Part of the lure of modern portfolio theory is that it provides a theoretical basis for a correct asset allocation to all assets. If we knew the expected return of bonds, equities and other assets, and we knew how their prices would move in relation to each other over time, and all these price movements were random, independent and uncertain, the MPT model would be the perfect model. For those who wanted to increase risk or reduce risk, just reduce the allocation to the market portfolio or leverage up. Everyone needs an allocation benchmark! Such a structure, assuming the satisfaction of the assumptions, would be optimal even for those drawing down on capital.
The problem is that correctly estimating the key inputs of these models is impossible given that the only way to do so, given the dynamics of the model, is to extract inputs from past relationships or current implied relationships.
Now, you might say, well, let us input what we consider to be the expected returns from our valuation models, and the risks and the correlations. The problem here, is that the MPT structure was not designed to be a tactical model. A tactical model is held together by its benchmarks and reference points, whereas MPT structures act like manic depressives when shown such sheer disparities in inputs. The only way MPT models can be used is if they, like a lunatic in an insane asylum, are bound, gagged and drugged.
And besides, the moment you start to input your own views on valuation, returns and relative price relationships you risk operating outside the boundaries of beta. You enter market timing territory in MPT parlance, taking on a de facto allocation to lagged beta.
In the end, MPT structures merely end up providing broad allocation structures, the main benefit of which is an allocation to bonds, equities and other assets like commodities and property. The illusory benefit in terms of visuals is that the statistical outputs provide a range of risk/return trade offs that can be used to match structure with client risk preference, although the linking of those structures to asset liability demands, in the absence of modelling, may at times be no more than guesswork.
Instead of being optimal risk/return structures, they become allocation management structures. There is a lot of research that has shown that basic 1/n allocation structures have produced superior risk adjusted returns to mean variance structures. What this shows, to my mind, is that markets are not efficient and that an allocation structure that sells high and buys low is able to harvest return and reduce risk by taking advantage of non market efficient relative price movements. Though, the fact that a rebalancing 1/n structure outperforms a rebalancing MV structure, is also only a matter of chance. 1/n is just as random a structure as an MVO allocation, although the bigger the n the greater the allocation to differing relative price movements, which enhances the opportunity for return management.
But, if economies and markets are not efficient, and more importantly future prices are to a certain extent influenced by past price movements (and hence demand flows), then you can have clear risks that should require management by structure and discipline. If MPT suggests that structure should take note of efficient markets, then the existence of non efficient market and non efficient market risks suggests that structure should likewise adapt.
The reason why MPT only assesses point in time standard deviation, is that this is the only risk within an efficient market model. Outside of such a precept, you end up with time varying risks, and where you have time varying risks and liabilities you need to at least consider liquidity, credit risk, duration and yield. All three have intersects with equity, bond and portfolio yield dynamics.
In a non efficient market world, we no longer have point in time standard deviation risks dominating, but short and long term risk return trade offs. If you knew that bonds would outperform equities by 20% over the next three years, would you really have any allocation to equities? In a sense this is the same dilemma that would strike an MVO structure dead, forcing it to weight wholly in the dominant asset class. But the non efficient world risk is still one of uncertainty over the timing and the magnitude of the outcome. We still need to hedge our bets and how much we hedge our bets is dependent on our risk preferences and valuation differentials. We should end up with a range of allocation structures with varying allocations to bonds, equities and/or other asset classes, that we can then illustrate via a standard deviation and return analysis. So the two intersect, using similar tools but for different reasons.
In the end, we may well end up with vary similar structures – MVO and disciplined valuation driven structures. Where this all stops making sense is when we start to include liabilities as determinants of portfolio structure. How do we adjust for the need to meet liabilities?
In an MPT world, the supply of bonds and equities may just as well be determined by the expected returns (and hence the cost of capital and the required return for bearing risk) and relative price sensitivities to new information. In a non MPT world, the supply and price of bonds and other low risk assets may be more related to the need for greater certainty of income and capital given the risks of risky asset classes with potentially higher returns. This itself is also impacted by the relative balance between the saving of capital (deferral of consumption) and the consumption of capital (sale of capital for cash). Hence the demand and supply and pricing of assets is also partly determined by economic consumption, saving, production and investment decisions as it is by portfolio risk/return.
If high return asset classes were not indeed risky and return was certain, you would weight exclusively in equities. So the existence of lower risk assets is partly due to aversion to price volatility but more importantly the need for certainty of income and capital. As such, given the nature of risk and the time frame of significant risk, you can build an ALM model that allocates to lower risk and risky assets based on the C,S,I and P decisions within an economy, given the nature of risk and expectations of risks and returns. This is the basis of portfolios split at one level between those assets that would be used to meet liabilities in a risk event and those assets to be used for longer term consumption. In reality, this portfolio acts as a normal portfolio in the absence of a significant risk event, in the sense that there are no individual pots allocated to specific goals (the whole is optimised), the portfolio balance, relative to risk/return inputs, is more or less stable (lower risk allocation would extend as economic and market risks extend), and the portfolio itself would only activate its dedication sequence once a significant market and economic risk event has impacted returns. It is worth noting, that a market does not enter a risk event as soon as it starts falling, given that a portfolio could still be holding significant realizable returns well after a market downturn. Also, the diversified benchmarked structure of an efficient vehicle of this type would provide enhanced management of risk and return.
But MPT has also developed its own human capital theory of portfolio structure that, at first glance, would appear to solve for the fact that portfolios, as they move towards retirement, should be increasing their lower risk asset allocation. This also introduces the argument that younger investors, with bond like human capital (practically everyone bar stock brokers and investment professionals) should invest pretty much exclusively in equities. While the theoretical framework is different, its premise, that younger investors should be in equities and gradually increase the allocation to a full low risk allocation as you near retirement, is no different from traditional theory that states the longer time of the investor the greater the allocation you can have to equities. It is odd that the two theories intersect, at least on this point. So traditional MVO structures that recommend a mix of equities and bonds have themselves become invalidated by the theory of human capital. So, whether or not you believe in MPT, modern MPT does provide an equity risk override assumption for long term investors.
As a point of interest, I had developed dedicated, integrated asset liability modelling and management systems in the very early 1990s and had put much of my early thinking in this area to pen in the first UK Institute of Financial Planning investment course during circal1992/1993.