value-at-risk in bank regulartion
Regulators typically use Value at Risk (VaR) as part of their regulatory framework to assess and monitor the risk exposure of banks.
Minimum Risk-based Capital Requirements: Regulators set minimum capital requirements that banks must maintain to ensure they have sufficient buffers to absorb potential losses. The Basel framework specifies minimum capital requirement as percentages of risk-weighted assets and methods to calculate RWA for credit risk, market risk and operational risk. VaR models approach is an option to determine the size of counterparty credit risk exposure.
Market Risk Measurement: Regulators require banks to calculate and report VaR on a regular basis, usually daily or at least weekly, for their trading activities. Banks use various VaR models, such as historical simulation, Monte Carlo simulation, or parametric models, to estimate potential losses at a specified confidence level (e.g., 99% VaR).
Back-testing of VaR models: Since there is no one single standardized VaR model, banks can create their own variety of VaR models. Back-testing allows bank regulators to verify risk models used by subject banks and assess whether VaR forecasts are properly calibrated or accurate. The essence of back-testing is to identify banks using models that may potentially underestimate risk, thereby not only endangering the financial health of an individual bank but the industry in general. Depending on the number of exceptions (actual observations over and above the expected level of losses), regulators could levy penalties or increase level of capital requirements.
Stress Testing: In addition to regular VaR calculations, regulators often require banks to conduct stress tests. Stress tests involve assessing the potential impact of extreme market conditions or macroeconomic scenarios on a bank's risk exposure and capital adequacy. Stress tests complement VaR by evaluating the resilience of a bank's balance sheet under adverse conditions.
There are known limitations of VaR:
Assumptions: VaR assumes normal distribution of returns and constant correlations, which may not hold during periods of financial stress or back swan events. Thus, Var model can underestimate tail risks
Historical data approach: VaR uses historical data which may not anticipate future market conditions or emerging risks when market dynamics changes rapidly
False sense of security: VaR doesn’t report the maximum potential losses. 99% percent VAR means that in 1% of cases (that would be 2-3 trading days in a year with daily VAR) the loss is expected to be greater than the VAR amount. VaR does not say anything about the size of losses within this 1% of worst trading days