The Review of Economic Studies, Ltd.

Myopia and Inconsistency in Dynamic Utility Maximization

Author(s): R. H. Strotz
Source: The Review of Economic Studies, Vol. 23, No. 3 (1955 - 1956), pp. 165-180
Published by: Oxford University Press
Stable URL: http://www.jstor.org/stable/2295722 .
Accessed: 25/06/2014 08:10

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

Oxford University Press and The Review of Economic Studies, Ltd. are collaborating with JSTOR to digitize,
preserve and extend access to The Review of Economic Studies.

http://www.jstor.org

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 Myopia and Inconsistency in
Dynamic Utility Maximization
"but you must bind me hard and fast, so that I cannot stir from the spot
where you will stand me . . . and if I beg you to release me, you must
tighten and add to my bonds."-The Odyssey.

I. INTRODUCTION
This paperpresentsa problemwhichI believehas not heretoforebeen analysed2and
providesa theory to explain, under differentcircumstances,three related phenomena:
(1) spendthriftiness; (2) the deliberateregimentingof one's futureeconomicbehaviour-
even at a cost; and (3) thrift. The sensesin whichwe deal with thesetopics can probably
not be very well understood,however,until after the paper has been read; but a few
sentencesat this point may shed some light on what we are up to.
An individualis imaginedto choose a plan of consumptionfor a futureperiodof time
so as to maximizethe utilityof the plan as evaluatedat the presentmoment. His choiceis,
of course,subjectto a budgetconstraint. Our problemariseswhen we ask: If he is free
to reconsiderhis planat laterdates,will he abideby it or disobeyit-even thoughhis original
expectations offuture desires and means of consumptionare verifed ? Our answer is that
the optimalplan of the presentmomentis generallyone whichwill not be obeyed,or that
the individual'sfuturebehaviourwill be inconsistentwith his optimalplan. If this incon-
sistencyis not recognized,our subjectwill typicallybe a " spendthrift,"a termwhich has
had no meaningin existing utility theory but which becomes explicatedin the theory
presentedhere. If the inconsistencyis recognized,the rationalindividualwill do one of two
things. He may " precommit" his futurebehaviourby precludingfutureoptions so that
it will conformto his presentdesireas to whatit shouldbe. Or,alternatively,he maymodify
his chosen plan to take account of future disobedience,realizingthat the possibilityof
disobedienceimposes a furtherconstraint-beyond the budget constraint-on the set of
plans which are attainable. It is in this way that the individualbecomes " thrifty"-a
termwhichalso acquiresmeaningin the contextof the analysis. Whatis crucialto all this
is that the discountappliedto a futureutilityshoulddependon the time-distancefrom the
presentdate and not upon the calendardate at whichit occurs. In a final section,we shall
have some thingsto say about the meaningof " consumersovereignty" in the framework
of dynamic choice.
1 I am indebtedto many colleagues at NorthwesternUniversity and elsewherewho have commented
helpfullyon this manuscript,and I am especiallyindebtedto Mr. Fred Westfieldand Dr. Alvin Martyfor
many sustainedand fruitfuldiscussionsof the subjectas well as to M. JacquesDreze, Prof. HarryJohnson,
Prof. R. Solow, Mr. S. A. Thore, Professor Gerhard Tintner, Professor H. Wold, and members of the
seminar of the Departmentof Applied Economics, CaTbridge University, for criticismsrendered. The
usual caveat protectingthese courteous people from furtherresponsibilityis, of course, in order.
The paper was completedwhile the author was visiting the Departmentof Applied Economics, Cam-
bridge University, on a Rockefeller Grant.
2
But it has been alluded to by Paul A. Samuelson," A Note on Measurementof Utility," Reviewof
EconomicStudies,IV, 2 (Feb., 1937),esp. p. 160. See also FriedrichA. Hayek, ThePure Theoryof Capital
(1941), p. 218, and M. Allais, L'Economieet Int&ret,Annexe 3.

165

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
166 REVIEW OF ECONOMIC STUDIES

II. THE UTILITY OF THE CONSUMPTION PLAN

The general problem of intertemporal utility maximization is that of an economic

decision-maker who must choose among various possible functions relating his economic
activities to time. The decision-maker may be a firm or an individual and the object of
choice may be the variation through tinmeof one or more economic magnitudes such as
corporate sales, profits, the supply of a labor service, the consumption of a given commodity,
or the expenditure of income. It suits our purpose, however, to deal with a quite simple
version of this problem and later to show the considerable extent to which our results can be
generalized. It is only as a standard illustration, then, that we have chosen to think of an
individual who must decide how a single economic magnitude, his consumption, is to
vary over some period of time subject to a simple budget constraint. As a further simpli-
fication we abstract from all considerations of risk and uncertainty. This may disturb
those who feel that the essence of dynamical problems is thereby ignored ; but I think it
will become clear as we go along that to introduce risk and uncertainty would only clutter
up the analysis and prevent our getting a clear view of the particularissues to be considered
here. Risk and uncertaintydo have some bearing on the topic, but that may best be examined
much later on.
To begin, assume that an individual must choose at time t =-: among a (possibly
infinite) number of alternative time paths of consumption each of which is certain. For
example, in Figure 1 the curves C1, . . . , C5represent functions relating consumption to
time. These curves, which may be called " consumption-time" curves, are the objects of
choice and are defined over a specified time period, 0 < t < T. The case in which the period
is infinite need not be excluded, but, to fix ideas, we suppose that it is finite. This is merely

\ AC Ci

- -- 3=c3(t)
%\ .C..'

