Description

'And what about the cash, my existence's jewel?' - Charles Dickens, Nicholas Nickleby

INTRODUCTION

In the wake of the infamous corporate scandals of the late 1990s, investors have redoubled their focus on companys' cash flow as a sanity check on reported earnings. This trend has been diffuse, touching nearly all investable sectors. We say 'nearly' because it appears nobody is looking at the cash flow statement, or beyond the reported numbers for that matter, for the newly created 'originating mortgage REITs,' most notably American Home Mortgage. American Home, like all other mortgage REITs, is required to pay out 90% of their REIT taxable income, but most egregiously, has not been able to internally generate the cash to do so. We are recommending a short position in American Home Mortgage (NYSE: AHM). But first, a little background.

BACKGROUND

In the mid 1990s, numerous mortgage REITs were capitalized to invest in mortgage securities. Although these companies varied in the types of mortgage assets in which they invested (prime residential, subprime residential, commercial), they shared a number of important characteristics. First, they funded themselves entirely on a wholesale basis, either through asset-backed securities or reverse repurchase agreements. Unlike banks or thrifts, these institutions had no deposits. Second, because they paid out 90% of their income, these institutions could not grow without external capital. Finally, like most non-fee generating financial institutions, mortgage REITs derived substantially all of their income from either interest rate or credit risk. In almost no case did an mREIT have a durable competitive advantage.

After the 1998 liquidity crisis, which precipitated 30-40% depreciation in four of the biggest mortgage REITs, Annaly Mortgage, Impac Mortgage, Novastar Financial, and Redwood Trust, mortgage REITs fell out of favor, and none were capitalized for another few years. But recently, with the help of Friedman, Billing, Ramsey, themselves an investment bank turned mortgage REIT, over a dozen new mREITs have been formed in the last 18 months. With near-zero credit losses for several years and, until recently, a historically steep yield curve, the environment couldn't have been more fecund for mortgage REITs. You see, spread lenders make money in only four ways: duration risk, credit risk, convexity risk, and distribution capabilities. Banks and thrifts rely heavily on the latter, through retail generated loans and deposits, but mortgage REITs, which are generally wholesale players, tend to rely exclusively on the first three sources for their net interest income.

The propitious environment of yesterday has not endured. Spread-based lenders have suffered compression, and passive REITs like Annaly have reduced their dividend accordingly. Warehouse spreads in the mortgage business have come in, and subprime originator REITs like New Century and Novastar have acknowledged a tougher environment in public statements. Credit trends have been so good that investors have become cavalier about risk, bidding down credit spreads, as Redwood has acknowledged in SEC filings. But mysteriously, American Home, seemingly subject to all the aforementioned risks, came out with results this week that blew away expectations, and raised earnings guidance for both this year and next, increasing the dividend for good measure. How can this be possible? What is American Home's secret sauce?

Well, as it turns out, AHM doesn't generate any cash and may not be appropriately amortizing their mortgage securities ..­

CASH FLOW DISCUSSION

In examining the cash flow statements for mortgage companies, it is unfair to rely on cash flow from operations as the linchpin of the analysis. Indeed, for mortgage companies like American Home, cash flow from investing and financing activities is part and parcel of the normal operations. So, let's focus on total cash flow.

2005.Q2 2005.Q1 2004.Q4 2004.Q3 2004.Q2
Net increase (decrease) in cash 34,613 (30,059)6,341 (247,438)336,361
Proceeds from preferred 48,414 - 83,425 52,057 -
Proceeds from common 587 331 776 426 329
Dividends (31,950)(28,931)(26,167)(24,468)(7,575)
Cash before equity deals (14,388)(30,390)(77,860)(299,921)336,032

Evidently, after the second quarter of last year, when the company generated a significant amount of cash, AHM has not generated any cash, even BEFORE the payment of dividends. So, on a cash basis, the company loses money. But, the company pays out dividends to shareholders, further exacerbating the cash drain. So, we ask, where is the disconnect?

We suspect the company would explain this cash conundrum by citing their investments in mortgage securities and mortgage servicing rights, which require cash investments today, but don't produce cash until tomorrow. And in all likelihood, that is a perfectly reasonable explanation. But where does the cash actually come from to pay the dividends? Well, the answer, we suspect, is very simple: it comes from the equity investors. After the proceeds from preferred and common shareholders, the cash flow drain is plugged. And in all likelihood, the company will return to the market this quarter with a common or preferred offering. Round and round the proverbial carousel we go.

