eThesis - Kauppakorkeakoulun sähköiset gradut

PURPOSE OF THE STUDY

The purpose of this thesis is to study whether highly overconfident or hubristic CEOs are randomly distributed across firms, or whether they are more often promoted in those companies where their optimistically biased beliefs could be expected to be relatively most useful. Previous studies have presented evidence of a number of ways in which overconfidence causes decision-making to deviate from rational first-best policies. I study the relationship between firm-specific risk prior to CEO selection and the level of hubris indicated by the eventual appointee specifically in their self-promotion tendencies. In this way, I add to the strand of literature that suggests CEO hubris would be more common in firms with a high degree of firm-specific risk.

DATA AND METHODOLOGY

I study a sample of 191 randomly selected CEOs who, according to data from Standard & Poor's Execucomp, were promoted to lead US-based public companies in 1991-2009. After extensive background checks, this sample excludes founder-entrepreneurs, heirs and individual investors whose selection is unlikely to represent a competitive CEO contest. The association or lack thereof between CEO hubris and idiosyncratic risk is assessed by employing the prominence of a CEO's photograph in annual reports immediately after promotion as a proxy for overconfidence and regressing it on idiosyncratic risk measures that pre-CEO selection stock market data from CRSP yields. The annual reports are collected from Thomson Financial Worldscope.

FINDINGS OF THE STUDY

High idiosyncratic risk during the five years prior to CEO selection does have a borderline significant statistical relationship with the level of the appointed CEO's hubristic tendencies. This finding is new, although the robustness of its significance is much lower when time effects are controlled for. Also, the statistical association is detected assuming major variations in a company's idiosyncratic vs. systematic risk. Therefore the magnitude of the risk effect cannot be considered very meaningful at least in the scope of this study, and call for further research.