?Arlington Asset Investment is a hybrid agency/non-agency investment company chaired by Eric Billings. The portfolio consists of 173MM in non-agency securities and 665MM in agency securities.

AI reports operating earnings using a very different convention than its peers.  This is overlooked by the market which gets lost in the accounting complexities of derivative hedges.  On an apples-apples basis for other mortgage REITs, runrate annual operating earnings at AI are 2.08 not the 4.48 which is reported.  Using the multiple of much more liquid and less complicated peers yields a target price of $14.56 which is over 30% below the current stock price.  Arguably given AI’s relative illiquidity and complex credit exposures, it should trade well below this peer multiple.

Investment Thesis?

Arlington Asset accounting differs from industry norms. Markets do not fully understand this and are valuing AI as if its accounting is apples-apples with peer REITs. In fact, AI accounting significantly overstates current period earnings relative to peers. When this catches up with Arlington, as it inevitably must, the stock should revalue.??

Mortgage REITs funded with short term lines typically hedge their cost of funds using a swap, which effectively converts their overnight borrowing rate into a longer term rate. In contrast, AI hedges by purchasing a strip of back-to-back 90 day Eurodollar futures extending out over five years. Economically a five-year term swap and a strip of five-year Eurodollar futures are equivalent. However, they are accounted for quite differently.??

For example, at 3Q11 Anworth (ANH) funded its balance sheet mainly with repos maturing in an average 38 days days at an interest rate of 26bps. However, using swaps, they extended the average maturity of their liabilities to 452 days (1.2 years) at an effective interest rate of 1.15%. The income statement for ANH shows the interest expense for its repurchase debt at 1.15%.??

AI, in contrast, uses very different accounting when it reports operating earnings. In 3Q, AI repo funding had a weighted average interest rate of 37bps and an average term of 14 days. The Eurodollar futures used to hedge this rate had a weighted average rate of 3% over the next five years. However, the income statement shows the interest rate paid on its repo agreements at just37bps. Over the next five years the funding costs of AI will rise to 3% and at some point exceed that level so that the weighted average cost is 3%. To compare AI and ANH on an apples to apples basis one should compare ANH with its 115bps cost of fund to AI with a 3% cost of funds (remembering that the AI cost is higher in part because they have hedged their interest expense much further into the future than ANH).  ??

Without getting too arcane, an OTC swap and a strip of Eurodollar futures are arbitrage equivalent from a real economic perspective (you can accept my word on this or read through this document for the gritty details: http://www.cmegroup.com/trading/interest-rates/files/Eurodollar_Futures_Interest_Rate_Building_Blocks.pdf ) .  In fact, an OTC dealer will consult the Eurodollar market for pricing data and often construct a hedged position between the two markets.  However, the mechanics of the two instruments are slightly different and that is reflected in the accounting.  The economics of an OTC interest rate swap flow through the interest expense line on the income statement.  The economics of the cash settled Eurodollar hedge flow through as securities marks.  While the true economics of the two instruments are for all intents and purposes exactly equivalent, this divergent accounting treatment is why AI reports higher operating income (as AI themselves define that non-GAAP term) than it would using a peer comparable methodology.   

To be very clear, operating earnings is non-GAAP and I am in no way accusing AI of falsifying its financial statements.   I simply note that the world does not understand that what AI calls operating earnings are not what anyone else calls operating earnings and that AI’s numbers are in no way real or sustainable.??  While it is an error to compare AI at 37bps with ANH at 115bps that is exactly what Street analysts on the name do.  Either these analysts do not understand the real economics of AI’s balance sheet or they are hungry to participate in the stream of capital market offerings that these types of companies provide over time.??

AI reported a core earnings run-rate of 1.12 in 3Q11. Adjusted for a 3% real cost of funds AI would have reported a .52 core earnings number or a 2.08 annualized run-rate. Much more liquid agency REITs such as NLY and ANH with far less hair trade around 6-7x run-­-rate earnings and 90-100% of book. Given the relative illiquidity of AI and the complications of its non-agency re-remic securities portfolio, it certainly should not trade above that level. CIM, a non-agency peer with a market cap 10x that of AI, trades at 5x run-rate earnings and 80% of book value. Using conservative 7x run-rate earnings multiple for AI yields a target price of 14.56, roughly 30% below today’s price. ??The market does not seem to understand that AI’s current earnings are not sustainable and indeed whatever they over-earn today below their real 3% cost of funds they will eventually under-­earn on the other side. Analysts however straight-line AI’s interest expense as constant through next year which demonstrates that they don’t understand the real dynamics at work. This should change as AI’s cost of funds continues to creep up over time forcing the Street to review its flawed projections.?? 

• AI has a portfolio of non-agency securities marked at approx. 65% of par.  The company touts that there is upside book value potential if these securities payoff closer to par (i.e., at 75% AI would accrete 3.58/share).  While I do not see any reason why one would expect these securities to trade up versus down in value, an investor with a strong directional view on non-agency housing credit should take this into account.

• AI is actually structured as a C-corp which is different from its predominantly REIT structured peers.  As a C-corp, AI dividends are tax-advantaged versus REIT dividends.  In theory this should make $1 of AI dividends worth more than $1 of a comparable REIT dividend.  However, this advantage is likely offset by the illiquidity of AI’s small float and the “hair” of AI’s hybrid agency/non-agency portfolio structure.

• AI is subscale and will almost certainly attempt an equity offering in the short-medium term.  While equity offerings are normally negative dilutive events for mortgage REITs, in this case the impact could be even greater as it might draw attention to AI’s unique accounting conventions.

• AI does pay a sizable (although unsupported) dividend which could dissuade short-sellers.  One potential approach for longer term investors might be to go long some vanilla mortgage REITs against AI to mitigate the negative carry while waiting for AI to revalue.  In the shorter term an awareness seems to be building among a critical mass of sector investors that all is not right with the market’s understanding of AI’s accounting.  This could force Street analysts, whose earnings models are based on the company’s flawed operating earnings, to shift tack, causing a revaluation.