The past fifteen months have not been kind to shareholders of mortgage REITs and especially Arlington Asset Investment (AI). (AI operates as a C corporation, but its business is similar to an mREIT. The company invests in agency residential mortgage-backed securities (MBS), which are issued by Fannie Mae and Freddie Mac, and riskier private-label MBS.)
Mortgage-backed securities (and thus mortgage REITs) do not offer an especially attractive risk-reward balance for investors. When interest rates fall, homeowners rush to refinance (i.e. prepay) their existing mortgages. MBS with attractive coupons are therefore often paid down or retired early. When interest rates rise, homeowners keep their existing mortgages, which lengthens the expected average lives on low coupon mortgages and MBS. Rising yields and longer maturities compound losses for MBS investors.
MBS therefore perform best in a steady economic and interest rate environment. Homeowners continue to pay their mortgages. Investors collect their interest and principal payments. Everybody is happy.
The Plunge. From the beginning of 2015 to today (March 8, 2016), the value of an investment in AI, including dividend payments, plunged 43.5%. By comparison, the Market Vectors Mortgage REIT Income ETF (MORT) fell 12.7%, the Dow Jones U.S. REIT Index (which I believe does not factor in dividends) slipped 4.5% and the total return on the S&P 500 was ‑1.4%.
The business models of mortgage REITs expose them to profit and cash flow declines under different (and opposing) scenarios. Net interest yields on most mortgage-related investment portfolios have fallen for several years now, due to the steady decline in mortgage rates. In 2014, 2015 and early in 2016, declining interest rates produced losses on the derivative financial instruments used by mortgage REITs to hedge their liability exposures (with little or no offsetting gains on investments). Lower net interest yields and losses on hedges squeezed cash flows, forcing many mREITs to cut their dividends.
Recent Performance of AI. In 2015, AI reported a net loss of $69.4 million or $3.02 per share. Net interest income was $135.7 million, up 21% from 2014. By comparison, the average balance of its MBS investments, both agency and private-label, increased 37% to $3.9 billion.
Any analysis of AI’s net interest income is complicated by its accounting policy of including amortization of premiums paid on MBS investments in investment gains and losses. Although this is apparently permissible under accounting rules, it is unorthodox. Amortization of premiums (or discounts) on investments should be included in interest income to give a more accurate and complete calculation of net interest yield. AI’s management said on its February 3 conference call that the company will begin including premium amortization in interest income and core operating income this year.
The current accounting policy for premium amortization also clouds the analysis of investment losses. Such losses were reported to be $152.4 million in 2015, more than the year’s net interest income of $135.7 million. Nearly two-thirds of that amount or $105.8 million represented losses on interest rate derivatives (i.e. hedges); but the company also recorded $64.4 million of losses on trading investments, which includes the amortization of MBS premiums. In 2014, the company reported investment gains of $84.2 million. (It should be noted that mark-to-market losses only reduce AI’s ability to pay dividends when the investments are sold. If held, they continue to produce the same level of interest income.)
Although to my knowledge AI has not disclosed the full year amount of premium amortization, it did say that premium amortization equated to $0.25 per share in the fourth quarter. That works out to about $5.8 million (assuming no tax effect) for the quarter. If I then assume that premium amortization was the same in the three preceding quarters, I estimate the full year figure at $23.2 million or a little over one-third of total losses on investments for the year. The difference of $41.2 million probably represents realized and unrealized losses on MBS, which are classified as trading investments.
AI also reported an income tax provision of $38.6 million in 2015, even though it showed a loss before income taxes of $30.8 million. The provision was due to an increase of $56 million in its valuation allowance against its deferred tax assets. Based upon current market conditions and valuations, AI believes that it will not be able to utilize fully its net capital loss carryforwards of $240.7 million before they expire in 2019 and 2020.
Despite reporting a 2015 GAAP net loss of $3.02 per share, AI said that its core operating income per share (a non-GAAP measure) increased 20% to $5.91 in 2015. Excluding realized gains and losses on private-label MBS, which are usually less predictable and lumpier, core operating income per share was $5.30, up 11% from 2014.
