AN EMPIRICAL TEST OF LEARNING IN MANAGEMENT EARNINGS FORECASTS

Abstract

My dissertation examines whether managers issuing earnings guidance learn from the forecast errors in prior earnings guidance issued by them. Using data on quarterly earnings forecasts issued by managers during the period from 2001 to 2016, I find results that are consistent with managers learning from their previous forecast errors to improve their forecast accuracy. However, the intensity of the managers' reactions to previous forecast errors is asymmetric. Consistent with prior literature that emphasizes the importance of meeting or beating forecasts for managers, certain managers that miss their own forecasts tend to be conservative enough in their future forecasts to avoid missing their own forecasts again. However, as expected, when the managers have met or beaten their previous forecasts, they have a smaller forecast error, but they still beat their previous forecasts. Additional analysis suggests that these effects persist even after controlling for potential earnings management to achieve these earnings targets. I also examine the impact of managerial attributes and board governance characteristics on the learning process. My analysis suggests that while CEO overconfidence and CFO overconfidence appear to impede learning, Managerial ability, CEO duality and outside CEO(s) as director(s) strengthen the learning effect. My findings shed light on an important aspect of management guidance and may have implications for users of this information such as financial analysts and investors.