The recent collapse of Silicon Valley Bank has brought mortgage-backed securities back into the public consciousness for the first time since the financial crisis. This is unfortunate, because the crisis taught us all a lot about one variant of the product: subprime mortgages packaged into other securities, some of which were investment-grade and some of which were not, with some of the non-investment-grade ones then packaged further into new securities. And that was all very exciting.

But the mortgage-back securities involved in Silicon Valley Bank, and other banks in headlines of the recent past or perhaps the near future, are a completely different animal. US financial regulations have spent a century-plus trying to deal with three interrelated facts:

1. Banks are a great way to aggregate lots of capital.

2. The single greatest demand for capital comes from mortgage origination. (Equities are a bigger market, but mortgage origination is an order of magnitude higher than annual IPO volume.)

3. While banks are natural participants in the mortgage-origination process, they're not the ideal holders of mortgages, because banks fund their balance sheets with demand deposits, and mortgages are hard to liquidate on demand.

The agency paper market exists to partly solve this problem. The basic idea is to remove two of the complicating features of mortgages: the risk that the borrower won't pay the mortgage back, and the fact that a single mortgage (or even a portfolio of mortgages) is an incredibly inconvenient thing to sell.

It works like this: quasi-government agencies (Fannie Mae and Freddie Mac) or an actual government agency (Ginnie Mae) buy up residential mortgages that conform to certain criteria. Then those agencies create financial products that are tied to the payback of the mortgages—but insured against defaults—and sell them as mortgage-backed securities (MBS) on the agency paper market.

That system partly solves the liquidity problem, since MBSs are easier to value and sell, and it solves the credit risk problem by making it somebody else's problem.

All that’s left is interest rate risk. (For a quick primer on the basic bond math of duration and convexity, see this older post.) Agency mortgage-backed securities are backed by long-term fixed-income streams of cash flow, and like any fixed-income product, this makes them sensitive to interest rates. But there's a twist! The US mortgage system, notoriously, allows borrowers to refinance if rates decline.

From a lender's perspective, this is an incredibly annoying feature: normally, if you buy a 30-year bond yielding 3.78% today, when rates go up 50 basis points to 4.28%, the value of the bond would drop by 8.5%, and if rates go down 50 basis points to 3.28%, the value of the bond would rise by 9.3%.¹ The present value of a 30-year MBS has roughly the same reaction to higher rates, but what it does when rates go down is much weirder: it basically turns into cash as the mortgages that backed it get paid off, mostly to be refinanced at newly-attractive rates.

So an MBS buyer is taking one-way risk on rates: if rates go up, the value of the MBS declines. If rates go down, the MBS doesn't get to tag along. Naturally, MBS buyers demand higher yields in exchange for this risk. How much higher is a tricky question: people have devoted their entire careers to increasingly complex models of how various households will refinance, and it's a moving target: the rise of increasingly sophisticated mortgage lenders has actually increased people's propensity to refinance mortgages when it saves them money. But fundamentally, the yield premium on agency MBS is the income from selling a call option. Specifically, the mortgage lender is giving the borrower the option to pay back the loan at any time.

There are many times when market participants are tempted to think that the cash they get from selling options is a form of "income", but it's not. Selling an option is selling insurance, and insurance companies definitely do not treat the premium as free money.

Unfortunately, there's a temptation, especially among companies aiming to hit some kind of target return on assets, to run their finger down the list until they find the asset that hits the desired rate of return, and then back their way into a thesis as to why it's a great opportunity right at this moment. Among institutional fixed-income investors, there are times to prefer treasury bonds to MBS, and times to prefer MBS to treasury bonds; there are times to lend for the short term and times to get the most duration you can. But if you're just looking for an opportunity that has minimum credit risk and maximum return, long-dated MBS will be at the top of the list.

But they're at the top for a reason: buying them means making one bet on interest rates and another bet on interest rate volatility. And the flip side of their higher immediate return is that in the circumstances in which they're worth more, the borrower refinances, and in the circumstances where they're worth less, the price goes down. There are circumstances when this bet is worth making, which is fortunate because Americans still want to take out lots of mortgages and there needs to be some lender somewhere on the other side of the trade. But when mortgages are packaged, simplified, and stripped of many of their risks, they actually turn into more of a pure interest rate derivative, and that's a product that's dangerous to speculate in.

Read More in The Diff

The Diff has covered matters related to agency paper a few times, but it’s topical today because of the role it played in the collapse of Silicon Valley Bank.

• Here’s an older post ($) from 2021 that went deeper into how SVB operates.

• Then about a month ago we talked about SVB’s attitude towards risk ($), where I mentioned losses on their mortgage-backed security investments but noted that this wouldn’t necessarily lead to a bank run and that depositors would be safe if it did.

• And recently, since SVB has been a go-to topic: this post ($) was sent out as the collapse unfolded, and this one was sent out on Monday.

1. The gains are slightly higher than the losses because as interest rates drop, the duration of the bond rises, i.e. the present value is weighted increasingly to the future. So bonds have the nice property that they get more sensitive to rates as they perform better.

Help us learn more about you!

We'd like your feedback so we know what to cover, so if you haven't already, please fill out this one-minute form.