Fooling Some of the People All of the Time, a Long Short (And Now Complete) Story, Updated With New Epilogue (6 page)

As an investment company (a BDC is a type of regulated investment company), Sirrom did not report or consolidate the results of its underlying investments. Instead, Sirrom marked its portfolio to “fair-value.” We took Sirrom’s SEC filings and built a database that tracked the cost and fair-value of every investment in each period. By tracking the performance of loans by the year of origination, we determined that although the overall portfolio statistics appeared appealing, rapid asset growth masked poor results. We estimated that from inception to final maturity, roughly 40 percent of the loans went bad. Moreover, the data indicated management had ample advance warning of problems and was slow to recognize them in the portfolio markings.

Many mezzanine lenders receive free equity warrants known as equity “kickers” because the free warrants kick up the returns. Sirrom marked the loan and the equity kicker separately for valuation purposes. The database revealed that when trouble arose, management would mark down the equity kicker, but would leave the loan at full value. Obviously, if the equity value is reduced, then the risk in the loan has increased, making the loan less valuable as well. Management did not take that into account and kept the loans at full value until it determined that a loss on the loan was inevitable. In looking at the history of the loans, not surprisingly, writing down the equity kicker proved a reliable predictor of future loan write-downs. Further, an initial loan write-down often preceded a further write-down until eventually there was a final loss, or write-off.

This should not happen. If Sirrom’s management marked the portfolio fairly, then future adjustments should be independent of prior adjustments. No pattern should exist. In trading markets, when bad news arises, the market resets the value of securities to the point where at the new price the securities are expected to generate a positive future return. If Sirrom did this, then write-downs should not beget further write-downs. The only explanation was the management was slow to fully acknowledge bad news. And there were other red flags. For example, Sirrom’s auditors, Arthur Andersen, wrote in the 1996 audit opinion that “We’ve reviewed the procedures used by the Board of Directors in arriving at its estimate of value of such investments and have inspected underlying documentation, and in the circumstances we believe the procedures are reasonable and the documentation appropriate.” In the 1997 audit letter, Arthur Andersen removed that sentence.

People began to ask questions and raise doubts. The company managed a final equity raise led by Morgan Stanley in March 1998. In July, Sirrom announced slightly disappointing quarterly results, with two bad loans losing around $10 million, or about twenty-five cents per share. The shares fell from a high of $32 in May to around $15 in July. We covered our short at $10 a share, just before the shares collapsed to under $3 in October.

We got wonderfully lucky in demutualized Summit Holdings Southeast (SHSE), a Florida workers’ compensation specialist. The combination of conservative accounting as a mutual, all the IPO proceeds going to the company and a management team with a large initial stock and option grant, made this appear to be a fat pitch. We invested about 15 percent of the fund at $14 per share in May 1997.

Even better, SHSE had recently begun reducing risk by purchasing reinsurance on very favorable terms. Essentially, reinsurance companies were willing to take most of the risk and pay SHSE a huge fee. We believed that when the market realized SHSE evolved from a risk business to a high-quality, predictable-fee business, both the earnings and multiple would expand. We were actually disappointed when SHSE announced its sale to Liberty Mutual for $33 per share in cash in June 1998.

It turns out that management was savvy and we were fortunate. At the IPO road show, the CEO made the offhand comment that he wanted to sell the company “before the warranty ran out.” I heard the comment, but it did not fully register until later that year when Unicover was exposed. Unicover was a reinsurance broker that induced reinsurance companies to reinsure workers’ compensation on uneconomic terms. Sometimes the same risk was passed around several times. With each transfer, Unicover charged a fee. When the reinsurers saw what happened, several refused to pay claims. Almost every workers’ compensation insurance stock collapsed as the scheme unwound. I suspect the favorable reinsurance Unicover offered enabled SHSE to change its business model. Unicover’s exposure probably would have caused SHSE to implode along with the other workers’ compensation companies. However, any problem belonged to Liberty Mutual rather than us. Sometimes it is better to be lucky than smart.

