Download View/Open

Research Review New Developments in Macroeconomic Theory: A Prospectus and Appraisal
By Roger K. Conway and James R. Barth·
I
Although the battles that are fought In
appear to be fought
[the economlCS lIterature]
wlth antique pop guns, the bullets are real and they
may'soon be fired at you
1\
I
i'
I
Robert Solow
Introduction
There lS no longer, If there ever was, any questIOn
but that the workIngs of the macroeconomy have
major lmpacts on the agricultural sector Indeed,
one of the Invlted paper seSSIOns at the 1982 Ameri­
can Agrlcultural Economlcs ASSOCiatIOn (AAEA)
meetIngs focused excluslvely on "Current Macro­
economlC Pollcles and U S Agrlculture" Further­
more, some of the lmportant lnfluences of macroeco­
nomlC pollcy on the farm sector have been recently
and extenslvely dlscussed In an excellent survey ar­
ticle by Prentice and Schertz (80) and In an artlcle
by Tweeten (108) m whlch the latter goes so far as
to state that "the economlC structure of the farmIng
Industry In the long run wlll depend more on fed­
eral taxatlOn, money supply and trade pollcles In
the Federal Reserve System, the Internal Revenue
Servlce, and Department of State than' on commodity
programs In the U S Department of Agrlculture"
(108, p 864)' Although the Prentice and Schertz
monograph provldes a useful d,scussIOn of standard
monetarist and Keyneslan theOries, both of whlch
to varyIng degrees provlded the foundatlOn for the
economlC poilcles lmplemented durIng the fift,es
through the mldseventIes, a new theory-the new
classlcal macroeconomlcs-also merits attentlOn
Th,S recent theory gamed promInence durIng the
·Conway IS an economist WIth the NatIOnal Economics Thvl­
81On, ERS Barth 18 a profe880r In the Department of Economics
at George Washmgton UnIversity and 8 VIsIting scholar W1th the
Federal Reserve Bank of Atlanta The authors Wish to thank
Lorna Aldrich, DIck Hwdacher. Dick Heifner, Nadme Loften,
Tom Lutton, Lloyd Teigen, DaVId. Torgerson, Paul PrentIce, Lyle
Schertz, and Paul Sundell, 811 of the U S Department of
Agnculture (USDA), Michael Bradley of George WaahlDg\on
Uruverslty. Robert Keleher of the Federal Reserve Bank of
Atlanta, and especially P A V B Swamy of the Board of Gov­
ernors, Federal Reserve Board, and George Washmgton Univer­
Sity, (or theu valuable comments ~d help on an earher verSIon
of thJS paper
IItahclzed numbers In parentheses refer to Items In the Refer
enceB at the end of thiS artlcle
turbulent seventIes, when the clalms that fiscal and
monetary pohcy could smooth the busmess cycle
were called mto questlOn Broadly contemptuous of
the fact that the predommant Views of the macro­
economy seemed to be unable to explam the recent
perIOd of slmultaneously hlgh InflatIOn and slugglsh
output growth, some economlsts clalmed thls new
theory could explam how to reduce lnflatlOn qUIckly
and wlthout loss of output
We belIeve It lS m the Interest of the agrlcultural
commumty to become famlilar wlth th,s new theo­
retical development In macroeconomlCS because It
may, as Tweeten suggests, have a substantlal lnflU­
ence eventually on what happens m the agrlcul­
tural sector Moreover, th,s theory already has
operational components directly-applIcable to the
agrlcultural sector (see, for example, 13, 33, 37, 41,
102, 105) Our artlcle wlll, therefore, attempt 'to
present, as well as to appralse, the new classlcal
macroeconomlCS One of our maJor conciuslOns lS
that the new classlcal economlCS lS far more comple­
mentary wlth tradltIonal macroeconomlC theory
than elther advocates or detractors have thus far
emphaslzed
The New Classical Macroeconomics
A cruclal element of the new classlcal approach lS
the modehng of bUSIness cycles wlthm an eqUlll­
brlUm framework As It lS assumed that expecta­
tlons are formed ratlOnally, these models are fre­
quently referred to as ratlOnal expectatlOnB models
However, th,s assumption lS but one of several lden­
tIfyIng th{s class of models' The other maJor
aSsumptlOns are (1) perfect wage and price flexlblh­
ty WIll occur, leading to contmuous market cleanng,
(2) ecOnOIDlC agents possess lmperfect InformatIOn
about some lmportant economlC varlables, (3) some
form of the "Lucas-RappIng supply functIOn" ap­
phes m whlch (as described' later) nomInal shocks
have only temporary lnfluences on real varlables,
and (4) a vanant of the natural rate hypothesls
·We shall deSCribe these models synonymously as new clasSical,
ratIonal expectatIons, or eqwhbnum cycle models
AGRICULTURAL ECONOMICS RESEARCHNOL 35, NO 3, JULY 1983
23
holds m which perceived nommal vanabies produce
no real effects The Importance of these assumptIOns
will become clear as we descnbe a basiC rational
expectations model
rational expectatlOns'model We start'by assumIng
that the demand for goods and services (an IS' equa­
tion) IS given by the followmg equatIOn (where all
subsequent vanables except the mterest rate are m
natural logarithms)
Rational Expectations Defined
Y~ =
The concept of ratIOnal expectatIOns was first
postulated by Muth (74) when he suggested that, as
expectatIons are Informed predictIOns' of future
events, It'seems only'reasonable to consider them
exactly like predictIOns from a relevant'theory of
economiC behaVior This postulate can be stated
more rigorously as follows Based on a given Infor­
matIon set, the subjective probablhty dIstnbutlOns
of outcomes are assumed to be distributed about the
objectIve probability chstrlbutlOns of outcomes This
equatIng of subjective and objective probablhty chs­
tributlOns IS based on the notIOn that, because Infor­
matIon IS a relatIvely scarce commodlty, economic
agents wlthm an economy Will use'lt optImally
More speCifically" the structure of the pertment'eco­
nomic model IS assumed to be'lnformation which IS
known and thus used m the formatIOn ,of expecta­
tIon' However, It IS not assumed that the predIc­
tions of mdlvlduals are always-correct (perfect fore­
Sight) or that expectatIOns are the same for all
agents (homogeneous expectations)
Malharls and Brock (70) contend that ratIOnal ex­
pectatIOns IS a plaUSible theory because (1) the
hypotheSIS IS apphcable ,to all dynanuc problems,
and expectatIOns m different markets would not
have to be treated m different ways, (2) If expecta­
tIOns were not moderately ratIOnal, there would be
opportumtIes to make profits, and (3) rational
expectatIOns IS a hypotheSIS that can be modified
With Its analytIcal methods remammg apphcable m
systems With mcomplete or mcorrect InformatIOn
(70, p !53)
A Basic Rational Expectations Model:
Its Policy Implications
We wlll use a Simple model developed by Sargent
and Wallace (90) to deSCribe the basiC features of a
IAdaptIve expectatIOns IB another wlCiely employed expectatIon
scheme, whereby the expectations of 8 vanable are based. solely
on past values of that variable Only In the unlIkely case that
pnces, when actually modeled, are found to mapley 8 random
walk Will rational and adaptive expectat1()D8 Yield the same
results
24
Y, - a(r, -
P1) + u lt
(1)
where
Y·,
Y,
r,
demand for goods and ,services,
the trend or natural level of output,
= the nommal rate of mterest,
~+1 = expected price for period t + 1 deter­
mmedln period t,
p., = The expected rate of mflatlOn deter­
milled In period t - I-that IS,
u lt
(P~ + 1 - P~), and
= a random error term such that
U It ' "
N(O,
"1) Turmng to the money market (or an LM equatIOn),
we can assume that the money demand functIOn IS
given by
M:' = p, + 'YY,
-
Tr,
+ u"
(2)
where
M:'
P,
= the demand of money,
= actual pnce level durmg period t,
''v,
u,.
eqUlhbnum output, and
= a random error term With
u,. '" N(O,
,,~)
whereas the money supply (M') IS exogenously
determIned by monetary authOrities and Will be ex­
plamed more fully
Aggregate supply IB assumed to be given by the
follOWIng equatIOn •
Y~
= Y, + fl(P, - P:) + u"
(3)
where
Y:
=
=
aggregate supply, Vat
a random error term'such that U 3t '" N(O, .r,),
and the other vanabies are as described above
.. Notice that when there IS no UDcertamty so that
P: • Pt , then the aggregate supply curve becomes vertical
and one obtaIns essentially the ClaSSICal model-hence, the
term new claSSIcal macroeconomu:s For the theoretical
development of the new cl&B8lcal aggregate supply curve, see
(68, 60, 62, 94)
Usmg equatIOns (1), (2), and (3) and the equlhbrIum
or market clearmg conditIOns, M' = M' = M and Y'
= Y' = Y, to solve for the reduced-form prIce equa­
tion, one obtams­
-lQy
Ta p.
P, -_ """"j')
,+""""j') , + 1
+
(JJ - a)1
D
+ (3-ya p. '
a
1
+ D (M, - 110,) + D
u ll
1+'ra
-
D
where U.
=
(J1
D u lt
(Ja
-
""""j') 110, -
(6)
(9)
18
a random
error'80
~
IL
D -"
(1
-
+ ora)
D
u" ­
+ a(M, - E(M,)
D
(7)
Substltutmg equatIOn (7) mto equation (3) Yields
=
a(J
Y, + """"j') (M, - E(M,» + U.
II
a(J
+ """"j') u" + U. or
assumed throughout that E(u 1 ~) = 0 for all I '¢ J where
4 This assumption IS somewhat dublouB as equations
(1·3) are Simultaneous so that Jomt probabilIty rustrlbutlollB are
Implied Sargent and Wallace make thiS aSBumptlOn for computa
tlOnal tractability, however. as thiS need not be·the case, thiS
particular appro~ch 18 hmlted
lIt
(8)
= Y,
The Importance of thiS result IS that the systematic
component h IS not present In,equatlOn (11), which
means that the monetary authorities cannot sys­
tematically affect output Only the random element
In the inoney supply, u... affects output, but thiS
component represents unanticipated movements,
and IB thua uncontrollable by the monetary
authOrIties On the.basls of thiS finding, Sargent
and Wallace estabhshed the "superneutrahty" pro­
poSitIOn of monetary pohcy differences between ac­
tual and full employment output follow a random
1,J
Y,
a~)
(11)
(6)
This equatIOn states that the conditIOnal expecta­
tion of the prIce level IS determIned by future prIce
expectatIOns and the expected money stock Next,
we subtract the expected prIce level from the actual
price level"Yleldlng.
