Professional Documents
Culture Documents
Markup Pricing Versus Marginalism PDF
Markup Pricing Versus Marginalism PDF
.
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].
M.E. Sharpe, Inc. is collaborating with JSTOR to digitize, preserve and extend access to Journal of Post
Keynesian Economics.
http://www.jstor.org
CATHERINE LANGLOIS
Introduction
127
128
of theallowancemadeforprofit.Thus,firmsweredescribedas applying
some"standard"markupoveraveragetotalcosts,a factthatraisedthe
issue of compatibility
betweenbusinesspracticeandthe predictionsof
marginalist
pricetheory.
In defenseof marginalist
principleswas the argumentthatrealworld
practicedid not implyrejectionof the profitmaximizingmodelif the
ThusRobinson(1950) arguesthat,in
facts werecorrectlyinterpreted.
orderto decideon the bestuse of its resources,a producercan makea
doesnotmakeexplicitthedynamicprocessof adjustment
of production
andpricethatmusttakeplaceto maintaininventoryat its targetlevel.
Thehypothesispresentedandempiricallytestedin whatfollowsis that
thelevel of targetinventoryis chosentogetherwithits sellingpriceso
thatprofitis maximizedoverthetimeit will taketo sell the inventory.
The firm achievesthis goal by maximizingprofitper unit of time,
optimizationwhichgeneratesa formulafor optimalmarkupover the
averagecost of goodsin inventory,whereaveragecost is the average
the
productioncost of the goods themselves.Thus,by transforming
130
the terms outlinedhere, then markupover the cost of goods sold should
coincide with the inverse of the price elasticity of the sales time of
inventory. This elasticity has been measured for General Motors,
Chrysler,and Ford, and its inverse does indeed coincide with markup
over the cost of goods sold as recorded by company accounts (see
Langlois, 1989). It is one of the objectives of this paperto provide yet
furtherevidence of the compatibilityof this approachwith the data.
The marginalist debate illustrates the marginalist school's lack of
emphasis on verifiable quantitativeprediction.Empiricalmeasures of
price elasticity of demand were available in the early 1960s, yet the
proponentsof themarginalistcontroversydidnot appealto themto refute
or confirmthe neoclassical approach.Yet, as will now be arguedusing
automobileindustrydata,the analysis of these elasticity measuresprovides interesting insights. The demand for automobiles has been the
subjectof numerouspost-warstudies,andvariousauthorshaveestimated
price elasticities for the shortandlong runs(Chow, 1960, is the standard
referencefor long-runestimates).As shown in Table 1, short-runprice
elasticity of demandfor automobilesin the U.S. has been estimatedto
lie in the range (-3.1, -1.32). In Machlup's (1946) interpretationof the
facts, the inverseof thepriceelasticityof demand1/IqlIshouldyield profit
maximizingmarkupover marginalcost.
FromTable 1, markupis then estimatedto lie between 32 percentand
76 percent,a promisingoutcome for marginalisttheory,given estimates
of marginalcost in the automobileindustry.White states that marginal
costs formanufacturinglie betweenone-halfandtwo-thirdsof wholesale
prices(White, 1971, p. 120).This estimateis corroboratedby Evans,who
132
JOURNALOF POSTKEYNESIAN
ECONOMICS
Table 1
Period
Estimate
Evans
1948-1964
-3.1
Westin
1. Stock AdjustmentModel
2. DiscretionaryReplacementModel
Weiserbs
1. DynamicLinearModel
2. DynamicQuadraticModel
1953-1972
-1.32
-1.65
1929-1970
-2.13*
-2.32*
* Uncompensateddemand elasticities.
Compensateddemand elasticitiesare also reportedin Phlips.
Sources: Evans, M., MacroeconomicActivity.Theory,Forecastingand Control.An EconometricApproach.Harperand Row, 1969.
Westin,R., "Empirical
Implicationsof InfrequentPurchase Behaviorin a
Stock AdjustmentModel,"AmericanEconomicReview,June 1975, p. 393.
