Looking at absolute risk parameters, investors’ main objective is to have an estimate of the potential loss they may incur. Many methodologies can be used, such as historical and forward-looking simulations, to assess the ultimate risk limits. An ultimate risk limit should be activated when realised performance becomes close to a shortfall or loss, corresponding to an institution’s risk tolerance limit.
Risk budgets can sometimes be defined in terms of an absolute amount expressed in local currency: for example, Danish pension fund ATP’s total risk (the risk consumption) is calibrated using a simulation of its benefits commitments. Others define risk tolerance as maximum shortfall over a certain horizon (for instance 15% for the medium-term asset allocation plan of a major Asian pension fund). A shortfall constraint can be translated into a Value at Risk (VaR) or Conditional Value at Risk (CVaR) constraint, with specific assumptions regarding horizon, frequency, confidence level, etc. CVaR represents the expected return of the distribution below a certain threshold. It is more precise than VaR as it captures the asymmetry of the distribution; our optimisation process at Amundi focuses on CVaR instead of standard volatility / VaR minimisation.
In the institutional world, VaR and CVaR are the most used measures to quantify potential loss, even though investors also typically monitor portfolio volatility and Sharpe ratio, which are then usually measured on an ex-post basis and are not forward-looking constraints.
Risk parameters can also be related to relative risk, and tracking error volatility (or TE), defined on an ex-ante or ex-post basis, is the most used and known indicator for this purpose. TE should be set as a function of the institution’s excess return target, which itself depends on the information ratio it aims to achieve over the long term, as well as on its investment beliefs regarding the potential value added to be generated through active management.3
The tracking-error of active management (including both Tactical Asset Allocation and security / fund selection contributions) relative to institutional investors’ SAA is generally modest. A Mercer survey cites a 2% average tracking-error, with a 0.3 to 0.35 information ratio, corresponding to an expected excess return of 50 to 75 bp.4
These tracking error targets should be viewed as indicative and not as strict limits. They should be monitored over a relatively long period, such as 3 to 5 years, as this is representative of a typical market cycle. Indeed there are times when investment opportunities appear more attractive, justifying devoting a higher risk budget to active strategies than the long-term average for example.