Cheaper Jumbo Loans Just Aren't That Weird

For the past few months the interest rates on jumbo mortgages -- ones too big to be guaranteed by Fannie Mae and Freddie Mac -- have been creeping closer and closer to the rates on conventional Fannie/Freddie-guaranteed mortgages, and yesterday they finally celebrated victory.

Hi! Welcome to my new blog at Bloomberg View, where I will be writing about the business, deals, personalities, scandals and delights of the financial industry. I am, as the little box to your right suggests, a former investment banker and corporate lawyer, and in this space I hope to explain a little of the complexities of the financial world, and convey how much fun they can be. We'll aim for a mix of short and long pieces depending on what's in the news -- this first offering is, admittedly, on the long side -- as well as linking to interesting stuff from elsewhere.

Also, there will be footnotes, we are fond of footnotes here. For now I am using asterisks; in the future they will even look good and be hyperlinked and stuff. So be excited for that.

Here we go: For the past few months the interest rates on jumbo mortgages -- ones too big to be guaranteed by Fannie Mae and Freddie Mac -- have been creeping closer and closer to the rates on conventional Fannie/Freddie-guaranteed mortgages, and yesterday they finally celebrated victory, with jumbo rates officially below conforming rates. From the Wall Street Journal:

In the past two weeks, the relationship has flipped, a combination of interest-rate volatility, government policy and banks flush with cash that are enjoying lower funding costs, making jumbo mortgages an attractive investment for them.
The average 30-year fixed-rate conforming mortgage was at 4.73% last week, according the Mortgage Bankers Association, compared with 4.71% for the average jumbo 30-year fixed-rate mortgage.

People enjoy being puzzled by this conundrum -- here are Matt Yglesias and Felix Salmon -- which is a little weird. Lending money to rich people to buy houses really should be a better business than lending money to poorer people to buy houses, on a variety of grounds, mostly related to the rich people having more money and nicer houses. So it's not that puzzling on first principles that big mortgages on big houses might have lower rates than small mortgages on small houses.

But we don't live in a world of first principles and the actual puzzle comes from the fact that conforming loans have no credit risk and jumbo loans have credit risk, because the conforming loans are guaranteed by the government, or rather by Fannie Mae and Freddie Mac, which are basically withered arms of the government,* [see first footnote below] so the banks don't have to worry about those borrowers defaulting.

But this is not quite right: The guarantee doesn't eliminate the credit risk of the conforming loans; it just shifts it to Fannie and Freddie. For a fee. And so if jumbo credit risk is more "expensive" than the guarantee fee -- that is, if expected losses have a higher present value than the fee -- then jumbo loans should be more expensive than conforming loans guaranteed by Fannie/Freddie, and vice versa.**

There's no a priori reason that the Fannie/Freddie guarantee fees should be lower than the credit risk of jumbo loans. The guarantee fee is of course the price of getting Fannie and Freddie to take on the credit risk, and if they were rational and competitive economic actors that fee would be roughly equal to the expected value of that credit risk. If jumbo loans had as much credit risk as conforming loans, then the guarantee fee for the conforming loans should roughly equal to the credit risk component of the jumbo loans, and so jumbos and conforming loans should have around the same rates. If jumbos were less risky -- if rich people were much better at paying off mortgages than less rich people -- then jumbos would of course cost less.

But of course, the whole reason for Fannie and Freddie's existence has long been to subsidize conforming mortgages, not to be rational and competitive economic actors, so the guarantee fee should be underpriced.*** But now that the government has announced a vague halfhearted goal of ending Fannie and Freddie's existence, its fantasy scenario would be to get rid of the subsidy -- by letting the guarantee fee rise to market levels -- without destroying the mortgage market.

One thing that's worth saying is: That kind of seems to be working! The jumbo-conventional convergence is, among other things, evidence of the subsidy's disappearance, as banks are becoming indifferent between taking mortgage credit risk on themselves and paying Fannie/Freddie to get rid of it. Bloomberg News last week had an article about Fannie's new risk-sharing bonds that cited a Barclays analyst report saying that "Fannie Mae and Freddie Mac's current guarantee fees are 'almost priced appropriately,' " so go team. That report**** is fascinating; here is what you might consider the money chart:

Source: Barclays.

This suggests that the total credit-risk-ish portion of a conforming loan held on a bank's books is about 94 basis points running, while the equivalent for a conforming loan guaranteed by Fannie or Freddie is 86 basis points. There is much to ponder here including that (1) Barclays expects average conventional mortgage losses to run about 4 basis points a year, which seems lowish?***** and (2) the "credit risk" component seems to be more a function of bank capital requirements -- how much capital bank regulators require banks to hold against the credit risk of unguaranteed loans -- than it is a function of banks' own risk assessment.

