Many investors, trading floor managers, fund of fund allocators and risk managers have at one time experienced unexpectedly large losses that were not predicted by their risk management systems. Some of these losses may have occurred in portfolios, which according to the risk reports only contained very little risk, perhaps they were predicted to be market neutral, well diversified and, until the large loss happened, had only generated very moderate return swings. Then, seemingly out of the blue, this safe portfolio generated large losses and for reasons that were not detected, nor predicted.
Risk systems are today one of the key technologies within financial firms. Risk management guides everything from market exposure to regulatory reporting. As a financial market participant, you cannot avoid being exposed to various risk management techniques and protocols. Therefore, it is important to know what a risk management system can capture, measure and predict and perhaps more importantly, what it does not capture.
The kernel of this article is to show some examples of risk exposures that can fly below the radar screen of risk systems.
Ever more complex and sophisticated risk models and systems, may give an illusion of control and safety. As we will exemplify risk systems can be blind to, and therefore portfolios vulnerable to, risks that systems are not measuring and do not have the capacity to imagine.
What gets measured can be managed and while we may feel safer by improving our risk systems and risk management techniques, they did not stop bank failures in 2008, nor are they likely to prevent or predict the next large dislocation or crash. It is therefore important to broaden our scope of risk awareness to what lies outside the risk management systems reach.
It is to be noted that some of the risks or biases we describe may be intended components of an investment strategy. If that is the case, we should expect this to have been clearly communicated to investors pre-investment, so they are aware and can fully understand what to expect in terms of return behaviour in different market environments. Such a discussion about pre-investment expectations and understanding between managers and investors lies outside the scope of this article. In a next article, we intend to revert to the topic of aligning investor expectations, through enhanced understanding, with actual performance.
Even if it is rare, we have come across cases where a risk or bias was not known or understood by the portfolio manager himself. In such cases, both manager and investor are equally exposed to possible negative surprises.
The point we want to make, by illustrating with some examples from our own experience, is that an investor should never rely too much on his risk systems. They can never cover more than a part of all risks. The investor needs to look beyond his risk reports and try to find what other risks and biases he may be exposed to in a portfolio or strategy. This is a never-ending task and requires experience, imagination and creativity along with the hard work of analysing the traditional risk reports.
The below list of hidden risks and biases are by no means exhaustive, but all examples are real and from our own experience. We focus mainly on market risk, rather than operational risk issues.
If these examples succeed in sowing a seed of sound doubt and this triggers a more holistic risk approach, then this article has achieved its' aim.
"Anything that can hurt my portfolios and I am in particular looking out for the things that have not yet happened or that I have not yet imagined"
Risk manager of large Global Macro Hedge Fund when asked about his risk focus.
Exploding/Imploding Betas
For any particular stock, its' Beta is more or less stable over time. The stability of the Beta is also more or less sensitive to various changes in markets. There are stocks where, in a certain market environment, its Beta increases a lot while in other stocks, their Beta "implodes".
A portfolio of stocks may contain long positions and short bets with opposite Beta dynamics. The longs and shorts can react very differently to a certain type of market downturn and may then cause larger losses than what the risk system's portfolio Beta estimates suggests. We have in particular observed this phenomenon during periods when investor behaviour could be characterised as risk-on-risk-off. To have different Beta biases on the short and long side of the portfolio can indeed be a conscious part of an investment strategy. It does however, result in a portfolio with a different and in some respects higher risk than the risk report shows.
For diversified, unleveraged traditional portfolios, this is likely to be a minor problem. For active more concentrated and leveraged strategies, this difference in Beta behaviour can result in outsized negative results. As an extreme example, consider the stock of Volkswagen, subject to a takeover bid by Poche combined with a so called short-squeeze.
Volkswagen’s Beta both imploded and then exploded during the financial crisis of 2008. For a portfolio manager using the stock as a hedge against other long-positions, the dynamic of the stock provided a very poor insurance. Needless to say, the possibility of a short-squeeze was not captured by most risk-systems.
For a concentrated, leveraged hedge fund portfolio with different Beta sensitivities on the short and long side, this hidden bias can translate into a downside potential that is multiples greater than the upside, even when, according to the risk system, the overall portfolio is Beta neutral.
Tipping point stocks
A stock will often trade within a certain band around an industry sub-index. But then it dramatically changes its price development and stop following its' index peers. This can for example happen when a company’s credit rating drops or when its' bond spread widens out significantly. Credit always matters in a stressed situation. A recent example of this is Chesapeake Energy. On February 8, 2016, it lost a third of its value one single trading session, as the market focused on the risk of default for the company.
A stock can, on a piece of bad credit news, start trading more like a junk bond. The deterioration of credit standing is just one example of an event that has not happened and is therefore not represented by historical price movements of the stock. Such new events are usually outside the domain of a risk system. This indirect credit risk may be more or less relevant for a particular portfolio depending on how many of its' positions are near a tipping point but also on the credit cycle when more companies may be close to junk bond credit spread levels.