C =C (t)

' t
0 T

Figure 1

to assume that the individual has only'idle curiosity in what happens after several millenia.
Each function C(t) is, of course, single-valued and bounded from both above and below,
and time is treated as a continuous variable.'
The next step is to assume that an individual who faces the alternatives C, (t), C2(t),. ..

1There is no difficultyin regardingtime as discreteand one can then employ the method of Lagrange
multipliersratherthan the Calculusof Variationsused here. I find that some issues can be treated more
smoothly by dealing with the continuous case, however, and prefer the approach taken here.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 MYOPIA AND INCONSISTENCY 167

can orderthemtransitivelyand that his orderingcan be representedby a utilityfunctionall

r T
c(t
(1) ~~~~~~~~Dlr=
1DN 0
or any monotonicincreasingfunctionthereof. The subscriptT appearson the functional
to indicatethat the preferencesexpressedare those at time 'r althoughchoicesare among
consumptionpaths for the entire period.
To derive any interestingtheoremsfrom the analysis, however, one must assume
morethanjust the transitivityof preferencesamongthe consumption-time curves. Accord-
ingly, I supposethat the orderingof these curvesis such that the utilityfunctionalcan be
writtenas
-1T

(2) Ox X (t-) ut[C(t), t] dt9

.0

whereu[C(t), t] is an " instantaneousutility function"2 assigningat each time t a value

u(t) to C(t), andwhereX(t- -) is a weightor discountfunctionwhosevaluedependsnotably
on the time-distancebetweena future(or past) date t and the presentdate r. Expression
(2) may be normalizedand this we do by settingA(O)= 1.
In our later effort at generalization(Section VII) we find it crucial that the utility
functionalshouldbe representable as an integralwithrespectto timeand thatthe integrand
functioncan, as in this illustration,be factoredinto two functions,one dependingonly on
the time distanceof future (or past) consumptionfrom the presentdate, and the other
beingindependentof the presentdate. What changesthen as time (X)marcheson is only
the discount function ?(t - r), and it undergoes only a linear shift.
If T.> 0, one may questionthe relevancefor choicemakingof the rangeof integration

from 0 to 'r. For the especially simple functional given by (2), the value of ?X(t-r)
"0
u[C(t),t] dt is historicallygiven. In the moregeneralcase, however,this is not technically
necessary,becausethe u functionmight contain as one of its argumentsa lead value of
C, e.g., C(t + 0). We prefer,therefore,to integratefrom 0, noting that the budgetcon-
straintto be introducedlater requiresthat CO) } be taken as given, anyway. This

also allows(trivially)for the possibilitythat a personis not indifferentto his consumption

history but enjoys his memoriesof it.
The relativeweightwhich a person may assign to the satisfactionof a futureact of
consumption(the mannerof discounting)may dependon either or both of two things:
(1) the timedistanceof the futuredate from the presentmoment,or (2) the calendardate
1 A "functional " is simply a function of the form of a function, or, more exactly, a function of an
infinite number of variables. In (1), for example, IT depends not on any particularvalue assigned C(t),
but on the form of the function C(t) as t goes from 0 to T. Changethe form of the functionand you change
DX. The economistwill find a good introductionto the topic and to the related subject of the " calculus
of variations" in R. G. D. Allen, MathematicalAnalysisfor Economists,ch. XX. For early applications
of the calculus of variationsto utility problems, see G. Tintner, " Distribution of Income Over Time,"
Econometrica,1936, and " Maximizationof Utility Over Time," Econometrica,1938.
2
This is doubtlessa misnomer,althoughthe termmay facilitateone's umderstanding in the firstinstance
of its use. It wouldbe morepreciseto say that a is a realnumberassignedto [C(t),t] such that the individual
may be said to maximizea weightedsum (integral)of numbersassignedin this way.
Otherwise,it is awkwardto think of u as a utility " experienced" at a point of time when in a later
section we allow the possibilityof its dependingon the consumptionof a later date.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 168 REVIEW OF ECONOMIC STUDIES

of the future act of consumption. The weight I assign to my pleasure (u function) in drinking
champagne next September 26 may depend either on the fact that that date is a certain
length of time away from the present or on the fact that that is my birthday. To the extent
that time-distance is important, I may assign a different (and probably higher) weight to
September 26 as it draws nigh ; if only the calendar date is important, the weight will not
change as that date approaches.1 Both bases for discounting a future date are included in
the functional (2). The importance of the calendar date enters through the appearance
of t in the instantaneous utility function, whereas the importance of time distance is given
by X(t - -). The distinction between these two causes of discounting is commonly over-
looked, because it has no consequence in those theories which regard the present date as
fixed.2 A truly dynamic theory of utility maximization must, however, assume that the
present date changes, and, as we shall see, the distinction is then an important one.
The reader may wonder whether the instantaneous utility function could be replaced
in (2) with any monotonically increasing function thereof; the answer is that it cannot.
It is determined up to a linear transformation.3 Just as the von Neumann-Morgenstern
utility function is determined up to a linear transformation by assuming that the individual
behaves so as to maximize the probability-weighted sum of the utilities resulting from the
various outcomes of a gamble, so too is this function specified by assuming that an indi-
vidual acts as if to maximize a weighted sum of instantaneous utilities arising at different
points of time.4

