to complete its divestiture of Palm approximately six months following this
offering by distributing all of the shares of Palm common stock owned by
3Com to the holders of 3Com’s common stock.” The statements often
mentioned IRS approval as a precondition of distribution; the specified
time frame for the distribution was usually six to twelve months.
We searched registrations starting in 1995, although since Edgar’s data-
base was incomplete prior to May 6, 1996, we were unable to find all firms
before then. As it happens, we find no firms in 1995 that satisfied our re-
quirements, so the final sample contains 18 issues from April 1996 to Au-
gust 2000. This sample, shown in table 4.1, consists of every carve-out of
less than 20 percent of subsidiary shares in which the parent declared its in-
tention to distribute the remaining shares.

B. Constructing Stubs

We define the stub value using the ratio of subsidiary shares to be given to
parent shareholders at the distribution date. The ratio is the parent’s hold-
ings of the subsidiary divided by the outstanding number of shares of the
parent on the record date of the distribution. Unfortunately, this ratio is not
known with certainty on the issue date because the number of parent shares
outstanding can fluctuate, for example, because of the conversion of con-
vertible debt or the exercise of options owned by insiders.
Let the parent stock have a date 0 price per share of and the sub-
sidiary stock. Let xbe the ratio of subsidiary shares that are given to
parent shareholders at the distribution date. A negative stub means that
We can also express the stub as a fraction of the parent,
which we do with a lowercase s:

Thus to calculate stub values, we have to estimate the expected ratio at
each point in time. We did this in two stages. First, we simply used the
naive ratio of the parent holdings in the subsidiary divided by the current
parent shares outstanding, using Center for Research in Security Prices
(CRSP) data on shares outstanding. Since the various contingencies gener-
ally raise the number of shares of the parent, this naive ratio likely over-
states the actual ratio and thus makes the calculated stub more negative.
Second, after examining the pattern of stubs in the eighteen cases, we more
carefully studied the cases of potential negative stubs.
We concentrate on negative stubs in order to consider only cases of
clear violations of the law of one price. Of course, there may be mispric-
ing in other situations, but in such cases there is no uncontroversial proof

s

PxP

P

S

P

PS

(^0) PP
00
0
0
0

−
=.
SPxP
PS
00 =− <0. 0
P 0 S
P 0 P
136 LAMONT AND THALER