Diversification across geographic mortgage pools created rather than removed correlation.

Evidence

In principle, the different tranches of the CDO were isolated from each other. If they went bad, they went bad in reverse order, and it was thought that they couldn’t all go bad since different people held different parts of the bond and the underlying markets were uncorrelated. Unfortunately, it turned out that the underlying markets were quite strongly correlated. Adding Manhattan to Arizona and Baltimore in a single security made them correlated. The sheer volume of cash invested in real estate created one big market in US real estate across the world that became increasingly correlated to equities, particularly to the equities of the banks that were trading real estate. When the income streams of the riskiest (junior) part of the bond dried up as NINJA mortgagers walked away from their debts, investors in the more-secure tranches took fright and dumped their assets as part of the general search for liquidity. — Mark Blyth, Austerity, 97

Categorization

  • DEEPTHINKING: The insight that financial diversification itself manufactured correlation — an inversion of the core risk-spreading premise — is dense and paradigm-relevant beyond this crisis.
  • XREF: Echoes the Champagne Fairs’ logic that institutions restructure risk in unintended ways; relevant to Mattei’s systemic diagnosis of financial fragility. — connects to: Correlation assumptions in CDO design collapsed when housing markets proved linked. — Mark Blyth, Austerity: The History of a Dangerous…, 91; Good diversification requires assets uncorrelated with each other and earning above cash returns — Mark Blyth, Austerity: The History of a Dangerous…, 715; Imitation-based microstructure can defeat the risk reduction promised by diversification — Donald MacKenzie, An Engine, Not a Camera: How Financial…, 1089

Broader Connections