Fannie Mae and Freddie Mac are really really big. This can mess with your intuitions. So is this a big number or a small number?
Fannie Mae and Freddie Mac could require another $190 billion in government support under a worst-case economic scenario, according to stress test results made public Wednesday by the firms' federal regulator.
I mean, it is a big number. Look at it: $190,000,000,000, it has so many digits. And yet. This table from the stress test report puts it in a kind of context:
So, yes, Fannie and Freddie would need $190 billion from the government in a "severely adverse scenario" in which housing prices decline by 25 percent. But that number:
- is less than 4 percent of the size of Fannie and Freddie;
- is driven mostly ($105.5 billion worth) by the purely non-cash accounting effect of marking down Fannie and Freddie's deferred tax assets -- they'd only have 2 percent losses on their actual credit portfolio; and
- would still leave Fannie and Freddie with tens of billions more dollars on its government funding commitment.
And this is the highly stressed scenario. The Federal Housing Finance Agency's stress tests also include three other scenarios (positive, baseline and negative but not as bad as the Dodd-Frank scenario); in all of them, Fannie and Freddie would have large positive income and would not require any money from Treasury. In fact they'd give Treasury back tens of billions of dollars:
What does this tell you? Well, as you may be aware, there is endless wrangling over Fannie Mae and Freddie Mac in courts and Congress and everywhere else. These stress tests might usefully inform that wrangling.
First of all, the reason that Fannie and Freddie would need $190 billion from the government is that they have essentially no capital at all right now, so $195.8 billion of losses translate into $190 billion of government support:
"These results...are not surprising given the company's limited capital," said Kelli Parsons, a Fannie Mae spokeswoman. "Fannie Mae is not permitted to retain capital to withstand a sudden, unexpected economic shock of the magnitude required by the stress test."
And the reason Fannie and Freddie have no capital is that that's what Treasury wanted: The controversial 2012 third amendment to the Fannie/Freddie bailout required Fannie and Freddie to give all of their profits to the government. So Treasury gets all of Fannie and Freddie's profits in exchange for bearing all of the risk of their losses.
In 2013, Treasury took out $130 billion of profits. So a worst-case scenario of $190 billion of losses in another crisis doesn't seem so bad; Treasury is doing very well off Fannie and Freddie in a non-crisis, and with big risks come big rewards. Or, put another way: If Treasury allowed Fannie and Freddie to retain their profits just from 2013, they'd be almost well capitalized enough to withstand another financial crisis.
I suppose that might inform your thinking about the fairness of the third amendment. Before the government changed the rules, Fannie and Freddie were required to pay a 10 percent return on the government's investment, with the rest of the profits accruing to Fannie and Freddie. Afterward, the government took all of the profits. The ostensible reason for this is that the government was worried that Fannie and Freddie would not have enough income to pay the 10 percent dividends and would have to keep borrowing more, depleting Treasury's commitment to support them. The shareholders suing the government over the amendment think that that's just a pretext.
So it's worth noticing that, in all of the FHFA's scenarios above, Fannie and Freddie keep paying dividends to the government. Even in "Scenario 3," the most negative FHFA projection, they pay $36.3 billion of dividends over two years -- just about a 10 percent annual return on the government's senior preferred stock. So the FHFA's own projections suggest that Fannie and Freddie are unlikely to need to borrow from Treasury to pay their dividends.
But the more important question is not so much how much sympathy you should have for the investors suing over the past treatment of Fannie and Freddie -- I have a certain amount, but reasonable people disagree -- but rather what should be done with housing finance in the future. The current stalking horse seems to be the Johnson-Crapo bill, in which the government would reinsure conforming-type mortgages as long as private capital was on the hook for the first 10 percent of the losses.
Nobody seems to like that proposal, and these stress tests can give you a sense of why: 10 percent is too much. As we've discussed before, it would require raising $450 billion of private capital to insure $4.5 trillion of single-family mortgages, significantly raising the cost of mortgages. But the stress tests found that Fannie and Freddie's credit losses, over two years of 25 percent further declines in house prices, would be just 2 percent. The extra 8 percent is unnecessary. Or rather, some of it is probably useful -- you don't want to get down to zero capital in a crisis -- but it's still too big.
