Rules of Thumb

The Economic Journal, 122 (May), 479–501. Doi: 10.1111/j.1468-0297.2012.02502.x. Ó 2012 The Author(s).
The Economic Journal Ó 2012 Royal

Economic Society. Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.
RULES OF THUMB IN LIFE-CYCLE SAVING DECISIONS*
Joachim K. Winter, Kathrin Schlafmann and Ralf Rodepeter
We analyse life-cycle saving decisions when households use simple heuristics, or rules of thumb,
rather than solve the underlying intertemporal optimisation problem. We simulate life-cycle saving
decisions using three simple rules and compute utility losses relative to the solution of the optim-
isation problem. Our simulations suggest that utility losses induced by following simple decision rules
are relatively low. Moreover, the two main saving motives reflected by the canonical life-cycle model –
long-run consumption smoothing and short-run insurance against income shocks – can be addressed
quite well by saving rules that do not require computationally demanding tasks, such as backwards
induction.
Much recent research on householdsÕ saving and investment behaviour has focused on
financial literacy, that is, on individualsÕ knowledge of such fundamental financial
concepts as compound interest, and their ability to apply such knowledge when making
financial decisions (Lusardi and Mitchell, 2007; van Rooij et al., 2011). In this article,
we take a different but related perspective on householdsÕ savings decisions. We start
from the – at least to us – natural assumption that even financially sophisticated
individuals do not solve intertemporal optimisation models when they make their
saving and investment decisions. Rather, they often use simple or sophisticated de-
cision rules, which we call rules of thumb in this article. The question we ask is: how
large is the utility loss that households incur when they use such rules of thumb rather
than solve an intertemporal optimisation problem?
It is a well-known finding from psychological research on decision making that
individuals use heuristics, or rules of thumb, in making judgements and decisions. In
economics, rule-of-thumb behaviour has been recognised as an important aspect of
bounded rationality since the seminal work by Simon (1955).1 Life-cycle consumption
and saving decisions are a case in point. There is a large literature on such models and
their solution. In realistic versions which incorporate income uncertainty, the solution
of the underlying intertemporal optimisation problem is rather complicated. It requires
backwards induction, and no closed-form solution for current consumption as a function
of the relevant state variables exists. Many authors argue that individuals are unable to
perform the calculations which are required to solve the underlying intertemporal
* Corresponding author: Joachim Winter, Department of Economics, University of Munich, Ludwigstr. 33,
D-80539 Munich, Germany. Email: winter@lmu.de.
This article is a substantially revised and extended version of unpublished research by Rodepeter and
Winter (1999). We thank Michael Adam, Axel Börsch-Supan, Thomas Crossley, Angelika Eymann, Werner
Güth, Silke Januszewski, Ronald Lee, Alexander Ludwig, Annamaria Lusardi, Daniel McFadden, Matthew
Rabin, Paul Ruud, Daniel Schunk, the special issue editor (Rachel Griffith) and referees, and participants at
numerous seminars and conferences for helpful discussions and comments. We gratefully acknowledge
financial support by the Deutsche Forschungsgemeinschaft (DFG) through SFB 504 at the University of
Mannheim (Rodepeter and Winter) and through GRK 801 at the University of Munich (Schlafmann).
Kathrin Schlafmann also gratefully acknowledges support by LMU Mentoring.
1
For surveys of this literature, see Camerer (1995), Conlisk (1996) and Rabin (1998). Gigerenzer and
Todd (1999) provide a psychological perspective on bounded rationality and heuristics.
[ 479 ]
480 THE ECONOMIC JOURNAL [MAY
optimisation problem by backwards induction (Wärneryd, 1989; Pemberton, 1993;
Thaler, 1994; Hey, 2005).
Standard economic theory is based on the notion that if individuals have preferences
over all possible states of nature at the current and any future date, and if their
behaviour is time consistent, there is some intertemporal utility function that individ-
uals maximise. The standard approach in the literature on householdsÕ life-cycle
behaviour is to assume that preferences are additively separable over time and that
there is some discounting of future utility. More specifically, it is standard to assume
that the rate at which individuals discount future utility is constant and that the within-
period utility is of the constant relative risk aversion (CRRA) type.2 A well-defined
intertemporal optimisation problem exists that corresponds to these intertemporal
preferences. This problem is well understood, and the standard model of life-cycle
saving serves as a powerful tool in applied research and policy analysis.3
Many important questions remain open, however. Do individuals behave according
to the solution of an intertemporal optimisation problem? If not, how do individuals
make intertemporal choices? Further, if the assumptions of the standard life-cycle
model are not empirically warranted, can it nevertheless deliver predictions that are
valid in practical applications? There is a large and still growing empirical literature
that addresses these issues from different perspectives.4 It is our reading of this liter-
ature that the question whether rational behaviour is an empirically valid assumption in
life-cycle models is still open.
A few examples serve to illustrate this point. In experimental studies of intertemporal
decision making, rational behaviour is frequently rejected. In the context of life-cycle
models, a series of experimental studies test whether individuals perform backwards
induction in cognitive tasks that involve some dynamic trade-off (Hey and Dardanoni,
1988; Carbone and Hey, 1999; Johnson et al., 2001). In their experiments, backwards
induction, and hence rational behaviour, is strongly rejected. In experimental studies
of search models (which is an intertemporal decision task slightly different from life-
cycle decision making, but more akin to experimental study), Moon and Martin (1990)
and Houser and Winter (2004) find that individuals use heuristics that are quite close
to optimal but still different from optimal search rules (which have to be computed by
backwards induction). Additional evidence on the prevalent use of heuristics in in-
tertemporal or dynamic decision problems comes from a series of experiments by Hey
and Lotito (2009), Hey and Knoll (2010) and Hey and Panaccione (2011). More closely
related to the topic of this article, Anderhub et al. (2000) and Müller (2001) document
that individuals use relatively sophisticated heuristics but do not use backwards
induction in experimental studies of a simple saving task. Finally, Binswanger and
2
There are many papers which depart from this standard model. An important example is the literature
on hyperbolic discounting which departs from additive separability and exponential discounting (Laibson,
1998).
3
Browning and Lusardi (1996), Browning and Crossley (2001) and Attanasio and Weber (2010) review this
vast literature. Low et al. (2010) present a recent, sophisticated version of this approach that encompasses
wage and employment risk. Blundell et al. (2008) study empirically whether households insure against income
shocks; they document partial insurance for permanent shocks and almost complete insurance of transitory
shocks among US households.
4
For reviews of the literature on choice over time, see Loewenstein (1992), Camerer (1995), Rabin (1998)
and Hey (2005).
Ó 2012 The Author(s). The Economic Journal Ó 2012 Royal Economic Society.
2012 ] RULES OF THUMB IN LIFE-CYCLE SAVING DECISIONS 481
Carman (2011) present survey data on the use of rules of thumb in actual household
decisions. They show that three types of households – planners, rule-of-thumb users
and ÔunsystematicÕ savers – each make up about a third of their sample. Interestingly,
the wealth outcomes of rule-of-thumb users are similar to those of planners.
In our analysis, we take a normative perspective, maintaining the assumption that
individuals have standard intertemporal preferences. Our point of departure is the
observation that individuals have limited computational capabilities. We characterise
individuals who follow simple heuristics, or rules of thumb, rather than use the decision
rule given by the solution to the dynamic optimisation model that corresponds to
maximising their preferences. More specifically, we compute life-cycle saving decisions
under three different exogenously specified saving rules and compare the outcomes
with the optimal solution. The criterion used for this comparison is a consumption
equivalent (CE): we express losses in life-time utility associated with rule-of-thumb
behaviour in terms of additional consumption required to give individuals the same
utility they would achieve if they behaved optimally. The life-cycle model we use is
characterised by both income and life-time uncertainty. Both the model and the
approach used to solve the dynamic optimisation problem numerically are standard in
the literature, following Deaton (1991) and Carroll (1992, 1997). Our approach is
related to earlier work on near-rational behaviour in intertemporal consumption and
saving problems by Cochrane (1989) and Lettau and Uhlig (1999).
Rules of thumb have been analysed in life-cycle saving models before, in particular in
tests of the life-cycle/permanent income hypothesis in the macroeconomics literature.
Starting with the seminal paper by Hall (1978), a series of studies assume that a fraction
of the population behaves according to some simple rule of thumb, such as Ôjust
consume your current income in every periodÕ whereas the rest of the population
behaves optimally.5 Estimates of the fraction of rule-of-thumb consumers in the
population range between zero and well above 50% and depend heavily on assump-
tions about householdsÕ preferences and econometric estimation approaches. In this
article, we explore such a simple consumption-equals-income rule and two other saving
rules that have been used in the economics literature on life-cycle saving behaviour.
More recently, Scholz et al. (2006) studied saving behaviour using data from a sample
of older American households. They show that observed saving decisions are closer to
the solution of an intertemporal optimisation model than they are to two simple rules
of thumb (one assumes a constant saving rate, whereas the second is based on age and
income specific average saving rates). Also related to our research question is a recent
study by Calvet et al. (2007) who quantify the welfare losses that result from sub-optimal
portfolio choices (which are perhaps driven by rules of thumb as well) using Swedish
data.
We should point out that we use the term rule of thumb for any decision rule that is
not the solution to an underlying utility maximisation problem and easy to derive and
apply for individuals with limited computational capabilities. This terminology is con-
sistent with the use in behavioural economics and psychology, as discussed by Goodie
et al. (1999). In most of the economics literature, the term Ôrule of thumbÕ is typically
5
The macroeconomics literature on rules of thumb includes papers by Flavin (1981), Hall and Mishkin
(1982), Cochrane (1989), Campbell and Mankiw (1990) and Weber (2000, 2002).
used in a much more narrow sense, namely, for households that spend a fixed fraction
or all of their income in every period (Deaton, 1992b; Browning and Crossley, 2001).6
This is only one of the rules of thumb we consider. As will become clear below, the
other rules of thumb that we analyse are forward-looking and therefore much closer to
the life-cycle framework than the Ôfixed consumptionÕ rule of thumb usually postulated
in economics.
The remainder of this article is structured as follows. In Section 1, we present a
version of the standard life-cycle model of saving decisions which allows for both life-
time and income uncertainty. Next, we describe three saving rules which can be used in
this framework (Section 2). In Section 3, we simulate and compare saving decisions
based on the these saving rules. Section 4 concludes.
1. The Benchmark Life-Cycle Saving Model

