JENA ECONOMIC
R
ESEARCH PAPERS
# 2007 – 049
Scenario-Based Satisficing in Saving:
A Theoretical and Experimental Analysis
by
Werner Güth
M. Vittoria Levati
Matteo Ploner
www.jenecon.de
ISSN 1864-7057
The JENA ECONOMIC RESEARCH PAPERS is a joint publication of the Friedrich-Schiller-
University and the Max Planck Institute of Economics, Jena, Germany. For editorial
correspondence please contact m.pasche@wiwi.uni-jena.de.
Impressum:
Friedrich-Schiller-University Jena Max-Planck-Institute of Economics
Carl-Zeiß-Str. 3 Kahlaische Str. 10
D-07743 Jena D-07745 Jena
www.uni-jena.de www.econ.mpg.de
© by the author.
Scenario-Based Satisficing in Saving:
A Theoretical and Experimental Analysis
Werner G¨uth
Max Planck Institute of Economics, Strategic Interaction Group, Jena, Germany
gueth@econ.mpg.de
M. Vittoria Levati
∗
Max Planck Institute of Economics, Strategic Interaction Group, Jena, Germany
levati@econ.mpg.de
Matteo Ploner
Max Planck Institute of Economics, Strategic Interaction Group, Jena, Germany
University of Trento, Italy
ploner@sssup.it
Abstract
Contrary to the models of deterministic life cycle saving, we take it
for granted that uncertainty of one’s future is the essential problem of
saving decisions. However, unlike the stochastic life cycle models, we cap-
ture this crucial uncertainty by a non-Bayesian scenario-based satisficing
approach. Decision makers first form aspirations for a few relevant sce-
narios, and then search for saving plans satisficing these aspirations. In
addition to formally specifying scenario-based satisficing in saving, we ex-
plore it experimentally. The results confirm that optimal intertemporal
allocations are difficult to derive, and suggest that satisficing allocations
can be reached easily when aspirations are incentivized.
JEL Classification: C91; D81; D90
Keywords: Intertemporal allocation decisions; Bayesian updating; Satisficing
behavior
∗
Corresponding author. Max Planck Institute of Economics, Kahlaische Str. 10, 07745
Jena, Germany. Tel.: +49 3641 686629; fax: +49 3641 686667.
Jena Economic Research Papers 2007-049
1 Introduction
Household decision making, e.g., the choice of an optimal consumption vector,
is usually analyzed without taking into account uncertainty of, for instance,
the quality of experience goods. Similarly, optimal saving plans were first an-
alyzed by abstracting from all the uncertainty of one’s future. Models of life
cycle (Ando and Modigliani, 1963) and permanent income (Friedman, 1957)
provide a description of optimal behavior when income is deterministic. Due
to decreasing marginal utility, allocation of resources to consumption in each
period follows an homogenous pattern. Later on, life cycle-permanent income
models were extended to account for uncertainty in income (Hall, 1978). When
income is stochastic and agents fully and perfectly exploit all available infor-
mation, consumption follows a random walk. In recent years, the literature on
intertemporal allocation of wealth has been further developed in the direction
of dynamic optimization (see, e.g., Pemberton, 1993, 1997; Adang and Melen-
berg, 1995).
Such Bayesian optimization exercises, though in principle very appealing,
rarely provide practical advice. The decisive reason is that agents do not have
well-behaved intertemporal preferences (Frederick et al., 2002), nor given prob-
abilities for the possible developments of our life experiences. Moreover, indi-
viduals usually do not engage in the often very complex task of intertemporal
optimization. Thus, the empirical research mainly tests qualitative and quanti-
tative aspects of optimal saving. Here, contrary to the rational choice tradition
of microeconomics, we rely on the satisficing approach (Simon, 1955). Our
specification of satisficing accounts for uncertainty of one’s own future, but it
does not require Bayesian reasoning. The central idea is that the decision maker
specifies likely scenarios of life experiences, forms scenario-specific aspirations,
and then searches for actions satisficing these aspirations. Depending on how
successful this search turns out, she decides whether to continue her search or
2
Jena Economic Research Papers 2007-049
to adapt her aspirations.
Usually, satisficing relies on forming different aspiration levels for different
basic goals. When, for instance, searching for an apartment, the decision maker
may specify an aspiration for the rent she is willing to accept, an aspiration for
the number of rooms she wants, and so on. In saving decisions, the decision
maker typically concentrates on one basic goal: guaranteeing herself sufficient
monetary means for consumption over the years she may experience. We as-
sume that the decision maker applies her aspirations to different scenarios of
possible life experiences capturing the uncertainty of her future. More specif-
ically, we suppose that the decision maker forms scenario-specific aspirations
specifying the financial means she wants to have available for consumption pur-
poses (e.g., how much she aims to spend when living rather long, losing her job
or having to retire early).
Similar to the rational choice approach, the satisficing approach with its
three constituent subprocesses (aspiration formation, satisficing, aspiration adap-
tation) provides a flexible and rich terminology to explain why one decides in
certain ways. Such flexibility, however, is a curse rather than a blessing when
it comes to predict behavior because making specific predictions requires im-
posing rather bold assumptions. In our view, such bold assumptions are better
developed in line with stylized facts of empirically observed behavior. Thus,
rather than speculating on how actors satisfice when making saving decisions,
we prefer to collect experimental evidence so as to learn whether and in which
ways people are satisficers.
1
We rely on a two-by-two factorial design distinguishing short and long life
horizon as well as low and high future income. Thus, each participant has to
form four scenario-specific aspiration levels. Our only experimental treatment
variable concerns how these aspiration levels are elicited: once as “cheap-talk”
data, and once as incentivized choices. In the latter case, for the randomly
1
Anderhub and G¨uth (1999) provide a survey of experimental economics studies on saving
behavior. For the psychological literature see W¨aneryd (1999).
3
Jena Economic Research Papers 2007-049
selected scenario, a participant earns her aspiration when this does not exceed
her actual monetary payoff; otherwise, she earns nothing.
Our method of incentivizing aspiration choices provides a new way to test
optimality without imposing risk attitudes. Optimal aspiration profiles do not
allow increasing the aspiration for a specific scenario without endangering the
aspirations for other scenarios. Hence, one can test optimality in a non-Bayesian
way by eliciting state-specific aspirations whose profile should be optimal. Of
course, optimal aspiration profiles have to be consistent, i.e., there should exist
consumption or saving plans satisfying the aspirations for all scenarios. An im-
portant result of our treatment with incentivized aspirations is that participants
learn to form consistent aspiration profiles, but not optimal ones, i.e., even after
becoming more experienced, participants remain only boundedly rational.
The specific saving tasks are illustrated in Section 2 where, after the tra-
ditional rational choice analysis (Section 2.1), we introduce our definition of
satisficing in savings (Section 2.2). The experimental protocol is described in
Section 3, and the data are statistically analyzed in Section 4. Section 5 con-
cludes.
2 The model
The basic approach can be illustrated by a simple situation in which the decision
maker knows her present income or wealth y (> 0), but has a stochastic future
income Y and an indeterminate length of life T . The present certain income y
can be spent for either present consumption c
1
∈ [0, y] or saving y − c
1
. The
latter, together with the stochastic future income Y , has to finance her future
consumption.
We assume that the future income and the number of future life (and, thus,
consumption) periods can assume only two values, i.e., Y ∈ {Y
,
¯
Y } with 0 <
Y
<
¯
Y , and T ∈ {T ,
¯
T } with 1 < T <
¯
T . To limit the number of parameters,
4
Jena Economic Research Papers 2007-049
