A reader of this blog contacted me to suggest that when I look at financial text, instead of trying to analyse stories, I could instead test for “readability”. The idea being that the more readable a text, the clearer the strategy then the better the company might perform. Or indeed the more jargon, obfuscation and evasive language, the worse a company might perform. He’s done initial analysis showing that great performers like Berkshire Hathaway, Google and Amazon have much more readable communications than the likes of Enron, Tyco and Global Crossing, which all failed.
It’s a good suggestion and encouraging to get feedback on ideas for writing about. There are some simple tests for “readability” such as a Flesch score. Like many good innovations (such as SatNav, message encryption and the early internet), readability scores were developed by the US military. The Flesch score is relatively simple = 207 – 1 x (total words / total sentences) minus 85 x (total syllables / total words). A high score means “easy to read.”
“The cat sat on the mat.” scores 116.
“The Australian platypus is seemingly a hybrid of a mammal and reptilian creature.” scores 37.5 as it has 24 syllables and 13 words.
Softening bad news
To support the idea, I’ve also found research published by S&P Global Market Intelligence (by Frank Zhao) suggesting that after their results are disappointing, management use more complicated language on conference calls. This is because management are trying to “soften” bad news by using obfuscating language.
This agrees with my own experience, disappointing trading statements often use the phrase “broadly inline” which translates into simple English as “profit warning”. The logic goes that if trading was inline, or better than expected, companies would spell this out clearly. When trading is disappointing, they don’t want to use the word “disappointing” so they just say “broadly inline”. Which in effect means “disappointing”. Market makers and professional investors almost always notice, and the share price falls anyway. Management are not fooling anyone; or perhaps only themselves, if they pay a financial PR company lots of their shareholders money to puff up the presentation of disappointing results.
That’s a simple example, and there is a much simpler way of testing for it (CTRL + F) than readability scores.
So instead I thought I’d try something more ambitious. I wanted to see if readability scores might contain a signal for speculative, innovative companies with little to judge between them. That is: a good readability score might provide a valuable signal where historic numbers were of little use. So I went back to 2013-14 and analysed the strategic commentary from 3 different graphene companies Admission Documents. Graphene, in case you didn’t know, is a new material with lots of wonderful properties and commercial potential – but the specific market opportunities are still unclear.
Despite appearing similar (loss making, not much financial history to go on) there was a huge range of outcomes. The first company has increased in value 12x, the second has fallen in value by 2/3 (ie lost around 66% of its value) and the third fell by 90%. Like the proverbial butterfly flapping its wing’s in Brazil diverting a hurricane in Texas, small differences in initial conditions can compound to have a big effect.
Difficult to read
But like many innovations, my readability score experiment didn’t work. All the companies were scored as “difficult to read”. I tried different readability scores like Gunning Fogg, and also text from different parts of the document. But really the companies had readability scores that were too similar, particularly when compared to the wide divergence in share price performance.
Even when an experiment doesn’t work – you learn something. In this case “readability” is probably too a crude measure, compared to analysing a story. For instance, two undeniably great stories are barely readable: “Trainspotting” by Irvine Welsh (written in urban Scotts slang) and “A Clockwork Orange” by Anthony Burgess (written in Nadsat, half Russian half English). The language is difficult, but the stories are great.
Reading the graphene companies documents with my own eyes, Versarien by far the best performing stock does seem to have a better story, about using porous metal technology, that can be used in medical devices or to dissipate heat away from computers Central Processing Units (CPUs). Haydale on the other hand talks about “functionalisation” of graphene to deliver “a tailored customer focussed solution” – whatever that is. The third company: Applied Grahene Materials, just talks vaguely about becoming a “value-added partner, rather than simply becoming a commodity producer of graphene”.
Easy, but not simple
In conclusion: readability is easy to test, but maybe not that useful for what I was trying to do (analyse a sample of 3 graphene companies’ Admission Documents). Perhaps a much larger sample would show readability scores could be useful, but I don’t think there were many more graphene companies listed at the same time.
For me it’s reassuring that at least there are some things that computers/Natural Language Processing still can’t do well: evaluating young, unproven businesses which want to IPO. There are just too many uncertainties, so experience and human intuition still have a big role to play. Possibly there are more sophisticated machine learning approaches that could work? Or maybe this is a challenge that is simply “undecide-able” –neither an expert human nor a machine can identify successful IPO’s in advance?
As a general observation: a more idiosyncratic style communication style, which avoids clichés and buzzwords, could be a signal that management are prepared to ignore the crowd, think for themselves and exploit unique opportunities (Burford Capital being an example). Alternatively companies that feel like they have to sound upbeat all the time, often do so because they know that they’ll have to return to investors and ask for more money in future. Whereas a tone which clearly communicates challenges as well as opportunities, might also signal a business with high returns on capital, where management are unlikely to need to ask for more money (Games Workshop being an example) so they communicate more realistically.
Ironically, despite having the clearest strategy of my sample, Versarien later changed course. But at least management had clear ideas about what they wanted to do, and were able to spot better opportunities and adjust accordingly. If you have a nebulous strategy, you don’t even know when you are drifting off course.
I feel like it’s an idea worth exploring, even if for now the conclusions are disappointing: it may make this blog less readable, but hopefully more interesting.
Photo by Aaron Burden on Unsplash