Rethinking Modern Portfolio Theory: Embracing the Unpredictable Realities of the Market

In collaboration with Hervé Kias .

The Fallacy of Modern Portfolio Theory

The financial system is largely based on one theory: that of Modern Portfolio Theory (MPT) and Market Efficiency (EMT). They advocate the idea that markets are always "rational", that they perfectly reflect all known information and always produce, in some way, the "right" price, or that they can easily predict a value towards which the price of an asset will tend. All this leads us to treat economics as an exact science, like Newtonian physics, when in fact it is a human or social science. Indeed, the reality is that markets are prone to unpredictable behaviour, overreaction or apathy, irrational mania or panic. The fact that they reflect these behaviours does not imply supra-human wisdom. Markets are merely a mirror of human psychological failings: fear, greed, delusion and loss of contact with reality.

This fallacy was reinforced by the consequences of the financial crisis of 2008 which resulted in the world being flooded with liquidity by central banks. In this environment, managing assets and creating value were relatively easy. However, over the past 40 months, the world has changed: central banks are withdrawing this liquidity, war, inflation, societal issues defended by one side or the other, sometimes vehemently, and so on. These hard-to-predict shocks and the resulting violent market movements that follow, contribute to a kind of exhaustion among professionals and investors alike. In this environment, MPT still pushes to rely on estimates of expected returns, volatilities, and correlations to predict portfolio risk. This has given rise to the practice of "optimization", whereby the best possible allocations are chosen based on these estimates.

The Overconfidence of Optimization

This is overconfidence, a well-known behavioural bias. The world is far more unpredictable than the models predict, so future estimates of volatilities and correlations are bound to be wrong (more so as the time horizon of the prediction lengthens). Optimal solutions are therefore only optimal for a single state of the world or for a limited time horizon. Thus, they are generally not robust.

Investment professionals therefore need to move away from volatility or, more generally, from the concept of correlations, to make genuine risk management and investment decisions. It is undoubtedly necessary to use na?ve models, i.e., with few parameters and without excessive "intelligence". However, one must make sure that they are used in a state of the world where they have a good probability of creating value.

The Virtue of Humility

The famous author Antoine de Saint-Exupéry summed it up well: "Perfection is reached, not when there is nothing left to add, but when there is nothing left to take away". Today, many asset managers are accumulating data and betting on Deep/Machine Learning... but is this the right path? The search for information is tending towards infinity, and no one can put infinite resources into obtaining and extracting all the information needed to describe and predict the behaviour of financial markets.

It is therefore urgent to implement processes that consider the latest academic research on the subject, including non-linearity and the science of complexity. Social sciences with population logic, just like the natural sciences (physics, biology, etc.), incorporate concepts that accept uncertainty, in addition to randomness. While the latter is easy to model, especially in finance, using models derived from MPT, uncertainty is impossible to describe in the current state of knowledge. Many asset managers dismiss these uncertain markets environments as "epiphenomena". Today's world, with its repeated shocks, proves that uncertainty is now a new "normal" and not the exception. The humility of not knowing is becoming a virtue. Indeed, in certain market configurations, the tools and models derived from MPT have a high probability of failing. So, investment professionals have to admit their limitations.

Timing Models to Market Environments

Few investment funds have processes that automatically adapt to the likely denormalization of financial market variations due to investor behaviours and irrationality. The temptation to try to describe these excesses remains too strong, even though it is impossible to do so. A simple indicator that characterizes with a good probability, the state of the world we are in can help us to better pilot our models and selection. If market timing is difficult, investment models timing is much simpler. In some environments, a model may likely work, but in others, it may more likely malfunction. Having such an indicator allows value to be created, but above all, enables the ability to design strategies that have the merit of being resilient and more stable: they say what they do, and do what they say. Ultimately, what interests an investor is that the strategy behaves in a continuous, linear fashion, something that strategies based solely on MPT cannot claim.

Having a pool of products with this kind of processes would undoubtedly enrich multi-management.