Core operating income per share, which is a metric used by other mortgage REITs, should be a predictor of AI’s ability to pay its dividends. Yet, despite reporting an increase in core operating income in 2015, the company cut its annual dividend from $3.50 to $2.50 in September. Clearly, there must have been other financially quantifiable factors that caused AI to determine that a dividend cut was necessary. Those factors should be included in its definition of core operating income.
The mortgage REIT business model. Mortgage REITs acquire mortgage-related assets (agency MBS, private-label MBS and whole mortgage loans) with borrowed money (typically with very short-term repurchase agreements or repos at loan-to-value ratios in excess of 90%). Thus, they earn the spread or difference between investment income and funding costs. That spread is available to cover loan losses, trading losses (if any) and operating expenses. Nearly all of the profit after these expenses is paid out to shareholders as dividends.
Because interest earned on MBS is usually fixed and interest paid on repos is variable, mortgage REITs are exposed to increasing financing costs in a rising interest rate environment. Accordingly, most mortgage REITs seek to lock in their financing costs through interest rate swaps and futures; but such hedges rarely cover the full repo exposure and have shorter durations than the hedged mortgage-backed securities. Thus, these hedges typically provide time-limited and less then complete protection against the risk of rising interest costs.
Furthermore, these hedges do not protect against widening MBS spreads. Spreads did widen in 2015, as funding costs fell faster than MBS yields. As a result, the hedges employed by AI and other mortgage REITs generated losses with no offsetting gains. Since most mortgage REITs do not employ hedge accounting, they recorded the full amount of those (mark-to-market) losses on interest rate derivatives, which reduced their profits significantly in 2015.
The combination of declining net interest income and losses recorded on interest rate derivatives forced many mREITs to cut dividends in 2015.
AI’s Business Model. AI invests in investment grade-rated agency MBS and high yield private-label MBS (which are usually subordinated MBS tranches in securitizations backed by Alt-A or subprime mortgages). In 2015, AI earned 4% on these investments before amortization of premiums paid on agency MBS. I estimate that amortized premiums reduced net interest yield by about 65 basis points, so net interest yield after premium amortization was probably around 3.35%.
Against this income, the company incurred interest costs on repos and FHLB advances of 0.4% and hedging costs of about 1.1%. Those liabilities covered 90% of total investments, the rest was covered by unsecured long-term debt and equity. So I estimate AI’s effective all-in interest cost for its entire MBS portfolio at about 1.35%, which is equal to a total financing cost of 1.50% times the 90% advance rate on MBS. (The 10% difference represents equity that does not carry a fixed interest cost.) Thus, under the assumptions of this simplified model, AI’s $4 billion investment portfolio earned a spread of 2.0% in 2015, generating net interest income of about $80 million.
That $80 million is theoretically available to cover losses on investments, operating expenses and interest on long-term unsecured debt. I will assume for my analysis that losses on investments are zero. That may be a heroic assumption considering that the company recorded investment losses of $152.4 million in 2015 and average annual investment losses were $29.1 million from 2011 to 2014.
AI’s reported investment losses include (mark-to-market) losses on interest rate derivatives, which can be reversed. They also include amortized premiums on MBS investments, which I have already factored into my calculation of net interest income. Both are therefore excluded from investment losses in this analysis.
Investments that can generate losses under this line item include loans, trading of securities and forward commitments to purchase MBS. Yet, AI may book gains as well as losses on those loans, trades and forward commitments, so predicting what the net impact of all of this activity will be in any given year is difficult. Consequently, I assume in this back-of-the-envelope analysis that investment gains and losses are zero.
From the $80 million of net interest income, I deduct other (operating) expenses, which totaled $14.2 million in 2015 and averaged $16.1 million over the past five years, and interest on long-term unsecured debt of about $4.5 million. This comes to a total of roughly $20 million for these two line items, leaving $60 million available for dividends. With 23 million shares outstanding and an annual dividend rate of $2.50, AI’s current annual dividend cost is $57.5 million.
