Stressed out but surviving

Philip Bayley
Standard & Poor's released a report last week that outlines the results of stress tests undertaken on 19 sub-prime and non-conforming RMBS transactions comprising 118 tranches of rated notes. While the report largely discusses the results of the stress testing in aggregate, the results for each transaction and rated tranche are also presented, making this the most comprehensive analysis of this type undertaken in the domestic market.    
 
Two scenarios were considered for the stress testing: Scenario 1 looked at the maximum level of arrears each underlying portfolio could withstand before the rating on each tranche is affected and specifically, the impact of an arrears level of 32.5 per cent in each portfolio; Scenario 2 considered the impact of a 10 per cent write down on the original value of security properties.

S&P found that under Scenario 1 a majority of rated tranches, down to the 'B' rating level, could withstand arrears of 32 per cent, as could about 90 per cent of the 'AAA' rated tranches, without being downgraded. Under Scenario 2, 77 per cent of tranches and 90 per cent or more of the 'AAA' and 'AA+' tranches would not be adversely impacted by a 10 per cent decline in original property values.

A critical reading of the report raises some obvious questions. Why were Scenarios 1 & 2 considered in isolation when in reality they would likely be concurrent events? And how were the stress test levels set? It seems it would have been more instructive to link arrears to rises in unemployment and interest rates and declines in house prices to those seen in other markets.

As always there are modelling constraints but S&P also made some good points in response to these questions. The key point to be made here is that S&P did not consider a decline in the current market value of properties in the underlying asset pool because there is no way to benchmark this. The only information available is the value of the property when it entered the pool. So any analysis must consider the impact of the value of the property having been overstated at the time it was mortgaged.

Given this, a combination of the two scenarios would result in a double counting of the impact. In Scenario 1 an increase in arrears at the rate contemplated flows through the rating model to result in higher default levels and a correspondingly higher loss given default.

Similarly, under scenario 2 a deliberate over-valuation of security property from day one means the original loan to valuation ratios were higher than originally considered. Adjustment of the model for higher LVRs again results in higher assumed default rates and loss given default.

As for the selection of the stress test levels, arrears of 30 per cent have already been observed in some transactions but the average peaked at 17 per cent in January this year. Moreover, the nature of Australia's subprime and nonconforming mortgage market is different from that of other countries, making benchmarking difficult.

The same can be said of Australia's residential property market and assuming a 10 per cent over valuation of security property from the start does not seem unreasonable.     
And finally, it would be logical to link changes in arrears to changes in unemployment and interest rates and if anyone has such a model… In the meantime of course, we may see how this plays out in reality.