Mark Saunders: Strategic Asset Allocation (SAA) is a process by which institutional investors can identify a diversified portfolio that meets their investment goals. Let’s imagine an insurance company needs to earn a 4% a year return on the premiums it has collected – many different investment strategies could achieve 4% a year. Insurers will use an SAA framework to identify the optimal portfolio that is expected to return 4% a year whilst also minimising the volatility, or uncertainty, of that return.
In reality, companies are not just concerned with the average expected return of their portfolios. They also care about aspects such as liquidity, earnings, return on capital, downside risk –how much the portfolio might lose in a market crash – and the outlook for future solvency levels. An advanced SAA process will take several factors into account when producing an optimal portfolio.
Mark: The sophistication of a company’s SAA is proportional to three things: the size of the company, the duration of its liabilities, and what you might call its management objectives – the metrics that its board monitors and targets.
"P&C business is fairly short-term, so companies writing these policies
have a correspondingly short investment horizon."
A small company doesn’t typically have the resources to dedicate to a sophisticated SAA, so will usually have a fairly basic approach – or might even outsource its investment management to a third party.
P&C business is fairly short-term, so companies writing these policies have a correspondingly short investment horizon. Plus, they can manage financial performance by re-pricing policies at renewal. So, these companies often have fairly simple SAA objectives, with most of their profits coming from underwriting.
In contrast, life insurance business can run over twenty years or more, so the investment strategy becomes a much more fundamental part of the business model. A company could end up investing billions of pounds of premiums over decades, and a significant element of the company’s overall profits may be coming from earning a return on those premiums.
This means they will usually look to incorporate a range of constraints and objectives within their SAA, rather than just targeting an average return. The choice of metrics they include will depend on what is most important to their board: producing cash flow, managing their solvency position, or maximising return on capital.
Mark: There are a lot of European mutual life insurers with strong balance sheets, where solvency is not something they need to manage in terms of their investment policy. In this case, they can focus on maximising return for a given level of investment risk, without worrying about other constraints. These companies are the minority, though – most insurers do monitor how their investment allocation will affect their solvency position.
"Managing solvency through an investment strategy is a common lever, but the
better question is where that should come into an SAA process."
Public insurance companies have different options when it comes to managing their solvency. For example, they can change the type of insurance risks they retain, sell back books of their business, or raise new capital. However, companies often intend to keep liabilities on the books over the long term: they don’t want to sell purely for solvency reasons and raising equity could upset existing shareholders. So, the simplest way to control the solvency ratio is often through their asset allocation.
Managing solvency through an investment strategy is a common lever, but the better question is where that should come into an SAA process: right at the end or as a constraint within the SAA – for example, limiting the allocation to 20% in equity? Or should it be optimised for a position within the SAA tool itself?
Mark: If a UK insurer uses a traditional mean-variance approach to SAA – that is, looking at the average and standard deviation of return – diverting a small proportion of their portfolio into US equity will increase returns with little impact on volatility, because of the diversification benefits.
This means that the optimal portfolio could allocate a proportion to US equity. However, if we include solvency in the optimisation, the additional capital requirement from investing in overseas equity may negate the additional returns you earn. The optimal portfolio may invest in domestic rather than overseas equity – or might not invest in equity at all. It all depends on the return assumptions and the investment horizon.
In other words, incorporating solvency metrics within the SAA forces you to not only think about the additional return that can be earned from investing in an asset class, but also the capital charge that would be incurred, and if that additional capital is worthwhile given the additional return.
Mark: Companies were thinking about how their investment choice impacts solvency back in the days of Individual Capital Assessment (ICA), pre-Solvency II. However, historically companies have evaluated aspects like solvency at the end of their SAA process, after performing the efficient frontier analysis. They select one or more portfolios from the efficient frontier and look at how they will impact a range of metrics. This approach can be cumbersome and is unlikely to result in the optimal portfolio.
We believe that if market risk is a material element of solvency capital, it should be brought into the optimisation function. So, the solvency capital is calculated over thousands of scenarios for every point on the efficient frontier. This method is more robust when it comes to evaluating different potential investment allocations.
"We now have the software modelling capability to calculate solvency capital
within the SAA optimisation and within the efficiency frontier process."
When testing hundreds of portfolios across thousands of scenarios, it’s a lot of computation. There are a many data points, and for each of those, the model is projecting forward over multiple time steps. Historically, there hasn’t been the software or the computing power to easily do this.
We now have the software modelling capability to calculate solvency capital within the SAA optimisation and within the efficiency frontier process – where these different portfolios can be tested and to find out on average how it performs over, for example, 10,000 stochastic scenarios, which are random scenarios with different variables for interest rates, inflation, spreads, and equity returns.
But it’s still a resource-heavy operation – which means tools to make it easier are beneficial.
Even today, a lot of companies don’t test that many portfolios: they'll have a stochastic model and test five or ten different portfolios, and that will take them weeks of work to run the models and analyse the results.
Mark: It’s difficult to give a precise number, but the majority of medium-sized life insurers managing £5 billion-plus in assets are using a more sophisticated approach to SAA. These more sophisticated approaches range from using a stochastic liability model to a dedicated SAA tool, which incorporates a range of metrics.
"It’s important to emphasise that, as with all models, the output is only
as good as the inputs."
Across different companies there are varying levels of management buy-in for these sophisticated SAA tools. When we introduce SAA, we start by showing the benefits of a stochastic approach to it before adding in liabilities and solvency capital.
Of course, it’s important to emphasise that, as with all models, the output is only as good as the inputs.
Also, no tool beats the knowledge and expertise of experienced human staff. This tool can provide a powerful analysis, but you still need the right people to interpret the results and make informed decisions.
Mark: One development is machine learning and neural networks. A dedicated SAA tool will model things at a product level – rather than a policy-by-policy level, which is what you get when using a liability model.
"A dedicated SAA tool can produce the metrics that the company needs
and allow them to test hundreds of portfolios."
This difference matters in areas such as With Profits policies, where there is a guaranteed level of bonuses that they might pay out, or, for example, when portfolio returns are above a certain level it means that the liabilities will increase.
The availability of neural networks packages within common coding languages means that the output from a liability model can easily be fitted to drivers such as asset returns and inflation. A dedicated SAA tool fitted with a function to use neural networks can produce all the metrics that the company needs and allow them to test hundreds of portfolios in a matter of hours, versus the weeks it can take to test only a handful of portfolios using a stochastic liability model.
This means our software can tell clients how their liabilities will change for different portfolios in much shorter timespans than has previously been the case.
