A recent (December 2011) paper published by CEPR offers a very interesting analysis of the macroeconomic risks propagation in the current crisis. The paper, titled Great Moderation or Great Mistake: Can rising confidence in low macro-risk explain the boom in asset prices? (CEPR DP 8700) by Tobias Broer and Afroditi Kero looks at the evidence on whether the period of Great Moderation in macroeconomic volatility during the period from the mid-1980s (the decline in macroeconomic volatility that is unprecedented in modern history) had an associated impact on the rise of asset prices that accompanied this period, setting the stage for the ongoing crash.
In recent literature, this rise in asset prices, and the crash that followed, have both been attributed to "overconfidence in a benign macroeconomic environment of low volatility" or to excessively optimistic expectations of investors that the lengthy period of macroeconomic stability and upward trending is the 'new normal'.
The study introduced learning about the persistence of volatility regimes in a standard asset pricing model of investor decision making. "It shows that the fall in US macroeconomic volatility since the mid-1980s only leads to a relatively small increase in asset prices when investors have full information about the highly persistent, but not permanent, nature of low volatility regimes." In other words, in the rational expectations setting with no errors in judgement and perfect foresight (investors are aware that volatility reductions are temporary), there is no bubble forming.
However, when investors "infer the persistence of low volatility from empirical evidence" (in other words when knowledge is imperfect and there is a probabilistic scenario under which the moderation can be permanent, then "Bayesian learning can deliver a strong rise in asset prices by up to 80%. Moreover, the end of the low volatility period leads to a strong and sudden crash in prices."
Specifically, calibrated model generates pre-collapse rise in asset prices of 77% and overvaluation of assets by 79% over the case of no learning. The subsequent collapse of asset prices is 84% in the case of imperfect information learning.
A pretty nice result!