An individual’s risk preferences are affected by the way investment choices are presented. The authors conclude that presentations of downside risk result in a lower rate of error than presentations of upside risk and that presentations in terms of probability of return in a given year result in a lower rate of error than presentations that describe the frequency of returns above or below some threshold.
The authors’ dataset is a sample of 1,200 Australian investors’ risk preferences divided between three portfolios. To analyze discrete choice experiments, the authors present hypothetical investment choices in order to describe the risk preferences and propensity for error of a population and the way they vary with sociodemographic variables and different risk presentations. They investigate the effects of four different risk levels and nine different risk presentation formats on risk preference. They define a theoretical framework based on prospect theory and a procedure for parameter estimation under that framework. The authors then outline a second layer of analysis based on regressing the maximum likelihood parameters of their behavioral model against sociodemographic variables. They present their findings, focusing on presentation effects and demographic effects, and compare them with those of previous researchers.
Within the limited cases studied, the authors find that presentations based on the percentage likelihood of returns below threshold result in more consistent risk preferences than those based on the frequency of returns below threshold. They also find improved consistency where downside variation is presented, which they ascribe to increased cognitive effort resulting from loss aversion. Their findings suggest that less financially literate individuals might benefit from a graphical presentation of risk.
How Is This Research Useful to Practitioners?
Awareness of the impact of presentation format on risk preference is vital to practitioners. The authors not only emphasize this point in terms of differences in outcome but also describe which risk presentations lead to more consistent decision making. In doing so, they present an approach for evaluating the quality of a risk presentation format. Furthermore, they demonstrate that different sociodemographic groups are affected in different ways by presentation effects.
The methodology of fitting behavioral models offers practitioners some insight into the way the behavioral finance community approaches risk preference. That some of the results contradict those in the literature the authors cite serves to highlight that such findings are experimental, model dependent, and not necessarily conclusive.
The authors’ results should be of great interest to professionals involved in the sale of funds or provision of investment platforms to individual investors. These groups should be concerned with the impact of their presentation format on investment decisions. The results should also be of interest to investors more broadly so that they can better understand potential biases present in their own decision making as a result of presentation effects. Finally, it is of significant interest to regulators, who can, deliberately or not, have an impact on individuals’ decisions through the presentation restrictions they choose to impose.
How Did the Authors Conduct This Research?
The authors take data from discrete choice experiments in which each participant is asked to rank portfolios containing a risk-free asset, a risky asset, or a combination of the two. The participant is asked to rank the portfolios with risky assets; the choice sets vary according to the level of risk and the presentation format. The formats include the chance of losses or gains, the number of years out of 20 when the portfolio would out- or underperform some threshold, and a graphical display of the 5%–9% confidence interval of returns. To model preferences, the authors use a prospect theory utility model with the added probability of an individual making some error in the calculation of the certainty equivalent value of a portfolio. This model allows for individuals to express unique risk preferences and for presentations of risk to result in different risk preferences across a population. The parameters include risk aversion in gains, risk aversion in losses, loss aversion, and standard deviation of the error term.
They examine the median best-fit parameters to demonstrate that the results are consistent with prospect theory and examine the differences in parameters between groups. The authors regress all of the model parameters against sociodemographic variables and examine the implications, comparing them with the results of previous researchers. Their results are consistent with previous findings that risk aversion is positively related to the amount of retirement savings and with proximity to retirement, but they contradict previous researchers who found risk aversion to be negatively related to employment and cognitive ability. They also present the model-specific component of the standard deviation of the error term to demonstrate which risk presentations give rise to the most consistent decision making under their model (that is, those with the lowest standard deviation of certainty-equivalent value error).
The authors themselves highlight the main drawback to their study: Experimental results are known to change when participants are incentivized with small cash amounts and would likely change much more if the amounts were to correspond to their net worth, as in the case of real investment decisions. The results may, therefore, not accurately reflect those made by individuals for their own investments. Other confounding effects could include priming (because each group answered questions on its own series of presentations) and availability bias, particularly given the timing of the study in the aftermath of the financial crisis. Further work of this type surveying different populations and exploring other presentation formats would be extremely valuable.