Social Sharing
Extended Viewer
On the White Board - July 2010
Jul 15, 2010
Systemic risk and liquidity shocks
We have worked recently to characterize the financial system’s exposure to one particular type of shock—a shock to funding liquidity.
Systemic risk, since the financial crisis of 2007-08, has become an active topic of discussion in the financial community, and in particular among regulators. Historically, bank capital standards arguably only addressed the micro-prudential problem of setting minimum capital levels in order to limit the likelihood of failure for individual institutions. They did not address the macro-prudential problem of the simultaneous failure of multiple institutions, an event which would require large government bailouts simply to keep the financial system functioning. It is this problem that lies at the heart of systemic risk analysis.
In the context of risk models, there have been a number of attempts to characterize systemic risk by considering the collection of banks in an economy as a portfolio, and the government’s exposure to bailing out the system as the tail risk of this portfolio. In concept, such an approach enables both a quantification of the level of systemic risk and an allocation of this risk to the individual banks. The allocation can then serve as the basis for a fee (or tax) on the individual banks to compensate for their contribution to the overall systemic risk. With this prior work, however, the attempts have been to quantify the portfolio risk statistically through the historical equity prices of the banks. This is clearly a lot to ask of the efficiency of the equity markets, and of our statistical techniques. More concerning, by embedding in a statistical structure all of the interconnections between banks and all of the exposures to common shocks, we lose the opportunity to characterize what the actual sources of systemic risk are.
Our approach is somewhat more fundamental. We begin by calibrating a structural (or Merton) model of credit risk to a set of large financial institutions, linking the market prices of the firm’s equity and equity options and its reported level of liabilities with its probability of default. We then define a systemic liquidity shock, and exploit the model to examine the forecasted changes in both equity values and default probability under this shock.
Our liquidity shock is inspired by the one-month liquidity stress proposed in December 2009 as part of the Basel Committee’s guidelines on liquidity risk. We utilize regulatory call reports (as provided through our CFRA division) to describe the makeup of each bank’s liabilities, and assume specific run-off rates for the different funding sources. For example, we assume that in the liquidity shock, banks lose 5% of their core deposit funding, 50% of their non-core deposit and repo funding, and face a drawdown of 5% of their total off-balance sheet commitments and letters of credit. The shock produces a funding shortfall, which we assume the banks fund by selling off assets on a pro-rata basis (so as to maintain the relative riskiness of the assets); we assume cash and equivalents can be sold at face value but that non-cash assets are sold with a liquidity haircut of 20%.
Overall, the funding shock stress test has a greater impact on banks who rely more on non-core deposits, repo and wholesale funding; who have less deposit funding relative to their lending activity; who have more off-balance sheet commitments to lend; and who have less of their assets in cash and equivalents. We display the liability profile for Citigroup, as well as our assumed stressed funding shortfall, in the figure below.
To characterize the risk of the system, we consider the sum across all of the financial institutions of the probability of default multiplied by the level of liabilities. In a sense, this can be thought of as the hypothetical cost of purchasing protection on all of the liabilities in the financial system. In the figure below, we examine the impact of our liquidity stress on this cost as of three dates—January 2008, November 2008 and December 2009. Not surprisingly, the greatest impact occurs when the stress is applied (to the already stressed market) of November 2009. More surprisingly, we see that the level of exposure is still (as of the end of 2009) quite high; looking more closely, however, we see that most of this exposure is coming from Citigroup, and that the other banks in the system are all significantly less exposed to the liquidity shock now versus at the height of the crisis.