Column overall title: A Mathematician on Wall Street

Column 12

A Theory of Inefficient Markets – Part II

by Edward O. Thorp
Copyright 2003

“Market prices are frequently nonsensical.” – Warren Buffett

On the morning after the PALM IPO the New York Times had a front
page story on the PALM/COMS price disparity.1 Was that enough to bring
crazed “investors” to their senses? Let’s see what happened.
Table 1 shows the high, low and last prices, taken from the Wall Street
Journal, for selected days. The excess premium paid by those who bought
1.35 shares of PALM instead of one share of COMS is shown in the last two
sections, for the cases when the COMS stub is valued at $0 and when it is
valued at $20 a share. Other cases can be calculated from these by
interpolation or extrapolation. Although the individual prices of the two
securities gyrated wildly during the day, we use the simplified approximation
of supposing they reached their high, low and last prices at the same time. In
fact, the actual fluctuations in premium were probably even wider than the
table indicates. Table 1 shows that even on day four the price disparity
ranged from around 25% to over 50%, depending on the assumed value of
the stub.

6/25/2010 1
 Table 1 The Initial Price Disparity Between 3Com (COMS) and its Palm Pilot (PALM) Spin-off

Palm Pilot Premiuma Palm Pilot Premiumb

3Com Market Prices c
Palm Pilot Market If 3Com Stub = 0 If 3Com Stub = $20
Pricesc
Trading Day High Low Las High Low Last High/Hig Low/Low Last/Las High/High Low/Low Last/Last
t h t

1 T 3-2- 117.2 80 81.2 165 92.6 95.0 105 45 47 125 65 67 (83%)

h 00 7 0 3 6 (90%) (56%) (58%) (107%) (81%)
2 F 3-3- 88.5 80.8 83.0 97.9 76.2 80.2 44 22 25 64 (72%) 42 45 (55%)
00 1 6 4 2 5 (49%) (27%) (30%) (52%)
3 M 3-6- 85 68.0 69.5 77 62 63.1 19 16 16 39 (46%) 36 36 (51%)
00 6 6 3 (22%) (23%) (23%) (52%)
4 T 3-7- 75.75 66.5 72.2 71.3 60 66.8 21 15 18 41 (54%) 35 38 (53%)
u 00 5 8 8 (27%) (22%) (25%) (52%)

a
Palm Pilot Premium is calculated as 1.35 ∗ PALM price minus COMS price, valuing COMS stub as zero, then converting the
difference to a percentage of the COMS price, consistent with the example in the text. Premiums are rounded to the
nearest dollar and to the nearest percent after the calculations.
High/High uses the high prices of the day for each stock. Similarly for Low/Low and Last/Last.
b
The same, except the stub is valued at $20. Simple interpolation or extrapolation then gives the premium for any other
assumed value of the COMS stub.
c
Prices are rounded to two decimal places but the calculations use full accuracy.

6/25/2010 2
Table 2 The Continuing Disparity Between COMS and its PALM Spin-
off
Tradin Day COMS PALM PALM PALM Premiumb
g Last Last Premiuma if COMS Stub =
c c
Price Price If COMS 15% of COMS
Stub = 0
5 W 3-8-00 70.44 64.75 24% 38%
1 W 3-15- 61.06 55.75 23% 37%
0 00
2 W 3-29- 63.20 49.69 6% 18%
0 00
3 W 4-12- 44.31 33.31 1% 13%
0 00
4 TH 4-27- 38.88 27.06 -6% 4%
0 00

a
Palm Pilot Premium is calculated as 1.35 ∗ PALM price minus COMS price,
valuing COMS stub as zero, then converting the difference to a percentage of
the COMS price, consistent with the example in the text. Premiums are
rounded to the nearest dollar and to the nearest percent after the
calculations.
High/High uses the high prices of the day for each stock. Similarly for
Low/Low and Last/Last.
b
Alternately, this column gives the premium assuming the street estimate of
1.50 shares of PALM will be issued to each share of COMS, with the stub
valued at zero. You can then increase the premium by adding your estimate
of the stub value.
c
Prices are rounded to two decimal places but the calculations use full
accuracy.

