Using an Eye Tracker to Examine Behavioral Biases in Investment Tasks: An Experimental Study

Abstract

Contrary to the premise of rational models, which suggests that investors’ aggregate portfolios are the appropriate informational asset for evaluating a file performance, we find, using an eye tracker, that investors spend more time looking at performances of an individual asset than at the performances of the overall aggregated portfolio and at the net value change more than the assets’ final value. We also find that investors look at the monetary value change longer than at change in percentages. Specifically, participants look longer at the value change of gaining assets than at the value change of losing assets. We propose the possibility that investors are not only engaged in judgment when evaluating their portfolio (leading to loss aversion and mental accounting) but may also be predisposed to looking for reassuring elements within it. Thus, it may be that humans use mental accounting by nature and not necessarily by judgment.

Keywords:

• Mental accounting
• Loss aversion
• Judgment
• Eye tracker

Notes

1. There are other “anomalies” in behavioral finance which can be explained by using the mental accounting principles (see DeBondt and Thaler [1995], Shiller [1998], Thaler [1999]).

2. The experimental situation follows real life situations, such as where investment fund managers decide on the investment amounts of each asset individually. Specifically, in this study we wanted to test and learn what would be the most important piece of information (e.g., positive or negative, change in value or in percentage) for investors, having to base one investment decision on a previous one.

3. We did not analyze the influence of the order of presenting the gain or the loss, since in this paper we tried to focus on the importance of each information variable on the investors’ decision making. (Any analysis of the order effect of the treatment or the losing/gaining asset would have moved this paper from its main focus.)

4. The recording of eye movement is quite complex. To prevent technical noise in the data we asked each participant to be involved in three trading days for each condition. The returns in each trading day weredifferent.

5. Averaging across repetitions has to do with the notion of measurement in the mathematical theory of measure, since there is always a random error component which can be of different size and is unknown. This makes relying on a single observation (measure) problematic from a methodological point of view.

6. For all the cases, the significance is lower than 0.05 except, in the losing asset, for low gain-high loss where: Z = 1.342, sig = 0.09 and for high gain-high loss: Z = 1.278, sig = 0.1.

7. The only case that is not significant is low-gain-high loss (Z = 0.635, sig’ = 0.262). For high gain-high loss the significance is lower than 10% but higher than 5% (Z = 1.344, sig’ = 0.09).

8. In the third phase, subjects invested more based on their “earnings” in the second phase, leading to higher gains. This indicates that subjects were tuned, involved and aware of the game's rules and constraints. If not involved they would have not had any reason to invest as they did, trying to boost performance.

9. We took the change in value since it is the most important variable, as shown in Tables 5 and 6.