= D1 -_ ul l
N(O,
(10)
Y,
P, - E(P,)
""
Subtractmg equatIOn (10) from equatIOn (9) Yields
M, - E(M,) = U" This means that equatlOn'(8)
reduces to
ApplYing the conditional expectatIOns operator to
equatIOn (4), one obtBlns
+ ~ E(M,)
that u4t
This equatlOn.mdlcates that the monetary authorl
ties Will mcrease (decrease) the money supply dur­
mg recesslOnary (mflatlOnary).perlOds Based on the,
prInCiple of rational expectatIOns, one can now solve
for the expected money supply Domg so Yields
where 0. _1 represents the Information set aVBllable
to every econOmIC agent at time t - 1
E(P,)
u"
Accordmg to equatIOn (18), output depends only on
the difference between the actual money supply and
the expected money supply To determme the ex­
pected money supply, we now postulate a pohcy
feedback rule wherem pohcymakers attempt to
stabilIZe output at the full employment level ThiS,
procedure IS captured m the followmg equatIOn
where u4t
Asswrung ,that expectations are formed rationally,
then
a~ + (J'r a
+ ora)
D
(4)
u"
where D = (J1 + a + 'r(Ja
(JJ -
(J(1
IS
= I,
25
walk, and thus are not subject to bemg stablhzed by
the monetary authorities'
Okun (77) and Schultze (92) crlhclze Sargent·and
Wallace for pretendmg that "all the world were a
wheat Pit" They argue that the real world IS pre­
dommantly' composed of customer markets, not auc­
hon markets, whICh are based on forward contracts,
thereby Implymg conSiderable wage and price rigid­
Ities Subsequent papers have, therefore, mtroduced
rigidities mto the Sargent-Wallace model and have
then exammed the resultmg ImphcatlOns'
Will not be fully and contemporaneously reflected In
current nommal wage rates Taylor (10ll and Gray
(46) also develop Similar models of overlappmg
wage contrac!'S m which, many glven,perlOd, some
contracts are,bemg negotiated while others are still
m force In thiS SituatIOn, monetary pohcy IS agam
effective because of the sluggishness of wages
relative to pnces
Whether or not the strong pohcy Impotence result
holds Will depend mamly on the specification or" out­
put given by'equatlOn (3) In one study, when Lucas
(57) assumes that productIOn IS withheld m antlcl­
Phelps and Taylor (79) speCify yet another model,
one based on the assumptIOn that firms set then
prices m the period prevIous to the one m which
they sell their output This type of advanced price
settmg enables monetary pohcy to affect real varia­
bles Without Influencmg prices In the current
period The followmg aggregate output equatIOn as
pOSited by Phelps and'Taylor mcorporates th,s price
pabon of hIgher prIces In future perIods, actual out·
stlcklneSB
put IS then Influenced by the systematic component
of the pol !cy feedback rule In a subsequent article,
however, McCallum (65) once agam obtams the
Sargent-Wallace neutrahty result usmg the Lucas
supply functIOn speCificatIOn, but only after d,s­
countmg E(Pt + ,) appropriately by the rate of
mterest
The pohcy Impotence hypotheSIS no longer holds
once certam mowficatlOns are ,made m the output
supply equatIOn For example, Fischer (36) speCifies
a model m which wages are constant for two periods
because of contracts In thiS case, the aggregate
supply functIOn becomes
y.,
=.!.
2
(P, • '" 1
E (P,)) + u"
(12)
"0
+ "1
E, _ 1 (P,) +
"2
P, -
1
+
", M, + u"
(13)
In thiS type of model, monetary pohcy can clearly
,affect real output f!:owever, McCallum (67) demon­
strates that slow prlce'adJustment per se IS not m­
compatible With the pohcy meffect,veness hypoth­
eSIs: One can demonstrate thiS hypotheSIS by pOSlt­
mg an aggregate supply equatIOn m which the
superneutrahty result holds no matter what rigidi­
ties are ,mtroduced The key factor guaranteemg
the policy meffect,veness hypotheSIS m, thiS situa­
tIOn IS that the length of the period durmg which
the wage IS rigid IS no longer than the penod for
pohcy actIOn to be' effective'
,- 1
Aggregate supply IS now no longer mfluenced only
by the current mformational error between actual
and expected price Instead, monetary pohcy, even
With the assumptIOn of ratIOnal expectatIOns, can
systematically Influence real wage rates and there­
by aggregate output because the authOrities' actions
'Note that'all right hand Side variables are considered or
thoganBI to the error term when rabonal expectations models
are estImated Thus, tests of cawiallty or, more properly, ex­
ogenelty are closely hnked to the time senes models of rational
expectatIOns theorists (see, for example, (84, 89» Unfortunately,
causahty teste are subJect to senoUB theoretical and empIrical
comphcabonB WhICh, In our View, hmlt their usefulness See (28)
for a ~diSCUSSIOn of thiS POInt Note also that the emphasiS 18 on
CYclIcal movements, which 18 why applied rational expecta­
tlOnIsts detre'nd and deseasonallze their data One attempts to
work With economiC time series which are, In other words,
covariance statIOnary and purely mdetennlnIstic
26
y: =
The Problem of Persistence
Stnctly mterpreted, the expectatlOnal errors
described above, should be randomly distributed
over time If expectatIOns are formed ratIOnally If
so, It wrectly follows that output should be uncorre­
lated over time Neei:lless to say, thiS IS not the case
'Readers are adVised to consult the survey article by McCallum
(68) for a more comf.lete diSCUSSIOn of alternative supply func
tiona and their Imp Icabons for policy effectiveness One ffilght
add that, although McCallum's counterexamples are effective
debBtmg POints, they appear somewhat artifiCial In descrlbmg
real world relative rigIdities between wages and policy Imple
mentation Perhaps ~recogruzmg thiS, McCallum (66) provl~es a
more recent and shrewder critique of stIcky wage/price models
when one examines the data Qwte the contrary, all
the avaIlable eVIdence indIcates that output and
unemployment are hIghly serIally correlated The
observed "persIstence" of posItIve output-prIce and
money-output relatIOnshIps over bUSiness cycles
would seem to contradIct polIcy neutralIty
Most rational expectatIOns supply functIOns are
modIfied to mclude a lagged output term as an ex­
planatory variable to account for thIS observed
8erIai correlatIOn However, th,S procedure for deal­
mg wIth persIstence has been crItiCIzed by
Modlgiiam (73) as being ad hoc, as no rIgorous
theoretIc framework Is'provlded to JustIfy the mclu­
SIOn of th,S addItIOnal varIable In response, ra­
tional expectatIOns adherents have subsequently of­
fered several explanatIOns to account for the lagged
term One explanatIOn IS based on the notion that
costs of adJustment prevent Instantaneous adJust­
ments Another, advanced by Lucas (55) and ex­
plICItly derIved by Sargent (88), IS' based on mforma­
tIon lags WhICh lead to serIal correlation. More
speCIfically, a model based on Phelps' (78) work
WIth dIspersed markets, called "Islands," and mfor­
matIOnal dIscrepancIes IS used to generate demand­
Induced prIce-employment correlatIOns m the
economy as well as to prOVIde a theory of the per­
SIstence effects asSOCIated WIth aggregate-demand
shocks. Lucas' basIC argument IS that economIC
agents engaged m market actIVIty are exposed to a
contmuum of dynamIC and unantiCIpated opportum­
tIes that wJ!1 not remaIn constant whIle an mdIVId­
ual conducts an mformatIOn search In such a SItua­
tIon, real and nommal effects cannot be ,mmed,­
ately dIsentangled 80 that serIal correlation occurs
Another explanatIOn for serIal correlatIOn or perSIS­
tence IS based on the eXIstence of mventorIes
BlInder and FIscher (14) demonstrate that a sudden
prIce change can be met by altermg mventorles 80
that changes m productIon wJ!1 be less than sales If
prIces remam the same m the next perIod, produc­
tIOn could be Increased to rebUIld mventorles to
theIr orIgInal level As a result, a,lagged mcrease
m productIon of varymg duratIon would occur whIle
the orIgInalmventory stock IS bemg restored Even
If no surprIse prIce change occurred m tIme t + 1,
the output gap m time t + 1 would stIll depend
poSItIvely on the output gap m preVIOUS perIods
Unfortunately, none of these explanatIons for
perSIstence appears entIrely satIsfactory
Rational Expectations, Stability, and
Multiple Equilibria
Thus far we have examined the ratIOnal expecta­
tIons hypotheSIS as well as the use of lagged
VarIables to explain observed prIce and output
movements Whenever lagged varIables appear,
however, the questIOn of convergence of the eco­
nomIC system naturally arIses When consldermg
thIs questIOn, Sh,ller (95) contends that "these (ra­
tIonal expectatIOns) models may explode rather
than converge
because the expectatIOns
mechamsm held at tIme t mfluences the behaVIor
of Y and hence the reVIsed expectatIons mechamsm
m tIme t + 1 Because of the arbItrary nature of
the adJustment mechamsm, we cannot generally
tell WhICh models WIll explode and whIch WIll not"
Furthermore, he argues that the pOSSIbIlIty eXIsts
that "we are left WIth a fundamental mdeter­
mmancy for the solutIon of ratIOnal expectatIOn
models and mfimty of potentIal solutIOns for all but
those degenerate models whIch YIeld zero order dIf­
ference equatIOns"
Both these Issues are serIous because they suggest
that ratIOnal expectatIon models may be speCIfied so
as to permIt the pOSSIbIlIty of self-fulfillIng unstable
expectatIOns As Flood and Garber (38) note, a prIce
bubble can develop when the actual market prIce
depends poSItIvely on Its own expected rate of
change Because economIC agents under ratIOnal
expectatIOns WIll not make systematIc predIctIOn er­
rors, prIce and Its expected rate of change are POSI­
tIvely related, suggestmg a smular relatIOnshIp be­
tween prIce and ItS actual rate of change Thus, one
may be m a SItUatIOn In whICh an "arbItrary self­
fulfilling expectatIOn of prIce change may dr,ve ac­
tual prIce changes mdependently of market funda­
mentals" (38, p 746).