Weiserbs, estimates reproducedin Phlips,L.,AppliedConsumptionAnalysis,
pp. 201 and 207.
onlyone-thirdof wholesaleprice,
percent).Marginalcost,representing
fallsshortof averagecost.Thesecostdatarelateto costsincurredgiven
industryoutput.Thefactthatmarginalcost is less thanaveragecost at
that outputlevel suggeststhat,over the whole rangeof outputs,the
downwardslopingaverage
marginalcostcurvelies belowapresumably
Studiesof automobileretailing
cost curve,at least for manufacturing.
suggesta similarpatternfor costs (see Pashigian,1961;Davissonand
Taggart,1974).Thus,thelow valuesfoundfor ITllarenotincompatible
with the industry'scost structure.However,in a worldof declining
averagecosts over the whole rangeof outputs,marginalrevenueand
marginalcost mayneverintersectto yield an optimalmarkupequalto
1 / ITl.The cost controversythatdevelopedalongsidethe marginalist
debatewas concernedwithpreciselythispoint,andit is to a summary
of this controversyand an analysis of its contributionthatwe now turn.
II. The cost controversy and the commensurability
of costs and demand
The debateoverthe shapeof the short-runcosts facedby firmsbeganwith
Eiteman's(1947) discussionof costs.Thecentralquestionwas thepractical
relevanceof the conceptof marginalcost. As Lee (1984) reports,Eiteman
"had never heard any price settereven mentionmarginalcosts." And,
accordingto Eiteman,this was attributableto the difficultiesinvolved in
measuringmarginalcost as soon as a firm used multipleprocesses. But
even if marginalcosts weremeasurable,the shapeof the cost curvesfaced
by a typicalfirmmay denymarginalcosts anypracticalrelevancein terms
of pricingstrategy.Marginalistprinciples,accordingto Eiteman,can only
be implementedif minimumaveragecost obtainsat outputswell below
capacityto guaranteean intersectionbetween marginalrevenueand the
upwardslopingportionof marginalcost.
Eiteman and Guthrie(1952) questioned 366 firms and found 316 to
believe averagecosts to decline with increasedoutput,reachinga minimum at or close to full capacity,with averageandmarginalcost curves
simply stopping at full capacity. The implicationis that marginalcost
lies below averagecost at all levels of operationand "cannot intersect
the marginalrevenuecurve(1) if the averagerevenuecurveis horizontal
or (2) if the average revenue curve is high and relatively elastic"
(EitemanandGuthrie,1952, p. 832). Ritter(1953) pointedout atthe time
thatthis did not invalidatemarginalismas a doctrineprescribingrational
134
doesnotrequireequalityof marginal
behavior,sinceprofitmaximization
revenueandmarginalcostif adiscontinuity
preventsit.Moretothepoint,
EitemanandGuthrie'sfindinglimitsthesetof quantitylevelsoverwhich
costsaredefinedandin mostinstancesimposesa comersolutionto the
neoclassicalprofitmaximization
problem.But,if profitmaximizingdoes
notoccurattheintersection
of marginalrevenueandmarginalcost,the
optimalmarkupovermarginalcost is no longerequalto the inverseof
thepriceelasticityof demand.Andthis sameconclusionholdsif firms
areassumedto makeintertemporal
tradeoffsto, for example,preserve
marketshare,since profitmaximization,thus constrained,no longer
impliesequalityof marginalcostandmarginalrevenue.
Thus,the failureof realworldfirmbehaviorto conformto the marruledoesnotnecessarilymeanthat
ginal-cost-equals-marginal-revenue
profitis notmaximized.Theimplicationis ratherthattheinverseof the
priceelasticityof demandneednotbeaproxyformarkupovermarginal
cost.Theissueis animportant
one,sincenumerousauthorsbaseempirical workon theassumption
that1 /1,9I
doesindeedcapturemarkupover
marginalcost (see,forexample,Sumner,1981,orRosse, 1970).
