The share market is driven by stories and during reporting season, the most popular stories are those about earnings. Whilst there are plenty of possible narratives attached to each earnings release, in this article the aspect I want to focus on is – how believable are the reported earnings?
In a mathematical sense, the believability of earnings would be captured using a probabilistic frame to determine their statistical significance. In my all time favourite article on this subject, Bloomberg columnist Matt Levine describes this idea in relation to Bank of America’s quarterly earnings:
“A bank’s earnings are a quantum event; they are entirely probabilistic, and the answer you get depends on who’s doing the observing. You make some guesses with some degree of statistical likelihood, and then you apply one of a half-dozen accounting regimes to the guesses, and you get a number, and then you’re like, ooh, look at this number, it’s so numeric.
But it’s not a thing in the world. It’s just the output of applying your rules to your inputs. But your inputs are fictions, and your rules are fictions, so it’s very silly indeed to treat your output as a fact.”
This lovely quote neatly captures the two ideas behind the statistical significance of earnings – how well do we know the true economics of what the business is doing and how good a job do we do of representing this with a reported accounting number? (In the Equity Toolkit, we analyse these ideas under our Financial Frame as Earnings Quality.) In this article I hope to use today’s earnings release from Genworth Australia ($GMA.AU) to highlight some of these ideas. As an insurer of bank earnings, Genworth’s earnings move beyond basic quantum mechanics into the realm of string theory.
Genworth provides lenders mortgage insurance. In return for a premium paid by a mortgagee, it agrees to reimburse the bank if the mortgagee cannot make good on their loan. It goes without saying that this is a more risky proposition than mortgage lending itself. Genworth receive on average 1.65% of the mortgage as a premium, paid upfront. In contrast, the banks net interest margin is >2% and they receive that for every year of the mortgage. In addition, we can be almost certain that Genworth get the priviledge of doing this for only the riskier portion of the banks new mortgages.
In working out how much profit they make from this process, Genworth must make three large estimates:
- As the premium is paid upfront but applies for the life of the loan, how much of the premium should be recognised as income (net earned premium) in each accounting period?
- What percentage of the premium should be set aside to cover losses where mortgagees default on their loan?
- What is the income generated on the insurance float?
For any company, one of these estimates on its own would be enough to make us cautious on the reliability of the end output. In the case of Genworth, the reliability is further clouded by the nature of the underlying insurance contract and their balance sheet structure. To understand why this is the case, let’s consider the process of estimating mortgage losses – by far the most important of the above estimates.
The Problem with Housing Statistics
All insurance contracts are written using actuarial assesment of probabilities. There is naturally a risk in this process, but in many cases, the probability distribution of outcomes is reasonably well understood. This is not the case with housing loans. Compare the distributions of losses on Australian Home Building Insurance with an international selection of those on housing loans: