The Machine Rebellion

By D. Thomakos

Delighted that my last post yesterday was well received in LinkedIn. Many thanks for Jonathan Regenstein for introducing the post to Alexander Fleiss at Rebellion Research AI Asset Management.  Alexander has written a post in his website about the methodology I presented in mine, and I gratefully acknowledge the support and comments. You can read Alexander's post here "What is the formula for predicting GDP".

Alexander raises two concerns about the methodology: the impact of external factors such as economic crises and the reliability of the proposed measure of my post. Very briefly (and this is a discussion that we can follow through in later posts) this is what I can suggest on Alexander's comments:

1. The Covid-19 crisis is reflected well in the movement of both series, and so is the 2008 financial crisis. The reason, I would venture to say, for this is the fact that capacity utilization is inversely related to unemployment and higher unemployment might lead to adverse effects on disposable income which in turn will have adverse effects in mortgage repayments. All-in-all the suggested measure should move with the economic cycle as its underlying variables are procyclical.

2. The reliability question is more subtle, because we have to define where reliability refers to. If it refers to financial trading, then the measure is certainly reliable (based on my backtesting) fully after 2015 - before it has tracked the index well but started generating excess returns well into the 2000s. If reliability refers to some other concept or economic measurement I would have to check with the data!

In any event, it is always great to have a discussion on the prospects and limitations of all research and I welcomed Alexander to write a guest post here at Prognostikon. Stay tuned!