RatIonal expectatIOns proponents vary In theIr
response to thIS CrItICIsm Muth (74) OrIgInally at­
tempted to d,sm,ss thIS Issue by pOinting out that
"for a bounded solutIOn, the coeffiCIent of the larger
(unstable) root vamshes, the mItIal condItIon 18
then fitted to the coeffiCIent of the smaller root" In
other words, If the general solutIOn for prIce IS
(14)
and A, > 1, then A2 = 0 and Al IS determmed by
the ImtIal condItIon However, Sargent and Wallace
27
argue against unstable roots because of "a ter­
minal condItion that has the effect of ruling out
'speculative bubbles' " Minford (71) expresses
perhaps the most popular defense against thIs par­
ticular weakness when he asserts that past hIstory
has not prOVIded examples of prIces exploding to In­
fimty Because thIs mformatlOn becomes a part of
the model Itself, It IS presumed that one need only
conSIder convergmg solutIOn • Nevertheless, those
who beheve the dynamICs for explOSIve behaVIor
stIll eXIst, have challenged thIS argument, although
admItting that some exogenous actIOn (for example,
pohcy) may serve to dampen the instabIlity,
(90)
BurmeIster (20, 21) beheves the non-umqueness
questIon may be more'seMOlla for equlllbrium
models than for those models WIeldIng "free param­
eters" and prodUCing dlseqwlibrlum prIce adJust­
ments whIch are anathema to the new claSSICIsts
In any event, the Issues of stablhty and multIple
eqwhbrlum have not yet been adequately resolved
m the new clasSIcal economICS hterature
The Lucas Critique and Time­
Inconsistency
Perhaps the most Important consequence of the ra­
tional expectatIOns hterature IS hypotheSIS, formu­
lated by Lucas (56), that the behaVIOr of economIC
entitIes IS not invarIant to changes In the econODllC
environment For example, when the Government
changes Its pohcy rules, It IS assumed that'IndIvld­
uals WIll observe thIS change and modIfy their
behaVIOr so as to maxImIze their objectIve func­
tIOns ThIs means that tradItIonal econometrIc
models may be serIOusly flawed when used for
policy analYSIS because these models SImply extra­
polate past behaVIOr based on old pohcy rules Into
the future In econometrIc parlance, the hnkage be­
tween behaVIoral and policy coeffiCIents (referred to
as "cross-equatlOn restrictIOns") needs to be taken
mto account Stated another way, the coeffiCIents In
those equatIOns deSCribing the behaVIOr of economIc
agents (such as supply and demand fuctlOns) should
be restricted by the coeffiCIents (or pohcy param­
eters) m t!Iose equatIOns modehng econODllC
poliCIes Hansen and Sargent (48, 49) are among the
'However pnce bubbles have occurred WIth InmV1dual com­
modJhes whether for 17th century Dutch tuhps or for world
sugar dunng the 1973 74 commodJty boom See Conway (26) for
8 cbscU8810n of true Issue
I
28
first to do empirIcal work taking these restrictIOns
mto account, partIcularly WIth respect to large
econometflc models Their analysIs suggests that m­
corporatmg the appropflate the appropriate restric­
tIons WIll be extremely dIfficult·
Kydland and Prescott (53) and Prescott (81) have
taken the "Lucas Cfltique" one step further by
arguIng that formulatmg Government pohcy uSing
optImal control theory IS Invalid when economIC
agents maxImIze their objectIve functIOns uSing ra­
tIonal expectatIOns ,because any policy result WIll be
eIther "tIme-InconsIstent" or suboptImal By tIme­
consIstency, they mean the followmg Suppose IndI­
VIduals formulate their optImal behaVIOr In an inI­
tIal peflod t where they have accurately antICIpated
aU future Government policy varIable magmtudes
for each time peflod over a fimte hOflZon If these
indIVIduals follow theIr optimal plan untIl a perIod J
and then recalculate an optImal plan for all suc­
ceedIng peflods, the orlgmal plan IS conSIdered to
be time-eonslstent when the recalculated optImal
plan is SImply the continuatIOn of the oflg:Inal plan
deflved In the first peflod for all perIOds after J
Kydland and Prescott further state that when ex­
pectatIOns are formed ratIOnally by economIc agents
and when they are aware of the authorIties' decI­
SIon rule, the Bellman pnnClple of optImahty IS In­
applicable To understand theIr pOint, conSIder the
optImal rule for patents as an Instance of tIme­
mconslstency Ideally, the Government would like
to break Its commItment WIth fIrmS ellJoylng
monopoly patent rights and thereby have the ad­
vantage of a competItIve environment WIth the cur­
rent technology now used under eXIstIng patents
However, the Government would also have to
pledge to honor future patents to stImulate new in­
ventIOns In'a future tIme penod, however, these
new inventIons would be old and, once again, the
Government could terminate all past patents As
expressed by Barro, "the pohcy that optImIzeS at all
dates subject to the ImtIal condItIons applYing at
those dates IS subject to a tIme-mconslstent proper­
ty where earher plans are not followed through" (4,
p_ 69) Once ' anticIpated by pnvate mdlvlduals, a
steady or fixed policy IS a tIme-eonslstent pohcy, but
-The lDSlght of Lucas 18 mdeptmdent of ratIonal expectations so
that one could concelV8 of "Irrational expectatIOns" produclDg
the same parameter instablhty In econometrIc models See (I2)
for estunatlon of a polIcy reaction functIon
It results ,In suboptimal technological advance
Analogous policy consIderatIOns can easily be
Imaged wIthin the context oHhe agncultural sec­
tor, such as international trade agreements, buffer
stock prIce stabIlizatIOn programs, and crop set­
aSIde programs Some cntlcs, like Modlgllam (73),
have argued that time-inconsIstency would emerge
as an obstacle only when pollcymakers and prIvate
mdlvlduals maxmuze heterogeneous and mcompatl­
ble objective functIOns However, Calvo (22) has pro­
VIded a counter example, demonstrating the
pOSSIbIlity of a tlme-mconslstent result emerging
even when pollcymakers and prIvate mdIvlduals
have the same objective functIOns
Kydland and Prescott's findmg apparently supports
the supenonty of rules rather than dIscretIOn In the
conduct of econom;c policy The longstandmg debate
between rules and dIscretIOn can thus be reVIsed In
the framework of the new claSSIcal models to dlstm­
gUlsh between rules WIthout feedback and rules
WIth feedback Now that "optimal" policy rules
mIght be suboptImal or time-inCOnsIstent, a dIlem­
ma emerges as to whether pohcymakers should
adhere to theIr onglnal optimal rule or subsequent­
ly revIse It Kydland,and PreBcott prefer adherence,
whereaB LucaB and Sargent (64) beheve thlB rule
doeB not neceBsanly follow, baBed on the aBBump­
tlOnB of theIr model InBtead, LucaB and Sargent
contend that It IB entIrely posBlble that pohcy­
makers could achIeve a hIgher level of total (addI­
tIve) mtertemporal utlhty by contmually revlsmg
theIr plans every penod'than they could by adher­
mg to an mltlal pohcy program The counter­
argument m favor of ruleB put forth by Kydland
and PreBcott IB that the Government runs the nsk
of "crymg wolf" They specifically clrum that pohcy
announcements by the authontles would lose theu­
credlblhty If the announcements were contmually
Bubject to revIsIon Lack of credIblhty, m turn,
would Increase the varIance of an mdIVldual's,reac­
tlon functIOn, make Its Identification more prob­
lematic, and thus lead to mdetermlnate conse­
quences To aVOId thIS SItuation, Kydland and
Prescott, as well as Lucas, suggest that one
SImulate over a number of dIfferent pohcy rules and
then Implement the one WIth the supenor operating
charactenstlcs The Blgmficance of thIS proposal IS
that authontles would no longer seek the optimum
opt.morum, but would be forced to settle for a sec­
ond beBt solutIOn
Credibility: Can Disinflation be Painless? Accordmg to ratIOnal expectatIOn proponents, credl­
blllty'by both monetary and fiscal authontles IS
neceBBary for Btable economIc growth In particular,
mflatlOn can be reduced only by a change m the
current and proBpectlve fiscal and monetary pohcy
regimes that wIll reduce mflatlOnary expectatIOns
Temporary reBtrICtlve fiscal and monetary actlOnB
wIll not Buffice, only a permanent change m the
pohcy regime from "what IS perceIved as a long­
term Government' policy Involvmg the, hIgh average
rates of government defiCits and monetary expan­
810n In the future" wIll keep (87, p 2) If Buch an
"abrupt" change m Government strategy "IS suffi­
cIently bmdIng as to be WIdely beheved," Sargent
argueB that the COBt m foregone output and unem­
ployment could be Bmall As eVIdence for hIS POSI­
tion, Sargent (86, 87) clteB events In AUBtna,
Hungary, Germany, Poland, France, and Czechoslo­
vakIa durmg the twenties as examples where
"credIble" poliCIes ended dramatIc hypermflatlOns
or prevented them from occurnng at near-zero out­
put cost
ImpoBmg a gold Btandard on a government IS one
way to eBtabllBh credIbIlIty, Sargent contends,
because the government IBsumg demand notes and
long-term debt would promlBe to convert any out­
Btandmg notes mto gold, on demand, or to levy suf­
fiCIent taxeB to honor Its debt That government
would be expected to meet ItB debt obhgatlOns whIle
SImultaneously reframmg from seignIOrage '"
Sargent and Wallace (91) and Sargent (85) rely on
the credIbIlIty Issue to argue for coordInatIOn of
monetary and fiscal poliCIes because of theIr connec­
tion VIa the Government budget restramt The Fed­
eral Reserve Board's polIcy regardmg open market
operations Wlllmfluence the flow of potentIal
seignIOrage revenues WhICh, aB noted below, IS a
part of the mtertsmporal budget conBtrammg fiscal
