Case studies · 2000
Palm and 3Com's negative stub
A parent company's stake in its own spun-off subsidiary implied a negative value for the rest of the business, and short-sale constraints stopped arbitrage from closing the gap.
What happened
A staged separation
In 1999, 3Com Corporation was a large US networking equipment maker that also owned Palm Inc outright. Palm made the Palm Pilot, the handheld organiser that was, for a few years around the turn of the millennium, one of the hottest consumer technology products in America. 3Com decided to separate the two businesses, but in stages rather than all at once. The plan was to float a small slice of Palm to the public first, and then distribute the rest of 3Com's Palm shares directly to 3Com's own shareholders later in the year.
On 2 March 2000, 3Com carried out the first stage. It sold about 5 per cent of Palm's shares to the public in an initial public offering (IPO), at an offer price of US$38 a share. 3Com kept the other roughly 95 per cent, about 532 million shares. Every 3Com shareholder was also promised a future payout: when the second stage happened, each 3Com share would convert into a claim on about 1.5 Palm shares, on top of whatever 3Com's own operating business was worth.
A number that should not exist
Demand for the IPO was intense. Palm shares traded as high as about US$165 on their first day before settling back to close at US$95.06. 3Com's own shares, which by construction now contained a claim on those same Palm shares plus 3Com's other businesses, closed the same day at US$81.81.
Those two numbers together produce something impossible. Each 3Com share carried a claim on about 1.5 Palm shares, and Palm closed at US$95.06, so the Palm claim alone inside one 3Com share was worth roughly US$145. Yet 3Com shares were trading at only US$81.81. The market was pricing the rest of 3Com, its networking business, its cash, everything else it owned, at roughly minus US$63 a share, about minus US$22 billion in total. A real, operating company, generating actual revenue, was being priced as if it were worth less than nothing.
This should not be possible if markets can freely arbitrage away the gap. Shareholders can never be forced to put in more money than they paid for a share; that is what limited liability means, so a share of any real business must be worth at least zero. An arbitrageur who spotted the mispricing had an obvious trade available on paper: buy the cheap bundle (3Com stock) and short the expensive piece it contains (Palm stock, in the 1.5-to-1 ratio), locking in the negative stub as a riskless profit once the shares were eventually delivered. Plenty of investors saw this. Very few could actually do it.
A short that could not be built
Palm's public float, the only shares that could be borrowed and sold short, was only about 5 per cent of the company. The other 95 per cent was locked up inside 3Com, not due to be released until the final distribution. Every arbitrageur who wanted to short Palm was fishing in the same tiny pond of borrowable shares. Demand to borrow overwhelmed supply, and the fee to borrow Palm stock spiked, reportedly above 25 per cent a year through April 2000, and higher still afterwards. Short interest in Palm eventually reached about 147.6 per cent of the floated shares, meaning the same shares had been lent, shorted, and re-lent more than once over. For many investors, borrowing Palm stock at any price was simply not available.
A calendar change instead of an arbitrage
Without a working short, the arbitrage that should have closed the gap could not be executed at scale, and the negative stub persisted for weeks. It did not vanish instantly, and it did not vanish smoothly either. On 8 May 2000, 3Com announced it would move up the final distribution date to 27 July 2000, well ahead of the year-end target it had previously flagged. 3Com's own share price jumped by more than 10 per cent the next day, while Palm's fell, because pulling the distribution date forward shrank the amount of time short sellers would have to keep paying that high stock-loan fee, which had itself been part of what was propping up Palm's price relative to 3Com's. The stub narrowed steadily over the following months and turned positive around the time 3Com finally completed the distribution on 27 July 2000, handing out its remaining Palm shares to its own shareholders at a ratio near 1.483 (fixed by the actual number of 3Com shares outstanding on the 11 July 2000 record date).
Not an isolated case
In 2003, the economists Owen Lamont and Richard Thaler published an analysis in the Journal of Political Economy documenting five other tech-sector carve-outs from the same era (uBid, Retek, PFSWeb, Xpedior and Stratos Lightwave) that showed the identical pattern: a parent's stake in a partially spun-off subsidiary implied a negative value for the rest of the parent, and in every case, a binding short-sale constraint was the reason the mispricing was not traded away immediately.
