(0.0327)

(0.0892)

(0.0566)

(0.142)

0.513*

0.651

1.789**

3.089**

(0.269)

(0.672)

(0.717)

(1.458)

State FE

yes

yes

yes

yes

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

yes

yes

yes

yes

Observations

900

900

900

900

486

486

486

486

0.002

0.010

0.001

0.007

0.019

0.043

0.013

0.030

Within R-squared

– 36 –

Table A4: Falsification Test: Small Business Employment – Non-Tech This table repeats the analysis in Table 4, but limits the sample to small firms (<20 employees) that are not in technology industries. The dependent variables are the natural logarithm of employment of college graduates (columns (1) and (2)) and employment of non-college graduates (columns (3) and (4)). Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

Ln(College Employment)

Patent Law

Ln(Non-College Employment)

(1)

(2)

(3)

(4)

-0.00203

-0.00174

-0.00706

-0.00654

(0.00451)

(0.00472)

(0.00584)

(0.00607)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

0.114*

0.0686

(0.0636)

(0.0801)

0.285**

0.410**

(0.128)

(0.164)

0.00330

0.00359

(0.00396)

(0.00527)

0.00851

0.00577

(0.0211)

(0.0255)

State FE

yes

yes

yes

yes

Year-quarter FE

yes

yes

yes

yes

Observations

745

745

745

745

0.001

0.035

0.004

0.034

Within R-squared

– 37 –

Table A5: Falsification Test: Employment by VC Activity – Non-Tech This table repeats the analysis in Table 7, but limits the sample to small firms (<20 employees) that are not in technology industries. The sample includes states with above median VC financing in 2012 in columns (1)-(2) and states with below median 2012 VC financing in columns (3)-(4). Patent Law is an indicator for whether a state has previously adopted patent reform. The list of states and corresponding signing dates is provided in Table A1. The control variables are defined in the text. Each specification includes state and year-quarter fixed effects. Robust standard errors are clustered by state. ***p<1%. ** p<5%, * p<10%.

High VC

Patent Law

Low VC

(1)

(2)

(3)

(4)

-0.00434

-0.00235

-0.00289

-0.00354

(0.00560)

(0.00459)

(0.00912)

(0.00986)

Ln(GDP)

Ln(Income per Capita)

Unemployment Rate

Ln(Patents)

0.264

0.0226

(0.184)

(0.0490)

0.546***

0.143

(0.180)

(0.200)

0.00902*

0.00644

(0.00468)

(0.00848)

0.0240

-0.0220

(0.0406)

(0.0309)

State FE

yes

yes

yes

yes

Year-Quarter FE

yes

yes

yes

yes

Observations

404

404

341

341

0.003

0.118

0.001

0.013

Within R-squared

– 38 –