We assume that an individual’s or household’s optimal life-cycle consumption and
saving behaviour can be derived from a well-defined intertemporal optimisation
problem, given additively separable preferences with constant exponential dis-
counting. We use a version of the standard life-cycle model with borrowing con-
straints and both life-time and income uncertainty which is an extension of the
model proposed by Carroll (1992, 1997). In the remainder, it is understood that the
decision-making unit is the household even though we usually refer to individual
decisions.
Households are assumed to maximise, at each discrete point s in time, the
expected discounted stream of utility from future consumption. The per-period utility
function is denoted by u(Cs), to be specified below. Future utility is discounted by a
factor (1 þ q)1, where q is the time preference rate. The interest rate is denoted by
r. The maximum age a person can reach is T and we define sts as the probability of
being alive in period t conditional on being alive in period s. To simplify notation, we
also use a binary random variable that indicates whether an individual is alive in
period t:

1 if the individual is alive in period t,
St ¼
0 if the individual is not alive in period t.
The individual’s intertemporal optimisation problem can be stated as follows. In the

planning period s, the maximisation problem is given by:
X
T
max Es ð1 þ qÞst sts uðCt Þ s.t.; ð1Þ
fCt gTt¼s t¼s
Atþ1 ¼ ð1 þ r ÞðAt þ Yt Ct Þ; ð2Þ
At 0 8t ¼ s . . . T ; ð3Þ
6
An exception is Deaton (1992a) who specifies a more complicated rule of thumb; see the discussion of
Rule 3 in Section 2.
Maximisation of expected discounted utility given by (1) is subject to a standard law of
motion for assets At (2) and a borrowing constraint (3). Note that while the
household will optimally hold zero assets at the end of terminal period T, the
individual might die before T with non-zero assets, that is, there are accidental
bequests in our model.
Income Yt is determined by an exogenous process which is widely employed in the
life-cycle literature (Carroll, 1992, 1997; Cocco et al., 2005). This process features a
deterministic hump-shaped pattern as well as both permanent and transitory shocks. In
particular, income Yt can be written as
Yt ¼ St Pt Vt ; ð4Þ
where Pt is the permanent income component and Vt is the transitory shock. Recall that
the random variable St reflects life-time uncertainty and takes the value 1 as long as the
individual is alive while it is set to zero thereafter. The permanent income component
Pt itself follows a random walk with a drift
Pt ¼ Gt Pt1 Nt ; ð5Þ
where Gt incorporates the deterministic and exogenously fixed hump-shaped life-cycle

pattern of income and Nt is the permanent shock. Both transitory shocks Vt and
permanent shocks Nt are i.i.d. log-normally distributed random variables with unit
expectation and variance r2v and r2n , respectively.
Finally, we assume that the within-period utility function is of the CRRA type,
Ct1c
uðCt Þ ¼ ; ð6Þ
1c
where c is the coefficient of relative risk aversion (and the inverse of the intertemporal
elasticity of substitution).
As in any model of intertemporal decision making, the individual’s decisions can be
described by a time-invariant decision rule, that is, a mapping from states into actions.
In the life-cycle saving model, such a decision rule will be a function Ct ¼ Ct(At, Yt) that
maps current assets and current income into saving decisions. As noted before, we take
the decision rule given by the dynamic programming solution to the intertemporal
optimisation problem as a benchmark. All other decision rules (i.e. any function that
maps states into actions) are interpreted as rules of thumb or heuristics. In the next
Section, we present three such rules.
Before we analyse how rules of thumb perform relative to the benchmark solution,
we conclude this Section by briefly sketching how the solution to the intertemporal
optimisation model given by (1)–(5) can be computed. While there is no closed-form
solution, the optimal allocation of consumption over time is characterised by the
following first-order condition:
1þr t
u 0 ðCt Þ ¼ s Et ½u0 ðCtþ1 Þ: ð7Þ
1 þ q tþ1
This is a modified version of the well-known standard Euler equation in which next
period’s expected marginal utility is weighted with the conditional probability of being
alive in period t þ 1. While the intuition of Euler equations such as (7) – balancing
marginal utility across periods – is clear that, in general, there is no closed-form
solution which would allow individuals to compute their optimal consumption decision
in each period. Rather, every consumer has to solve the entire life-time optimisation
model by backwards induction in each decision period. As noted by many authors
before, this procedure is computationally demanding and we can safely assume that
individuals do not actually solve this problem when making their consumption and
saving decisions (Hey and Dardanoni, 1988). Moreover, Pemberton (1993, p. 5) points
out that the intuition behind the Euler equation does not help us find simpler
behavioural rules that would generate as if behaviour.
To solve the intertemporal optimisation problem for the case with implicit borrowing
constraints numerically, we apply the cash-on-hand approach by Deaton (1991) in the
version developed by Carroll (1992). Cash-in-hand denoted by Xt, is the individual’s
current gross wealth (total current resources), given by the sum of current income and
current assets,
Xtþ1 ¼ ð1 þ r ÞðXt Ct Þ þ Ytþ1 : ð8Þ
To reduce the number of state variables we follow the approach in Carroll (1992)
and standardise all variables by the permanent income component. The solution to the
optimisation problem is then computed by backwards induction over value functions
starting in the last period in life T, where households consume all remaining wealth.
Taking this into account the optimal behaviour in period T 1 can be computed and
so on.
From a more technical point of view, we employ equally spaced grids for both
standardised cash-on-hand and standardised savings. Piecewise Cubic Hermite Inter-
polation is used for points not on the grid. Shocks to the income process are
approximated by Gauss–Hermite quadrature.
2. Three Rules of Thumb for Life-Cycle Saving Decisions