Another short that did well was Century Business Services (CBIZ), a “rollup” of accounting service firms with lousy accounting itself. In a rollup, a consolidator buys small, private companies at a lower multiple than the consolidator receives in the public markets. Every acquisition is accretive to earnings, which drives the stock price higher and enables the consolidator to use its currency to do more acquisitions of private companies in a never-ending virtuous cycle. Michael DeGroote, a famous and wealthy former partner of H. Wayne Huizenga of Blockbuster fame, led CBIZ. Like most roll-ups, CBIZ claimed to improve the operations of the acquired companies and generate 15 percent internal annual revenue growth. In fact, the sellers tended to be entrepreneurs toward the end of their careers. They sold their businesses to CBIZ and hit the golf course.

CBIZ’s accounting was not compliant with generally accepted accounting principles (GAAP) in several ways. First, CBIZ recognized revenue on newly acquired firms starting on the “effective” acquisition date, which occurred before they actually closed the deals. Second, CBIZ valued the stock it issued as currency at a 40 percent discount to the market value. These tricks enabled it to recognize revenue prematurely, understate goodwill, and mislead investors about the multiples it paid for acquired companies.

For the first time in Greenlight’s history, we wrote letters to the SEC. We critiqued CBIZ’s accounting and asked the SEC to insist on clearer disclosures in future filings. The SEC never responded to us. However, a year later CBIZ restated its acquisition accounting to use “closing” dates rather than “effective” dates to begin recognizing revenue and to increase its goodwill. It reduced the “internal” growth rate, missed budget by a wide margin, and replaced the management team. The shares collapsed from $25 in August 1998 to less than a dollar each in October 2000.

But because we covered CLCX and Sirrom just before the market began a rapid descent into the Russia default, Long-Term Capital Management, and Asian economic crises, we had greater-than-usual net long exposure at the wrong time. As a result, we lost money for five consecutive months, from May to September 1998.

August was our worst month ever. The market had a huge sell-off at the end of the month. I was on vacation in upstate New York that week. The prices were crazy, and there was nothing to do about it. We couldn’t look at the screen and say these were “fire sale” prices generated by other investors going through “forced liquidations.” As detailed later, these were the type of excuses Allied would use to hold its impaired investments at inflated values. The price was the price, and we marked the portfolio to reflect that. We lost more than 8 percent that month. Ouch.

One of our largest partners was a semi-retired, well-known hedge fund manager with a fine, lengthy track record. We were proud to have him as a partner. As the story goes, he claims he called our office on that last day of August and no one returned his call. Keswin, who was not on vacation, said it was absolutely impossible that he would have ignored this partner or his call. The partner soon summoned us to his office. He said he could not believe how irresponsible he thought we had been. He asked about our portfolio. We described our largest investment in Agribrands, an animal feed manufacturer that had been spun off from Ralston Purina. With its Asian market exposure, it had fallen in the sell-off that summer. It was still a great opportunity and would be an enormous winner in 1999. Ultimately, Cargill bought the company a couple of years later for twice where it traded in late 1998. As we told the story, this semi-famous investor scoffed. “I thought you were moneymakers!” At his first opportunity, he fully redeemed his investment in Greenlight. For the next five years, every few months someone told me they’d met him and heard how lousy we were.

He was not alone. A good number of our partners reconsidered their investments after our bad five-month run. While we had worked hard to explain our program and the related risks, some of our partners probably paid more attention to the short-term track record. Our partners redeemed about half their accounts between August and January. It would have been worse except our portfolio of depressed securities recovered with the market in the fourth quarter. Some partners maintained confidence and even increased their investments, and we attracted some new partners. Overall, new investments matched redemptions. We finished the year up 10 percent and ended with $165 million under management.