III. THE OPTIMAL PLAN AS SEEN TO-DAY

In this section we explore the following preliminary problem. A consumer at time

X- = 0 wishes to maximize his utility functional:
rT

- X(t - 0) - u [C(t), t] dt
(3) (Do0

1Of course, it could drop to zero once the date is passed. This possibilityis excludedin the functional
(2), as is any interactionbetween the time-distanceand calendar-datebases of discounting; otherwise
either X or u or both would have both t - T and t as arguments.
1 rather assume that in speaking of economic myopia (The Economicsof" Welfare)Professor Pigou
2
had the time-distanceconcept in mind. CertainlyJevons did when he wrote: people of good sense will
not discount the future except for uncertainty-but people do discount the future in accordancewith its
remoteness." (The Theoryof Political Economy,pp. 77-80), and Bohm-Bawerk: " It is one of the most
pregnantfacts of experiencethat we attach a less importanceto future pleasuresand pains simply because
they are future, and in the measurethat they are future." (ThePositiveTheoryof Capital,tr. by W. Smart,
p. 253). For other examplessee also Alfred Marshall,Principlesof Economics,8th ed., p. 120, Adolphe
Landry,L'Interetdu Capital(1904),ch. X, §150and E. C. K.Gonner, Interestand Saving(1906), p. 36.
3 See, on this point, Paul A. Samuelson, op. cit.
4There is no presumptionthat these two measuresof utility are the same. Measurementis arbitrary
and for differentpurposesdifferentmeasuresmay be the most convenient. Cf. my statementin " Cardinal
Utility," AmericanEconomicReview,XLIII, 2 (May, 1953), p. 397 : " Furthermore,the acceptanceof
the von Neumann-Morgensternmeasure does not preclude the definition of still other measures. It is
true that the von Neumann-Morgensternmeasureis convenientand manageablefor the class of problems
involving risk, but it need not prove convenient for all classes of utility problems that may conceivably
arise. Nothing rules out the usefulnessof anothermeasurefor anotherpurpose."
If we were to introducerisk explicitly into the present analysis, we would have to write:
rT r00
·^D= X -(t - [) .u [ p(C,t) - v(C) dC, t] dt,

wherep(C,t) is the probabilitydensityat t of consumptionC and where v(C) is the von Neumann-Morgen-
stem measureof the utility of C. Probabilitiesare, of course, introducedhere in a very special way, being
made to depend only on C and t.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 MYOPIA AND INCONSISTENCY 169

with respect to C(Ot) and subject to the constraint:

0
T

(4) C(t) dt - K(O)

0
where K(0) is a constant stock at t == 0. The constraint here is a simple one, but it will
do for our purpose.'
The present problem is one in the calculus of variations and, assuming piecewise
differentiability, first-orderconditions for a maximum may be found as follows.
Define:

(5) y(t) - C(t) dt

J o
so that:
y(O)= 0,
(6) y(T) = K(0)
and:
(7) y = C(t).
(The dot above the y indicates a derivativewith respect to time).
The problem then is to maximize:

(8) | X(t- 0) -u[yf(t), t] dt = X(t) u[p(t), t] dt

0 0
subject to the fixed end points given by (6). This is an elementary problem now in standard
form, and the solution is provided by Euler's differential equation:

(9) dt [ a- X(t)u[p(t), t] ] = y X(t)u[y(t), t]

which, upon substituting C(t) for y(t) and letting Uc(t) = a u[C(t), t]/KC, is:
(10) X(t)uc(t) + uc(t) X°(t) =- 0, 0 < t < T.
(10) may be written as:
(11) X°(t)/ X(t) = - lC(t)/Uc(t) 0 < t < T,
or as:
(12) d In X(t)/dt = -d In uc(t)/dt, 0 < t < T.
The solution to this differential equation is simply:
(13) A(t) * uc(t) = constant, 0 < t < T,
"
with the constant dependent on K(0). This i's to say that the stock of consumption"
K(0) must be distributed over the interval 0 to ITso that the discounted marginal utility of
consumption is the same for all dates. This is, of course, a quite obvious condition, and,
in view of the absence of intertemporal complementarity, will assure us of a maximum for
0oprovided that the instantaneous utility function displays a diminishing marginal rate
of utility. This would be the case for any consumer who does not concentrate his entire
consumption for the period at a single point of time.
1 Often, for a constraintof this sort (as in the case of a man who must ration fresh water to himself
during an ocean voyage), there is the additionalrequirementthat C(t) > 0 for all t, but such a condition
is not imposed in what follows.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
170 REVIEWOF ECONOMICSTUDIES
IV. THE QUESTION OF INCONSISTENCY
Equation(11) along with (4) thereforepicks out from all possibleconsumption-time
curvesthat one whichis optimalfor the consumerat - 0. But it would be a mistake
to conclude that, even under conditions of certainty, the optimal curve is the one which
the individualwill actually follow. The difficultyarises becauseall we really know is
that this is the curve he will start to follow. It is his best consumption plan at 'X = 0. At
a later date, X > 0, he may (or must, if he is to maximise 'D, x > 0) reconsiderhis plan,
and x cannot then be droppedfrom (8) becauseit is no longerzero.' The problemthen
is to maximize:
rT

14) X (t - T) u[C.(t), t] dt

with respectto C(O)} and subjectto

0

(15 { C(t) } given, and

I C(t)dt= K(T),

where:

(16) K(r) = K(O) -

C(Q)dt.
Define:

