آیا تئوری آینده نگری می تواند برای پیش بینی تمایل سرمایه گذار به پرداخت مورد استفاده قرار گیرد؟
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|11163||2013||14 صفحه PDF||سفارش دهید||12199 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Banking & Finance, Volume 37, Issue 6, June 2013, Pages 1960–1973
Cumulative prospect theory (CPT) is widely considered to be the most successful descriptive theory for decision making under risk and uncertainty. Sophisticated methods have been developed to reliably elicit CPT parameters on an individual basis. The aim of this paper is to analyze whether such methods are suited to be applied in real world situations, particularly in the context of investment counseling for retail investors. Specifically, we examine whether CPT parameters elicited via standardized computer tools are successful in predicting an individual’s preference for different structured financial products. Surprisingly, we find only low predictive power of the elicited CPT parameters on the WTP. Using a second set of experiments, we examine possible explanations for the low prediction quality. Overall, we have to conclude that it is too much of a leap to draw conclusions about the attractiveness of complex financial products from CPT parameters elicited via simple lotteries.
Consider an investment advisor who sells financial products to retail investors. It is in his interest to recommend products that fit the risk preferences of his clients well. First, doing so might establish a comparative advantage. Second, the interest might be driven by legal requirements such as those defined in the MiFID, the Markets in Financial Instruments Directive (The European Parliament and the European Council, 2004). While the MiFID asks the banks to collect “information on the length of time for which the client wishes to hold the investment, his preferences regarding risk taking, his risk profile, and the purposes of the investment” (Article 36(4), The European Parliament and the European Council, 2006), the directive is silent about how such an ambitious goal can be achieved. The common practice among financial advisors is to use simple questionnaires to gain insight into their customer’s preferences, with risk aversion and investment experience self-reported on a five- or seven-point scale (Barclays Wealth, 2008). From an academic perspective, such a procedure leaves much to be desired. Extensive research shows that risk behavior follows complex patterns that cannot be described by a simple five-point scale. Furthermore, empirical evidence suggests that non-expected utility theories, such as cumulative prospect theory (CPT) (Tversky and Kahneman, 1992), explain risk behavior better than the traditional expected utility approach (see Starmer (2000) for a general survey, and Camerer (2000) for a survey of field evidence on the descriptive relevance of prospect theory). Accordingly, over the last few years, elaborate elicitation mechanisms for CPT have been developed (see, e.g., Wakker and Deneffe, 1996 and Bleichrodt and Pinto, 2000, and Abdellaoui et al., 2007). These methods allow for a non-parametric elicitation of the value function and the probability weighting function. The aim of this paper is to analyze whether such methods are suited to be applied in real world situations, particularly in the context of investment counseling. Through experimental research we examine whether CPT parameters derived via such elicitation procedures are successful in predicting an individual’s willingness to pay (WTP) for different investment products. Because of the difference in the domains used for eliciting the parameters and deriving the predictions, our experiment can be considered an out-of-task evaluation of CPT. Out-of-task tests are important because they shed light on the applicability of decision theories in practice. The saying “time is money” is probably nowhere more relevant than in a banking environment. Therefore, one fundamental requirement for an implementation of such elicitation procedures in everyday bank business is the economic utilization of resources. As advisory time is a scarce resource, we analyze whether the whole procedure can be implemented using standardized computer tools, especially without time-consuming investor–advisor interactions. A transfer to real world banking on a large scale would only be conceivable from the viewpoint of a financial intermediary if such a computerized procedure would produce promising results. To answer our research question, we bring the same participants to the lab twice. During the first visit, we elicit CPT preference parameters on an individual basis by applying a modified version of the elicitation procedure of Abdellaoui et al. (2007).1 During the second visit, participants state their WTP for various investment products with different risk profiles. Our analysis compares theoretical predictions based on the CPT parameters from the first part with the actually stated WTPs from the second part. We use the method of ABP07, because it is not very susceptible to the influence of decision errors (Blavatskyy, 2006). As a second nice property, the method provides a completely parameter-free elicitation that allows for both parametric and non-parametric predictions. To cover the full range of complex risk profiles, we use structured financial products as investment opportunities. Structured financial products allow issuers to form almost any type of tailor-made payoff profile to serve clients with specific preference structures. This is achieved by combining an underlying, typically a stock or a stock index, with one or more derivatives on that underlying. CPT, designed to capture complex patterns of risk behavior, should be especially suited to explain differences in the WTPs of these financial products. Surprisingly, we find that CPT has virtually no predictive power if we run the procedures in a fully computerized environment. Based on these findings we explore alternative means of conducting the elicitation in bank practice (e.g. through personal interviews), which we test in a second set of experiments. Our main findings are: (1) more personal forms of interactions (i.e. personal interviews) enhance the internal consistency but do not improve the prediction quality, (2) competence effects seem to play a role but only a minor one, (3) an explanation based on the propagation of decision errors only can be ruled out, (4) the prediction quality increases substantially when we examine binary choices for simple prospects instead of WTPs for elaborate investment products. Overall we have to conclude that trying to predict an individual’s WTP for complex financial instruments based on individually elicited CPT parameters seems too much of a leap. The remainder of the paper is structured as follows. In Section 2, we introduce the procedure to elicit individual CPT parameters. In Section 3, we describe the structured products that we use to determine WTPs. Section 4 presents the design and the main results from our first experiment, Section 5 documents the results from the second set of experiments, and Section 6 contains the discussion of potential explanations for our findings. Section 7 concludes.
