CEO Overconfidence, Trading Activity, and Performance Evidence from REITs

The effects of overconfidence have been mainly investigated investorsí decisions regarding stock portfolios, and far less so the decisions of managers at corporations. The few exceptions that do investigate the effect of behavioral biases within a corporate setting include most notably the empirical studies by Malmendier and Tate (2005a, b, 2008), Malmendier, Tate, and Yan (2006), and Crane and Hartzell (2009). Malmendier and Tate (2005a) relate overconfidence to investmentñcash-flow sensitivity of CEOs and show that CEO overconfidence has a positive effect on investment, which is measured using firm capital expenditures. Crane and Hartzell (2008) investigate a different behavioral aspect, the disposition effect, using a REIT sample. REITs offer an ideal setting to investigate professional managersí investment decisions. Investment and divestment decisions can easily be identified since REITs mainly purchase and sell buildings, so most of the investment decisions can be followed by the researcher. Additionally, REIT asset values are transparent, making it possible to observe how much a firm spent when investing in a project. We investigate the effect of CEO overconfidence on trading activity using REITs. Using a REIT sample gives us the opportunity to identify each investment decision and observe how frequently REIT managers trade and to relate trading activity to overconfidence, and do that separately using purchases and sales. We define a CEO as overconfident if he buys more of his own company stock than he sells throughout the whole sample period, following Malmendier and Tate (2005a). We find that overconfident CEOs significantly purchase more properties than non-overconfident counterparts. Between 6% and 13%, we also document that overconfident CEOs are significantly more likely to hold properties than their non-overconfident counterparts. A CEO being a net buyer of his own companyís stock might either indicate overconfidence or access to private information. However, with private information, we expect managers to perform better, while overconfident CEOs trade non-optimally so must have worse performance. We evidence that they have worse operating performance than their non-overconfident counterparts. To separate overconfident managers from those who have access to private information, we combine being a net buyer of own company stock with the performance of the company. First, we interact net buyer with bad performance to have an extended measure of overconfidence. Additional to this, we also interact a net buyer dummy with a good performer dummy to proxy for managers having private information. An interaction dummy of being a bad performer and being a net buyer of own company stocks gives stronger results than the net buyer dummy alone. Overconfident CEOs purchase more and sell less. When we use an interaction dummy of being a good performer and a net buyer of own company stocks, it even has opposite impacts on purchasing and sales activities indicating that CEOs with private information behave and trade different from overconfident CEOs.