In this Section, we present three decision rules that allow individuals to make their life-
cycle saving decisions. Such rules of thumb might be used by individuals, which are
either unwilling or unable to compute optimal decision rules, such as those derived in
the previous Section. In Section 3, we use these decision rules to simulate intertem-
poral consumption-saving decisions. All these decision rules have been proposed in
previous literature. Table 1 contains an overview.
The first decision rule is the standard Ôconsume your current incomeÕ rule by
Keynes (1936). The second rule corresponds to Friedman’s (1957) Ôpermanent
incomeÕ decision rule. The third rule is taken from Deaton (1992a). As we explain
below, Deaton designed this rule with the explicit goal that it should be easy to
compute but still match optimal behaviour closely. All the rules of thumb we con-
sider are relatively easy to compute, although some might seem to be quite involved.
Most importantly, however, these rules do not require using backwards induction.
Each rule provides a closed-form solution for current consumption given expecta-
tions about future income (i.e. given survival probabilities and the expected path of
future income).
Table 1
Alternative Decision Rules in a Life-Cycle Savings Model
Decision rule Description Source

Benchmark Solution to the underlying intertemporal Carroll (1992), Rodepeter
optimisation problem and Winter (1998)
Rule 1 Consumption equals current income Related to Keynes (1936)
Rule 2 Consumption equals permanent income Friedman (1957)
Rule 3 Consumption equals cash on hand up to Deaton (1992a)
mean income, plus 30% of excess income
2.1. Rule of Thumb No. 1 (Keynes)

The first rule of thumb we consider is the simplest rule one can think of – just consume
your current income and do not save at all:
Ct ¼ Y t : ð9Þ
This rule is related to the famous consumption function by Keynes (1936). We
should note that a (macroeconomic) Keynesian consumption rule will typically allow
for some constant fraction of current income to be saved in each period. Such a rule
does not directly translate to the microeconomic life-cycle saving problem since it
lacks an element of dissaving (either in rainy days or in retirement).7 Here, we use
the extreme version that rules out any saving as a benchmark case. In the formal
analysis of life-cycle consumption and saving decisions, this rule has been used by
inter alia, Hall (1978), Flavin (1981) and Campbell and Mankiw (1990). As simple as it
is, this decision rule seems to be natural from a psychological perspective (Wärneryd,
1989).
2.2. Rule of Thumb No. 2 (Permanent Income)

The second decision rule is the permanent income rule proposed by Friedman (1957).
This rule is much more complicated than the Keynesian consumption rule but still
much easier to apply than the optimal decision rule. Friedman hypothesised that
consumption is a function of permanent income which is defined as that constant flow
which yields the same present value as an individual’s expected present value of actual
income. In Friedman’s original work, individuals use a weighted average of past income
to compute permanent income. In our simulations, we impose rational expectations
about future income so that we can compute permanent income based on the
realisations of calibrated income processes. Specifically, we start with the identity
X
T
YtP ð1 þ r Þti ¼ At þ Ht ; ð10Þ
i¼t
where YtP is permanent income as of period t, At are current assets and Ht is the present
value of (non-asset) income given by
7
We return to the issue of whether rules of thumb could be modified or combined to account for different
savings motives and stochastic shocks in the concluding Section.
" #
X
T
Ht ¼ Yt þ E Yi ð1 þ r Þti : ð11Þ
i¼tþ1
Assuming that individuals consume their permanent income in every period, and
re-arranging these identities, we obtain the Ôpermanent incomeÕ decision rule,
r 1
Ct ¼ YtP ¼ ðAt þ Ht Þ: ð12Þ
1 þ r 1 ð1 þ r ÞðT tþ1Þ
Setting the interest rate to zero for the moment, this reduces to
1
Ct ¼ YtP ¼ ðAt þ Ht Þ: ð13Þ
T t
Here, one can see that individuals distribute their (expected) total wealth equally
over their remaining life-time, smoothing consumption but not insuring themselves
against utility losses from negative income shocks as in the life-cycle model presented in
Section 1. However, individuals update their expectations about future realisations of
the income process. If the stochastic component shows persistence or follows a random
walk, permanent income reflects all past and current shocks.
Note that in the absence of income uncertainty (or in the case of certainty equiva-
lence), there is no need for precautionary saving, and this rule of thumb corresponds
to the solution of the underlying optimisation problem (if one further ignores time
preference). In the life-cycle model with income uncertainty presented in Section 1,
the permanent income rule deviates from the benchmark solution. However, as Pem-
berton (1993) argues, this rule is both forward-looking and easy to compute. Therefore,
it might be reasonable to assume such a decision rule for individuals which are Ôfar-
sighted rather than myopicÕ and whose Ôconcern is for ÔÔthe futureÕÕ rather than with a
detailed plan for the futureÕ (p. 7, emphasis in the original). Pemberton refers to the
underlying concept as Ôsustainable consumptionÕ. Our simulations allow us to evaluate
how such forward-looking behaviour performs relative to the benchmark solution.
2.3. Rule of Thumb No. 3 (Deaton)

Deaton (1992a) considers a static consumption rule which is relatively easy to compute.
It is much simpler than the permanent income rule, but it is not forward-looking. As
Deaton states, his goal was to approximate the solution of the underlying optimisation
problem (a life-cycle model similar to ours) with a rule that Ôshould be simple, simple
enough to have plausibly evolved from trial and errorÕ (p. 257). Deaton assumes that
individuals consume cash on hand, Xt, as long as cash on hand is less than expected
income. If the income realisation exceeds expected income, individuals save a constant
fraction, f, of excess income (and consume the rest right away). Formally, Deaton’s
decision rule can be written as:
8
< Xt if Yt Et ðYt Þ and Xt Et ðYt Þ
Ct ¼ Et ðYt Þ if Yt Et ðYt Þ and Xt > Et ðYt Þ :
:
Et ðYt Þ þ f½Yt Et ðYt Þ if Yt > Et ðYt Þ
Next, we follow Deaton in setting this fraction to 30%. Deaton explicitly states that he
specified this decision rule, including the choice of f ¼ 30%, entirely ad hoc. The
intriguing feature of this rule is that while being based on just easy-to-compute
expected income, it approximates the optimal solution quite well in Deaton’s
application. We will show below that this is also true in our slightly more involved life-
cycle model.
2.4. Saving Motives Reflected by Simple Rules of Thumb

Before we turn to simulating life-cycle saving decisions using these three rules of
thumb, it is useful to briefly review the central motives for saving that operate in our
benchmark life-cycle model. Table 2 contains an overview of how these saving motives
are reflected in the three simple decision rules that we consider. It is important to
recognise that decision rules differ in the motives for saving they reflect. As we will see
in the next Section, in some specifications of the individual’s stochastic environment,
not all saving motives are relevant. This implies that there is no universal ranking of
these decision rules in terms of their usefulness to individuals.
In a world with uncertainty about the length of life and stochastic income, the
intertemporal optimisation problem of Section 1 is designed to capture both risk
aversion (i.e. consumption smoothing over time) and precautionary motives (i.e. self-
insurance against negative income shocks). Among our three rules of thumb, Deaton’s
rule is the only one which allows for a precautionary saving motive, while the Ôcon-
sumption equals incomeÕ rule includes no saving motive at all. The permanent income
rule is forward-looking in the sense that individuals use their expectations about future
income, and in the case of persistent shocks also information about past and current
shocks, in their consumption and saving decisions. Therefore, this rule reflects the
consumption smoothing motive of saving.
3. Simulation and Evaluation of Rule-of-Thumb Behaviour