Though we generated attractive risk-adjusted returns in 1998, we didn’t hit our goal. Growth stocks and large-capitalization companies were the flavor of the day. Many of the huge stocks leading this advance would take years to grow into the nosebleed valuations they achieved that year. The S&P 500 shrugged off the Asian crisis to return an eye-popping 28.3 percent. Coca-Cola led the market, trading around fifty times earnings. The earnings were low quality because the company divested its bottling operations one at a time, creating gains that counted in the earnings. I did not have the guts to short Coca-Cola, but I should have. I figured with a company that large I couldn’t possibly have a unique insight. Instead, I contented myself with explaining Coca-Cola’s problems in investor meetings and quizzing prospective hires about their views on the subject.

CHAPTER 4

Value Investing through the Internet Bubble

From the 1998 low, the Internet bubble launched to its full glory. I believe the battle between America Online and the short-sellers catalyzed the bubble.

America Online traded at a high multiple of what many short-sellers believed to be low-quality earnings. America Online spent heavily on marketing or “customer acquisition costs” to generate monthly fee-paying subscribers. Short-sellers believed America Online inflated its income statement by capitalizing these costs and writing them off over the expected life of the subscriber relationship. America Online’s accounting did not comply with GAAP, which required the costs to be expensed as incurred.

I evaluated shorting America Online and determined that even if the accounting were wrong, it was a lousy short because the true economics of the business were incredibly compelling. The stock was inexpensive considering the company’s economic profits. I calculated the net present value of a subscriber by comparing the up-front cash customer acquisition costs to the subscription payments over the expected life of the customer relationship. America Online was adding so many new customers that it would not take long to justify its seemingly lofty stock price. Add in the possibility of new revenue streams, including advertising, and I saw it was a really bad short idea. Perhaps this is what “value investor” Bill Miller saw that convinced him to step out of the box and take a large long position. I did not have the guts to buy America Online, but contented myself by not shorting it and arguing with those who did.

When America Online bit the bullet and took a huge write-off of its capitalized customer acquisition costs and agreed to expense them as incurred in the future, the short-sellers had nothing left to criticize. Though America Online had lower earnings under the more conservative accounting, the market understood the high return America Online generated on its investment in customer acquisition costs and looked through the lower reported earnings. As the market appreciated America Online’s powerful model, after a small initial decline, the stock soared. America Online traded at a higher multiple than Coca-Cola and was still a buy! Coke lost its market leadership. Now, no multiple was too high to pay for a leading “new economy” stock. Many misunderstood the real chain of events and interpreted America Online’s soaring stock as proof of the dubious theory that traditional valuation measures no longer applied.

By early 1999 the market saw that the best and the brightest of the short-sellers had been proved wrong on America Online. If they were wrong about America Online, they could be wrong about every other Internet stock. Never mind that only a handful had viable, let alone robust, business models. Bearish arguments were no longer considered. I believe the hubris from the victory over the America Online shorts was a primary cause of the Internet bubble.

For the most part, we avoided the damage in the short portfolio by refusing to sell short anything just because its valuation appeared silly. We reasoned that twice a silly valuation is not twice as silly. It is still just silly. Kind of like twice infinity is still infinity. Instead, we concentrated on selling short companies with high valuations combined with misunderstood fundamentals and deteriorating prospects. As always, frauds were preferred.

We found several good frauds in 1999. One of them, Seitel, had a multi-client library of seismic data used to find hydrocarbons. Energy companies partnered with Seitel to “shoot” (shaking the ground and measuring the reaction) data—a costly investment. The energy partner received an exclusive period to use the data. After that, Seitel could re-license it to other energy companies. Seitel capitalized the investment in shooting the data and expensed it in proportion to the
expected
licensing and re-licensing revenue. Seitel assumed that a dollar invested in data would generate $2.50 in revenue. As a result, under Seitel’s accounting it was guaranteed a 60 percent margin on any license or re-license revenue.

