All Together in the Quicksand?
The recent Wall Street meltdown has shown how contagion can spread quickly from one bank to another during a financial crisis. As a consequence, what really scares markets is not the stand-alone risk of insolvency, but rather the likelihood of a domino effect which would endanger a significant portion of the banking system.
Xin Huang (University ofOklahoma ), Hao Zhou (Federal Reserve Board), and Haibin Zhu (Bank for International Settlements) have looked at the risk of joint default in a recent research report published by CAREFIN, the Bocconi research center on finance. Their results show that the level of correlation between variations in net worth of big banks oscillates over time, and that past values cannot be used as accurate predictors of future risk. Their econometric model, which forecasts future correlations, obtains a better fit than simple extrapolation of past correlations, with respect to actual losses in big banks. The model also shows that correlation increases when stocks plummet, which means that systemic risk spreads more rapidly in a depressed market than otherwise.
Huang, Zhou and Zhu have also calculated a new indicator of risk in the banking sector, defined as the cost of insuring oneself against financial disasters (i.e. a situation in which more than 15% of a bank's liabilities with third parties turn out to be insolvent). When individual banks become riskier and/or the correlation with other banks augments, the cost of such insurance increases rapidly. The indicator can thus be interpreted as a measure of banking fragility. It can also be seen as the expected cost that tax-payers must shoulder when government decides to bail out banks, after their capital has been wiped out by strong losses.
Such cost oscillates wildly across time. In "normal" market conditions, such as between 2004 and 2006, it stays well below one dollar for every thousand dollars of banking liabilities, while it can go above ten dollars in times of financial turbulence. It reached $12 before the collapse of Bear Stearns, for an absolute cost of $110 billion. "A huge sum of money, although less than a sixth of the $700 billion obtained from Congress by Paulson to bailout US banks," comments Andrea Resti, director of CAREFIN.
In addition of checking for the past evolution of systemic risk in the banking industry, CAREFIN has also conducted a what-if analysis, in order to assess how such risk would respond to radically worsening macroeconomic conditions. Using sophisticated econometric techniques, Huang, Zhou and Zhu simulated how the risk of contagion would respond to a replay of past market shocks, such as 9/11, or the collapse of Long Term Capital Management hedge fund and the Russian financial crisis, both occurred in 1998. In such cases, the cost of insuring $1,000 of banking liabilities against financial disaster increases by 40-50 dollars. Most of this increase can be attributed to a deterioration in the ratings of individual banks, but the sudden jump in correlation among credit institutions proves being decisive, too.
Comparing their forecasts and simulations with actual market data, the authors of the CAREFIN report have verified the goodness-of-fit of their model. Actual risk is worse than estimated risk in only 3.5% of cases; this is what happens for instance at the start of the subprime mortgage crisis (August-September 2007) or on the eve of the Bear Stearns bailout. In these cases, the increase in the risk of bankruptcy and the correlation among banks was higher than predicted, based on the market quotes of the previous days. Summing up, this research study enables us to better predict contagion risk, but also shows that there's no econometric model that can adequately capture the irrational peaks and troughs of financial markets.