In this Section, we present simulation results and compute the utility losses associated
with using three alternative rules of thumb relative to the benchmark solution, taking
preferences as given. More specifically, to compare utility losses across different
decision rules, we compute a CE measure, that is, the percentage increase in
consumption needed in every period and every state which would provide an individual
with the same expected life-time utility under a given behavioural rule as they would
obtain had they solved the underlying optimisation problem. In a different context,
this measure of welfare loss has been used by Krueger and Ludwig (2007).
Table 2
Savings Motives Captured by Alternative Decision Rules
Benchmark Rule 1 Rule 2 Rule 3

Consumption smoothing Yes No Yes No
Precautionary saving Yes No No Yes
Life-time uncertainty Yes No No No
3.1. Calibration of the Life-Cycle Model and Welfare Measure
In Table 3, we report the benchmark parameter values used to calibrate the model. We
chose these values to be in the range of values typically used in the literature.8 The
parameter values in the income process have been taken from Cocco et al. (2005). The
authors report estimates of the deterministic, hump-shaped life-cycle profile as well as
the estimated variance of permanent and transitory shocks. Furthermore, they differ-
entiate between three education groups: households whose head does not have a high
school degree, whose head has a high school degree, or whose head has a college
degree, respectively. We stratify our simulations by these three education groups. This
approach allows us to analyse how the performance of a rule of thumb is affected by the
characteristics of the income process faced by the household. Moreover, stratification
by education addresses some of the heterogeneity in saving and income insurance that
has been documented by Blundell et al. (2008), inter alia.
To quantify the welfare loss associated with using rules of thumb instead of the
optimal solution to the maximisation problem, we compute a CE measure. To be
precise, we compute the percentage increase in consumption a household following a
given rule of thumb would need in each period and in each state to have the same
expected life-time utility as if it was using the optimal solution. As the within-period
utility function is of the CRRA form, there is a closed-form solution for this measure
once the expected life-time utility has been computed under both the rule of thumb
and the rational behaviour.
The analytic solution for the CE can be derived in the following way. Under CRRA
preferences, the expected life-time utility of a household following a given rule of
thumb EUROT is given by
" #
XT ROT 1c
t 0 ðCt Þ
ROT
EU ¼E ð1 þ qÞ st ; ð14Þ
t¼0
1c
where st0 is the ex ante probability of being alive in period t. The CE is then defined as
( )
XT ROT
t 0 ½Ct ð1 þ CEÞ1c
E ð1 þ qÞ st ¼ EUOPT ; ð15Þ
t¼0
1 c
, EUROT ð1 þ CEÞ1c ¼ EUOPT ; ð16Þ

OPT
where EU is the expected life-time utility of a household who has solved the
maximisation problem. Rearranging (16) yields the closed-form solution for the CE
measure:
OPT 1=ð1cÞ
EU
CE ¼ 1: ð17Þ
EUROT
To obtain the CE we thus simulate the discounted life-time utility of 100,000
households both under the optimal solution and under the given rule of thumb. The
8
The parameter values used in this study also lie in the range of parameter values obtained using structural
estimation by Gourinchas and Parker (2002) and Cagetti (2003).
Table 3
Parameter Values Used for Calibration of the Life-Cycle Model
Parameter Benchmark value

Preference parameters
Relative risk aversion coefficient c 3
Rate of time preference q 4%
Interest rate r 2%
Life-time parameters
Conditional survival probabilities sts Life-table values
Retirement age 65
Maximum age at death* T 100
Income process (i.i.d.)
Variance of transitory shock r2e;IID 0.25
Variance of permanent shock r2u;IID 0
Income process (random walk) No high school High school College
Variance of transitory shocky r2e;RW 0.1056 (0.0080) 0.0738 (0.0034) 0.0584 (0.0045)
Variance of permanent shocky r2u;RW 0.0105 (0.0011) 0.0106 (0.0004) 0.0169 (0.0006)
Notes. *In specifications without life-time uncertainty, the age of death is fixed at 80. yParameters values are
from Cocco et al. (2005). Standard errors in parentheses.
expected life-time utility under each rule is then computed as the average simulated
discounted life-time utility. Equation (17) then gives the CE measure directly.9
3.2. Simulation Results

In the following, we first show, for each of the three education groups, the age profiles of
consumption spending and asset holdings that arise for the alternative decision rules and
in three different scenarios: income certain, life length certain; income stochastic with
i.i.d. shocks, life length uncertain; and income stochastic as a random walk, life length
uncertain. We then summarise the implied utility losses that arise, expressed as CEs.
Figure 1 shows consumption spending (left column) and asset holdings (right col-
umn) for the three educational groups (from top to bottom: no high school, high
school, college) when there is no uncertainty – the income path is deterministic and
the length of life is known. In this case, realised income is equal to expected income in
every period and so Deaton’s Rule 3 collapses to Rule 1, the Keynesian rule. Both rules
imply that households consume their deterministic income in every period. Thus, the
dashed line that depicts consumption of these households in the graphs in the left
column also shows the deterministic income path.
In the absence of impatience (i.e. interest rate r equal to discount rate q), behaviour
according to Rule 2 (Permanent Income) would be identical to the optimal behaviour
of the rational consumer. However, in the benchmark parameterisation, households
9
As an alternative welfare measure, we also computed the percentage change in income which would give
households the same expected utility under a given rule of thumb as under the optimal behaviour. In this
case, however, there is no closed-form solution since households adjust their behaviour as income changes.
The computation is therefore more involved. But under the rules of thumb considered here, households
spend most (or even all) of their income anyway, in particular in earlier periods of life. The two welfare
measures (percentage increase in consumption and in income, respectively) thus turn out to be almost
identical.
Panel (a): No High School
45 100
40 90
80
35 70
Consumption
30 60
Assets
50
25
40
20 30
20
15
10
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age
Panel (b): High School

45 100
40 90
80
35 70
Consumption
30 60
Assets
50
25 40
20 30
20
15
10
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age
Panel (c): College