So the reason that jumbos are getting cheaper than conforming mortgages is a market segmentation reason: Banks are pricing the credit risk of jumbo mortgages, while the quasi-government is pricing conforming credit risk. And Fannie and Freddie are getting more risk-averse and expensive, for political reasons, just as banks are getting a bit more risk-welcoming and cheaper, for profitability and short-memory reasons.

There's another market segmentation issue here too. The nice thing about the market for (1) jumbo mortgages and (2) Fannie/Freddie mortgage-backed securities is that they have largely overlapping buyers: banks. So the question is less "why are banks more gung-ho on jumbo mortgages than creepy hedge funds are on agency mortgage-backed securities?" than it is "why are banks more gung-ho on jumbos than the same banks are on agency mortgage-backed securities?"

Here's one possible answer. As interest rates go up, the market value of existing 30-year fixed-rate mortgages should go down. Interest rates have been going up recently (though not today!). If you are a bank, a thought you might have in your banky brain is that you should be positioning yourself for a rising interest rate environment. On very first principles this is hard, because the business of a bank is to borrow short-term and lend long-term, so rising interest rates cause all of your long-term loans to lose value without causing an equivalent drop in the value of your short-term borrowings.

There is a solution to this problem, which is to ignore it, and that is precisely how banking works: banks lend money long-term and then, through the miracle of accounting, if those loans lose value because interest rates go up, the banks keep them on their books at their initial value. The market value of those loans is, for important accounting and regulatory and profitability purposes, irrelevant.

This is so core to banking -- so fundamental to the ability to borrow short-term and lend long-term -- that most of the time no one thinks about it. On the other hand, people do tend to get exercised about things like banks doing things, such as buying and trading securities, that are not so core to banking. Also they tend to get exercised about banks hiding losses, which is a thing these days.

Possibly related is the fact that banks that hold "available for sale" securities -- gambling with your money, etc., and particularly including agency mortgage-backed securities****** -- do have to show the changes in value of those securities in their financial statements. Those changes don't flow through net income, but they do affect capital levels: if those securities drop in value, the capital ratios of the banks that hold them are reduced. This happened last quarter as rates rose, leading to billions of dollars of phantom-ish losses on agency mortgage-backed securities at various banks this year.

Even ignoring credit risk: If you were a bank in a rising-ish interest rate environment and you could choose between a mortgage that loses value and requires you to raise more capital as interest rates rise, and an identical mortgage that loses value but you can ignore it and not increase capital as interest rates rise, which would you choose? That seems to me to be an adequate explanation for jumbo loans' attractiveness all on its own.

* They are. I mean, people will tell you that they're not -- especially people who own Fannie and Freddie stock! -- but they're lying.

** Incidentally there's at least potentially a third category of conforming-sized loans that are not guaranteed by Fannie and Freddie -- where the bank keeps the non-jumbo credit risk instead of paying the fee. But this category is pretty underpopulated; here's an amazing bit from the Journal story:

"I've had situations where I've told clients, 'You don't need to borrow within the [conforming] limit. I can get you a lower rate if you borrow a little more,' " said Rolan Shnayder, director of new-development lending at H.O.M.E. Mortgage Bankers in New York.

Well but just. There is an arbitrage here, like, borrow more and then just hand some of it right back, Rolan, give me a call, you can refinance my mortgage, we'll get along great.

*** Which still doesn't prove that jumbos should cost more. Like the credit risk of conforming mortgages might be 100bps, the guarantee fee might be a healthily subsidized 50bps, and the credit risk of jumbos might be 25bps -- still cheaper than conforming. This does not seem to fit the facts though.

**** "How high can g-fees go?," Barclays Securitised Products Research, 23 August 2013, not online sadly.

***** For comparison, Fannie Mae's conventional single-family loan default rates and loss severities for 2007, 2008, and 2009 were 0.32% & 11%, 0.59% & 26%, and 1.07% & 37%, respectively (Table 43 here). So the high one-year loss in 2009 was around 40bps (1.07% x 37%), or 10 times Barclays' number. On the other hand in 2007, sort of a mixed-bag year for mortgages I guess, Fannie had conforming credit losses of around 3.5bps, or just inside Barclays' number. So it's not nuts as a going-forward guess. It's not pricing in another 2008-2009, though, is the thing.

****** Take Wells Fargo, a bank cited by the Journal as giving out cheap jumbo mortgages. Wells has some $323bn of residential mortgages on its books as "loans," and some $110.5bn of Fannie/Freddie mortgage-backed securities on its books as "securities available for sale."

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

    To contact the author on this story:
    Matthew S Levine at

    Before it's here, it's on the Bloomberg Terminal.