There are lots of things to worry about in the world and somewhere on the list is the fact that, while yields on agency mortgage-backed securities are really really really low, the rate you’ll pay for a new mortgage is only really low, so a couple of reallys have fallen off a truck somewhere. This worry isn’t at the top of my personal list – my mortgage rate is low enough I guess? – but it seems to make many other people’s list for two intersecting reasons. First, if the primary desire of Fed policy is to get people to buy houses, be rich, etc., and if its primary mechanism for doing so is buying MBS, then the inefficiency in transforming that mechanism into that desire is rather macroeconomically important and bad. Second, if money is coming out of the Fed and not ending up in homeowners’ pockets, that leaves only so many pockets it could be ending up in, and it is easy enough to observe that big banks (1) sit between the Fed and the homeowners and (2) have lots of pockets. So you can see how it might be fun to worry about money going to big multipocketed banks, because if it does, you get to be mad at them.

Anyway the New York Fed is doing a conference on it today; here’s the background paper and it’s really interesting; I recommend it, particularly if, like me, you have a hazy understanding of agency mortgage securitization. Everything in this space is predicated on somewhat fake math but their math is less fake than the simple spread math, which basically assumes that banks make a profit of:¹

• Annual Profit = Mortgage Rate – MBS Yield

By that math, as William Dudley points out, the spread was 30-50bps in the ’90s and early 2000s, but rose to 150bps in September and is around 120bps now. The Fed’s paper, on the other hand, walks through the actual securitization process to get cash flows into and out of the mortgage lender, and computes its profit (technically, profit plus non-interest-y costs like underwriting and hedging) as roughly:

• Up-Front Profit = Sale Price Into MBS – Origination Price + 4 x (Mortgage Rate – MBS Coupon – GSE Guarantee Fee

Why 4? I dunno it’s in the paper.² Anyway by this measure here is what has happened in the world:

So historically banks captured about 1 to 3% of the notional amount of each conforming loan – which they used to pay for underwriting and stuff; in ’05-’08 that number was mostly south of 2%, and for the last year of quantitative easing via MBS-buying it’s crept closer to 5%, where it is now. So where banks used to make $ 10,000 on a $ 500,000 mortgage, now they make $ 25,000.³

The Fed’s authors look at a bunch of potential explanations for this and shrug at most of them, ultimately “conclud[ing] that the growth in the market value of originated mortgages remains something of a puzzle,” which I guess is what the conference is for.

Still a couple of the explanations suggest that the banks should be looking to their pockets. A frequently suggested explanation for rising spreads is that, in the olden days, Fannie and Freddie never made banks take back bad mortgages, and now they’ve alerted banks that they will, so banks need to keep more spread to self-insure against that. But the Fed’s authors note that new loans are originated with much better documentation and credit scores than the old loans, and so the putback likelihood is modest – they calculate 19bps on pretty rule-of-thumb-y assumptions. But they add:

That said, perhaps the “true” cost of putback risk comes from originators trying to avoid putbacks in the first place by spending significantly more resources on underwriting new loans or on defending against putback claims. Furthermore, the remaining risk on older vintages is larger than on new loans, and many active lenders are also still subject to lawsuits on non-agency loans made during the boom. It is unclear, however, why these claims on vintage loans should affect the cost of new originations.

Is it? In a world where I could start up a mortgage lender and make new loans at my cost, untainted by previous putback claims, it would be unclear: I would just do that, and compete down the spreads charged by the legacy lenders with piles of old-loan putbacks to work through. In a world where Bank of America (1) makes all new mortgages and (2) also has a lot of bad mortgages in its closet, it’s perhaps less unclear why BofA would use one set of mortgages to subsidize the other, or how they’d get away with it. In a different context – talking about capacity constraints as driving these higher profits – the authors note:

New originators can enter the market, but entry requires federal and/or state licensing and approval from Fannie Mae, Freddie Mac and Ginnie Mae to fully participate in the origination process. To the extent that training now may take longer than in the past, or that approval delays for new entrants are longer (as anecdotally reported), the speed of capacity expansion may have declined compared with earlier episodes.

Similarly, the Fed’s authors point to interesting evidence that much of the spread differential comes specifically from refinancing loans, not just federally subsidized HARP refinancings but also “GSE streamline refinancing programs,” which, like HARP,

when done through the institution that currently services the loan, relieve[] the lender from representation and warranties relating to the borrower’s creditworthiness and home value, while a different-servicer refinancing requires the new loan to be fully underwritten. As a consequence, for borrowers eligible for a streamlined refinancing, there is an advantage to staying with the same servicer/lender, as doing so will reduce the documentation the borrower will need to submit. This in turn again creates some pricing power for the current servicer (although likely less so than for high-LTV loans).

So if you’re an individual with a loan, you’re not stuck stuck with the lender you have, but you’re kind of sticky. And they’ll charge you for the difficulty you’d have going elsewhere. And on a macro level, the Fed and the GSEs aren’t stuck stuck with the lenders they have to originate mortgages and transmit monetary policy, but they’re kind of sticky. Sticky enough that those lenders can still charge for the cost of making up for their past mistakes.

The Rising Gap Between Primary and Secondary Mortgage Rates [NY Fed]
The Spread between Primary and Secondary Mortgage Rates: Recent Trends and Prospects [NY Fed]
Opening Remarks at The Spread between Primary and Secondary Mortgage Rates: Recent Trends and Prospects Workshop [NY Fed / William Dudley]

1. This is vague for several reasons, including how you capitalize that annual spread and also:

While the primary-secondary spread is a closely tracked series, it is an imperfect measure of the pass-through between secondary market valuations and primary market borrowing costs for several reasons. First, the yield on any MBS is not directly observed, because the duration of the security is uncertain. The calculation of the yield is based on the MBS price and cash-flow projections from a prepayment model, itself using projections of conditioning variables (e.g. interest rates and house prices). For TBA contracts, the projected cash-flows and thus the yield also depend on the characteristics of the assumed cheapest-to-deliver pool. The resulting yield is thus subject to model misspecification. Second, the primary-secondary spread typically relies on the theoretical construct of a “current coupon MBS.” While at each point in time TBA contracts with different coupons trade, the current coupon is a hypothetical security that is trading at par and is meant to be representative of the yield on newly issued securities.

2. Nah, it’s an industry-standard assumed multiple for capitalizing excess servicing rights (4x), with base servicing rights a bit higher at 5x, although both numbers have declined in recent years. The guarantee fees now average 40bps. Further explanation is in the paper; which has an appendix that does the math:

3. That’s just like gross profit, not counting the fact that someone has to review your paperwork and stuff. The paper cites somewhat dubious statistics that “total loan production expenses averaged $ 4,717 per loan in 2008 and $ 5,128 per loan in.””

Dealbreaker