نتیجه گیری انگلیسی
In this paper, we use a sophisticated method to estimate CPT preference parameters on an individual level. Based on these estimates, we predict the individuals’ preferences in an investment decision context. The predictions and the stated preferences show almost no correlation. Although the internal consistency can be enhanced by replacing the fully computerized procedure by more personal forms of interaction, the prediction quality remains low. We analyze a number of potential causes for the low prediction quality, but none is able to fully explain our findings. A substantial increase in prediction power is only obtained if we replace the task of stating WTPs for complex investment products by a task that is based on binary choices between simple prospects, similar to the decisions in the preference elicitation process. We conclude that the complexity of the evaluation task is a major obstacle for the attempt to predict preferences for specific investment products from carefully elicited CPT parameters. There are different venues for follow-up research, most importantly an extended analysis of further potential drivers of our key findings. The low predictive power might be caused, for instance, by wrong assumptions regarding the reference points chosen by our subjects in their CPT evaluations. The formation of the reference point is crucial for every CPT valuation and can be influenced by the framing of the decision situation. We have carefully designed our experiment to minimize ambiguity with respect to reference point choice: Our product descriptions strive for a neutral presentation of the outcome distribution in reference to the product price; the instructions and the additional information provided are also carefully chosen in an attempt to prevent subjects from perceiving some required rate of return as their reference. Therefore, we do not suspect that heterogeneous reference points are a key driver of our results. Nevertheless, an extended analysis of such issues is definitely a worthwhile and important endeavor. For the general aim of improving the investment counseling process, one should also look at alternative preference theories that might be better compatible with real world investment decisions. Although CPT combines many attractive features of decision making under risk and uncertainty, there is a growing literature on violations of CPT in individual decision making (see, e.g., Birnbaum, 2006 and Wu and Markle, 2008). Hence, alternatives to CPT might be more successful in predicting the WTP for investment products. Similarly, alternative elicitation procedures might yield better results. Our experimental design assumed preference parameters to be stable over time. Harrison et al. (2005) present evidence that the simple Holt and Laury (2002) measure of risk aversion is indeed stable over a time horizon of approximately 6 months. Whether this also holds for the much more subtle CPT parameters is an open question. Zeisberger et al. (2012), for example, show that there is substantial variation in the parameters over time for about one third of the subjects in their sample. One way to address this issue would be to conduct the full elicitation procedure twice to measure the changes over time. However, even if our results could be explained to a large extent by unstable preferences, this would only highlight the difficulty of applying the prediction based on CPT in a real-world advisory process. Further research should also not exclude the possibility that it is not CPT and the specific parameter elicitation methods that produce the problems, but the flawed assumption that individuals have clearly shaped preferences and WTPs with respect to complex financial products in the first place. Decision analysis starts from the assumption that most people are unable to state meaningful preferences in complex decision situations and that decomposition and simplification are needed to guide the process. This provides many further perspectives and interesting approaches to the research question. As an example, Rieger (2012) shows the influence of probability misestimation on the attractiveness of structured financial products with complex payoff profiles and discusses potential ways to debias investors. We hope that more work will be done on the out-of-task-performance of CPT and other decision theories to better understand what is behind the virtually non-existing relationship between CPT predictions and stated WTPs. At least, the few statistically significant results from our experiments provide some indication for a basic relationship between behavioral decision preferences and investment decisions. For the time being, we have to conclude that the implementation of an academically appealing elicitation procedure in a practical investment context is difficult. Hopefully, further research will provide guidance on how to overcome the obstacles.