Selecting a VaR Model and Benchmark for SEC Rule 18f-4
A primary requirement of the SEC 18f-4 Regulation is to perform a daily value at risk (VaR) for a portfolio with derivatives, compared against the designated reference portfolio for the fund. On the surface this may seem straightforward, but there are nuances that the fund must consider in its decision making process. This post discusses VaR Model selection and how to choose and analyze the appropriate benchmark.
Value at Risk Models
There are 3 approaches to VaR that can be considered when complying with SEC Rule 18f-4; these are parametric models, historical simulation, and Monte Carlo simulation.
The parametric method is the simplest of all three approaches, relying heavily on the the assumption that the returns for a security are coincident with a normal distribution and that past returns are indicative of future returns. These assumptions may hold for a subset of non-linear securities but do not capture the complexities for those with embedded optionality or return distributions with excess kurtosis or skew.
To perform historical simulation, a security’s actual time series of returns is used to bootstrap a distribution of projected returns. This method, unlike the parametric, can accommodate any distribution for linear and non-linear securities provided a sufficient time series is available.
The same benefits of the historical method apply to the Monte Carlo simulation but with the additional ability to have fine grain control over all inputs that drive returns. Though it is computationally intensive, this granularity makes it possible to simulate returns given a particular view on the market and stress test scenarios.
Each method has advantages and disadvantages relative to the fund’s strategy and investments. Considerations should be made for differences in techniques as the final rule will require that any VaR model used incorporates all material risks associated with the fund’s investments. This allows funds to specify applicable market risk factors without prescribing to a standard VaR methodology.
A non-exhaustive list of common risk factors that a fund must account for in its VaR model are: equity price risk, interest rate risk, credit spread risk, foreign currency risk and commodity price risk. The VaR methodology must ensure these factors are captured and applied appropriately to vanilla securities, options, and positions with embedded optionality.
How Benchmarks are Selected and Analyzed
A fund must establish its leverage risk outer limit based on a relative VaR test that compares the fund’s VaR to that of a designated reference portfolio (benchmark). The benchmark can be either an index that meets certain requirements or its own investments (excluding derivatives). This selection is at the discretion of the fund manager.
Important Consideration: Funds are not required to apply VaR models consistently when calculating the VaR of its portfolio and that of its designated reference portfolio (i.e. if Monte Carlo simulation is used for the portfolio, Historical can be used for the benchmark). However, the VaR calculations individually must be applied appropriately, in definition and requirements. This allows funds to utilize less-costly approaches and obtain VaR from a third-parties instead of analyzing both portfolios and benchmarks in-house.
Applying Key Requirements
A key requirement of VaR is the availability of relevant historical data. Both the quality and available history of observations have impact on calculated results. The rule states that a fund’s chosen VaR model must be based on at least three years of historical market data. Funds with newer or novel investment exposures may experience challenges in collecting appropriate data sets.
Once a fund has decided on its approach, it must determine compliance with the applicable VaR test at least once each business day. Daily testing is required to ensure consistent adherence to the rules and ensuring inputs are consistently sourced whether it’s at market open, close, or intraday. The requirement is designed to help a fund highlight the potential for changes in market risk factors associated with a fund’s investments to result in losses.
To be compliant, the VaR calculation is required to use a 99% confidence level and a time horizon of 20 trading days. In doing so, a fund can measure and seek to limit the severity of more-extreme but less frequent negative returns.
Back Testing
Periodically, the calculated VaR for the fund should be compared to actual losses incurred, to determine if the VaR model used is appropriate given realized performance. A proper system for calculating, monitoring and maintaining compliance should include real-time analysis of the models themselves.
Conclusion
A fund’s risk manager will need to determine the appropriate VaR model that aligns with the fund’s strategy and investments, and obtain the data to support the VaR test. With the implementation of daily VaR, the risk manager will need to evaluate the effectiveness of the VaR model, and reflect changes that need to be made to capture relevant market risk factors as part of their periodic review.
