Measuring Systemic Liquidity Risk and the Cost of Liquidity Insurance. By Tiago Severo

Measuring Systemic Liquidity Risk and the Cost of Liquidity Insurance. By Tiago Severo
IMF Working Paper No. 12/194
Jul 31, 2012
http://www.imfbookstore.org/ProdDetails.asp?ID=WPIEA2012194

Summary: I construct a systemic liquidity risk index (SLRI) from data on violations of arbitrage relationships across several asset classes between 2004 and 2010. Then I test whether the equity returns of 53 global banks were exposed to this liquidity risk factor. Results show that the level of bank returns is not directly affected by the SLRI, but their volatility increases when liquidity conditions deteriorate. I do not find a strong association between bank size and exposure to the SLRI - measured as the sensitivity of volatility to the index. Surprisingly, exposure to systemic liquidity risk is positively associated with the Net Stable Funding Ratio (NSFR). The link between equity volatility and the SLRI allows me to calculate the cost that would be borne by public authorities for providing liquidity support to the financial sector. I use this information to estimate a liquidity insurance premium that could be paid by individual banks in order to cover for that social cost.

Excerpts:

Introduction

Liquidity risk has become a central topic for investors, regulators and academics in the aftermath of the global financial crisis. The sharp decline of real estate prices in the U.S.  and parts of Europe and the consequent collapse in the values of related securities created widespread concerns about the solvency of banks and other financial intermediaries. The resulting increase in counterparty risk induced investors to shy away from risky short-term funding markets [Gorton and Metrick (2010)] and to store funds in safe and liquid assets, especially U.S. government debt. The dry-up in funding markets hit levered financial intermediaries involved in maturity and liquidity transformation hard [Brunnermeier (2009)], propagating the initial shock through global markets.

Central bankers in major countries responded to the contraction in liquidity by pumping an unprecendented amount of funds into securities and interbank markets, and by creating and extending liquidity backstop lines to rescue troubled financial intermediaries.  Such measures have exposed public finances, and ultimately taxpayers, to the risk of substantial losses. Understanding the origins of systemic liquidity risk in financial markets is, therefore, an invaluable tool for policymakers to reduce the chance of facing these very same challenges again in the future. In particular, if public support in periods of widespread distress cannot be prevented—due to commitment problems—supervisors and regulators should ensure that financial intermediaries are properly monitored and charged to reflect the contingent benefits they enjoy.

The present paper brings three contributions to the topic of systemic liquidity risk:

1) It produces a systemic liquidity risk index (SLRI) calculated from violations of “arbitrage” relationships in various securities markets.

2) It estimates the exposure of 53 global banks to this aggregate risk factor.

3) It uses the information in 2) to devise an insurance system where banks pre-pay for the costs faced by public authorities for providing contingent liquidity support.

Results indicate that systemic illiquidity became a widespread problem in the aftermath of Lehman’s bankruptcy, and it only recovered after several months. Systemic liquidity risk spiked again during the Greek sovereign crisis in the second quarter of 2010, albeit at much more moderate levels. Yet, the renewed concerns regarding sovereign default in peripheral Europe observed in the last quarter of 2010 did not induce global liquidity shortfalls.

In terms of exposures of individual institutions, I find that, in general, systemic liquidity risk does not affect the level of bank stock returns on a systematic fashion.  However, liquidity risk is strongly correlated with the volatility of bank stocks: system wide illiquidity is associated with riskier banks. Estimates also show that U.S. and U.K. banks were relatively more exposed to liquidity conditions compared to Japanese institutions, with continental European banks lying in the middle of the distribution. More specifically, the results indicate that U.S. and U.K. banks’ stocks became much more volatile relative to their Asian peers when liquidity evaporated. This likely reflects the higher degree of maturity transformation and the reliance on very short-term funding by Anglo-Saxon banks. A natural question is whether bank specific characteristics beyond geographic location reflect the different degrees of exposure to liquidity risk.

I start the quest for those bank characteristics by looking at the importance of bank size for liquidity risk exposure. Market participants, policymakers and academics have highlighted the role of size and interconnectedness as a source of systemic risk. To verify this claim, I form quintile portfolios of banks based on market capitalization and test whether there are significant differences in the sensitivity of their return volatility to the SLRI. The estimates suggest that size has implications for liquidity risk, but the relationship is highly non-linear. The association between size and sensitivity to liquidity conditions is only relevant for the very large banks, and it becomes pronounced only when liquidity conditions deteriorate substantially.

Recently, the Basel Committee on Banking Supervision produced, for the first time, a framework (based on balance sheet information) to regulate banks’ liquidity risk. In particular, it proposed two liquidity ratios that shall be monitored by supervisors: the Liquidity Coverage Ratio (LCR), which indicates banks’ ability to withstand a short-term liquidity crisis, and the Net Stable Funding Ratio (NSFR), which measures the long-term, structural funding mismatches in a bank. Forming quintile portfolios based on banks' NSFR, I find that, if anything, the regulatory ratio is positively associated with the exposure to the SLRI. In other words, banks with a high NSFR (the ones deamed to be structurally more liquid) are in fact slightly more sensitive to liquidity conditions. This counterintuitive result needs to be qualified. As noted later, the SLRI captures short-term liquidity stresses, whereas the NSFR is designed as a medium to long-term indicator of liquidity conditions. Certainly, it would be more appropriate to test the performance of LCR instead. However, the data necessary for its computation are not readily available.

The link between bank stock volatility and the SLRI allows me to calculate the cost faced by public authorities for providing liquidity support for banks. Relying on the contingent claims approach (CCA), I use observable information on a bank’s equity and the book value of its liabilities to back out the unobserved level and volatility of its assets. I then estimate by how much the level and volatility of implied assets change as liquidity conditions deteriorate, and how such changes affect the price of a hypothetical put option on the bank’s assets. Because the price of this put indicates the cost of public support to banks, variations in the put due to fluctuations in the SLRI provide a benchmark for charging banks according to their exposure to systemic liquidity risk, a goal that has been advocated by many experts on bank regulation. 


http://www.imf.org/external/pubs/cat/longres.aspx?sk=26131.0