Withreferenceto automobile
industrydata,thefactthatmarginalcosts
lie belowaveragecostscouldexplainthelow valuesfoundfor Irl.But
forthesevaluesto confirmthehypothesisthatfirmssetmarginalrevenue
equalto marginalcost,it mustbe arguedthatthe schedulesintersect.In
whatfollows,it is shownthat,as longas demandforthefirm'soutputis
not horizontal,an intersectionbetweenmarginalcost and marginal
revenuealwaysexists if the temporalcommensurability
of costs and
demandis bypassed.Moreover-andmoreimportantly-itturnsoutthat
thisdissociationsuggestsamodelof firmbehaviorthat,whenconfronted
withthe data,canbe reconciledwiththe empiricalevidenceon 11 for
the automobile
industry.
Costs definedin the standardmanneras productioncosts have a
temporaldimension.Itis thecostof daily,weekly,ormonthlyproduction
thatis described.The demandandmarginalrevenuecurvesassociated
to suchcostsmust,of course,havethesametemporaldimension.To the
cost of dailyproduction
is associateddailydemandforoutput.It is this
of costs and demandthatprenecessarytemporalcommensurability
cludesanyguarantee
of existenceof a pointof contactbetweenmarginal
costandmarginalrevenuein thecostconfiguration
revealedby Eiteman
andGuthrie(1952).
betweencostsanddemandis a constraint
Temporalcommensurability
as a level of inventory.
producedperunitof time,it canbe interpreted
An inventorylevel is just datedandcanthereforebe associatedto demand
relationshipswith varying temporaldimensions. A stock of outputcan
be sold within a week or a month,dependingon the price chargedfor it.
In other words, an arbitraryprice-quantitypair can always be fitted on
a demandschedule if the temporaldimension of demandis allowed to
vary. If marginalproductioncosts lie below average costs, the pattern
can be reproducedin the cost of goods in inventory.Thus marginalcost
associatedto variouslevels of inventorymay lie below averagecost. But
the possibilityof choosingthe temporaldimensionof demandguarantees
existenceof anintersectionbetweenmarginalcost andmarginalrevenue.
The possibility of choice raises the questionof the significance of the
particulartime frame chosen. In the model developed below, the firm
chooses thattime frameforthe sale of its inventorythatmaximizesprofit
made over the period.Thus the temporaldimensionchosen for demand
is generatedendogenously and is non-arbitrary.To clarify the above
ideas and introducethe model (which will be developed in section III
below), a formalpresentationof demandis necessary.
Inthe standardmonopolymodel,demandis approachedas a one-to-one
relationshiplinking selling price to quantitydemanded.According to
Debreu(1959), who providesa precisedefinition,thetime intervalwithin
which demand will actually be realized is a "compact elementary
interval" whose length "is chosen small enough for all the instantsof
an elementaryintervalto be indistinguishablefromthe point of view of
the analysis" (Debreu, 1959, p. 29). Thus, in defining the concept of
demand,for example, a subdivisionof the elementaryintervalwill not
revealvariationsin quantitydemandedthataremeaningfulfromthepoint
of view of the decisionmaker'sstrategy;day-to-dayvariationsin demand
for a productmightbe meaningful,while the fluctuationsobservedfrom
one minuteto the next aresurelynot.The elementaryintervalis therefore
chosen, in what follows, as the unit within which flow demand is
measured.Following Debreu'sterminology,the elementaryintervalwill
be referredto as a date. Demandat date t is referredto as flow demand
q. At pricep, flow demandof date t can be written
q= q (p, )
Within any date T, flow demandis a one-to-one relationshiplinking
price to quantity.
136
ECONOMICS
JOURNALOF POSTKEYNESIAN
q(p@,)dx
+
s-
) p+18 +sin
q(T)d= -
-2t - c- s+
p+18t-cost+1
( Q, , 7) 0, then thereexists a
neighborhood
U, q,, i) such thatp
p(q, t)
Similarlyfor the othervariables.
t) in U. Similarly
neighborhood
U(p,
thatp = p(q,
*
*
138
ECONOMICS
JOURNALOF POSTKEYNESIAN
(ii)
I (p' ,Pq
t) - c (q(p, t))
X (p, q) = p q- c (q)
t (p, q)
p, q - Ct (q)
rt,(p,C)
MARKUPPRICING VERSUSMARGINALSM
139
(1)
1+ 1/n (p, t)
(2)
P
l-l/(p,
q)
recordedin Moody'sIndustrial
Manuals.
