Injunctions destroy innovation---statistically significant effect

In the previous sections, I showed that the Supreme Court decision led to an increase in patenting, both at the intensive and extensive margins. Furthermore, this change in enforcement also positively affected patent quality and R&D investments. Overall, this evidence suggests that patent litigation had real distortive effects on firms’ ability to innovate in 2006, and the decision was able to reduce some of this burden faced by innovative firms. In this section, I explore why patent litigation affects innovation by firms.

6.1 Litigation lowers innovation returns: evidence from the composition of innovation

Firms exposed to litigation may reduce innovation for different reasons. The most intuitive channel is that patent litigation lowers the returns from investing in innovation. Since direct involvement in patent litigation can be very expensive (Bessen and Meurer, 2013), firms will take into account this risk when assessing whether to invest in a project. As a result, when the risk of patent litigation is too high, firms may choose to forgo some good investment opportunities. If this channel is quantitatively important, I should expect to find two results in the data. First, firms operating in more intensively litigated areas should be more positively affected. This is what I found in the main results. Second, within a firm, projects in an area where patent litigation is more intense should become relatively more valuable. This reshuffle should happen in every firm, irrespective of whether it is more or less exposed to litigation. In other words, every firm should perceive the investment in riskier patents to be more valuable.

In order to provide evidence in favor of this idea, I study whether firms experienced a relatively increase in risky patents after the decision. In order to focus on within-firm resource allocation, I sort patents applied for by each firm across two categories – risky and non-risky – depending on whether they belong to one of the USPTO technology classes in the top 10% (or 25%) of litigation. This reshape of the data implies that each firm has two observations per period. Since I am interested in the within-firm allocation, I can now test whether risky patents increased relatively more after the decision conditional on a full set of firm-by-time fixed effects. In practice, I estimate the following equation:

[[CHART OMITTED]]

where αjt is a set of firm-time fixed effects, αjr is a set of fixed effects at the firm-group level, 1{Riskr} is a dummy for riskier groups. As mentioned above, I group patents in two classes, such that r = {high risk; low risk}. If the return channel is the driving force behind the response of innovation to the ruling, I would expect risky patents to grow substantially more than non-risky patents within the firm portfolio, which is β > 0. Standard errors are clustered at firm-level.

In this analysis, I consider two outcomes: first, I explore the intensive margin of the effect by looking at ln(patjtr), which is the logarithm of the grant patent that firm j applied for during time t in the class of risk r. To obtain a purely intensive margin, I estimate this regression with a subset of firms that are simultaneously active in both risk classes around the decision time. Second, I look at the extensive margin with yjtr equal to 1{P atjtr > 0), which is a dummy equal to one if the firm j applies for any subsequently granted patent in risk-group r at time t. In this case, my sample is much larger, since I consider every firm that has applied for at least one patent in the ten years before the decision.

Results are reported in Table (6). When I look at the intensive margin, I find that patents belonging to more intensively litigated patent classes do not appear to increase relatively more within a firm’s portfolio. Estimates are very small and noisy.54 On the other hand, I find that firms are more likely to patent in a risky class in the two years after the decision, rather than in the two before. The results are similar whether risky patents are defined as being in the top 10% or the top 25%. Lastly, in Table (A.12) I use the full panel dimension – without collapsing pre- and post – and I estimate the effect differentially for pre- and post- relative to the quarter before the decision. This analysis shows that this effect is not driven by differential trends in patenting before the decision.

At least partially, these results are consistent with the return channel: the decision also shifted the patenting behavior of firms across classes, in particular by making companies more likely to patent in a more risky area after the decision. While a similar effect is not identified at the intensive margin, these results are in line with the reshuffle idea that should occur if the decision were to increase the perceived returns of R&D investment.55

6.2 Litigation exacerbates financial constraints

Operating in a high-litigation environment can also hinder innovation by reducing the amount of resources available for R&D. The idea that exposure to litigation can deplete corporate resources is intuitive. Firms in sectors where litigation is more intense are more likely to pay large settlements or overpays for licensing agreements. This happens because companies want to avoid the escalation of legal conflicts to courts or just limit their negative consequences, as in the BlackBerry case discussed previously. Furthermore, ex-ante these companies may be forced to devote larger resources to monitor potential threats and modify their products to minimize the risk of litigation. For example, eBay in the 2006 10-K recognizes that litigation claims “whether meritorious or not, are time consuming and costly to resolve, and could require expensive changes in our methods of doing business, or could require to enter into costly royalty or licensing agreements.” In response to this, companies may invest more intensively in defensive tools, such as a large legal department within the company, which seems to have some effects on deterring attacks (Cohen et al., 2019).