However, Seitel did not generate anything close to $2.50 of revenue per dollar of investment. Seismic data do not have an indefinite life. If the data shows a high probability of hydrocarbons, somebody drills to find out. After that, who needs the data? Most of the license revenue came from licensing, rather than re-licensing. As a result, the 60 percent margin assumption inflated Seitel’s earnings.

Worse, the initial licenses covered only a small fraction of the cost of shooting the data. Under Seitel’s accounting, shooting data and the related initial licensing fee generated earnings but burned cash. Re-licensing generated cash, but re-licensing sales were harder to make. In an effort to maintain accounting profits, Seitel increased uneconomic, cash-burning investments in new data shoots.

Based on this analysis, we sold Seitel short during a strong period for energy service stocks. When that cycle ended, Seitel’s shares fell sharply and the short contributed to our 1999 return. However, we did not cover because the accounting story had not played out. Seitel was heavily leveraged and we thought it would go bankrupt. Seitel shares made a strong recovery in 2000, and we stuck with it for a three-year fight until it did finally go bankrupt in the next downturn in the spring of 2002. The CEO was eventually sentenced to five years in prison.

We also found a good long idea that year. Reckson Associates, a large real estate owner, spun off a nondescript entity called Reckson Services. In early 1999, I met the CEO Scott Rechler, who was young, aggressive, and smart. I left our two-hour meeting with only a vague sense of the strategy, but a strong sense that Rechler was going to do exciting things. Reckson Services was an assortment of opportunities, including speculative real estate ventures in student housing and gaming; a shared-office space business; a concierge services provider; and OnSite, which was a money-losing start-up that wired office buildings for Internet access. I calculated there was enough intrinsic value in the traditional businesses to justify the current share price of about $5 without giving any value to OnSite. I wanted the free option on OnSite and felt that Rechler would do something great, so Greenlight invested 3 percent of the fund in Reckson Services.

As Rechler and his team huddled for a few months, the stock began to rise as investors seeking new Internet stocks noticed OnSite, and the stock took off when an Internet tout under the name “Tokyo Joe” highlighted Reckson Services. Reckson Services hired a fellow from General Electric. GE’s management was so well regarded that hiring anyone who worked at GE lent instant credibility. Reckson announced that it would lever OnSite and the shared-office-space company to transform the company into an incubator of Internet companies, catering to small and medium business. An incubator is essentially a publicly traded venture-capital company. (The leading Internet incubator CMGI had a $10 billion market capitalization at the time.) Reckson changed its name to Frontline Capital Group. The market loved the hire from GE, the new strategy, and the new name. The stock soared, reaching $60 a share by the end of the year, valuing the company at $2 billion. We sold one-third of our Frontline Capital position near what would be the top. This was not a brilliant call on the stock: Frontline had simply become too big a percentage of the fund given its valuation. The market had begun to properly (in hindsight, excessively) value the option we had acquired for nothing.

The explosive rally in Frontline Capital, the recovery of Agribrands, the first collapse of Seitel and the early success of a spin-off investment, Triad Hospitals, fueled a 39.7 percent return in 1999. At this point, Greenlight hit “critical mass.” Having survived the 1998 market meltdown and capitalized on the 1999 market melt-up, our results were much better than most value-oriented funds, many of which spent 1999 “fighting the tape.” Seemingly overnight, everyone wanted to invest. Our assets under management hit $250 million.

But in early 2000, things became difficult for us. First, Frontline began to fall. OnSite had trouble over a non-compete clause in its CEO’s contract with his former employer that delayed its planned IPO. Though the Internet bubble still had several weeks to its final top, Frontline did not exceed its late 1999 highs.

We also lost money shorting Chemdex, a publicly traded start-up setting up a business-to-business (B2B) network for companies to sell chemicals to one another. Chemdex paid for its customers to install computers and software to use its network. Chemdex induced a couple of large chemical companies to test the service at a discounted commission by giving them stock. While Chemdex hoped to earn a small commission on each transaction, it booked the entire value of the goods exchanged as revenue. Chemdex had almost no chance to generate enough commissions to cover its enormous up-front investments or its operating expenses, which were plainly not being controlled. We invested 0.5 percent of our capital to short Chemdex at $26 in September 1999, which was up substantially from its $15 IPO price in July.