polIcy They Bhow that thIS BltuatlOn Btrongly m­
fluenceB the deslrablhty of Fnedman'B "k-percent
rule" for the monetary base ThuB, one can enVIsIOn
a SItuatIOn where such a rule IS undeSIrable,
whether over the long run or'not FlrBt, It could
happen when the amount of outstandIng Govern­
I'Se.gnlOrage 18 revenue from money creatIOn For an analYSIS,
see (100
29
-
-" .,
ment Interest-bearmg debt IS so large that a pretax
real Interest rate greater than the rate of growth m
fiscal output IS needed to place the debt Second, a
sltuatlOn could occur m whlCh the monetary and
fiscal authorIties clash If fiscal pohcy dommates
monetary pohcy so that there IS an Ironclad rule for
Government expenditures and exphclt taxes, then a
concomitant monetary accommodatlOn may be needed
to finance the growth rate of total Government m­
debtedness Fmally, there IS the case m which the
deficit path the authorIties select Is'slgmficantly m
deficit m a present value sense ThiS conditIOn
seems to represent the economIC SItuatIOn In many
Latm AmerIcan countrIes today
Arguments In support of the F!ledman k-percent
rule hmge on whether or not the dommatmg fiscal
regIme Just descrIbed occurs A k-percent rule dog­
gedly adhered to by the monetary authorities IS a
way to dlsclplme,fiscal pohcy Thus, a monetary
authOrIty that IS stronger than Its fiscal pohcy
counterpart wlil announce and admlmster a
k-percent rule for the monetary base, thereby 'dlc­
tatmg the flow of seignIOrage revenues for the fiscal
authOrIties, who are then forced to lower Govern­
ment expendItures, Increase tax revenues, and/or
employ bond financmg so as to balance the budget
ThiS scenarlO seems to descrIbe the current situa­
tIOn m West Germany where the Bundesbank has
adopted such a rule and appears to have successful­
ly adhered to It
New Classicism: Empirical Results
EconometrIc mvestIgatlOns of the rational expecta­
tIOns hypothesIs, although numerous, suffer from
severe hmltations One problem IS that ratIOnal ex­
pectatIOns and adaptive expectations are "observa­
tIOnally eqUivalent," as Sargent (89) demonstrated
ThiS means that both expectatlOnal theOrIes may
produce results that are equally consistent With the
data As a result, economists cannot test directly
which theory IS best supported by the eVidence
Indirect tests have therefore been attempted, but
With mixed results Barro (7, 8, 9), m particular,
has produced many studies attemptnig to determme
whether only unanticipated money (or money
shocks) affects output HIS findmgs mdlcate that an­
tiCipated money growth IS neutral, whereas unan­
tiCipated money growth has real effects However,
by respeclfymg the way expectations are formed,
30
Small (96) and Gordon (42) generated results that
contradict Barro's findmgs This reversal of results
pomts out a weakness m testmg ratlOnal expecta­
hons models, namely, that there IS a Jomt hypoth­
eSIs mvolvmg the speCificatIOn of the model as well
as the expectatIOns mechamsm
Furthermore, only If strong assumptlOns hold can
parameters be found that remam mvarlant under
policy changes (as Lucas deSires) In thiS SituatIOn,
a varymg parameter approach IS the most general
because It makes fewer a prIOr< assumptIOns about
the economiC environment if the coeffiCients do
vary, then It IS ImpOSSible "to roll back econometrIC
models to the levels where parameters do remam
mvarIant" (106, p. 12) Swamy, Barth, and Tmsley
(99) have also demonstrated that there are serIOus
problems,wlth the ARMA models used to estimate
the unobservable exogenous processes m the new
claSSical models They POint out specifically that
statIOnarity has been mapproprIately assumed and
that there are IdentificatIOn problems aSSOCiated
With statIOnary models and unrestricted reduced
forms which have not yet been addressed U
One very suggestive study by Neftci and Sargent
(75) argues that a change m the distrIbuted lag
relatIOnship of output on the actual money supply
wlil occur when the feedback policy rule of the
monetary: authOrities changes at some pomt m time
If the neutrahty' hypotheSIS IS true, whereas no
change Will ,occur If a relatIOnship between output
and mnovatlOns I~ the money supply IS found They
hypotheSized a policy break durmg 1964 and, usmg
a Chow test, found the eVidence supported the neu­
trality propOSitIOn Yet the Chow test IS not robust
aa there IS a difficulty m correctly characterlzmg
pohey behaVlor by time perIOds Furtherlllore, Neftcl
and Sargent's alternative hypotheSIS for thiS test IS
only a smgle arbitrary break m policy regIme In
fact, a structural break IS pOSSible m every perIod
and, on that baalS, Swamy and Tmsley (100) show
that the Chow test IS statistically mconslstent, and
Resler, Barth, Swamy, and DaVIS (82) diSCUSS prob­
lems With detectmg structural change based on ex­
Isting tests
llConway and Fryar (27) prOVide a further dISCUSSIOn of
econometric estimatIOn difficulties With ratlOnal expectatIOns
models
Theoretical Weaknesses of the Rational
Expectations Approach
One fundamental problem analyzed by Swamy,
Barth, and TInsley (99) IS the contradIctIOn between
the Muth-LuCRs-Sargent ratIOnal expectatIons
hypotheSIS and the BayeSIan defimtIon of ratIonalIty
based on coherence and other consIstency reqUIre­
ments " As they POInt out, the problem IS that the
ratIonal expectatIOns hypotheSIS requIres the
equahty of subjectIve ,and ObjectIve probabIlItIes
(they rely on three drlferent defimtIonsof objectIve
probabIlItIes for comprehenSIveness) of events In
contrast, the subjective BayeSian prInCiple of ra·
tIOnality rehes entirely on subjectIve probabIlItIes,
thereby reqUIrIng no such equahty They also argue
that the ratIonal expectatIOns hypotheSIS may be
conSIdered a lInguIStIC axIOm, thIS beIng too restrIc­
tIve for practIcal use As an alternatIve, Swamy and
others bUIld a competItIve, lOgIcally conSIstent,
mIcroeconomIC model based on a BayeSIan frame­
work of sUbjectIve expectatIons WhICh, when aggre­
gated across Ind,V,duals, becomes a StochastIC coeffi­
CIent macroeconomIc model SUItable for empIrIcal use
RatIonal expectatIons proponents are CIrcumspect In
dISCUSSIng how the "true" underlYIng structure of
the economy becomes known, how fast agents can
obtaIn InformatIon abo\lt thIS structure, and what
the InformatIOn costs are The aBBumptIon,by pro­
ponents appears tradItIonally to be that IndIVIduals
acqUIre knowledge of the structure oLthe economy
rapIdly and freely for epIstemolOgIcal ease A more
realIstIc VIew IS that It takes a long tIme for
economIC agents to d,scover bf they ever do) the
true structure of the economy and hence to adJust
theIr expectatIons approprIately For thIS reason,
ratIonal expectatIons may be more approprIately
conSIdered to be a longrun, steady state phenome­
non However, If a break In structure occurs durIng
the tranSItIon perIod toward a longrun eqUIlIbrIum,
problems arIse A more reasonable assumptIon,
noted by FrIedman (39), IS that knowledge of the
new structure IS costly to obtaIn and, therefore, WIll
be acqUIred adaptIvely If so, actIVIst polICIes under
th,S regIme WIll generate real effects
'-see (31, chapters 6 and 7) for a r~goroU8 defimtlOD of theseu­
10matlc pnnclpies Coherence, as defined by de FlDettl (30), 18 the
most Important consistency reqwrement It amounts to restrict
109 a gwen heher system 80 that It would be lmpomnble for any
mdtvldual to lay multiple fSlr bets lD such a way that the mdl
VIdual WInS under aU posSible outcomes When coupled WIth
other 8.Xloms, the suggestion 18 that the probabilities of elemen­
tary events will sum to 1 and that cODJunctIve aod mSJunctlve
events Will obey both the product and addItion rules
Usmg HICks' (50) termInology, real tIme IS
abstracted from the new claSSICal models, and
lOgIcal, not hIstOrIcal, tIme becomes the cnterIOn, as
ratIOnal expectatIons proponents'conslder the
length of temporal expectatIo!,s formatIOn to be
"whatever It takes" to allow mdIvIduals the oppor­
tumty to acqUIre suffiCIent, InformatIOn to assure
that expectatIOns are fulfilled
FInally, a contrad,ctIOn to the Sargent-Wallace
pohcy Impotence result deSCrIbed above anses when
one takes Into account the portfoho balance theory
as put forth by BUIter (19), Tobm,('103), and Had­
Jlmlchalakls (47) Perfectly antICIpated mcreases m
the money growth rate WIll mcrease the perfectly
antICIpated InflatIOn rate and thereby mduce substI­
tutIOn of capItal for real balances Th,S mcreases
the capItal stock and hence generates Impacts on
output and employment so that money IS not
neutral after all
Implications of the New Classical
Macroeconomics for the Government
Budget
The new claBBIcal macroeconomICS prOVIdes m­
terestmg polIcy prescrIptIons when one conSIders
the Government budget An eXamInatIon of ChrIst's
(24) formulatIOn of the Government budget restramt
(GBR) clarIfies thIS Issue As ChrIst noted, the
Federal Government must satIsfy the follOWIng con­
stramt
G+IB=T+dM+dB-dA
(15)
DefiCIt = G + IB - T = dM + dB - dA
(16)
or
where 1 IS the nominal rate of Interest (1 = r + 11"),
,.. is the expected InflatIon rate, and r IS the real
rate Th,S equatIon states that' Government expen­
dItures on goods and servIces (G) mInUS taxes (T)
plus Interest prud on the Government debt out­
standIng (IB) must be financed by ISSUIng more out­
SIde money (M) and/or bonds (B) and/or sellIng
assets (A) EquatIon 15 clearly shows that the
Government does not have a free hand to determIne
all the'macroeconomIc polIcy varIables SImul­