If a portfolio has a significant component of stocks whose debt is close to non-investment grade territory, this bias will amplify the effects of a downturn in credit markets and thus give such a portfolio a credit exposure that will not normally be captured, as stocks are not generally regarded as credit instruments.
Hidden credit carry
Complex and diversified portfolios across global Fixed Income, Currencies, Commodities, Equity indices etc. may, in spite of great diversification and active trading of positions, contain a persistent credit spread, which can be only found if the actual positions are analysed over time through a rating filter. Lesser country credits may be overrepresented on the long side whilst the shorts are in more highly rated countries.
Such persistent rating differences can be the source of a significant credit carry income. This may or may not be an intended part of the strategy. However, if this carry is persistent, the performance and risk of the portfolio should be viewed as one part "passive" credit spread and one part the result of active investment decisions.
The investor should also be aware that a portfolio with a persistent credit bias will react more violently to global financial market turmoil when credit spreads widen out asymmetrically, i.e. the lesser credits widen a lot more and become illiquid sooner. It is not uncommon to see that bond investors substitute high-quality government bonds with corporate or even high-yield bonds. In this case, the investor has, in a matter of speaking written a financial crisis put to the market against receiving a premium in the shape of credit spread. This is unlikely to be clearly expressed in the risk report.
Option premium pick-up
If options are allowed in an investment strategy alongside its' underlying assets, there are numerous ways in which positions can be created which create a premium income for taking on un-limited risk. Non-linear instruments such as options are difficult to integrate in traditional risk system's reports. What may look like steady revenue can carry huge downside risks that may fly below the radar screen of the risk system.
Figure 1 CBOE PUT Index, an index that systematically sells put Option on the S&P 500
This risk is closely related to the credit carry strategy. It has a return pattern of slowly accruing profits, only to suddenly lose years of profits during one volatility expansion. The feeling the investor may get by looking at his performance and risk report is that the strategy is safe and has a consistent way of generating steady returns.
"A turkey is fed by the farmer every morning for 1,000 days. Eventually the turkey comes to expect that every visit from the farmer means more good food. After all, that’s all that has ever happened so the turkey figures that’s all that can and will ever happen. But then Day 1,001 arrives. It’s two days before Thanksgiving and when the farmer shows up, he is not bearing food, but an axe. The turkey learns very quickly that its expectations were catastrophically off the mark. And now Mr. Turkey is dinner.”
Nassim Nicholas Taleb - Author of "The Black Swan"
Selling volatility can be a profitable strategy over the long term, but requires deep pockets to stay solvent through the drawdowns. A position of net short volatility is likely to generate its' largest drawdown when equity market decline.
Depending on the sophistication of the risk system, even some basic option positions may not be adequately captured. Straddles (selling put and calls around a strike) or calendar spreads (selling options for one expiry and buying it for another) are hard to capture by anything other than a dedicated option risk system or capable of handling path dependency. In a more traditional risk system, these positions may well look like a riskless exposure, while they certainly are not.
Catching bubbles
When the dotcom bubble burst in 2000, many investment strategies that were not investing in Internet stocks were as badly affected as those that did. There was little or no diversification effect to balance out the large losses in Internet stocks. Apart from the obvious reason that market panic affects all risk assets, there was also a phenomenon, which was more extreme than ever before; single sector concentration in stock markets.
The extreme valuations of Internet stocks buoyed the entire stock market. Without Internet stocks, many of the large stock markets were actually declining even though the overall index was moving up on the back of the hysteria in Internet stocks. This in turn led to a corporate behaviour, which was relatively new and therefore not caught by risk systems.
Companies would re-engineer or just re-label part or all of their business as Internet related to piggyback on the boom in all Internet related stocks.
We can today see something similar, with the so called “FANG”-stocks (Facebook, Amazon, Netflix and Google). While the underlying companies have a clearly different business model than what was dominant during the dot com years, the price increase in these four stocks have had a disproportionately large impact on the S&P 500.
Risk systems were unable to adequately represent the new risk inherent in the Internet bubble. It is one of the weaknesses of all risk systems that they believe the markets are correctly priced even when there is a no connection with normal business logic and rationality. Risk systems are, as mentioned before, not very good at catching things that haven't happened before.
Style-drift
Style drift is when a manager is changing the ways in which risk is taken away from the original investment strategy. Not all style drift is bad. However, it may create a different risk profile and sensitivity to market events than what was intended and communicated to investors pre-investment.
An example; A highly systematic investment strategy, based on quantitative models may start to give the portfolio manager more discretion to take active investment decisions and override what the models tell him. This may go undetected in spite of dramatically altering the risk taking of the portfolio and leading to a very different behaviour in relationship to market events. For investors who were looking to have exposure to a particular investment style, adherence to the original trading methods is crucial. Style drift is difficult to detect over short-term periods and even with high degrees of transparency as it requires both deep analysis as well as pre-set and explicit expectations on how a manager should implement trades and take risk.