In sectionIV, furtherempiricalevidenceof the compatibilityof the
abovemodelwiththedatais presented.Theliteratureprovideseconometricestimatesof r forthe automobileindustry.Dataon markupover
Manuals.
averagecost of goodssoldis recordedin Moody'sIndustrial
The theoreticallink betweenelasticitiesrl ande providesthe tool for
empiricaltestingof themodel.
LinkingT1ande involvesdefinitionof a thirdconceptof elasticity:r =
(3 q / t) x (t/q). {, defined holding price constant,measures the percentage change in quantitydemandedand sold for a 1 percentchange in
the time span allowed for demandto be realized.As shown in Appendix
140
or
=-
Iril
MARKUPPRICINGVERSUSMARGINALISM141
Table 2
Per-unit-timesales patterns:threenumericalexamples
A:declining
per-unit-timesales
B: constant
sales
per-unit-time
C: increasing
per-unit-timesales
1
2
3
10
9
8
10
19
27
1
2
3
10
10
10
10
20
30
1
2
3
10
11
12
10
21
33
whilethesecondpresentstheresults,andthethirdconfrontsthemwith
the availableevidenceon markupoverdirectcostsandrl.
Methodology
C is an elasticity concept associatedto the integral of flow demand. In
t), where
otherterms, = q
q (p, t) = o q(p, ) d. Thus, Cwill
at
q
be descriptiveof the progressionof cumulativesales over time, not of
the change over time of per-unit-timesales. To fix the ideas, consider
the numericalexamples in Table 2.
To each date t is associatedsales at that date, q, and cumulativesales
since date 1, q. The elasticities are computedbetween dates 2 and 3. If
per-unit-timesales areconstant(exampleB), the elasticityof the change,
over time, in per-unit-time sales (A q /A t) x t/ q = 0, since this elasticity
Aq t
10
5/2
At
50/2
If per-unitsalesdecline,-.2941growthfromdate2 to 3 in exampleA,
r dropsbelow 1 to 0.8696, while for sales growingat + .2174 as in
exampleC, r risesabove1to 1.1111.Thusbarringdramaticfluctuations
142
Table 3
Estimatesof C:Methodology
Sample: 1953.3-1973.5
MultiplicativeModel: DependentVariableLNSALE
IndependentVariables
1
C
LNTRND
LNPRIX
LNGPY
Coefficient
12.730330
0.0183887 -2.0451384 0.6263535
(T)
(6.9467845) (0.3694518) (-4.6618805) (6.3647833)
Adjusted R2= 0.598306
D.W. = 1.600705
MultiplicativeModel: DependentVariableLNQNT
IndependentVariables
2
C
LNTRND
LNAPRC
Coefficient
(T)
LNCGPY
AdjustedR2 = 0.999878
D.W.= 1.850189
= 0.5185682
Sample 1953.3-1972.4
MultiplicativeModel: DependentVariableLNQNT
IndependentVariables
3
C
LNTRND
LNAPRC
A
= 0.4899031 -2.3714331
Coefficient
15.272142
(T)
(6.82194796) (2.3536793) (-49388468)
LNCGPY
0.5562831
(3.8418860)
AdjustedR2 = 0.999868
D.W.= 1853300
5 = 0.4899031
to mayturnoutto be
elasticity,theregressioncoefficientit corresponds
modestly positive, but the confidence intervalsurroundingit is likely to
contain zero. In regression 1 of Table 3, the logarithm of quarterly
automobile sales (LNSALE) is regressed against the logarithm of the
time trend (LNTRND), the logarithm of the implicit new auto price
deflator(LNPRIX),andthelogarithmof personalincome (LNGPY).The
144
Ps
'u,
variable.A timetrend(TRND),theaverage
units,QNT,is thedependent
new auto implicitprice deflatorseries (APRC),and the cumulative
The
variables.4
currentdollarnationalincome(GPY)aretheindependent
coefficients
and
5.