taneously Instead, at least one pohcy tool must be
endogenously determIned Another POInt made clear
31
by equatIon (15) IS that a government's budget con­
stramt IS not qUlte the same 'as an mdlvldual's
NotIce speclfically that by mcreasmg the money
supply and generatmg InflatIOn, the Government
can reduce Ita real defiCIt thIS It does by reducmg
the real value of Government bonds and other out­
standmg lIabIlItIes However, because the nommal
mterest rate mcorporates mflatlOnary expectatIons,
thIS InflatIonary tax may be moderated to the ex­
tent the InflatIOn IS fully antICIpated and IS thereby
embodIed m the rate of mterest (that IS, .- = "-)
The speclfic Influence of InflatIOn on the,GBR IS
shown by the folloWlng equatIOn"
+ rBIP + .-MIP - TIP - "AJP
d(BlP) + d(MIP) - d(AJP)
GIP
=
(17)
The new classIcal school belIeves that the Influence
of tax cuta on InflatIOn and output depends on
whether or not the defiCIt IS bond-financed If It IS,
there WIll be neIther InflatIOn nor real effecta If m­
dlvlduals engage m "ultraratIonal" behaVIor It In
thIS case, as Barro (1, 6), DavId and Scaddmg (29),
and others argue, there IS no essentIal dIfference
between debt,and tax finance thIS vIew IS also
referred to as the "RIcardIan eqUlvalence" theorem
because It traces back to RIcardo's tract on the full
capltahzatIon of the tax when bond sales are used
to finance- the debt A Government sale of securItIes
whIch IS fully antIcIpated by mmvlduals means that
prIvate spenmng Wlll be reduced by exactly the
same amount as the bond-financed defiCIt because
mdlvlduals Wlll reahze that the Issuance of Govern­
ment securItIes means a hIgher future tax lIabIlIty
for whIch savmgs must be mcreased Under thIS
behavorIal assumptIOn, aggregate demand remams
constant when taxes are cut, leavmg PrIces and real
output unchanged Moreover, the IS-LM curves re­
mam unchanged, thus leavmg the mterest rate
unaffected However, the composItIon of output has
changed, prIvate spenmng has dropped by an
amount equal to the mcrease m Government
spenmng
Of course, strong assumptIOns underhe thIS
analysIs In partIcular, the model depends on fimte­
hved mmVlduals lInked across generatIons by be­
1I8arth and Cordes (10), Barth and Morrell (11), and Turnovsky
(107) diSCUSS these and other dynamIc cOnBlderabona- of the GBR
l.rrhlS behaV10r 18 used to explalO "DemsoD's Law"-the ob­
served stable relationship between gross private BaVIng and
gross national product lD the Uruted States
32
questa" Because of thIS assumptIOn and other fac­
tIlrs, many mdlvlduals have crItIcIzed these models,
mclumng Feldstem (35), Buchanan (17), BUlter (18),
Drazen (32), CarmIchael (23), Stokey (98), Brewley
(16), and Tobm (104). The mam CrItIcIsms are
(1) altrUIsm may be absent between parent and
chIld, (2) the generatIonal cham may be broken by
people Wlthout chIldren, (3) constramts on wealth
transfers among generatIOns may arIse from Imper­
fect rental, mortgage, and annUlty markets, and
(4) survIval IS uncertrun so that the prIvate rIsk of
death wIll be greater than the socIal rIsk Empmcal
eVIdence on the ultrarat10nalIty hypothes1s, as d1S­
cussed m a survey artJcle by Stevens (97) and a re­
cent study by Feldstem (34), seems to md1cate that
dlscountmg of bond-financed defic1ts 1S less than
complete
In contrast to taxes, the output effects of Govern­
ment expenmtures, as demonstrated by Barro (5),
depend on whether they are perceIved as temporary
or permanent In the case of permanent Govern­
ment expend1ture changes, the assumptIOn IS made
that tax rates w1ll be adJusted so as to generate suf­
fiC1ent revenues to match the change m permanent
Government expend1tures Thus, concomItant
changes m tax rates WIll only result'from a new
and permanent level of Government expend1ture
Based on thIS reasomng, Barro (2, 5) and Kydland
and Prescott (52) argue that 1f all changes m
Government spendmg were permanent, then tax
rates should be constant or nearly constant over the
busmess cycle so that the budget would be balanced
on average, WIth surpluses and defic1ts arIsmg only
from random output fluctuatIOns
In contrast, accormng to Barro (3), temporary tax
and Government expendIture changes Slgmficantly
mfluence output because incent1ves for mtertem­
poral substItutIOn of work and productIOn are
created The greatest mcentlves arIse m the case of
transItory expend1tures Wh1Ch are not cons1dered as
close substItutes for prIvate spendmg (for example,
wart1me expenmtures) Ind1v1duals recogmze that
mcome IS temporarIly mcreased and w1ll react by
"A recent and grOWlDg mterest has developed 1D extendtng
Samuelson's (83) overlapPing generations model to examine B
number of macroeconomic IBsues lDvolvmg the mtertemporal
allocation of,resources We pOInt thIs out because thiS type of
model IS an eqwhbrlum one and assumes perfect foreslght or ra
tlOnal ex~tabons See McCallum (69) and Kareken and
Wallace (51) for a ruscUBslon of,thl8 i1terature
temporarIly mcreasmg theIr labor supply At the
theoretIcal level, the basIc result IS that aggregate
demand IS unaffected whIle aggregate supply m­
creases As far as empIrIcal work IS concerned,
Barro (5) found support for the view that temporary
movements m defense purchases, generally assocI­
ated With war epIsodes, almost doubled the response
m output as compared With an equal and perma­
nent shIft m defense purchases
Conclusions
A remarkably persuasIve recent artIcle by Boland
(15) puts forth the proposItion that an empIrIcal test
can be "true" only If the theory behmd It has
lOgical valuiIty More than 2,300 years ago, ArIstotle
constructed necessary conditions m the form of
axIOms for admIssIble arguments (that IS, the prm­
clples of deductIve reasonIng) An argument IS
lOgical when two rules of mference hold (1) modus
ponens, whICh states that the truth of every
assumptIOn Imphes the truth of all conciuslOns, and
(2) modus tollens, which states that when one con­
clUSIOn IS false, at least one assumptIOn IS false It
IS unfortunate that discernIng the truth of assump­
tIons m economic theory IS ImpossIble because the
problem of mductlOn has never been solved Th,s
problem means that the new claSSICists are unable
to have their proposItIons log~cally venfied from the
truth of economIc particulars
As a result, the sophIsticated falSificatIOn cnterlOn
developed by Popper and emphasIZed m economIcs
by Fnedman" conSiders the absolute truth of com­
petmg theones less Important than testmg them to
discover which one has the most successful lOgically
denved conclUSIOns or predIctIOns Th,s process may
be performed by use of statistical techmques and
h,stoncal evaluatIOn The new clasSICists have ,thus
far succeeded m demonstratmg observatIOnal eqUiv­
alence With traditional models as well as m provld­
mg hlstoncal examples that seem to support some
of their arguments Yet others, hke Gordon (43, 45),
have produced h,stoncal counter examples support­
mg the view that a credible reductIOn of Inflation
by the authontles WIll be paInful both m terms of
unemployment and lost output
IGSee (26) for 8 brief dISCUSSion of Popper, Friedman, and
Lakatos, as well as other mfluences on economIC methodology
The contnbutIon of th,s new economic program may
ultimately he not so much m explammg the h,ston­
cal record as m ascertammg the mlcroeconomlC
foundatIOns used to produce Its macroeconomIc con­
clUSIOns The pnmary value of the new claSSIcal
school may be ItS attempt to develop a coherent,
economIc explanatIOn of subjective expectation for­
mation by use of optimIzation and game-theoretIc
techmques so as to explam cychcal movements m
key macroeconomic vanabIes better In thiS
respect, the Lucas cntIque and the tIme mconsls­
tency proposItIon have already emerged as poten­
tIally Important contnbutlOns to mterpretIng and to
reformulating econometnc pohcy analYSIS
11
Econ­
omists are now more aware that authontles are not
plaYIng a game agamst nature whose response IS
mvarlant to pohcy changes Instead, the game IS
played agaInst alert and ratIOnal agents who Will
react to pohcy stlmuh As a result, the new claSSI­
cal macroeconomIcs has been valuable In demon­
stratmg how the length of real pohcy Influence
could be bnefer and that the Impact and predlcta­
b,hty of those pohcles may be weaker than pre­
VIOusly beheved Rather than vlewmg the new
claSSical approach as a dlstmct macroeconomIc pro­
gram, we beheve It may be more appropriate to con­
Sider It a part of the most Important new SCientIfic
research program to emerge over the past decade
InformatIOn theory and probablhstlc economICs 18
Th,s Important research IS bemg rapIdly mcor­
porated mto all areas of economICS, Includmg pubhc
finance, mdustrlal organIzatIOn, labor economiCS,
and pubhc chOIce Although the assumptIOns of new
clasSical models tend to comclde WIth the underly­
mg monetarist models, they can also be modified to
Yield d,seqUlhbnum KeyneSian results 10 The
dramatic pohcy neutrahty conclUSIOn that resulted
m such a highly controversIal mtroductlOn of the
new claSSical models IS bemg toned down conSider­
ably by Its originators As Sargent and Lucas now
state, "the pomt of thiS example was to show that
l"We beheve recent challenges to the Lucas critique by pro
ponents of Vector Autoregre88IOD (V AR) models are misplaced In
addition to McCallum's (66) rebuttal to their argument, time
senes models such 8S VAR suffer from an Incomplete theoretical
development See (28) and (100) for 8 more complete explanatIOn
(54) for aD Ilium mating dISCUSSIon of thiS research
program
