Machine learning could help firms select more effective board directors

Lea SternA growing chorus of institutional investors and economics critics is finding flaws in the way that corporate boards of directors advise and monitor senior management.

A key reason is that CEOs effectively control the board selection process, tending to choose directors who are unlikely to offer opposition or provide the diverse perspectives necessary to represent shareholders’ interests.

Emerging technology could help, according to new research by Léa Stern, an assistant professor of finance and business economics at the University of Washington Foster School of Business.

Stern’s analysis demonstrates that machine learning algorithms trained on the performance and attributes of potential candidates could help mitigate cognitive biases and complement CEO’s efforts to pick effective corporate directors.

“A striking result of our study is that machine learning models consistently suggest alternative directors who would have been likely both to accept the directorship and to outperform the directors actually chosen by the firms,” Stern says.

The one commonality of those alternative directors? They don’t resemble the CEO.

Defining a ‘good’ director

Measuring the performance of corporate directors is no simple task. Boards largely work behind closed doors where the actions of any individual are difficult to observe or isolate.

But Stern and her co-authors—Isil Erel and Michael Weisbach of Ohio State University and Chenhao Tan of the University of Colorado—devised a novel benchmark for the effectiveness of individual directors: popularity with shareholders. Specifically, they measured the fraction of votes each director receives in annual shareholder re-elections.

“These votes reflect the support the director personally has from the shareholders and should, in theory, incorporate all publicly available information about the director’s performance,” says Stern.

With this common metric as a guide, the research team set out not only to diagnose existing deficiencies in corporate boards, but also to develop a more disciplined method of selecting board directors in the future.

Training the model

They began by examining the directors of a large set of publicly traded corporations between 2000 and 2011. This data set “trained” the machine learning algorithm they developed to accurately predict the performance of directors.

They tested the algorithm on a separate set of directors who joined firms between 2012 and 2014. The algorithm proved well calibrated: directors predicted to do well performed much better than those predicted to do poorly.

Stern and her colleagues constructed a pool of potential directors for each board opening from among the directors who joined the boards of smaller nearby companies.

What they found was eye-opening.

Crony capitalism

The difference between the directors suggested by the algorithm and those actually selected by firms allowed the researchers to assess the features that are overrepresented—and overrated—in the director nomination process.

Compared to the algorithm’s recommendations, they found that the directors who were actually hired but predicted (accurately) to underperform tended to be male, with large networks, lots of past and present board experience and a finance background.

“In a sense,” Stern says, “the algorithm is telling us exactly what institutional shareholders have been saying for a long time: that directors who are not old friends of management and come from different backgrounds both do a better job in monitoring management and are often overlooked.”

Building better boards

Stern suggests that there are two possible reasons why real-world firms persist in appointing sub-optimal directors who they could predict will be unpopular with shareholders.

The first is that CEOs do not want effective directors on their boards. When they maintain management-friendly boards, they are able to consolidate power over their firms.

The second is that management is not able to select effective directors due in part to behavioral biases.

A well-designed machine learning algorithm, on the other hand, makes decisions based on rules rather than discretion.

Stern cautions that machine learning is a developing technology. “We advocate for tools built on algorithms as decision aids, not substitutes for human judgement,” she says. “Achieving the right balance in the division of labor between humans and machines to take advantage of their relative strengths is key.”

“For the purpose of our study, though, it is clear that algorithms are not prone to agency conflicts and the biases that occur when boards and CEOs meet together to select new directors. Institutional investors are likely to find this attribute particularly appealing and encourage boards to rely on an algorithm for director selections in the future.”

How well this approach to selecting directors will be received by management, Stern admits, is an open question.

Selecting Directors Using Machine Learning” is the work of Isil Erel, Léa Stern, Chenhao Tan and Michael Weisbach.