IZA DP No.

2993
Choi ce Over Ti me
Paol a Manzi ni
Mar co Mar i ot t i
D
I
S
C
U
S
S
I
O
N

P
A
P
E
R

S
E
R
I
E
S
For schungsi nst i t ut
zur Zukunf t der Ar bei t
Inst i t ut e f or t he St udy
of Labor
August 2007

Choice Over Time



Paola Manzini
Queen Mary, University of London
and IZA

Marco Mariotti
Queen Mary, University of London




Discussion Paper No. 2993
August 2007



IZA

P.O. Box 7240
53072 Bonn
Germany

Phone: +49-228-3894-0
Fax: +49-228-3894-180
E-mail: [email protected]





Any opinions expressed here are those of the author(s) and not those of the institute. Research
disseminated by IZA may include views on policy, but the institute itself takes no institutional policy
positions.

The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center
and a place of communication between science, politics and business. IZA is an independent nonprofit
company supported by Deutsche Post World Net. The center is associated with the University of Bonn
and offers a stimulating research environment through its research networks, research support, and
visitors and doctoral programs. IZA engages in (i) original and internationally competitive research in
all fields of labor economics, (ii) development of policy concepts, and (iii) dissemination of research
results and concepts to the interested public.

IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion.
Citation of such a paper should account for its provisional character. A revised version may be
available directly from the author.
IZA Discussion Paper No. 2993
August 2007









ABSTRACT

Choice Over Time
*

In the last twenty years a growing body of experimental evidence has posed a challenge to
the standard Exponential Discounting Model of choice over time. Attention has focused on
some specific anomalies, notably preference reversal and declining discount rates, leading
to the formulation of the model of hyperbolic discounting which is finding increasing favour in
the literature. In this paper we provide a survey of both the theoretical modelling and the
experimental evidence relating to choice over time. As we will show, a careful analysis of the
mapping between theoretical models and experimental investigations raises questions as to
whether some of the most focused upon anomalies should be indeed classified as such, or
whether they are really the most challenging ones for conventional theory. New
developments are emerging both at the theoretical and empirical level, opening up new
exciting avenues for investigation


JEL Classification: C91, D9

Keywords: time preference, choice over time, theory, experiments


Corresponding author:

Marco Mariotti
Department of Economics
Queen Mary, University of London
Mile End Road
London E1 4NS
United Kingdom
E-mail: [email protected]







*
We wish to thank Steffen Andersen, Glenn Harrison, Michele Lombardi, Efe Ok, Andreas Ortmann,
Daniel Read for useful comments and guidance to the literature. We are also grateful to the ESRC for
their financial support through grant n. RES000221636. Any error is our own.
1 Introduction
Many economic decisions have a time dimension. Hence the need to describe howoutcomes
available at future dates are evaluated by individual agents. The history of the search for
a rational model of preferences over (and choices between) dated outcomes bears some
interesting resemblances and dissimilarities with the corresponding search in the eld of
risky outcomes. First, a standard and widely accepted model was settled upon. This is
the exponential discounting model (EDM) (Samuelson [50]), for which the utility from
a future prospect is equal to the present discounted value of the utility of the prospect.
That is, an outcome r available at time t is evaluated now, at time t = 0, as o
t
n(r),
with o a constant discount factor and n an (undated) utility function on outcomes. So
according to the EDM, r at time t is preferred now to at time : if
o
t
n(r) o
s
n()
Similarly a sequence of timed outcomes r
1
, r
2
, ... r
T
is preferred to another sequence
1
,

2
, ...
T
if
T
X
i=1
o
i1
n(r
i
)
T
X
i=1
o
i1
n(
i
)
Subsequently, an increasing number of systematic anomalies were demonstrated in ex-
perimental settings. This spurred the formulation of more descriptively adequate non-
exponential models of time preferences.
This mirrors the events for the standard model of decision under risk, the Expected
Utility model, in which case observed experimental anomalies led to the formulation of
Non-Expected Utility models. However, unlike the case of choice between risky outcomes,
for choice over time no normative axioms of rationality were formulated which had the
same force as, say, the von Neumann-Morgenstern Independence axiom of utility theory.
Perhaps for this reason, economists have been readier to accept one specic alternative
model, that of hyperbolic discounting.
In this chapter we review both the theoretical modelling and the experimental evidence
relating to choice over time. Most of the space is devoted to choices between outcome-
date pairs, which has been better studied, especially experimentally, but in section 4 we
also discuss choices between time sequences of outcomes. In the next section we examine
the axiomatic foundation for models based on discounting, exponential or otherwise. In
section 3 we review the new breed of models emerged as a response to experimental
observations. Section 5 looks in more detail at the empirical evidence, while section 6 is
devoted to evaluating the explanatory power of the various theories. Section 7 concludes.
2
2 Axiomatics of exponential discounting for outcome-
date pairs
We begin by describing a basic axiomatisation of exponential discounting for outcome-
date pairs due to Fishburn and Rubinstein [11]. This will help us in giving a sense of the
type of EDM violations that one may expect to observe in practice.
We should make clear at the outset that we follow the standard economic approach
of taking preferences (as revealed by binary choices), as the primitives of the analysis.
Any utility emerging from the analysis will simply describe the primitive preferences
in a numerical form. We are not, therefore, considering experience utility (i.e. the
psychological benet one gets from experience) as a primitive, an approach which is more
typical in the psychology literature. Also, we focus on time preferences as if the agent
can commit to them: this is in order to avoid a discussion of the thorny issue of time-
consistency
1
, which would deserve a treatment on its own.
Let A _
+
, with 0 A, represent the set of possible outcomes (interpreted as gains,
with 0 representing the status quo), and denote by 1 _
+
the set of times at which an
outcome can occur (with t = 0 1 representing the present). Unless specied, 1 can
be either an interval or a discrete set of consecutive dates.
A time-dependent outcome is denoted as (r. t): this is a promise, with no risk attached,
to receive outcome r A at date t 1. Let < be a preference ordering on A 1. The
interpretation is that < is the preference expressed by an agent who deliberates in the
present about the promised receipts of certain benets at certain future dates.
As usual, let ~ and ~ represent the symmetric and asymmetric components, respec-
tively, of <. Fishburn and Rubinstein [11]s characterisation uses the following axioms:
2
Order: < is reexive, complete, and transitive.
Monotonicity: If r then (r. t) ~ (. t).
Continuity: (r. t) : (r. t) < (. :) and (r. t) : (. :) < (r. t) are closed sets.
Impatience: Let : < t. If r 0 then (r. :) ~ (r. t), and if r = 0 then (r. :) ~ (r. t).
Stationarity: If (r. t) ~ (. t + t
0
) then (r. :) ~ (. : + t
0
), for all :. t 1 and t
0

