It is quite remarkable that in Australia today we have neither (a) a standardised debt serviceability benchmark that is reported by all lenders or (b) any serviceability metrics for investors in, for example, RMBS portfolios beyond LVRs. The first time the regulator or an RMBS investor really learns about increases in risk is after the event via default rates (when it is already too late).
In a better world, we would have access to a common loan serviceability measure which, in combination with LVRs, one could use to cross-sectionally compare the quality of every single loan in the country.
Importantly, this could be reported both at the origination of the loan and over its life (the use of automated property valuation models, which are gaining currency and are far more scientific than human valuers, also allows lenders to dynamically (and cost-effectively) recompute their LVRs every week as opposed to relying on the original LVR from when the loan was advanced, which was typically several years ago).
Here is an extract of what Luci had to say today:
"[T]he events of the crisis emphasised that financial stability analysis is not the same as most of the macroeconomic analysis done in advance of monetary policy decisions. In particular, for financial stability analysis, it isn't enough to look at the aggregate or average data. We need to track the tails of the distribution, where the worst risks are. So where my colleagues in Economic Group are mainly focussed on what is likely to happen, the team and I in Financial Stability Department tend to focus on the less-likely things that might go wrong. We often find ourselves looking at the same data, but in different ways.
For an example of a case where aggregate data can be outright misleading, consider this graph (Graph 1). It shows how much the ratios of housing prices and household debt to household income increased, in percentage points of household income, over the boom period of 2000 to 2006. Comparisons like this drew some observers to conclude that the United States was less likely to have had a bubble then than some of the other countries shown on the graph. Of course, that doesn't ring true given how things turned out. Understanding why requires looking at much more disaggregated data. These would have shown that lending standards had eased much further in the US mortgage market than in those in other countries. More US households were taking out loans with very small deposits, or for larger amounts than they could reasonably pay back. But because most borrowers were still taking out conventional ‘prime’ loans, this wasn't obvious from aggregated data.
Here is an example of both disaggregated data and the kind of new data we find useful. This graph shows a slightly more detailed version of some APRA data we have already presented in the last few Financial Stability Reviews (Graph 2). They are the flow of newly originated owner-occupier housing loans, disaggregated by the loan-to-valuation ratio at origination. These data are quite new – they are only available since the start of 2008, though planning for that collection started several years before that. These data provide an insight into loan-to-valuation ratios on new loans. In particular, they help us understand whether an increase in the average loan-to-valuation ratio on new loans has occurred because there are more high-ratio loans, or fewer low-ratio ones. It is not hard to see that the implications for risk in the household sector would be different depending on which was true."
