Eric Avenel has pointed out to me that my proposition 1 (Colangelo [19
95, p. 329]) holds only if goods are not very close complements (gamma
> -0.85). In fact, if goods are very close complements, both bidders
prefer to see their rival integrating rather than integrating themselv
es, and both prefer the situation with an integration in which they ar
e not involved to no integration at all. It is possible to check that
for gamma less than or equal to -0.85, pi(U,m) > pi(v) > pi(U) and pi(
D,v) > pi(D)(m) > pi(D). Hence there are two equilibria: one in which
the upstream bidder bids a strictly positive epsilon while the downstr
eam bidder bids zero (vertical integration takes place); the other in
which the upstream bidder bids zero while the downstream bidder bids a
strictly positive epsilon (horizontal integration takes place). Both
types of integration are thus possible for gamma less than or equal to
-0.85. The essence of the conclusions reached in the paper remains un
changed.