(FORTUNE Magazine) – The rabbit-hole went straight on like a tunnel for some way, and then dipped suddenly down, so suddenly that Alice had not a moment to think about stopping herself before she found herself falling down a very deep well. --Lewis Carroll, Alice's Adventures in Wonderland

General Electric is without a doubt one of the most beloved stocks in history. About the worst criticism ever leveled at the illustrious company is that its stunning run of profit growth--101 straight quarters--is somehow artificial, the result of "managed earnings." After all, the argument goes, GE never seems to have had a loss in one division that wasn't happily offset by a gain in another; it never seems to have encountered a windfall profit that wasn't smoothed away by a "one time" restructuring charge. Can such an extraordinary record really be the result of an uncannily canny management--or is there a bit of accounting wizardry going on behind the curtain? While there's nothing illegal per se about managing earnings--and indeed, many big-name companies are alleged to be doing just that--GE, for one, denies that it engages in such artful accounting. As longtime CEO Jack Welch likes to say (and as a company spokesman recently reiterated to FORTUNE), "GE manages businesses, not earnings."

And besides, demand legions of investors, what would be so terrible if the company did help smooth out the ride? Isn't such earnings consistency why GE stock trades at a lofty premium to its competitors'? Good question. And the answer isn't exactly cut and dried. But such accounting games do make it difficult to gain a clear picture of how GE's businesses are actually performing. That means, if the company's core operations were ever to hit a rough spot--in a jarring and jumbled economy like this one--investors might not discover it until too late.

Until very too late indeed.

So I decided to take a look for myself. On a tip from a friendly money manager, who served as my White Rabbit of sorts, I dove into GE's financial statements and wound up having an adventure worthy of Lewis Carroll (or Carol Loomis, as the case may be). When I landed, I was in a place where little was obvious, nothing was simple, and no one--not even the number crunchers at GE--seemed to know the difference between reality and fantasy. Welcome to Accounting Wonderland.

To start with, it is a labyrinthine realm. GE's 12 operating businesses--including television network NBC and GE Capital, as well as the companies that make light bulbs and aircraft engines--generated $129.9 billion in revenue last year. If this large, diversified portfolio of businesses is what enables GE to generate a consistent earnings stream, it's also what makes it so obtuse. "It is an extremely difficult company to evaluate because there are so many moving parts," says William Fiala, GE analyst at Edmond Jones. Perhaps that's why only 28 Wall Street analysts cover GE, compared with the 47 who follow Gillette, which has a market cap one-thirteenth the size of GE's.

By far the company's largest and most profitable division is GE Capital Services. It routinely accounts for more than 40% of GE's earnings. Capital includes five of GE's operating segments, involving some 28 subsidiaries. Since GE Capital Services powers GE's earnings, I thought I'd begin there. But almost immediately things started getting strange. Just trying to figure out the corporate structure of all these insurance and financing operations is, well, trying. It turns out there are more incestuous relationships among GE Capital's business units than among the original cast members of The Brady Bunch.

GE Capital Services is divided into two divisions: GE Capital Global Insurance Holdings, which owns GE's insurance companies, and GE Capital Corp., which owns everything else. One of GE Global's insurance businesses, Employers Reinsurance--one of the largest reinsurance outfits in the world--has issued almost $1.2 billion in preferred stock to GE Capital, stock on which it pays dividends. Then GE Global itself has issued another $150 million of dividend-paying preferred stock to GE Capital.

So why would one wholly owned GE entity have to pay dividends to another? I called a half-dozen GE analysts to find out. None of them were aware of these cross-holdings or what purpose they might serve. Think about that for a minute. These people read the company's financial statements for a living. In Accounting Wonderland, analysts who cover the stock are much like the guests at the Mad Hatter's tea party--blissfully oblivious to the illogic swirling around them. Luckily, GE was happy to set me straight. Take a deep breath: It seems that in 1994, GE Capital transferred Puritan, a life insurance company it owned, to Employers Reinsurance in exchange for the preferred stock on which Employers Reinsurance now pays dividends that are set to approximate the earnings of Puritan's assets at the time of the transfer; the stock GE Global issued GE Capital was in exchange for a loan that GE Capital gave it in 1995 to help it buy two German insurance companies for $1 billion. Got that?

