Prepayment Risk and CPR

A regular bond pays you back on a schedule you can write down in advance. A mortgage-backed security doesn't, because every homeowner in the underlying pool has the right to pay off their loan early — by refinancing, selling their house, or just paying extra — and none of them ask the bondholder's permission. That unpredictable early return of principal is prepayment risk, and CPR is how the market quantifies its speed.

What drives the speed

Prepayments accelerate mainly when it becomes attractive for homeowners to refinance — interest rates fall well below the pool's average coupon — and also, independently, through ordinary "housing turnover" (people moving, selling, and paying off their old mortgage as part of the sale) which continues even when rates aren't especially low. Because refinancing incentive depends on where rates are relative to the pool's coupon, prepayment speeds move sharply with the rate cycle, and a pool's CPR early in its life (before most owners are "in the money" to refinance) looks very different from later.

Worked example

A pool has $200 million in remaining principal and a CPR of 12%. The approximate dollar amount expected to prepay over the year:

$$200{,}000{,}000 \times 0.12 = 24{,}000{,}000$$

Roughly $24 million of principal returns to investors early, on top of scheduled amortization, spread unevenly across the year as SMM (the monthly equivalent of CPR) compounds. If rates then fall sharply and CPR jumps to 30% because refinancing becomes attractive, the expected early paydown roughly rises to $200{,}000{,}000 \times 0.30 = 60{,}000{,}000$ — investors get their principal back much faster, right when reinvesting it at a comparably high rate is no longer possible, because rates just fell.

What this means in practice

That last point is the core pain of holding MBS: prepayments speed up exactly when rates fall (principal comes back and can only be reinvested at the new, lower rate) and slow down exactly when rates rise (principal that investors want back gets locked in longer at the old, now-below-market coupon). This asymmetry is why MBS have negative convexity, and why pricing them correctly requires modeling prepayment behavior, not just discounting a fixed schedule.