such that : + t
0
. t + t
0
1.
The rst four axioms alone guarantee that preferences can be represented by a real-
valued utility function n on A1 with the natural continuity and monotonicity property
1
Initiated by Strotz [54].
2
Fishburn and Rubinstein [11] consider the general case where the outcome can involve a loss as well
as a gain, that is x < 0, and they do not require that 0 X. Here we focus on the special case only to
simplify the exposition.
3
(that is, n is increasing in r and decreasing in t, and it is continuous in both arguments
when 1 is an interval). The addition of Stationarity allows the following restrictions:
Theorem 1 (Fishburn and Rubinstein, 1982) If Order, Monotonicity, Continuity,
Impatience and Stationarity hold, then, given any o (0. 1), there exists a continuous and
increasing real-valued function n on A such that
(r. t) < (. :) =o
t
n(r) _ o
s
n(t)
In addition, n(0) = 0, and if A is an interval, then n is unique (for a given o) up to
multiplication by a positive constant.
The representation coincides formally with exponential discounting, but note well the
wording of the statement. One may x the discount factor o arbitrarily to represent
a given preference relation that satises the axioms, provided the utility function n is
calibrated accordingly. In other words, for any two discount factors o and o
0
, there exist
two utility functions n and such that (n. o) preferences in the representation of theorem 1
are identical to (. o
0
) preferences in the same type of representation. In order to interpret
o as a uniquely determined parameter expressing impatience one would need an external
method to x n. This is an important observation, often neglected in applications, which
naturally raises the question about what then exactly is impatience here. Benoit and Ok
[3] deal with this question by proposing a natural method to compare the delay aversions
of time preferences, analogous to methods to compare the risk aversion of preferences over
lotteries. As they show, in the EDM it is possible that the delay aversion of a preference
represented by (n. o) is greater than that represented by (. o
0
) even though o o
0
.
Moreover, given the uniqueness of n only up to multiplication by constants, and the
positivity of n for positive outcomes, an additive representation (at least for strictly
positive outcomes) is as good as the exponential discounting representation. That is,
taking logs and rescaling utilities by dividing by log o, one could write instead
(r. t) < (. :) =n(r) t _ n() :
Coming back to the axioms, Continuity is a standard technical axiom. Order is a
rationality property deeply rooted in the economic theory of choice. Cyclical preferences
for example are traditionally banned from economic models. Monotonicity and impatience
are also universally assumed in economic models, which are populated by agents for whom
more of a good thing is better, and especially for whom a good thing is better if it comes
sooner: certainly these are reasonable assumptions in several contexts, though as we shall
see, not in others.
4
Stationarity, however, does not appear to have a very strong justication, either from
the normative or from the positive viewpoint. So it should not be too surprising to
observe in practice violations of this axiom and in fact, as we shall see later in some
detail, plenty of them have been recorded. What is surprising is rather the willingness of
economists to have relied unquestionably for so many years on a model, the EDM, which
takes stationarity for granted. Indeed Fishburn and Rubinstein themselves explicitly
state that we know of no persuasive argument for stationarity as a psychologically viable
assumption ([11], p. 681). This led them to consider alternative separable representations
that do not rely on Stationarity. One assumption (which is popular in the theory of
measurement) is the following:
Thomsen separability: If (r. t) ~ (. :) and (. :) ~ (.. t) then (r. :) ~ (.. :).
This allows a dierent representation result:
Theorem 2 (Fishburn and Rubinstein, 1982) If Order, Monotonicity, Continuity,
Impatience and Thomsen separability hold, and A is an interval, then there are continuous
real-valued functions n on A and o on 1 such that
(r. t) < (. :) =o (t) n(r) _ o (:) n()
In addition, n(0) = 0 and n is increasing while o is decreasing and positive.
This is therefore an axiomatisation of a discounting model, in which the discount
factor is not constant. However, while Thomsens separability is logically much weaker
than stationarity and it is useful to gauge the additional strength needed to obtain a
constant discount factor, one may wonder how intuitive or reasonable a condition it is
itself. One might not implausibly argue, for example, that if exactly ( r) is needed
to compensate for the delay of (: t) in receiving r, and if exactly (. ) is needed
to compensate for the delay of (t :) in receiving , then exactly (. r) is needed to
compensate for the delay of (: :) in receiving r. This argument does not seem to us
introspectively much more cogent than stationarity,
3
though it permits the elegant and
exible representation of theorem 2.
It should be clear from the above results and discussion that the EDM for outcome-
date pairs is best justied on the basis of its simplicity and usefulness in applications.
Violations especially of the stationarity aspect of it are to be expected, and while they
have captured most of the attention, it is perhaps violations of other properties, such as
Order, which would appear to be more intriguing, striking as they do more directly at
the core of traditional thinking about economic rationality.
3
Fishburn and Rubinstein [11] also provide a dierent argument for Thomsen separability, based on
an independence condition when the domain of outcomes is enriched to include gambles.
5
3 Recent models for outcome-date pairs
3.1 Hyperbolic discounting
As we mentioned already, over the past twenty years or so a body of empirical evidence has
emerged documenting that actual behaviour consistently and systematically contradicts
the predictions of the standard model. As we discuss more fully in section 5, various
exponential discounting anomalies have been identied.
4
As we explain further in section
6, in a sense some of these are not anomalies at all: they do not violate any of the axioms in
the theorems above, but only make specic demands on the shape of the utility function.
Among those that do violate the axioms in the representations, one particular eect has
captured the limelight: preferences are rarely stationary, and people often exhibit a strict
preference for immediacy. Decision makers may be indierent between some immediate
outcome and a delayed one, but in case they are both brought forward in time, the formerly
immediate outcome loses completely its attractiveness. More formally, if r and are two
possible outcomes, situations of the type described above can be summarised as
(r. 0) ~ (. t) and (. t + t) ~ (r. t)
Note that this violates jointly four of the ve axioms in the characterization of theorem
1, with the exception of Impatience. Let r
0
,= r be such that (r
0
. 0) ~ (. t) (such
an r
0
exists by Continuity). It must be that r
0
< r (for otherwise if r
0
r then by
Monotonicity (r
0
. 0) ~ (r. 0) ~ (. t) and by Order (r
0
. 0) ~ (. t)). By Stationarity
(r
0
. t) ~ (. t + t). Then by Monotonicity again (r. t) ~ (. t + t), a contradiction with
the observed preference.
However this is commonly interpreted as a straight violation of Stationarity, since
the latter is sometimes dened in terms of strict preference as well as indierence. It is,
however, compatible with the weaker requirement of Thomsen separability.
As a matter of fact, many researchers observing these phenomena do not pay at-
tention to any axiomatic system at all, preferring rather to concentrate directly on the
EDM representation itself (sometimes implicitly assuming a linear utility). In the EDM
representation the displayed preferences are written as
n(r) o
t
n() and o

n(r) < o
t+
n()
which is impossible for any utility function n and xed o.
4
For a survey of these violations, see Loewenstein and Thaler [29] or Loewenstein and Prelec [26]; for
a thorough treatement of issues concerning choice over time see Elster and Loewenstein [9].
6
This present time bias (immediacy eect) is a special case of what is known as pref-
erence reversal (or sometimes common ratio eect in analogy with Expected Utility
anomalies in the theory of choice under risk), expressed by the pattern:
(r. t) ~ (. :) and (. t + t) ~ (r. : + t)
Strictly speaking, as the agent is expressing preferences at one point in time (the
present), nothing is really reversed: the agent simply expresses preferences over dierent
objects, and these preferences happen not to be constrained by the property of station-
arity. The reason for the reversal terminology betrays the fact that often, especially in
the evaluation of empirical evidence, it is implicitly assumed that there is a coincidence
between the current preferences over future receipts (so far denoted <) and the future
preferences over the same receipts to be obtained at the same dates. In other words, now
dating preferences explicitly, (r. t) <
0
(. :) is assumed to be equivalent to (r. t) <

(. :),
where <

with t _ :. t is the preference at date t. If today you prefer one apple in one
year to two apples in one year and one day, in one year you also prefer one apple imme-
diately to two apples the day after. It is far from clear that this is a good assumption.
In this way, the displayed observed pattern can be taken as a reversal of preferences
during the passage of time from now to date t. Whether this is a justied interpretation
or not, the displayed pattern does contradict the EDM. This is, though, a somewhat soft
anomaly, in the sense that it does not contradict basic tenets of economic theory, and it
can be addressed simply by changes in the functional form of the objective function which
agents are supposed to maximize. Notably, it can be explained by the now popular model
of hyperbolic discounting (HDM)
5
(as well as by other models). In the HDM it is assumed
that the discount factor is a hyperbolic function of time.
6
In its general form, o : 1
is given as
o (t) = (1 + ct)

b
a
with c. / 0
In the continuous time case, in the limit as c approaches zero, the model approaches the
EDM, that is
lim
a!0
(1 + ct)

b
a
= c
bt
5
See for instance Phelps and Pollack [39], Loewenstein and Prelec [26], Laibson [22], Frederick, Loewen-
stein and ODonoghue [12].
6
For evidence documenting behaviour compatible with this functional form see for instance Ainslie
[1], Benzion, Rapoport and Yagil [4], Laibson [22], Loewenstein and Prelec [26] and Thaler [55]. It
is important to stress that Harrison and Lau [18] have argued against the reliability of the elicitation
methods used to obtain this empirical evidence. They argue that this evidence is a direct product of the
lack of control for credibility in experimental setting with delayed payment.
7
For any given / (which can be interpreted as the discount rate), c determines the departure
of the discounting function from constant discounting, and is inversely proportional to the
curvature of the hyperbolic function.
Hyperbolic discount functions imply that discount rates decrease over time. The
hyperbolic functional form captures in an analytically convenient way the idea that the
rate of time preference between alternatives is not constant but varies, and in particular
decreases as delay increases. So people are assumed to be more impatient for tradeos
(between money and delay) near the present than for the same tradeos pushed further
away in time. It can account for preference reversals.
This model ts in the representation of theorem 2 in section 2. Preference reversal
can be easily reconciled within an extension of the EDM, in which the requirement of
stationarity has been weakened to Thomsen separability.
The present time bias can be captured even more simply in the most widely used form
of declining discount model, the quasi-hyperbolic model or (,. o) model. In it, the rate
of time preference between a present alternative and one available in the next period is
,o, whereas the rate of time preference between two consecutive future alternatives is o.
Therefore (r. t) is evaluated now as n(r) if t = 0 and as ,o
t
n(r) if t 0, where , (0. 1]
(the case of , = 1 corresponds to exponential discounting). So we may have
n(r) ,on() and ,o
t+
n() ,o

