Oct 24

Which Strategic Asset Allocation Model is Best?


We have all heard that choosing the asset allocation for your insurer will basically define your investment results. In other words, strategic asset allocation explains about 90% of the variance in performance.

Tactical asset allocation can add value, but is difficult to achieve in the long run for a host of reasons. These include transaction costs, taxes and issues in the ability of managers to add material value over passive benchmarks over long periods of time. Quite frankly, common sense dictates that the more efficient the market for the asset class, the more difficult it will be to add value with tactical asset allocation.

Thus, we are back to starting our discussion about the importance of asset allocation with strategic asset allocation

But, where should we start?

Most practitioners will reach for their Markowitz efficient frontier models (or MPT, Modern Portfolio Theory), plug in historical values for return, standard deviation and correlation amongst asset classes, beginning the discussion in that manner.

Those same practitioners, if prompted, may point to the inadequacy of using historical values and suggest more reasonable expected returns, standard deviations and correlations. Assuming these projected values are reasonable, we are still left with some of the major issues in MPT, including:

  • Asset returns are normally distributed. Alas, they are not. Even something that may seem to be normally distributed, like US equity returns are not. Upon further inspection, a log normal distribution (with thicker ‘tails’ on both the far upside and far downside) seems to do a better job of approximating historical returns. And even that can understate strong downward moves.
  • Correlations are fixed and do not change. Alas, that is not the case. There is actually a web site that helps prove the point. Just focus on a simple correlation between US Large Cap (e.g. S&P 500) and the Barclays’ Aggregate Bond Index, and you will see correlations go from .05 to -.05 to 0 to .18, depending upon the time period chosen. A big difference? Not much by itself, but when combined into the MPT model, especially with other assets, it can make a big difference.
  • All investors are rational and risk-averse. In other words, all investors will want to be on the efficient frontier (e.g. highest return per unit of risk) at all times. I don’t know about you, but I have yet to meet a completely rational human being. Importantly, we see ‘irrational exuberance’ to some degree in many situations, especially the financial markets.
  • If all investors followed MPT’s recommendations, it would probably invalidate some of the assumptions in the model. In other words, if everyone used MPT because it provided the ‘best’ asset allocation mix, it would then not provide the ‘best’ asset allocation mix. In other words, model inputs change if everyone follows MPT.

There have arisen changes to the MPT that answer some of these major issues.

For example, post-modern MPT uses non-normal distributions. And, with the Black-Litterman model, the investor is required to state how his or her assumptions about expected returns differ from the market. And, then the investor must state his degree of confidence in the alternative assumptions.

Also, another look at strategic asset allocation is provided by an approach that uses regime changes in market turbulence, economic growth and inflation to forecast expected asset returns. Determine the regime and you will get a better idea of returns, risks and correlations. There is much more work to be done on this approach, although it does move a step closer to a tactical asset allocation approach than a strategic one.

Of course, the major problem with 100% of these models is the nature of the beast: Financial market returns, risk and correlations are all somehow related to that which has occurred in the past. This works with things that do not change and only have to be ‘discovered,’ such as we find in traditional physics (less so in quantum physics), chemistry and biology.

However, how many of us remember the last time central banks played such a major role in world economies and in the financial markets? I don’t see any hands being raised, so I think you see what I mean.

We are operating in an environment where all the fancy models in the world may not be a very close analogue to what we may see going forward…in addition to the major issues and approaches in those models, already noted. Thus, we must be careful to shy away from ‘physics envy.’

With that in mind, we must think creatively when it comes to strategic asset allocation, and that probably starts with a simple mindset.

What are the reasonable returns we can expect on a given asset class over the next 3-5 years (assuming that is your relevant timeframe)? How might correlations change over that time frame (we can perform a sensitivity analysis to see if and how much that might matter)? And, really, what is the ‘worst case’ returns one can expect in a given asset class performance, given what we know about today?

These are very, very difficult questions that should lead to very productive discussions with your investment consultant and investment manager.

Are you ready for some Fumbling?

Return to “From the Northwest Quadrant”

For More of the Latest Insurer Investment Insight, please visit InsurerCIO.com