Periodic Reporting for period 1 - BROAD (Learning to bracket broadly)
Okres sprawozdawczy: 2017-04-01 do 2019-03-31
Reforms of social security systems in many EU countries have led to greater personal responsibility of consumers for their retirement savings and financial well-being. Consumers increasingly face numerous decisions that involve risk, such as how to allocate their retirement savings between risky and safe assets. However, there is ample scientific evidence that consumers often fail to make optimal choices. The overall objective of this project was therefore to better understand aspects of how consumers perceive and deal with these risky decisions.
Standard theory in economics assumes “broad bracketing” of risky choices: an individual considers the consequences of all her decisions under risk together. However, accumulating evidence suggests that people often bracket narrowly, i.e. they view each choice in isolation and not in the context of the other choices they face. This has consequences for the amount of risk an individual is willing to take and might lead to detrimental consequences (e.g. buying overpriced smartphone insurance). From a policy and an academic perspective, an important open question is whether people can “learn” to bracket broadly through a clever design (or “framing”) of the decision environment. The objective of Work Package 1 was to investigate whether a certain display of information induces learning to bracket broadly. In particular, do people continue to bracket broadly over time and in other related contexts? This is crucial for understanding the potential long-term effects of policy measures.
Work Package 2 also focused on risk perception in consumers’ investment decisions. Many otherwise puzzling investment phenomena (e.g. underdiversification) can be explained by assuming that consumers have a preference for skewness. Put simply, an investor with a preference for skewness likes assets which she expects to have a right-skewed, i.e. lottery like, payoff distribution, and (everything else equal) dislikes those which she expects to have a left-skewed distribution. However, testing this in the field is quite difficult because consumers’ expectations about the performance of an asset are usually unknown. The objective of Work Package 2 was to measure return expectations for several assets for a representative sample of the population in the Netherlands and to use these to test whether consumers’ investment choices reflect a preference for skewness.
Standard theory in economics assumes “broad bracketing” of risky choices: an individual considers the consequences of all her decisions under risk together. However, accumulating evidence suggests that people often bracket narrowly, i.e. they view each choice in isolation and not in the context of the other choices they face. This has consequences for the amount of risk an individual is willing to take and might lead to detrimental consequences (e.g. buying overpriced smartphone insurance). From a policy and an academic perspective, an important open question is whether people can “learn” to bracket broadly through a clever design (or “framing”) of the decision environment. The objective of Work Package 1 was to investigate whether a certain display of information induces learning to bracket broadly. In particular, do people continue to bracket broadly over time and in other related contexts? This is crucial for understanding the potential long-term effects of policy measures.
Work Package 2 also focused on risk perception in consumers’ investment decisions. Many otherwise puzzling investment phenomena (e.g. underdiversification) can be explained by assuming that consumers have a preference for skewness. Put simply, an investor with a preference for skewness likes assets which she expects to have a right-skewed, i.e. lottery like, payoff distribution, and (everything else equal) dislikes those which she expects to have a left-skewed distribution. However, testing this in the field is quite difficult because consumers’ expectations about the performance of an asset are usually unknown. The objective of Work Package 2 was to measure return expectations for several assets for a representative sample of the population in the Netherlands and to use these to test whether consumers’ investment choices reflect a preference for skewness.
In Work Package 1, I studied how individuals bracket over time in a series of investment decisions. In the time dimension, narrow bracketing would imply that an investor evaluates the performance of an investment frequently even though she may be investing for the long run. Earlier experiments (e.g. Gneezy and Potters 1997) showed that subjects, who receive information on returns aggregated over time (infrequent feedback), are more willing to invest in a risky asset with positive expected returns than subjects who receive information on the return in each period (frequent feedback). My own previous work provided tentative evidence that individuals are able to learn bracket broadly at least within the same decision task: Subjects going from more to less frequent return feedback increased their investment while the opposite was not the case.
To find out whether learning to bracket broadly applies to different tasks and whether it persists over time I analyze data from two laboratory experiments. In Experiment 1, subjects play the standard investment task in part 1 with either frequent or infrequent feedback. In part 2, they play a different investment task for which the type of bracketing should make a difference. If subjects learn to bracket broadly across tasks, those subjects who played Infrequent in Part 1 should display behaviour more in line with broad bracketing in Part 2. Overall, the evidence therefore suggests that learning to bracket broadly carries over to related tasks.
Experiment 2 addresses the question whether learning to bracket broadly is confined to a brief period after exposure to a framing that encourages broad bracketing, or persistent over time. I study changes in feedback frequency (treatments Infrequent-Frequent and Frequent-Infrequent) but with a time lag of one week between the two parts. I find evidence for a persistent effect of learning to bracket broadly in the sense that behavior in part 2 does not differ between the two treatments.