Not that you'd know any of this unless you went back to read the 1994 and 1995 annual reports. After that the tangled structure remains in place, but GE's filings omit any explanation. GE assured me that none of this has any impact on its bottom line. Indeed, when you look at the consolidated results at GE Capital Services or GE itself, the dividend expenses from one entity and the dividend income realized by the other cancel out. Still, the arrangement makes it harder to figure out exactly how each operating unit is performing. And why continue such a complicated structure if it serves no purpose? I thought it might be a tax dodge, but GE says no. These cross-holdings are merely vestigial, GE's version of an appendix. And guess what? GE says it is now considering eliminating them.

If, in the above case, who owns what matters not a whit, in other cases GE has profited greatly by shifting assets from one corporate cousin to another. Take NBCi, the less than exceptional NBC-affiliated Web portal. In 1999 GE helped create NBCi by taking an odd assortment of Internet investments--a 19% interest in Snap.com, a 10% interest in CNBC.com, all of AccessHollywood.com, plus a few additional sites-sans-business-model--and donating them to a publicly traded Internet direct marketer called Xoom.com. GE received a 39% stake in Xoom, which was then renamed NBCi. But contributing these assets to NBCi also allowed GE to record a $388 million accounting gain, simply because these sites were now valued at a much higher multiple than they had been as part of GE. Why? Because they were suddenly part of a pure-play Net company with a $3.9 billion market cap and no earnings, naturally.

GE had hoped its stake in NBCi would give it some Internet funny money to play with. It hasn't worked out that way. Instead the stock has lost 97% of its value since its debut. GE's share of NBCi's losses last year came to $258 million. Some of the company's other Internet investments have proved troubling lately too. GE Equity, one of GE Capital's subsidiaries, is essentially an in-house hedge fund. It owns stock in hundreds of companies, many of them former new-economy Wunder-stocks, like Commerce One, Internet Capital Group, Digital Think, and iVillage. When Net stocks were hot, GE quietly used these investments to goose earnings--realizing almost $300 million in after-tax gains in the last quarter of 1999 and another $400 million in the first quarter of 2000. Even with the sales, at one point GE Equity was sitting on $4 billion in unrealized gains in publicly traded stocks plus a $2 billion portfolio of private investments. Then the Internet bubble burst, and billions in unrealized gains turned to billions in unrealized losses. (From March to December, the total value of GE's investments in public companies dropped from about $5 billion to $1.6 billion).

Exactly how great these losses are, however, is a bit difficult to calculate, owing to what in GE's Accounting Wonderland might be called the Cheshire Cat effect: Certain investments suddenly appear, disappear, and then reappear in GE's filings with the SEC. Meanwhile, the value of some investments float, mischievously disembodied from reality. For instance, look at GE's stake in Internet shopping network ValueVision (in another of the company's keiretsu-like deals, ValueVision owns options to buy NBCi stock as well). In one quarterly SEC filing last year, GE claimed its 16 million-plus ValueVision shares were worth $222.6 million when the markets closed on March 31. This simply isn't true. Based on the close of ValueVision's stock that day, they were actually worth $662 million. In the next quarterly filing ValueVision disappears completely from a list of GE's investments, only to reappear the following quarter, when GE claims--correctly--the stock was worth $413 million. Similarly, GE's 8.6 million shares of Internet Capital Group are listed as being worth $923 million at the end of March, when in fact they were worth just $782 million. And then there's GE's investment in NBCi, which it first acquired in November 1999. We see neither hide nor hair of this until September 2000--when it miraculously materializes on a list of investments filed with the SEC and boosts the overall portfolio by $161 million.

Just what the heck is going on here? "There were errors in our methodology for calculating value," says GE spokesman Gary Sheffer somewhat sheepishly. In other words: my bad. And what about NBCi's sudden appearance in GE's September filings? Oh, that was a mistake too. Turns out GE doesn't even account for NBCi like its other Internet stocks--instead it's what GE defines as an "affiliated company," and it consolidates its portion of NBCi's earnings on its own income statement--so NBCi should never have been included in that filing. (Is that Britney Spears I hear? "Oops! We did it again.") GE has already filed amended statements with the SEC, correcting some of these errors, and (after our conversation) it began doing so for the others. But, okay, how many accountants missed these mistakes? GE couldn't provide a number, but the answer is: lots.

