Should Directors Have Term Limits? – Evidence from Corporate Innovation

Abstract

This paper examines the effect that directors with extended tenure have on corporate innovation based on a sample of US firms from 1996 to 2006. Using the propensity-score matched-pair research design, I find that firms with a higher portion of outside directors enjoying extended tenure produce significantly fewer patents and that these patents receive fewer subsequent citations. These firms also have lower research and development (R&D) productivity and exploration intensity than their matched control firms, although I found no significant difference in their R&D investment intensity. Difference-in-differences tests based on director deaths and regulatory changes in the early 2000s suggest that the adverse effect of long director tenure on innovation performance is causal. I also find that the effect is mitigated when long-tenured directors have more years of overlap in service with CEOs, and when long-tenured directors are executives at other firms. Finally, I find that boards with extended tenure attenuate the contributions of innovation outputs to future firm value and performance. These findings shed new light on the debate over length of board tenure and provide another justification for imposing term limits on directors.

JEL Classification:

• G34
• O31
• O34

Acknowledgements

I am grateful for helpful comments from the editor Laurence van Lent and an anonymous referee. I remain responsible for any remaining errors or omissions.

Notes

1 Regulations in the UK, Australia, India, Hong Kong, Singapore, among other countries and regions, have imposed term limits on directors of between 9 and 10 years.

2 I thank the referee for suggesting this research design.

3 Compared with exploration innovation, exploitation innovation tends to have less impact and benefits, as evidenced by fewer patents and subsequent patent citations, as well as lower contributions of R&D investments to the firms' future revenues.

4 Prior studies have examined various market and firm characteristics that affect innovation, including the participation of venture capital (Kortum & Lerner, 2000; Tian & Wang, 2014), especially corporate venture capital (Chemmanur, Loutskina, & Tian, 2014), CEO overconfidence (Hirshleifer, Low, & Teoh, 2012), analyst coverage (He & Tian, 2013), private ownership (Ferreira, Manso, & Silva, 2014), involvement of institutional investors (Aghion, Van Reenen, & Zingales, 2013; Ferreira et al., 2014), conglomerate form (Seru, 2014), competition (Aghion, Bloom, Blundell, Griffith, & Howitt, 2005), stock liquidity (Fang et al., 2014), debtor-friendly legal environments (Acharya & Subramanian, 2009), stringent labor laws and lower union powers (Acharya, Baghai, & Subramanian, 2013; Bradley, Kim, & Tian, in press), and financial market development (Hsu, Tian, & Xu, 2014).

5 The sample period starts in 1996 because it's the first year that director data become available in the RiskMetrics database. The sample period ends in 2006 due to availability of patent data from the NBER Patent Citation database at the time of this research.

6 I tried to fill in missing values for research quotient whenever possible by doing the calculations myself.

7 As a robustness check, I also reconduct the analyses using firm-year observations without missing R&D data. In the baseline analysis, the mean R&D intensity for the pooled treatment group (T) and control group (C) is 0.068 and 0.061, respectively. The difference is statistically insignificant, which is consistent with the results reported in Table 4 Panel A.

8 The reason for using a patent's application year rather than its grant year is that previous studies (such as Griliches, Pakes, & Hall, 1988) have shown that the former is superior in capturing the actual time of innovation.

9 The truncation problem arises because the patents appear in the NBER patent database only after they are granted. I observe a gradual decrease in the number of patent applications as I approach the last few years of my sample period. This observation is because the lag between a patent's application year and its grant year is significant (about two years on average) and many patent applications filed during these years were still under review and had not been granted by 2006. To adjust the truncation bias in patent counts, I supplement the NBER database with the Harvard Business School (HBS) patent database, which contains patents granted through 2010. To the extent that the patent application outcomes have been announced by 2010 for the patents filed by 2005 (the last year of my sample period), this approach largely mitigates the patent truncation concern. Neither the NBER nor the HBS patent database is likely to be affected by the survivorship bias. As long as a patent application is granted by the USPTO, it is attributed to the applying firm at the time of application even if the firm later gets acquired or goes bankrupt.

10 I thank the referee for suggesting this variable.

11 See Armstrong et al. (2010) and Lu et al. (2001) for a detailed discussion of the nonbipartite matching algorithm.

12 See for example, Rosenbaum and Rubin (1983) and Lemmon and Roberts (2010) for a more detailed discussion of the matching method and cautionary notes. I used random forest as an alternative matching algorithm based on Wager and Athey (2015), and results remain qualitatively unchanged.

13 I match firms on the pre-event four-year averages of innovation variables because many of these variables have values of zero, which makes it difficult to calculate meaningful percentage growth measures. Therefore, to satisfy the parallel trends assumption, I match firms on both the numerator and denominator of a hypothetical ‘percentage growth rate’ for innovation outputs.

14 This identification strategy offers at least three advantages over identification based on sudden deaths of directors. First, the number of firms that are affected by the regulatory changes is larger than the number of firms who experience director death. Second, regulation-induced changes in board composition can potentially lead to changes in the percentage of long-tenured directors in both directions (i.e. increase and decrease) whereas deaths only result in a decrease of percentage of long-tenured directors. Third, regulations affect more directors at a given firm whereas director deaths only affect a small group of directors. I thank the referee for suggesting this identification strategy and for pointing out the advantages of this identification over the sudden deaths of directors.

15 Several high-profile intellectual property-based transactions provide good anecdotes. For example, Google purchased Motorola Mobility for $12.5 billion, primarily to obtain 14,600 granted patents and 6700 pending patents. Similarly, Nortel Network's patent portfolio was purchased for $4.5 billion by several companies, including Apple, Microsoft, and Research in Motion.

Additional information

Funding

The financial support of the National Natural Science Foundation of China (Project 71372049) and BNP Paribas-Tsinghua SEM Center for Globalization of Chinese Enterprises is appreciated.