n(r)
rationalising the present time bias. As we expand on further below, this same approach
can be applied in the case of sequences of outcomes (see section 4).
3.2 Relative Discounting
Ok and Masatlioglu [36] have recently proposed an interesting and challenging axiomatic
model which, though retaining a certain notion of discounting, dispenses with the usual
notion of evaluating future outcomes in terms of their present value. In their relative
discounting model (RDM), in other words, it is not possible in general to attribute a
certain value to outcome-date pairs (r. t) and state that the outcome-date pair with the
higher value is preferred. More precisely, their representation (axiomatized for the case
where the set of outcomes A is an open interval) is of the following type: there exists a
positive, real valued and increasing utility function n on outcomes and a relative discount
function o : 1 1 dened on date pairs such that
(r. t) % (. :) =n(r) _ o (:. t) n()
The relative discount function o is positive, continuous, decreasing in its rst argument
for any xed value of the second argument (with o (. t) = 0), and o (:. t) = 1,o (t. :).
8
The model is axiomatised in terms of a set of axioms which includes some weak (but
rather involved) separability conditions.
The authors own interpretation of the preference (r. t) % (. :) is that the worth at
time t of the utility of that is to be obtained at time : is strictly less than the worth
at time t of the utility of r that is to be obtained at time t. They argue that one of the
the main novelties of the RDM is that the comparison between the values of (r. t) and
(. :) is not made in the present but at time t or :. However it seems hard to tell when
a comparison between atemporal utilities is made. When comparing outcome-date pairs,
and not utilities, it is certainly at time 0 that the agent is making the comparison. So
one could as naturally say that the comparison between the utilities n(r) and n() is also
made at time 0, but instead of discounting the utility of the later outcome by the entire
delay with which it is to be received, it is only discounted by a measure of its relative
delay with the earlier outcome, whose utility is not discounted at all (psychologically this
corresponds to projecting the future into the present, which seems reasonable). While
this might appear a little like splitting hairs, the issue might become important if the
present agent were allowed to disagree with his later selves on the atemporal evaluation
of outcomes, that is on the function n to be used (in the existing model this disagreement
between current and future selves cannot happen, by an explicit assumption made on
preferences). A nal, and in our opinion appealing, interpretation of the model is as a
threshold model with an additive time-dependent threshold in which the term o (:. t) is
not seen as a multiplicative relative discount factor but just as a utility fee to be incurred
for an additional delay. In fact, just as we did for the EDM representation in section 2,
here, too, we can apply a logarithmic transformation to obtain a representation of the
type
(r. t) % (. :) =n() _ n(r) + o (:. t)
Whatever the interpretation, one virtue of the RDM is that it can explain some hard
anomalies, notably particular types of preference intransitivities (although no cycle within
a given time t is allowed - contrast this with the vague time preference model discussed
below). The relative discounting representation includes as special cases both exponential
and hyperbolic discounting. Therefore, beside intransitivities, it can also account for every
soft anomaly for which the HDM can account. In this sense the model is successful. On
the ip side of the coin, one might argue that it is almost too general, and many other
special cases are also included in it. For example, the subadditive discounting or similarity
ideas discussed in the next section can also be formulated in this framework.
A similar model has been studied independently by Scholten and Read [52], who call
it the discounting by interval model. Their interpretation, motivation and analysis is
however quite dierent from that of Ok and Masatlioglu [36]. In their model, the discount
9
function is dened on intervals of time, which is equivalent to dening it on pairs of dates
as for the RDM. But the authors argue for comparisons between alternatives to be made
by means of usual present values, for which the later outcome is rst discounted to the
date of the earlier outcome (using the discount factor which is appropriate for the relevant
interval) and then discounted again to the present (using the appropriate discount factor
which is appropriate for this dierent interval). So, formally: for : t,
(r. t) % (. :) =o (0. :) n() _ o (0. :) o (:. t) n(r) =n() _ o (:. t) n(r)
Scholten and Read do not axiomatise their model, but focus on interesting experimental
evidence suggesting some possible restrictions of the discounting function.
3.3 Similarities and subadditivity
While not proposing fully edged-models, two contributions by Read [42] and Rubinstein
([47], [48]), put forth some analytical ideas to interpret certain types of anomalies. We
consider the contributions by these two authors in turn.
Subadditivity. Read [42] suggest that a model of subadditive discounting might apply.
This means that the average discount rate for a period of time might be lower than the
rate resulting from compounding the average rates of dierent subperiods. Furthermore,
he suggests that the ner the partition into sub-periods, the more pronounced this eect
should be. Formally, if [0. 1] is a time period divided into the intervals [t
0
. t
1
] . .... [t
k1
. 1].
Let o
T
= exp
r
T
T
be the average discount factor for the period [0. 1] (where :
T
is the
discount rate for that period), and o
i
= exp
r
i
T
the average discount factor that applies
to the sub-period beginning at i (where :
i
is the discount rate for that period). Then,
if there is subadditivity, for any amount r available at time t
k
, and letting n denote an
atemporal utility function, we have that
n(r) o
T
n (r) o
0
o
1
... o
k1
More abstractly, this general idea could even be dened independently of the existence of
an atemporal utility function. Given preferences < on outcome-date pairs, if
(r. t
k
) ~ (r
k1
. t
k1
) ~ ... ~ (r
0
. 0)
and (r. t
k
) ~ (r
0
0
. 0)
subadditivity could be taken as implying that
r
0
0
r
0
10
It is important to note, though, that in the absence of further assumptions on preferences
the existence of a separable discount function is not guaranteed. The RDM discussed in
the previous section characterises subadditive discounting by o (t. :) o (t. :) o (:. :).
This is reminiscent of some empirical evidence for decisions under risk, according to
which the total compound subjective probability of an event is higher the higher the
number of sub-events into which the event is partitioned (e.g. Tversky and Koehler [58]).
Preferences for which discounting is subadditive may not be compatible with hyperbolic
discounting, that is, discount rates may be constant or increasing in time, contradicting
the HDM, while implying subadditivity. This is precisely the evidence found by Read
[42].
Similarity. Rubinstein ([47], [48]) argues that similarity judgements may play an im-
portant role when making choices over time (or under risk). He also shifts attention to the
procedural aspects of decision making. He suggests that a decision procedure he originally
dened for choices under risk (in Rubinstein [46]) can be adapted to model choices over
time, too. Let -
time
and -
outcome
be similarity relations (reexive and symmetric binary
relations) on times and outcomes respectively. So : -
time
t reads date : is similar to date
t and r -
outcome
reads outcome r is similar to outcome . Rubinstein examines the
following procedure to compare any outcome-date pairs (r. t) and (. :):
Step 1) If r _ and t _ :, with at least one strict inequality, then (r. t) ~ (. :).
Otherwise, move to step 2.
Step 2) If t -
time
:, :ot (r -
outcome
) and r then (r. t) ~ (. :). If r -
outcome
,
:ot (t -
outcome
:) and t < : then (r. t) ~ (. :).
If neither the premise in step 1 nor the premise in step 2 applies, the procedure is
left unspecied. Rubinstein used this idea to show how it serves well to explain some
anomalies, some of which run counter to the HDM as well as to the EDM.
Of course, once the broad idea has been accepted, many variations of this procedure
seem also plausible. For example Tversky [56] had suggested a lexicographic semiorder
procedure according to which agents rely on their ranking of the attributes of an alter-
native in a lexicographic way when choosing between dierent alternatives. The rst
attribute of each alternative is compared. If, and only if, the dierence exceeds some
xed threshold value then a choice is made accordingly. Otherwise, the agent compares
the second attribute of each alternative, and so on. Yet another procedure reminiscent of
Tverskys lexicographic semiorder is described in the next section.
7
7
Kahneman and Tversky [21], too, discuss the intransitivities possibly resulting from the editing
phase of prospect theory, in which small dierences between gambles may be ignored.
11
Finally, Rubinsteins [47] experiments show that precisely the same type of decision
situations that creates a diculty for the EDM may also be problematic for the HDM,
while they may be easily and convincingly accounted for by similarity-based reasoning.
He argues that, in this sense, the change to hyperbolic discounting is not radical enough.
3.4 Vague time preferences
Manzini and Mariotti [30] introduce the notion of vague time preferences as an applica-
tion of their general two-stage model of decision-making
8
. The starting consideration is
that the perception of events distant in time is in general blurred. Even when a decision
maker is able to choose between, say, an amount r of money now and an amount of
money at time t, it may be more dicult to confront the same type of alternatives once
these are both distant in time. This diculty in comparing alternatives available in the
future may blur the dierences between them in the decision makers perception. In other
words, the passage of time weakens not only the perception of the alternatives (which are
perceived, in Pigous famous phrase
9
on a diminished scale because of the defectiveness
of our telescopic faculty), but the very ability to compare alternatives with one another.
In the vague time preferences model, the central point is that the evaluation of a
time dependent alternative is made up of two main components: the pure time preference
(it is better for an alternative to be available sooner rather than later, and there exists a
limited ability to trade o outcome for time); and vagueness: when comparing dierent
alternatives, the farther they are in time, the more dicult it is to distinguish between
them.
For (r. t) to be preferred to (. :) on the basis of a time-outcome tradeo, the utility
of r may exceed the utility of by an amount which is large enough so that the individual
can tell the two utilities apart. The amount by which utilities must dier in order for the
decision maker to perceive the two alternatives as distinct is measured by a the positive
vagueness function o, a real valued function on outcomes. When the utilities dier by more
than o, then we say that the decision maker prefers the alternative yielding the larger
utility by the primary criterion. Formally the primary criterion consists of a possibly
incomplete preference relation on outcome-date pairs, represented by an interval order as
follows:
(r. t) ~ (. :) =n(r. t) n (. :) + o (. :)
where n is monotonic increasing in outcomes and decreasing in time. When neither alter-
8
See Manzini and Mariotti [31].
9
See Pigou [40].
12
native yields a suciently high utility, the decision maker is assumed to resort to some
additional heuristic in order to make his choice (secondary criterion). Since each alter-
native has a time and an outcome component, two natural heuristics are distinguished.
In the outcome prominence version, the decision maker will rst try and base his choice
on which of the two available ones is the greater outcome; and only if this comparison is
not decisive will he resolve his choice by selecting the earlier alternative. On the contrary,
in the time prominence version of the model, the decision maker rst compares the two
alternatives by the time dimension. If one comes earlier, then that is his choice; otherwise
he looks at the other dimension, the outcome, and selects based on which is higher.
Formally, let ~ be dened as in the display above, and let c ~ / if and only if neither
c ~ / nor / ~ c. Assume that 1 and 1 are the asymmetric and symmetric parts,
respectively, of a complete order on the set of pure outcomes A. Finally, let <

(with
~

and ~

the corresponding symmetric and asymmetric parts, respectively) denote a

complete preference relation (not necessarily transitive) on the set of alternatives (i.e.
outcome-date pairs) A 1, and let i = (r
i
. t
i
) A 1 for i c. /. Then the two
alternative models are as follows:
Outcome Prominence Model (OPM):
1. c ~

/ =
(a) c ~ / (Primary Criterion), or
(b) (c ~ /. r
a
1r
b
) or (c ~ /. r
a
1r
b
. t
a
< t
b
) (Secondary Criterion)
2. c ~

/ =(c ~ /. r
a
1r
b
. t
a
= t
b
)
Time Prominence Model (TPM):
1. c ~

/ =
(a) c ~ / (Primary Criterion), or
(b) (c ~ /. t
a
< t
b
) or (c ~ /. t
a
= t
b
. r
a
1r
b
) (Secondary Criterion)
2. c ~

