Everything is backed by the mathematics of game theory and proofs showing that managers (well, CEOs primarily) will behave better if they are controlled. That makes the advice an undisputable fact except for one thing. The equations are not exactly the same as the practices, and there are many ways actual human beings will either fail to act according to predictions, as when boards do not implement the practices well enough, or will have counter-measures against the predictions, as when CEOs manipulate the board. So in a world of actual humans acting as directors on boards and CEOs of firms, does the advice hold true?
This is what we (Andrew Shipilov, Yeonsin Ahn, Timothy Rowley, and me) examined in research published in Journal of Organization Design. Our approach was simple. We had data on firm adoptions of 11 different governance practices and a series of firm outcomes, and we focused on Return on Assets, Debt, and Dividends distributed to shareholders. These are outcomes that shareholders care about because they concern profitability, risk, and money returned to the owners. We used modern machine-learning techniques to find out which of the practices predicted these outcomes best.
So, what did we find? The best way to summarize our findings is that what works in equations does not work in practice. Hardly any of the 11 practices had any effect on the three outcomes. Two that had effects – and independent audit committee improved profitability and director evaluations increased dividends – suggest that a major mechanism contributing to impact is whether directors have reason to pay close attention to the firm and counter-act selfish CEO actions. That is at least some encouraging news, though overall our findings suggest that much governance advice is hot air.