Monday, March 26, 2007

Trembling with greed

According to Whitney Tilson, the best time to buy a stock is when you are trembling with greed. That's exactly how I was feeling last week as I set my limit order on First Marblehead each evening. Obviously, given a choice, I prefer to pay at the low end of what the market offers each day. But as the price crept up day after day during the week, I realized that I'd be willing to pay a lot more to get into a position and that my window of opportunity might be closing up. As the sub-prime lending story peeks, its negative effect on financial companies' shares could subside.

Interestingly, the sub-prime meltdown could actually help First Marblehead. In the past, consumers might be tempted to pay for college with a mortgage equity withdrawal (MEW), but that option may be less attractive in the future. Banks in the midst of tightening housing loans, might see education loans as a partial replacement for growth. Bond investors might see education loans (which can't be wiped out by bankruptcy and are backed by third-party guarantees) as a safer alternative investment. It's kinda like how pushing down on one part of a water bed causes another part to be pushed up.

One item to watch as time goes on is the trend of the "trust updates" lines in the income statement. According to the 10-K: "Trust updates reflect changes resulting from the passage of time, which results in accretion of the discounting inherent in the fair value estimates of additional structural advisory fees and residuals, as well as changes in the assumptions, if any, underlying our estimates of the fair value of these service revenue components." In 2002 (the first year I can track down those numbers), the updates represented 1.23% of the total "service receivables" line on the balance sheet, which is First Marblehead's share of the trusts. Since then, the updates have fluctuated between 4.44% and 6.25%. If I understand it correctly, this number represents the sum of all previous years' conservative accounting. If the company gets too aggressive in the coming years or if lots of borrowers prepay their loans, that number could go down or become negative.

In a sense, First Marblehead's conservative estimation of it's trust assets is akin to other under-valued assets such as real estate carried at cost or Oracle's installed base. Analysts tend to discount these assets altogether since they can't be precisely valued. The Standard & Poor's report on First Marblehead says, "Cash generated from operations totaled 21% and 68% of net income in FY 06 and FY 05, respectively. We believe the relatively low percentage of net income converted into cash stems from the large portion of FMD's revenues that are derived from residuals from securitized loans." Now these statements are true, but I'm not sure they are the negative that the report implies. If the company, can't convert earnings to cash, there's an obvious problem. But it seems like management has gone out of its way to structure the trusts so that First Marblehead gets the residuals, which makes sense because it doesn't really need cash and nobody else can value them as well as First Marblehead can. They are eating their own dog food, which must be encouraging to lenders and bond investors.

Friday, March 23, 2007

Why I bought First Marblehead

First Marblehead is an out-sourcer of private student loans. In order to understand why I believe Wall Street undervalues this company, it's important to understand what a difficult asset private student loans are to value:

  1. Student loans are not backed by an asset that can be repossessed and sold as are mortgages and car loans.
  2. Borrowers often have no income or even credit histories.
  3. Student loans do not normally need to be repaid until after the student graduates and enters the workforce.
  4. Loans do not normally carry a pre-payment penalty.
  5. Loans may take 20 to 25 years to be repaid in full.
  6. Private student loans are not backed by a government guarantee, but must compete (at least indirectly) with state and federal government-backed loans.
  7. There is great variation in the ability and willingness of students to repay their loans upon graduation.
  8. Student loans are highly regulated.
  9. Securitization is required for the original lenders to transfer student loans to investors.

These "borrower-friendly" loans (as First Marblehead's CEO calls them) present many difficulties to lenders who would like to enter the market. The only way to be comfortable issuing these loans is to gather a few years (at least 10, I'd imagine) of experience and borrower data. That is exactly what First Marblehead offers. It's important to understand that the company doesn't originate or guarantee any loans, but provide outsourcing to lenders so that they can establish there own "private label" student loan programs.

Currently, First Mablehead provides these services at or near cost, and in exchange, the loan originators agree that if they sell the loans, they do it through First Marblehead. The process of selling loans to investors, called securitization, can be complicated and risky, but banks are generally eager to transfer loans to third parties so that the proceeds may be reinvested. Again, FMD has many years of experience. Once the loans are sold to investors, the originating banks are generally free of them altogether.