The central figures in GE's alleged earnings management--Accounting Wonderland's King and Queen of Hearts--are the company's use of various one-time gains and losses from asset sales and its use of restructuring charges for layoffs and plant closings. Most recently, GE recorded a $1.3 billion gain last year when UBS bought Paine Webber, in which GE had a large stake. But according to a GE press release this gain was "more than offset by one-time charges." The company did something similar in 1993, and again in 1997--when it sold part of its Paine Webber stock to make up for a $200 million charge it took to write down loans GE Capital had granted bankrupt retailer Montgomery Ward, and then took a $2.3 billion restructuring charge to offset gains from a deal with Lockheed Martin. Thank God for coincidences.

Finding these restructuring charges and one-time gains within GE's financial filings is about as easy as using flamingos as croquet mallets (another Alice reference). GE mentions them in the discussion accompanying its quarterly reports but doesn't break them out in its financial statements. In its 1997 annual report, an unnumbered footnote advises investors that the $2.3 billion in restructuring charges have been lumped in with normal operating expenses on the "Cost of Goods Sold," "Cost of Services Sold," and "Other Costs and Expenses" lines of GE's income statement. That makes some sense when you start to read what the charges are for. Some items, such as employee severance, lease termination, and building demolition, certainly sound extraordinary. But then there are others, such as "higher estimated manufacturing cost," which one would be hard-pressed to describe as anything but routine. (It's curious why GE, supposedly the best-managed company in the world, announces big restructuring charges almost yearly--but that's a question for another day.)

And finding out just which of GE's businesses are affected by these one-time charges poses yet another challenge, since the company has excluded restructuring from its segment results since 1998. In fact, this and several other accounting changes and revisions GE has made over the past decade make it nearly impossible to assess the results of any business segment for an entire ten-year period. In many cases these accounting changes make the performance of business segments appear much more positive--and consistent--than before. For instance, the 1997 annual report states that GE's Industrial Products and Systems segment earned a profit of $1.5 billion in 1997, a decline of 6% from the $1.6 billion it had earned the year before. But in its 1998 annual report the same segment's 1997 profits are now listed as $1.8 billion, a 12% increase from the prior year. GE says it is simply complying with SEC rules that require it to present investors with information in the same way it is reported to the company's executives. And, at GE, managers aren't penalized for restructuring charges.

But again, this explanation is a head scratcher. Why aren't managers held accountable for these expenses? After all, how does a manager evaluate his business without being aware of the costs associated with layoffs and factory closings? How does he determine his return on capital? To be fair, GE says it tries to outline the way in which restructuring costs are divvied up--it just doesn't do so in its segment results. Instead the company leaves it up to investors and analysts to do the math. And GE says it ignores restructuring charges when evaluating operational performance because business-unit managers have no control over them. Such decisions are made at HQ.

When I called General Electric and said I was doing a story on how confusing its financial filings are, the company informed me that it had won all sorts of awards for transparency. Last year a trade magazine named its investor-relations department the best in the country. And in a survey of analysts and portfolio managers, GE was selected as the publicly traded company that does the best overall job of disclosing financial and corporate information (only 6% of those surveyed picked GE, but this was twice as many as chose the runner-up, Intel). Analysts and portfolio managers, however, get to spend a lot of time on the phone asking GE to clarify and expand on its financial statements. And in many cases GE provides analysts with detailed information that it doesn't reveal to the general public. GE can do this despite a new SEC rule--rule FD (for fair disclosure)--that is supposed to prohibit a company from making important information available to only a select group.

The reason is, once again, a function of GE's size. Rule FD says only that "material" information cannot be disclosed selectively. But materiality is a somewhat subjective term, and in a $129 billion company it takes a pretty big number for something to become "material" to consolidated earnings. Yet isn't there a point at which several immaterial events--a $100 million gain there, a $50 million charge there--might add up to something material? As in the last quarter, perhaps, in which GE's profit growth from the previous year's fourth quarter totaled $496 million. "We are trying to provide investors with as much information as they need to make a valid decision on whether to buy the stock," GE CFO Keith Sherin says. "We have to reach a balance between line-item detail and providing information that will be useful to our stockholders."

Perhaps the company is simply too big to achieve this balance--or maybe the details are too small. In Accounting Wonderland, you sometimes can't tell at all.

FEEDBACK: jkahn@fortunemail.com