Speculation was that once PALM stock was available in quantity, and
options were listed which could be used as substitutes for shorting stock, the
arbitrage and hedge fund community would manage to overpower the crazed
speculators and reign in the spread. Table 2 shows that it took more than a
month for the mispricing to mostly disappear. In retrospect it might have
taken longer except that Friday, March 10, 2000, just seven trading days
after the IPO, marked the end of the U.S. tech stock “bubble,” as measured
by the NASDAQ index. Table 2 shows that the price of both COMS and PALM
declined sharply after this, probably causing the crazed buyers of PALM to
sell enough to drive the spread down close to fair value.
With the PALM/COMS example in mind, let’s take another look at the
efficient market theory.

A Portrait of the Ideal Efficient Market (You Can’t Beat It)

For a perfectly efficient market, one you can’t beat, we can ask ideally
that:
1. All information is instantly available to all participants. In
the case of COMS/PALM, we had an extreme, where the information
was front page news for weeks. Any EMH that needs to exclude
this from the information set holds no interest for anyone.

6/25/2010 3
 2. All participants are financially rational, for example they
always prefer more money to less money, other things being equal.
3. All participants can and do instantly evaluate all available
relevant information and determine the current fair price of every
security.
4. New information causes prices to immediately “gap” to
the new fair price, preventing anyone from gaining an excess
market return by trading at intermediate prices during the
transition.
Note: Supporters of this theory realize, in varying degree, that some or
all of these conditions are unrealistic, but claim that the conditions still hold
well enough to make the theory a good approximation. The COM/PALMS
example rebuts each of these assumptions.
The investors who held PALM could have sold it and bought COMS,
thereby increasing their payoff for virtually every future outcome. It appears
as though the PALM investors, if they were aware, either were betting they
could wait and resell to a “greater fool,” or that they did not prefer more
money to less money! Prast (2003) cites Fama as acknowledging that
“individual behavior may be affected by biases, but the market is efficient
thanks to arbitrage by a few rational individuals.”
Fama makes a reasonable point. As I am, in my opinion, a rational
investor who was willing to (and did) arbitrage the spread, why did the
disparity persist? It was because, although the safe way to profit required a
hedge via selling PALM short, buying COMS, and waiting to capture the
spread, only small amounts of COMS were available to be lent to short
sellers, so the arbitrageurs were unable to trade enough to correct the
relative mispricing. Note, however, that investors who were long PALM could
buy the hedge by swapping their PALM for COMS, according to the equation
1 ∗ COMS = 1.35 ∗ PALM + ( 1 ∗ COMS − 1.35 ∗ PALM ) where the last term is the hedge
that an arbitrageur would like to own. Any swap has this same structure:
SWAP=LONG+HEDGE , so the same ideas and arguments apply to any relative
mispricing situation which arbitrageurs can’t fully exploit because there
aren’t enough shares available to sell short.

A Portrait of the Real Inefficient Market (Some of You Can Beat It)

The real world of investing is an inefficient market that some of us can

beat where:
1. Typically, some information is instantly available to the
minority that happen to be listening at the right time and place.
Much information starts out known only to a limited number of
people, and then spreads to a wider group in stages. This
spreading could take from minutes to months, depending on the
situation. The people who act on the information earlier capture
the gains. The others get nothing or lose.
Note: The use of early information by insiders can be either legal or
illegal, depending on the type of information, how it is obtained, and how it’s
used.
2. Each of us is financially rational only in a limited way.
Individually we vary from those who are almost totally irrational to

6/25/2010 4
 some who strive to be financially rational in nearly all their actions.
Thus in the real markets, the financial rationality of the participants
is bounded. Each of us at any one time is able to, at best, rationally
analyze only a small portion of the available investment
opportunities.
3. Participants typically have only some of the relevant
information for determining the fair price of a security. For each
situation, both the time to process the information and the
willingness or ability to analyze it generally varies widely among
individuals.
4. The buy and sell orders that come in response to an item
of information sometimes come in a flood within a few seconds,
causing the price to gap or nearly gap to a new level. But more
often the orders are spread out over minutes, hours, days or
months, as the academic literature on “anomalies” documents.2

6/25/2010 5
 _________________________
1
The New York Times, March 3, 2000, page A1, “Offspring Upstages
Parent In Palm Inc.’s Initial Trading.”
2
See, for example the extensive literature on how the market typically
takes weeks or months to fully adjust the stock price after earnings
surprises, stock buy back announcements, and spin-off
announcements.

6/25/2010 6
 References

1. Prast, Dr. Henriette, Presentation at 2003 Quantitative Finance Review,

Wilmott and 7 City Learning, London, Nov. 11, 2003.

6/25/2010 7