If you are observant, you'll notice that so far First Marblehead can't make any profit since it provides services to lenders at cost. When it structures a loan secutization, First Marblehead creates a trust where lenders deposit loans and divides the into tranches with different risks and returns. If the underlying loans are paid off as scheduled, the junior tranches will earn substantial returns on investment. If many of the loans go bad or are paid early, the junior tranches may earn less than expected or even lose principle. Setting up and administering these trusts is a difficult and complicated task, so First Marblehead is able to change substantial fees for these services.

The bulk of the fees are paid upfront by the trust. In the most recent securitization, First Marblehead claimed 13% of the trust balance upfront. It also receives ongoing fees (~1%) and residuals (~5%) of the trust. The later fees are estimates of the present value over the life of the trust, so if the loans default or get paid early, the result could be much worse. Residuals are the most junior tranche in the trust, so in theory, First Marblehead takes the greatest risk and earns the greatest return.

The beauty of the First Marblehead model is that it requires very little upfront and ongoing capital expenses. Most of its value to customers arises from its borrower database, the expertise of its employees, and the strong history of its securitizations. The market for private student loans is growing so quickly, there isn't any need to spend money to expand operations. In my opinion, this investment is a bit like buying Oracle a the start of the internet revolution. As the market increases, lenders are going to be more interested in taking advantage of the intellectual property First Marblehead already owns.

Based on explosive earnings growth, it wouldn't be too surprising to find that First Marblehead is priced to perfection. But my DCF model suggests that the shares are fairly priced even if earnings remain level for the next ten years. Part of the problem seems to be that current earnings are somewhat uncertain since a large portion are estimates of future cash flows from residuals and ongoing fees. The flip-side is that the estimates might under-value these future cash-flows. So far, the estimates seem to be on the conservative side.

One concern is the possibility that lenders will decide to hold onto loans rather than securitize them. If that happens, there's no profit for First Marblehead and the lenders might develop their own borrower databases. This (and many other) concerns are addressed by an excellent analysis from Tom Brown. (To be honest, I never would have found this company without reading his website.)

Wednesday, March 21, 2007

Is organic growth better?

One response to Oracle's 3rd Quarter earnings pointed out that we don't really know how much of Oracle's recent growth has been due to acquisitions and how much is organic. All other things being equal, organic growth is the best sort of growth. For one thing, acquisitions tend to be more expensive and can mask problems in the acquiring company.

Suppose, for instance, that Coke noticed problems in its flagship product. So management decides to acquire Budweiser at a steep premium. Then they refuse to break out the portion of revenues that was beer related so that investors won't notice the slowdown in Coke sales. Another issue is that this sort of growth isn't sustainable. Who would Coke buy next?

But not all acquisition strategies are created equal. Right now, Exxon Mobil is raking in huge amounts of cash for delivering gas to American drivers. But it's clear the ride won't last forever. Eventually existing reserves will be tapped and new sources of energy will be required. Although Exxon could use it's current resources to research and develop alternative energy, it might be cheaper and less risky to wait for a smaller company to develop a winning solution and buy that company.

This, in fact, is Oracle's strategy. Actually, the database giant does one better—they supply the platform that upstart companies use to develop new products. Most modern "Enterprise" applications use some sort of relational database to store vast amounts of information about an enterprise and its connections to the outside world. Generally, the Oracle database must be at least among the target platforms that applications support. Therefore, as new companies emerge to write software for specialized purposes, they are likely to target Oracle's database. And Oracle's R&D efforts to improve the Oracle platform will encourage more startups to target it.

But small software companies have several problems that a large company, such as Oracle, can solve. Small companies are required to spend a significant portion of their revenues selling to customers, supporting customers, and providing a productive work environment for employees. These activities scale particularly well, so larger companies have an edge over smaller ones. As a result, a new product has the potential to become significantly more profitable as part of the Oracle stack then it does on its own.

Ideally, Oracle would develop new products for new markets. But it's unreasonable to demand a company grow organically when it has the opportunity to acquire growth for a lower cost and lower risk.

