A Portfolio Manager Questions a Risk Engine I: Maximum Portfolio Loss and Risk Contribution

Yannis Sardis, 30 March 2021

The evaluation of a portfolio’s overall risk exposure to normal or Black Swan market conditions cannot be adequately covered by a sole risk number and its variations. Instead, one needs to deploy a selection of metrics and calculation methodologies, to ensure that the risk one assumes to produce an expected performance, complies with the given investment policy criteria.

The frequent absence of a holistic approach to risk assessment is amplified by the growing number of diversified investment vehicles, fund types and less liquid alternative asset pools such as derivatives, private equity, private debt, real estate, and tailored structured products. To address the benefits of a multi-faceted portfolio risk analysis, we launch a series of notes, where an intelligent Risk assessment Engine (RE) answers the questions of a representative Portfolio Manager (PM), each time focusing on a specific piece of the risk puzzle.

PM: I manage a basket of multi-asset portfolios. How can I quantify the potential risks that my portfolios are exposed to?

RE: A first line of assessment can utilize the notion of risk built around the prediction of the maximum potential portfolio loss over a selected time horizon, under a certain confidence level, known as Value at Risk. VaR is a regulatory approved ex-ante risk metric, an estimate of the maximum amount that a portfolio could lose overnight, over a month or a year, depending on the probability confidence level and the time horizon that an investor chooses as the most representative for his portfolio objectives.

PM: Are there various VaR methodologies? What is the best VaR model that I can use for my portfolio risk estimation?

RE: The most frequently used models for a VaR estimation are the Historical Simulation, the Monte-Carlo Simulation and the Parametric (Variance-Covariance) Method. There is not an absolute answer to which is the ‘best’ estimation model, as each method has its pros and cons. If one wants to easily communicate the VaR estimation to a not so quantitatively prone audience, the Historical Simulation may be the preferred route. For a most advanced methodology which does not explicitly rely on how the markets have behaved historically, the Monte-Carlo simulation could be the method of choice. The Parametric method is easy to implement and requires no significant amount of data, but it may not be as widely utilized due to its assumption that market returns follow a normal probability distribution.

PM: How do I identify large risk concentrations to specific portfolio segments, so that I can actively and timely manage such a diversifiable risk?

RE: Like the weight of a segment which is the percentage of the portfolio allocation to that segment, the Component VaR represents the amount of the portfolio risk exposure due to the asset allocation into that segment. For example, let us assume that your balanced multi-asset-class portfolio is allocated to several industrial sectors, as follows: Financials 30%, Energy 30%, Information Technology 40%, whilst the Component VaR that I estimate for each sector is: Financials CVaR 60%, Energy CVaR 25%, Information Technology CVaR 15%. This example displays that although the Financials and Energy sectors have the same portfolio weight, Financials amount for 60% of the total portfolio risk exposure, whereas Energy only for 25%, indicating a possible rebalance to mitigate your concentrated risk. Furthermore, I can produce for you a risk decomposition for a list of categorizations such as asset class, sector, risk country, reference currency, issuer credit rating and underlying security holdings, thus help you identify any profound asymmetries between the performance contribution of each categorized segment and their associated risk exposures.

PM: What about unexpected extreme events and the ‘tail’ risks that they frequently introduce? What if the markets move beyond my 99% confidence level?

RE: A good estimation method for the tail risks of a portfolio is the Conditional VaR, also known as Expected Shortfall. Conditional VaR tells you what the portfolio’s potential loss could be if the realized market returns occupied the tails of the return distribution curve (extreme events). It thus indicates what can happen to your portfolio if the markets move so excessively and unexpectedly that the 1% probability case (i.e. the ‘residual’ in the 99% confidence level) of my VaR estimation does indeed occur.

PM: In our next Q&A session, we will be addressing the notions of diversification and cross-correlations in multi-asset portfolios and groups of portfolios.

FINVENT Software Solutions is a provider of financial software applications and custom engineering services. The award-winning KlarityRisk platform specializes in investment risk analytics and fixed income performance attribution reporting and it is offered to Private Wealth institutions, Asset and Hedge Fund Managers and Family Offices.

Would you like to know more about our Portfolio and Risk Management solutions? Please contact us at info@finvent.com and we will get in touch.