How Safe are the GSE's?

Freddie Mac's earnings release from last week created quite a buzz. It was initially viewed as an unmitigated positive, but upon further review, we all realized Freddie's accounting is too opaque to draw any reasonable conclusions.

I do feel that there have been some mischaracterizations of reality around the blogosphere, and I thought Accrued Interest could help shed a little light on the situation. So here is a little Q&A on what this development really means to real investors.

Q: Freddie Mac is actually insolvent, right? I heard they had negative net worth.

A: If you define insolvency as negative net worth on paper, then yes. I don't know to whom such a calculation is relevant. Its a quirky statistic that is emblematic of their recent woes. But it isn't relevant to investment valuation.

Q: But if they are insolvent, that would mean the tax payers might have to bail them out!

A: Tax payers will have to bail out the GSE's if it comes to that. I have no doubt about that. But it only will come to that if the GSE's cannot fulfill their function as liquidity providers to mortgage originators. Right now that isn't a problem, despite the negative net worth.

One could argue that the GSE's are only able to fulfill that liquidity function because of implicit government support. I think that's probably true. But if you believe that, then it wouldn't matter how much money the GSE's lost as long as the market believed in their government support.

Q: Freddie Mac's senior debt rating is still AAA, which is about as much bullshit as Ambac. How can these guys keep losing money quarter after quarter and retain that rating?

Actually, through the miracle of subordination, debt holders are probably relatively safe, even without government support. To be sure, there is no way either GSE would earn a AAA rating without the implied government support. But consider the debt/asset situation at Freddie Mac:

Total Assets: $786 billion (eliminating their deferred tax asset)
Total Senior Debt: $755 billion
Total Sub/Preferred Equity: $19 billion

So in order for senior liabilities to be greater than assets, the asset pool would have to decline by about 4%. Freddie Mac's credit loss was at a 12bps rate in Q1 and their forecast is 18bps in 2008 and 20-25bps in 2009. Take the company's estimates with whatever brand of Kosher Salt you like, but consider the odds of them being wrong by a factor of 16. (See note at the end of this post)

Q: You just want to be lied to, don't you? How can you trust any of their asset valuations anyway? I heard its all Level 3 assets!

A: They have $157 billion in Level 3 assets.

Q: Every one on my message board knows that Level 3 assets are toxic waste. That would more than make up your 4%!

A: First of all, the concept of Level 1, 2, and 3 assets stems from FAS 157, which is summarized here for those who like primary sources. The idea was to categorize the means by which assets have been valued by management. Level 3 assets are those that have been priced using "unobservable" inputs.

Q: Aha! Unobservable means mark-to-make-believe!

A: Part of requiring the Level 3 disclosure was to allow investors to consider how much they want to trust asset valuations based on models, especially in a market like this. So if you want to discount the valuation of Level 3 assets, the new disclosure allows you to do so.

Q: OK, so how much should I discount the assets? 100% or just 80%?

A: Unfortunately, there is some debate as to what constitutes an unobservable input. In Freddie Mac's case, they had classically valued their ABS portfolio by getting dealer quotes, and therefore believed that suggested a Level 2 designation. However given the wide variance in dealer quotes, Freddie decided to move the assets to Level 3. I'd think of it this way: if the model inputs being used by dealers were "observable either directly or indirectly" (Level 2) it stands to reason that the various dealers would have similar observations, and thus similar prices. Since they didn't have similar prices, you have to conclude the model inputs are not readily observable.

Q: Sounds like you are leaning toward 100%.

A: The reality of the bond world isn't that simple. The fact is that the overwhelming majority of fixed income instruments rarely trade. Therefore almost all bonds held on any company's balance sheet are valued by a model. For that matter, bonds that are held in your run-of-the-mill investment-grade mutual fund are similarly valued by model. One could make a case that a very wide swath of bonds are valued with "unobservable" inputs.

For example, there are 1,082 tax-exempt municipals bonds rated below investment-grade by Moody's. Of these, only 307 have traded any time this year. Now I grant that there is some correlation among junk-rated muni spreads, but how comfortable would you feel about the valuation of some struggling nursing home deal in Wisconsin by examining the trading level of a convention center in Texas? According to the FASB "Adjustments to Level 2 inputs that are asset specific... might render the measurement a Level 3 measurement." Sounds like the valuation of rarely traded municipals would fall into Level 3.

Q: But Freddie Mac is getting their quotes from dealers! And Freddie Mac is one of the 5 or so best accounts to have as a bond salesman. The dealer firm is obviously biased.

A: Granted. But what's the alternative? You are talking about positions for which there is no trading market. The best you can do is ask someone what they might pay for it, and value it that way. Its biased, but the alternative would be for Freddie Mac to create their own model. Can you imagine the outrage on the interweb if that's how they valued their positions?

Besides, I'd bet that Freddie Mac thought that getting quotes from dealers was the only way to avoid Level 3 designation. Asking for a theoretical bid from a dealer could reasonably be considered an "observable input" thus allowing a Level 2 categorization. Only when it became obvious that the dealer community had no idea what to bid did Freddie move the assets to Level 3 designation.

Q: So when the dealer quotes were too low, Freddie changed their methodology! Its Enron all over again!

A: Actually Freddie Mac didn't change their methodology, merely moved their ABS portfolio into Level 3. They always valued their positions with dealer quotes. You are better off not obsessing over the Level 3 assets themselves, but rather the fact that no one seems to know what Freddie's assets are actually worth.

Q: I still don't trust their accounting.

A: Neither do I. Its clear that derivative accounting according to GAAP doesn't reflect the reality of Fannie Mae or Freddie Mac's business. My best guess is that Freddie's recent figures were aided in a non-economic way by accounting practices. But who knows? I really don't feel like I have a good handle on it.

And guess what? I'd feel exactly the same way if they had zero Level 3 assets.

So what's the point here? Any financial firm involved in fixed income securities and related derivatives is likely to have significant Level 3 assets. Reflexively assuming this means the firm is involved in shady securities is lazy analysis. The hysteria over Freddie's Level 3 assets is misplaced. Thoughtful analysis as to why Freddie Mac felt compelled to move their assets into Level 3 is what's needed. Its a little spooky to consider that Freddie Mac can't get a good value on their securities, that their dealer evaluations varied so much. That's the more important point in analyzing Freddie's balance sheet.

Note
This calculation as presented here is not entirely accurate, primarily because I erroneously equated the credit loss percentage as if it were a percentage of assets, but in fact it is a percentage of Freddie Mac's guarantee portfolio. That's what I get for trying to put together a back-of-the-envelope example, but there is no excuse for publishing something like this.

I probably shouldn't have included the example at all, as it was a very simplistic calculation, and honestly, it probably took away from my bigger point that FRE and FNM's accounting is a mystery. I mean, I tried to argue the opacity of their books, then I used their books to make a point.

On top of that, I quoted their credit loss percentage of their guarantee portfolio vs. their asset base, which was totally wrong on my part. My idea was to value the company's debt from an asset liquidation perspective but the introduction of the credit loss percentage wasn't the right metric.

The more accurate way to look at it is that the company expects $3.1 billion in credit losses in 2008 vs. the gap between assets and debt of $31 billion. So if you imagine credit losses as requiring cash to flow out the door, we'd need 10x the 2008 credit loss level (with no positive cash flow in the interim) for the debt/asset ratio to fall below 1.

Now that could be coupled with losses in their investment portfolio. That's an area that's difficult to forecast, because I just don't know whether they've properly written down their assets or not.