| Date | Event |
|---|---|
| 1999 | 3Com, which owns 100 per cent of Palm Inc, announces a plan to separate the two companies in stages. |
| 28 Feb–2 Mar 2000 | 3Com prices and completes the IPO of about 5 per cent of Palm on NASDAQ, at an offer price of US$38 a share. |
| 2 Mar 2000 | Palm's first trading day: intraday high about US$165, closes at US$95.06. 3Com closes at US$81.81. The implied stub value works out at about −US$63 a share, about −US$22 billion in total. |
| 10 Apr–9 May 2000 | Palm stock-loan fees run above 25 per cent a year; short interest in Palm eventually reaches about 147.6 per cent of the floated shares. |
| 8 May 2000 | 3Com announces it will accelerate the final spin-off distribution to 27 July 2000. 3Com's shares jump over 10 per cent the next day; Palm's fall. |
| 11 Jul 2000 | Record date for the spin-off: the final distribution ratio is fixed at about 1.483 Palm shares per 3Com share. |
| 27 Jul 2000 | 3Com completes the distribution of its remaining roughly 532 million Palm shares. The stub, deeply negative in March, has narrowed to around zero. |
| 2003 | Lamont and Thaler publish "Can the Market Add and Subtract? Mispricing in Tech Stock Carve-Outs" in the Journal of Political Economy, documenting Palm/3Com alongside five similar cases. |
The mechanics, in course language
Lecture 10 develops the no-arbitrage bounds on option prices and put-call parity, \(c + Ke^{-rT} = p + S\). Every one of those bounds rests on the same assumption: if a mispricing appears, someone can trade against it, typically by taking a short position, directly or through options, and hold that position until the mispricing closes. Palm and 3Com is the case that shows what happens when that assumption fails, and it fails on a bound even more basic than put-call parity.
Each 3Com share, after the IPO, was a bundle of two claims: a forward claim on about 1.5 Palm shares, to be delivered at the eventual spin-off date, stapled to a residual claim on 3Com's own operating business, its stub. Because 3Com's shareholders were not required to pay anything further for the business, and could never be asked to, the stub had to be worth at least zero. That is limited liability, the most basic no-arbitrage floor available: a share in a real company cannot have negative value, because nobody can be forced to pay to get rid of it.
The market's day-one prices ignored that floor. Buying 3Com stock should have been a strictly better trade than buying an equivalent Palm position directly: the same claim on Palm shares, plus a piece of 3Com's other business, for a lower total price. The textbook arbitrage was to buy 3Com and short Palm in the 1.5-to-1 ratio, then let the future spin-off shares close out the short automatically, leaving a free stub with a value that was, on paper, negative, meaning a guaranteed profit at no risk.
The trade broke down at the short leg. To sell Palm stock short, an arbitrageur first needed to borrow it, and only about 5 per cent of Palm's shares were available to borrow; the rest sat inside 3Com, waiting for the later distribution. Demand to borrow Palm to run this arbitrage vastly outstripped that tiny lendable float. Borrowing costs spiked and, for many would-be arbitrageurs, shares to borrow simply were not available at any price. This is the clearest illustration of a general point: a no-arbitrage bound only binds while the trade that enforces it is actually executable. Remove cheap access to the short side, as a severe short-sale constraint does here, and even a bound as basic as equity cannot be worth less than zero can be violated in the market for months at a time.
The mathematics
The day-one implied stub value fixes the exact size of the violation, using the closing prices from 2 March 2000 and the precise 1.525 distribution ratio from the academic reconstruction of the deal terms (narrative accounts round this to about 1.5).
$$V_{\text{claim}} = r \times P_{\text{Palm}} = 1.525 \times 95.06 = \$144.97$$
$$\text{Stub} = P_{\text{3Com}} - V_{\text{claim}} = 81.81 - 144.97 = -\$63.16 \text{ per share}$$
Scaling this per-share figure up to 3Com's roughly 350 million shares outstanding gives the total implied value the market was placing on the rest of 3Com's business:
$$\text{Total stub} \approx -63.16 \times 350{,}000{,}000 \approx -\$22.1 \text{ billion}$$
P_palm = 95.06 # Palm close, 2 Mar 2000, USD
P_3com = 81.81 # 3Com close, 2 Mar 2000, USD
ratio = 1.525 # precise distribution ratio (Palm shares per 3Com share)
shares = 350_000_000 # 3Com shares outstanding, approx.
claim_value = ratio * P_palm
stub_per_share = P_3com - claim_value
stub_total = stub_per_share * shares
print(f"claim_value = ${claim_value:,.2f}")
print(f"stub_per_share = -${abs(stub_per_share):,.2f}")
print(f"stub_total = -${abs(stub_total)/1e9:,.1f} billion")Output
claim_value = $144.97
stub_per_share = -$63.16
stub_total = -$22.1 billionThis recomputation lands at −US$63.16 a share, right on top of the "about −US$63" figure quoted above. The small gap against a looser back-of-envelope version comes from using the precise 1.525 ratio here, rather than the rounded "about 1.5" that narrative accounts use for readability. The US$95.06 and US$81.81 close prices, and the resulting minus-US$63/minus-US$22bn figures, are the exact numbers reported in the academic literature on this case.