On a mildly related note, there's news today that Oracle is suing SAP. The complaint accuses TomorrowNow, a company SAP bought shortly after the PeopleSoft acquisition, used support login information from customers who had or were about to switch away from Oracle support to download documents, patches and software. It seems there were numerous unnamed SAP employees involved in the project who are included among the defendants. I can't imagine legal documents normally are very good reads, but Oracle lawyers seem to have a knack for producing entertaining briefs. I throughly enjoyed the anti-trust briefs from a few years ago as well.

I am a bit concerned that the lawsuit will backfire in the court of public opinion, since Oracle is attempting a similar support contract end-around against Red Hat. Assuming Oracle has obtained and is using its support material legally, there won't be an actual lawsuit, but there might be questions within the open source community and the media.

Monday, March 19, 2007

Why I sold Alberto-Culver

A few weeks ago, I sold a call option against Alberto-Culver and on Friday, the option was exercised. I can't think of a better way to sell my shares: I sold for $22.50, the price on Friday was $22.51, and I got paid 30¢ for the privilege. As a result, my total combined ratio (after taking a penny a share loss) is 25.92%.

I still have a few shares left, but I got out of the bulk of my position in Alberto, because my original thesis is no longer valid—Sally Beauty was spun off. As it happens, I estimate that I earned 17.47% (38.90% annualized) on the first purchase and 16.32% (46.72% annualized) on the second purchase. "New Alberto" did better than I prediced, so I'm quite pleased with the overall transaction.

I still think Alberto will outperform the market over the next 5 to 10 years, but I don't think it will be a dramatic overachiever. I'll maintain a minimal stake in the company for a long time, if only because it isn't cost-effective to sell. Based on the dividend discount model, I guesstimate the shares are worth closer to $25 apiece, but I'm happy to sell in order to free up cash for an even better opportunity that I hope to buy into within a few weeks.

Thursday, March 15, 2007

Final possession on Alberto-Culver option

Today was the penultimate trading day on my Alberto option and (coincidently) the first day of the NCAA Men's Basketball Tournament. Watching my option come down to the wire is a bit like watching the final possessions of a basketball game. Early in the game, "fundamental analysis" such as per-possession offense and defense are the best ways to predict the outcome. By the end of the game, however, nearly everything that has happened to that point in the season is meaningless. Often games come down to a handful of possessions in the final minutes that determine which team will come out ahead. In other words, close games are determined by luck.

Yesterday, the odds of avoiding a "loss" on the option where roughly the same as a team up by 2 points playing defense with the shot clock turned off. Today, Alberto-Culver shares jumped 1%, which is sort of like hitting a three pointer. Tomorrow is the final possession of the game, since the option will expire if not exercised. If I'm to record a profit on the option, I need Alberto-Culver to drop by 0.22 or more in tomorrow's session.

Odds of -0.22% or greater gain of Alberto-Culver in one day

Notice that the curve is much smoother around the zero line than the graphs I showed yesterday. That's because I corrected the issue with excess 0% gains. (Basically, I use the adjusted close from Yahoo to avoid miscalculations due to splits and dividends. But that sacrifices precision, especially early in a stock's history. Therefore, if the two adjusted prices are the same and the actual closing prices are only a few percentage points from 0%, I now use the actual closing prices.)

Writing call options is a bit like playing a slightly better team. Over a long period of time, the stock market in general goes up. This graph is the odds that the S&P 500 will advance in a particular day.

Odds of 0% or greater gain of S&P 500 in one day

On the other hand, call options pay a premium, which is a little bit like being spotted a few points (or betting with a point spread). For the ACV option, the game is coming down to the wire. I think I made the right decision based on fundamentals, but I still might be on the wrong side of the trade tomorrow.

The odds the option will be exercised tomorrow are roughly 80%, so by Monday I should have the cash to make another purchase.

Wednesday, March 14, 2007

Odds of my ACV option getting exercised

As of the market close today, Alberto-Culver stands exactly at $22.50—my option's strike price. I would expect the option to be exercised if the price ends a penny or two higher on Friday. This graph shows that (if the past is any guide) there is a better-than-even chance I'll be selling my shares.

 Odds of 0% or greater gain of ACV over 2 days

The horizontal axis shows the range of gains and losses experienced by ACV shares over any two day period. The vertical axis is the odds that the price will increase by a certain percentage or greater in two days. So for a 0% or greater increase, the odds are 56% or so. Actually due to rounding errors there are more 0% gains in my data then there ought to be. (You will notice that the curve becomes vertical on the 0% line. That's not an optical illusion, but an artifact in the data.)