If the financing of innovation were frictionless, this shift of monetary resources should not affect firms’ ability to invest in good projects. In reality, firms face constraints in funding innovation (Brown et al., 2009; Hall and Lerner, 2010), and therefore a reduction in internal resources has an impact on firms’ ability to innovate. When this is the case, intense patent litigation exacerbates this financing problem and therefore it increases the inefficiency in funding R&D. To test whether this theory is true in the data, I examine the heterogeneity of the decision effects across firms characterized by a differential likelihood of being financially constrained. If this channel is relevant, I expect companies that are more likely to be financially constrained to react more positively to the shock. In other words, this story would predict a higher elasticity between investment in R&D and a reduction in litigation costs for companies facing more financial frictions.

In order to study this, I modify the standard model described by equation (1) by adding an interaction with a dummy F inConj , which is equal to one for firms that are more likely to be financially constrained. More specifically, I estimate:

[[CHART OMITTED]]

Furthermore, I separately study the behavior of the two groups of firms. In line with previous discussion, I would expect β1 > 0.

Following the relevant literature in finance, I identify firms that are more likely to be financially constrained in three different ways. First, I study the differential behavior of small versus large firms. Previous research has found that smaller firms tend to have a harder time accessing external funding (Fazzari et al., 1988; Chodorow-Reich, 2014; Bottero et al., 2015). In my setting, I focus on smaller firms within the public firm sample. In particular, I construct two definitions of small firms, looking at whether they are below the median of employment or revenue. Second, I identify firms with no rating on public debt as companies that are more likely to be financially constrained (Kashyap and Lamont, 1994; Almeida et al., 2004). More specifically, I look at firms with no rating reported in the three years before the Supreme Court decision. Lastly, I examine the heterogeneity across firms that pay and do not pay cash dividends, looking at the three years before the decision.

The results are reported in Tables (7) and (8). The decision led to an increase in R&D intensity only for firms more likely to be financially constrained. When splitting the sample across the two groups, I systematically find that the coefficient is positive and significant for the financially constrained group, while non-significant and small for the other group. When using the full sample, more financially constrained firms increase R&D intensity more. This is true across all the measures, although it is not statistically significant in some cases. Lastly, I find that more financially constrained firms did not respond more than non-financially constrained firms in terms of patent applications.56

As a robustness, I show that, in my case, the results are not simply capturing heterogeneity across firms in growth (Farre-Mensa and Ljungqvist, 2015). To rule this out, I augment equation (6) by fully interacting measures of firm growth in the two years before the decision to my treatment. In particular, in Table (A.15) I report the results looking at revenue growth. I find that, if anything, the main coefficient β1is estimated more precisely when I add the growth controls. In an unreported Table, I find the same when looking at asset growth. Overall, my analysis is not just capturing a spurious correlation of these measure of financial constraint with different growth trajectories.

These results suggest that a decline in R&D returns is not the only channel through which patent litigation may affect innovation. Instead, financial constraint is an important dimension to consider when evaluating the effect of operating in area where litigation is intense.

Even if the largest negative returns are experienced in the day in which the news became public, the portfolio continues to experience negative returns for the following month, reaching the bottom in mid-June. Overall, these results hold when using alternative models (Figure A.4).

In summary, these facts confirm that public NPEs suffered a great deal around the Supreme Court decision. In particular, the shock led to a large drop in market value, which did not revert back in the weeks that followed. The results are robust to the removal of each of the NPEs considered in the sample.21 Overall, this evidence demonstrates that the ruling was a critical event in patent enforcement and greatly affected the players in this market. Furthermore, these results confirm that the decision was not completely anticipated by market participants.