Union Effects on Productivity, Profits, and Growth: Has the Long Run

NORC at the University of Chicago
Union Effects on Productivity, Profits, and Growth: Has the Long Run Arrived?
Author(s): John T. Addison and Barry T. Hirsch
Reviewed work(s):
Source: Journal of Labor Economics, Vol. 7, No. 1 (Jan., 1989), pp. 72-105
Published by: The University of Chicago Press on behalf of the Society of Labor Economists and the
NORC at the University of Chicago
Stable URL: http://www.jstor.org/stable/2534990 .
Accessed: 19/04/2012 18:30
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
The University of Chicago Press, Society of Labor Economists, NORC at the University of Chicago are
collaborating with JSTOR to digitize, preserve and extend access to Journal of Labor Economics.
http://www.jstor.org
Union Effectson Productivity,
Profits,and Growth:Has the
Long Run Arrived?
JohnT. Addison,UniversityofSouthCarolinaand
Universitit
Bamberg
BarryT. Hirsch,University
ofNorthCarolinaat Greensboro
This articleinterprets
literature
on economic
examiningunioneffects
performance.
Productionfunctionstudiesindicatesmalloverallunion
impactson productivity;
positiveeffects,
wheretheyexist,appearto
resultfrommanagementresponseto decreasedprofitexpectations
and froma naturalselectionprocess. Lower profitability
among
unionizedfirmsis well established;moreinteresting
is thepossibility
thatunionsappropriatequasi rentsderivingfromlong-livedtangible
and intangiblecapital.The connectionbetweenunions,investment
behavior,and productivity
growthemergesas a particularly
fruitful
line of empiricalinquiry,althoughit does not encouragea sanguine
view of unionism'slong-runimpact.
I. Introduction
Controversycontinuesto surroundthe natureand directionof union
effects
on economicperformance.
Much of theimpetusbehindthisdebate
stemsfromtheworkof RichardFreemanand JamesMedoff(1984), who
pointto potentialincreasesin productivity
resultingfromunion-induced
changesin theworkplace.Such increasesareeffected
throughtheexercise
We havebenefited
fromthecomments
of seminar
and readersof
participants
earlier
ofthepaper.
versions
[JournalofLaborEconomics,1989,vol. 7, no. 1]
?) 1989by The University
of Chicago. All rightsreserved.
.50
0734-306X/89/0701-0005$01
72
UnionEffects
73
of "collectivevoice," coupled withan appropriateinstitutional
response
frommanagement.
Unions,itis argued,lowerturnover
and establishmore
inworkplacescharacterized
efficient
structures
governance
bypublicgoods,
in production,and long-term
complementarities
contractual
relations.Althoughrecentliteraturecontinuesas beforeto investigateproductivity
is beinggivento an examinationofunion
differentials,
increasingattention
on profitability,
and growth.
effects
investment,
This articleinterprets
recentliterature
on ecoexaminingunion effects
nomicperformance.
We firstaddresstheambiguitysurrounding
theproductionfunctiontestitself,priorto offering
an interpretation
of observed
productivity
differences.
The newerliterature
examiningunioneffects
on
profitability
is nextanalyzed.Issuesyetto be resolvedin thisarea include
the magnitudeof the union profiteffect,
the consistencyof union productivityand profitability
estimates,and the sourcesfromwhichunions
captureprofits.
Finally,we turnto whatmaybe themostinteresting
aspect
of the literature;namely,the dynamiceffectof unions on productivity
growth,investment
activity,and long-runperformance.