In Work Package 2, we use an innovative method with financial incentives for accuracy to elicit a distribution of return expectations for two assets at the level of the individual investor. We elicit these expectations for a large sample from the Dutch population. Participants can then decide how much of 100 Euro they want to invest in either of these assets or a safe asset. Their payment in one year depends on the performance of the thus created portfolio. After six months respondents were again asked about the expectations and allowed to change their portfolio allocations. Combining the data about stock market expectations and investment behavior we are able to test for a preference for skewness.
Our data analysis shows that stock market expectations vary substantially across investors, that they are not in line with historic values, and that variation in expectations matters for actual investment choices. Consistent with a preference for skewness, individuals who expect higher skewness for an asset invest more into that asset.
The results of both work packages were presented at various conferences and seminars. To facilitate exchange between academics and practitioners I organized a practitioner session on financial education at the 2018 Maastricht Behavioral Economic Policy Symposium and gave a course at the Finance Summer School of the University of Applied Sciences of the German Sparkassen Banking Group. The research component of this grant was accompanied by extensive training measures and mentoring regarding grant acquisition, media relations, and Ph.D. supervision.
To find out whether learning to bracket broadly applies to different tasks and whether it persists over time I analyze data from two laboratory experiments. In Experiment 1, subjects play the standard investment task in part 1 with either frequent or infrequent feedback. In part 2, they play a different investment task for which the type of bracketing should make a difference. If subjects learn to bracket broadly across tasks, those subjects who played Infrequent in Part 1 should display behaviour more in line with broad bracketing in Part 2. Overall, the evidence therefore suggests that learning to bracket broadly carries over to related tasks.
Experiment 2 addresses the question whether learning to bracket broadly is confined to a brief period after exposure to a framing that encourages broad bracketing, or persistent over time. I study changes in feedback frequency (treatments Infrequent-Frequent and Frequent-Infrequent) but with a time lag of one week between the two parts. I find evidence for a persistent effect of learning to bracket broadly in the sense that behavior in part 2 does not differ between the two treatments.
In Work Package 2, we use an innovative method with financial incentives for accuracy to elicit a distribution of return expectations for two assets at the level of the individual investor. We elicit these expectations for a large sample from the Dutch population. Participants can then decide how much of 100 Euro they want to invest in either of these assets or a safe asset. Their payment in one year depends on the performance of the thus created portfolio. After six months respondents were again asked about the expectations and allowed to change their portfolio allocations. Combining the data about stock market expectations and investment behavior we are able to test for a preference for skewness.
Our data analysis shows that stock market expectations vary substantially across investors, that they are not in line with historic values, and that variation in expectations matters for actual investment choices. Consistent with a preference for skewness, individuals who expect higher skewness for an asset invest more into that asset.
The results of both work packages were presented at various conferences and seminars. To facilitate exchange between academics and practitioners I organized a practitioner session on financial education at the 2018 Maastricht Behavioral Economic Policy Symposium and gave a course at the Finance Summer School of the University of Applied Sciences of the German Sparkassen Banking Group. The research component of this grant was accompanied by extensive training measures and mentoring regarding grant acquisition, media relations, and Ph.D. supervision.
Both work packages contribute to the bigger effort to understand “the way consumers understand and choose financial services” (European Consumer Agenda). Existing work on choice bracketing shows an effect of the framing of information for a specific task involving risk in a static setup, i.e. when group A is exposed to a different framing than group B. Work Package 1 moves to learning across different tasks in a dynamic context and shows that framing can be used strategically, in order to systematically change the way an individual brackets her choices and thus deals with risks. In particular, people can learn to take a broader perspective on the risks they face. The results suggest, for example, that banks or pension providers could influence how much risk consumers take in their investment decisions by changing the frequency of sending out performance information.
The results of Work Package 2 fill a gap between laboratory and aggregate level observational evidence for skewness preferences. Our findings call for more research on decision processes that go beyond the mean-variance model. Exercises similar to the one conducted in this work package readily extend to other contexts where individuals have to form expectations over potentially highly skewed outcomes. For example, measuring expected skewness could contribute to a better understanding of insurance choice or certain types of gambling such as sports betting.
The results of Work Package 2 fill a gap between laboratory and aggregate level observational evidence for skewness preferences. Our findings call for more research on decision processes that go beyond the mean-variance model. Exercises similar to the one conducted in this work package readily extend to other contexts where individuals have to form expectations over potentially highly skewed outcomes. For example, measuring expected skewness could contribute to a better understanding of insurance choice or certain types of gambling such as sports betting.