• CEOs are often expected to be more rational and objective than others, writes Wharton professor Marius Guenzel.
  • But leaders are equally vulnerable to biases, his new research paper points out.
  • The paper challenges the notion that managers’ experience on the job will improve their decision-making over time and ultimately de-bias them.

When a celebrity CEO decides to move on, as Amazon’s Jeff Bezos announced he would later this year, it’s an opportunity to revisit what makes people in those roles successful, or flounder, as the case may be. CEO appointees often have a tailwind of strong performance and are expected to be more rational and objective than others. However, CEOs are equally vulnerable to biases, according to a new research paper titled “Behavioral Corporate Finance: The Life Cycle of a CEO Career,” by Wharton finance professor Marius Guenzel and Ulrike Malmendier, professor of finance and economics at the University of California, Berkeley.

“Behavioral biases and mistakes that people make are not because of, let’s say, a lack of education,” Guenzel said in a recent interview on the Wharton Business Daily show on SiriusXM. (Listen to the podcast above.) “These are mistakes that even the most sophisticated, most educated people make, so this really speaks to the hard-wiring of biases in people, and that biases are significant elements of human decision-making.”

The “rational-manager paradigm” makes presumptions on three fronts: (a) selection, (b) learning, and (c) market discipline, the paper stated. In the “selection” aspect, it is presumed that corporate managers “are smart and highly educated,” and therefore not susceptible to the biases of consumers and investors. In the “learning” aspect, they are expected to learn from “occasional mistakes,” and then “update rationally and optimize” their decisions in the future. The “market discipline” aspect is where managers are closely monitored by their boards and the market, “keeping any bias-driven errors at bay.”

However, that line of reasoning is flawed, according to behavioral corporate finance research over the past 15 years or so, the researchers wrote in their paper. “A convincing body of evidence documents systematic and persistent biases in managerial decision-making, including overconfidence, reference-dependent thinking, and reliance on cognitive shortcuts, and reveals that managers’ character traits and past experiences shape their decisions.”

“We focus on managerial biases and how these biases play a role in each of the different career phases – CEO appointments, the CEO being at the helm of the firm, and then being dismissed eventually,” said Guenzel. “The traditional arguments for why a CEO, we would think, is rational are CEO selection, learning and market discipline. Our contribution is to say that it’s not clear that these arguments are sufficient to prevent biased decision-making at the very top of organizations.”

In the first phase, the authors examined why selection mechanisms fail to filter out biased candidates, or worse, favor candidates with certain biases. In the second phase where a CEO is in charge of a company, they looked at the “systemic biases that CEOs exhibit” and why they fail to learn from mistakes. In the third phase, they questioned if boards and markets are aware of CEO biases, how biased CEOs are incentivized, and when they are replaced.

Research has shown that biased decision-makers — overconfident decision-makers — are actually more likely to be appointed to the CEO position, said Guenzel. If a CEO’s abilities cannot be directly observed, past performance is a guide. “But good past performance can happen for a lot of reasons; for example, if the person just made some risky moves and implemented some risky strategies because they are overconfident. That leads to overconfident people eventually being filtered into the top of organizations, rather than being filtered out.”

As an example of biased CEO decisions that could prove costly, the paper cited the 2019 announcement from Airbus that it would discontinue its flagship A380 aircraft “after years of persistent cost explosions, development failures, and canceled orders.” The decision immediately lifted shareholder wealth by $1.9 billion but also caused 3,500 job cuts, the paper noted.

Pitfalls in CEO Selection

The paper explored the selection criteria for CEOs in two scenarios: when their biases are not observable, and when they are observable. In cases where biases are unobservable, boards considering a set of risk-averse managers but looking to maximize value would appoint “the manager with the highest payoff as CEO since ability has to be inferred from payoffs,” the authors write. Overconfident CEOs tend to choose projects with a high-risk-high-reward profile, which makes it more likely that they will be appointed as the CEO.

M&A deals are examples where CEO overconfidence “can have detrimental effects on shareholder value,” said Guenzel. Things could go wrong “if a CEO overestimates the synergies they can create by acquiring another firm where, in fact, there is low potential for synergies, and the long-term performance of that combined firm being negative relative to if they had just been keeping going on as stand-alone businesses.”

CEO overconfidence could also seep into decisions on hiring other members of the executive team. “If I’m an overconfident CEO, I might prefer a chief financial officer who has similar viewpoints as me,” said Guenzel. “Having people next to you that think like you and behave like you can accelerate or intensify bad decision-making at the top, when we have biased decision-makers.”