/ =(c ~ /. r
a
1r
b
. t
a
= t
b
)
In its simplest specication, the (o. o)-model, there are just two parameters, with
o taken as the individuals discount factor (which embodies the pure time preference
component of preference), o a positive constant measuring the individuals vagueness,
and n assumed linear in outcome.
13
4 Preferences over sequences of outcomes
When it comes to sequences of outcomes available at given times, the standard exponential
discounting model still widely used is that introduced by Samuelson [50], whereby sequence
((r
1
. t
1
) . (r
2
. t
2
) . ... (r
T
. 1)) is preferred to sequence ((
1
. t
1
) . (
2
. t
2
) . ... (
T
. 1)) whenever
the present discounted utility of the former is greater than the present discounted utility
of the latter:
T
X
t=1
n(r
t
) o
t1

T
X
t=1
n(
t
) o
t1
Similarly to the case of outcome-date pairs, Loewenstein and Prelec [26] highlighted
that there exists a number of anomalies which cannot be accommodated within the stan-
dard framework. We will discuss these anomalies in greater detail in section 5, while here
we limit ourselves to present the functional form that Loewenstein and Prelec [26] intro-
duce to account for these phenomena. They propose that the utility of some sequence
r = ((r
1
. t
1
) . (r
2
. t
2
) . ... (r
T
. 1)) should be represented by
l (r) =
T
X
i=1
(r
i
) o (t
i
)
where o is a discount function assumed to be a generalised hyperbola, o (t) =
1
(1+at)
b
a
, as
in the general case of hyperbolic discounting we saw earlier, and is a value function on
which the following restrictions are imposed:
V1: the value function is steeper in the loss than in the gain domain:
(r) < (r)
V2: the value function is more elastic for losses than for gains:

v
(r) <
v
(r) for r 0, where
v
=
rJ (r)
(r)
V3: the value function is more elastic for outcomes that are larger in absolute magnitude:

v
(r) <
v
() for 0 < r < or < r < 0
Manzini, Mariotti and Mittone [32] pursue a dierent approach, where building on
Manzini and Mariotti [30] they postulate a theoretical model which extends the one
for outcome-date pairs to sequences. In order to rank monetary reward sequences, the
decision maker looks rst at the standard exponential discounting criterion. However,
preferences are incomplete, so that sequences are only partially ordered by the criterion.
Here too they are completed by relying on a secondary criterion. Sequence r is preferred
14
over another sequence if the discounted utility of r exceeds the discounted utility of
by at least o (). When sequences cannot be compared by means of discounted utilities,
the decision-maker is assumed to focus on one prominent attribute of the sequences. This
prominent attribute ranks (maybe partially) the sequences and allows a specic choice
to be made. This latter aspect of the model is in the spirit of Tversky, Sattath and
Slovic [60]s prominence hypothesis. The attribute may be context dependent, so that for
instance in the case of outcome-date pairs case, as we saw above, each alternative has two
obvious attributes that may become prominent, the date and the outcome.
We stress that, at a fundamental level, the only departure from the standard choice
theoretic approach is that the decision makers behavior is described by combining sequen-
tially two possibly incomplete preference orderings, instead of using directly a complete
preference ordering. In the case of monetary sequences we use the following represen-
tation for preferences. Let ~

denote the strict binary preference relation on the set of

alternatives (sequences) , where a typical sequence has the form i = (i
1
. i
2
. .... i
T
). For
given n, o, o with the usual meaning, and secondary criterion 1
2
:
For all c. / , we have c ~