I'NotIce that deVIatIOns of variables from thelr trend values
may not be a SIgn of market faIlure, and furthermore, even If
they are, the most appropriate way for the Government to Inter
vene, If at all, may be through prOVIdIng of mformatIon One
way IS through the development of contingent claIms markets,
such as optIOns and futures markets See (40) and (76) for 8 full
diSCUSSion
I·see
33
wIthIn ,preCisely that model used to ratlOnahze reac­
tIve monetary pohcles, such pohcles could be shown
to be of no value It hardly follows that all pohcy IS
Ineffective 10 all contexts" (63, p 60) and "[t]hIS IS a
narrow special case, reqwrIng arbItrary assump­
tions about InformatIOn flow whIch has found httle
or no empIrIcal support" (93, p 11) TIllS Important
quahficatlOn has only recently been adffiltted openly
In summary, we contend that research based on the
new claSSIcal macroeconomIcs approach may be
vIewed as, extendIng, rather than replaCIng, tradI­
tIOnal models Greater recogmtlOn of the stochastiC
nature of macroeconomIcs can only add to pohcy­
makers' understandIng of how the economy works
and can thereby lead to Improved policy recommen­
dations Lucas (59) perhaps put these new contribu­
tIOns 10 the proper context "[tjhere 18 no POInt 10
lettmg tentative and, I hope"promlsmg first steps
harden lOto pOSItIOns that must be defended at all
costs" (59, p 713)
References
0) Barro, Robert J "Are GovernD}ent Bonds Net
Wealth? Journal of PolitICal Economy, Vol
82, Nov -Dec 1974, pp 1095-1117
(2) "UnanticIpated Money Growth
and Unemployment, In the Umted,States,"
Amencan EconomIc ReVIew, Vol 67, Mar
1977, pp 101-15
(8) "Unanticipated Money, Output,
and the Pr,ce Level 10 the Umted States,"
Journal of Polltlcal'Economy, Vol 86, Aug
1978, pp 549-80
(9)
, and Mark Rush" UnantIcIpated
Money and Economlc ActiVIty," In Ratwnal
ExpectatIOns a",d EconomIc Policy (ed
Stanley FIscher) ChIcago Umv of ChIcago
Press, 1980
(10) Barth, ,James R , and Joseph J Cordes "Sub­
stltutablhty, Complemen4nty"and the Im­
pact, of Government SpendIng on EconomIc
ActiVIty," Journal of EconomICs and BUSI­
ness, SprIng/Summer 1980, pp 235-42
(11) Barth, James R , and Stephen 0 Morrell "A
PrImer on Budget DefiCIts," Federal Reserve
Bank of Atlanta 'EconoirHc ReVieW, Aug 1982,
pp 6-17
(2) Barth, James R "Robm Sickles, and Phlhp
WeIst "AssesslOg the Impact of VarylOg Eco­
nomIc CondItIOns on Federal Reserve
BehavlOr," Journal of Macroeconomtcs, Vol 4,
No. I, Wmter 1982, pp 47-70
(3) BelongIa, Mike "Unanticipated Inflation, Ra­
tional ExpectatIOns, and the Relative PrIce of
Food" Paper presented to AmerIcan Agricul­
tural EconomIcs ASSOCIatIOn Annual Meet­
Ings"Logan, Utah, Aug 1982
(3) "Federal DefiCIt Pohcy, and the
Effects of Pubhc Debt Shocks," Journal of
Money, Banking, and Credit, Vol XII, No 4,
Part 2, Nov 1980, pp 747-62
(4) "Intertemporal Sub,stltutlOn and
the Busmess Cycle," Carnegie-Rochester Con­
ference Senes on,Publlc PolICY, Vol 14,
SprIng 1981, pp 237-68
(14) Bhnder, Alan S , and Stanley FIscher "In­
"Output Effects of Government
Purchases," Journal of Pollhcal Economy,
Vol 89, No 6, Dec 1981, pp 1122-41
(15) Boland, Lawrence A "A CrItique of FrIed­
"Reply to FeldsteIn and
Buchanan," Journal of Pol;t!cal Economy,
Vol 84, Apr 1976, pp. 343-50
(16) Brewley, Truman "The IndetermInancy of
Interest Rates" The Center for MathematIcal
StudieS In EconomIcs and Management
(5) (6) 34
"The Eqwhbrlum Approach to
Busmess € y cles," Money, ExpectatIOns, and
BUSiness Cycles Essays In MacroeconomIcs
New York AcademIc Press, 1981
(7) ventorIes, RatIOnal Expectations, and the
Busmess Cycle," Journal of Monetary Eco­
nomICS, Oct 1981, pp 277-304
man'~
CrItIcs," Journal of Economl,C LLtera­
ture, Vol XVII, June 1979, pp 503-22
ExplaIn), Agricultural EconomIcs Research,
Vol. 34, No I, Jan 1982, pp 34:35
SCience D,scusslDn Paper No 491 Evanston,
Il! Northwestern Umv ,Aug 1981
(17) Buchanan, James "Barro on the R,card,an_
Eqwvalence Theorem," Journal of PolitIcal
Economy, Vol 84, Apr 1976, pp 337-42
(18) BUlter, WIllem "Government FInance m an
Overlappmg-GeneratlDns Model wIth G,fts
and Bequests," m SocIal Security Versus
PrIVate SaVIng (ed George M von Fursten­
berg) CambrIdge, Mass Ballmger, 1979,
(19) "Tbe MacroeconomIcs of
Dr Pangloss A CrIttcal Survey 'of the New
ClaSSIcal EconomIcs,'J EconomLc Journal,
Vol 90, Mar 1980, pp 34-50
(27) (28) Conway, Roger K, P A V B Swamy"and
John F Yanagtda "The ImPOSSIbIlIty of
Causahty Testmg " SpecIal Stud,es Paper
Board of Governors, Federal Reserve Board,
forthcomIng
(29) DaVId, Paul A , and John ScaddIng "PrIvate
SavIngs U1trarattonallty AggregatIOn and
"DenIson's Law," Journal of PolltLcal Econ­
amy Vol 9, Mar IApr 1974, pp 225-49
(20) Burm,"ster, Edwm CapItal Theory and
DynamICs New York CambrIdge Umv
Press, 1980
(21) "On Some Conceptual Issues In
RatlOnat ExpectatIOns ModellIng," Journal of
Money, Cred,t and BankIng, Vol XII, No 4,
Part 2, Nov 1980, pp_ 800-16
, and Edward G Fryar, Jr
"RatlDnal ExpectatlDns In AgrIcultural
EconomIcs Research and PolIcy AnalysIs
Some PItfalls" Umv of Arkansas AgrICul­
tural EconomIcs Staff Report, FayetteVIlle,
Ark,1983
(30) de FInettl, B The Theory of ProbabIlity
Vol 1 New York John WIley & Sons, 1974
(31) DeGroot, M H OptImal StatIStIcal DeCISIOns
New York McGraw-HIll Publ Co, 1970
(22) Calvo, GUIllermo "On the TIme-ConsIstency
of Optimal PolIcy In a,Monetary Economy,"
Econometrica, Vol 46, No 6, 1978, pp
1411-28
(32) Drazen, Allan "Government Debt, Human
CapItal, and Bequests In a LIfe-Cycle Model,"
Journal of Pol,tIcal Economy, Vol 86, Apr
1978, pp 505:16
(23) CarmIchael, Jeffrey "On Barro's Theorem of
Debt NeutralIty The Irrelevance of Net
Wealth," American EconomIc ReVieW, Vol.
72, Mar_ 1982, pp 202-13
(33) EcksteIn, ZVI "RatIOnal ExpectatIOns Model­
mg of AgrIcultural Supply" Paper presented
' at AmerIcan EconomIc ASSOCIation Annual
MeetIngs, Washmgton, DC, Dec 1981
(24) ChrISt, Carl F "On FIscal and Monetary
PolICIes ,and the Government Budget
(34) FeldsteIn, MartIn "Government DefiCIts and
Aggregate Demand," Journal of Monetary
EconomICS, Vol 9, Apr 1982, pp 1-20
RestraInt," Amencan Economl.c RevU!w,
Vol 69, Sept 1979, pp 526-38
(35) (25) Conway, Roger K "A Reduced Form PrIce
EquatIOn from a Walraslan Model of the
World Sugar Market" Staff Report No
AGES821115 US Dept of Agr , Econ , Res
Serv, Dec 1982
(26) ReVIew of Mark Blaug's The
Methodology of EconomICs (or How Econoffilsts
"PerceIved' Wealth m Bonds and
SocIal SecurIty A Comment," Journal of
PolitIcal Economy, Vol 84, Apr 1976,
pp 331-36
(36) FIscher, Stanley "Long-Term Contracts, Ra­
tional Expectations, and the OptImal Money
Supply Rule," Journal of PolitIcal Economy,
Vol 85, Feb 1977, pp 191-206
35
(37) FIsher, Brian S "Rational ExpectatlOns'ln
AgrIcultural Economics Research and Pohcy
AnalysIs," AmerIcan Journal of Agricultural
EconomIcs, Vol 64, May 1982, pp 260-65
(38) Flood, Robert P , and Peter M Garber
"Mark!,t Fundamentals versus Price-Level
Bubbles, The First Tests," Journal of PolitI­
cal Economy, Vol 88, No 4, Aug 1980,
pp 745-70
(39) Friedman, BenJamm "Optimal ExpectatlOns
and the Extreme InformatIOn AssumptIOns of
cRatIonal ExpectatlOns" Macromodels:" Jour~
nal of Monetary EconomIcs, Vol 5, Jan 1979,
pp 23-41
(40) Gardner, Bruce "Commodity OptIOns for
Agriculture," American Journal of Agrlcu/.­
tural'EcoTwmlcs, Vol 59, Dec 1977,
pp 986-92
(41) GoodWin, Thomas H , and Steven M Sheffrln
"Testing the RatIOnal Expectations Hypoth­
eSIs In an Agncultural Market;" ReVieW of
EconomICS and Statistics, Vol 64, Nov 1982,
pp 658-67
(42) Gordon, Robert J "New EVldence'That Fully
Anticipated Monetary Changes Influence
Real Output After All," The Center for
Mathematical StudieS In EconomICS and
Management SCIence DISCUSSion Paper No
369 Northwestern Umv , Evanston, TIl , Apr
1979
(43) "Output FluctuatIOns and
Gradual Price AdJustment," Journal of Eca­
nomIC Literature, Vol XIX, No 2, June 1981,
pp 493-530
(46) ,Gray, Jo Anna "Wage IndexatIOn A Macro­
economic Approach," Journal of Monetary
EconomIcs, Vol 2, Apr 1976, pp 221-36
(47) HadJlmlchalakls, Michael G Modern Macra­
economICS, Englewood ChITs, N J Prentlce­
Hall Inc, 1982
(48) Hansen, Lars,Peter, and Thomas J Sargent.
"Formulatmg and Estimating Dynamic Lmear
RatIOnal ExpectatIOns Models," m RatIOnal
ExpectatIOns and Econometric PractIce (ed R
E Lucas, Jr and T J Sargent) Mlnneapohs
Umv of Minnesota Press, 1981
(49) "Linear RatlOnal ExpectatIOns
Model. for Dynamically Interrelated Vari­
ables," Journal of EconomIC DynamICs and
Control, Vol 2, Feb 1980, pp 7-46
(50) HiCks, John R Causahty
York BaSIC Books, 1979
In
EconomICs New
(51) Kareken, John H , and Ned Wallace (ed.)
Models of Monetary EconomICs Federal
Reserve Bank of Mlnneapohs, Jan 1980
(52) Kydland, Finn, and Edward C Prescott "A
Competitive Theory of Fluctuations and the
Feaslblhty and Deslrablhty of Stablhzatlon
Pohcy," RatIOnal ExpectatIOns and EconomIc
Policy (ed Stanley Fischer) Chicago Umv of
Chicago Press, 1980, pp 169-98
(53) "Rules Rather than DiscretIOn
The InconSistency of Optimal Plans," Journal
of PolitIcal Economy, Vol 85, June 1977, pp
473-91
(44) "Recent Developments m the
Theory of InflatIOn and Unemployment,"
Journal of Monetary EconomIcs, Vol 2, Apr
1976, pp 185-219
(54) Lippman, Steven A , and John J McCall
"The Economics of Uncertainty Selected
TopICS and Probablhstlc Methods," H~ndbook
of MathematIcal EconomIcs, Vol 1 (ed
Kenneth J Arrow and MIchael D Intnhga­
tor) Amsterdam North-Holland Pubhshlng
Company, 1981.