Now, we suspect naysayers will point out that in periods of growth, mortgage companies are large consumers of cash. Such naysayers will point to the results of Countrywide Financial, which consumed a lot of cash in the peak of the refinance boom of 2003. And such an argument is quite astute. But to that argument we offer two explanations. The first is that American Home actually SHRUNK their REIT portfolio from $7.4 billion in Q1 to $6.8 billion in Q2. So, they should be generating cash not consuming it. Second, as you will all agree, Countrywide is not a REIT. As such, Countrywide is not required to pay out its non-cash earnings as a cash dividend. In fact, Countrywide (both then and now) has a relatively meager dividend yield, presumably for just that reason. Furthermore, after a brief stint of creating an originating REIT subsidiary in the 1990s called Indymac, Countrywide had IndyMac convert back to a C corp. Mortgage companies, especially growing mortgage companies, make no sense to be structured as REITs, despite what Eric Billings may tell you. The cash, quite simply, is just not there.

EARNINGS DISCUSSION

If cash were the only problem, American Home would be interesting, but unlikely to be compelling. The catalyst, and indeed the degradation in the economic model, relates to earnings, not cash. As we mentioned earlier, American Home makes money in one of three ways: duration, convexity, or credit. But, as you can plainly see in their filings, the net duration of their book is about a month, so there is no duration risk to speak of. Credit, while potentially problematic long-term, couldn't be better in todays environment of excess liquidity and rising home prices. As such, the economic degradation surrounds convexity. The company paid lip service to increased convexity costs in their press release and quarterly conference call this week, but they barely scratched the surface. Let's take this opportunity to delve into more vivid detail.

Since the REIT conversion early last year, American Home has sold loans through two channels. For the loans they wish to discard, they sell loans for cash to Wall Street, Fannie and Freddie, or correspondent mortgage lenders. They are paid for these loans in cash. In the second method, American Home securitizes loans into securities, and repurchases their own securities from their sponsored, securitizing qualifying special purpose entity (QSPE). In this method, there is no net cash exchanged, as American Home is essentially purchasing loans from an affiliated entity. But, per GAAP, they recognize a gain on sale, even though there is no cash received. The gain is intended to represent the future earnings associated with the mortgage security created, discounted back to today. Let¡¯s consider an example.

If AHM creates a loan for $100, records a gain on sale of $2 (consistent with the figures for the past few quarters), and then repurchases the security from the QSPE, it has in essence purchased a package of a loan plus an interest-only strip. Unlike other originating mREITs, there is no look-through to the actual loan on AHM's balance sheet. Rather, they list the securities as trading securities, and tell you the carrying value as a percentage of ¡°par.¡± But par, in this case, is not the par value of the loan, but the par value of the security. This is critical because the security is held at a premium to the loan, but at what premium, we can not discern.

What we do know is the following: American Home has been reporting gain on sales of approximately 200 bps for loans securitized. Furthermore, the company has said that they expect the loans to have 3-4 year lives with regard to the historical production of 3/1 and 5/1 hybrid ARMs. So, if true, then on an economic basis, they should amortize their securities at 50 bps per year, assuming prepayments came in line with expectations. If done correctly, there will be no premium remaining when the loans all prepay (at par).

Recent deals, however, have prepaid much, much faster than expectations. Mortgage ABS tends to prepay very slowly near inception (most likely because few borrowers get a loan and then prepay it within the first six months) and prepayments tend to rise as the ABS ages. We have examined the first three American Home deals from 2004 (only ones with 8 months of data), and the speeds are troubling. We encourage interested parties to view them on Bloomberg (write me if you don¡¯t know how), but the conclusion is that both deals that are at least a year old have prepaid such that half of the senior securities remain outstanding, implying useful lives of 2 years. But, because the prepayment speeds have accelerated (currently above 60%), the useful lives may in fact be less than that.

If the trading securities were being amortized appropriately, the quarterly net interest margin would be significantly lower than it is today. Moreover, the AOCI account that tracks any unrealized gains and losses in their mortgage securities would have sustained heavy damage. As of the end of the second quarter, it had only recently gone negative. We suspect further negative marks are likely. Although there is no way to say exactly how much American Home has overstated the value of their securities (which they will tell you are authenticated by a third party), we think it could be as much as 1.5-2 pts. Taking the lower number, and at 10x leverage, that would be a reduction in their equity capital of 15%.