(17) y(t) C(t)dt

so that
y(O) = 0
(18) y(r) = K(O)-K(X)
y(T) K(O)
and:
(19) A(t) - C(t).
The solution is given by:
(20) XO(t- )/ X (t - -uc(t)luc(t), t < t < T,

with K(X)determiningthe constantof integration. This solutionmay be entirelydifferent

from that of (11) becausethe discountfunctionhas been shifted. To continueto obey a
fixed consumptionplanjust becauseit was otpimalwhen viewedat an earlierdate is not
1 Following a suggestion by P. N. Rosenstein-Rodan(" The Role of Time in Economic Theory,"
Economica,Feb., 1934, p. 84), we may want the economic horizon to move uniformlywith an increasein
t. Although T remainsconstant, any horizon in the utility functionalmay be consideredto change under
the following assumptions. Let the horizon at X = 0 be H, 0 < H < T, with H definedby the condition
that k(t) = 0 for H < t < T. Then at time t, the horizon may be taken as H + ' with X (t - t) always
zero for t - T > H, provided H + X < T. Our analysis is valid then over the period 0 < 'X < T - H.
It is to be noted, however,that there is no movinghorizon possiblein the budgetconstraintconsideredhere.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 MYOPIA AND INCONSISTENCY 171

rationalif that planis not the optimalone at the presentdate. The bestplanwill generally
change with a change in -x, and there is nothing patentlyirrationalabout the individual
who findsthat he is in an intertemporaltusslewith himself-except that rationalbehaviour
requireshe take the prospectof such a tussle into account.
Should the individualre-evaluatehis plan periodically,his actual behaviourcould
be describedin termsof the graphbelow. C0 may be his best plan at -r= 0; C1 may be
his best plan at X = r ; C2 at r =-r2 ; etc. If he does not reconsiderhis originallybest
planduringthe period0 < r < r,, he abidesby it and follows C0duringthat periodas is
indicatedby the heavypart of the curve. At xr, however,he reconsidersand, takinginto
account that the " stock of consumption" availablefrom then until T is the original
amountminuswhat he has alreadyconsumed,he chooses C1as his best plan and follows
it (along the heavyportion)until -r2when he reconsidersonce more, etc. This meanshis
actualbehaviouris representedby the sequenceof heavy arcs shown in the figure. The
integralof this sequenceof arcs must, of course,be K(0).

Cl
~---C2

0 T, T2 13 T
Figure 2

If the plan is re-evaluatedcontinuously,any single plan chosen has validityonly at

= xr,and actual behaviour is then given by the locus of the C(t) for t = C as determined
by (15) and (20) as r proceedsfrom 0 to T.
In this sectionwe have consideredthe actualdynamicsof utilitymaximization,as dis-
tinct from the mere plan for the future which is made at a given moment, and have
questionedwhetherthe actual path of consumptionover the period would be the same
as that whichis chosenas optimalat the begirningof the period. In the next sectionwe
show that it need not be.

V. CONSISTENCY,INCONSISTENCY,AND THE NATURE OF THE

DISCOUNT FUNCTION
Underwhatcircumstances will an individualwho continuouslyre-evaluteshis planned
course of consumptionconfirmhis earlierchoices and follow out the consumptionplan
originallyselected?

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
172 REVIEW OF ECONOMIC STUDIES

This requiresthat if { C*(t) and C*(t) mazimixe

r't rT
(21a) X(t-O)-u[C(t), t]dt + -\ (t-O) u[C(t), t]dt

rD ,* f T
subject to C(t)dt
( = K(0), then { C*(t) 1

should maximize
.T
(21b) I X(t-,')-u[C(t)]dt
JT
Tr r r T )
subject to | C(t)dt = K(O) - C*(t)dt, or that { C*Qt)

should be the solution both to

(22a) °(t- 0)X(t - 0) = - i(t)c(t), T < t < T,
and
- -f t(t)uc(t), ' < t T.
(22b) X°(t ')/(t- T') <
This must hold for all T'. Equating the left-hand sides of (22a) and (22b), it is clear that
as a necessary and sufficientcondition the logarithmic rate of change in the discount function
must be a constant, so that for T = 0.
(23) d In (t)/dt = In k (a constant), 0 < t < T,
or:
(24) X(t) kt, 0 < t < T.
This we shall call the harmony case. It requires that the discount function be of a very
special form, namely that all future dates should be discounted at a constant rate of interest.1
This is what our intuition should lead us to expect. In the language of the discrete
case, a discount function of this sort means that the relative importance of 1957 and 1958
is the same in 1957 as in 1956. Consequently, when in 1956 one decides how to apportion
consumption between 1957 and 1958, this is the same decision one would make in 1957.
Thus, in 1957 the plan laid down in 1956 is confirmed.
But so far we have adduced no reason why an individual should have such a special
discount function, i.e., no reason why the defect in the telescope that Professor Pigou
spoke of should be logarithmically linear with respect to the distance of the object being
viewed. Indeed, if it is believed that this special case is realistic, a rationale is needed.
We provide one in the next section.
1 The reader is cautioned that this is not the same as the familiarproposition that the marginalrate
of substitution between present and future consumption is determinedin equilibriumby discounting at
the market rate of interest. The marginalrate of substitutionbetweenconsumptionat time 0 and at time
t is, for the utility functionalused here, equal to:
x (t) /
X(0) 3C(0)
and not to X(t)/ x(0). If one unit of consumptionat time 0 can be exchangedfor (1 + r)t units at time t,
where r is the rate of interest,utility maximizationrequiresthat:
1 _ k(t) au / C(t)
(1+r)t (0) au / C(0) '
which is quite differentfrom the requirementthat: t
X(t) = (1 r) X(0)= kt )(0).