It is concluded
thatthedebateoverwhatunionsdo will focusincreasingly
on thelongertermconsequencesof unionrentseeking.
and the ProductionFunctionTest
II. Productivity
MeasurementIssues
The majority
oftheunionproductivity
studiesfollowBrownand Medoff
somevariantoftheCobb-Douglasproductionfunction
(1978) in employing
Q = AK'(L, + cL )l-a,
(1)
whereQ is output,K is capital,Lamand L, are union and nonunionlabor,
A is a constantof proportionality,
and a and (1 - a) are the
respectively,
withrespectto capitaland labor.The parameter
c reflects
outputelasticities
differences
betweenunionand nonunionlabor.Ifc > 1,then
productivity
in line withthecollectivevoice model;if
unionlaboris moreproductive,
c < 1, thenunion labor is less productive,in line with conventionalargumentsconcerningthedeleteriousimpactof such thingsas unionwork
rules and constraintson merit-basedwage dispersion.Manipulationof
equation(1) yieldstheestimating
equation(we ignoretheerrorterm)
In (QIL)
- In A + a In
(KIL)
+ (1 -a) (c -1)P,
(2)
whereP is uniondensity(L,/L).
Equation (2) assumesconstantreturnsto scale. This assumptionmay
be relaxedby includinga In L variableas a measureof establishment
size.
The coefficient
differential
on P measuresthe logarithmicproductivity
74
Addison/Hirsch
if it is assumedthatthe union productivity
of unionizedestablishments;
of labor inputs,the union
efficiency
effectsolelyreflectsthe differential
on P
effectis calculatedby dividingthe coefficient
labor productivity
by (1 - a). Also, to anticipatewhat follows,c > 1 impliesthatunions
(TFP), as obtainsaftersubtracting
have highertotalfactorproductivity
a ln(K/L) frombothsidesof (2),
In Q -a In K -(1 -a)ln
L = TFP - In A + (1 -a)(c
-1)P.
(3)
test.As Brown
A numberoflimitations
attachto theproductionfunction
and Medoffthemselves
noted,theuse ofvalue added as an outputmeasure
confoundspriceand quantityeffects.
That is, partof the measuredunion
differential
may resultfromhigherpricesin the unionized
productivity
effectstend to be
sector.Not surprisingly,
estimatedunion productivity
are made (e.g., Allen 1984b)and are rarely
lowerwhenpriceadjustments
largein studieswhereQ is measuredexplicitlyin physicalunits.Union
firms
can moreeasilypassthroughhighercostsinproductmarketssheltered
is
fromnonunionand foreigncompetition.Use ofvalue added,therefore,
most likelyto confoundprice and outputeffectsin aggregateanalyses
uniondensity.It is less likelyto
value added to industry
relatingindustry
analysesthatat least allow the possibilityof
be a problemin firm-level
controllingfora firm'sunion statusand industrydensity.A relatedposin equation(2) maycrudelybe tracking
sibilityis thattheunioncoefficient
wage differential.'
theunion-nonunion
The mostpointedcriticismleveledat theBrown-Medoff
approachhas
come fromMorganReynolds,who argues:"Estimatesof c are of no consequencein discoveringthe independentimpactof unionizationon probecausethereis no wayto statistically
separateobservedmarginal
ductivity
differentials
fromunion/nonunion
price[wage] differentials"
productivity
(Reynolds 1986, pp. 445-46). He points out that if profit-maximizing
firmsoperateon the labor demandcurve,theywill adjust employment
such thatthe marginalvalue product,OY/OL, equals the wage. If union
establishments
face a higherwage, theywill lower employmentso that
highermarginalvalue product.Accordingto
theyhavea correspondingly
Reynolds(1986,p. 445), we thuswill observe
w,/Wn= (OY/OLa)/(OY/OLn)=
c,
(4)
Itrelies
tousbyGreggLewisincorrespondence.
Thisargument
wassuggested
valueaddedis thesumofunionandnonunion
ontheargument
that,bydefinition,
thereturn
thatchiefly
to ownedcapital.Follaborcostanda residual
comprises
ofsuchan equationyieldsa ln(Q/L) function
lowingLewis,a Taylorexpansion
inconjunction
withlabor'ssharetimestheunion
inwhichuniondensity
(P) enters
is thatthecoefficient
on P fromeq. (2) varies
The suggestion
wagepremium.
withtheunion-nonunion
wagedifferential.
directly
UnionEffects
75
Rewheresubscriptsu and n denoteunion and nonunion,respectively.
gardlessof whetheror not capital and labor qualityare controlledfor
unionfirmsfacinghigherwage ratesmusthavea correspondstatistically,
inglyhighermarginalvalue productin long-runequilibrium.The productionfunction
test,accordingto Reynolds,provesonlythatunionwages
are higher,not thatunionsincreaseproductivity.
