Asset-Liability Management: Science or art?
When I qualified as an actuary, the so-called financial economics debate was just erupting within the actuarial profession. For pension schemes it overturned the orthodoxy that equities were the best investment and replaced them with fixed income. Both approaches were wrong: they were too simplistic, and ignored the investment golden rule of diversification.
But in sense both approaches were correct. At short durations equities are a poor match to inflation-linked liabilities, but at long durations they are a decent match. There is a term-structure which is observed in many economies to the correlation between equity and inflation from around zero, or even negative, at the short end, to around 0.5 at the long end.
Strategic Asset Allocation
The starting point for making investment decisions is the strategic asset allocation (SAA). For institutional investors such as insurance companies and pension schemes something better than simplistic models is required. Asset-liability modelling (ALM) is the key tool to use to determine the SAA. ALM combines three elements: the assets, the liabilities and a stochastic economic scenario generator (ESG). An ESG is a model of the economy which creates multiple (at least a thousand, and often many more) simulations of the future path of the economy and its financial markets, over a given time horizon.
At its simplest the assets are the SAA under consideration, a set of asset classes that add up to 100%. But it can be extended to include strategic derivative policies, for example. More advanced approaches include dynamic SAA where the SAA changes on a trigger, be it time or funding level, for example. Typically, the actuary will consider several alternative portfolios, often with the addition of new asset classes, alongside the current SAA.
The liabilities and their associated metrics use the output from actuarial valuation, ideally the yearly liability cashflows (which I call an “armadillo”) rather than their discounted value. ALM can be seen as a natural extension of actuarial valuation. A roll-forward algorithm is required, which is straightforward for assets but more complicated for liabilities, as for example you may need assumptions for new policyholders, and various approximations are usually used.
Economic scenario generation
Creating an ESG is straightforward, starting from mean (expected return) and standard deviation (risk) for each asset class, and the correlations between them. However, creating and maintaining a meaningful ESG from a coherent economic model to produce realistic simulations is technically demanding. The ESG needs to provide yearly return figures for each asset class under consideration over the time horizon for each simulation. For fixed income it is better that the ESG provides the yield curve (or curves, if multiple currencies) from which returns are derived based on the characteristics of the fixed income asset class.
Art and science
ALM provides much valuable information on the risk-return trade off to determine how the investments of an institutional investor should be implemented. But it is dangerous to see ALM as a black box which spits out an optimal answer. The value-add in ALM is not just in the quantification of the trade-off off between conflicting objectives and risks but also in the conversations it facilitates on objectives for risk tolerances. There is both art and science in ALM. The science will determine how the assets can be made closer to the liabilities, for example by buying longer dated fixed income or through a complex swaption strategy. The art is determining which is best for the characteristics of an investor and its decision-makers.