All
modelis usedforregressions,1, 3,
multiplicative
aresignificantandhavethe expectedsign. An averagepriceincrease
decreasescumulativesales while an increasein incomeor in the time
allowedforsalesto be madeincreasessales.Theadditivemodelis used
for regressions,2, 4, and6. Again,all coefficientsaresignificantand
havethe expectedsign. All regressionshavebeen correctedfor serial
correlation
AdjustedR2 is of the
procedure.
usingthe Cochrane-Orcutt
orderof 99.9 percentfor all regressions.
Test of the behavioralhypothesis
If firmschoose inventorylevels andpriceso as to maximizeper-unit-time
profit,then at the optimalpoint the following relationshipholds:
or
1
m=m+
where m is markupover average direct costs. Thus, given a range of
values for , computationof the ratio-C/m will yield an intervalwithin
which Tris expected to lie.
146
Table 4
Estimatesof 5
Sample 1953.3-1973.4
MultiplicativeModel: DependentVariableLNQNT
IndependentVariables
1
C
LNTRND
LNAPRC
LNCGY
Coefficient
(T)
14.029798
= 0.3691767 -2.1531173 0.6385347
(6.1594962) (2.0023913) (-4.3748699) (5.1321090)
A
Adjusted R2 = 0.999888
D.W. = 1.897562
5 = 0.3691767
2
Coefficient
(T)
IndependentVariables
TRND
APRC
LNCGY
44532.420
1422.4177
-667.78975 3.8673892
(1.7941245) (4.0623119) (-4.4233425) (3.0547783)
A
Adjusted R2 = 0.999964
D.W. = 2.066540
, = 0.8789
Sample: 1948.4-1969.4
MultiplicativeModel: DependentVariableLNQNT
IndependentVariables
3
LNTRND
C
LNAPRC
LNCGY
Coefficient
(T)
21.570651
(7.133456)
Adjusted R2 = 0.999874
D.W. = 2.115463
r = 1.1056435
IndependentVariables
TRND
APRC
CGY
90346.118
1053.7905
-1203.1536 6.466371
(5.3364858) (6.4495549) (-5.0331825) (6.2957888)
D..
D.W. == 2.081139
2.081139
0.7438
= = 0.7438
Table 4 (continued)
Sample: 1948.4-1973.4
MultiplicativeModel: DependentVariableLNQNT
IndependentVariables
5
C
LNTRND
LNAPRC
Coefficient
(T)
9.4495282
(5.1430586
LNCGY
Adjusted R2 = 0.999811
D.W. = 2.057361
{ = 0.7657064
IndependentVariables
TRND
APRC
CGY
1055.7221
-351.69931 5.3452385
24667.608
(1.9598794) 10.434661) (-2.2546144) (8.5465201)
A
Adjusted R2 = 0.999970
D.W. = 2.178076
{ = 0.7151
estimatesof Cthanthemultiplicative
model,whilethereverseoccursfor
the sample1953 to 1974. As shownin Table5, the rangeof values
generatedfor11includesEvans'(1969)estimateforthesample1948to
1969.The 1948to 1973sampleprovidesa narrowrangefor l, whichis
close to the values foundby Weiserbs(in Phlips, 1983). Weiserbs,
however,usespre-waras well as post-wardatato generatehis demand
elasticities.Oursample,therefore,doesnotoverlaphis.
in Table5, of estimatedpriceelasticitiesof
The fit, as summarized
demandto the rangeof values for iT generatedindependentlyusing
estimatesof C and values for m, providesevidencethat the model
withthedata.
presentedin thispaperis notincompatible
148
JOURNALOF POSTKEYNESIAN
ECONOMICS
Table 5
Summaryof results
Regression Results
RelevantComparisonStudy
Sample 1953.3-1973.4
Range for {: [.369, .879]
Markupm: .296
Range foriT:[-2.97, -1.25]
Author:Westin(1975)
Sample: 1953-1972
Price elasticityestimates: -1.65, -1.32
Sample 1948.4-1969.4
Range for r: [.744, 1.106]
Markupm; .302
Range foril: [-3.66, -2.46]
Author:Evans (1969)
Sample: 1948-1964
Price elasticityestimate:-3.1
Sample 1948.4-1973.4
Range for {: [.715, .766]
Markupm: .296
Range forrT:[-2.59, -2.42]
Conclusion
This paperreviewedthe marginalistandcost controversiesin light of the
post-warstatisticalestimatesof short-run
priceelasticityof demandfor
automobiles.
