LO 5.1: Explain the following lessons on VaR implementation: tim e horizon over

LO 5.1: Explain the following lessons on VaR implementation: tim e horizon over which VaR is estimated, the recognition o f time varying volatility in VaR risk factors, and VaR backtesting.
There is no consensus regarding the proper time horizon for risk measurement. The appropriate time horizon depends on the risk measurement purpose (e.g., setting capital limits) as well as portfolio liquidity. Thus, there is not a universally accepted approach for aggregating various VaR measures based on different time horizons.
Time-varying volatility results from volatility fluctuations over time. The effect of time- varying volatility on the accuracy of VaR measures decreases as time horizon increases. However, volatility generated by stochastic (i.e., random) jumps will reduce the accuracy of long-term VaR measures unless there is an adjustment made for stochastic jumps. It is important to recognize time-varying volatility in VaR measures since ignoring it will likely lead to an underestimation of risk. In addition to volatility fluctuations, risk managers should also account for time-varying correlations when making VaR calculations.
To simplify VaR estimation, the financial industry has a tendency to use short time horizons. This approach is computationally attractive for larger portfolios. However, a 10- day VaR time horizon, as suggested by the Basel Committee on Banking Supervision, is not always optimal. It is more preferred to instead allow the risk horizon to vary based on specific investment characteristics. When computing VaR over longer time horizons, a risk manager needs to account for the variation in a portfolios composition over time. Thus, a longer than 10-day time horizon may be necessary for economic capital purposes.
Page 56
2018 Kaplan, Inc.
Topic 5 Cross Reference to GARP Assigned Reading – Basel Committee on Banking Supervision
Historically, VaR backtesting has been used to validate VaR models. However, backtesting is not effective when the number of VaR exceptions is small. In addition, backtesting is less effective over longer time horizons due to portfolio instability. VaR models tend to be more realistic if time-varying volatility is incorporated; however, this approach tends to generate a procyclical VaR measure and produces unstable risk models due to estimation issues.
In t e g r a t i n g L i q u i d i t y R i s k i n t o V a R M o d e l s