Fear of economic disasters, such as depressions, is more frequent than their actual occurrence. People tend to perceive a growing risk of disaster as they see economic conditions deteriorate. A new Federal Reserve paper illustrates that this pro-cyclicality of fears can trigger fluctuations in equity prices that go well beyond the actual changes in economic conditions, consistent with actual historical experience. Disaster fears also can make asset returns partly predictable.

Learning, Rare Disasters, and Asset Prices, Yang K. Lu and Michael Siemer
Finance and Economics Discussion Series Paper 2013-85
Divisions of Research & Statistics and Monetary Affairs, Federal Reserve Board, Washington, D.C.
http://www.federalreserve.gov/pubs/feds/2013/201385/201385pap.pdf 

The below are excerpts from the paper. Cursive lines and emphasis has been added.

“We examine how learning about disaster risk affects asset pricing in an endowment economy [a simplistic model where a representative agent pursues steady consumption and efficient allocation of wealth]… The model generates time variation in the risk premium through Bayesian [data-dependent] updating of agents’ beliefs regarding the likelihood and severity of disaster realization.”

“We define a rare disaster as an infrequent large shock that has long-lasting negative effects on aggregate consumption. Examples…include the Great Depression and World War II. Anecdotal evidence shows that much more often than an actual disaster occurring is that individuals fear that the economy might be headed toward a disaster. During the 2007-2009 financial crisis many commentators, including Nobel Prize-winning macroeconomists, highlighted the possibility that the U.S. economy could fall into another Great Depression. The markets responded with a large decline in stock prices and increased volatility. The nuclear accident in Japan in early 2011 is another example. This event was followed by a 22% decrease in the Japanese stock market within two days as the country wondered whether the nuclear meltdown could be contained.”

“Agents are not always suspicious that there may be a disaster unless they observe some evidence of it. Although in theory… [one] should always consider the possibility that there is a disaster with some probability, we find it more reasonable to assume that agents will…[neglect this probability in ] a booming economy.”

We assume that the disaster state is unobservable and that the parameters governing the severity of rare disasters are unknown to agents. Agents observe real-time consumption data and update their beliefs regarding the occurrence and the severity of a disaster over time. When agents price assets, they are rational in the sense that they account not only for the uncertainty associated with the occurrence and severity of a disaster but also for future updates of their beliefs concerning this uncertainty.”

When the severity of a disaster is uncertain, it becomes more difficult for agents to distinguish a temporary shock from a persistent shock, which causes their beliefs about being in a disaster to become more responsive to temporary shocks to consumption. Furthermore, the updating of agents’ beliefs regarding the severity of a disaster leads to additional time variations in the perceived disaster risk. Hence, our model endogenously generates volatile stock returns even in the absence of any disaster realization.”

“Our model yields an equity premium that matches the observed one using U.S. return data from 1948 to 2008, a period during which the U.S. economy did not experience any disaster. More importantly, the volatility of model-implied stock returns is 17.3%, accounting for 93.5% of the observed volatility. Our model also yields sizable variation in risk-free rates that matches the data. Beyond matching basic asset price moments, our model based predictive regression of future excess returns on the dividend price ratio generates results that are largely consistent with the data.”