Board independence and operating performance: analysis on (French) company and individual data

ABSTRACT

This article studies the relationship between board independence and firm operating performance in French listed companies. We take advantage of an original database, with a time-series dimension that can be used to mitigate heterogeneity and dynamic endogeneity issues. In addition, this database can be disaggregated at the individual (director) level. This design enables us to introduce firm fixed effects and individual fixed effects in firm performance equations, thereby controlling for heterogeneity at the firm and individual levels. Our main result is to document a significant negative relationship between independence and accounting performance. This result suggests that, in the French context, the costs of independence (i.e. the informational gap supported by independent directors compared to insiders and affiliated directors) outweigh the benefits of independence (i.e. the reduction in agency costs).

KEYWORDS:

• Board structure
• firm performance
• independent directors
• GMM estimator
• individual heterogeneity

JEL CLASSIFICATION:

• G30
• G34
• C23

Acknowledgment

We thank Loïc Dessaint, from Proxinvest (http://www.proxinvest.com/index.php/fr/page/index.html), for giving us free access to the Proxinvest database on directors, as well as precious advices on the use of this dataset. We acknowledge financial funding from the Chaire FDIR Finance Durable et Investissement Responsable and the Institut Caisse des Dépôts et Consignations pour la Recherche. We thank in particular Isabelle Laudier and Didier Janci for their support during this research. This research has been conducted as part of the project Labex MME-DII (ANR11-LBX-0023-01). We are finally grateful to Tristan Auvray, Andrea Bassanini and Vincent Bouvatier for useful comments and suggestions. All errors are, obviously, ours.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Two empirical studies examine, directly or indirectly, independence in the French context. Ammari, Kadria, and Ellouze (2014) estimate the relation between the fraction of independent directors and accounting performance, for 40 large French listed companies. They find that it is negatively correlated with the ROA and positively correlated with the ROE. Ginglinger, Megginson, and Waxin (2011) examines employee board-level participation. The fraction of independent directors is introduced as a control in OLS regressions: this proportion is not correlated with ROA.

² There are pieces of evidence in support of this view. In particular, independence is associated with a greater CEO performance–turnover sensitivity (Weisbach 1988; Bhagat and Bolton 2008).

³ Proxinvest is the leading proxy voting agency in France.

⁴ All our results are robust to the exclusion of the medium-sized companies (76 companies out of 331 never appeared in the SBF250 over the period).

⁵ We control for 12 industries: Agri-food industry, energy and mining, consumer goods industry, other manufacturing, construction, wholesale and retail trade, insurance and real estate, IT, media and communication, health and social services, transportations, other business and personal services.

⁶ Turnover is defined as the number of shares traded in a given year divided by the total number of outstanding shares.Both stock price volatility and stock turnover rate are extracted from the Proxinvest database. They are missing for some observations (see Table 1). To avoid reducing the sample size, we set missing observations of both variables equal to zero and include for each variable a dummy that equals one if the information is available, zero otherwise. This dummy allows the intercept term to capture the mean of both variables for missing values.

⁷ Note that it is in contrast with Ammari, Kadria, and Ellouze (2014), who do not make the distinction between affiliated and inside directors.The proportion of women is a driver of a firm performance in the French context, as shown by Sabatier (2015).

⁸ Basically, this means creating one instrument for each variable and lag distance, rather than one instrument for each time period, variable, and lag distance. For more details, see Roodman (2009).

⁹ Note that we do not intend to use variation in independence status across companies for a given year as a source of variation. To do so, director–year fixed effects (ψ_it) should be introduced in Equation 2 instead of director fixed effects (ψ_i). We do not follow this strategy for two reasons. First, using director–year fixed effects (ψ_it) dramatically increases the number of variables on the right-hand side of Equation 2, thereby reducing the precision of the estimates: even aggregate variables that where significant in Equations 1 and 3 lose their explicative power. Second, business cycles are captured in Equation 2 by year fixed effects µ_t.

¹⁰ To identify this unconnected sub-group, we use the STATA command felsdvregdm (Mihaly et al. 2010).

¹¹ We reproduce our previous regressions (GMM) on this sub-sample: results are all consistent (with a slight decrease in the significance level due to the smaller number of observations).

Additional information

Funding

This work was supported by the Chaire FDIR Finance Durable et Investissement Responsable; Institut Caisse des Dépôts et Consignations pour la Recherche.