In November, Chemdex announced a strategic alliance with IBM Global Services, where IBM would sell technology to Chemdex’s customers. I could not figure out why that was exciting, but the shares doubled in a week. I missed the joke and doubled the position at $71 per share. In mid-December, Morgan Stanley’s star Internet analyst, Mary Meeker, reiterated her “outperform” rating, writing, “We think Chemdex has got what it takes.” There really was no more substance to her analysis than that. As a result, the shares soared another 50 percent that week. In late February, Chemdex changed its name to Ventro. The name change indicated Chemdex would expand its network to other industries—“verticals”—and needed a new name to express its bolder ambition.

I got a clue and gave up. We covered at $164 per share on February 22. This made Chemdex/Ventro our biggest short loser of all time, costing us 4 percent of capital. Did I feel smart when the shares hit $243 on February 25? No. That was not ten times as silly as $26. Both were stupidly silly. Of course, after the bubble popped, the shares touched $2 later that year . . . on their way lower.

At the top of the bubble, technology stocks seemed destined to consume all the world’s capital. It was not enough for all the new money to go into this sector. In order to feed the monster, investors sold everything from old economy stocks to Treasuries to get fully invested in the bubble. Value investing fell into complete disrepute. Julian Robertson’s Tiger Fund, which had an extraordinary multi-decade record and became the largest hedge fund in the world, performed poorly while holding a variety of old-economy stocks. Robertson liquidated.

February 2000 was our second worst month ever. We lost 6 percent, mostly in our longs, as capital fled traditional industries. We lost several percent more in early March until the Nasdaq peak on March 10. We lost a little bit of money every day for five weeks. Other than cutting our losses in Chemdex, there really was not much to do about it.

Then . . . the market reversed. Just like that. Partially informed by our Chemdex/Ventro experience, I believe the Internet bubble made its ultimate top the day the last short-seller could no longer afford to hold his position and was forced to cover. Market extremes occur when it becomes too expensive in the short-term to hold for the long-term. John Maynard Keynes once said that the market can stay irrational longer than you can stay solvent. From the peak, the market returned to rationality. The leading stocks suffered a devastating bear market and value investing made a “bottom.” These enormous excesses would be completely reversed over the next few years.

This was a good environment for our strategy, and we recovered from our bad start to the year. However, the Frontline Capital stock we held fell sharply and cost us, and we lost money shorting the mail-order contact lens seller 1-800 Contacts. We believed it sold lenses without properly verifying the prescriptions, as required. The Food and Drug Administration (FDA) investigated, but decided not to act. Again, we could not stand the pain and covered the short at a large loss. Subsequently, the shares collapsed when its supplier, Johnson & Johnson—possibly fearing legal consequences of improper consumer use of its lenses—cut off 1-800 Contacts’ lens supply.

We successfully shorted CompuCredit, a credit card issuer to customers with poor credit. Its rapid asset growth masked the losses in its reported results. We looked at the losses on a “lagged” basis. This allowed us to analyze the credit performance without the influence of fresh loans that had not yet had time to default. We saw that CompuCredit’s losses adjusted for growth were 18 percent a year, rather than the 10 percent touted by management. The company held an analyst day in Atlanta and, aware of our bearish view, pointedly told us that we could only listen by phone. One brokerage firm analyst helpfully pointed out, “Buy the stock. They are having their first analyst day ever. If the news weren’t good, they wouldn’t be calling the meeting.” The stock doubled in our face. This time, we stayed patient. Weeks later, the company announced disappointing results, and our losses turned to gains. As credit losses mounted, CompuCredit’s next quarter’s results were even worse.