Close this search box.
Close this search box.

Does Causality Predict Markets Movements Or Is It A Flawed Overhyped Tool?

Does Causality Predict Markets Movements Or Is It A Flawed Overhyped Tool?

Artificial Intelligence & Machine Learning / Trading & Investing

The financial sector has bowed down to the seminal work of Harry Markowitz and colleagues!

Moreover, who in the mid-20th century introduced a new way of thinking about investment portfolios. One that emphasized diversification and the balancing of risk and reward. Their theory became a cornerstone of financial risk management. As time passed, however, this innovative approach transformed into a rigid axiom. Furthermore, today, its validity for contemporary finance is now becoming re-evaluated.

A central flaw in this traditional financial theory is its foundation on historical market data to forecast risk. The familiar warning that “past performance is not indicative of future results” underscores a deep flaw—not a mere academic oversight but one with tangible, sometimes disastrous, outcomes. Numerous financial institutions have foundered, having placed undue reliance on historical data that lacks context.

In response to the inherent risks, the concept of “causality” has become introduced into financial forecasting.

Causality, in its essence, should link occurrences in a verifiable way that allows for the anticipation of future events, something that retrospection can’t assure. Yet, efforts to apply causality often lapse back into the same old patterns of relying on historical market data, essentially circling back to the original problem.

Classical causal models, like the ones proposed by Judea Pearl, utilize diagrams to map the dependencies between various factors. These models rely heavily on conditional probabilities.

However, the methodology typically uses limited historical market data to establish these probabilities, a method riddled with shortcomings.

It’s critical to realize that the stock market is a minute part of the global economy’s complex web, susceptible to an array of unforeseen factors.

How can causality work if its looking at a mere fraction of the underlying data?

Any attempt to derive predictive causality needs to consider a vast array of external economic variables!

The challenge is that the totality of these external factors is too complex to be fully understood at any given moment.

Additionally, the global economy is not static; it’s a living, evolving entity that frequently births new trends and causal dynamics.

This continuous evolution of economic factors suggests that causality, when based on past data, may offer limited insight into future market behavior. The fluid nature of the economy, fueled by innovation and human creativity, perpetually generates new variables and shifts existing ones in ways that historical data cannot anticipate.

In conclusion, the financial markets’ dependency on causality derived from historical data is fundamentally flawed due to the dynamic, interconnected, and unpredictable nature of the global economy. Acknowledging this complexity and unpredictability is crucial in developing more robust financial prediction tools that can withstand the tests of time and uncertainty.

Yann LeCun – Facebook Director of Ai & Courant Prof. sits down with Alexander Fleiss

This graph demonstrates why causality as a prediction tool can be flawed when based on a limited sample set. The complete data set (light blue) shows a clear trend line (green), representing the true relationship between the independent and dependent variables. However, when we draw conclusions from a limited sample (red), the predicted trend line (orange) deviates significantly from the true relationship. This discrepancy illustrates the potential for error when using small samples to infer causality, which can lead to incorrect predictions about future events.

Does Causality Predict Markets Movements Or Is It A Flawed Overhyped Tool?