/ if and only if
1. either
P
T
t=1
n(c
t
) o
t1

P
T
t=1
n(/
t
) o
t1
+ o (/), or
2.
P
T
t=1
n(c
t
) o
t1
_
P
T
t=1
n(/
t
) o
t1
+o (/),
P
T
t=1
n(/
t
) o
t1
_
P
T
t=1
n(c
t
) o
t1
+o (c),
and c1
2
/
The above obviously begs the question of which secondary criterion one should use.
This can be suggested by the empirical evidence available so we postpone examining this
issue further, to explore suggestions from data (see sections 5 and 6).
We should nally note that although positive discounting of some form or the other is
deeply ingrained in much economic thinking and in virtually all economic policy, the issue
of whether this is a justied assumption is open. Fishburn and Edwards [10] axiomatise,
in a discrete time framework, a discount neutral model of preferences over sequences
that dier at a nite number of periods. Their general representation takes the following
form:
c < / =
X
ft:a
t
6=b
t
g
n
t
(c
t
) _
X
ft:a
t
6=b
t
g
n
t
(/
t
)
where the n
t
are real-valued functions on an outcome sets A
t
that may possibly vary
with the date. The axioms they use for this model express conditions of order, continuity
sensitivity (every period can aect preference) and of course (given the additive form)
independence across periods. When it is also assumed that the outcome sets A
t
are the
same, further separability assumptions of a measure-theoretic nature allow the following
specialisation of the model:
15
c < / =
X
ft:a
t
6=b
t
g
o (t) n(c
t
) _
X
ft:a
t
6=b
t
g
o (t) n(/
t
)
where o (t) is a positive number for any period t. It is not required to be included in the
interval (0. 1), and therefore it is consistent with negative discount rates. Finally, a form
of stationarity yields a constant, but possibly negative discount rate model:
c < / =
X
ft:a
t
6=b
t
g
o
t1
n(c
t
) _
X
ft:a
t
6=b
t
g
o
t1
n(/
t
)
where o is a uniquely dened positive number.
5 Assessing empirical evidence
Our starting point has been to underline how some observed patterns of choice are irrec-
oncilable with the standard theoretical model. So far, in assessing the theories, we have
taken the empirical evidence at face value. However, a more rigorous assessment of the
reliability of the empirical evidence itself is called for.
Indeed, assessing time preferences is a non trivial matter. A common theme emerging
from the huge literature is that their reliable elicitation poses several methodological prob-
lems, and it results in vastly dierent ranges for discount factors estimates.
10
Although
a plethora of studies exist which elicit time preferences, these hardly have proceeded in
a highly standardised way. Many confounding factors occur from one study to another,
which hamper systematic comparisons to determine to what extent these dierences de-
pend on the elicitation methods themselves as opposed to other dierences in experimen-
tal design. Moreover, as we shall explain, some recent empirical advances even put into
serious question certain results of the traditional evidence.
5.1 Psychological eects
To begin with, there are two families of possible psychological eects which act as con-
founding factors in the evaluations of time preferences: hypothetical bias and aective
response. The rst term refers to the fact that a substantial proportion of experimental
subjects makes dierent choices when answering hypothetical questions as compared with
situations where the answer determines the reward of the responder. For instance, one
thing is to ask a subject how much he is prepared to pay for a cleaner environment in
the abstract, and quite another is to ask the same question as part of a policy document
10
See e.g. Table 1 in Frederick, Loewenstein and ODonoghue [12].
16
that is going to determine the amount of taxation.
11
Because of this, it would seem rea-
sonable to want to rely on experimental evidence arising from designs which are incentive
compatible, that is such as that the respondents reward for participation depends on the
answer he or she has given.
By aective response we refer to the emotive states that might be evoked when experi-
mental subjects have to evaluate the delayed receipt of a good or a service, as compared to
money. For instance, Loewenstein and Prelec [27] explain by a preference for improving
sequences the behaviour of a consistent proportions of decision makers that, after having
chosen a fancy French restaurant over a local Greek one, and prefer it sooner rather than
later, also choose the sequence (Greek dinner in one month and French dinner in two
months) over the opposite sequence (French dinner in one month and Greek dinner in two
months). This preference for increasingness can be motivated by savouring: a decision
maker might like to postpone a pleasant activity so as to enjoy the build up to it. As a
ip side to this, dread would be reduced by anticipating an unpleasant task and reducing
the time spent in contemplation of this unsavouring activity.
12
More generally one can
think of a plethora of potential relevant attributes of a sequence which might inuence
choice (see e.g. Read and Powell [44], who study subjects stated verbal motivation for
their choices). These aective responses do not only involve sequences, of course, e.g. in
the case of the choice of the optimal timing of a kiss to your favourite movie star. Goods
and services may possess characteristics which make them idiosyncratically attractive or
repulsive to respondents, and evoke feelings quite other than pure time preferences.
5.2 Soft anomalies
In addition to the psychological eects mentioned earlier, framing eects may be rather
substantial, too. Loewenstein [24] observed a delay/speed up asymmetry, i.e. a dier-
ence in the willingness to pay for anticipating receipt of a good and that to postpone it.
He showed that when subjects were asked to imagine that they owned a good (a video
recorder in the experiment) available in one year, they would be prepared to pay only
$54 on average in order to anticipate receipt, and obtain it now. On the other hand,
when asked to imagine they actually owned the videorecorder, subjects were asking on
average a compensation of $126 in order to delay receiving it for a year. Loewenstein
11
The literature on whether or not the payment of experimental subjects has an eect on response is
huge, see e.g. Plott and Zeiler [41], Read [43], Hertwig and Ortmann [20] and Ortmann and Hertwig
[37], just to cite a few. Cummings, Harrison and Rutstrm [8] have examined this in the context of the
type of dichotomous choices that are asked in time preferences elicitation, though in a dierent domain.
Manzini, Mariotti and Mittone [32] instead deal with the time domain.
12
An early formal model of this kind of eects has been proposed by Loewenstein [23].
17
interpreted this as a framing eect,
13
a purely psychological phenomenon. He conjectured
that if prompted to imagine that he owns a good that is immediately available, when
asked how much he would have to be paid to delay receipt of the good, a decision maker
frames the delay as a loss. If instead the decision maker is prompted to imagine that
he owns a good available at a later date, when asked how much he would be willing to
pay in order to anticipate collection he would frame this last occurrence as a gain. Note
that this type of results were found in both purely hypothetical scenarios as well as an
incentive compatible one. If, as in prospect theory,
14
losses count more than gains, then
there is an asymmetry in discount rates elicited from the two choice frames. Agents are
less willing to anticipate the gain than to postpone a loss, i.e. they are more patient for
speeding up than for delaying an outcome. As we explain more in detail in section 6,
however, these phenomena are not really a violation of standard discounting theorems, as
they only impose restrictions on the shape of the utility function.
While the delay/speed-up asymmetry refers to dierences in the implied discount rates
depending on the time when the good is available, the so called magnitude eect refers
to dierences in the implied discount rates between large and small outcomes. It was rst
reported by Thaler [55], who found that, in an hypothetical setting, on average subjects
were indierent between receiving $15 immediately and $60 in a year, and at the same
time indierent between receiving $3. 000 immediately and $4. 000 in a year. While the
rst choice (assuming linear utility) implies a 25% discount factor, the second implies a
much larger implicit discount factor, of 75%. Shelley [53] carried out a study of both
the delay/speed up asymmetry and the magnitude eect. She carried out a test for the
possible combinations of gain, loss and neutral frames with either a receipt or a payment.
For receipts, she found that implied discount rates are higher for small amounts ($40 and
$200) than for large amounts of money ($1000 and $5000), and for speed up than for
delay (time horizons considered were of 6 months and one year for the small amounts,
and 2 and 4 years for the large amounts). From an economists perspective, the problem
is that all these experiments were based on hypothetical choices, without real payments.
However, they have been replicated also with real monetary payments (see e.g. Pender
[38]). Similarly to the previous anomaly, this can be reconciled with the EDM.
We have already discussed one of the main phenomena that violates one of the axioms
(stationarity), namely preference reversal. Intriguingly, in addition to the direct prefer-
ence reversal we have considered, recently Sayman and ncler [51], have found evidence
of what they dub reverse time inconsistency, whereby subjects who prefer a smaller-
sooner reward when both options are in the future switch to the larger-later reward when
13
See Tversky and Kahneman [57].
14
See Kahneman and Tversky [21]
18
the smaller option becomes imminent.
Thaler [55] also observes evidence consistent with discount rates declining with the
time horizon. That is, subjects were asked questions of the following type: What is the
amount of money to be received at dates t
1
. t
2
. .... t
K
that would make you indierent to
receiving r now? The implied discount rates (assuming linear utility) were declining as
the dates increased (for example, they were 345% over a one-month horizon and 19% over
a ten-year horizon).
15
There is a certain air of unreality about these values, and we shall
say more about this aspect later, when we consider the issue of risk-aversion and of eld, as
opposed to hypothetical experiment, data. However, we emphasize now that even within
the realm of experimental observations within an assumed linear utility model, Read [42]
uncovers contrary evidence. Discount rates appear to be constant across three consecutive
eight-month periods. Rather, his evidence is consistent with subadditive discounting, as
discussed in section 3.3.
5.3 Source of data and other elicitation issues
Since our focus is on the rationality or otherwise of decision makers, we ought to con-
sider whether it is possible to reconcile economic theory with either experimental evidence
arising from experimental designs which are incentive compatible, or with empirical evi-
dence from eld data (which, using real-life choices, automatically avoid any worry about
incentive compatibility); and with data that involve only monetary outcomes.
While the discrepancies between observations, and the unrealistic values found, suggest
that some problems must be addressed in the elicitation procedures, the point is that
paying subjects is in itself not necessarily enough to produce reliable data. What an
incentive compatible elicitation mechanism must do to be dependable is to induce people
to reveal (what they perceive to be) their true evaluation of the good in question. Various
methods have been used in domains dierent from time. In fact, the literature on the
elicitation of home grown values for all sorts of goods is vast. Traditionally, experimenters
induced preferences (i.e. valuations of specic goods) in experimental subjects in order to
assess the validity or otherwise of a given theoretical model. As the interest has moved
towards assessing and eliciting subject preferences in choices among dierent goods, or
in their valuation for some goods, various mechanisms have been introduced to tease out
home grown preferences from experimental subjects.
The most popular methods relied upon in the literature on the elicitation of prefer-
ences other than time preferences are English auction, second price auction and Becker-De
15
See also Benzion, Rapoport and Yagil [4] for an example in the case of hypothetical choices and
Pender [38] for actual choices.
19
Groot- Marschak procedure (BDM). For each of them bidding ones true value is a domi-
nant strategy, and in many experimental settings instructions encourage bidders explicitly
to understand and learn the dominant strategy (see e.g. Rutstrm [49]).
Let us consider them in turn:
English (or ascending) auction: agents compete for obtaining a good. With the so
called clock implementation of the auction, the price of the good increases steadily
over time. As time passes participants can withdraw. When only one is left, he
wins the object and he alone pays the price at which he won.
Vickrey (i.e. second price sealed bid) auction: subjects submit a single bid, secretly
from all other participants. The one with the highest bid wins the object, but pays
only the second highest price. This is why it is strategically equivalent to the English
auction, since in the latter the winner is the one who stays when the second highest
bidder gives in.
BDM: this is also equivalent to the two previous auctions, although here bidders
play against a probability distribution, rather than other subjects. Because of this
the BDM procedure has the objectionable diculty that it introduces a probability
dimension to the problem. Subjects have to declare their willingness to pay for a
good. Then a price is drawn from a uniform distribution, and if this price is higher
than the willingness to pay, the agent gets nothing, whereas if it is lower, the agent
pays the price drawn (so for a winning bidder it is as if he put forward the highest
bid in a second price auction, with the price drawn playing the role of the second
highest bid from a ctitious bidder).
All the above are strategically equivalent: so would it make any dierence which one
is used in the lab?
In auctions with induced preferences (i.e. where subjects are told what their valuation
for a good is) Noussair, Roibin and Rueux [35] nd that Vickrey auctions are more
reliable in eliciting preferences than the BDM procedure. Again with induced values,
Garratt, Walker and Wooders [13] nd that in the comparison with the usual student
population, when using experienced e-bay bidders as experimental subjects, the dierence
between over and under bidding is no longer signicant (while the proportion of agents
bidding their value is indistinguishable from standard lab implementation with students).
On the other hand, when preferences are not induced (i.e. they are home grown),
Rutstrm [49] nds that (average) bids are higher in the second price auction than in
either BDM or rst price auctions. Moreover, as noted by Harrison [16], these elicitation
methods suer from serious incentive properties in the neighbourhood of the truth telling
20
dominant strategy: deviations may be cheap enough that experimental subjects do not
select the dominant strategy.
Although none of these auction methods has been applied until recently (see below) to
time preferences, the systematic discrepancies between alternative methods to elicit pref-
erences for goods suggest that dierent elicitation methods might also produce dierent
estimates when applied to the time domain. The most relied upon elicitation technique for
time preferences at the moment consists in asking a series of questions, in table format, of
the type Do you prefer: A) A today or B) A +r at time 1, where r is some additional
monetary amount which increases steadily (from a starting value of zero) as the subject
considers the sequence of questions (see Coller and Williams [6] and Harrison, Lau and
Williams [19]). A decision maker would start switching from selecting A to selecting B
from one specic choice onwards, making it possible to infer the discount factor.
16
This
table method has been used with additional variations, namely an additional piece of
information (e.g. giving for each choice the implicit annual discount/interest rate implied
by each choice and the prevalent market rate in the real economy) in order to reduce the
extent to which subjects anchor their choices to their own experience outside the lab and
unknown to the experimenters. Coller and Williams [6] found discount rates to be much
lower than previously found once this kind of censoring is taken into account.
A very recent experimental study by Manzini, Mariotti and Mittone [33] has made the
rst comparative analysis of the table method, the BDM and the Vickrey auction in a
choice over time setting. Preliminary results show a similarity of elicited values between
the latter two methods, but a marked dierence between them and the table method.
However, one must be aware that all choice experiments involving questions about
money-date pairs only reveal discount factors for money in an unequivocal way. It is
often implicitly assumed that such experiments also reveal the discount factor for con-
sumption, but this interpretation requires the assumption that the money oered in the
experiment is consumed immediately: subjects do not use capital markets to reallocate
their consumption over time. This assumption is not outrageous (especially for small
amounts, it may not be implausible that capital markets considerations are ignored) but
certainly it cannot be taken for granted without further study. Coller and Williams [6]
were the rst to point out the possible censoring eects of capital markets on experimen-
tal subjects responses. Cubitt and Read [7] explore in great detail what exactly can be
inferred from responses to the standard laboratory tasks on choice over time once it is
admitted that subjects are able and willing to access imperfect
17
capital markets, so that
16
To be precise, one can only infer a range for the discount factor, whose width depends on the size of
the progressive increments of the additional monetary increments x.
17
I.e., with the borrowing interest rate possibly diering from the lending rate.
21
the implicit laboratory rate of interest competes with the market ones. They point out
that the choice between two money-date pairs in the presence of capital markets is not
really the choice between two points in the standard Fischer diagram
18
but rather the
choice between two whole consumption frontiers. As is intuitive, this fact greatly reduces
the possibility of inference about discount factors for consumption.
A dierent but conceptually related reservation about the correct inferences to be
drawn from experimental results comes from the recent work by Noor [34]. He observes
that nothing excludes that experimental subjects integrate the laboratory rewards with
the anticipated future levels of wealth. The striking implication is that if such future
wealth levels are expected to change, all main documented soft anomalies, including pref-
erence reversal, turn out to be compatible with the EDM. Intuitively, if the subject is
more cash constrained now than he expects to be at a later date, so that his marginal
utility for money is higher now than it is expected to be in the future, he may well choose
according to the pattern of the preference reversal phenomenon, while still making his
choices on the basis of a constant, and not declining, discount factor. In a precise sense,
the EDM is shown in this work to have no empirical content unless the integration with
expected future wealth is excluded a priori. However, Noor also suggests an experimental
design including risky prospects as outcomes, which is suciently rich to test the EDM.
Furthermore, from eld data, Harrison Lau and Williams [19] nd that, unlike pre-
vious claims of non-constant discount factors, although discount factors do depend on
household characteristics, within each homogeneous group discount factors over one and
three years horizon are indeed constant. But this is not all: as we mentioned earlier, the
concavity of the utility function may explain apparent anomalies, thus calling for both the
time preferences and the preference for the good whose receipt is delayed to be elicited
simultaneously. If a single utility value n(r. t) is elicited, rather then a separate assess-
ment of the time and outcome components, it seems fair to argue that the concavity of
the utility function might conate into the estimate of discount factors. Starting from this
consideration, Andersen, Harrison, Lau and Rutstrm [2] show that the implausibly high
estimates of discount factors previously uncovered fall substantially once the concavity of
the utility function is taken into account in the estimation, and both risk and time prefer-
ences are elicited from experimental subjects. The dierence is quite dramatic: whereas
under the assumption of risk neutrality the point estimate of the yearly discount factor is
roughly 25%, it falls six-fold to about 4% once risk aversion is (correctly) accounted for.
19
18
In which consumption levels at two distinct dates are represented on each axis on the plane.
19
Field data from the retiriment options oered to retired military personnel in the US (see Warner
and Pleeter [61]) suggest higher than expected discount rates. However see Harrison and List [17] for a
critique of the heroic extrapolation method used.
22
Summing up, then, when it comes to preferences over outcome-date pairs, once the
correct estimation techniques are used and concavity of the utility function is allowed for,
the wildly varying discount factor estimates fall to more manageable ranges of variation
and realistic values.
5.4 Hard anomalies
There are other observed violations of the EDM which are more fundamental in the
sense that, unlike preference reversal, they seem to contradict the basic assumption of
maximization of any economically reasonable objective function. One notable instance
is that human decision-makers have been shown to make intransitive choices. Although
most data in this direction come from choices under risk, the evidence available for time
preferences though limited is clear in suggesting that violations of transitivity are more
frequent in this domain. Tversky, Slovic and Khaneman [59] show that a substantial 15%
of subjects exhibited cyclical patterns of choice that could not be explained by framing
eects - and in an experiment which was not designed to uncover cycles. When the issues
of cycles in choice is addressed directly, the evidence is even more striking: Roelofsma
and Read [45] found that the majority of intertemporal choices were intransitive. Cyclical
choice is thus one solid anomaly that cannot be accommodated within any discounting
model.
While incentive compatible experimental investigations of choices over outcome-date
pairs form a small but non-negligible literature, experimental investigations of choices over
reward sequences are extremely thin on the ground in the economics literature, especially
with nancially motivated subjects. Arguably, this is because the diculties highlighted
above are exacerbated by payment of experimental subjects having to take place over
weeks if not months. The unreliability of data from experiments based on hypotheti-
cal choices seems to be driving the recent increase in incentive compatible experimental
designs.
The rst experimental paper on preferences over monetary sequences of outcomes is
due to Loewenstein and Sicherman [28]. In a survey of members of the public entering
a museum, interviewers asked participants to choose among hypothetical alternative pro-
les of either wages of savings plan over a number of years. Loewenstein and Sicherman
[28] found evidence of preference for sequences of increasing monetary payments (ver-
sus constant or decreasing ones). They explain this nding by pointing to a preference
for maintaining the current consumption level, so that wages should be non-decreasing.
Admittedly, though, these were hypothetical questions, and some respondents motivated
their preference for increasing sequences with ination. In addition, the framing of these
23
questions as salary proles might evoke an improvement in ones career, or just be what
one would generally expect (i.e. aective response). However other authors have found
evidence of preferences for constant and even decreasing sequences of outcomes over time
(e.g. Chapman [5], Gigliotti and Sopher [14], Guyse, Keller and Epple [15]). The domain
of choice seems also to be important (e.g. there are dierences in observed choices depend-
ing on whether or not the sequences are of money, or health or environmental outcomes
see e.g. Chapman [5] and Guyse, Keller and Epple [15]).
Manzini, Mariotti and Mittone [32] asked subjects to make binary choices among all
possible pairs of monetary sequences, with an increasing, constant, decreasing or jump
(i.e. end eect) pattern, both in a paid condition (where subjects do indeed receive the
sums corresponding to the sequence chosen) and an unpaid condition (where choices are
hypothetical). Previous experimental evidence on reward sequences suggests that the
general trend of the sequence (increasing or decreasing) is relevant to make decisions.
However, in this case the data provide much weaker evidence than Loewenstein and Pr-
elecs [25] in support of their view that sequences of outcomes that decline in value are
greatly disliked (p. 351). It is found that, even in the simple decision problems studied,
where monetary sequences can be clearly ordered according to their trends, simply choos-
ing according to the heuristics that favours the increasingness of the trend does a rather
poor job at explaining the data. The modal subject and choice is rational, in the sense
of being compatible with positive time preference combined with preference for income
smoothing (concave utility function). Therefore while choice incompatible with EDM is
observed, it is not to the extent that the existing literature suggests. When there are
no aective factors involved (such as, for example, the sense of dread for choices relating
to health, or the sense of failure involved in a decreasing wage prole), some theory of
positive discounting can provide a rough approximation of the choice patterns. However a
non negligible proportion of our subjects (around 30%) choose in ways that are incompat-
ible with any form of positive discounting (exponential, hyperbolic or otherwise). These
subjects violate the basic economic assumption that for a good, the sooner the better,
suggesting that other mechanisms beyond discounting are at work. That is, Loewenstein
and Prelecs pioneering ndings do capture, beside aective factors, some of the heuris-
tic considerations that people use when evaluating neutrally (without aects) money
sequences. Moreover the study nds that irrational choices present a systematic pat-
tern, not encountered previously. Of these, the most striking are the association between
certain types of rational choices and irrational choices (those who prefer a decreasing to
a constant sequence are disproportionately concentrated among those who also prefer a
constant to an increasing sequence); and the association between irrational choices of
a dierent type (choosing an increasing over a decreasing sequence is very strongly as-
24
sociated with choosing an increasing over a constant sequence). Such patterns cannot
be generated by any discounting model, nor by such a model augmented with random
independent mistakes.
One last puzzling experimental nding that we wish to highlight is due to Rubinstein
[48]. In a hypothetical setting he nds a single outcome/sequence of outcomes type of
preference reversal of the kind highlighted by Loewenstein and Prelec [27] in a dierent
domain: in a classroom experiment a majority of students preferred to receive a payment
of 997 monetary units at a later date t

