Evaluating probabilities subjectively
Uneducated guesses by the overly-educated
On Twitter, where pretend investors present their pretend theses to a pretend audience, a well-followed account laid out his thesis for owning Russian equities throughout the shuttering of the Russian stock market post-Ukraine conflict.
Despite drawing down substantially from their peak, the thesis was summarized by “90% chance Russia opens back up and everything gets back to normal, 10% chance that it goes to zero” which is exactly the sort of lazy thinking that you’d expect from an equity investor that is dabbling their feet in investing internationally. Below is the correct response to this analysis from a good analyst:
Credit investors and MBA students are familiar with loss-given-default analysis as well as decision tree analysis. Both of these practices seek subjective probabilities with “informed” analysis, based on past and current data. But at the end of the day you’re still applying your opinion to set of data and coming up with a number that’s (hopefully) well-informed.
What is under-discussed is the subjective practice of using and weighting historical data. A credit investor might look at default probabilities based on several different financial metrics and hone in on one data point as their “fulcrum” metric (“the company has EBITDA coverage of 6x, this implies a default probability of less than 5% over a 10 year period” where another credit investor might say “the company has less than 6 months worth of cash on hand, this implies a default probability of 20% over a 10 year period”). A market is made between these two prices.
Fundamental equity investors must understand that regardless of how well-detailed their analysis of a business might be, they are counting on a quantitative practice that can be backtested and analyzed with historical data. Too many equity investment theses are predicated on “this is a unique scenario, there’s no close comparison or analogous business model”. One historical analog in these situations is worth one thousand “yeah, but my judgement is [x]”.
A business might be cheap trading at 10x forward estimates, and maybe historically it has traded at 15x. Maybe the market is pessimistic on the business model and maybe the forward earnings estimates are misperceived. Show that this works- show that the market does this from time-to-time with historical analysis. Show that you can buy temporarily cheap businesses and show that they can re-rate, and use an exhaustive data set. Most businesses that re-rate lower will never re-rate higher, and deservedly so. Show how many businesses rated from 15x to 10x and then back again, and how likely that is on a probabilistic basis. Inform your qualitative assessment of the business with historical occurrences and you might be shocked to learn that the market gets things right more often than not.
Russia is a good example of this- I hope to show more examples in future writings. There are historical analogs out there, you must do the work and explore them. Start with businesses that have 100-year histories. Show how Coke’s multiple fluctuates over time and why, show how the insurance industry’s multiples expand and contract. Only then can assigning subjective probabilities be done with confidence.