3. The Concept of Correlation

Correlation describes the failure of multiple institutions resulting from the collapse of asset prices due to an exogenous event (e.g., the fall of housing prices in the period prior to the 2008–2009 financial crisis). Correlation can also refer to the herding instinct of asset managers that can result in market crashes and instability, or in irrational asset bubbles. In addition the academic literature discussing “indirect” connectedness is really discussing correlation. Although correlation played an important role in the recent crisis, contagion is what transformed $100–200 billion in losses on subprime mortgage products1 into the destruction of roughly $8 trillion of equity market capitalization between October 2007 and October 2009.2

Recently, the risk of the asset management industry’s herding behavior has come to the forefront of the discussion about correlation risk, as it may result in market crashes and instability, particularly during periods of distress.3 Herding behavior involves the tendency of asset managers to move out of a particular security or asset class in a correlated manner. The concern is that if most large asset managers sell at the same time, the market for that security or asset class may collapse, putting stress on all holders of such assets. As a consequence of this herding concern, regulators have considered the idea of designating large asset managers as systemically important financial institutions (SIFIs), though managers may have escaped the spectre of SIFI designation for the time being due to recognition that any risks are industry-wide and not firm specific.4

SIFI regulation would likely be unsuccessful in preventing adverse herding behavior. Herding by definition involves a number of firms across the industry acting in a coordinated fashion.5 SIFI designation and the consequent regulation of the SIFI is firm specific, so would be ineffective in combatting an industry-wide problem. The only potentially effective solution to correlated market declines that result from herding behavior is to impose a temporary form of circuit breaker that may help slow the price drops in an asset class or in particular securities, at least give time for a deep breath. During the market crash of October 1987, specialists halted trading in some of the most severely affected stocks.6 In addition the Federal Reserve issued a public statement indicating its intent to foster stability by providing liquidity to the market.7 The Federal Reserve followed through by both increasing the level of reserves in the system through open market operations and by reducing the target federal funds rate by 50 bps.8 When asset prices plunge, the Federal Reserve could also become the buyer or market-maker of last resort.

This book does not focus on correlation risk and herding. Obviously, if the correlated losses in housing had never occurred, contagion would most likely not have occurred either. But the prevention of correlated risks and herding behavior, while important, may be extremely difficult—this is the task for so-called macroprudential regulation. At the very least it involves policies like the detection and prevention of bubbles, which are beyond the scope of this book. For a more detailed discussion of the academic literature on correlation, see the appendix.

The three distinct C’s of systemic risk, connectedness, correlation, and contagion, are not mutually exclusive. They overlap to some extent. Thus, for example, a funding connectedness problem may set off a contagion, or fire sales of assets may intensify a contagion due to correlated holdings of assets subject to fire sales. Despite these overlaps, I believe the concepts are distinct enough to be very useful in analysis.