(45) "Why Stopping Inflation May Be
Costly EVidence from Fourteen Historical
Episodes" Conference Paper No 108. Na­
tional Bureau of Economic Research, 1981
(55) Lucas, Robert E ,Jr "Econometnc Pohcy
EvaluatIOn A Critique," The PhIlips Curve
and Labor Markets, Vol I (ed Karl Brunner
and Allen Meltzer) CarnegIe-Rochester Con­
36
PrOposItIon," Journal of Monetary Economws,
Vol 4, Jan 1978, pp 121-30
ference Series, Supplement to Journal of
Monetary Econom.cs, 1976
(56) "Econometric TestIng of the
Natural Rate HypothesIs" In The Econome­
tncs of Pnce DeterminatIOn Conferencded
Otto EcksteIn) Board of Governors of,the
Federal Reserve System, WashIngton, DC,
1972
(57) "An EqUlhbrIum Model of the
BUSIness Cycle," Journal of Polttwal Economy,
Vol 84, Dec 1975, pp 1113-44
(58) "ExpectatIOns and the Neutrahty
of Money," Journal of Econom.c Theory, Apr
1972, pp 103-24
(59) "Methods and Problems In BUSI­
ness Cycle Theory ," J'!urnal of Money, CredIt
and Banking, Vol 12, No 4, Part 2, Nov
1980, pp 696-715
(60) "Some InternatIOnal EVIdence on
Output-InflatIOn Trade-qtTs," Amencan Eco­
nomIc Rev.ew, June 1973, pp 376-34
(61) "Tobm and MonetarIsm A
RevIew ArtIcle," Journal of EconomIC Lltera­
ture, Vol XIX, No 2, ,June 1981, pp 558-85
(62) , and Leonard Rappmg "Real
Wages, Employment, and InflatIon," Journal
of Polttlcal Economy, Vol 77, Sept IOct 1969,
pp 721-54
(63) Lucas, Robert E , and Thomas J Sargent
"Mter KeyneSIan MacroeconomIcs" After the
Phlltps Curve PersIStence of HIgh InflatIOn
and HIgh Unemployment The Federal
Reserve Bank of Boston Conference SerIes,
No 19, June 1978
(64) "IntroductIOn," In Ratwnal Ex­
pectatlOns and EconometrIC Practice, Vol 1
(ed Robert E Lucas, Jr and Thomas J
Sargent) Mmneapohs Uruv of M,nnesota
Press, 1981
(65) McCallum, Bennett T "Datmg, DIscountmg
and the Robustness of the Lucaa-Sargent
(66) McCallum, Bennett T "MacroeconomIcs after
a Decade of RatIOnal ExpectatIOns Some
CrItIcal Issues," Federal Reserve Bank of
R.chmond Econom.c Rev.ew, Dec l!i82, pp
3-12
(67) "Pr,ce Level Adjustments and
the RatIOnal ExpectatIons Approach to Macro­
economIc StabIlIzatIon PolIcy," Journal of
Money, Cred.t and Banking, Vol 10, Nov
1978, pp 418-36
(68) "RatIonal ExpectatIOns and
MacroeconomIc StabIlIzatIon PolIcy An Over­
vIew," Journal of Money, Cred.t and Bank­
lng, Vol XII, No 4, Part 2, Nov 1980,
pp 71646
(69) "The Role of OverlappIngGeneratIons Models m Mo_netary EconomIcs,"
WorkIng Paper No 989 NatIOnal Bureau of
Econonuc Research, Sept 1982
(70) MallIarIs, A G, and W A Brock Stochast.c
Methods In EconomIcs and Finance New
York North-Holland PublIshmg Company,
1982
(71) MInford,
A P L SubstItutIOn Effects, Specu­
latlOn and Exchange Rate Stablltty Amster­
dam North-Holland PublIshmg Company,
1978
(72) ModIgharu, Franco "The MonetarIst Con­
troversy or, Should We Forsake StabIlIzatIOn
Pohcles," Amertcan Economtc Reuz,ew,
Vol 67, Mar 1977, pp 1-19
(73) "Should Control Theory Be Used
for EconomIc StabIlIzatIOn? A Comment,"
Supplement to the Journal of Monetary Eco­
nomIc. (ed K Brunner and A H Meltzer)
Vol 7 CarnegIe-Rochester Conference SerIes
on PublIc PolIcy, 1977, pp 85-91
(74) Muth, John F "RatIonal ExpectatIOns and
the Theory of Pr,ce Movements," Econometn­
ca, Vol 29, July 1961, pp 315-35
37
(75) NeftcI, Sahh, and Thomas Sargent "A Little
Department Staff Report No 77 Federal Re­
Bit of EVidence on the Natural Rate Hypoth­
serve Bank of MInneapohs, Feb 1982
eSIs from the US," Journal of Monetary Eco­
(85) nom.cs, Vol 4, Apr 1978, pp 315 0 19
"A ClaSSical MacroeconometrIc
Model for the United States," Journal of
(76) Newberry, David M , and Joseph E StIghtz
Poh/.cal Economy, Vol 84, pp 207-37
The Theory of Commod.ty Pnce Stab.itzatwn­
"The Ends of Four BlgJnfla­
(86) A Study In the Econom.cs of RISk New York
tlOns," WorkIng Paper No 158 Research
Oxford UniV Press, 1981
Department, Federal Reserve Bank of MIn­
neapohs, May 1981
(77) Okun, Arthur Pnce and Quant./tes A Macro­
econom.c AnalySts WashIngton, DC The
"France's RestoratIOn of the Gold
(87) BrookIngs InstItutIOn, 1981
Standard In 1926," Wall Street Journal, Sept
1981
(78) Phelps, Edmund S (ed) Mterofoundatwns of
Employment and InflatIOn Theory New York­
(88) Macroeconom.c Theory New
Norton,,1970
York Academic-Press, 1979
(79) , and John B Taylor "StablhzIng
Powers'of Monetary Pohcy Under Rational
ExpectatIons," Journal of Polttteal Economy,
Vol 85, Feb 1977, pp 163-89.
(89) "The ObservatIOnal EqUIvalence
of Natural and Unnatural Rate TheorIes of
Macroeconomics," Journal of Poht.cal Econ­
omy, Vol 84, June 1976, pp 631-40
(80) Prentice, Paul T , and Lyle P Schertz Infla­
(90) , and Nell Wallace "RatIOnal
Iton A Food and Agncultural Perspecltve
ExpectatIOns, The OptImal Monetary Instru­
AER-463 U S Dept of Agr ,Econ Res
ment, and the Optimal Money Supply Role,"
Serv ,Feb 1981.
Journal of Polttteal Economy, Vol 83, Apr
1975, pp 241-54
(81) Prescott, Edward C "Should Control Theory
Be Used for Economic StablhzatlOn?" Supple­
(91) "Some Unpleasant MonetarIst
ment to the Journal of Monetary Econom.cs
ArithmetiC," Federal Reserve Bank of MIn­
Vol 7 Carnegte-Rochester Conference SerIes
neapohs Quarterly ReVieW, Fall 1981,
on Pubhc Pohcy (ed Karl Brunner and Allan
pp 1-17
Meltzer), '1977, pp 13-38
(92) Schultze, Charles E "Some Macro Founda­
(82) Resler, D H, J R .Barth, P A V B. Swamy,
tions for Micro Theory," Brookings Papers on
and W D DaVIS "DetectIng and EstimatIng
Econom.c ActIV.ty Vol 2 WashIngton, DC,
Changtng Ecpnomlc RelatIOnships The Case
1981
of Discount Wmdow Borrowmgs," SpeCial
Studles,Paper No 165 Board of Governors,
(93) Sims, Christopher A "Pohcy AnalYSIS With
Federal Reserve Board, WashIngton, DC,
EconometrIc Models" PresentatIOn to the
Aug 1982
BrookIngs Conference. on Economic ActiVity,
WashIngton, DC, Apr 1-2,1982
(83) Samuelson, Paul A "An Exact ConsumptlOn­
(94) ShefTrIn, Steven Ratwnal Expectatwns New
Loan Model of Interest With or Without the
York CambrIdge UniV Press, 1983
SOCial ContrIvance of Money," Journal of
Poltt.cal Economy," Vol 6, Dec 1958,
(95) ShIller, Ronald J "RatIOnal ExpectatIOns and
pp 467-82
the DynamiC Structure of Macroeconoml~
Models," Journal of Monetary Econom.cs,
(84) Sargent, Thomas J "Beyond Demand and
Vol ·2, Apr 1978, pp 1-44
Supply,Curves In MacroeconomICs," Research
38
(96) Small, DavId H "UnanticIpated Money Growth and Unemployment m the Uruted States' Comment," American Economic Review, Vol 69, Dec 1979, pp 996·1003 (97) Stevens, Nell A "Govl'rnment Debt Fmanc·
mg-Its Effects m VIews of Tax DIscount·
mg," Federal Reserve Bank of St LoUIS
ReVieW, July 1979, pp 11·19
phcatlOn to the U S Dau"y Economy" Paper
presented to American Agricultural Eco·
nomlcs ABBoclatlon Annual Meetmgs, Logan,
Utah, Aug 1982
(103) Tobm, James "Money and EconomIc
Growth," Econometrica, Vol 33, Oct 1965,
pp 671·84
(104)
"Government DefiCIts and
(98) Stokey, Nancy "On the Non·Robustness of CapItal AccumulatIOn," Chapter 1V m Asset
Barro's Result on SoCIal SecUrIty," The Accumulatwn and EconomiC ACtIVity
Center for Mathematical Stud... In Eco­
ChIcago Uruv of ChIcago Press,
nOmtCB and Management Scr.ence D1scusslon
1980
Paper No 426 Northwestern Uruv Evanston,
(105) Todd, RIchard M "Inventories m a Rational
m,1980
ExpectatIons Model of the Corn·Hog Sector,
and ImphcatlOns for AgrIcultural Subsector
(99) Swamy, P A VB, J R. Barth, and P A
Modehng" Unpubhshed manuscript Uruv of
T!nsley "The RatIonal Expectations Approach
Mmnesota and Federal Reserve Bank of Mm·
to EconomIc Modehng," Journal of EconomiC
DynamiCs and Control, Vol 4, May 1982,
neapohs, 1983
pp 12547
(106) Turner, Thomas H , and Charles H WhIte·
(100) Swamy, P A V B., and P. A T!nsley.
man "Econometric Pohcy Evaluation Under
"Lmear PredIction and EstimatIOn Methods
RatIonal ExpectatIOns," Federal Reserve
for RegressIOn Models WIth StatIOnary
Bank of MinneapolIS Quarterly ReVieW,
StochastIC CoeffiCIents," Journal of
Sprmg·Summer 1981, pp 6·15
Econometrics, Vol 12, Feb 1980, pp 10342.
