סמינר באסטרטגיה

: Raffaella Sadun, Harvard Business School

Abstract: 

We develop a new measure of CEO behavior based on detailed activity level data collected for over 1,100 CEOs in six countries (Brazil, France, Germany, India, UK and US). 

We use machine learning to project the high dimensionality of the data into one measure of CEO behaviour. 

According to this measure CEOs are at or between two extremes: those who are directly involved in production and engage in unplanned, one-on-one meetings, and those who mostly engage in large, pre-planned meetings with several high-level functions and participants. 

A firm-CEO matching model makes precise that under optimal matching CEO behaviour is uncorrelated with firm performance, while when information frictions are sufficiently severe some CEOs get hired by the wrong firms,thereby generating a non-negative correlation. 

We reject the null of zero correlation as firms led by CEOs who coordinate several functions have higher productivity, which is consistent with this CEO behavior being relatively rare and the existence of significant frictions. 

CEO behavior does not predict performance before the CEO is hired or in the first two years of CEO tenure: the productivity effect occurs after two years, thus allaying endogeneity concerns. 

We also find that CEO-firm matching frictions are particularly acute in developing countries, and may account for a large fraction of the cross country productivity differential observed in the data.