If investment yields, funding costs and operating expenses remain at or near the same levels going forward and if it can avoid investment losses, the current dividend is sustainable. Those are obviously big ifs. Yet, AI’s stock, which currently carries a 19% dividend yield implicitly assumes something worse.
The current market price appears to assume that the company could not even sustain a 50% dividend cut from the current level. That would translate into a 9.8% yield, which is still quite high when compared with the average dividend yield of 2.3% on the S&P 500. A 33% dividend cut would bring AI’s yield in line with the average mREIT yield of 12.8% and reduce its annual payout to $38.5 million.
Valuation. 2015 was clearly a tough year for most mortgage REITs and especially for AI. AI’s stock fell more sharply than the average mortgage REIT. At the current price, AI is valued at 62% of book value, compared with 80% for the average mREITs. AI’s stock also carries a divided yield of 19.0% vs. the group average of 12.8% and trades at less than 4.0 times forward earnings based upon consensus estimates, compared to a group average of 8.2.
Price-to-book value is a widely-used valuation metric for mortgage REITs. Most mREITs classify their MBS investments as trading securities, so mark-to-market gains and losses on them are included in earnings. Equity book value therefore represents the residual fair market value (after deducting liabilities) of an mREIT’s assets and is a proxy for the liquidation value of a REIT, except that it does not factor in the cost of liquidation.
AI’s book value includes a deferred tax asset worth $4.25 per share; but mREITs do not have deferred tax assets because their earnings are not taxable. After excluding this $4.25 for the deferred tax asset, AI’s equity book value per share is $16.83. At the current (March 8) price of $12.75, the stock trades at 80% of this adjusted book value. On this metric, AI is in line with its peers.
Outlook. After a generally dismal showing and poor relative performance in 2015, mortgage REITs are trading at 80% of book value and have an average dividend yield of 12.8%. AI’s valuation is in line with peers on book value, but cheap on forward earnings multiples and dividend yield.
The discount to book value reflects investor concerns that mREITs will face a more challenging environment as interest rates rise. This could result in additional realized and unrealized losses that will reduce future book value. Low forward earnings multiples and high dividend yields suggest considerable doubt about whether mREITs can achieve and sustain a recovery in earnings from 2015 levels.
Yet, as long as the economic remains steady and the pace of recovery stays sluggish, the Federal Reserve is likely to raise short-term interest rates at a slow pace. A gradual increase in rates could limit the downward pressures on MBS spreads and give mREITs the opportunity to make portfolio adjustments that will allow them to sustain their profits and dividends.
In the near-term, as fears of recession subside, the snapback in medium- and long-term rates may allow AI and other mREITs to win back a large portion of the losses that they have suffered on their interest rate derivatives without incurring one-for-one matching losses on their MBS investments. That could provide a boost to equity book values and reverse much of the negative investor sentiment on AI and mREIT stocks.
Accordingly, I believe that AI and its peers are a good speculative bet at current levels. AI could have upside into the low-$20s, if its financial performance improves. (AI’s $3.50 dividend was a key factor in the stock’s ability to reach an all-time high of $32.96 in 2011. Consequently, under the current business model, the cut in the dividend to $2.50 will most likely limit the stock’s recovery potential. Even so, it is not hard to imagine that the stock could get back to $22.50, which represents a potential gain of 76% from the current level.
Going forward, I am pleased that AI will begin including amortization of premiums on MBS investments in interest expense, but I would also like to see the company redefine its primary, non-GAAP financial metric, core operating income, to include as separate line items realized losses on trading investments, hedging costs and any other income statement or cash flow line item that contributes to management’s assessment of the dividend rate. Both AI and its peers could also do a better job of separating mark-to-market gains and losses from the cash flow streams that help determine the size and sustainability of dividend payments.
March 8, 2016
Stephen P. Percoco
Lark Research, Inc.
P.O. Box 1543
Linden, New Jersey 07036