Innovation activity

This section presents our empirical results. First, we show the results obtained on the number of patents successfully filed by the firms. Second, we analyze the quality of innovation, by exploiting information on the citations that each patent received from subsequent patent applications. We proceed by estimating the dynamic effects of banking deregulations and discussing a number of robustness checks.

5.1 Innovation outputs

Table 3, Panel A, shows OLS estimates using the logarithm of patent counts as the dependent variable. In Column (1), we show that the interstate deregulations had a positive effect on the innovation outputs. In particular, allowing out-of-state banks to enter the state increased innovation activity for a firm located in the state by 21%.

Meanwhile, the coefficient of intrastate deregulations is not statistically different from zero.

While in Column (1) we only control for industry and year fixed effects, in Column (2) we add the logarithm of sales and capital-to-labor ratio. Controlling for these firm characteristics reduces the magnitude of the interstate coefficient, which is however more precisely estimated and becomes significant at 1%. In Column (3), we further control for R&D stock. As expected, the stock of R&D has a positive and significant effect on patenting; however, the interstate deregulation coefficient remains significant at 1%. In Column (4), we confirm our findings by including a host of controls that may potentially affect innovation, such as firm age, HHI, ROA, tangibility and cash holdings. The interstate deregulations coefficient remains both statistically and economically relevant, indicating a 19% increase in patenting.

In Columns (5) - (8), we adopt a more restrictive specification that instead of industry fixed effects includes firm fixed effects. As expected, restricting the identification to within-firm variations leads to sensibly smaller deregulation coefficients, but the statistical significance is confirmed at 1% level. The economic magnitude of the effect is relevant as well: the most restrictive specification (Column 8), indicate a 12.7% increase in patenting. As found above, the intrastate deregulations had no relevant effects.

In Table 3, Panel B, we provide estimates from fixed-effect Poisson QMLE regressions, which take into account that patent counts are skewed to zero.⁷⁰ Similar to our OLS results, the most restrictive specification (Column 4) indicates a 14% increase in patenting following interstate deregulations.

70 An alternative approach could be to use transformations of the dependent variable to avoid losing observations with zero patents (as in Table 3, Panel A). For example, in unreported analyses we have estimated OLS regressions using the logarithm of (1 + patent counts) as dependent variable. However,

5.2 Innovation quality

Our results so far suggest that interstate deregulations caused an increase in firms’

innovation activities as measured by the raw number of patents granted. However, the existing research has demonstrated that patents differ greatly in “value” and that simple patent counts do not capture the relative importance of the underlying inventions (Harhoff et al. 1999). In this section, we measure innovation by weighing each patent using the number of future citations received from subsequent patents (Trajtenberg 1990). Forward citations reflect the economic and technological

“importance” as perceived by the inventors themselves (Jaffe et al. 2000) and knowledgeable peers in the technology field (Albert et al. 1991). Because forward citations suffer from truncation problems, we adopt patent counts weighted by truncation-adjusted citation counts from the NBER data (see e.g. Hall et al. 2001; Hall et al. 2005).⁷¹

Results reported in Table 4, where we use cite-weighted patent counts as dependent variable, indicate that interstate deregulations lead to a significant and economically relevant increase in the quality of patenting, whereas intrastate deregulations have an insignificant effect. Hence, not only the number of patents have increased but their average quality has risen as well, suggesting that the effect did not purely come from the larger supply of financing and thus lower rationing of projects being financed. We further argue that the average increase in the quality of innovations stems from a rise in the risk of innovative projects being financed.

5.3 Dynamic effects

Although U.S. states passed deregulation legislations at specific points in time, the real consequence of interstate deregulations on credit supply caused by the actual entry of banks in the new states may manifest after several years. Even patenting an innovation 71 The problem arises from the fact that “citations to a given patent typically keep coming over long periods of time, but we only observe them until the last date of the available data” (Hall et al. 2005).

Besides the use of truncation-adjusted citation counts, the problem is mitigated by the inclusion of year fixed effect. In fact, our results are robust to the adoption of unadjusted citation counts.

is the outcome of a process that sometimes can require several years. A specification that compares raw outcomes before and after deregulations, as the one used in the previous sections, may not be well-suited to capture these potential dynamic effects.

