Effective Investment Risk measurement is not about box-ticking or meeting Compliance requirements, rather this most critical function expresses a fundamental understanding of a manager’s approach to portfolio management and is to be found at the heart of a professional investment process.
The asset management industry is littered with risk rating agencies which provide a “number” that can pass the compliance or manager selection test for advisory processes. Investment firms stamp the logo and associated number on factsheets and marketing materials as if it were some sort of kitemark or product endorsement. The real question, is how reliable are these models and “numbers”, particularly when markets fall heavily? Sadly, the answer is that the application of a kitemark or “number” on a factsheet or marketing material rarely provides the real insight that is required to deliver a thorough understanding of the fund manager’s approach to investing, and how performance will be delivered when markets turn lower. After all, an investor’s hard-earned money is on the line here.
The different models we have encountered can be positioned in 2 major categories:
1) built upon a static asset allocation using historical returns with an economic scenario generator output.
2) Investment horizon measured by time-dependent asset allocation models using long-term averages for asset class performance, volatility, correlation, etc. An economic scenario generator offers a view of likely future trends.
Regardless of the type of models used, volatility or standard deviation remains the common denominator and singularly the most important determinant of the “number” awarded to the fund or portfolio from a risk point of view. Of course, the big advantage of relying on the standard deviation is its simplicity. Whilst we understand that this approach dovetails into the current client advice framework, it is not necessarily the best measure of risk.
There are solid reasons for this:
Frequency of historical data used in the analysis. Quarter end readings over the past 5 years will produce a much lower volatility number than monthly readings over the past 1 year in an increasingly volatile investment environment. This methodology masks intra quarter drawdowns and volatility events.
Historic volatility at an aggregate level may give a false sense of security, like an airbag fitted to a car but fails to deploy in an accident. Ask yourself whether historic volatility numbers were relevant in the recent Covid market drawdown? Of course not. Cross-asset correlations tend to increase in a stressed market and individual asset behaviour changes.
The analysis assumes the asset class proxies used by the analyst in risk modelling closely align to the fund and its assets held in live portfolios. A UK equity asset class may come in many different shapes and perform very differently in real life to the analysis undertaken!
Critically the analysis used assumes no change in economic regimes. This is super-important. The deflationary environment of the past few decades provoked quite a different asset class returns to those of the 1970’s for example.
Finally, (for this short paper) a single measure of risk does not help professional advisers or fund selectors understand the decomposition or the effects of different types of risk in a multi-asset portfolio.
The industry prefers using standard deviation at a research level, at optimisation level, portfolio composition and then at a marketing level – we understand this “one-stop shop” from a practical perspective, but the failings are significant and perhaps dangerous? An institutional investor would be unlikely to accept such a framework – why therefore is it considered suitable for smaller retail investors?
A difference in approach and re-framing the question may help.
Let’s start with Investment risk and volatility. Volatility is often an aggregate (fund level), after the fact measure. Knowing about the portfolios' previous levels of volatility is interesting but offers little assistance for the future. By approaching the portfolio construction as a “portfolio of risks” rather than a portfolio of different asset class returns helps pigeon-hole the headline volatility number into its critical elements, Liquidity risk, Market risk, Credit risk, Interest Rate Risk, and Concentration risks. Each of these elements of overall risk must be measured independently before they form part of the portfolio. This approach helps the portfolio manager, the adviser, or the fund selector to understand the type and suitability of risk in each economic regime.
Next, we will address the widespread use of historic data. To address this problem, institutional and more sophisticated investors usually use different variations of the Value at Risk (VaR) methodology. Unfortunately, as with any model, one cannot blindly use VaR without understanding its shortfalls. VaR, if measured using historic data only, does not add value. At Alpha Beta Partners we recognise VaR is not a new methodology but when correctly deployed using powerful systems and judgments and always looking one year ahead the outputs offer significant value. Bringing this foresight to the risk equation sets us apart.
So how does one crack this risk puzzle?
A Value at Risk approach or an Expected Tail Loss (a variation of VaR), using a mix of historical and simulated returns with a future view, can help produce a robust and decidedly more useful risk measure.
Risk decomposition and portfolio construction based upon these risks may well help limit drawdowns. Regime change or assumptions need to be clearly thought out and should cover a period that one can forecast with a high level of probability. Then at the aggregate level, without assuming normal probability, a simulation exercise results in a VaR and an Expected Tail Loss measure which is more robust.
The above exercise produces some interesting results. A standard deviation that one can use to measure against the traditional risk corridors deployed in the financial planning process, an Expected Tail Loss, and a simulated VaR that can be used as a yardstick to compare actual returns. Where VaR measures “the best of the worst” case scenario, Expected Tail Loss measures “in case things do go wrong, what is the expected loss?” This is much more useful to advisers, fund selectors, and investors than a view of simple volatility which happened in the past.
Following analysis using the current model employed the observation reports findings such as: “This fund has a historic 6.5% volatility and is categorised as a Cautious Balanced Fund”
Reframing the analysis and the output we believe a simple statement such as: “This fund has annualised VaR of 4.3%. This means there is a 95% chance the fund will not lose more than 4.3% in the case of a drawdown. The funds' risk decomposition means the fund has a low market risk, credit, and interest rate risk. In case of a liquidity event, the fund’s concentration and liquidity risks are low”.
To complete the loop, of course, these risk measures should be consistently used to track a portfolio’s performance and its risk-adjusted performance.
Taking a step further, the Expected Tail Loss approach can be used in portfolio optimisation - but that’s a topic for another day…
If you have any feedback or questions, please get in touch.