45 100
90
40
80
35 70
Consumption
30 60
Assets
50
25 40
20 30
20
15 10
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age
Fig. 1. Behaviour Due to Different Decision Rules: Income Certain, Life Length Certain
Notes. The Figure plots the behaviour generated by the different decision rules: rational behaviour
(solid line), Keynes Rule 1 and Deaton Rule 3 (dashed line), Permanent Income Rule 2 (dash-
dotted line)
are impatient, so optimal behaviour (solid line) shifts more consumption to earlier
periods in life than permanent income rule (dash-dotted line) would suggest. Note also
that since households are borrowing constrained, Rule 2 households consume less than
their permanent income early in life. Finally, due to impatience, rational households
would like to have a decreasing consumption profile over their life cycle; however, due
to the borrowing constraint, these households cannot consume more than their
income early in life.
The differences between education groups are also interesting. With increasing
education, the income profile becomes steeper. The implications are straightforward.
Marginal utility of consumption is high for low levels of consumption at the beginning
of life and lower for higher consumption levels later in life. For given education and
income profiles, consumption levels are identical for all rules as long as households are
still borrowing constrained. Between households that follow different rules, differences
in consumption arise only later in life and these differences are more pronounced the
higher education and the steeper the income profile.
The right column of graphs in Figure 1 shows asset holdings for the three education
groups and alternative decision rules. As noted above, Rule 1 and 3 (Keynesian and
Deaton) households consume their current income in every period so they do not
build up any assets. Asset holdings are larger for Rule 2 (permanent income) house-
holds than for rational households since the latter prefer earlier consumption because
of their impatience.
In Figure 2, we introduce uncertainty: income follows a stochastic process with a
deterministic component and i.i.d. shocks, and the length of life is uncertain. In this
environment, rules of thumb can now show whether they succeed in helping house-
holds to insure against income shocks and life-time uncertainty in the absence of
insurance markets. As noted above, we simulated realisations of the stochastic income
process for 100,000 households; the graphs in Figure 2 show averages across these
simulations, again by education group and decision rule. The graphs in the left column
again show consumption spending, those in the right column show asset holdings.
Note first that differences in consumption and, thus, also in asset holdings arise
already early in life when households are borrowing constrained, in contrast to the
deterministic income case in Figure 1. This is because some households realise positive
shocks which are, depending on the decision rule used, not consumed fully. Also for
this reason, consumption decisions implied by Rule 1 (Keynesian) and Rule 3 (Deaton)
are now different. Another difference to the earlier results is that because of uncertain
lifetime, households effectively become more impatient and impatience increases with
age. Thus, consumption is shifted towards younger ages. In particular, rational con-
sumers save less, so their behaviour moves closer to that of Rule 1 and 3 households;
conversely, it differs more from that of Rule 2 (permanent income) households. The
asset holdings reflect these effects of uncertainty.
A peculiar feature of Rule 3 is that it does not allow for asset decumulation once
households are retired. This drawback of this rule is due to its simplicity. One could
also imagine that households who follow this rule switch to a different one after
retirement. However, we refrain from modelling rule switching in this article.
Finally, Figure 3, which is similarly structured, shows the average age profile of
consumption expenditure and asset holdings when income follows a random walk so
45 200
180
40
160
35 140
Consumption
Assets
30 120
100
25 80
20 60
40
15
20
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age

45 200
180
40
160
35 140
Consumption
Assets
30 120
100
25 80
20 60
40
15
20
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age
Panel (c): College

45 200
180
40
160
35 140
Consumption
Assets
30 120
100
25 80
20 60
40
15 20
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age
Fig. 2. Mean Behaviour Due to Different Decision Rules: Income I.I.D., Life Length Uncertain
Notes. The Figure plots the mean behaviour generated by the different decision rules (calculated
from 100,000 simulations): rational behaviour (solid line), Keynes Rule 1 (dashed line), Perma-
nent Income Rule 2 (dash-dotted line), Deaton Rule 3 (dotted line)
50 200
45 180
40 160
140
Consumption
35 120
Assets
30 100
25 80
60
20
40
15 20
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age

50 200
45 180
160
40
140
Consumption
35 120
Assets
30 100
25 80
60
20
40
15 20
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age
Panel (c): College

50 200
45 180
160
40
140
Consumption
35 120
Assets
30 100
25 80
60
20
40
15 20
10 0
20 30 40 50 60 70 20 30 40 50 60 70
Age Age
Fig. 3. Mean Behaviour Due to Different Decision Rules: Income Random Walk, Life Length
Uncertain
Notes. The Figure plots the mean behaviour generated by the different decision rules (calculated
from 100,000 simulations): rational behaviour (solid line), Keynes Rule 1 (dashed line), Perma-
nent Income Rule 2 (dash-dotted line), Deaton Rule 3 (dotted line)
that shocks have much more severe consequences. The main implication is that
rational households save more and have larger asset holdings than in the i.i.d. case. For
the three rules of thumb, the differences are minor.
We now turn to the central question of this article: how does rule-of-thumb behav-
iour translate into utility losses, relative to using the optimal decision rule? Table 4
shows utility losses, expressed as CEs, for a variety of scenarios. More specifically, the
numbers in the Table are the percentage increases in consumption in each period that
would compensate an individual for using a non-optimal decision rule rather than
solving the corresponding intertemporal optimisation problem.
In Panel (a), we report results using the benchmark calibration described in Table 3.
When there is no income risk, CEs and thus utility losses are generally very low (<1%).
The CE is lowest for college graduates since their income profile is the steepest. This
implies that their marginal utility is very high early in life relative to low marginal utility
later in life when their consumption level is very high. But differences between the
decision rules only occur once households are no longer credit constrained and their
consumption level is already high. Hence, college graduates require only a small
increase in consumption in the early periods to compensate for the relatively low utility
losses later in life. The CEs are relatively high for households in the middle education
group that follow Rules 1 and 3: these households face a large drop in income at
retirement which is not smoothed by these rules. These results provide a first important
insight: utility losses implied by rules of thumb depend on the income profile, both in
early life and around retirement, and they are not necessarily ranked by education.
Table 4
Life-Time Utility Loss from Using Alternative Decision Rules
No high school High school College

Length
Income of life Rule 1 Rule 2 Rule 3 Rule 1 Rule 2 Rule 3 Rule 1 Rule 2 Rule 3
Panel (a): benchmark calibration

