Flashcards in performance measurement 3 Deck (38)
risk-free rate of return can be defined as
the rate at which money is borrowed or lent when there is no credit risk, so that the money is certain to be repaid.
Assets issued by national governments in their own currency were traditionally assumed to be “risk free”, and so the risk-free rate was often referred to as the “Treasury rate”.
In practice, and depending on the particular context, the risk-free return might therefore be interpreted as the rate of return offered by:
short-term conventional government bonds
long-term conventional government bonds
index-linked government bonds.
This assessment was based on the thinking that national governments can always print money to ensure that their debts will be paid. Thus the concept of national debt being risk-free was only valid in a closed economy, where all financial transactions are in the currency of that economy. This feature was seen in much of the early part of the twentieth century, where large economies of China and the Soviet Union were effectively closed to outside influence.
However, as recent events have demonstrated, the ability of governments to print money is a route not available to governments that use a shared currency (such as the euro), or those who wish to tie their exchange rate closely to something else (like the dollar).
However in practice at the current time, all economies interact with each other, and therefore currency risk needs to be taken into account. Printing money deflates its value against other currencies, and thus the holder of overseas government debt may not receive a return that is free of currency risk, or is risk free in real terms if the inflation rates in the overseas territory and in the domestic economy of the holder differ.
For this reason the main rating agencies assign a credit rating to the government debt of most countries, generally assessed against the US Dollar. One key measure to be considered is the ratio of outstanding debt to GDP, as well as the maturity of the existing debt, since this will determine the cost of servicing the debt and the need to raise refinancing capital as existing debt matures.
Asset liability modelling
Perhaps the most important practical actuarial technique for use in determining an investment strategy nowadays is asset liability modelling.
Asset models and the fundamentals of Asset Liability Modelling were discussed in the Core Reading for Subject A301.
investor’s objectives will normally be stated with reference to both
assets and liabilities. In setting an investment strategy to control the risk of failing to meet the objectives a method of taking into account the variation in the assets simultaneously with the variation in the liabilities is required. This can be done by constructing a model to project the asset proceeds and liability outgo into the future.
In other words, by constructing and using an asset liability model.
The outcome of a particular investment strategy is examined with the model and compared with the investment objectives. The investment strategy is adjusted in the light of the results obtained and the process repeated until the optimal strategy is reached.
The optimal strategy being that which best meets the objectives of the investor.
Models can either be deterministic or stochastic. In both deterministic and stochastic models it is important that the assumptions made in constructing the model are consistent with each other.
It is essential to ensure that the assumptions and parameter values underlying both the assets and the liabilities are consistent if sensible results are to be obtained. In practice, this may be achieved by modelling of the appropriate dynamic links between assets and liabilities. For example, in many models inflation is assumed to influence both the asset returns and the liabilities.
Advantages of asset liability modelling
encourages investors to formulate explicit objectives. The objectives should include a quantifiable and measurable performance target, defined performance horizons and quantified confidence levels for achieving the target.
a typical objective might be to:
maximise the expected solvency level
at the end of a 3-year period
subject to the probability of insolvency at any time over that period being no more than 0.1%
For an institution such as a pension fund or insurance company the objectives might be specified in terms of the results of a valuation carried out at a specified time in the future. In practice there is likely to be feedback between the model output and the setting of the objectives. This may be because the additional insight provided by the model projections may help the investor to refine its investment objectives.
The success of the strategy is monitored by means of regular valuations. The valuation results will be compared with the projections from the modelling process and adjustments made to the strategy to control the level of risk if necessary. The model itself may also be refined in the light of how the actual experience compares to that which it predicted.
Deterministic asset liability modelling
A deterministic model is based on a single set of assumptions about future experience. Using the assumptions we project the asset proceeds and liability outgo for each future time period. We can then compare the incidence of income and outgo to check that the outgo can always be met. We will also want to know the value of assets and the value of liabilities at each future time period according to one or more valuation bases.
Stochastic asset liability modelling
In contrast to a deterministic model, a stochastic model treats the key parameters as random variables with a given mean and a defined probability distribution. For example, future inflation might be defined by a normal distribution with mean of 5% and standard deviation of 2%, and real interest rates (ie the excess of nominal short- term interest rates over inflation) might be distributed between –2% and +8%, say.