Data and facts
| Quantity | Value | Source |
|---|---|---|
| Palm IPO offer price, 2 Mar 2000 | US$38/share | Forbes, 2 Mar 2000 |
| Palm shares floated to the public | ≈5% (≈23m shares) | Forbes, 2 Mar 2000 |
| Palm intraday high, day one | ≈US$165 | Contemporary press; Chicago Booth Review |
| Palm close, 2 Mar 2000 | US$95.06 | Chicago Booth Review (Lamont & Thaler, 2003) |
| 3Com close, 2 Mar 2000 | US$81.81 | Chicago Booth Review (Lamont & Thaler, 2003) |
| Distribution ratio (narrative) | ≈1.5 Palm shares/3Com share | Lamont & Thaler (2003) |
| Implied stub value, day one | ≈−US$63/share, ≈−US$22bn total | Chicago Booth Review; recomputed above |
| Peak Palm short interest | ≈147.6% of floated shares | Chicago Booth Review (Lamont & Thaler, 2003) |
| Palm stock-loan fee, 10 Apr–9 May 2000 | >25% per year | Convenience-yield/short-sale-constraint literature (NBER lineage) |
| Final distribution ratio, 27 Jul 2000 | ≈1.483 Palm shares/3Com share | SEC EDGAR, Palm Inc Form 8-K, 2000 |
| Similar negative-stub carve-outs in the same sample | 5 (uBid, Retek, PFSWeb, Xpedior, Stratos Lightwave) | Lamont & Thaler (2003) |
The lesson
- A no-arbitrage bound is a floor set by the trade you can actually do, not by economic common sense alone. Limited liability guarantees a share cannot be worth less than zero, but the market briefly ignored that guarantee because almost nobody could execute the short that would have enforced it.
- A "riskless" arbitrage still needs a security to borrow. The Palm short was visible to any first-year finance student, yet it required borrowing shares that barely existed in public float. The arbitrage was blocked by market plumbing, not by risk.
- Parity-type identities and no-arbitrage bounds generally hold only when the offsetting trade is executable. When a short-sale constraint binds hard enough, a mispricing that looks like free money can persist for months.
- Scarcity of a security to borrow is itself priced. Stock-loan fees above 25 per cent a year on Palm show that the cost of shorting is a real financing cost that belongs in any arbitrage calculation, not an afterthought.
- Corporate actions can move prices as much as fundamentals do. When 3Com accelerated its spin-off timetable, 3Com's shares jumped and Palm's fell, purely because the new calendar shrank the future stock-loan-fee advantage of holding Palm rather than 3Com.
Where it appears in the course
Think about it
- If you had spotted the negative stub on 2 March 2000 but could not borrow Palm shares, what alternative trades, if any, could you have used to express the same view without a direct short?
- 3Com's share price jumped when it announced an earlier spin-off date, even though nothing about its underlying networking business had changed. What does that tell you about what was really being priced into 3Com stock at the time?
- Short-sale constraints do not appear only in obscure corners of the market. Where else in modern markets might a thin lendable float make an obvious-looking arbitrage impossible to execute at scale?
Sources
- Lamont, Owen A., and Richard H. Thaler. "Can the Market Add and Subtract? Mispricing in Tech Stock Carve-Outs." Journal of Political Economy 111, no. 2 (April 2003): 227-268. Working paper precursor: NBER Working Paper No. 8302 (May 2001). nber.org
- Chicago Booth Review, "Can the Market Add and Subtract?" (summary of Lamont and Thaler 2003). chicagobooth.edu
- Forbes, "Palm Charges Out Of The IPO Gate," 2 March 2000. forbes.com
- CNN Money, "Palm a hit, but not home run," 2 March 2000. money.cnn.com
- Forbes, "3Com Spins Off Palm," 28 February 2000. forbes.com
- US Securities and Exchange Commission, EDGAR filing system, Palm, Inc. (CIK 0001100389), Form 8-K filings, 2000. sec.gov
- Cherkes, Martin, and Christopher S. Jones. "A Solution to the Palm-3Com Spin-off Puzzles." Working paper. econ.yale.edu
