Yesterday the RBA released what might turn out to be one of the most influential papers ever on Australian mortgages, mortgage stress, arrears and defaults.
Titled, “Mortgage-related Financial Difficulties: Evidence from Australian Micro-level Data” the authors Matthew Read, Chris Stewart and Gianni La Cava investigated, “the factors associated with the incidence of mortgage-related financial difficulties in Australia.”
The most exciting part of what might otherwise seem a seriously dry topic was the combination of two data sets which look at raw “loan-level data on residential mortgages from two Australian banks”, together with the more behavioural data from the HILDA survey.
The loan level data that the researchers used came from MARQ Services* and enabled analysis at a granular level hitherto unavailable but which will become the new standard once the RBA’s CLF reporting framework comes into place in July 2015.
The authors said, and the RBA understands given its new requirements, that the more granular MARQ Services data “allows us to analyse the factors associated with 90+ day arrears, which are a precursor to default and possible loan losses for lenders.”
Not to put too fine a point on it, the revolution in data transparency which powers this paper is that as Australia moves towards more granular data first, to the RBA under its CLF and then, at an individual ADI level under APRA’s CPS 220, it allows a better understanding of the true risks associated with mortgages in the Australian financial system.
Indeed the authors are somewhat bullish of their own paper saying, “to the best of our knowledge, this is the first paper to use micro-level data to quantitatively analyse mortgage-related financial difficulties in Australia.”
But here is why it is so important, at a time when there is much discussion around the systemic risk of Australia’s housing boom.
The authors say that:
This paper provides a useful input into the analysis of housing finance in Australia for a few reasons. First, the micro-level analysis provides a new ‘bottom-up’ assessment of the risks associated with housing lending. Second, the information could be used as an input into stress tests of the housing lending exposures of authorised deposit-taking institutions and mortgage insurers. Third, it could be useful in informing decisions about the design of the prudential policy framework. More broadly, the information could help to inform decisions about the level of risk that lenders, their investors and regulators are willing to accept.
Indeed it does. What they find is that high LVR loans are more risky, that LVR’s at origination matter and Low Doc loans are riskier still. They also find that the higher the proportion of debt servicing to income, the higher the probability of a missed payment. And once you’ve missed one you have an increased chance of missing more.
This paper also shows prepayments (the amount to which borrowers are ahead of schedule) are a solid insulation against risk.
So at a time when the OECD is recommending Australia raise rates to curb a housing boom and thus lower systemic risk in the economy and banking system, the use of big data suggests their recommendation of the use of a blunt instrument to deal house price appreciation a blow is naive and hamfisted.
Big data is telling the RBA, APRA and Australia’s banks exactly what they need to know and from July 2015, when the RBA’s new data requirements kick in, they will have access to this data each and every month.
That’s a big data-driven revolution in risk monitoring and with it management.
Now it is just up to “supervisors carefully monitoring changes in lending standards, as well as the importance of borrowers exercising prudence when taking on mortgage debt.”
No need for a rate hike to cool housing.
Here is a link to the RBA Research paper.
*Disclaimer: Greg McKenna does some consulting work for Morgij Analytics the parent of MARQ Services.