Technical correction on ABS housing finance data

An old co-conspirator of mine correctly pointed out something that I was not fully aware of regarding the ABS housing finance data. When banks lend to borrowers, the loan purpose of "purchasing" a property is very different to "refinancing" an existing loan where there is no purchase. It affects a range of things, including the valuation requirements (eg, there is greater valuation risk with a refinancing since there is no market price). The default probabilities can also be different. But more importantly, providing finance for the "purchase" of a property is mutually exclusive to supplying finance for a refinancing of an existing home that has already been purchased: the two can never be the same. The ABS acknowledged this when I spoke to them, and said the data they receive from the banks is separated out. Nevertheless, the ABS mixes the two together in its housing finance data under the label of "purchases" of dwellings. Coming from a lending/asset-backed finance background, the two loan purposes are so inherently different I never conceived that lending for the "purchase" of existing dwellings could be confused with the refinancing of explicitly old (ie, non-purchase) dwellings. In point of fact it actually is under the ABS system. I have therefore re-cut my chart from yesterday with and without the refinancing data. It does not change anything I said, as you would expect (refi's only represent about 35% of the total): financing for purchases has been recovering since March, and, furthermore, the average growth rate has been running at a double-digit annualised rate. You can see the differences in charts below.