Last month I reviewed the excellent book, Superforecasting by Dan Gardner and Philip E. Tetlock. Now while we can’t predict the future, we certainly can create a model for the probability of an event occurring and this book uses the Bayesian probability theory.
I have briefly touched on this but want to go into more detail on the concept. It is actually a complex formula initially proposed by Thomas Bayes, an 18th-century mathematician and theologian. I’m not delving into the formula directly, rather the concept.
First, we need to accept that nothing is binary in that we never (well, never say never, so very seldom) have a possible future that is 0% or 100% certain. We live in the murky grey bit in-between 0% and 100%.
To use the Bayesian theory, you first need to decide what you’re looking to assign a probability to and also define a time period.
I’m going to use the possible downgrade of our sovereign debt to junk status by Standard and Poor’s (S&P) and whether it will happen by the end of 2016 or not.
You need to start with a base percentage rate and I’d use the following: What percentage of countries rated by S&P, that were one notch above junk status, got downgraded to junk within 12 months?
This data is not easy to find, but lots of Googling and digging suggests that pretty much two out of three countries one notch above junk status got downgraded to junk within one year.
So we start with a 66% probability of being downgraded. Now we add other factors and adjust this initial base rate accordingly.
Finance minister Pravin Gordhan’s Budget in February was good, but was it good enough? It certainly ticked most boxes but I have two concerns.
Firstly, it lacked any big gestures, such as a VAT increase or total privatisation of SAA. Secondly, it really just carried on from previous budgets and the mid-term budget policy statements.
Good, but how much does it move the dial? I would say it improved our odds of not being downgraded by 15%.
So now we’re on a 51% chance of a downgrade.
The big issue with the downgrade threat is GDP growth and here S&P expects our growth for the 2014 to 2017 period to average 2% a year. 2014 GDP growth was 1.5%, while 2015 was 1.3%, and nobody is expecting 2016 to be above 1%. So we’d need to knock it out of the park in 2017 to get a 2% average for the four years.
The bottom line is that we will miss the hopeful 2% growth for 2014 to 2017 and that adds to the downgrade risk. I’d say this is a biggie and will increase the odds of a downgrade by 20%.
So now we’re on a 71% chance of a downgrade.
You would continue with this process, adding and subtracting from the percentage as you look at other data points, ultimately coming to a conclusion of the percentage chance of the event happening.
There are two very important things to remember:
You need to keep adjusting the model; let’s say GDP does suddenly swing higher (maybe commodity prices move up swiftly), then you’d adjust the number accordingly.
Further, while this is frankly guessing at each chance and impact of each event, you can research more and refine the data. Then you can start to get to a fairly decent percentage chance, but watch out for personal biases.
We can use this theory for almost anything and, like other skills, you’ll get better the more you work with the Bayesian probability theory.
This article originally appeared in the 17 March 2016 edition of finweek. Buy and download the magazine here.