Imagine a bank could tell, at the time of lending you money, exactly what the chances were of your paying it back. It is the banking equivalent of insurers trying to determine the chances of your making a claim. If the chance is too high, both banks and insurers will reject your business.
Well, both try to do exactly that. Banks have built enormously powerful models that suck in huge amounts of data to forecast loan defaults. Bad debt is an inevitable part of lending, a cost as sure to be incurred as any other. A good model allows a bank to fine tune its strategy, working out what types of lending would make for the most profitable business.
A certain form of lending may have a high expected default rate but that is fine if the profit margins justify it. A bank treats the forecast defaults as a cost to be taken as soon as the loan is written, so when the defaults arrive as expected, they have already been provided for.
The success of models is crucial for the stability of the whole financial system. When they get it wrong, disaster can follow.
The models are proprietary to each bank — they consider them a crucial competitive advantage. Every now and then they hit on a useful new bit of information to include. Analysts have found, for example, a strong correlation between the petrol price and bad debts in unsecured lending books — an increasing petrol price means you will have higher bad debts.
Such straight causal modelling does not capture the complex adaptive nature of the economy. If a bank increases the rate of foreclosure on defaulting home-loan borrowers, it will have an effect on their ability to service their other loans, including those to other banks. If you foreclose on vehicle owners you are often removing their means of getting to work and earning a salary.
There is no end to the data mining that can be done. Insurer Discovery, for instance, has found that people who score highly in its Vitality wellness programme also have lower claims.
But occasionally they go wrong. When loan books perform much worse than expected, the banks suddenly have to make exceptional additional provisions, hitting them as a once-off cost.
The worst such recent episode was in 2008 and 2009, when home-loan lending in South Africa was hit with a worse than expected economic environment and weakening house prices. The models seemed to all be based on perpetually rising house prices and a strong economy. The big four banks collectively lost billions when that scenario changed and large additional provisions had to be made.
A more recent example of models going wrong was when African Bank told the market in May it had got its expectations wrong in the unsecured lending space, having to make an additional R445m write-off.
The behaviour of other banks and the behaviour of regulators are very difficult to model. The current difficulties in the unsecured lending space have been caused by the rapid increase in lending undertaken by three of the big four banks — FirstRand, Standard Bank and Nedbank — such that while each individually could have modelled the risk appropriately, the simultaneous entry of competitors threw out the forecasts. Now, all the banks are pushing up provisions and slowing lending in light of the revealed behaviour of their competitors.
The last microloan crisis was caused by the state’s decision to axe lenders’ access to the public service payroll for deductions. This led to the collapse of Saambou and forced sale of BoE. With regulators now pondering sweeping regulatory change, there is a risk of what seem like small changes having a big effect. When the Treasury last year mooted the possibility of ending garnishee orders, for example, there was panic among certain banks that rely on them for collections.
When it comes to analysing banks and setting their policy environments, we should be sensitive to the health of their models. Making the world even more unpredictable is the surest route to another financial crisis.
But regulators can do the one thing that would have a positive effect on banks’ models: provide for better information that can be incorporated into them.
Better and quicker disclosure of lending behaviour would enable each bank to be more responsive to the behaviour of their competitors. The aim should be to remove chaos, not cause it.
BY STUART THEOBALD
STUART THEOBALD is a Rhodes University graduate
Article Source: Business Day