Over-confidence and the Probability Assessment Ability of Retail Investors

Abstract

This study directly tests of the ability of retail investors to assess the probability of pay-offs to various financial assets over short term (7 days) and medium term (3 month) periods. Psychology studies have demonstrated that decision makers are typically over-confident in their ability to assess uncertainty (Lichtenstein, Fischhoff and Phillips 1982, Yates 1990). Daniel, Hirschleifer and Subrahmanyam 1998, Odean 1998, Scheinkman and Xiong 2003, and others, have studied the effect of investor over-confidence on financial market equilbria. However, it is questionable whether the results of the pyschology studies are relevant to decision making by investors in financial markets. The psychology studies show that the level of over-confidence exhibited by decision makers is highly dependent on the context of the assessment of uncertainty. The level of over-confidence varies with: the difficulty of the assessment; the rarity of the event in question; and the level of training of the assessors. Moreover, the psychology studies of over-confidence have not studied the assessment of uncertainty of financial asset pay-offs. Instead, they have focused on weather forecasting (Murphy and Winkler 1977 ), medical prognosis (Murphy and Poses 1993) and other contexts. We collect from 309 participants (mostly MBA students) their subjective assessments of the probabilities that particular assets will increase in price over a defined period. Each Monday, before markets open, participants assess the probabilities of 3 events that could occur over the next week (the SP the 10 year US Treasury yield rises; the Brent price of crude oil rises) and 3 events that could occur over the next 3 months (a specified industry and 2 individual stocks would outperform the S&P500 over 3 months). A total of 15,576 probability assessments are collected over 26 weeks. The participants exhibit a high level of over-confidence in their ability to assess the probabilities of directional changes in asset prices. However, in contrast to assumptions in the finance literature, the over-confidence is not more severe in males than females. Participants who have worked in the investment management industry exhibit less over-confidence. Over 80 percent of the participants own stocks directly. Our sample of retail investors do not provide probability assessments that have more refinement than assessments based on the long-run average of weekly changes. Thereby providing further support for the efficient market hypothesis.