6.4 Preference Reversals and Multiselves
There is evidence that individuals resolve the same intertemporal trade-off differently
depending on when the decision is made.
14
Researchers, starting with the work of Strotz
(1955), have argued that this phenomenon requires modeling the individual as a collection
of distinct selves with conflicting interests. Such models represent a major departure from
standard economics conception of the individual as the unit of agency. For example, if the
individual cannot be identified as a coherent set of interests, then the economists’ welfare
criterion is not well-defined. Hence, for neuroeconomists, preference reversals constitute
an empirical validation of the psychologist’s — as opposed to the economist’s — view of the
individual.
Consid er the following example: in period 1, the agent c hooses the consumption stream
(0, 0, 9) o ver (1, 0, 0) and chooses (1, 0, 0) over (0, 3, 0). Inperiod2theagentchooses(0, 3, 0)
over (0, 0, 9). Suppose the agen t faces the following decision problem: he can either choose
(1, 0, 0) in period 1 or leave the choice bet ween (0, 0, 9) and (0, 3, 0) for period 2. Confronted
with this choice, the agent pic ks (1, 0, 0).
In Gul and Pesen d orfer (2001), (2004) and (2005), we propose a standard, single-self
model that accounts for this behavior. To illustrate the approach, define C to be the set of
second period choice problems fo r the individual; that is, an element C ∈ C consists of con-
sumption streams with identical first period consumption levels: (c
1
,c
2
,c
3
), (c
0
1
,c
0
2
,c
0
3
) ∈
C ∈ C implies c
1
= c
0
1
. In period 2, the individual chooses a consumption stream from
some C. In period 1, the individual chooses a choice problem C for period 2. Choosing
(1, 0, 0) in period 1 corresponds to {(1, 0, 0)} while the option of leaving it to period 2 to
choose between (0, 3, 0) and (0, 0, 9) is described as
C = {(0, 3, 0), (0, 0, 9)}
With this notation, we can summarize the (period 1) behavior as
{(0, 0, 9)}Â{(1, 0, 0)}ÂC = {(0, 3, 0), (0, 0, 9)} ∼ {(0, 3, 0)}
14
See Loewenstein, et al. for a recent survey of the experimental evidence. In the typical experiment,
subjects choose between a smaller, date 2 reward and a larger, date 3 reward. If the choice is made at
date 2, then the smaller-earlier reward is c hosen. If the choice is made earlier (i.e., at date 1) then the
larger-later reward is chosen. This phenomenon is sometimes referred to as dynamic inconsistency or a
preference reversal.
36