Miss America and Murder
This week, we are going to spend a little time discussing the issue of correlation vs. causation. Before your eyes glaze over in a fog of – say what now? – we are going to endeavor to make this both brief and hopefully mildly entertaining. We are prompted to go down this path because of a number of recent discussions we have had with both clients and market pundits regarding the dismal market events of 2022 (the worst year for a 60/40 investor in ~90-years) and the likely glide path for markets in 2023 given that many folks believe that a recession both in Canada and the U.S. is the base case.
Okay, before we dive into 2023, we would first like to recall a research piece written about a quarter century ago by Dr. David Leinweber, who was a professor at Caltech, which is an institution known for making lots of smart people. Dr. Leinweber in a piece entitled Stupid Data Miner Tricks: Overfitting the S&P 500 posited that the single best indicator for the glide path of the S&P 500 was (drumroll) butter production in Bangladesh. Now, your first reaction might be – I had no idea Bangladesh was known for butter. Once you got past this bit of knowledge enhancement, your next reaction might be – seriously?
Okay, let’s look at one of Dr. Leinweber’s charts and then comment:
Here, we are looking at the S&P 500 vs. the butter production in Bangladesh (as promised) from 1981 through 1993 and, as you can see, it’s a stellar fit. In fact, butter production explains about 75% of the moves in the S&P 500 over that period, so all you really had to do to predict where the stock market was going back in the 80s and 90s was to reach out to your contacts in Dhaka (the capitol of Bangladesh in case you were wondering) and get a gauge on where butter production was heading.
But Dr. Leinweber was not satisfied with 75% - he wanted better. So, after much research and we suppose conversations with experts in a variety of fields, he introduced an even better fit:
Here, Leinweber added U.S. cheese production and raised the predictive value to 95%. Obviously, butter and cheese are in the same family (we suppose), so maybe it was not all that surprising that when the two combined forces, they would become and even more powerful predictive tool. Now, not only would you want to talk to your contacts in Dhaka, but you would also want to pull in your friends in Wisconsin to get a sense of how U.S. cheese production was going. Big fire at the Gouda factory in Green Bay? Sell your stocks!
Still not satisfied (Caltech never is), Dr. Leinweber went back to the drawing board to see if he could get to the ever elusive 100% correlation. Sadly, while he got close, just as with the perfect 4-leaf clover, he was unable to achieve dairy perfection:
Not surprisingly, sheep played the pivotal role (we are guessing because butter and cheese come from sheep), bringing correlation to 99% or near perfection. Now, if you called Dhaka, spoke to Green Bay and then spoke to ranchers in both in Bangladesh and the U.S. to get a sense of what sheep pregnancies looked like, you had a very strong sense of what the stock market was likely to do the following year.
Okay, clearly Dr. Leinweber was kidding (he did have “stupid” in the title). While his data was all accurate, it goes to the issue of correlation vs. causation. Investors are constantly seeking explanations (causation) or predictive tools for why the market went up or down or whether it is likely to go up or down in the future, but often times even the best fitting explanations are meaningless in the grand scheme of things.
Which brings us back to 2022 and 2023. We have a simple rule when it comes to markets – U.S. recessions tend to be bad for stocks (worldwide), but they also tend to produce great long-term opportunities. Good businesses rarely go on sale, but when we run into the teeth of the U.S. recession, most things will tend to go on sale for a brief period of time, which allows one to add good businesses at distressed prices. Now, U.S. recessions do not cause stocks to go down, but rather the myriad of factors that lead to recessions – higher interest rates and borrowing costs, slowing demand, increasing layoffs, falling confidence and animal spirits, lower butter production (we jest) – that contribute varying pieces of the weakness that investments experience.
That said – while most years that have seen U.S. recessions have also seen stocks go lower, there are some exceptions – 2020 the most recent example as the Covid-induced recession saw a sharp selloff, but then a rapid recovery with stocks actually closing higher for the year. Thus, while we have our playbook and our playbook suggests that 2023 is likely to be challenging because of the varying factors that are likely to contribute to a U.S. (and Canadian recession), we are also cognizant of the fact that stocks often zig when we expect them to zag.
Okay, one last chart before we call it a day:
We have heard anecdotally that hot vapour murders are on the rise, so if you know any beauty pageant contestants – please warn them.