…the Sensex goes up 0.65% (on average) the next trading day! With the 2019 World Cup kickstarting in a few days, that’s awesome news for investors, right?
But what if I told you that actually the Sensex goes up only 0.16% the day after India wins a match - but an incredible 1.45% the day after India loses a World Cup match! Hmmm, what now? As an investor I seek returns but I’ve also been bleeding blue long-before I knew what stocks (or even money) was - so how does the rational adult in me reconcile this?
Well, simply by recognising that these statistics don’t mean anything.
Not all data is meaningful - and I feel that understanding this has become increasingly important in a world where data & data modelling have become the core of most industries.
Econometrics is the branch of economics that uses statistical/mathematical models to study & describe the relations between different economic systems. And one of the things we learn in econometrics (and also in statistics courses) is something called spurious correlations.
That’s the term used to define instances when two things/events appear to be related, but in reality are not. Like India playing a cricket World Cup match, and the markets going up by 0.65%. You know what’s even more incredible? The long-term trends between the Sensex index value and the runs scored by Sachin and Virat Kohli.
There is an 80% correlation between Sachin scoring runs and the Sensex index level over the course of his entire career. And Captain Kohli’s relationship with the markets since the start of his career is even more impressive, with a correlation of 94% with the Sensex!
Does the market really go up because Sachin or Kohli scored runs? Basic correlation thinks so - but the more probable answer is that both these unrelated data sets are influenced by different variables that have made them follow similar trends over time. There might be other issues too - is this the right data-set, is the sample large enough, etc.
You may have heard people say the phrase “correlation does not imply causation”.
This statistical mantra basically means that just because two things are related, it doesn’t mean that one thing leads to another. A causal relation between two events exists if the occurrence of the first leads to the other - the first event is the “cause” and the second is called the “effect”.
Let’s take another example - in the past 20-30 years, the price of cricket match tickets has increased as has the number of people attending cricket matches. Does this mean that an increase in ticket prices leads to an increase in the number of fans going to the stadium? Surely not.
However, the opposite might be possible - that is, an increase in the number of people going to cricket matches (which has increased due to population & popularity of cricket) might have directly caused an increase in the ticket prices of cricket matches.
One of the most common types and reasons for spurious correlations is (believe it or not) just mere coincidence
- if you look hard enough, there are plenty of coincidences & patterns in nature and life. Just look at this website
created by then Harvard Law student Tyler Vigen highlighting some of the weirdest coincidences in the U.S. economy!
Testing for causality is complicated but important - especially if one is building financial/mathematical models, to ensure that the model has real predictive power. Everyday investors don’t really need to learn this in order to manage their investments.
However, knowing that not all things that seem related necessarily are, is a powerful insight.
Especially in finance where there are thousands of variables impacting the market & many such relations being constantly thrown around. This will help you at least ask the right questions when such seemingly related events are mentioned.
Like, do election results really impact the long-term performance of Indian stock markets? What do you think?