For comparison, big banks are required to have enough capital that a "severely adverse scenario" would still leave them with capital equal to about 3-4 percent of their mortgages. If you apply that rule to Fannie and Freddie -- which would have credit losses of about 2 percent in a "severely adverse scenario" -- then you get an appropriate capitalization of 5 or 6 percent. If you required private capital to take the first 5 percent of mortgage losses, with the government reinsuring the rest, then even in another 25 percent downturn the government would not be on the hook for any losses, and there'd still be a 3 percent private capital cushion left.
So you can read these stress tests as good news: They suggest that the government's risks from Fannie and Freddie aren't that big relative to the rewards, that Fannie and Freddie are in better shape than some people have claimed, and that a future mortgage market doesn't need gigantic amounts of private capital to function effectively. Yeah, in a future crisis Fannie and Freddie would lose $190 billion, but in the grand scheme of things that's just not very much money.
That table is from page 6 of the report, and shows the Dodd-Frank Act Stress Test results for Fannie and Freddie combined. Pages 13 and 14 have results broken out for Fannie and Freddie separately, which are not especially interesting. Bottom lines are:
According to the press release, the FHFA has been running its own scenario projections since 2010; the Dodd-Frank stress test is new this year. Next year the FHFA ones will be phased out, leaving only the much more pessimistic Dodd-Frank stress test.
There's also the deferred tax asset stuff, which is not that informative, but very comical. Basically, Fannie and Freddie have large tax losses that they can use to shield future income from taxes. If you expect them to have lots of future income, that tax shield has a present value of about $105.5 billion. If you expect them to have no future income, it has a present value of about $0. Much of the recent history of Fannie and Freddie consists of people changing their mind about these DTAs: They were written down to zero in 2008, when Fannie and Freddie went into conservatorship, and were written back up in 2013 as conditions improved and Fannie/Freddie turned back to making money. That write-up allowed Fannie and Freddie to borrow tens of billions of dollars to give back to Treasury as accounting profits. In another crisis they would ... be written down to zero again? Or not? None can say, so FHFA just ran the numbers both ways. Either there'd be a $105.5 billion accounting loss, or there wouldn't be. Nothing would be economically different between the two scenarios, but the math would change by a hundred billion or so. Okay!
That's $130 billion of senior preferred dividends: actual cash out to Treasury, not just accounting profit. The breakdown is $82.5 billion from Fannie (page F-6 here) and $47.6 billion from Freddie (page 172 here).
Err, they'd have $130 billion of the $84 or $190 billion they'd need, depending on what you do with the deferred tax asset. Obviously you don't want to end up with zero capital even in a crisis, but you get the idea.
Really like 9.6 percent. There's $117 billion of Fannie "senior preferred stock" (page F-3) and $72 billion at Freddie (page 171). So 10 percent for two years would be $37.9 billion, so in a moderately stressed scenario there'd be a $1.6 billion dividend shortfall. Pocket change!
Of course, that's in 2014, and the amendment happened in 2012. Which makes this chart interesting:
The amendment happened in August 2012. In October 2012, based on June 2012 financial statements, FHFA concluded that even in an adverse scenario Fannie and Freddie would need to draw only about $22 billion from Treasury, dramatically less than its previous estimate.
Ha! Let's unpack that. Mortgages are risk-weighted at 50 percent for risk-based capital. Stressed capital requirements range from 4 percent (leverage ratio -- not a risk-weighted measure) to 8 percent (total risk-based capital). So a bank stress test is essentially, assume you lose a bunch of money, do you still have capital of at least 4 percent of your assets, and at least 4.5 to 8 percent (depending how you count) of your risk-weighted assets? So, for a mortgage portfolio, that works out to at least 2.25 percent (common equity tier 1) to 4 percent (leverage, total risk-based capital) of your credit exposure. I say "3-4 percent" in the text for simplicity, and because just, like, common equity tier 1 is probably not a fair comparison.
I mean, statistically. If you have lots of smaller pools of private capital, instead of two giant aggregates like Fannie and Freddie, I suppose some of them could idiosyncratically have more than 2 percent credit losses.
Also, I mean, who knows if the stress tests are right, etc. etc. etc., as Bank of America so vividly reminded us. Here is a good Steve Randy Waldman post from 2010, titled "Capital can't be measured," which points out that "capital positions reported by 'large complex financial institutions' are so difficult to compute that the confidence interval surrounding those estimates is greater than 100%." That may be less true for monoline mortgage thingies than for big banks, but only somewhat less true.
To contact the author on this story:
Matt Levine at firstname.lastname@example.org
To contact the editor on this story:
Tobin Harshaw at email@example.com