than 1000 at an even later date t

+ 1, but when
choosing between sequences of four payments of a constant amount of either 997 starting
at t
0
or 1000 starting at t
0
+ 1, the latter sequence was now preferred, contradicting the
theory of hyperbolic discounting.
That is, subjects exhibited the following type of behavior: they chose r to be received
at some date t

versus the larger sum r + . to be received at date t

+ 1 (they were
impatient and preferred smaller reward earlier rather than larger reward later) but chose
the sequence
c = ((r + .. t
0
+ 1) . (r + .. t
0
+ 2) . (r + .. t
0
+ 3) . (r + .. t
0
+ 4))
over the sequence
/ = ((r. t
0
) . (r. t
0
+ 1) . (r. t
0
+ 2) . (r. t
0
+ 3))
where t
0
+4 _ t

+1. This contradicts not only the EDM but also the HDM and in fact any
model of discounting based on diminishing impatience (declining discount rates): if the
subject were impatient at the late date t

and not willing to trade o one unit of delay for

an additional reward ., he should have been unwilling to perform all four trade-os of this
type involved in the comparisons between sequences. Rubinsteins explanation is based
on similarity: given choices between sequences of alternatives with two distinct attributes,
the decision maker uses a procedure whereby rst of all he tries and rank alternatives based
on dominance (i.e. greater outcome and earlier time of receipt); if this is not decisive, he
looks for similarities between the two dimensions, trying to discriminate based on evident
dierences. If none can be found, then he chooses based on some dierent criterion.
20
As
we will see in section 6, however, other theories, too, can explain this phenomenon.
6 Empirical anomalies and theory
As we have already made clear, some of the anomalous types of behaviour discussed in
the empirical literature are not hard anomalies after all, in the sense that they can be
20
See also Rubinstein [46].
25
easily accommodated within the axiom set of theorem 2 in section 2. This is the case
for the main phenomenon of preference reversal described in section 3.1. So we begin
by tackling soft anomalies rst (delay/speed-up asymmetry, magnitude eect and inverse
preference reversal), and then move to the hard anomalies (cycles in choice and the
single outcome/sequence of outcomes preference reversal).
The delay/speed-up asymmetry does not even violate by itself any of the Fishburn-
Rubinstein axioms for exponential discounting and in this sense it is not an anomaly.
Imagine for notational simplicity that the object whose receipt is to be delayed or an-
ticipated is an amount of money r. So the eect can be written within the EDM, in a
non-contradictory way, as:
n(r 1) = o
t
n(r)
n(r) = o
t
n(r + 1)
where 1 and 1 are, respectively, the amount of money that the agent is willing to pay
to anticipate the receipt of the money which he is entitled to receive at date t, and 1 is
the amount of money that the agent requires to delay to date t the receipt of the amount
of money to which he is entitled now.
The magnitude eect, like the delay/speed-up asymmetry, also does not violate any
of the Fishburn-Rubinstein axioms for exponential discounting. As noted by Ok and
Masatlioglu [36],
21
for example, the exact Thalers numbers reported before are compatible
with the EDM with a o = 0.95 and a concave utility function n(r) = r
:42
+ 45.9 dened
on the positive real line.
However, although formally this phenomenon is indeed compatible with exponential
discounting, some observations are in order. One might argue that although some utility
function can necessarily be found (given the non-violation of the axioms) that ts the (few)
observations, additional constraints might be desirable for the utility function, and these
constraints might create an incompatibility with the observed eects. For example, it is
often informally argued (and often simply taken for granted) that for small amounts the
utility function ought to be linear. In this case, the EDM is incompatible with the mag-
nitude eect (using Thalers numbers, it would require for example o =
15
60
=
250
350
=
3000
4000
).
However, in this case the magnitude eect (which involves only two dates) is also incom-
patible with the HDM, and indeed with any separable discounting model axiomatised in
Theorem 2 of Fishburn and Rubinstein: we would still obtain a contradiction of the type
r