DESCRIPTION OF NEW STRATEGY

The historical economics notwithstanding, American Home will reassure you that they are in the process of changing their asset retention strategy. The motivation behind this change is putatively that shareholders have grown skittish with the company's gain on sale accounting, and would prefer they account for the same strategy as a financing. In both cases, we think you'll see that the economics are similar, and it's only the accounting that's different. We think they will be unable to mask the accelerating prepayment speeds for much longer, in either case. At 2x current book value, the company's valuation leaves less room for error. But we will leave it to you to come to your own conclusions.

For our purposes, we will spare you the company's turgid description of their newfangled strategy, and cut right to the numbers. In essence, the company will be retaining short-reset MTA loans, which adjust monthly. In demystifying the economics of these loans, we can rely on the voluminous information contained in American Home's most recent asset-backed prospectus, coupled with company comments. The company will tell you the economics are as follows for loans that, on average, yield MTA + 260 bps:

-Upfront origination costs (2 pts ¨C 1 direct, 1 indirect)

-Sub-servicing costs (4 bps annually, which equates to appx $100/loan on $325,000 avg size)
-Amortization of origination costs (75 bps annually ¨C assuming 2.5 year life)
-Deferred bond issuance costs (7 bps annually, which assumes it costs them $2 million to securitize $1 billion in loan receivables, including trustee, bankers, and attorneys fees)
-Credit enhancement for 80+ LTV loans and negative amortization (15 bps annually)
-Asset-backed securities premium (30 bps)

Thus, as the company would have you believe, the returns on equity (assuming 13x leverage) can be decomposed as follows:

Weighted average gross margin (bps) 260
Less: amortization (75)
Less: bondholder premium to index (30)
Less: external credit enhancement (15)
Less: deferred bond issuance costs (07)
Less: sub-servicing costs (03)
Weighted average net margin 130

Net Margin * Asset Leverage (13x) 16.90%
+ Index Portion of Return ~3.00%
Total Return on Equity ~20%

Terrific, right? Well, terrific if accurate, that is¡­ You see, there are just a few minor inaccuracies with the company¡¯s new strategy that deserve mention.

First, who do you think is taking out a short-reset loan today, when the yield curve is flatter than at any time in the last three years? In other words, why would someone pay LIBOR (3.43%) + 2.60% = 6.06% for an adjustable-rate mortgage, bearing the risk of rising rates, when the same person can get a 30-year fixed rate mortgage for 5.66% (per Freddie Mac release for week ending 7/14/2005)??

The answer is twofold. The first relates to the often substantial teaser rates originators offer. A teaser, simply put, is the mortgage company¡¯s equivalent of offering you a temporary sale. The originator offers you a lower rate for a period of time to ¡°tease¡± you into selecting the product they are selling. So, while the company will tell you that the average MTA loan they originate yields MTA + 260 bps, the weighted average mortgage rate today is 2.74%, or MTA + ZERO! The loans will one day have gross margins of 3.10%, but if the teaser period is six months on average (the company says teasers are 1, 3, 6, or 12 months), then it will take three years for the gross margin to be ~260 bps (0.5 yrs*0% + 2.5yrs*3.1% ¡Ö2.6%).

Second, and probably most importantly, MTA is not LIBOR, as you can quite easily see in the following data table; rather, MTA stands for ¡°moving average treasury¡± and is a lagged index relative to LIBOR. As such, in times like these, where the Fed is tightening, MTA is at a significant discount to LIBOR:

2005 Jan Feb Mar Apr May Jun
LIBOR (1mo) 2.59% 2.71% 2.87% 3.09% 3.13% 3.34%
MTA 2.02% 2.17% 2.35% 2.50% 2.63% 2.74%

2004 May Jun Jul Aug Sep Oct Nov Dec
LIBOR (1mo) 1.11% 1.37% 1.50% 1.67% 1.84% 2.00% 2.29% 2.40%
MTA 1.29% 1.38% 1.46% 1.52% 1.60% 1.68% 1.77% 1.89%

Data: Bloomberg (US0001M Index for LIBOR, 12MTA Index for MTA)

As you can readily see, the spread between LIBOR and MTA is at it widest point in the last 14 months (actually since November 2001, but who's counting). But, inconveniently enough, American Home can't fund in the capital markets at MTA; instead, they fund their securitizations at LIBOR like everyone else. But, who, you may ask, absorbs the spread compression between MTA and LIBOR? Well, American Home does, to the tune of 60 bps in today's environment. To those who would suggest American Home will make up the spread when it converges back to historical levels, we offer you AHM's prepayment speeds as a counterpoint. Truly, it seems unlikely that the loans will last that long...

