The Hitchhiker’s Guide to Investment Research, Part 5: The Two Razors

Everyone who read the Hitchhiker’s Guide to the Galaxy knows there are five parts in this trilogy of four. So, naturally, my Hitchhiker’s Guide to Investment Research has to end in part five as well. This time, I discuss how I differentiate between methods, theories, and approaches. It all boils down to two razors that incidentally not only help me in my research but also in everyday life.

The first razor is – you guessed it – Occam’s Razor. Named after the scholastic philosopher and friar William of Ockham (1287-1347), it simply states that when confronted with two or more theories of how the world works, the one that requires fewer assumptions is more likely to be true.

Applying this razor will prevent you from adopting overly complicated and overfitted models to forecast markets or the economy. In economics, central banks and academics have a tendency to build enormously complex models like DSGE models (Dynamic Stochastic General Equilibrium) of a Keynesian economy with sticky wages and prices. The hope is that by adding more and more variables and interactions between these variables, one can capture more and more nuances of how the economy and forecasts of inflation, growth, and other important variables will become more accurate. I admit I am not sure if the Bank of England still uses the COMPASS model for its forecasts, but it described the full model in 2013 including all the parameters and variables that flow into it. It is a DSGE model of a Keynesian economy with stocky wages and prices (nope, I didn’t make that one up) and if you are a masochist, I dare you to look through the appendix that explains all variables and equations. But if you want to have a glimpse of the model, take a look at Table 3 in the main note which shows the 52(!) variables and their distribution that are fitted in the model.

John von Neuman famously said, “with four parameters I can fit an elephant and with five I can make him wiggle his trunk”. And as this webpage shows, one can indeed fit an elephant with four parameters. The problem is that a better fit of past data often means you start overfitting the past and your forecasting performance is becoming worse.

The best demonstration of this problem in economics has been given by former Chief Economist of the Bank of England Andrew Haldane in his “The dog and the frisbee” speech (if you haven’t read it, do so now).

I have written an entire series of blog posts under the title “Dumb alpha” for the Enterprising Investor blog where I show how simple investment techniques often work much better than complex models. Since then, I have added a few more musings on the topic of overfitting data on my blog.

In my work, I genuinely distrust all complex models and prefer to use simple models with as few parameters as possible. Did I mention that in economics and finance, it is easy to be a busy fool?

Finally, Occam’s Razor also is very helpful outside of economics and finance. It will allow you to avoid falling trap of conspiracy theories, which have gained so much traction these days. Every conspiracy theory is identifiable by the very fact that if you try to enumerate how many things have to happen at the same time and how many people have to conspire to keep an action a secret, you very quickly come to a large number of individuals and entities. Meanwhile, the assumption that things are just happening by accident or because individuals follow their incentives requires zero or one assumption (i.e. that people follow their incentives). Which one do you think is more likely to be the true explanation? It just goes to show that while there is a 700-year-old tool available to distinguish truth from fiction, many people are still too uneducated to use it and fall prey to conspiracy theorists.

The second razor I use is the lesser-known Hitchens’ Razor. Named after Christopher Hitchens (1949-2011) it simply states that “what can be asserted without evidence can also be dismissed without evidence”. It helps me not get overly exhausted about people using theories in finance and economics that have little to no empirical evidence to support them. I have mentioned yesterday in Part 4 that people in finance and economics tend to formulate grand theories that then become gospel. Even in the face of numerous falsifications, they still pretend that extraordinary claims require extraordinary evidence – and then they continue to use a disproven theory. Both economics and finance are ripe with theories and models that have either only weak evidence in their favour (e.g. the notion that inflation expectations matter for future inflation or the CAPM) or theories that once worked but no longer have any supporting evidence in their favour as the world has changed (e.g. quantity theory of money or the value factor).

If somebody comes with a theory or a claim, the burden of proof is with the person who makes the claim. Once a claim has been proven to make useful forecasts in real life, you can accept the claim. But always look for falsifications. And pay attention to them. Strong opinions lightly held, not weak arguments strongly held.

Hitchens’ Razor is by the way the best tool to gain peace of mind outside of finance and economics. The world is full of trolls who want to “own the libs” or seek attention by making outlandish claims. By applying Hitchens’ Razor, you don’t fall into the trap of trying to convince them that you are right with the use of data or logic. Whenever I hear these trolls, I shrug my shoulder and move on. They have no evidence to support their claim so I can dismiss their claims without evidence as well.