Our main result on -contract paths is the following development of Theorem 3.
We first note that the lower bounds
(where defined) have been phrased in terms of the function as opposed to
used in the various results on
-contract paths in Section 2.2. It is, of course,
the case that the bound claimed for
will also be a lower
bound on
when
and
holds whenever the deal
is IR. The statement of Theorem 6, however, claims rather
more than this, namely that a specific resource allocation setting
can be defined for each
and each
, together
with an IR deal
in such a way that:
can be achieved by a single
-contract and cannot be realised by an IR
-contract path. Recalling that
is a partial function, the latter property is equivalent to the
claim made in part (c) for the
deal
of the theorem statement.
Furthermore, this same deal although achievable by an IR
-contract path can be so realised
only by one whose length is as given in part (b) of the theorem statement.
Regarding the proof itself, there are a number of notational complexities which we have
attempted to ameliorate by making some simplifying assumptions concerning the
relationship between - the size of the resource set
- and
- the number of agents
which are needed to realise
in a single IR deal. In particular, we shall
assume that
is an exact multiple of
.
We observe that by employing a similar device
to that used in the proof of Theorem 4 we can
deal with cases for which
does not have this property: if
for integer values
and
,
we simply employ exactly the
same construction using
resources with the ``missing''
resources from
being
allocated to
and never being reallocated within the
-contract path. This approach accounts
for the rounding operation (
) in the exponent term of the lower bound.
We shall also assume that the number of agents in
is exactly
.
Within the proof we use a running example for which
and
to illustrate specific features.
We first give an outline of its structure.
Given
a resource allocation setting
involving
agents and
resources, our aim is to define an IR
-contract path
In order to simplify the presentation we employ a setting in
which the agents are
.
Recalling that
, the resource
set
is formed by the union of
pairwise disjoint sets of size
. Given distinct
values
and
with
, we use
to denote one of these
subsets with
the
resources
that form
.
There are two main ideas underpinning the structure of each -contract in
.
Firstly, in the initial and subsequent allocations, the resource set
is partitioned between
and
and any reallocation of resources between
and
that takes place within the deal
will involve only
resources in this set. Thus, for every allocation
and each pair
,
if
then
. Furthermore,
for
should both
and
be involved, i.e.
, then this reallocation of
between
and
will be an
-contract. That is, either exactly one element of
will
be moved from
to become a member of the allocation
or
exactly one element of
will be moved from
to become a member of the allocation
.
In total, every
-contract
in
consists of
a simultaneous implementation of
-contracts: a single
-contract for each of the
distinct pairs
of agents from the
agents in
.
The second key idea is to exploit one well-known property
of the -dimensional hypercube network:
for every
,
contains a
Hamiltonian cycle, i.e. a simple directed cycle formed using only the edges
of
and containing all
vertices.7
Now, suppose
We have noted that each -contract,
that occurs in this path
can be interpreted
as a set of
distinct
-contracts. An important property
of the utility functions employed is that unless
there will be no individually rational
-contract path that realises the deal
, i.e. the
-contract deals must occur simultaneously in order for the progression
from
to
to be IR.
Although the required deal could
be realised by a sequence of
-contracts (or, more generally, any suitable
-contract path),
such realisations will not describe an IR contract path.
The construction of
utility functions to guarantee such behaviour provides the principal component in showing that
the IR deal
cannot be realised with an IR
-contract path: if
is
any allocation for which
is an
-contract then
is not IR.
We now proceed with the proof of Theorem 6.
Proof. (of Theorem 6)
Fix
.
consists of
pairwise disjoint sets of
resources
For and
these yield
and
The sequence of allocations in is built as follows.
Since
is the immediate successor of the initial allocation
, it suffices to describe how
is formed from
(when
) and
from
.
Let
be the allocation to be formed from
. The deal
will be an
contract in which
. For each pair
we have
in the allocation
. In moving to
exactly one element of
is reallocated between
and
in such a way that in
,
, since
and
are tracing complementary Hamiltonian cycles with respect to
this ensures that
, thereby maintaining the invariant property.
Noting that for each distinct pair
, we either have
allocated to
in
or
allocated to
in
,
the description just outlined indicates that the allocation
is completely specified
as follows.
The cube position,, satisfies,
For each, the subset of
that is held by
in the allocation
is,
(where we recall thatThe tables below illustrates this process for our example.-bit labels in the hypercube
are identified with subsets of
.)
Now if and
are distinct allocations generated by the process above then the
deal
is an
-contract if and only if for some
,
. It follows that if
is an
-contract
for some
, then for some
and all
,
.
To determine the minimum value of with which
, we observe that
without loss of generality we need consider only the case
, i.e. we determine the minimum number
of deals before
reappears. First note that in each round,
, if
then
, i.e. each round advances the cube position
places:
and
.
We can also observe that
for any
with
, since
Recalling that is the index of the
-bit label
corresponding
to
in the relevant Hamiltonian cycle - i.e.
if
,
if
- we note the following properties of
the sequence of allocations defined by
that hold for each distinct
and
.
The utility function is now given, for
, by
We now show that is the unique IR
-contract path continuation of
Suppose
is a deal that deviates from the contract
path
(having followed it through to the allocation
). Certainly both of the following
must hold of
: for each
,
; and there
is a
-tuple of pairs
with which
, for if either fail to be the case for some
, then
with the consequent effect that
and thence not IR. Now, if
is the allocation that would succeed
in
then
, and
thus for at least one agent,
. It cannot be the
case that
corresponds to an allocation occurring strictly later
than
in
since such allocations could not be realised by an
-contract.
In addition, since
it must be the case that
since exactly
cube positions in the holding of
must change. It follows that there
are only two possibilities for
:
reverts to the allocation immediately
preceding
or advances to the holding
. It now suffices to observe
that a deal in which some agents satisfy the first of these while the remainder proceed in
accordance with the second either does not give rise to a valid allocation or cannot be realised
by an
-contract. On the other hand if
corresponds to the allocation
preceding
then
is not IR. We deduce, therefore, that the only IR
deal
that is consistent with
is that prescribed by
.
This completes the analysis needed for the proof of part (b) of the theorem. It is clear that since the system
contains only agents, any deal
can be effected with a single
-contract, thereby
establishing part (a). For part (c) - that the IR deal
cannot be
realised using an individually rational
-contract path, it suffices to observe that
since the class of IR
-contracts are a subset of the class of IR
-contracts, were it the
case that an IR
-contract path existed to implement
, this would
imply that
was not the unique IR
-contract path. We have, however, proved that
is unique, and part (c) of the theorem follows.
We obtain a similar development of Corollary 1 in
Proof. As with the proof of Corollary 1 in relation to
Theorem 3, in each case we employ the
contract path from the proof of Theorem 6, varying the definition
of
in order to establish each result.
Thus let
In the remaining case,
and