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 MYOPIA AND INCONSISTENCY 173

VI. TWO STRATEGIES IN THE FACE OF INCONSISTENCY

An individual who because he does not discount all future pleasures at a constant
rate of interest finds himself continuously repudiating his past plans may learn to distrust
his future behaviour, and may do something about it. Two kinds of action are possible.
(1) He may try to precommit his future activities either irrevocably or by contriving a
penalty for his future self if he should misbehave. This we call the strategy of precommit-
ment. (2) He may resign himself to the fact of intertemporal conflict and decide that his
"optimal" plan at any date is a will-o'-the-wisp which cannot be attained, and learn to
select the present action which will be best in the light of future disobedience. This we
call the strategy of consistentplanning. These possibilities will be discussed in order.
1. The Strategy of Precommitment. To-day it will be rational for a man to jettison
his " optimal" plan of yesterday, not because his tastes have changed in any unexpected
way nor because his knowledge of the future is different, but because to-day he is a different
person with a new discount function-the old one shifted forward in time. Yet it is also
rational for the man to-day to try to ensure that he will do tomorrow that which is best
from the standpoint of to-day's desires. Unpleasant things which to-day we want to do
sometime in the future are continually put off until tomorrow (the " manana effect ")
unless we can find some way of precommitting ourselves to actually doing the task tomorrow.
Consequently, we are often willing even to pay a price to precommit future actions (and
to avoid temptation). Evidence of this in economic and other social behaviour is not
difficult to find. It varies from the gratuitous promise, from the familiar phrase " Give
me a good kick if I don't do such and such " to savings plans such as insurance policies
and Christmas Clubs which may often be hard to justify in view of the low rates of return.
(I select the option of having my annual salary dispersed to me on a twelve- rather than on a
nine-month basis, although I could use the interest!) Personal financial management
firms, such as are sometimes employed by high-income professional people (e.g., actors),
while having many other and perhaps more important functions, represent the logical
conclusion of the desire to precommit one's future economic activity. Joining the army
is perhaps the supreme device open to most people, unless it be marriage for the sake
of " settling down." And, of course, regretting either course later on (at least for the
moment) is to be expected, for otherwise precommitment would have had no purpose.
The worker whose income is garnished chronically or who is continually harassed by
creditors, and who, when one oppressive debt is paid, immediately incurs another is
commonly precommitting. There is nothing irrational about such behaviour (quite the
contrary) and attempts to default on debts are simply the later consequences which are to
be expected. Inability to default is the force of the precommitment.
What needs to be explained is not that people do precommit their future actions, but
that the practice is not still more wide-spread. The reason it is not, I believe, is because
of the presence of risk and uncertainty, both as to future tastes and future opportunities.
Because of risk and uncertainty, people are also willing to pay for options permitting them
a greater range of choice at future dates' and this is of overwhelming importance, especially
as it affects the detailed aspects of future behaviour.
2. The Strategy of ConsistentPlanning. Since precommitment is not always a feasible
solution to the problem of intertemporal conflict, the man with insight into his future
unreliability may adopt a different strategy and reject any plan which he will not follow
through. His problem is then to find the best plan among those that he will actually
follow.
1 See Tjalling C. Koopmans, " Utility Analysis of Decisions Affecting Future Well-Being,"(abs.),
Econometrica,April, 1950, 18: 174-175, anld " La notion d'utilit6 dans le cas de d6cisions concernant
le bien-etre futur", Cahiersdu Seminaired'Ecoio,ne'trie,ed. Rene Roy, 1953.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
174 REVIEW OF ECONOMIC STUDIES

Returning to the narrow framework of our consumption problem, let the best plan

of consumption be given by the function 4 (t) . As a solution, N(t) has the property

that at the limit where AT ---- 0:

X( -r) u[2f(t), ] df + (t-r)'u [(), t] dt

(25) 2o T
'T
+ u(t -) * U[2(t), t] dt

is a maximum of:
.TT - t T

i(t- T) 'u [(t), t] di + X(t- )u [(t), t] dt

(26)o .T
"t

+ u(t-T) u [Z(), t] di
*T+AT
rF+Atq
with respect to the function i y(t) ) when the individual at T can force himself to
I T J
follow during the period T to T + A' any plan which he may select subject to the con-
-T+ A I'T ,'T
straint that y(t)dt = K(O) _- 2(t)dt - z(t)dt, a constant.
·/T ,0 ,'T+AT
To maximize (26) requires only that the middle term be maximized. This gives:
(27) X°(t- T)/ X(t - T) - - uc(t)/uc(t), T < t < r+AT,
so that as AT --> 0, the condition on 2(t) becomes:
(28) X°(o)/ x(0) = --t 'c /uc(:).
Since this must hold for every T it may be written:
(29) X°(0)/ X(0) = - uc(t)c(t), 0 < t - T,
which is the solution to the " harmony case," given by:

(30) kt-T u [y(t), t] clt.

Again, this should not be in any way surprisingbecause only those plans that maximize
a functional such as (30) are attainable (will be obeyed). The individual must, therefore,
first substitute for his true discount function one that is linear in the logarithm and then
maximize. The appropriate value for k has already been determined by the analysis which
has preceded. At each point of time the individual equates - uc(t)/Uc(t) to X°(0)/X(0),where
X(t - r) is his true discount function. Since, when acting in this way he also equates
dkt'-r
-uc(t)/uc(t) to -,t-- /kt-T it follows that:

(31) cl ln kt-/dt = d In x(t - ) /dt

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 MYOPIA AND INCONSISTENCY 175

Consequently:
(32) In k = X0(O)/X(O)= X'(0)
(the second equality by normalization) or:
(33) k = antilog 'A(0).

The individual therefore abides by the consumption-time function which maximizes

T
(34) D= (antilog XP(O)
)t-t u [C(t), t] clt.

