Evolving the investment risk profession
30 October 2019
Investment processes and portfolios which are managed by humans are, by nature, prone to errors and biases.
Michael Blair, Head of Investment Risk and Portfolio Analytics, asks if the investment industry is looking in the wrong places to advance risk management techniques.
- The industry is looking in the wrong places to advance investment risk management techniques.
- Analysis of manager behaviour should be given equal prominence with portfolio analysis.
- Separating luck from skill serves to focus on real value-add and risk control.
- Recent, widely publicised liquidity and portfolio issues in high-profile funds serve as examples.
What the industry is missing
For more than 20 years the techniques used to measure and attribute portfolio risk, and consequently to control and manage that risk, have barely changed. The rise of tracking error analysis in the 1980s and 1990s and factor risk models to measure, decompose and attribute this based predominantly on mean variance techniques has held fast despite many shortcomings.
As such, the focus of understanding risk in investments remains entirely focused on understanding the portfolio, or the portfolio in the context of the index. Investment risk model research development has targeted improving the forecasting of volatility, tracking error (or value at risk) of a portfolio, measurement of factor exposures and tilts, stress testing, and decomposition of such measures.
But what about pilot or driver error? Investment processes and portfolios which are managed by humans are, by nature, prone to errors and biases. Despite the extensive volume of literature and analysis on behavioural bias in asset management, very little of this has found its way into the toolkit of the investment risk manager, or indeed the portfolio manager directly.
Learning from others
I have found in my career that much can be learned about risk management from other industries.
Take Formula One for example. For a large part, this sport is an exercise in risk management combined with optimising performance. Compared to investment, the timescales are much shorter and you could argue the penalty for failure much greater. However, if we compare the car in F1 to the portfolio in asset management, a huge amount of data and analysis goes into understanding both the car and the portfolio ex ante and ex post. However, as F1 consultant Mark Gallagher explains, in top-level motor racing there is an equivalent level of analysis on the driver providing feedback on performance and behaviour to incrementally improve outcomes:
‘F1 cars are mobile versions of the Internet of Things (the IoT), in that they are constantly transmitting data back to the pit team and the team HQ, who review it in real-time and make quick decisions with the driver on how best to proceed. This is similar to what many businesses are doing in their industries, albeit at not quite such a fast pace!’1
I don’t want to stretch the analogy too far but imagine this type of process slowed down to the pace of active investing. I believe this is one of the most important ways investment risk management should evolve. That is, with risk measurement and feedback on the investment behaviour of the portfolio manager – the driver – in combination with portfolio analysis.
Ahead of the curve
At Martin Currie in 2010 we added specific questions on the reasons for trades to our order management system and by 2015 we started to exploit a rich data set of decisions taken and the reasons for them.
By decomposing these decisions in individual tranches (think every trim of a stock weight treated as a separate short position and every top up an individual buy) we built a large data set against which to test for driver error in the form of behavioural biases.
We tested for the recognised biases found in prior studies and some of our own. The results brought confidence when no evidence of bias existed, but also highlighted some things to consider and work on. In some cases, this resulted in an enhancement to the investment process with evidence in later years of improved outcomes, attributable in part to this adjustment. We continue to work on and develop our behavioural analytics toolkit.
F1 cars are mobile versions of the Internet of Things (the IoT), in that they are constantly transmitting data back to the pit team and the team HQ, who review it in real-time and make quick decisions with the driver on how best to proceed.
Understanding the risk and the value-add
Considering behavioural analysis as an additional layer of risk control can lead to a wider examination of the sources of risk and value-add in investment management products and processes. The graphic below illustrates this.
We have assumed that in a human process, behavioural risks cannot be eliminated completely but can and should be identified and managed as effectively as possible.
It is useful to consider ‘good’ risks, in green below, that are embedded, and part of the process and philosophy sold to the client, relative to ‘bad’ risks, in white, that are not. Risk management should involve controlling the size of ‘good’ risks, relative to the mandate and risk appetite and avoiding or eliminating ‘bad’ risks wherever possible.
Understanding the risk spectrum
Risk management should involve controlling the size of ‘good’ risks, relative to the mandate and risk appetite and avoiding or eliminating ‘bad’ risks wherever possible.
The return, or value-add side of the coin can also be considered in this way and an illustration of this is given below. This allows a more detailed assessment of the value being delivered for clients and a clearer appraisal of the value over passive alternatives. Analysis of this kind goes beyond simply comparing returns between active and passive products. It considers the inputs to both risk-control and risk-return seeking activity that is intended to deliver additional value. It also recognises the elements of absolute return delivered by active managers that come simply from being invested in the market or following a particular investment style.
Scaling the value-add of active management
Advancing risk-management techniques
We have seen some recent high-profile examples in asset management where a lack of process and insufficient control of ‘bad’ risks have led to serious issues for clients, including both capital loss and lack of access to remaining capital – at least, in part, some of these issues appear behavioural. Analysis and understanding of driver error, as well as better standards of risk governance processes, would have served to avoid the situations that arose.
At Martin Currie, we believe understanding portfolio and market behaviour in the context of risk and potential ‘bad’ outcomes is crucial for optimising the value-add for our clients. In addition, the evolution of portfolio manager behavioural analysis as a core part of our risk-management toolkit will remain a key focus for us. We believe this is long overdue in the asset management industry overall. This, coupled with a clear assessment of the true added value we are delivering and how that relates to our process inputs and risk control, is our goal in building strong partnerships with our clients.
We believe understanding portfolio and market behaviour in the context of risk and potential ‘bad’ outcomes is crucial for optimising the value-add for our clients.
This information is issued and approved by Martin Currie
Investment Management Limited (‘MCIM’). It does not
constitute investment advice.
Market and currency movements may cause the capital value
of shares, and the income from them, to fall as well as rise and
you may get back less than you invested.
The opinions contained in this document are those of the
named manager(s). They may not necessarily represent the
views of other Martin Currie managers, strategies or funds.
The information provided should not be considered a recommendation to purchase or sell any particular security. It should not
be assumed that any of the security transactions discussed here were or will prove to be profitable.