Category Archives: decision under uncertainty

9/1/20: Herding and Anchoring in Cryptocurrency Markets

Our new paper, with Daniel O'Loughlin, titled "Herding and Anchoring in Cryptocurrency Markets: Investor Reaction to Fear and Uncertainty" has been accepted to the Journal of Behavioral and Experimental Finance, forthcoming February 2020.

The working paper version is available here:

Cryptocurrencies have emerged as an innovative alternative investment asset class, traded in data-rich markets by globally distributed investors. Although significant attention has been devoted to their pricing properties, to-date, academic literature on behavioral drivers remains less developed. We explore the question of how price dynamics of cryptocurrencies are influenced by the interaction between behavioral factors behind investor decisions and publicly accessible data flows. We use sentiment analysis to model the effects of public sentiment toward investment markets in general, and cryptocurrencies in particular on crypto-assets’ valuations. Our results show that investor sentiment can predict the price direction of cryptocurrencies, indicating direct impact of herding and anchoring biases. We also discuss a new direction for analyzing behavioral drivers of the crypto assets based on the use of natural language AI to extract better quality data on investor sentiment.

19/4/15: New Evidence: Ambiguity Aversion is the Exception

A fascinating behavioural economics study on ambiguity aversion by Kocher, Martin G. and Lahno, Amrei Marie and Trautmann, Stefan, titled "Ambiguity Aversion is the Exception" (March 31, 2015, CESifo Working Paper Series No. 5261: provides empirical testing of ambiguity aversion hypothesis.

Note: my comments within quotes are in bracketed italics

When an agent makes a decision in the presence of uncertainty, "risky prospects with known probabilities are often distinguished from ambiguous prospects with unknown or uncertain probabilities… [in economics literature] it is typically assumed that people dislike ambiguity in addition to a potential dislike of risk, and that they adjust their behavior in favor of known-probability risks, even at significant costs."

In other words, there is a paradoxical pattern in behaviour commonly hypothesised: suppose an agent is facing a choice between a gamble with known probabilities (uncertain, but not ambiguous) that has low expected return and a gamble with unknown (ambiguous) probabilities that has high expected return. In basic terms, ambiguity aversion implies that an agent will tend to opt to select the first choice, even if this choice is sub-optimal, in standard risk aversion setting.

As authors note, "A large literature has studied the consequences of such ambiguity aversion for decision making in the presence of uncertainty. Building on decision theories that assume ambiguity aversion, this literature shows that ambiguity can account for empirically observed violations of expected utility based theories (“anomalies”)."

"These and many other theoretical contributions presume a universally negative attitude toward ambiguity. Such an assumption seems, at first sight, descriptively justified on the basis of a large experimental literature… However, …the predominance of ambiguity aversion in experimental findings might be due to a narrow focus on the domain of moderate likelihood gains… While fear of a bad unknown probability might prevail in this domain [of choices with low or marginal gains], people might be more optimistic in other domains [for example if faced with much greater payoffs or risks, or when choices between strategies are more complex], hoping for ambiguity to offer better odds than a known-risk alternative."

So the authors then set out to look at the evidence for ambiguity aversion "in different likelihood ranges and in the gain domain, the loss domain, and with mixed outcomes, i.e. where both gains and losses may be incurred. …Our between-subjects design with more than 500 experimental participants exposes participants to exactly one of the four domains, reducing any contrast effects that may affect the preferences in the laboratory context."

Core conclusion: "Ambiguity aversion is the exception, not the rule. We replicate the finding of ambiguity aversion for moderate likelihood gains in the classic However, once we move away from the gain domain or from the [binary] choice to more [complex set of choices], thus introducing lower likelihoods, we observe either ambiguity neutrality or even ambiguity seeking behavior. These results are robust to the elicitation procedure."

So is ambiguity hypothesis dead? Not really. "Our rejection of universal ambiguity aversion does not generally contradict ambiguity models, but it has important implications for the assumptions in applied models that use ambiguity attitudes to explain real-world phenomena. Theoretical analyses should not only consider the effects of ambiguity aversion, but also potential implications of ambiguity loving for economics and finance, particularly in contexts that involve rare events or perceived losses such as with insurance or investments. Policy implications should always be fine-tuned to the specific domain, because policy interventions based on wrong assumptions regarding the ambiguity attitudes of those targeted by the policy could be detrimental."