What Happens When Bad Luck Gets Recorded

This chapter tests the history-to-price mechanism on a medical expense. The specific question is whether a credit system should treat the expense itself as evidence about a borrower, or treat only the borrower’s response—paying, delaying, or becoming delinquent—as informative. That distinction supports a practical policy judgment: hiding information can protect people from being priced for bad luck, but it can also change repayment incentives and redistribute the cost of hidden debt.

The shock is not the response

The Philadelphia Fed paper adds a medical expense shock to its credit model. At the beginning of the period, a person may receive no medical expense or a medical expense of a fixed amount. The shock is exogenous and independent of the person’s hidden type. A high-patience and a low-patience person are equally likely to be hit by it. In that narrow sense, the expense is bad luck rather than evidence that one person is more likely to default than another.

After receiving the expense, the person can pay it immediately or become delinquent. Delinquency moves the medical cost into the next period as medical debt. The choice is subject to the same kind of unobserved decision noise used elsewhere in the model, so the response does not perfectly reveal type. A low-patience person is nevertheless more likely to prefer current resources over a future payment and therefore more likely to become delinquent, all else equal.

The separation between event and response is the central fairness question. If a medical expense is independent of type, observing that the expense occurred should not lower a lender’s assessment of the borrower. Observing delinquency can be informative under the model because it reveals how the borrower responds when repayment is costly. The two facts can coexist: the shock is not a signal, while the response to the shock is a noisy signal.

The response also changes the borrower’s financial state even before a lender updates a score. Carrying the medical expense forward reduces effective net assets in the next period. If the borrower also seeks an ordinary loan, the same nominal loan size represents greater leverage when medical debt is present. A lender can therefore change the price for two related reasons: the borrower’s effective repayment burden is larger, and the delinquency choice may have lowered the inferred probability of repayment.

The model treats the medical record as the expense and the response together. When records are observable, the lender can condition on whether there was a medical shock and whether the borrower became delinquent. When records are unobservable, neither medical event is available for conditioning. The borrower still experiences the expense and still carries debt if delinquent, but the lender sees only the other observable parts of the credit history.

This creates three economically different visible states. With no shock, the borrower has no medical liability. With a shock and immediate payment, the borrower has experienced bad luck but does not carry medical debt forward.