What assumptions would be required in an asset model used to estimate the future income stream and asset values of an investment portfolio?
In practice, we might develop a stochastic asset model in which each of the key variables is modelled as a random variable, typically depending upon
past values of itself
present and past values of other economic and/or investment variables.
Having run the model a large number of times, we would scan the results from each simulation to gauge how likely it is for an unsatisfactory result to emerge. For example, we might define an unsatisfactory result as the company becoming insolvent.
If say seven out of 1,000 simulations resulted in insolvency at some point over the projection period, then we might conclude that the probability of insolvency or ruin over that period is about 0.7%. A further set of simulations could then be performed based on a different investment strategy to see how the probability of ruin compared. By repeating the projections for a suitable number of different investment strategies, we can ultimately determine which is optimal in terms of best meeting our investment objectives.
It is important that the particular structure of the model and value of the parameters depend on the purpose of the model. Although there are powerful statistical techniques to aid in the construction of the model, it will be necessary to apply judgement in establishing the most appropriate form and appropriate value of the parameters. If the results are to be used in an intelligent and critical manner, the actuary should be aware of the strengths and weaknesses of a particular model before interpreting the output.
Using a stochastic model is, arguably, the most appropriate way of allowing for the volatility and uncertainty underlying the assets and liabilities. However, this is highly dependent on whether we can specify probability distribution functions that reflect the true level of uncertainty.
Asset liability modelling is discussed further in Chapter 18.
What is the main problem with relying solely on past data to develop a stochastic asset model?
What are the two main risks involved in the development of an asset liability model?
List five “actuarial” techniques that may be used to develop an appropriate investment strategy and that take into account the liabilities.
The financial risks faced by an institutional investor can be considered under five headings, according to the source of the risk:
1. Market risk is the risk relating to changes in the value of the portfolio due to movements in the market value of the assets held.
2. Credit risk is the risk that a counterparty to an agreement will be unable or unwilling to fulfil their obligations.
3. Operational risk is the risk of loss due to fraud or mismanagement within the fund management organisation itself.
4. Liquidity risk is the risk of not having sufficient cash to meet operational needs at all times. It is related to market risk in as much as the liquidity of the overall portfolio is need to be taken into account in portfolio selection.
5. Relative performance risk is the risk of under-performing comparable institutional investors.
Monitoring and controlling market risk Introduction
For many institutional investors, variations in the market values of its assets represent the main risk of failing to meet its investment objectives. For example, a fall in asset values to levels far below those anticipated may threaten the financial integrity of the institution. It is therefore crucial to try and control market risk. The first steps in the control process will be to:
identify and define the risks
develop techniques and/or models by which to quantify such risks.
12 sources of operational risk for a small domestic company.
Defining market risk
The first stage in monitoring and controlling market risk is to define what is meant by risk. A suitable measure might be the variance of the return on the portfolio over a specified period of time or the maximum loss that could be suffered with say 95 or 99% probability within the timescale. The latter measure is commonly known as Value at Risk (VaR).
We discuss the use of Value at Risk further in Chapter 21, when we consider measures of downside risk.
The returns and losses may be measured in absolute terms or relative to some suitable performance benchmark such as an index, an industry median fund or the value of the liabilities.
The timescale chosen will depend on the institution. For a long-term investor such as a pension fund the period may be measured in months or years, whereas for a bank it is likely to be measured in hours.
For example, the estimation of market risk for the purposes of the Basel regulations that are intended to ensure the solvency of banks is based on a 2-week timescale and a 99% probability level. However, banks are likely to use different time periods and probability levels for their own internal risk assessment calculations. In contrast, insurance companies and pension funds typically use time periods of several years.
Having defined risk it will be necessary to establish a mathematical model which will allow the risk at any point to be calculated. For example, in the mean variance framework, given the expected returns on all the assets in a portfolio and the variance/covariance matrix, the mean and variance of the return on the portfolio can be calculated. These can be used to calculate the desired risk measure.
