Supply Side Mortgage Financing

On the recommendation of Mike Konczal, I checked out Adam Levitin’s account of the housing (with Susan Wachter). It’s a pretty long paper with a single core idea: to figure out what happened with the financial crisis, look at prices and quantities. Levitin observes that the yields (ie, interest rates above other interest-bearing assets) on mortgage-backed securities fell dramatically during the subprime boom, while the quantity of such products grew dramatically. The lesson to draw from that is that a rising supply of mortgage finance must be the chief culprit for the crisis, not demand. If the demand for mortgages was rising (say, from federal housing policy), the price of risk in mortgage markets would steadily rise. The fact that they instead fell points to the role of increasing supply in lowering the risk premia on mortgages. Levitin attributes this falling cost to mispriced mortgage securities.

Here’s a graph from a Fed paper by Diana Hancock and Wayne Passmore I'll get to in a bit that illustrates this pretty well:

If you compare the pre-boom era "Normal" with the subprime boom, it’s clear that mortgage yields fell dramatically relative to Treasury rates (ignore the crisis era stuff).

This is a clever idea, but attributing all of that fall in interest rates to mispricing seems a pretty bad assertion. There are many reasons why yields could fall. As I just blogged, large foreign inflows of capital could lead to a lower price of risk, though this scenario isn’t mentioned in the paper. Though Levitin dismisses monetary policy quickly, Rajan has argued that periods of absolutely low interest rates might generate a pattern of increased leverage and yield-seeking that would describe this data equally well.

But what I found most troubling was the use of simple market yields. It’s true that a fall in mortgage yields could imply a lower risk premium, all else equal. But when pricing a mortgage, a lot more goes in than just credit risk. Borrowers of mortgages typically have the ability to pay back the mortgage in full — for instance, by refinancing their mortgage to a lower rate. This is bad from the point of view of an investor, since borrowers are most eager to "prepay" their mortgage once interest rates are low. So, at exactly the point when interest rates fall (and you have fewer good investment options), mortgage borrowers give you all this cash. To deal with this, when you lend money in the mortgage market, you receive a prepayment premium for taking on that risk, and that goes into pricing a mortgage. The factors determining prepayment risk, as well as other risk determinants, changed dramatically during the subprime boom era, so it's not clear whether a falling risk premia explains everything.

Fortunately, the Fed paper from above does break this out. While this can get very complex, one simple check is to compare the MBS yield over Treasury (graphed above) to the option-adjusted spread (OAS) over Treasury (graphed below).

The OAS here is designed to pick up the portion of the mortgage yield attributable to the prepayment risk. You see that the OAS spread went down quite a bit during the subprime boom years as well. Overall, the MBS yield over Treasury (which Levitin interprets only as credit risk) fell by ~.7%. Yet ~.55% of that fall is attributable not to changes in risk preferences, but instead due to changes in the premia for carrying prepayment risk.

And the fault for that goes exactly to the Fed/global investors. The Federal Reserve promoted a climate of low interest rates, while global investors too pushed interest rates down. In response, carrying the prepayment risk of mortgages became lower (as borrowers didn’t have much room to refinance at substantially lower rates), and mortgages became more attractive as an asset class. The net change in mortgage yields due factors other than the prepayment risk should be something like 15 basis points; and some of that may also be attributable to other changes in the financing climate (the authors claim a sizable fall in the roll-over risk in this period associated with having to refinance debt holdings. This, too, can be attributed to interest rates that were persistently low). Though the authors of the Fed paper don’t seem to break this out — it’s very possible that the credit risk for mortgages actually went up during the subprime years.

Not only that, but the stock of borrowers grew quickly in response to these shifts in interest rates, suggesting that homeowners in this period were highly responsive to the price of mortgages. It’s possible that factors like land use regulations, a bubble mentality, lower downpayments, etc. may have been responsible for that, and that responsiveness is as important as the initial shock.

To be sure, the fact that the risk premium seems to have remained roughly constant even as worse borrowers took out mortgages may suggest that the risk premium may not have been high as it should have been. And I don’t understand this field nearly as well as I should, so please check out both papers and tell me where I’m getting things wrong. But I call this as another win for the camp blaming the Fed/foreign investors.

Edit: Adam Levitin has responded in the comments. He makes the good point that the Fed paper only looked at Agency MBS; while the pattern may well be different for private-label MBS, many of which were given out to loans on an adjustable-rate mortgage.