How To Find The Fittest IPOs
As investors chased initial public offerings in the late 1990s, they concocted some odd ways to gauge the prospects of companies with lots of buzz but little or no profits. Among the wackiest: the "eyeball ratio," calculated by dividing a company's projected market capitalization by the number of visitors to its Web site. Now, with big names such as Google planning to go public and with the overall number of IPOs rising, after a three-year drought, is there a better way to figure out whether today's hot new stock is more likely to become a runaway success like eBay (EBAY ) or a dog like Pets.com?
A recent academic study suggests it's possible to boost your odds of picking IPO winners and avoiding losers. The secret is a formula that draws on common financial data, such as revenues, return on assets, and debt. Nearly all the information you need is in the S-1 statement that companies are required to file with the Securities & Exchange Commission before going public. "It's a relatively simple model that individual investors can use," says Wayne Shaw, professor at the Cox School of Business at Southern Methodist University in Dallas, who developed it with Steven Henning, a former Cox professor who is now a director at accounting firm Marks Paneth & Shron.
By screening out the IPOs likely to be the worst performers, Shaw says his model, applied to 3,941 IPOs filed between 1995 and 2001, earned returns a year later akin to those of a similar portfolio of already-public small-cap stocks. That's a big deal, given the strong tendency of IPOs to lag after the initial euphoria wears off. In contrast, those IPOs that the model labeled losers trailed peers by more than eight percentage points a year after going public. (Because it's hard for individual investors to get IPO shares at the offering price, Shaw measures returns using the price at the end of the first trading session.)
HITTING PAY DIRT
Why invest in IPOs at all if, on average, winners fare no better than their established peers? For one thing, there's always the possibility you might pick the next Microsoft or Cisco Systems. But even without a big hit, a basket of IPOs selected by the model should outperform the Standard & Poor's (MHP ) 500-stock index by three to four percentage points a year to compensate for the higher risk level, says Shaw.
Of course, no formula is foolproof. From 1998 to 2001, the accuracy rate of Shaw's model ranged from 56.6% to 85.1%, depending on the year. As a result, it's risky to use it to forecast the success of any one IPO. But if you're willing to spread your money among a diversified portfolio of new issues from a cross-section of industries, you just might hit pay dirt. The model "plausibly separates companies that are less likely to do poorly in the long run," says Jay Ritter, a finance professor at the University of Florida who studies IPOs. Still, he adds, "no screen is necessarily going to work in any given quarter or year."
Moreover, the model doesn't take market valuation into account. But you should. If an IPO starts trading more than 40% above its offering price, steer clear. A sudden big rise, Shaw explains, suggests "all the good news is already priced in." It's also a bad sign when a stock starts trading below its offer price, since this indicates demand is soft.
You have plenty of opportunity to test-drive Shaw's model. Through June 18 this year, 171 companies have filed to go public, the most since 2000, says Thomson Financial (TOC ). But as the market heats up, so does risk. "When everyone is filing to go public, you get many more dogs completing IPOs," says Shaw. "This is a time when investors could get burned."
What are some good bets? The model likes Google, Domino's Pizza, and the electronic stock exchange, Archipelago. Among the pans: nutritional-supplement retailer GNC.
LOTS OF NONSTARTERS
Some companies never even get into the screen. Shaw excludes financial companies: Their financial structures don't work with his method. He also disqualifies companies expecting to raise less than $1 million, as well as offerings priced below $1 a share. "They're the most speculative," he says.
Then, he looks in the company's S-1 filing for the nine inputs the model needs to make a recommendation. The next step: Multiply the nine by the coefficients that give the data the proper weighting in the formula, and add the results. As a general rule, if the sum is below +0.5, the company's a good bet. Shun the deal if the sum is above +0.5. Obviously, the closer the outcome is to 0.5, the less clear the argument is for buying or avoiding a stock.
Try Shaw's formula on the Google deal. Start with its return on assets -- calculated by dividing the company's 2003 profits of $105 million by its $871 million in assets. Both numbers (and it's important to take the annual ones) are listed in the amended registration statement Google filed with the SEC on May 21. The answer, 12.1%, gets input as 0.121. Next, move to debt. This figure, measuring long-term debt as a percentage of total assets, is 3.8%, or 0.038.
Another input, "price range," looks at whether the company disclosed a price range for its stock. The idea, says Shaw, is to gauge how complete the prospectus is: "The less complete, the more risk," he says. Indeed, according to Shaw, 98% of successful IPOs included a price range in the initial filing, compared with just 70% of the losers. Because Google did not specify a range, enter zero here. If it had, you would enter a 1. (In fact, if Google had done a conventional offering instead of a Dutch auction, it might have lised a price range.)
If the company changes its price range, put a 1 next to "change in range" -- regardless of whether the price rises or falls. In the case of an increase, pencil in another 1 next to "increase in range." This suggests strong demand for the stock -- a plus in calculating a score.
The final pieces of the puzzle -- the company's revenue, age, the offering's projected proceeds, and the underwriter's rank -- have to be adjusted so they don't disproportionately influence the result. First, divide the millions and billions by a million. So Google's $961 million in revenue becomes 961, and its $2.7 billion of projected proceeds become 2,700. Then put those numbers through the logarithmic function in Excel or a financial calculator. Using a "natural logarithm" formula, Google's $961 million in revenue is 6.868, the $2.7 billion "proposed maximum" offering price is 7.901, and 70, the company's age in months, is 4.249.
"Underwriter ranking" rates the lead investment bank underwriting the offering by its market capitalization. Google's backers -- Credit Suisse First Boston (CSR ) and Morgan Stanley (MWD ) -- both have large market caps. Use the natural log formula here, too. Why is this important? "It's a way to measure risk," says Shaw. For example, "a Goldman Sachs (GS ) is not going to take on a dog and risk its reputation. But an unknown underwriter trying to break into the market may."
The verdict on Google? A thumbs-up with an overall score of -1.02. That, of course, doesn't tell you how much you should pay, but at least now you know whether it's worth your time even considering the question. And it's probably a lot more useful than the eyeball ratio.
By Anne Tergesen