Theserangefrom-1.32to -3.1.Dataonautomobile
industry
costssuggestthatmarginalcostslie belowaveragecosts.Sincemarkup
over averagecost at the retaillevel is on the orderof 30 percent,this
explainsthelow valuesfoundforir. Buttheissueis thevalidityof the
of 1 / Irllas markupovermarginalcost.Thiswouldimply
interpretation
that firms indeedchoose price so as to equatemarginalrevenueto
marginalcost,butin thecostconfiguration
typicalof the autoindustry,
theseschedulesmaynot evenintersect.Thus,in the absenceof precise
dataon marginalcostsattheretaillevel, confrontation
of themarginalrule
to
the
data
on
would
haveto remain
11
cost-equals-marginal-revenue
inconclusive.This papersuggestsa modificationof the standardfirm
model,whichprovidesadditionaltools to analyze
profitmaximization
thedata.If firmschooseinventoryandpriceso as to maximizeperunit
time profit, formulaefor optimalmarkupsover averageas well as
marginalcostsaregenerated,andtheirratiois predictedto match(, the
percentagechangein cumulativesales thatresultsfrom a 1 percent
Proof:
Let
q+
d=d.P=[.E=
dp
a3p at
. dt 1. P
dp
and
A
dt .
dp
at d+at
t=
ap aq dp_
By definitionof l, , and e
A
(1)
+A
=rA
and
A
150
Eiteman,W. J., and Guthrie,G. E. "The Shapeof the Average Cost Curve." American EconomicReview, December 1952,42.
Evans, M. MacroeconomicActivity:Theory,Forecasting and Control.New York:
Harperand Row, 1969.
Feldstein,M., andAuerbach,A. "InventoryBehaviorin DurableGoods Manufacturing: The TargetAdjustmentModel." BrookingsPapers on EconomicActivity,1976, 2.
Gordon,R. A. "ShortPeriodPriceDeterminationin Theory and Practice."American
EconomicReview, June 1948, 38.
Hall, R. L., and Hitch, C. J. "PriceTheory andBusiness Behavior." OxfordEconomicPapers, May 1939, 2.
Langlois, C. "A Model of TargetInventoryandMarkupwith EmpiricalTesting
Using AutomobileIndustryData." Journal of EconomicBehavior and Organization,
1989, 11, 47-74.
Lee, F. "The MarginalistControversyand the Demise of Full Cost Pricing." Journal
of EconomicIssues, December 1984, 28 (4).
Maccini, L., and Rossana,R. "Investmentin FinishedGoods Inventories:An Analysis of AdjustmentSpeeds." AEA Papersand Proceedings,May 1981.
Machlup,F. "MarginalAnalysis and EmpiricalResearch."AmericanEconomicReview. September1946,36 (4).
Pashigian,B. TheDistributionofAutomobiles,an EconometricAnalysis of the Franchise System.Englewood Cliffs, NJ: PrenticeHall, 1961.
Phlips, L. AppliedConsumptionAnalysis. Amsterdam:North Holland, 1983.
Ritter,L. "The Shape of the Average Cost Curve:Comment."AmericanEconomic
Review, September1953,42 (4).
Robinson,A. E. "The Pricingof ManufacturedProducts."The EconomicJournal,
December 1950, 60 (240).
Rosse, J. "EstimatingCost FunctionParameterswithoutUsing Cost Data."
Econometrica,March 1970, 38.
Sumner,D. "Measurementof Monopoly Behavior:An Applicationto the Cigarette
Industry."Journal of Political Economy, 1981, 89 (5).
Sylos-Labini,P. Oligopoly and TechnicalProgress. Cambridge,MA: HarvardUniversity Press, 1962.
Westin, R. "EmpiricalImplicationsof InfrequentPurchaseBehaviorin a Stock AdjustmentModel." AmericanEconomicReview, June 1975.
White, L. TheAutomobileIndustrysince 1945. Cambridge,MA: HarvardUniversity
Press, 1971.