Certain Certain 0.2 0.7 0.2 0.9 0.4 0.9 0.1 0.2 0.1
Certain Uncertain 0.0 3.0 0.0 0.4 2.7 0.4 0.0 1.1 0.0
i.i.d. Certain 24.7 4.3 6.9 21.8 6.7 6.2 15.9 8.3 3.7
i.i.d. Uncertain 24.2 5.0 6.4 21.1 7.1 5.6 15.6 8.5 3.5
Random Certain 19.9 2.8 11.7 14.2 2.7 9.8 14.4 4.3 11.4
Random Uncertain 16.9 3.4 8.8 11.0 3.6 6.5 11.1 4.2 8.2
Panel (b): high risk aversion (c ¼ 9)
Certain Certain 1.5 0.1 1.5 1.7 0.0 1.7 0.0 0.0 0.0
Certain Uncertain 0.9 1.1 0.9 1.2 0.4 1.2 0.0 0.0 0.0
i.i.d. Certain 40.5 32.4 32.0 32.9 30.4 29.2 4.4 4.0 2.0
i.i.d. Uncertain 39.4 32.1 32.0 32.8 30.2 29.2 4.4 4.0 2.0
Random Certain 179.2 55.5 141.1 140.3 46.8 135.1 170.8 72.0 162.4
Panel (c): high variance of permanent shocks (þ3 standard errors)
Random Certain 26.4 4.8 17.9 16.3 3.3 11.7 17.3 5.4 14.2
Notes. Life-time utility losses are expressed as the percentage increase in consumption in each period and state
that would compensate an individual for using non-optimal decision rules rather than solving the corre-
sponding intertemporal optimisation problem. Parameter values for the benchmark model are reported in
Table 3.
There are also differences in CEs depending on whether the length of life is
uncertain. As mentioned above, life-time uncertainty effectively makes agents more
impatient. Hence, rational households shift more consumption to earlier periods in
life. This implies that rational behaviour differs more from the consumption profile
implied by the permanent income rule, and thus the utility loss of following this rule
increases.
Next, consider the case with stochastic income (the third and fourth lines of Panel
(a)). Rule 1 (Keynesian) households do poorly now since not smoothing transitory
income shocks is very costly: their CE goes up to between 15% (high education) and
25% (low education). CEs for the other two rules also increase relative to the certain
income case, but they remain below 10% for all education groups. Deaton’s Rule 3 does
quite well – this is exactly the environment for which it was designed. Rule 2 (perma-
nent income) does relatively poorly for high education households whose income
profile is steep so that borrowing constraints bind longer. Whether lifetime is uncertain
or not does not have a major effect.
The random walk case (the last two lines of Panel (a)) is perhaps most realistic. Even
though income shocks now have permanent effects, CEs are still relatively low (mostly
below 10%), in particular for Rules 2 and 3. Here, the permanent income rule does
well because it takes shocks into account by updating permanent income. However,
there is no smoothing of small shocks early in life because of borrowing constraints, so
utility losses relative to optimal behaviour still arise. Only the Keynesian Rule 1 does
poorly, even though CEs are smaller than in the i.i.d. case.
Before we turn to a more general interpretation of these results, we briefly discuss
Panels (b) and (c) of Table 4. In these simulations, we use alternative parameter
values to investigate the sensitivity of our results with respect to householdsÕ risk
aversion and the variance of the income process. In Panel (b), we increase risk
aversion by setting the coefficient c equal to 9 rather than 3. If risk aversion is higher,
the curvature of the utility function is larger, so the effects of steep income profiles
are more pronounced. As long as income is certain, high-school graduates that follow
any of the rules of thumb come close to optimal behaviour in utility terms. For the
other education groups, CEs tend to increase when risk aversion increases for most
decision rules but they remain low. When income is uncertain, higher risk aversion
affects rule-of-thumb households quite negatively unless they have very steep income
profiles. For the low and middle education groups, CEs go up considerably in the
i.i.d. case. Finally, when income follows a random walk, high risk aversion households
fare very poorly under any rule of thumb; Rule 2 (permanent income) does slightly
better.
In Panel (c), we increase the variance of the permanent income shock in the random
walk case by 3 standard errors. These numbers should be compared with those in the
last two lines of Panel (a). It is apparent that increasing income variance makes rules of
thumb perform less well; CEs increase across the board. However, they typically remain
below 20% which is perhaps surprisingly low given that the income process has now a
rather large variance. As before, Rule 2 (permanent income) performs best in the
random walk case since it updates permanent income after each shock.
We can draw several conclusions from the numbers in Table 4. In the case of
income certainty, life-time utility losses resulting from following some rule of thumb
rather than solving the underlying intertemporal optimisation problem are relatively
small. When income is uncertain, individuals who follow rules of thumb in their
consumption and saving decisions suffer considerable utility losses relative to the
optimal decision rule. Not surprisingly, the magnitudes of these utility losses depend
on preferences (here, we studied risk aversion) and the specific structure of the
income process. There are also many cases in which rules of thumb do not imply
substantial utility losses and rules of thumb which are simpler than others (such as
Deaton’s rule) do not necessarily perform worse. Unless risk aversion is very high
(q ¼ 9), utility losses as expressed by CEs are below 20% and the permanent income
rule stays even below 10%.
The main conclusion from this comparison is that there is considerable variation in
the life-time utility loss associated with using rules of thumb. There is no uniformly
best rule of thumb, and for most stochastic environments analysed in this article,
there is some rule of thumb which yields relatively small utility losses (<10% of life-
time income). In the case of uncertain length of life and a random walk income
process, however, utility losses are substantial for all rules of thumb. When risk
aversion is very high, using a rule of thumb is a particularly bad idea. This observa-
tion is interesting since one might speculate that in real life, households with high
risk aversion might also be those who are financially less sophisticated and thus more
likely to use rules of thumb. Similarly, high education households which might be
more likely to use sophisticated savings strategies would actually perform relatively
better using rules of thumb than low-education households since their income
profiles are steeper.
What makes a rule of thumb perform well? Based on our results, we conclude that
the key factor that makes a rule of thumb successful is its ability to generate a measure
of life-time income that correctly reflects movements in future income. If the life-time
income process exhibits a strong deterministic trend and modest shocks with low
persistence, this might not be too difficult. We discuss some implications of this finding
in the concluding Section.
4. Conclusions
In our simulations of life-cycle consumption and saving decisions under three heuristic
decision rules, we found that losses in total life-time utility can, in general, be sub-
stantial compared with optimal behaviour (i.e. using the solution of the underlying
intertemporal optimisation problem for given standard life-cycle preferences). The
magnitudes of these losses vary with the assumptions about preference parameters and
the properties of the income process. An important result is that for most environ-
ments we simulated, there is some simple rule of thumb which results in only modest
utility losses (expressed by compensating variations, these losses are equivalent to be-
tween 5% and 10% of life-time income).
We conclude this article with a discussion of possible extensions of our analysis and
implications for future research in the life-cycle consumption and saving framework.
The saving rules we analysed in this article reflect two distinct motives for saving
(consumption smoothing and precautionary saving). A natural extension of our
approach would be to combine two or more of these rules of thumb. For example,
individuals could use a simple static rule, such as Deaton’s rule (Rule 3) to insure
themselves against adverse income shocks and, at the same time, they could do some
consumption smoothing by using forward-looking rules, that is, the permanent income
rule (Rule 2) which focuses on income uncertainty. By combining several rules of
thumb, individuals should be able to improve their total life-time utility considerably
and they might actually come quite close to using the solution to the underlying
optimisation problem in utility terms. The concept of mental accounting introduced to
the life-cycle saving literature by Shefrin and Thaler (1988) is closely related to the
notion of using multiple rules of thumb. For instance, mental accounting implies that
individuals make saving decisions with different time horizons and saving goals in
mind. Combining rules of thumb with different objectives, such as those analysed in
this article, with a mental accounting framework, appears to be a fruitful direction for
future research.
Unfortunately, combining two or more saving rules in our framework would com-
plicate the analysis considerably, both technically and conceptually. The optimisation
and simulation problem is obviously much more involved but more importantly, one
would have to make assumptions of how individuals allocate funds to different saving
rules. This allocation might change over the life cycle. This would imply that a second
class of behavioural decision rules would have to be specified for the allocation of
saving into different mental accounts.
It would also be interesting to compare the welfare consequences of suboptimal
savings decisions (the subject of this article) with those of suboptimal portfolio choices,
as in Calvet et al. (2007) when both are driven by rules of thumb. To perform such a
comparison, both simulation models would have to be specified and calibrated con-
sistently, or perhaps even embedded in a unified model. Such an analysis would allow
us to address the question of whether it is more important to use good rules of thumb
for the consumption-saving decision or for the decision of how to invest whatever
amount is saved.
Another important research question that we have not addressed in this article is:
how do rules of thumb actually arise? How do individuals decide which behavioural
decision rule they use? In our analysis, we have taken the rules of thumb as
exogenously given because our main objective was to evaluate the utility loss asso-
ciated with using some heuristic rather than computing the optimal solution to the
underlying decision problem. We did not model the choice between the optimal
strategy and using rules of thumb. Rule-of-thumb behaviour could be derived from
some meta problem if the cost of computing the solution to the underlying life-
cycle optimisation problem was taken into account.10 However, such an approach
might run into the conceptual problem of an infinite regress, as discussed by
Conlisk (1996).
A promising approach is to explore how rules of thumb arise endogenously from
learning behaviour; for example, Lettau and Uhlig (1999) investigate a model of
learning rules of thumb in intertemporal decision problems. However, in a life-cycle
saving setting, learning from own mistakes is impossible. Every life-cycle decision is
10
Computation costs have been considered in models with rule-of-thumb behaviour by Shi and Epstein
(1993) and Hindy et al. (1997).
made only once – all decisions are conditional on the planning period (i.e. on age)
and cannot be repeated with a different ÔtrialÕ decisions in the future. Therefore,
learning in the life-cycle saving problem is likely to be based on social interactions with
other people (family, neighbours or friends). Brown et al. (2009) present evidence for
the relevance of learning: in an experimental study with artificial life-cycle decisions,
subjects saved too little initially but learned to save optimally as life cycles were
repeated. In the domain of entrepreneurial financial decisions, Drexler et al. (2010)
conducted a field experiment with micro-entrepreneurs in the Dominican Republic
and found that simple rule-of-thumb training improved business practices and out-
comes. An important question for future research thus seems to be how individuals
acquire the rules they use for making financial decisions and whether policy inter-
ventions can help individuals to improve the rules they use. This research agenda is
obviously related to the ongoing discussion of how householdsÕ financial literacy can