(107) Turnosvsky, Stephen J Macroeconomic
(lOll Taylor, John "Staggered Wage Settmg m a
AnalYSIS and StablllZatwn PoliCY New York
Macro Model," American EconomiC ReVieW,
Cambndge Umv Press, 1977
Vol 69 Papers and Proceedmgs, May 1979,
(108) Tweeten, Luther G "MacroeconomIcs m
pp 108·18
CriSIS Agriculture m an Underachlevmg
(102) Thraen, Cameron S "Rational Expectations
Economy," American Journal of Agricultural
and AgrIcultural Pohcy An Econometric Ap.
EconomiCs, Vol 62, Dec 1980, pp 853·65
39
The Food and Agricultural Policy Simulator: An Addendum By Larry SaIathe, J. Michael Price, and Kenneth Gadson·
In a recent article, "On the MisUse of Theil's In­
equahty CoeffiCient," by Leuthold In thiS Journal
(3), the author Criticizes the use of a versIOn of
Theil's Inequality coefficlent'to vahdate the,perfor­
mance of an econometric model' Referring to the
artICle, "The Food and, AgrICultural Policy Simula­
tor," by Salathe, Price, and Gadson (4), Leuthold
suggests that "The'[Theil mequality coeffiCients]
U's reported by Salathe, Price, and Gadson are
probably much too low, reflecting greater accuracy
than IS really the case" Leuthold also raises a
number of Issues regarding the use of the Theil In­
equahty coeffiCient presented by Saiathe, Price, and
Gadson, and he Indicates they should report addi­
tional validation statisticsJor the Food and Agricul­
tural Policy Simulator (FAPSlM) The'purpose of
thiS note IS to address some of the Issues Leuthold
raIses and to present additIOnal vahdatlOn statistics
for FAPSIM
Three versIOns of the Theil Inequality coeffiCient _
are frequently used to evaluate econometric models
These statistics are defined as follows
T
"
T
T
'"
U = (l: (Y - Y )2)1121[( l: Y')1I2 + (l: Y')1I2]
t=1
t
L
t=1
t
transformations of the data as pOinted out by Theil
(6) In other words, If c denotes an arbitrary' con­
stant, then In general
T
...
T
(l:, (Y, - Y,)')1I2 1[( l:
t-l
T
(l:
t~l
T
t"'l
«Y, +
~
(l: (Y
t=l
t
~
~)112
T
+ (l:
t=-1
....
~)''']
"*
T
c) - (Y, + c)'»112 !'[( l: (Y, + cl'Jli2 +
t=1
+ c)'
)1"),
Leuthold suggests that th,S property IS a mllJor
cause for concern, stating that, "If one moves the
decimal pOint to the nght for each variable, one can
generate lower [Theil U] coeffiCIents" (3) However,
the type of transformatIOn deSCribed by Leuthold IS
not an addltwe transformation, but a multlpllcatwe
transformatIOn of the data The Theil U statistic IS
not senSItive to multlphcate transformatIOns
because
(1)
t=l!
T
....
T
T
"
(l: (cY,-cY.l'Jli2'1[(l: (cYJ')l"+(l: (cY,)'Jli2]
t-l
1-1
t~l
for any nonzero constant c ThiS means that the
umts used to measure Y, will not affect the value'of
the Theil U statistic Therefore, the results pre­
sented In (4) do not reflect a "umt problem" as
Leuthold suggests
where Y, IS the actual value of the Variable at time
t, Y, IS the predicted value at time t, and T IS the
number of time perIOds used for validation The ver­
sIOn used by Salathe, Price, and Gadson was the
Theil U statistic
Leuthold Criticizes the use of thiS statistic for two
reasons First, thiS statistic IS sensitive to additive
·The authors are agricultural economISts With the National
Economics DIVISion, ERS
'Itallclzed numbers 10 parentheses refer to Items lD the
References at the end of thIS note
40
One also has to wonder why It IS necessarily inap­
propriate to use a validation statistic which IS not
invariant With respect to an additive transforma­
tion, especially when thiS statistic IS used to
measure the predictive power of a model In terms of
a pnce or a quantity series For example, suppose
we have a data series [Y,I for the price of some com­
modity, and suppose 'c' IS an arbitrary constant Is
there some meanmgful mterpretatlOn that can be
IKost (1) cl81ms that an addItive transformation of thiS type
WIll always reduce the value of U However, trus 18 not necessari­
ly the case
AGRICULTURAL ECONOMICS RESEARCHNOL 35, NO 3, JULY 1983
attached to the new senes [Y, + cl, and, If not, why
would anyone be mterested m determmmg how
well the model predicts this variable?
A second cnticlsm of the Theil U statistic which
Leuthold raises IS that this statistic IS bounded
from above by 1 Leuthold seems to suggest that
this property somehow leads to "low [Theil] U coeffi·
clents," thereby overstatmg the predictive power of
a simulatIOn model As Theil showed, both, the
Theil U and U I statistics are constramed to he be·
tween zero and 1 Both,statlstics will attam a value
of 1 only If the predicted values represent the worst
poSSible forecasts of the actual values (Il' the sense
defined by Theil) Both t1"e Theil U and U I statistics
Will equal 0 only If the model prediCts the historical
values perfectly over the entire validatIOn period
Thus; both Theil's U and UI statistics are constramed
to lie WIthm the umt mterval, and the end pomts of
the mterval' may be mterpreted as correspondmg to
opposlte:extremes m model forecast accuracy The
fact that Theil's U statistic IS bounded from above
by 1 IS a 'useful property because the value of 1 pro·
vldes a benchmark representmg the worst pOSSible
model forecasts This property m no way causes thiS
statistic to overstate the predictive ability of a
simulatIOn model
Leuthold also criticizes the use of Theil's U statistic
because "A U m actual values Will always be less
than a U when data are measured as changes"
This statement Implies that U Will always be
less than U I and U, However, usmg the example
below, we can show that U IS not necessarily less
than U I or U, Suppose we have the follOWIng data
series for Y, and 'It
For all the data series mcluded m the model valida·
tlOn of FAPSIM, however, the Theil U statistics are
less than the~r correspondmg Theil UI and U2
statistics (see followmg table) But the Theil Uland
U, statistics do not appear unreasonable, thereby
confirmmg the earlier model validatIOn results
reported m (4) The Theil U I and U, statistICs do
result m changes m relative results among commo·
dlties as compared With the Theil U statistic, as
Leuthold suggests This result IS not surprlsmg
because of the differences m the properties of the
Theil U, U" and U 2 statistics
As Kost pomts out, each of Theil's mequahty coeffi·
clents has dIfferent properties and mtefprEitatlOns
(2) This means that each coeffiCient prOVides dlf·
ferent mformatlOn about the forecast accuracy of a
model The same IS true of other statistics that have'
been proposed for model validation (1) No smgle
measure prOVides a complete ,summary ofla model's
forecast ability as each has speCific limitatIOns For
thiS reason the mean absolute relative error, the
turnmg pomt error, and the plots of actual versus
forecasted values were presented m conjunction
WIth the Theil U statistic m the vahdation of
F APSIM (4) Leuthold appears to be overly concerned
With the properties of the Thed U StatiStiC, fruling
to recognize the shortcommgs of other validatIOn
statistICs (such as Theil's U, and U2 statistics) He
also seems to have overlooked the additional mfor·
mation beyond the Theil U statistic presented m
the validation of F APSIM
References
(1) Kost, Wilham E "Model ValidatIOn and the Net
Trade Model," AgrIcultural EconomICS Research,
Vol 32, No 2, Apr 1980, pp 1-16
(2) "Comment on the MISuse of Thed's
inequality CoeffiCient," AgrICultural EconomICs
Research, Vol 34, No 4, Oct 1982, p 40
Time perIOd
1
Variable
o
Y,
0000
Y
t
- :: Not
+11 +2
0000
000
0081
000
I
+3
0052
000
I
+4
0001
000
8pphc8bl~
then U = 1, U I = 049045, and U, = 096246 •
Thus, there IS no reason to expect that the Theil U
statistic WIll always be less than either U I or U,
'Although the values chosen for Yt and Yt In thiS example may
appear artificial, they were, In fact, obtsmea from a vahdatlOn
study of a darry sector model (5)
(3) Leuthold, Raymond M "On the MISuse of
Theil's inequality CoeffiCient," AgrIcultural Ecr;.
nomles Research, Vol 34, No 4, Oct 1982, p 39
(4) Salathe, Larry E ,J Michael Price, and Kenneth
E Gadson "The Food, and Agricultural Pohcy
Simulator," AgrIcultural EconomICS Research,
Vol 34, No 2, Apr 1982, pp 1-15
41
I
/
(5) "The Food and Agncultural Pohey
Simulator The Drury-Sector Submodel,;' AgrICUl­
tural EconomIcs Research, Vol 34,
3, July
1982, pp 1-14
No
(6) Theil, Henn EconomIc Forecasts and Pol&ey
Amsterdam North-Holland Pubhshlng Co ,
1961
Theil inequality coefficients, for the Food and Agricultural Policy Simulator
Theil U
VahdatlOn statistic
Theil U ,
I
I
Theil U,
0'024
019
014
021
013
006
0282
347
243
300
552
247
0562
714
495
560
1157
516
048
042
058
036
272
267
257
275
512
601
587
562
Farm price of turkeys
Farm price of,eggs
Farm price of milk
046
052
025
245
353
277
490
673
591
Acreage
Acreage
Acreage
Acreage
Acreage
Acreage
Acreage
030
011
037
085
035
042
047
276
177
350
542
412
449
252
546
389
728
1770
845
943
499
062
063
083
082
036
060
062
230
300
320
421
184
258
302
456
637
804
853
366
595
679
016
010
077
019
,003
126
087
279
205
036
259
181
584
490
073
Variable ,descriptIOn
Pork production
Beef productIOn
BrOIler prpductlOn
Turkey productIOn
Egg productIOn
Milk production
Price
Price
fuce
Price
Farm
Farm
Farm
Farm
Farm
Farm
Farm
of barrows and gllte
of slaughter steers
of utilIty cows
of brOIlers
planted
planted
planted
planted
planted
planted
planted
of wheat
of corn
of barley
of sorghum
of oats
of soybeans
of cotton
price of wheat
price of corn
prICe of barley
price ,of sorghum
price of oats
price of soybeans
price of cotton
Total cash receipts from farm marketings
Total farm productIOn expenses
Net farm Income
Consumer price Index, all food
Consumer price Index, all Items
42