When managers’ biases are observable, and if they have a “bright side” where the benefits outweigh the costs, “value-maximizing boards may deliberately seek managers with observable advantageous biases and character traits,” the paper stated. Overconfidence is a bias that can have such an apparent “bright side,” at least in certain industries or environments, that “it counteracts risk aversion and thus induces risk averse CEOs to choose investment levels closer to the first-best.” Biases among board members may be another reason why the CEO selection process gets distorted, the researchers noted in their paper. For instance, biased boards might be prone to appoint CEOs of the same gender, with a similar cultural background, or other salient similarities.

Why learning isn’t a sufficient remedy

The paper challenged the notion that managers’ experience on the job will improve their decision-making over time and ultimately de-bias them. It offered four reasons why that is a limited option. First, many measurable corporate events, such as acquisitions, “occur at low frequency” and are rare in a CEO’s tenure. “Thus, opportunities to learn from previous mistakes are few and far between.”

Second, learning from past decisions is limited as it is difficult to distinguish between causality and correlation of managerial decisions and outcomes. “Output is hard to measure, hard to attribute to specific individual performances, and hard to disentangle from other firm-specific or economy-wide events,” the paper explained.

Third, people with a “self-attribution bias” tend to attribute successes to their own actions but failures to external circumstances, the paper stated, citing prior research. “Even if performance evaluations were accurate, managers might draw wrong inferences and discount information that could induce learning.”

Finally, certain biases might even be reinforced, rather than ameliorated, as top managers overestimate the causal impact of their decisions, the paper stated. “CEOs likely believe they are in control, and they are personally invested because firm performance determines their reputation and pay.”

“There’s always the question — can we link [a biased CEO’s] performance directly, causally to some outcome, or is this just correlation?”

—–Marius Guenzel

When to fire a CEO

In the third phase, the authors examined the question of when biased CEOs are dismissed. Much depends on whether boards watch out for CEO biases, whether corporate governance interventions work with changes in compensation or dismissal, and whether boards themselves are biased, they stated.

CEO biases do not necessarily imply a higher rate of dismissal, and much depends on whether those biases are value-destroying or value-enhancing, according to the paper. Governance mechanisms and whether a board deliberately appointed a biased CEO are other factors that influence decisions on CEO dismissals. “If board members are biased, they might misjudge a CEO’s performance and make suboptimal retention and dismissal decisions, independently of whether the CEO is biased or not,” the authors noted.

Counteracting biases

Guenzel and Malmendier identified some lessons for organizations in each of the three phases of a CEO career life cycle. In preventing biases from influencing decisions on hiring CEOs, a first step would for both the manager candidate and the employer to have increased awareness of each other’s biases.

Another measure may require CEO candidates to be tested for biases in the selection process. “For example, in fast-changing environments such as the fashion industry or the renewable energy sector, selecting a CEO who systematically under- or overreacts to new information could be particularly costly,” they wrote. Interestingly, they noted that managers with a financial background “appear to exhibit fewer biases, at least in certain investment and financing contexts.”

In counteracting biased decision-making by CEOs, organizations could attempt “corporate repairs,” the paper stated. Those would include actions like identifying common mistakes, organizational redesign, procedural changes or different hiring practices, the authors recommended, drawing upon previous research. CEO overconfidence could be particularly harmful at firms with abundant cash flows, they noted. Here, they recommended processes where, for instance, the five most critical assumptions in projects are evaluated by two “uninvolved managers.” In settings where a firm may want to avoid a CEO’s social biases, it could require a two-stage process for project funding requests, anonymizing first-round proposals, it added.

Compensation is “a key tool to align the interests of managers and shareholders,” and modifying contract design could help check biases such as overconfidence or aversion to losses, according to the paper. Significantly, CEOs with low or no risk aversion tend to receive more performance-contingent compensation than those who are “highly risk-averse,” they noted, citing prior papers. Governance mechanisms such as regulation could also help curb biases. After the passage of the 2002 Sarbanes-Oxley Act that aimed to elevate accounting standards and increase board independence and governance stringency, firms with overconfident CEOs showed “improved acquisition performance,” they added, citing prior research.

That said, traditional governance mechanisms to align managerial and shareholders’ incentives may be largely ineffective in curbing CEO biases, the paper noted. The researchers found more promise in tools such as “the strategic use of debt overhang or procedural changes” to tame CEO biases. They also called for board members to “account for their own potential biases and mistakes in their judgment and evaluation of CEO performance, such as attributional errors and hindsight bias.” Corporate repairs and training for those who monitor managers would also help, they added.

It isn’t easy to build a cast iron case linking CEOs’ biases to adverse outcomes and fire them. “Ideally, when we see a CEO who is biased and makes bad decisions, we want to immediately fire them,” said Guenzel. “But then, there’s always the question – can we link their performance directly, causally to some outcome, or is this just correlation? And that’s a big impediment to why we cannot just easily fire biased decision-makers.”