Risk, the data and analysis, and the context

by Graham Andersen

Starting with the 2008 financial crisis there have been a number of global events that seem to defy the predictions of the established players but with the benefit of hindsight analysis now seem obvious and predictable. Clearly the Big Short, Brexit and Trump are such events but why were these events occurring so misread beforehand? Events that have huge consequences for the risk of a multitude financial assets, one could be forgiven for thinking that best and most reliable analysis available would be trained on the likely outcomes. But no, even the humble bookies got it wildly wrong.
Even in these days of big data and fintechs, risk would seem to be widely misunderstood and what data there is is subject to inadequate analysis. My contention is that evidence will show that the conventional view of the risk of the Australian mortgage market is at odds with a more disciplined and unbiased view.
A particular dubious analytic indicator is for analytic firms to put probabilities on single events like the election of President-elect Trump. The outcome of the election is a single event set in today’s circumstances which will occur or not and never be repeated. There may be an expectation of a particular result based on the data at hand but to put a probability on that event occurring is very dubious science. Its actually akin to putting a probability on the reliability of the data which we know is not complete so we don’t know what we don’t know.
So proposition one on risk is that besides being misunderstood it generally lacks proper analysis on inadequate data.
My next proposition on risk is that the perceived importance of the risk by us humans is based on the setting even though an equivalent or greater risk may be taken every day.
The tragedy at Dream World would seem to support this point. Although terrible, it clearly is a rare event. Society does not react the way they do to the Dream World tragedy as they do to car accident tragedies which unfortunately occur almost every day. Why the difference? It’s the context and control. On a theme park ride, you are totally putting your safety in the hands of complete strangers with no control whatsoever (same for aircraft), whilst we also need to travel in or drive cars which carries considerable risk. But we have some control over what’s happening or trust the person in control. Does this make a big difference in the perception of risk and the reaction to a tragic risk event? As flawed humans, I think it does.
Applying propositions 1 and 2 to the Australian mortgage market may allow us to have a different view of the risk than is conventional wisdom.
The availability of data on mortgages in Australia to do detailed analysis to understand where the risk is and the extent of that risk is not available in detail. Reporting standards and data transparency have not been improved in any meaningful way for quite some time. During that time, Australia’s macro settings have changed considerably, notably with rising household debt levels, flattening incomes, rising house prices and rising offshore debt levels.
So with inadequate data during a time of increasing stress in the macro-economic settings, risk analysis also has not kept up. Most internal models, regulator analysis and rating agency analysis looks backward on borrower performance (ie good) when macro settings would indicate rising risk. Even house prices rising above incomes increases risk and disguises performance. The idea of determining the probability of default of a single mortgage based on past performance is as much pseudo-science as putting a probability on a Trump win.
Shouldn’t alarm bells be ringing? Well no, because proposition 2 on the risk context kicks in. Regulators and the government are driving the bus and think they are in control, so whilst they understand that there is risk in there somewhere, its only going to occur somewhere else. Borrowers, overstretched or not, have been driving their own debt cars and have been rewarded with capital gains and so have learnt that more debt is good and government and regulators will do their best to support borrowers from defaulting. So in that context and even with the macro settings and external influences which could smash the party, the conventional analysis concludes that our banks are strong and the system is safe.
Any good analyst who can take themselves out of the context and properly analyse the available Australian mortgage and macro data in a deductive manner may not predict wholesale defaults but they’d certainly shine a spotlight on the risk of that occurring.

Leave a Reply