Forecast Fallacies, Tesla and Base Rates
How wrong analysts can be, and usually are, tends to be overlooked
Late last Friday evening ARK Invest published its latest Tesla model with the stupendous assumption that their new price target was $3000 for 2025, and even a bear case was $1500, double current share price levels. The modelling behind this, run on Monte Carlo simulations, has been torn apart on Twitter by Christopher Bloomstran with regard to its insurance assumptions, and FT Alphaville called it a “gish gallop” (or just a clever marketing ruse).
What I wanted to focus on however was the revenue outputs in the model where the CFA credited analyst has $31b in 2020 rising to $700b in 2025, that’s a CAGR of 117%! One of the things that always staggers me is the way analysts forecast revenue growth without any context to the base rate. The reason base rates are critical is that they help provide a statistical grounding on which to make decisions using historical precedent. The problem that we all have as financial analysts is that we tend to hugely over-estimate our level of competence without an eye on the reality. This excerpt from a study in 2016 which tracked global earnings forecasts from 2002 to 2014 concluded that on average analysts were 25% over-optimistic (with a healthy positive skew).Dunning-Kruger eat your heart out (more on that below).
Michael Mauboussin is one of the leaders in the area of base rates and this old Knowledge Project podcast is well worth diving back into. His Base Rate Book published (in 2016) while at Credit Suisse is downloadable online and contains a number of useful charts and tables like the one below, which shows that analysts tend to be far too narrow in their expectations — usually anchored to 5% and 10%.
Thinking back to the earlier Tesla assumptions we can also look up the base rates for what proportion of companies grow at certain rates with sales >$25b (the Tesla bucket). Unsurprisingly those that CAGR at >45% over 5 years are 0%. This is not to say it is completely impossible but when making assumptions and using these to underpin valuations we need to be aware of how likely the scenario is.
Across Europe at the moment we can see that analysts’ forecasts for Sales CAGR out to 2023, from 2019, (so avoiding the distortion of Covid from a low base) have some positive correlation with the historical sales CAGR delivered from 2016 to 2019, although outliers abound.
In the current era of buy and hold, and supra-normal returns from a select few equities, it is tempting to be drawn into the comfort that what has happened before will be repeated. With this in mind I thought it might be instructive to end with an excerpt from Adam Grant’s new book, Think Again, which I recently purchased and am thoroughly enjoying reading. In my view we are always amateurs at investing with so much to learn…
“It’s when we progress from novice to amateur that we become over confident. A bit of knowledge can be a dangerous thing. In too many domains of our lives, we never gain enough expertise to question our opinions or discover what we don’t know. We have just enough information to feel self-assured about making pronouncements and passing judgment, failing to realize that we’ve climbed to the top of Mount Stupid without making it over to the other side.”
Historical vs Forecast sales CAGR in European Stocks (>€1bln mkt cap)