= o (t) =
r
0

0
but also
r
0

0
<
r

21
See Ok and Masatlioglu [36] in the (2003) version.
26
So it seems to us that the real point about such eects is that either they do not
constitute an EDM anomaly, or if they do (because of the linearity of utility) they also
constitute an anomaly for (much) more general discounting models, notably including
HDM. It is incompatible with the linear version of relative discounting, too, and it can
be made compatible with a linear utility version of Manzini and Mariottis (o. o) model
only in a variant in which the vagueness term o is made to depend on the outcome (see
Manzini and Mariotti [30] for details).
In addition, it is fair to say that the evidence is still too scant. In order to t a
utility function and check its reasonableness (or compatibility with independent data on
concavity-convexity) one would need many more observations in dierent regions of the
time and outcome space. At best the existing observations might be simply suggestive of
the fact that human decision makers use certain yet to be discovered heuristic procedures
when judging outcome-date pairs. Certainly the magnitude eect, even together with
the assumption of exponential discounting, is not necessarily and intrinsically related to
diminishing marginal utility. For example, suppose n(r. t) = o
t
r

, where c (0. 1). Then

let J and 1 be the compensations required by a decision maker to delay by one period
from now the receipt of r and , respectively, that is o (r + J)

= r

and o ( + 1)

.
This implies
o ( + 1)

o (r + J)

=
r
r + J
=

+ 1
so that the magnitude eect (which requires that, if say < r, then
y
y+K
<
x
x+J
) never
obtains. It seems unlikely that simple changes of functional forms with respect to the
EDM will be descriptively adequate in general. However, as we noted already, a new
literature is emerging which attempts to estimate both the shape of the utility function
and discount factors at the same time (see Andersen, Harrison, Mortensen and Rutstrm
[2]): this type of research may in due course shed additional light on the issue of magnitude
eects.
Sayman and ncler [51]s negative preference reversal is a soft anomaly that cannot
be accommodated within the class of HDM. Indeed, if r < , the preferences (r. t) ~
(. t + t) and (. t) ~ (r. 0) correspond to o (t) n(r) o (t + t) n() and o (t) n()
n(r) (in a separable discounting model). This is obviously incompatible with HDM since
it requires decreasing (rather than increasing) discount rates. To the contrary, this type
of preferences can be accommodated within the model of vague time preferences.
For instance in the simple (o. o) representation with the TPM we need
ro

_ o
t+
+ o and o
t+
_ ro

+ o
o
t
r + o
so that (r. t) ~

(. t + t) by the Secondary Criterion while (. t) ~

(r. 0) by the Primary

27
Criterion. Negative preference reversal is also compatible with Ok and Masatlioglus
RDM. Here we need
o (t. 0) =
1
o (0. t)

n(r)
n()
o (t + t. t)
When interpreted as revealed preferences, observed cycles in binary choice are a harder
anomaly to deal with because they violate the fundamental axiom of Order. For outcome-
date pairs, the HDM cannot explain cycles, whereas alternative theories (most notably
Ok and Masatlioglus RDM and our own theory of vague time preferences) can. Similarly
Reads subadditive discounting and Rubinsteins similarity based decision making are also
consistent with the phenomenon.
Consider three alternatives (r. :), (. :) and (.. t), with r < < . and : < : < t. In
the RDM it is perfectly possible to have that n(r) o (t. :) n(.), n(.) o (:. t) n(),
but n() o (:. :) n(r). This could happen for example if there is very little perceived
dierence between t and : and between : and :, but the dierence between t and :
is perceived as signicant (imagine : < : < t and r < < .). Then the latter two
inequalities attest to the fact that the dierences . and r are enough to compensate
for the the small delays, but the dierence .r is not enough to compensate for the large
delay. Note however that in the model this compounding of small dierences eect is
not allowed to hold for the outcome dimension, but just for the time dimension. With
these assumptions it is also easy to see how Rubinsteins similarities might work: if t and
: are similar, and so are : and :, while all other comparisons are perceived as dierent,
then (.. t) is preferred over (. :), (. :) is preferred over (r. :) and because : and t are
not similar, nor are r and ., it is enough for the unspecied completing criterion to pick
(r. :) over (.. t) to complete the cycle.
Subadditive discounting, too, can explain cyclical choices. Consider the three time
intervals [0. :], [:. :] and [:. t], and let the discount factors over an interval [c. /] be denoted
by o
ab
. In the presence of subadditivty we have that o
0t
o
0r
o
rs
o
st
. So to generate the
cycle in choice where (. :) is chosen over (.. t) which is chosen over (r. :) which is chosen
over (. :), assume that the decision maker splits the time intervals involved in each binary
comparisons taking into account the common delay. Then we need:
n() o
0s
n(.) o
0s
o
st
n(.) o
0r
o
rt
n(r) o
0r
n(r) o
0r
n() o
0r
o
rs
28
which simplies to
n() n(.) o
st
n(.) o
rt
n(r)
n(r) n() o
rs
the last two inequalities imply n(.) o
rt
n() o
rs
, which is compatible with the rst
inequality as long as n(.) o
rt
n (.) o
st
o
rs
, that is if o
rt
o
rs
o
st
, which is precisely what
subadditive discounting entails.
The vague theory of time preferences is also consistent with intransitivities. For in-
stance in the OPM we need
ro
r
.o
t
+ o
o
s
_ .o
t
+ o and .o
t
_ o
s
+ o
o
s
_ ro
r
+ o and ro
r
_ o
s
+ o
so that (r. :) ~

(.. t) but (r. :) ~ (. :) and (. :) ~ (.. t) by the Primary Criterion.

However by the Secondary Criterion (. :) ~

(r. :) and (.. t) ~

(. :), thereby producing

a cycle. Obviously any theory that can cope with cycles can cope with direct preference
reversal, so we skip the details.
Next, consider the outcome-sequence type of preference reversal presented by Ru-
binstein [48] for the case of monetary sequences, and Loewenstein and Prelec [27] in a
dierent domain. Here once again the caveat must be that both these papers present
results from hypothetical questions. This notwithstanding, we report it here because as
we discussed earlier, these pose a harder challenge to conventional theories. Obviously
Rubinsteins own similarity considerations provide an explanation to this observed phe-
nomenon. In addition, the model of vague time preferences for sequences can provide an
alternative explanation. For the sequences described in the previous section suppose for
example that the secondary criterion is the natural one proposed by Rubinstein himself,
namely Pareto dominance between the outcome sequences, and that
o
t

