In this post we’ll go behind the scenes, to see how professional firms construct investment portfolios for clients, beginning with portfolio optimisation.
Portfolio optimisation
After the investor understands their client’s goals, they will usually begin with a process called portfolio optimisation.
The aim of portfolio optimisation is to find the combination of investments that will have the lowest risk for a given return objective. Or sometimes, to find the combination that will have the highest return for a given level of risk.
For example, imagine a client has asked its investment manager to beat the return on cash by 3% per year, using a mix of equities, bonds and alternative investments (real estate, private debt, hedge funds etc.).
The investment manager will use portfolio optimisation to help decide how much to allocate to each investment to achieve cash + 3%. The optimisation will tell the investment manager which combination is expected to be able to do this with the least amount of risk.
What goes into an optimisation?
The investment manager will start by estimating the following for each type of investment:
- Its future volatility. Initially, this is likely to be based on how volatile the investment has been in the past – say the last 10 years. Some managers may then adjust this up or down based on their forward-looking views for the investment.
- Its future returns. As discussed in this post.
- How its returns will be related to (or correlated to) the returns of other investments. Again, this is likely to be based on how correlated the investments have been in the past.
These inputs can then be used to estimate the future returns and volatility of any combination of investments1.
Results of the optimisation
A computer algorithm will then forecast the return and volatility of hundreds or thousands of different combinations of investments. This will produce a chart like the one below.
The line in the chart is known as the efficient frontier. Each point on the line represents the forecasted return and volatility of a particular combination of investments. The combination will be the one that the algorithm has found to have the lowest volatility (horizontal axis) for the level of return (vertical axis).
On the very bottom left of the line will be the least volatile portfolio. This is usually just a single investment – cash.
On the very top right of the line will be the most volatile portfolio. This is also usually a single investment; in this case, the one expected to be the most volatile – perhaps an investment like private equity, or emerging market equities.
The investor can then begin to build a strategy based on the portfolio on the efficient frontier that corresponds to their desired level of return. This is because the investor knows that no portfolio is forecasted to deliver the same level of return, with lower volatility.
Constructing an investment portfolio: improving the efficient frontier
A skilled investment manager will be on the look out for ways of pushing the efficient frontier up and left, for a better risk / return trade-off.
For example, investment managers are always researching new types of investments. Adding a new type of investment to the mix may make it possible to construct portfolios with better risk / return trade-offs.
They will also look to squeeze costs wherever possible. This could be by negotiating better fee deals or finding more cost-effective ways of accessing an investment. Reducing unnecessary costs should increase the return of a portfolio, without increasing risk (pushing the efficient frontier up).
Challenges with portfolio optimisation
As with all mathematical modelling, investors need to be cautious when using the results of an optimisation exercise.
All models suffer from the garbage-in-garbage out problem2 and optimisation models are particularly vulnerable to this given how subjective the choice of inputs is.
The results of the optimisation will depend on the estimates for returns, risks and correlations fed in. As noted earlier, the volatility estimates tend to be based on historic patterns of returns. This is a reasonable starting point, as the relative volatility of different types of investments has tended to be fairly consistent over time; the most volatile investments have tended to remain the most volatile investments etc. However, there is no guarantee that the future will look like the past.
Correlations are trickier. These are less stable than volatilities. As discussed in this post, riskier investments can become much more highly correlated during periods of market turmoil. Also, government bonds usually move in the opposite direction to equities because of their flight to safety characteristics. However, sometimes, bonds and equities can behave much more alike. For example, high inflation is generally bad for both bonds and equities.
For these reasons, investment managers may run their optimisation algorithms using different correlation and volatility inputs, to check the results still hold up.
The same applies to return forecasts, which are also subjective. Newer optimisation approaches may add an extra input for each investment, which places a higher weight on those investments that the investment manager feels more confident in their return forecasts for3.
Next steps when constructing an investment portfolio
The optimisation process produces what’s known as a Strategic Asset Allocation (SAA). This is like the first sketch of a painting. It’s a good starting point, but the final product may look quite different.
Firstly, the investment manager will want to know how resilient the portfolio is in tail-risk scenarios4.
They can do this by shocking the forecasted returns of the different investments, based on how they think the investments will behave in different situations. This is called stress-testing. The investment manager may modify the SAA after stress-testing it, to make it more resilient.
Also, not all the possible variations of an investment strategy can be easily captured in an optimisation exercise. Some investments will not have enough data to produce reliable return, risk and correlation estimates. It’s also difficult to capture different investment management styles.
Investment managers may also use sophisticated strategies to target certain risks, such as the risk of large losses in the equity portfolio. These are difficult to model in an optimisation, although other modelling approaches – including stress-testing – may be used to assess the effectiveness of these strategies.
Constructing an investment portfolio in practice
Finally, the SAA is fixed in time. An investment manager will usually want to layer on their views on markets and the economy, to adjust clients’ allocations over time. We’ll walk through how investment managers translate SAAs into actual client portfolios in the next post.
- We won’t cover the maths in detail here; however it’s like the formula for combining random variables found in a typical introductory statistics course. ↩︎
- = the answers you get from a model will only be as good as the data you have fed into it. ↩︎
- This is sometimes called “Robust Optimisation“. ↩︎
- I.e. extremely negative markets, like those in the 2008 credit crisis or the pandemic. ↩︎
Leave a Reply