In practice, banks often perform market risk calculations based on the assumption that the individual investment returns are normally distributed. Using estimates of means, variances and covariances of the returns for individual securities, it is then a simple matter to calculate the distribution of portfolio returns, which will likewise be normally distributed, and hence calculate the Value at Risk.
normality of returns may not be an appropriate assumption in practice, as
investment returns may demonstrate fat tails or negative skew.
An alternative approach is to carry out use simulations based on time series modelling. Thus an insurance company or a pension fund is more likely to use a suitable asset liability model to estimate the variance of expected surplus or the Value at Risk based on the surplus of assets over liabilities.
One desirable feature of any model used is that the factors that determine risk level are understandable.
Systems, reporting and benchmarks
The next stage is to ensure that the computer systems and data inputs are in place to calculate the risk exposure as often as is required.
The model can be used to assess the market risk associated with different asset allocation strategies and thereby determine which strategy is optimal given the objectives of the investor. The levels of market risk should be monitored at regular intervals, and the results of the analysis disseminated to all parties involved in the investment decision process, so as to enable them to make properly informed decisions.
Individual fund managers will need to be continuously aware of their risk exposures and the reports should also be made regularly to senior management.
In practice, fund managers will not only need to know their risk levels but will also need the tools to understand the effect of their actions on the risk of the portfolio. Guidelines will need to be established which can be translated into practical benchmarks and limits for departure from the benchmarks. A typical risk control system might therefore give the fund manager a benchmark asset distribution, expressed as a percentage of the portfolio in specified asset categories.
For a risk control system to be effective it is also desirable that it is clearly documented and that the personnel responsible for monitoring risk are independent of the fund managers.
State two ways in which allowance might be made for non-normal investment returns.
Load differences and load ratios
Limits for departure from the benchmark would also be set. Limits can be expressed in two ways.
A load difference specifies the range over which the percentage allocation to a specific class can vary, for example limiting overseas equities to between 5 and 15% of the total portfolio.
A load ratio specifies the maximum variation of the allocation to a specific asset class expressed as a percentage of the benchmark allocation to that class. This has the advantage that a constant load ratio permits a smaller absolute variation in the lower weighted asset classes.
A fund manager is set a benchmark for overseas investment of 20%, with a permissible range of 15% to 25%. Express this benchmark as a load ratio.
What type of base output would you expect from a simple risk control system for an institutional investor?
You have been asked to develop a risk monitoring system. List the desirable features and requirements of such a system.
How might risk be defined or measured for a manager responsible for a unitised fund, eg the pension managed fund run by a life insurance company?
Monitoring and controlling credit risk
key factors in managing credit risk are:
the creditworthiness of the counterparties with which an institution deals;
the total exposure to each counterparty.
Creditworthiness can be controlled by placing limits on the credit ratings (as published by the major rating agencies) of counterparties with which an institution may deal. Different limits might be imposed depending of the potential length of exposure. For example investing in a company’s corporate debt is a different risk to a settlement period risk with a broker or market maker. Credit risk can be also controlled in derivatives transactions by dealing on a recognised exchange with a central clearing house that stands as counterparty to all deals, rather than over-the-counter. The clearing house will seek to protect itself by requiring the counterparties to deposit “margin” with it. These margin payments are a particular example of the use of collateral provided by a counterparty as a tool against credit risk. More generally, any institution may demand collateral or margin payments as a way of protecting itself against the credit risk it faces under a particular transaction. The operation of a clearing house was described in more detail in Chapter 1.
It is important to monitor and place limits on the credit exposure to any single counterparty.
An example of such a limit might be that the investor is limited to a total investment in a bond with a credit rating of BBB of no more than the maximum of 5% of the total fund size or $250m. Of course the upper limit could be zero for bonds with a low credit rating.
This will involve aggregating exposures in different areas. For example a pension fund may hold both equity and debt issued by a bank as well as having cash on deposit with the same bank and having them as a counterparty to a derivatives deal. It will also be necessary to be aware of the particular relationships between different companies within the same group.
In addition, credit risk can be hedged using the credit derivatives discussed in Chapter 3.
Allowance for credit risk is often made in the model used to analyse Value at Risk. In practice, many quantitative investment models allow for both elements of risk simultaneously.