The next bogus assumption relates to bond premiums. Mortgage companies (other than Fannie Mae and Freddie Mac) don't sell asset-backed debt at LIBOR; rather, they sell at a premium to LIBOR to compensate bondholders for credit and convexity risk. AHM suggests that this premium averages 30 bps, but only the AAA portion of their most recent deal traded that tight. If you fully load the deal for the subordinated bond tranches, the weighted average bond premium is typically 5-15 bps wider than the AAA premium. While it's difficult to do a precise calculation in the most recent AHM deal because of pool cross-collateralization, it seems evident that the weighted average premium is at least 10 bps higher than 30 bps, so let's use 40 bps for discussion purposes.

The final important assumption relates to prepayment speeds. Although 25 percent of the loans have three year prepayment penalty, and are thus unlikely to prepay before the third anniversary, the balance of loans have either a one-year or no prepayment penalty. The amortization assumption of 75 bps assumes implicitly that the loans have a 2.5 year life, which seems dubious in light of American Home's speeds on recent deals that suggest that their securitizations have lives below two years (type AHM Mtge GO on Bloomberg). Either way, we expect higher prepayment speeds when the teaser rates burn off, and higher rates still as prepayment penalties sunset and the index adjust upwards (recall that MTA is a lagged index so is certain to rise near-term).

While we can't prove that 15 bps is too low an assumption to use for the credit enhancement, it seems unlikely to be below 20 bps, given that the financial guarantor, FGIC, has generally been a prudent pricer of risk. Bulk mortgage guaranty rates tend to be in the 30-40 bps range for Radian and MGIC, as a means for comparison. This most recent AHM pool of loans was over 25% investor properties and second-homes, traditionally the most susceptible properties to default (though we should concede the the average FICO is 709 and LTV is 74%). The vast majority of MTA loans are subject to negative amortization, which in essence means that the loan balances go up, not down. Much has been written in the financial press about the credit risk of these loans that we will not repeat here, but suffice to say, that loans with negatively amortizing balances are suggestive of borrowers who can't make the necessary monthly payments to service their mortgage, so are forced to rely on PIK-like residential mortgage debt.

So, let's recalibrate the economics, in light of the real bond premia and index mismatch, but ignoring credit, prepayment speeds, and teaser rates (big ignorance). Now, the economics work as follows:

Management net margin estimate 130
Less: LIBOR/MTA mismatch (60)
Less: wider bond premia (10)
Recalibrated net margin 60

Net Margin * Asset Leverage (13x) 7.80%
+ MTA index value 2.74%
Total Return on Equity 10.54%

So, assuming all goes according to plan, and there is neither credit, convexity, nor teaser problems, the ROE here is 10%. Is it really prudent for you to give this company additional equity if they are going to invest it at 10% gross ROE (remember, that's after indirect overhead, but before corporate expenses and any taxes that non-qualified shareholders must incur)?

VALUATION

American Home trades at 1.9x Q2 stated book value of $20.21. The company trades at a 8.7% dividend yield, annualizing the recently declared quarterly dividend of 86 cents/share. Based on the midpoint of company guidance for 2005 and 2006 of $4.60-4.80 and $4.85-5.15, the company trades at 8.4x and 7.9x, respectively.


CONCLUSION

We suspect that the recent phenomenon of originating mortgage REITs will be short-lived. Like Indymac, these companies will likely realize they do not generate the cash to meet the distribution requirements of a REIT. Perhaps they will convert back to C corporations. Under that scenario, after tax-affecting the qualified earnings from American Home, and applying a reasonable mortgage banking multiple of 8x, this company is worth between $25-30 today assuming the returns are as they say. With returns more similar to the ones we illustrate, the company is worth $15-20 WITHOUT dissolving the REIT, or right around adjusted book value, after recognizing the unamortized premium on the balance sheet.
""

Catalyst

-continued adverse mark-to-market changes in securities portfolio
-continued high (or accelerating) prepayment speeds on AHM deals
-any further federal reserve rate increases which will drive MTA/LIBOR spread
-credit deterioration on negative amortization portfolio
-any liquidity event that dries up wholesale funding
-prolonged inability of American Home to raise equity