Most liens on a seasoned second pool carry an AVM value with no interior inspection

Consider a small pool of seasoned second liens. Twenty two loans, unpaid balances from 18k to 96k, weighted average coupon 9.4, seasoning of 14 to 31 months, seller asking 88 of UPB. Collateral documentation is where a tape like that gets hard. Say eight files carry a full interior appraisal from origination. Fourteen carry an AVM output with a confidence score, and on nine of those there is a property condition report from a third party inspector who never went inside. That mix is what the market is doing now, with AVMs and exterior condition reports carrying a large share of home equity originations, and the originator's logic holds up. Underwriting a 34k second against a 410k house at 62 combined LTV, a 500 dollar appraisal is real money against the margin. The buyer's problem is a different problem from the originator's. A buyer prices at a point in time and cannot reorder those valuations without contacting borrowers, and an AVM confidence score describes model dispersion across that zip code rather than anything about that specific house. Two of the fourteen may sit in a county with low transaction counts, where the model is leaning on sales a long way out. The two ways through are a flat haircut applied to the AVM valued loans with the whole tape priced lower, or carving out the four weakest files and asking the seller to hold them back. Holdbacks on a 22 loan pool usually kill the trade. Ordering drive bys pre close on somebody else's borrowers carries its own problems. The federal quality control rule on AVMs applies to the users of those models, so it says something about the originator's process and nothing about any individual value. The open question is what a defensible haircut looks like when the dispersion in question is concentrated in two of the fourteen rather than spread evenly across all of them.