Trustees are familiar with the world of triennial actuarial valuations, and the accompanying review of contribution rates and investment strategy.
Typically, time would be allocated to consider the investment strategy (strategic asset allocation, or policy portfolio) in order to ‘get it right’. It would then be fixed for three years, sometimes longer.
In the last two decades of the 20th century this process served us well. The three-year bear market that followed was a challenge, but not necessarily a sign that the process was flawed. In fact, the whole point of a strategic allocation was to give a fixed point in navigating through choppy markets.
But that notwithstanding, there are now good reasons to make this process more dynamic, to reflect faster changes on both sides of the balance sheet and a greater focus on risk allocation.
Working from modern portfolio theory, with its efficient market and random walks, holding a static allocation and rebalancing back to it is the right thing to do. We know the expected outperformance of equities over bonds, and the risk associated with this. So when equities outperform and increase their weighting, we rebalance and take our profits; when equities underperform, we rebalance so that we are not underweight when equities next outperform.
However, we are now claiming that this is not the right thing to do.
The main reason relates to our questioning of the assumptions underlying modern portfolio theory. Is it true that asset returns follow a random walk, in other words that after going up a lot they are just as likely to go up next period as down? Or do they follow a mean reverting process, in other words after going up a lot are they more likely to go down next period than to go up?
Or indeed, do they experience distinct phases of high returns followed by low returns and vice versa, what we might term ‘regimes’? If this is true, and we believe it is, then it would be natural to change the strategic allocation as market conditions evolve.
Opponents of this view can claim that this is nothing more than glorified market timing, and there is little evidence to suggest that market timing can be consistently successful. There is, however, some high profile support to the contrary.
In the spring of 2003, the influential US investment strategy expert Peter Bernstein caused much debate by suggesting that strategic asset allocation was misused and often leads to an unhelpful focus on a single measure of risk (tracking error). His concerns were broader than this but are beyond the scope of this article.
For our purposes we agree with his concern that policy portfolios have become a substitute for thinking. Funds have had a false optimism that it is possible to have a simple policy that is designed for all weathers. The real benchmark should be the return required by the liabilities.
In this light it is logically better to use an active portfolio, in other words a form of dynamic strategic asset allocation, to adjust to market conditions and get the best from the opportunity set of assets rather than a fixed policy portfolio. This certainly follows if you believe as most do that equities and other asset classes are constantly changing their prospective risks and returns.

The nature of the changes in market conditions is important. We need to differentiate between the volatility associated with random movements that can go either way - up or down - and changes that will persist in one direction only.
If, however, the world is subject to changing regimes, each lasting a number of years, in which the return difference between equities and bonds can vary markedly, then equities become genuinely risky – even for long-term investors. In this case, the concept of an equity risk premium only really makes sense over the very long run - say 100 years - as the succeeding regimes revert around the ‘true’ mean value.
Essentially, investors only have partial information about the companies and markets in which they invest and therefore misprice assets; cycles of optimism and pessimism push markets to overvalued and undervalued levels.
It follows that if this theory is an accurate description of the world, then we will be interested in identifying regimes and altering our allocation accordingly.
Having argued the case for dynamic allocations being appealing in principle, we need to consider whether we can achieve this in practice. We will need to be wary of accusations of market timing, and therefore will need to find a disciplined process – one that can be implemented in largely ‘passive’ form according to predefined rules with limited need for subjective judgements.
Any dynamic allocation approach that seeks to enhance returns will need to follow some form of buy-low, sell-high process. This in turn requires an assessment of value to be made. Buying low requires the process to identify the asset as cheaper than ‘normal’ or average.
As we have indicated, proceeding down this route will require us to have a belief in markets varying in their prospective pay-offs, which in its turn implies some level of mean reversion. The theory of random walks embedded in classical finance theory gives us no support for such an approach.
Intuition would appear to be on our side - trees do not grow to the sky and prices cannot stray from fundamental value indefinitely - but belief is necessary as hard evidence is in short supply. There is weak support for mean reversion in the empirical data but it is not totally conclusive. We simply do not have long enough price histories to give us sufficient data points to examine. But we do have a logical theoretical underpinning that also sits comfortably with most people’s intuition and observed market behaviour.

If we can make a logical case for a more frequent approach to strategic asset allocation, how exactly should we proceed? We think trustees should consider three distinct elements for successful implementation:
q Establishing the risk budget policy: this framework calls for agreement on a risk budget amount – the value at risk (VaR). This target should be expressed in terms of an appropriate range and an associated target range for principal asset groupings. This step is not meant to constrain unduly, but makes the annual process more effective;
q Deciding the actual strategic asset allocation (SAA) at regular intervals: applying market-consistent assumptions - our best information on market risks and returns likely in the next few years - we apply the risk budgeting process to develop our optimal SAA within the risk budget policy, making full allowance for the transaction costs necessary;
q Applying a rebalancing process quarterly or monthly until the next risk budgeting analysis: rebalancing takes on board two inputs: first the degree of divergence in actual allocations from the strategic weights; second, the extent to which the VaR has moved outside the policy range; both inputs have to be viewed together. Rebalancing asset allocation in particular is best set up as a largely automatic process based on triggers at threshold points. On the other hand, breaches of VaR budgets have to be considered individually based on context.
We believe further automation may be possible to make this process suitable for use by trustee boards.
We have suggested that a static strategic asset allocation may not be in the best interests of pension funds and that many trustees will do better to adopt an asset allocation policy that changes in the light of shifting market conditions following a disciplined investment process. This approach is most often going to be referred to as dynamic strategic asset allocation (DSAA).
Tim Hodgson is a senior consultant at Watson Wyatt in the UK and a member of its Thinking Ahead Group

Topics