We test for dynamic effects by drawing on specifications similar to Kerr and Nanda (2009). First, we construct a dynamic difference-in-differences model employing a set of dummies that measure the distance in years from each deregulation passage, using as reference group the period three years or earlier before deregulations.

Results, reported in Table 5, Panel A, show that the coefficient prior to the interstate deregulation is small and statistically insignificant, thus confirming that our results are not driven by diverging pre-deregulation trends. By contrast, the post-deregulation coefficients are all positive and significant at conventional levels. Importantly, they become larger as we move forward from the reform year, with the largest effect corresponding to six and seven years after interstate deregulations.

Second, we allow the effect of deregulations on innovation to linearly grow over time using a variable equal to zero up to the deregulation year and then equal to the number of years since a deregulation was passed, capping the treatment effect at 8 years. Results, reported in Table 5, Panel B, confirm that interstate deregulations had a positive growing impact on firms’ patenting activity.

5.4 Robustness checks

We test the validity of our findings in several ways. We do not tabulate the results described in this section, but they are available upon request. We start by addressing the concern that other policies potentially affecting innovation were adopted around the same period of the banking deregulations. In the late 1980s, 30 U.S. states passed a set of business combination (BC) laws that reduced the threat of hostile takeovers thus weakening the governance role of the market for corporate control (Giroud and Mueller 2010; Bertrand and Mullainathan 2003). These laws might affect our analysis

through the effect of corporate governance on the managerial incentives to innovate⁷², and that effect would not be captured by our specification since BC laws impacted firms at their state of incorporation. To mitigate this concern, we control for a dummy equal to one if firms were incorporated in the states that passed a BC law, from the year of the passage onwards, and zero otherwise. Our results indicate that the positive effect of banking deregulations on firm innovation is not confounded by the passage of BC laws.

Re-examining findings in Black and Strahan (2002), Wall (2004) shows that the effect of deregulations on entrepreneurship was positive in some U.S. regions but significantly negative in others. We check in several ways how our results depend on regional effects. First, we our findings are unchanged if we augment our specifications with regional trends, computed as year averages of the dependent variables by region excluding the firm in question.⁷³ Second, we estimate region-specific deregulation effects. Results from this last exercise show that interstate deregulation coefficients are all positive, though their statistical and economic significance is not homogeneous across U.S. regions.

We perform a number of additional robustness checks that further validate our findings. First, we include state-level time varying characteristics as additional controls. In particular, we include lagged GDP growth and the logarithm of population, obtained from the U.S. Bureau of Economic Analysis. Second, we exclude observations corresponding to the year of interstate deregulations. Third, we show that our results are robust to the inclusion of linear state trends centering the identification on discontinuities surrounding the interstate deregulations (Kerr and Nanda 2009).

Fourth, to better isolate the effect of interstate deregulations, we exclude those states 72

The effect of corporate governance on innovation is ambiguous. Some empirical studies indicate that worse corporate governance reduces the incentives to innovate (Atanassov 2009). Chemmanur and Tian (2011) argue that some degree of managerial entrenchment isolates CEOs from short-term pressures, thus inducing them to focus on long-term value creation and innovate more. Sapra et al. (2011) show that the effect of corporate governance on innovation follows a U-shaped relationship.

73 Regions are defined according to the four-grouping classification provided by the U.S. Census: west, midwest, northeast and south (http://www.census.gov/geo/www/us_regdiv.pdf).

that passed intrastate deregulations within a year of interstate deregulations. Fifth, we restrict the analysis to firms that remain in the sample for at least 5 (10 or 15) years to purge the analysis from firm entry and exit. Sixth, we use contemporaneous rather than lagged controls. Seventh, we exclude firms headquartered in California and Massachusetts, since these states have a particularly high innovation activity. Eighth, we extend our sample up to 1997, i.e. the year when the implementation of the IBEEA finally enacted a nation-wide deregulation of the banking sector. Ninth, we allow for heterogeneous time and state effects by interacting all the covariates with year and interstate treatment dummies.