A Texas A&M University researcher calls into question the common notion that CEOs have a large effect on firm performance. Although previous research contends that the so-called “CEO effect” is substantial, a study by Professor of Management Markus Fitza, whose research centers on firm performance, suggests that most of the performance attributed to CEOs could actually be due to chance—to factors outside of CEOs’ control.

In his paper “The use of variance decomposition in the investigation of CEO effects: How large must the CEO effect be to rule out chance?” published in the Strategic Management Journal (a leading research journal for management studies), Fitza used performance data from the 1,500 largest U.S. firms from 1993 to 2012 to estimate the fraction of firm performance that can be attributed to CEOs.

“Differences in the performance of a firm during the tenures of different CEOs can be caused by at least two things,” explains Fitza, a professor at Mays Business School. “There are differences in CEO abilities as well as events that are outside of the CEO’s control—chance events.”

He says it was this chance factor that interested him and encouraged him to delve deeper. “I wanted to know how big the effect of chance on CEO performance might be.” Chance can have negative or positive effects on a firm’s performance, he notes. “For example, a scandal at a major competitor can help a firm, while an accident at an important supplier can have negative consequences. Over a long enough time period such effects tend to cancel out (a phenomenon called ‘regression to the mean’), thus it is unlikely that a firm is consistently high performing just because of chance events.”

However, Fitza says today the average CEO’s tenure is only four years, a period of time that might be too short for luck to even out. Fitza says he wondered how much of a firm’s performance during such short tenures might be due to chance, because if, for example, a lucky event early in a CEO’s tenure is not balanced by an unlucky one in such a short time period, then that CEO could be wrongfully credited for high performance that would have happened no matter who was leading the company.

To investigate this, Fitza applied a method that has been widely used to investigate the degree to which CEOs affect firm performance: variance decomposition. This method is often used to determine the percentage of the total variance of a variable that can be attributed to certain effects (such as the industries to which firms belong or the CEOs of firms).

Fitza created sets of simulated firm data in which the level of performance was left entirely up to chance. He then conducted a variance decomposition analysis on real firm data as well as on the simulated data and compared the results asking: “how big would the measured CEO effect be if all performance differences between different CEO tenures were caused by chance?'”

Doing so he showed that over 70 percent of the CEO effect measured by past studies could be due to chance.

Fitza’s findings reveal how previous studies tend to “wrongfully attribute the effect of random fluctuations to CEOs” suggesting it is difficult to differentiate between the effect of chance and real leadership skills. Thus the performance effect of CEO leadership “is much smaller than previously thought,” he explains.

Fitza says his findings could have important implications as they add to the current debate about the role and the importance of CEOs. “For example, they draw attention to the volatile environment CEOs are working in and question the current trend towards shorter CEO tenure,” he says, adding, “It might not be a good idea to replace CEOs because of low firm performance over short time periods, such as after one bad year.”

Therefore, he concludes, this work also has applications for CEO compensation. “If we do not want CEOs to be rewarded or punished for luck, then understanding their true contribution to company performance should be an important part in determining the level of their compensation.”

Source: Texas A&M University