Overconfidence in investment decisions: An experimental approach

Abstract

By experimentally inducing risk aversion, overconfidence in an investment setting is investigated, comparing the evaluation of actual investment decisions with alternative choices. After selecting their own investment, subjects confront three alternative investment choices, including the optimal one, and are asked about their willingness to pay and to substitute their own for alternative choices. Overconfidence is defined as the persistent overevaluation of the own investment decision. Results indicate that overconfidence increases (i) with the absolute deviation from optimal choices, (ii) with task complexity involving the number of risky assets, and (iii) decreases with individual perceived uncertainty.

Keywords:

• Risky decision making
• behavioural finance
• portfolio choice
• experimental economics

Notes

1. According to (A.1) the decision maker chooses i such that it maximizes W. By (1) and (A.2) the probability is, however, determined by (4). Denote by the utility of earning M with certainty and by the corresponding utility for . Due to (A.1), . Since cardinal utilities are only uniquely defined up to positively affine transformations, we can set and , so that .

2. One will not want to switch from i to î if one considers i as better than î. In such a case one wants to be rewarded for giving up i. On the other hand, one might regret one's choice and prefer î over i. In such a case one would be even willing to pay for switching from i to î. Our test of over(under)confidence requires that all three limits and are positive (negative).

3. If is the optimal limit we classify a decision maker as persistently overconfident when l(i, î) is significantly larger than . On the other hand, we will speak of persistent underconfidence if l(i, î) is significantly smaller than .

4. More specifically, we asked them: ‘How many lottery tickets are you willing to pay in order not to have to switch to the alternative investment level?’ and ‘How many lottery tickets do we have to pay you so that you are willing to switch to the alternative investment level?’

5. We have to subtract 50 from the number of points generated by two ten-sided dice to get the interval [−49, 50].

6. However, we observe a significant difference between the main and the control treatment at the 5% level for the WTA. For final classification this result does not matter, since we require consistency regarding both measures.

7. In the one asset treatment 30 out of 72 participants were classified as consistently overconfident, whereas in the two asset treatment this is true for even 58 out of 77 participants. The latter proportion is significantly higher ( p<0.001).

8. In an empirical study based on data from a large discount brokerage firm, Barber and Odean (2001a) show that males trade 45% more than females. Theoretical models, for instance by Odean (1998), predict that overconfidence is positively related to trading volume. Therefore their results suggest the opposite to our findings; males are more prone to overconfidence than females.

9. This asymmetry may be explained by the historical fact that in the early stage of financial markets investors were rather wealthy and small in numbers.

10. Such regulations exist already for pension funds, which often promote risky investments, e.g. by granting tax allowances on new shares.