be improved.
Certain institutions might also provide individuals with saving rules so that they do
not need to figure out optimal or heuristic saving rules themselves. An important
example is social security which replaces the need for discretionary long-term saving to
some extent. Another mechanism that provides rules for long-term saving is housing
expenditure. In countries, such as the US and the UK, many households buy their first
family homes relatively early in their life cycle and this decision determines a large
fraction of their consumption and saving pattern over future years. In Germany, due to
its favourable tax treatment, the acquisition of life-insurance policies with substantial
saving components during the early stage of the active working life has been quite
common in the past (Sauter and Winter, 2010). The acquisition of a family home or a
life-insurance policy is a one-time decision that fixes a substantial part of life-cycle
saving and it reduces the scope for discretionary saving over remaining years substan-
tially. There is also some recent evidence that individuals are quite willing to follow
saving rules provided by institutional arrangements (Thaler and Benartzi, 2004; Choi
et al., 2005). The behavioural patterns mentioned in this paragraph (based on direct
imitation, social traditions or institutional arrangements) can be interpreted as fol-
lowing a heuristic saving rule and they might result in decisions that are quite close to
optimal life-cycle behaviour.
Any empirical analysis of rule-of-thumb saving behaviour would have to account
for the possibility that individuals are heterogeneous with respect to the decision
rules they use. Once we give up the fiction of optimal behaviour based on the
solution to a (unique) underlying optimisation problem, the result that all individuals
follow the same decision rule does not need to hold any more. Some individuals
might care more about short-term precautionary saving, some for long-term con-
sumption smoothing, while others might rely on saving rules provided by institu-
tional arrangements. As noted above, combinations of rules might arise as well. To
our knowledge, there exist no econometric studies that try to identify individual
decisions rules in a life-cycle saving context, but this is clearly an important area for
future research that could build on results from laboratory experiments on decision
rules in dynamic problems by Müller (2001), Houser and Winter (2004) and, more
recently, Hey and Lotito (2009), Hey and Knoll (2010) and Hey and Panaccione
(2011).
Based on our results on the performance of simple saving rules, we would argue that
an important direction for research on life-cycle consumption and saving behaviour is
how social learning and institutional factors generate simple decision rules, and to test
whether individuals actually use such rules with micro data on household consumption
and saving.
University of Munich
BMW Museum, Munich
References
Anderhub, V., Güth, W., Müller, W. and Strobel, M. (2000). ÔAn experimental analysis of intertemporal
allocation behaviorÕ, Experimental Economics, vol. 3, pp. 137–52.
Attanasio, O.P. and Weber, G. (2010). ÔConsumption and saving: models of intertemporal allocation and their
implications for public policyÕ, Journal of Economic Literature, vol. 48, pp. 693–751.
Binswanger, J. and Carman, K.G. (2011). ÔHow real people make long-term decisions: the case of retirement
preparationÕ, unpublished manuscript, Tilburg University.
Blundell, R., Pistaferri, L. and Preston, I. (2008). ÔConsumption inequality and partial insuranceÕ, American
Economic Review, vol. 98, pp. 1887–921.
Brown, A.L., Chua, Z.E. and Camerer, C.F. (2009). ÔLearning and visceral temptation in dynamic saving
experimentsÕ, Quarterly Journal of Economics, vol. 124, pp. 197–231.
Browning, M. and Crossley, T.F. (2001). ÔThe life-cycle model of consumption and savingÕ, Journal of Economic
Perspectives, vol. 15(3), pp. 3–22.
Browning, M. and Lusardi, A. (1996). ÔHousehold saving: micro theories and micro factsÕ, Journal of Economic
Literature, vol. 34, pp. 1797–855.
Cagetti, M. (2003). ÔWealth accumulation over the life cycle and precautionary savingsÕ, Journal of Business and
Economic Statistics, vol. 21, pp. 339–53.
Calvet, L.E., Campbell, J.Y. and Sodini, P. (2007). ÔDown or out: assessing the welfare costs of household
investment mistakesÕ, Journal of Political Economy, vol. 115, pp. 707–47.
Camerer, C. (1995). ÔIndividual decision makingÕ, in (J.H. Kagel and A.E. Roth, eds.), Handbook of Experimental
Economics, pp. 587–703, Princeton, NJ: Princeton University Press.
Campbell, J.Y. and Mankiw, G.N. (1990). ÔPermanent income, current income, and consumptionÕ, Journal of
Business and Economic Statistics, vol. 8, pp. 265–79.
Carbone, E. and Hey, J.D. (1999). ÔA test of the principle of optimalityÕ, Theory and Decision, vol. 50, pp. 263–
81.
Carroll, C.D. (1992). ÔThe buffer stock theory of saving: some macroeconomic evidenceÕ, Brookings Papers on
Economic Activity, vol. 2, pp. 61–135.
Carroll, C.D. (1997). ÔBuffer-stock saving and the life cycle-permanent income hypothesisÕ, Quarterly Journal of
Economics, vol. 112, pp. 1–55.
Choi, J.J., Laibson, D., Madrian, B.C. and Metrick, A. (2005). ÔPassive decisions and potent defaultsÕ, in
(D.A. Wise ed.), Analyses in the Economics of Aging, pp. 59–78, Chicago and London: Cambridge University
Press.
Cocco, J.F., Gomes, F.J. and Maenhout, P.J. (2005). ÔConsumption and portfolio choice over the life cycleÕ,
Review of Financial Studies, vol. 18, pp. 491–533.
Cochrane, J.H. (1989). ÔThe sensitivity of tests of the intertemporal allocation of consumption to near-rational
alternativesÕ, American Economic Review, vol. 79, pp. 319–37.
Conlisk, J. (1996). ÔWhy bounded rationality?Õ, Journal of Economic Literature, vol. 34, pp. 669–700.
Deaton, A. (1991). ÔSaving and liquidity constraintsÕ, Econometrica, vol. 59(5), pp. 1221–48.
Deaton, A. (1992a). ÔHousehold saving in LDCs: credit markets, insurance and welfareÕ, Scandinavian Journal
of Economics, vol. 94, pp. 253–73.
Deaton, A. (1992b). Understanding Consumption, Oxford: Clarendon Press.
Drexler, A., Fischer, G. and Schoar, A. (2010). ÔKeeping it simple: financial literacy and rules of thumbÕ,
Discussion Paper No. 7994, CEPR, London.
Flavin, M.A. (1981). ÔThe adjustment of consumption to changing expectations about future incomeÕ, Journal
of Political Economy, vol. 89, pp. 974–1009.
Friedman, M. (1957). A Theory of the Consumption Function, Princeton, NJ: Princeton University Press.
Gigerenzer, G. and Todd, P.M. (1999). ÔFast and frugal heuristicsÕ, in (G.Gigerenzer, P.M. Todd and the ABC
Research Group, eds.), Simple Heuristics That Make Us Smart, pp. 3–34, New York, NY and Oxford: Oxford
University Press.
Goodie, A.S., Ortmann, A., Davis, J.N., Bullock, S. and Werner, G.M. (1999). ÔDemons versus heuristics in
artificial intelligence, behavioral ecology, and economicsÕ, in (G. Gigerenzer, P.M. Todd, and the ABC
Research Group, eds.), Simple Heuristics That Make Us Smart, pp. 327–55, New York, NY and Oxford:
Oxford University Press.
Gourinchas, P.-O. and Parker, J.A. (2002). ÔConsumption over the life cycleÕ, Econometrica, vol. 70, pp. 47–89.
Hall, R.E. (1978). ÔStochastic implications of the life cycle-permanent income hypothesis: theory and evi-
denceÕ, Journal of Political Economy, vol. 86, pp. 971–87.
Hall, R.E. and Mishkin, F. (1982). ÔThe sensitivity of consumption to transitory income: estimates from panel
data on householdsÕ, Econometrica, vol. 50, pp. 461–81.
Hey, J.D. (2005). ÔDo people (want to) plan?Õ, Scottish Journal of Political Economy, vol. 52, pp. 122–38.
Hey, J.D. and Dardanoni, V. (1988). ÔOptimal consumption under uncertainty: an experimental investiga-
tionÕ, Economic Journal, vol. 98, pp. S105–16.
Hey, J.D. and Knoll, J.A. (2010). ÔStrategies in dynamic decision making: an experimental investigation of the
rationality of decision behaviourÕ, Unpublished manuscript, University of York.
Hey, J.D. and Lotito, G. (2009). ÔNaive, resolute or sophisticated? A study of dynamic decision makingÕ, Journal
of Risk and Uncertainty, vol. 38, pp. 1–25.
Hey, J.D. and Panaccione, L. (2011). ÔDynamic decision making: what do people do?Õ, Journal of Risk and
Uncertainty, vol. 42, pp. 85–123.
Hindy, A., Huang, C. and Zhu, S.H. (1997). ÔOptimal consumption and portfolio rules with durability and
habit formationÕ, Journal of Economic Dynamics and Control, vol. 21, pp. 525–50.
Houser, D. and Winter, J. (2004). ÔHow do behavioral assumptions affect structural inference? Evidence from
a laboratory experimentÕ, Journal of Business and Economic Statistics, vol. 22, pp. 64–79.
Johnson, S., Kotlikoff, L.J. and Samuelson, W. (2001). ÔCan people compute: an experimental test of the life-
cycle consumption modelÕ, in (L.C. Kotlikoff, ed.), Essays on Saving, Bequests, Altruism, and Life-Cycle
Planning, pp. 335–85, Cambridge, MA, and London, UK: MIT Press (first circulated in March 1987 as
NBER Working Paper No. 2183).
Keynes, J.M. (1936). The General Theory of Employment, Interest and Money, London: Macmillan.
Krueger, D. and Ludwig, A. (2007). ÔOn the consequences of demographic change for rates of returns to
capital, and the distribution of wealth and welfareÕ, Journal of Monetary Economics, vol. 54, pp. 49–87.
Laibson, D. (1998). ÔLife-cycle consumption and hyperbolic discount functionsÕ, European Economic Review, vol.
42(3-5), pp. 861–71.
Lettau, M. and Uhlig, H. (1999). ÔRules of thumb versus dynamic programmingÕ, American Economic Review,
vol. 89, pp. 148–74.
Loewenstein, G. (1992). ÔThe fall and rise of psychological explanation in the economics of intertemporal
choiceÕ, in (G. Loewenstein and J. Elster, eds.), Choice Over Time, pp. 3–34, New York: Russell Sage
Foundation.
Low, H., Meghir, C. and Pistaferri, L. (2010). ÔWage risk and employment risk over the life cycleÕ, American
Economic Review, vol. 100, pp. 1432–67.
Lusardi, A. and Mitchell, O.S. (2007). ÔBaby boomer retirement security: the roles of planning, financial
literacy, and housing wealthÕ, Journal of Monetary Economics, vol. 54, pp. 205–24.
Moon, P. and Martin, A. (1990). ÔBetter heuristics for economic search: experimental and simulation evi-
denceÕ, Journal of Behavioral Decision Making, vol. 3, pp. 177–93.
Müller, W. (2001). ÔStrategies, heuristics, and the relevance of risk-aversion in a dynamic decision problemÕ,
Journal of Economic Psychology, vol. 22, pp. 493–522.
Pemberton, J. (1993). ÔAttainable non-optimality or unattainable optimality: a new approach to stochastic life
cycle problemsÕ, Economic Journal, vol. 103, pp. 1–20.
Rabin, M. (1998). ÔPsychology and economicsÕ, Journal of Economic Literature, vol. 36, pp. 11–46.
Rodepeter, R. and Winter, J. (1998). ÔSaving decisions under life-time and earnings uncertainty: empirical
evidence from West German household dataÕ, Discussion Paper No. 98-58, Sonderforschungsbereich 504,
Universität Mannheim.
Rodepeter, R. and Winter, J. (1999). ÔRules of thumb in life-cycle saving decisionsÕ, Discussion Paper No.
99-81, Sonderforschungsbereich 504, Universität Mannheim.
Sauter, N. and Winter, J. (2010). ÔDo investors respond to tax reform? Evidence from a natural experiment in
GermanyÕ, Economics Letters, vol. 108, pp. 193–6.
Scholz, J.K., Seshadri, A. and Khitatrakun, S. (2006). ÔAre Americans saving ÔÔoptimallyÕÕ for retirement?Õ,
Journal of Political Economy, vol. 114, pp. 607–43.
Shefrin, H.M. and Thaler, R.H. (1988). ÔThe behavioral life-cycle hypothesisÕ, Economic Inquiry, vol. 26, pp.
609–43.
Shi, S. and Epstein, L.G. (1993). ÔHabits and time preferenceÕ, International Economic Review, vol. 34, pp. 61–84.
Simon, H.A. (1955). ÔA behavioral model of rational choiceÕ, Quarterly Journal of Economics, vol. 69, pp. 99–118.
Thaler, R.H. (1994). ÔPsychology and savings policiesÕ, American Economic Review, Papers & Proceedings, vol.
84(2), pp. 186–92.
Thaler, R.H. and Benartzi, S. (2004). ÔSave more tomorrow: using behavioral economics to increase employee
savingÕ, Journal of Political Economy, vol. 112, pp. S164–87.
van Rooij, M.C.J., Lusardi, A. and Alessie, R. (2011). ÔFinancial literacy and stock market participationÕ, Journal
of Financial Economics, vol. 101, pp. 449–72.
Wärneryd, K.-E. (1989). ÔOn the psychology of saving: an essay on economic behaviorÕ, Journal of Economic
Psychology, vol. 10, pp. 515–41.
Weber, C.E. (2000). ÔRule-of-thumb consumption, intertemporal substitution, and risk aversionÕ, Journal of
Business and Economic Statistics, vol. 18(4), pp. 497–502.
Weber, C.E. (2002). ÔIntertemporal non-separability and ÔÔrule-of-thumbÕÕ consumptionÕ, Journal of Monetary
Economics, vol. 49, pp. 293–308.

