When loan-level data is withheld, credit quality is still knowable. The underwriting policy as it changed over time, described by the people who wrote and applied it, predicts vintage performance well enough to test a seller's loss assumptions.
Why is loan-level data withheld?
Often for entirely legitimate reasons — data protection obligations, or a staged disclosure process where loan tape arrives late. It is not by itself a signal about the book.
It does mean the loss assumption has to be tested another way, and the fallback of accepting management's aggregate numbers is not a test — least of all in a diligence running to an IC date.
- 01
Reconstruct policy
Build the underwriting policy timeline from the people who wrote and applied it.
- 02
Interview collections
Establish how recoveries behave and where forbearance has been used.
- 03
Test vintages
Compare cohorts to separate policy effects from economic conditions.
- 04
Find the quiet cohort
Identify the vintage nobody volunteers to discuss and ask why.
What does underwriting policy reveal?
How the book was built. Policy changes — a loosened income multiple, a new channel, an expanded geography — explain why one vintage performs differently from another, and they are describable by the people who made them.
Reconstructing the policy timeline is often more informative than a tape would be, because it tells you why the numbers differ rather than only that they do.
Policy is also easier to reconstruct than it sounds, because the changes were decisions rather than drift. Someone approved a new channel, someone signed off a loosened multiple, and the people involved remember both the change and the argument around it. What is hard to reconstruct is the effect, which is why the policy timeline has to be read against whatever performance data does exist.
What do collections staff know?
How recoveries actually behave, which borrower segments cure and which do not, and where forbearance has been used to manage arrears. That is where an optimistic loss curve usually breaks.
They also know which vintages generated disproportionate work, which is a reliable proxy for underwriting quality and is not in any report.
Collections staff are also the easiest population to reach, which is worth knowing when a process is tight. They are numerous, they turn over regularly, and the questions being asked are about their own work rather than about anything confidential to a transaction.
| Loan tape diligence | Policy and people | |
|---|---|---|
| Reads | The loans themselves | How the loans came to exist |
| Strongest at | Concentration and exposure | Explaining why vintages differ |
| Blind to | Why policy changed when it did | Single large exposures |
How do you read vintage behavior?
By separating the effect of underwriting changes from the effect of the economy. Cohorts originated under the same policy in different conditions isolate the second; cohorts originated in the same conditions under different policies isolate the first.
Aggregate loss figures deliberately blend the two, which is why a seller's headline number can be accurate and uninformative at once — the same problem a quality of earnings exercise runs into on the revenue side.
What can this not tell you?
Concentration. You will not find a single large exposure through policy and people, and concentration is the risk that most often surprises a buyer after close — which is an argument for talking to the customers themselves.
Aggregate exposure data is the minimum to insist on, and it is usually available even where loan-level detail is not.
The same limitation applies to anything about the specific terms of individual facilities. Policy and people describe how a book was built and how it behaves in aggregate, and they will not tell you what a particular borrower negotiated. Where a small number of exposures carry the outcome, aggregate methods are the wrong instrument however well they are run.