The subscript X is deleted from (D because (F is the same functional for all -X under these
circumstances.
A graphic interpretation of this result is given by Figure 3. For the one-parameter
family of functions kt-r, that one is chosen which is tangent at t = X to the true discount
function X(t - r). For the " strategical" man who cannot precommit his future conduct
it is now clear that the only relevant characteristic of his true discount function is the rate
at which it changes at the present moment (at t 4'

-
[antilog P(o)tT

1
~~~~(t-T)
t
r ~~~~~~~~T
Figure 3

'A word on the rationaleof this solution is in order. We think of all deliberateaction through time
as involving precommitmentsover successive intervals,if the intervalsare sufficientlysmall ; and in the
mathematicaltreatmenthere we are simply consideringthe limit form of such behaviourwhere those small
intervalsapproachzero. Actually, it is impossiblefor an individualto choose his rate of consumptionat
every time independentlyof his consumption-timepath immediatelyprecedingand still have piecewise
X

differentiability,for then we brush up against Taylor's Theorem. (Consider the person who can vary
manuallythe flow of waterout of a tap and the deliberationof which he is capable while letting waterflow
out.) The solution we have obtainedin this section is, however,entirelyconsistentwith Taylor'sTheorem.
In the discretecase two differentinterpretationsof the problem are possible: (1) Each period the
individual selects his consumptionfor that period and there is no intra-perioddiscountingor allocation
problem. (2) At alternativeperiods the individual selects his consumption for that period and the next
and allocatesthis amount betweenthe two periods. It is only the latter interpretationwhich is comparable
with our continuous model.
A theory based on interpretation(1) appearsto be more complex and less fruitful in its implications,
although similarto the presenttheory in that strategiesof precommitmentand of consistentplanningarise.
The author's work on this alternativeapproachis, however, not yet complete.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
176 REVIEW OF ECONOMIC STUDIES

VII. GENERALIZATION
What becomes of our results if more general utility functionals and budget constraints
are considered ?
Suppose that:
(T
(35) D = *-
X(t-T-) U[..., t]dt

is to be maximized with respect to a variety of functions Ct), C2(t), . .., each

relating an economic variable to time. Let u depend on these functions in as complicated

a way as we please, providedonly that T does not appearin the u function. u may, for example
T )
have as its arguments various integrals, derivatives, and lead or lagged values of Cl(t) f,
0
r T }
C2(t) , . . . Moreover, let the maximization problem be subject to constraints of what-
0 J
ever complexity, provided only that there exists a solution and that T does not enter into

these constraints, except that {C(t) , C2(t), ... are taken as historically given.

What is the necessary and sufficient condition on X(t - r) in order that planning
should be consistent ? This requires that whatever maximizes hD should maximize
+ iA X, for any A,7 0 < r + A: < T. For this to occur D+ A , must be a monotonic
increasing function of ID. Since T does not enter in u [ . . . , t] we need not concern
ourselves with that function. The question then is, how must X(t -M - A'r) be related
to X(t - r)? It is not enough that these should be monotonically increasing functions
of one another because of the variational nature of the problem. Each must be a positive
scalar multiple of the other, so that the scalar can be factored out of the integrand function
and placed in front of the integral sign. Thus, if x(t - T - A') = constant x ;(t - 4),
the solution to which is x(t - x) = kt-T, with the constant ratio equal to k-A T, we have:

-
:+A l X(t - - -A) u[. . , t]dt
(36) Jo

- k- A
= k-AT x(t - T) u[..., t]dt 0,

and this is the same solution as was obtained for the illustrative case treated earlier. The
important thing is that the relative weights of different dates should be invariant, and no
harm is done to the analysis by considering this more general type of functional and
constraint.
The qualifications underlined two paragraphs above are the significant ones. Should
T enter the problem in a more general way than by merely causing a linear shift of the
discount function, the problem of consistent planning becomes more involved. However,
even though the instantaneous utility function or the basic constraints do change with the
passage of time, if the individual does not take these future changes into account in deciding
what to do now, our analysis will still have validity for any fixed point of time.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 MYOPIA AND INCONSISTENCY 177

VIII. THE DISCOUNT FUNCTION

Special attention should be given, I feel, to a discount function, such as that shown
in Figure 3, which differs from a logarithmically linear one in that it " over-values" the
more proximate satisfactions relative to the more distant ones.' Such a funiction suggests
that individuals who precommit their future actions or who naively resolve now what
they " will do " in the future, commonly do not schedule the beginning of austerity until
a later date. How familiar the sentence that begins, " I resolve, starting next . . ."' ! It
seems very human for a person who decides that he ought to increase his savings to plan
to start next month, after first satisfying some current desires ; or for one to decide to
quit smoking or drinking after the week-end, or to say that " the next one is the last one."
It has been customary for the United States Army to offer voluntary enlistees a
furlough starting with the date of enlistment. This practice is not needed to enable a
man to put his affairs in order-he can do that first and then enlist-but it does serve
as an enticement to those who want the paternalism (" security ") of the army, but do not
want it right now.2 The many schemes for instalment buying (notably of used automo-
biles in the U.S.) which require " no down payment and nothing due for two months "
are evidence of the effectiveness of enticements of this same kind. Indeed, all purchases
on credit can be viewed as precommitments that often (although not always) exchange
future costs for a present pleasure.3
My own supposition is that most of us are " born " with discount functions of the
sort considered here, that precommitment is only occasionally a feasible strategy (because
of risk and uncertainty), and that we are taught to plan consistently by substituting the
proper log-linear function for the true one. Children are known to discount the future