Rules of Thumb

Uploaded by

Copyright:

Available Formats

Rules of Thumb

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rules of Thumb

Uploaded by

Copyright:

Available Formats

The Economic Journal, 122 (May), 479–501. Doi: 10.1111/j.1468-0297.2012.02502.x. Ó 2012 The Author(s).

The Economic Journal Ó 2012 Royal

RULES OF THUMB IN LIFE-CYCLE SAVING DECISIONS*

Joachim K. Winter, Kathrin Schlafmann and Ralf Rodepeter

1. The Benchmark Life-Cycle Saving Model

The individual’s intertemporal optimisation problem can be stated as follows. In the

Atþ1 ¼ ð1 þ r ÞðAt þ Yt Ct Þ; ð2Þ

where Gt incorporates the deterministic and exogenously fixed hump-shaped life-cycle

2. Three Rules of Thumb for Life-Cycle Saving Decisions

Decision rule Description Source

2.1. Rule of Thumb No. 1 (Keynes)

2.2. Rule of Thumb No. 2 (Permanent Income)

2.3. Rule of Thumb No. 3 (Deaton)

2.4. Saving Motives Reflected by Simple Rules of Thumb

3. Simulation and Evaluation of Rule-of-Thumb Behaviour

Benchmark Rule 1 Rule 2 Rule 3

, EUROT ð1 þ CEÞ1c ¼ EUOPT ; ð16Þ

Parameter Benchmark value

3.2. Simulation Results

Panel (b): High School

Panel (c): College

Panel (b): High School

Panel (c): College

Panel (b): High School

Panel (c): College

No high school High school College

Panel (a): benchmark calibration

You might also like