Minsky Moments and VaR
A fund that is governed by a VaR framework is mostly better risk managed than one that is not. VaR risk measures are trying to predict risk based on past information. The underlying assumption is that historical data contain information with predictive value for assets future price behaviour. When there is a change in market behaviour or a new market phenomenon appears, the predictive power derived from historical data will be greatly reduced. Unfortunately, such uncommon or new events are not so uncommon.
Uncommon events are often the reason for large market drawdowns or dislocations. So when a measure for how much a portfolio can lose is most needed, then VaR is the least useful measure for that. VaR is a necessary tool, but should not be confused with giving the answer to how much of the invested capital is actually at Risk.
Even in normal markets VaR has some serious flaws which the risk report reader must be aware of to not put too much trust in the numbers the VaR models produce.
The way VaR calculates risk means that risk is at its lowest after periods of benign correlations and declining volatility. This means higher than usual absolute positions are allowed within a given VaR limit. Most market participants use VaR for position limits and they are thus all allowed bigger nominal positions at the same time. If they actually take positions up to this higher limit, they all become vulnerable to VaR limits shrinking. This overall vulnerability to a deterioration of benign correlations and volatility is not captured by an individual VaR based risk system since it does not know the positioning of the overall market. Thus the risk for market wide selling pressure is not taken into account when estimating the risk for an individual portfolio. The risk system can only see its' own portfolio, but not how likely other investors are to turn into forced sellers.
This means that the real risk of large losses may be at its' greatest when the VaR risk report says it is very low. In the words of Hyman Minsky, “stability is destabilizing”.
The data needed to know overall market positions is not available, so the only tools available to handle this risk are personal insights and common sense coupled with experience. The risk system's VaR measure can lead to an erroneous impression of low risk compared with the real risk of significant losses.
Intraday trading
For any one instrument in a portfolio, risk systems typically use opening and closing mark-to-market prices in combination with the number of instruments held at the same times to calculate the profit or loss of that day. It is quite common that there are discrepancies between this method of calculating and the actual daily profit and loss. The portfolio risk analysis is generally based on the end of day positions.
Discrepancies between actual profit and loss and that calculated by systems can be difficult to break down and may remain unexplained and even build up over long periods. The profit and loss should go into the risk systems to calculate the efficiency or risk adjusted return of the portfolio. Without a correct and reconciled profit and loss, the risk efficiency calculation of the portfolio is flawed.
One investment activity, which may cause such discrepancies is intra-day trading. Good interaction between the risk department and accounting/operations will detect this. This may be difficult if front office and risk measurement are not in the same organization. Intra-day trading can therefore be unrecognized as the performance, and risk contributor it is.
Intra-day trading may be a way for the investment manager to generate additional returns, but it also generates additional commission expenses and can mean that higher risks are being taken during the trading day than what is seen in the end of day risk report. Intra-day trading activities are hard to detect and require a different approach to ensure correctness and accuracy of risk and performance attribution.
Risks hiding between front and back-office
We have described examples of unseen risks that are originating in front office practices and market changes. It is not the focus of this article to give examples of risks originating in operations or administration, even if these can be as large and costly. We would however like to point out that there is an in-between space where the investment manager may be the only source of information for back-office and administration. The following two examples are mentioned to highlight this in-between space where risk that goes undetected by the risk system may also arise.
Volatility is traded explicitly or indirectly on many underlying instruments and markets. It is not uncommon that volatility, i.e. the price of "risk" may be set differently in different markets, even when the underlying asset is the same or very similar. A listed stock option price may deviate from the volatilities embedded in the same issuer's convertible bonds.
This creates a problem for the risk system. Which price on single stock volatility should the risk system use? That of the listed stock option or that derived from the instrument held in the portfolio, i.e. the convertible bond itself? The investment manager may be the only source of input to answer this.
Some market instruments are not standardized enough to exclude the risk that they may not behave as expected when the event they are bought to protect against actually occurs. Credit Default Swaps (CDS), Exotic OTC options and structured products with embedded leverage and counterpart exposure are examples of where there may be contract and documentation risks. It is not always the case that documentation is screened before an investment is made. The risk system is unaware of this and will probably approximate the instrument with a standardised instrument.
Summary
Measurement and management of investment risk have taken quantum leaps over the past decades. The increased sophistication of risk systems has given rise to a new risk; that of having too much confidence in the risk reports. We may be blind to the risks that are not in the system.
“As we know, there are known knowns, these are things we know we know. We also know there are known unknowns. That is to say we know there are some things we do not know. But there are also unknown unknowns, the ones we don't know we don't know...”
Donald Rumsfeld – Former U.S. Secretary of Defence
As we have illustrated, risk systems are blind to everything they have not yet seen, to what they currently do not or cannot measure and to input that lie outside the data they capture.
To find the unknown unknowns requires a human process where experience and creativity are combined. With our examples, we have illustrated the existence of risks that can stay undetected by risk systems. The good news is that most of them are possible to find through the use of human input.
The prudent investor will always be on the lookout for any and all risks to his investments, including those are not yet known. He continuously attempts to detect, explore, monitor and manage the unknown unknowns. We hope this article in some small way can contribute to this.