most precipitously and the " virtue " of frugality is something to be instilled when building
" character." True discount functions become sublimated by parental teaching and
social pressure, and the inconsistency problem considered in this paper becomes lost from
sight. There is a rationale for discounting at a constant rate of interest. In some cases
training may be so effective that the individual's original discount function can no longer
be said to be his " true" one. His tastes have changed and his discount function has
become log-linear or perhaps even constant. His is the " harmony " case. Precommitment,
therefore, is never attractive to him-even under certainty. In other cases,4 however, we
may say instead that a person has been taught to plan and behave consistently and not
that his tastes have been molded. His is not the harmony case. Such a person will, from
time to time, depart from the consistent pattern of behaviour, sometimes because pre-
commitment becomes feasible (and this is always his preferred strategy under conditions
because precommitment becomes feasible (and this is always his preferred strategy under
conditions of certainty) and sometimes because of lapses that result when the true weight

1 As an empirical supposition, there is a precedent for this in Bohm-Bawerk,op. cit., pp. 257-258,
the originalsubjectiveundervaluationsare, in the highestdegree,unequaland irregular. In particular,
so far as the undervaluationis caused by defects of will, theremay be a strong differencebetweenan enjoy-
ment which offers itself at the very moment, and one which does not; while, on the other hand, there may
be a very small difference,or no differenceat all, betweenan enjoymentwhich is pretty far away, and one
which is fartheraway." In context it is clear that he is referringhere to changes in the rate of discount.
This same possibilitywas consideredby Marshall,op. cit., MathematicalAppendix, Note V.
2
Since marriage was mentioned earlier alongside joining the army as a possible precomm-niitmenit
strategy, I cannot avoid remarkingfacetiously that marriagetoo is commonly preceded by a period of
engagement!
3 There are, of course, other importantreasons for buying on credit.
4 For the distinctionmade here I am indebtedto Mr. W. B. Reddawayand others in attendance at a
seminar on this paper at the Departmentof Applied Economics, CambridgeUniversity.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
178 REVIEW OF ECONOMIC STUDIES

function becomes momentarily ascendant. These lapses are the splurges, binges, and
extravagances which we all know.1
This picture is typical, I suppose for most of us, but there are no doubt some who,
either through lack of training or insight, have never learned to behave consistently and
for whom the intertemporal tussle remains unsolved. These people we call " spend-
thrifts."2 By contrast, those who have taken on log-linear discount functions have
learned to be " thrifty."
Spendthriftiness, in the general sense of inconsistent or imprudent planning,3 is by
no means insignificant. It is especially among the lower-income classes, where education
and training are commonly blighted, that one would expect to find imprudent behaviour
of this sort.4 In America, lower-income people tend to gorge themselves with food after
pay-day ; overheat their homes when they have money for a bucket of coal ; are extrava-
gant, going on sprees on pay-day, not budgeting their money, and engaging in heavy
instalment buying ; do not keep their children in school ; and are freer in the expression
of their sexual and aggressive impulses.5 Their high birth rate is well-known. All these
behaviour characteristicscan be explained as a failure to cope intelligently with the problem
of the intertemporal tussle. Obviously, this is not the entire story; but the observations
are consistent with the hypothesis presented, and it would be upsetting if the facts werc
otherwise.
The character of behaviour under precommitment is more difficult to label. Its
results are somewhat ambiguous. Sometimes precommitment causes an individual to
sacrifice future pleasures heavily for the sake of present ones, e.g., to go into debt to make
possible an expenditure providing mostly present gratification. But at other times precom-
mitment seems more " wholesome,"6 as when a person contracts to save a certain amount
each month or goes into debt to buy a house. The distinction, I feel, is this: in the one
case the present is heavily favored at the expense of the future; in the other an allocation
is made among various dates of the future in accordance with weights given by the lower,
but more level portion of the X(t- r) function, as depicted in Figure 3. Precommitment
then has the effect of precluding grossly unequal allocation withinthat future period of time
once it moves into the present.
1 To one who would say that to discount the future for remotenessat all scems to him foolish and
irrational,I should reply that he is one who received very strong trainingas a child which went beyond
simply teaching him the strategy of consistent planning and effected such a change in his tastes that he
now finds it unnaturalto discount the future on this account. Moralizingagainst discountingthe future
in this way has, of course, found its way into the prominentliteratureon this subject. See, e.g., Bohm-
Bawerk,op. cit., pp. 253-255.
2
If the x(t--) function rose to the right of -r,inconsistencybetweenplans and behaviourwould lead
to " miserliness" with the individualsaving for a future plannedexpenditurewhich he continuouslypost-
pones. True miserlinesswould, however, appear to be better explained in other terms, money or wealth
becoming an object of desireper se.
* The " spendthrift" might, of course, also be definedas anyone who discounts the future because of
its remoteness,but I think this catches somewhatless satisfactorilythe essence of the term because such a
person may behave quite prudently-empty moralizingaside. Mr. Thore has pointed out, however, that
my spendthriftmight display more conservativebehaviourthan anotherperson who has a lower k even
though he plans consistently.
4 Perhaps in some underdevelopedeconomies this problemwould take on its most serious dimensions.
The notion that the incidenceof the sharp discounting of the future is greater among " primitiveraces,
children, and other uninstructivegroups in society" was asserted clearly by Fisher in The Theoryof
Interest,ch. IV, § 9, and earlier by Jevons irt The Theoryof Political Economy,ch. II and Bohm-Bawerk,
op. cit., p. 244.
5 See W. Allison Davis, "Child Rearingin the Class Structureof AmericanSociety " in The Familyin
a DemocraticSociety: AnniversaryPapersof the CommunityServiceSocietyof New York,1949and Robert