n(r) o
t

+1
n(r + .) + o(r + .. t

+ 1)
3
X
i=0
o
t
0
+i
n(r) _
3
X
i=1
o
t
0
+i
n(r + .) + o(c)
In this case the preference between outcome-date pairs can be explained by present dis-
counted utility (primary criterion) and the preference between sequences can be explained
by the secondary criterion.
29
Finally, the evidence of moderate preference for increasing sequences discussed in
Manzini, Mariotti and Mittone [32] is also consistent with the model proposed therein, as
well as obviously with the model by Fishburn and Edwards [10].
7 Concluding remarks
In the last twenty years a growing body of experimental evidence has posed a challenge to
the standard Exponential Discounting Model of choice over time. Attention has focused on
some specic anomalies, notably preference reversal and declining discount rates, leading
to the formulation of the model of hyperbolic discounting which is nding increasing favour
in the literature. As we have seen, it is debatable whether some of the most focused
upon anomalies should be indeed classied as such, or whether they are really the most
challenging ones for conventional theory. If they violate any axiomatic property, this is
Stationarity, which is not strongly defendable even on normative, let alone descriptive,
grounds. A group of theoretical ideas is beginning to emerge which can address not only
violations of Stationarity, but even more challenging observed phenomena.
At the same time, at the empirical level much progress is being made on two fronts:
sophisticated estimation of discount factors (e.g. considering censoring factors) and
simultaneous presence of discounting and risk aversion, a traditionally much neglected
issue until very recently. The results are quite stunning, implying as they do a serious
reconsideration of the previous estimates. On the other hand, other recent work challenges
the conventional interpretation given to responses to standard experimental choice tasks.
While in the theory of choice under risk there exist rationality axioms that exert a
strong normative appeal, this is less clearly the case for choice over time. Thus, it is natural
to step towards models of procedural rationality as opposed to normative rationality, given
that the latter lacks a clear notion. Some work already exists in this direction, but much
more remains to be done in this fascinating area, as we are still quite far from a clear-cut
best theory.
References
[1] Ainslie, G., 1975 Specious reward: a behavioral theory of impulsiveness and impul-
sive control, Psychological Bulletin, 82: 463-496.
[2] Andersen, S., G. W. Harrison, M. I. Lau and E. E. Rutstrm, 2007, Eliciting Risk
and Time preferences, mimeo, University of Central Florida.
[3] Benoit, J-P. and E. Ok, 2007, Delay Aversion, Theoretical Economics, forthcoming.
30
[4] Benzion, U., A. Rapoport and J. Yagil, 1989, Discount Rates Inferred from Deci-
sions: an Experimental Study", Management Science, 35: 270-284.
[5] Chapman, G., 1996, Expectations and preferences for sequences of health and
money, Organizational Behavior and Human Decision Processes, 67, 59-75.
[6] Coller, M. and M. B. Williams, 1999, "Eliciting Individual Discount Rates", Experi-
mental Economics 2: 107-127, 1999.
[7] Cubitt, R. and D. Read, 2007, Can Intertemporal Choice Experiments
Elicit Time Preferences for Consumption, Experimental Economics, online:
http://www.springerlink.com/content/v47u1v31w5306870/.
[8] Cummings, R., G. W. Harrison and E. E. Rutstrm, 1995, Homegrown values
and hypothetical surveys: is a dichotomous choice approach incentive compatible?",
American Economic Review, 85: 260-266.
[9] Elster, J. and G. Loewenstein, 1992, Choice Over Time, Russel Sage Foundation,
New York.
[10] Fishburn, P. J. and W. Edwards, 1997, Discount Neutral Utility Models for Denu-
merable Time Streams, Theory and Decision 43: 139-66.
[11] Fishburn, P. J., and A. Rubinstein, 1982,Time Preference, International Economic
Review, 23: 677-694.
[12] Frederick, S., G. Loewenstein and T. ODonoghue, 2002, Time discounting and time
preferences: a critical review, Journal of Economic Literature, 40: 351-401.
[13] Garratt, R., M. Walker and J. Wooders, 2007, Behavior in Second-Price Auctions
by Highly Experienced eBay Buyers and Sellers, mimeo, UCSB.
[14] Gigliotti, G. and B. Sopher, 1997, Violations of Present Value Maximization in
Income Choice, Theory and Decision, 43: 45-69.
[15] Guyse, J., L. Keller and T. Epple, 2002, Valuing Environmental Outcomes: Pref-
erences for Constant or Improving Sequences, Organizational Behavior and Human
Decision Processes, 87: 253-277.
[16] Harrison, G. W., 1992, Theory and Misbehavior of First Price Auctions: Reply",
American Economic Review, 82: 1426-1443.
31
[17] Harrison, G. W. and J. A. List, 2004, Field Experiments", Journal of Economic
Literature, XLII: 1009-1055.
[18] Harrison, G. W. and M. Lau, 2005, Is the Evidence for Hyperbolic Discounting in
Humans Just An Experimental Artefact?", Behavioral and Brain Sciences, 28: 657.
[19] Harrison, G. W., M. Lau and M. B. Williams, 2002, "Estimating Individual Discount
Rates in Denmark: A Field Experiment" American Economic Review, 92: 1606-1617.
[20] Hertwig, R. and A. Ortmann, 2001, Experimental practices in economics: A
methodological challenge for psychologists?, Behavioral Brain Science, 24: 383-451.
[21] Kahneman, D. and A. Tversky, 1979, Prospect Theory: an Analysis of Decision
Under Risk, Econometrica, 47: 363-391.
[22] Laibson, D., 1997, Golden eggs and hyperbolic discounting, Quarterly Journal of
Economics, 112: 443-477.
[23] Loewenstein, G., 1987, Anticipation and the Valuation of Delayed Consumption,
Economic Journal, 97: 666-684.
[24] Loewenstein, G., 1988, Frames of mind in intertemporal choice. Management Sci-
ence, 34: 200-14.
[25] Loewenstein, G. and D. Prelec, 1991, Decision Making Over Time and Under Un-
certainty: A Common Approach, Management Science, 37: 770-786.
[26] Loewenstein, G. and D. Prelec, 1992, Anomalies in intertemporal choice: evidence
and interpretation, Quarterly Journal of Economics, 107: 573-597.
[27] Loewenstein, G. and D. Prelec, 1993, Preferences for sequences of outcomes, Psy-
chological Review, 100: 91-108.
[28] Loewenstein, G. and N. Sicherman, 1991, Do workers prefer increasing wage pro-
les?, Journal of Labour Economics, 9: 67-84.
[29] Loewenstein, G. and R. Thaler, 1989, Intertemporal Choice, Journal of Economic
Perspectives, 3: 181-193.
[30] Manzini, P. and M. Mariotti, 2006, A Vague Theory of Choice Over Time, Advances
in Theoretical Economics, 6 (1): article 6.
[31] Manzini, P. and M. Mariotti, 2007, Sequentially Rationalizable Choice, American
Economic Review, forthcoming.
32
[32] Manzini, P., M. Mariotti and L. Mittone, 2006, Choosing monetary sequences: the-
ory and experimental evidence, CEEL WP 1-06 and IZA DP No. 2129.
[33] Manzini, P., M. Mariotti and L. Mittone, 2007, The Elicitation of Time Preferences,
research report for ESRC grant RES-000-22-1636, mimeo.
[34] Noor, J., 2007, Hyperbolic Discounting and the Standard Model, mimeo, Boston
University.
[35] Noussair, C., S. Roibin and B. Rueux, 2004, Revealing Consumers Willingness
to Pay: A Comparison of the BDM Mechanism and the Vickrey Auction, Journal
of Economic Psychology, 25: 725-741.
[36] Ok, E. A. and Y. Masatlioglu, A General Theory of Time Preferences, Journal of
Economic Theory, forthcoming.
[37] Ortmann, A. and Hertwig, R., 2005, Monetary Incentives: Usually Neither Neces-
sary Nor Sucient?" CERGE-EI Working Papers n. 307
[38] Pender, J., 1996, Discount Rates and Credit Markets: Theory and Evidence from
Rural India, Journal of Development Economics, 50: 257296.
[39] Phelps, E. and R. Pollack, 1968, On second best national savings and game-
equilibrium growth, Review of Economic Studies, 35: 201-208.
[40] Pigou, A. C., 1920, The Economics of Welfare, London, McMillan.
[41] Plott, C. and K. Zeiler, 2005, The Willingness to Pay/Willingness to Accept Gap,
the Endowment Eect, Subject Misconceptions and Experimental Procedures for
Eliciting Valuations", American Economic Review, 95: 530-545.
[42] Read, D., 2001 Is Time-Discounting Hyperbolic or Subadditive?, Journal of Risk
and Uncertainty, 23: 5-32.
[43] Read, D., 2005 Monetary incentives, what are they good for?, Journal of Economic
Methodology, 12: 265-7
[44] Read, D. and M. Powell, 2002, Reasons for sequence preferences, Journal of Be-
havioral Decision Making, 15: 433-460.
[45] Roelofsma, P. H. and D. Read, 2000 Intransitive Intertemporal Choice, Journal of
Behavioral Decision Making, 13: 161-177.
33
[46] Rubinstein, A., 1988, Similarity and Decision Making Under Risk (Is there a utility
theory resolution to the Allais Paradox?), Journal of Economic Theory, 46: 145-153.
[47] Rubinstein, A., 2001, A theorists view of experiments, European Economic Review,
45: 615-628.
[48] Rubinstein, A., 2003,Is it Economics and Psychology ? The case of hyperbolic
discounting, International Economic Review, 44: 1207-1216
[49] Rutstrm, E. E., 1998, Home-grown Values and Incentive Compatible Auction De-
sign", International Journal of Game Theory, 27: 427-441.
[50] Samuelson, P., 1937 A Note On the Measurement of Utility, Review of Economic
Studies 4: 155-61.
[51] Sayman, S. and A. ncler, 2006, Reverse time inconsistency", mimeo, INSEAD.
[52] Scholten, M. and D. Read, 2006, Discounting by Intervals: A Generalized Model of
Intertemporal Choice, Management Science, 52: 1424-1436
[53] Shelley, M., 1993, Outcome signs, question frames and discount rates Management
Science, 39: 806-15.
[54] Strotz, R. H., 1956, Myopia and inconsistency in dynamic utility maximization,
Review of Economic Studies, 23: 165-180.
[55] Thaler, R., 1981, Some Empirical Evidence on Dynamic Inconsistency", Economics
Letters, 8: 201-207.
[56] Tversky, A., 1969, Intransitivity of Preferences, Psychological Review, 76: 31-48.
[57] Tversky, A., and D. Kahneman, 1981, The framing of decisions and the psychology
of choice, Science, 211: 453458.
[58] Tversky, A. and D. Koehler, 1994, Support Theory: Nonextensional Representation
of Subjective Probability Theory, Psychological Review, 101: 547-67.
[59] Tversky; A., P. Slovic and D. Kahneman, 1990,The Causes of Preference Reversal,
American Economic Review, 80: 204-217
[60] Tversky, A., Sattath, S. and P. Slovic, 1988, Contingent weighting in judgement
and choice, Psychological Review, 95: 37184.
34
[61] Warner, J. T. and S. Pleeter, 2001, The Personal Discount Rate: Evidence from
Military Downsizing Programs", American Economic Review, 91: 33-53.
35