If the increase is 0.89% or greater, my option will lose money compared to simply selling the underlying stock on Friday. This graph shows that the odds are roughly 34% of taking a "loss" on the transaction.

Odds of 0.89% or greater gain of ACV over 2 days

Ideally, the shares will end somewhere between $22.50 and $22.70 so that I will sell the shares at a gain (odds ~ 22%). If that happens, I am ready to roll the proceeds into a new investment. If the option is not assigned, I might sell outright or (more likely) sell another covered call for April.

Monday, March 05, 2007

Why I doubled-down on Oracle

I've been planning on buying more Oracle off and on ever since I sold a portion a few years ago. Lately, I've held off because of price or because I had better ideas for the cash. As recently as last November, I was willing to sell another portion of my holdings. But since then, the price has fallen off far (15.97%) enough for me to switch back to buy mode.

I've borrowed some key ratios from Reuters:

Ratio                    Oracle  Industry*
-----                    ------  --------
P/E (TTM)                 24.02    31.08
Price to Sales (TTM)       5.39     5.96
Price to Cash Flow (TTM)  18.54    26.05
Price/FCF (TTM)           19.70    31.53
Operating Margin (TTM)    32.40    24.49
Net Profit Margin (TTM)   23.03    19.17
Return on Equity (TTM)    26.79    22.22

* Software & Programming 

The point is, Oracle is cheap compared to other software companies. But none of these ratios consider Oracle's superior growth. Year over year, Oracle has increased sales by 24.72% compared to 18.66% for the industry as a whole. Now much of the growth has come from the purchase of Siebel systems in January 2006, so we can't expect that growth to be repeated this year. But it's also becoming clear that the acquisition strategy is working. In particular, when a customer decides to buy one portion of the Oracle Suite, there is a strong incentive to buy all of their enterprise software from Oracle. Oracle the platform company is an even better business than Oracle the database company.

I also must admit that part of the reason I bought more Oracle was to round up an odd lot. Since Oracle doesn't currently pay a dividend, I intend to sell call options against my position. Of course, that requires Oracle to increase in price to where it is overvalued. That might take a few years, however.

Friday, March 02, 2007

Odds of successful call writing

Earlier today, I wrote a call option on my ACV stake. At the time, I was more concerned about getting paid to sell than loosing out on gains. This afternoon, I took a look the odds of exceeding my break-even price at the end of the two weeks. I calculated that if miss out on any increase of beyond 1.89%. First, I grabbed all the closing prices for ACV and loaded them into an SQLite database. Then I counted the total number of two week periods in my dataset:

sqlite> select count(*)
   ...> from (select * from acv_prices a
   ...>       join acv_prices b
   ...>            on (julianday(b.date) = julianday(a.date) - 14));

5503

Next I counted the number of those periods in which the closing price increased by 1.89% or more:

sqlite> select count(*)
   ...> from (select (a.close-b.close)/b.close increase
   ...>       from acv_prices a 
   ...>       join acv_prices b
   ...>            on (julianday(b.date) = julianday(a.date) - 14))
   ...> where increase > 0.0189;

2088

Therefore, if the past is any indication of the future, there is a 38% (2088/5503) chance my option will be called for a loss.

One of my goals in writing the option was that I would like to sell my Alberto-Culver shares. It's likely that a 1% increase will result in my option being assigned, so I also took a look at the number of fortnights in which the stock increased by that percentage or more. I won't show the code, but it turns out that 2551, or 46% of the periods resulted in greater than 1% increases. And just for kicks, I looked at the odds ACV will loose value over the fortnight, which is 43%.

Overall, the odds for each scenario shakes out like this:

ACV Option Odds
DownExpire 43%
Up Expire 11%
Up Excerised for gain8%
Up Excerised for loss38%
Total 100%

Why I sold an Alberto-Culver call option

Today I sold a call option on part of my Alberto-Culver position. Between now and March 16, I may be required to sell shares at $22.50. I'd be happy with a higher price, obviously, but I'm getting eager to use that cash to pick up shares of other companies that I have more confidence in. My current combined ratio is 24.92%.