F. Winch, The ModernFamilv, 1952, pp. 93-94. These authors ascribe these behaviourpatternsto anxiety
feelings stemming from the insecuritythat results from low and irregularincome. It is not clear to me
why this anxietyshould lead to spendthriftinessratherthan to miserlinessor to painstakingfamilybudgeting.
If the facts were the opposite, would one feel any less comfortablewith the explanationgiven?
6
The word is in quotes becausewe do not intend to provide moraljudgmentsat this point.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 MYOPIA AND INCONSISTENCY 179

IX. CONSUMER SOVEREIGNTY

What becomes of the concept of consumer sovereignty for dynamic decision-making
problems? To the extent that consumer sovereignty is one of our values, ought we to
allow people to behave imprudently? Should we permit them the strategy of precom-
mitment ?-e.g., should a man be allowed to sell himself into bondage for the sake of anl
immediate gratification of desires ?1 What ought to be our view of the irrevocable trust,
especially if the maker tries to revoke it? At which date should sovereignty inhere in the
maker? And ought we to instruct people to substitute a log-linear discount function for
the true one ? If so, should the constant k be selected so that the derivatives of the log-
linear function and the true one are equal at t = - ? Why should this be the appropriate
value ?
My view is that these questions are difficult to answer mainly because consumer
sovereignty has no meaning in the context of the dynamic decision-making problem. The
individual over time is an infinity of individuals, and the familiar problems of interpersonal
utility comparisons are there to plague us. The interpersonal aspect of the intertemporal
problem becomes clear if we think of a similar problem involving a family of brothers
where each has a utility functional depending not only on his own utility but upon a weighted
sum of the utilities of all of them. Suppose the oldest brother always has the power to
allocate the annual proceeds of an estate, but with it being foreknown that each year one
brother will die off, the oldest next. The shifting of the discount function of the family head
gives rise to the danger of inconsistent planning ; and the family head of the moment may
consider the alternative strategies of (a) an irrevocable trust, or (b) playing his favorites
extra heavily now knowing that they will be out of favor at a later date. What can the
detached view of consumer sovereignty be in this context !
Pigou2 and others have regarded " myopia" as an excuse for state intervention in
determining allocations over time (investment). But on what basis ought the state to
make these decisions ? Ramsey3 contended that all weights should be equal ; but at
least for those problems involving the allocation of a limited stock of goods over time
(i.e., the problem of " conservation "), this proposal becomes meaningless if one contem-
plates future generations ad infinitum.
More questions have been raised here than I am prepared to answer ; but somewhat
out of practical considerations I would suggest the following: The individual can probably
do as good a job as the state or any other agency in determining allocations for himself
as of future dates, provided the future dates include none which are proximate. That is
to say that I would have confidence in the judiciousness of a person to-day, if he is not
ignorant of future facts, to decide how much to save and how much to spend for the rest
of his life, starting a couple of years from now. Mr. Smith is probably in as good a position
to evaluate the relative importance of the Mr. Smith of 1960 and the Mr. Smith of 1970
as anybody else, and I would therefore permit him, in the absence of risk and uncertainty,
to precommit his future allocations provided the period of precommitment did not begin for
a couple more years. The real decisions to worry about are those where an immediate
or proximate satisfaction is gained at the expense of still-more-future costs. Precom-
mitment may be regarded as either good or evil depending upon whether the period of
precommitment begins now or later.

1 Typically,our commonlaw has held manycontractsthat partakeof this characterto be unenforccable.

The problemhere, ethically,is akin to that of whethera body politic should be permittedto vote itself into
a dictatorship.
2
A. C. Pigou, TheEconomicsof Welfare,ch. 2, pp. 23-30.
3 F. Ramsey, " A MathematicalTheory of Saving,"EconolnicJoutrnal,1928.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions
 180 REVIEW OF ECONOMIC STUDIES

The strategy of using a log-linear function also seems to have some appeal from the
" social" point of view. It means that the individual always decides what to do now
on the assumption that he has no authority over his future self. The individual cannot
decide what C(t) shall be for t > r, except that he can decide now what K(t) shall be for
t -X + dt. If the derivative of A(t- r) equals zero at t = r, then, of course, this case
reduces to that of" no myopia," the situation which Ramsay regarded as best. But, these
remarks notwithstanding, my own view is that the ethical issues presented by the problem
of dynamic choice remain basically unanswered if not unanswerable.

X. SUMMARY
To summarize, we have said that the optimal plan of future behaviour chosen as of a
given time
(A) may be a plan which will be followed under conditions of certainty (the harmtiony
case), or
(B) may be inconsistent with the optimizing future behaviour of the individual (the
intertemporaltussle). In this latter case
(1) the conflict may not be recognized and the individual will then be spendthrifty
(or miserly), his behavior being inconsistent with his plans, or
(2) the conflict may be recognized and solved either by
(a) a strategy of precommitment,or
(b) a strategy of consistent planining.
We have, moreover, hypothesized that the typical discount function has the shape
of X(t - r) as shown in Figure 3, and have argued that this hypothesis is consistent with
observed behavior.
Finally, we have challenged the meaning of the concept of consumer sovereignty
in this context of dynamic utility maximization.'

Evanston, Illinois R. H. STROTZ

1We have here treatedthe problenm of the intertenmporal

tussle only in the context of microeconomics
Similar issues may arise, however, in the aggregatecase where a group of persons or an economy must
decide the distributionof economic activity over time. Political decisions to eliminate a foreign trade
deficitor to balancea budgetnot this year, but next may serve as illustrations.

This content downloaded from 188.72.96.141 on Wed, 25 Jun 2014 08:10:16 AM

All use subject to JSTOR Terms and Conditions