Adding to American Capital Agency (AGNC)

Summary

• At 84% of book value, AGNC is trading near its historically low.
• Upside is collect 14% dividend and perhaps more with a pull to par. 
• Downside is controlled. There are many risks, but they are mostly uncorrelated. The biggest threat in the near term is book value decline due to rising rates and mortgage spreads. 
• Comparing the current environment to 2Q13 (the worst quarter in AGNC history), book value hits should be limited. 
• I added to my positions this past week.

Situation

American Capital Agency Corp (AGNC) is a mortgage REIT. It earns a spread by buying mortgage backed securities (MBS) with low cost funding, and then magnifying that spread with leverage. In the past few years this allowed AGNC to pay a consistently high dividend yield.

AGNC’s stock price has declined steadily the past 2 months, with price to book now at 84%, among the lowest in its history. My base case outlook is a 12% dividend in the next year, while a P/B recovery (less likely) would add 16% for a total return of 28%. But what is the downside, and the likelihood of that downside?

Many Risks, But Uncorrelated

AGNC is cheap for a reason. Many reasons actually. Among its many risks, the main three are: 1) book value hit due to rising rates and mortgage spreads, 2) flatter yield curve leading to spread compression and threaten the mREIT business model, 3) dividend sustainability with dollar rolls. The last point about dollar rolls is really a unique subset of spread compression. It is relatively obscure and deserves a separate discussion, but for now I want to focus on the first two risks.

The worst case would be a bear flattener, which is a combination of 1) and 2) - rising rates and compression spreads. Rising rates would hurt asset prices and decrease book value in the near term, while compressed spreads hurt P/B multiples by lowering future income. Theoretically, these can happen at the same time. For mortgage REITs though, these are conflicting risks that are unlikely to happen at the same point in time:

o Unlike other financial institutions that constantly have money coming and risk investing with lower spreads, a mortgage REIT’s exposure to spread compression mostly comes from mortgages prepayments, where investors have to re-investment into a lower spread environment. However, if rate/ mortgage basis are shocking upward, prepayments would likely to be muted.

o For AGNC, the most relevant rates are repo funding cost and MBS yields, where repo funding costs are unlikely to spike upward short of a banking crisis. High prepayments (both voluntary and involuntary) are unlikely in that scenario, given the current state of housing markets.

Since these risks are unlikely to happen at the same time, I will focus on the risk of rising rates and spreads hurting book value. This is the more immediate risk, and is also the one that hurt AGNC more historically.

Quantifying the Downside

Historically, AGNC’s worst performance came during 2Q13, when the stock dropped 31% in the quarter. About 2/3 of that price drop was simply due to market multiples. Price to book ratio flipped from a premium of 110% to a discount of <90%, which magnified a 12% decline in book value.




The ~12% in book value (both on a total and per share basis) was due to a confluence of multiple factors:

o A violent 63bps up move in 10yr UST, while mortgage yields were up more than rates.

o A collapse in specified pool pay-ups. AGNC owned prepayment protected MBS such as low loan balance and HARP loans. Normally, these trades a premium (“pay-ups”) to more generic MBS, but as rates go up and people are less worried about prepays, those premiums shrank dramatically.

Here’s an old 2Q13 presentation slide explaining the collapse in pay-ups. Note that the 30year, 4% coupon pay-up dropped from 3.28 in 1Q13, to 0.91 in 2Q13. A decline of 237bps!




As of now, many these factors are simply not present. Comparing the present situation to end of 1Q13 (the start of 2Q13 meltdown), AGNC already trades at ~15% discount to book value as opposed to a 10% premium. Yes, it’s very possible that rates may shock upward, but probably not as violently as in 2Q13 when talks of Fed “tapering” dominated news headlines. Finally, as the below table from Markit shows, specified pool pay-ups are nowhere near where they were in 1Q13, which could be greater than 3 points. As of January 2015, the higher pay ups are for 30 year low loan balance (LLB) pools with >4% coupon, and those are under 3% of AGNC’s portfolio.

Therefore I believe AGNC’s downside is limited - any book value deterioration from rate increases should be less than the 12% drop as seen in 2Q13, and thus well within the current 84% P/B buffer. I do not think P/B will drop much further because that would have to come from spread compression, which as I explained previously, is incompatible with a “rate up” scenario.

Finally, if all else fails management always have the option to do buybacks, as they have done before.