The Two Slabs

Put a Ginnie Mae pool and a Fannie Mae pool side by side at the same coupon and the first thing a new analyst notices is that the coupons match. The second thing, a few prints later, is that nothing else does. In the sample month on this site, Ginnie speeds ran 2.1 CPR faster than Fannie at the same coupon. CPR is how fast borrowers prepay, annualised, and 2.1 of it is a real difference on a bond you paid a payup for. The gap is not a mystery. It has three sources, and each one leaves a mark in the public files.

Same coupon, different borrower

• ScoreCredit scores run lower.
• LTVLoan to value ratios run higher, often above 95 percent at origination for FHA.
• BalanceBalances run smaller.
• Down paymentDown payments are thinner, so the equity cushion that lets a conventional borrower take cash out is thinner too.

A Fannie pool is made of conventional loans. A Ginnie pool is made of government insured loans, mostly FHA and VA, with a smaller share from USDA. Those programmes exist to lend to borrowers the conventional market prices out, so the borrower base differs on every field the disclosure file carries. Credit scores run lower. Loan to value ratios run higher, often above 95 percent at origination for FHA. Balances run smaller. Down payments are thinner, so the equity cushion that lets a conventional borrower take cash out is thinner too.

Each of those cuts a different way. A lower score and a higher LTV make a rate and term refinance harder to qualify for in the conventional market, which slows things down. A smaller balance means less dollar saving per refinance, which slows things down again. So on the borrower base alone you might expect Ginnie to be slower. It often is not, which is where the other two sources come in.

The servicer is part of the speed

A Ginnie servicer may purchase a loan out of the pool once it is 90 days delinquent. To the investor in the pool that purchase looks exactly like a prepayment: the balance leaves the pool at par. So Ginnie speeds carry an involuntary component that has nothing to do with rates and everything to do with credit and with servicer behaviour. When delinquencies rise, those purchases rise with them, and a Ginnie pool can print fast in a month when nobody refinanced anything.

The servicer's identity matters on the voluntary side as well. Ginnie servicing is concentrated in nonbank servicers, and a nonbank servicer that also originates has every reason to recapture its own borrowers when rates move. The disclosure file names the servicer on every pool. It is one of the most informative columns in it and one of the least used.

The refinance channel

Two squared stacks, one taller. Sample photograph.

The third source is the channel a government borrower refinances through. The FHA Streamline and the VA Interest Rate Reduction Refinance Loan need no new appraisal and limited re-underwriting, so the frictions that slow a conventional borrower with a high LTV or a thin file are mostly absent. There are seasoning rules, but once they are met a government borrower who is in the money can be moved quickly and at scale by a servicer that wants the business. That is why Ginnie speeds can outrun Fannie speeds in a rally even though the Ginnie borrower base looks slower on paper.

So the honest summary of the three sources is: the borrower base pulls one way, the delinquent purchase channel and the streamline channel pull the other, and which wins depends on rates and on credit conditions in the month you are looking at. A 2.1 CPR gap is the net of three forces, not a property of Ginnie.