He is correctin statingthatfirms
How valid is Reynolds'scriticism?
suchthat
operatingon theirlabordemandcurveswill adjustemployment
OY/OL = w. Reynolds'scase, then,restson the questionof what in fact
intheproductionfunction
studiesand whether
is measuredeconometrically
settlements
are locatedon labordemandcurves.
unionwage-employment
Because the productionfunctionstudiescontrolfor capital inputsand
of thelabor
laborquality,it can be arguedthattheymeasureverticalshifts
demand schedule resultingfromunion productivityeffects,namely,a
comparisonof unionand nonunion(OY/OL)I K. By contrast,movements
along a nonshifteddemandscheduleimplya changingratioof labor to
otherinputs.The productionfunctionapproach,then,need not be an
effectif otherinputsin the
inappropriatetestof the union productivity
productionprocesscan be measuredaccurately.(Other problemsmayof
But, as Brown and Medoff,as
coursemake the approachinappropriate.)
well as Reynolds,pointout,unionfirmsfacinghigherwage ratesmustbe
moreproductiveif theyare to survivein the long run.Hence the union
sample
effectis not being measuredacross a representative
productivity
of firms.Rather,onlythoseunion firmsthathave survivedby increasing
to offsethigher
productivity
(a shiftin thedemandschedule)sufficiently
differentials
ofunionproductivity
unionwagesareobserved.Measurement
fromamonga sampleof survivingfirmsthusoverstatesthe productivity
firm.It does measurea productivity
effectof unionson a representative
ratherthan a wage effect,
however,althoughin the long run thesetwo
effects
shouldtendto converge.
The Reynoldscriticismmustbe further
qualifiedifwage-employment
outcomesarenotlocatedon thelabordemandschedule.As is well known,
settlements
(McDonald and
on thedemandcurveare generallyinefficient
offthelabordemandcurve,at lower
Solow 1981). Thereexistsettlements
levelsthanwould obtainthroughsequential
wage and higheremployment
preferred
by both the union and
wage and employmentdetermination,
A settlement
on thecontractcurve(formedbythetangencies
management.
and firmisoprofitcurves)requiresexplicitor
of the union indifference
oththat
wages and employment;
implicitcontracting jointlydetermines
firm
the
could
maximize
short-run
once
a
was
determined,
wage
erwise,
to thecorresponding
pointon the
profitability
by adjustingemployment
demandschedule.No staticinefficiency
would resultin theeventof setto as the"strongefficiency"
tlements
on a verticalcontractcurve,referred
case (Brown and Ashenfelter
1986). Both partiesagreeto maximizethe
76
Addison/Hirsch
to ownersand laborand thenbargainoverits
jointvalueof theenterprise
division.Despite skepticismas to the actual union politicalmodel and
offthe demand
contractualmechanismsthatwould producesettlements
schedule(Oswald 1984), empiricalstudiesgenerallyrejectthehypothesis
outcomeson the demandscheduleand findmixed
of wage-employment
(Brown and
settlements
supportforthe hypothesisof stronglyefficient
1986;Card 1986;EbertsandStone1986;MaCurdyandPencavel
Ashenfelter
1986;Svejnar1986;and Abowd 1987).
offthe labor demandscheduleobtain,the link
If contractsettlements
relativewagesand marginalproductsis severed.
betweenunion-nonunion
We do know, however,thatthe marginalvalue productof union labor
(measuredon the demand schedule) will be less than the union wage,
whereasthetwowill be equatedin nonunionfirms(ignoringthepossibility
wherebynonunionoutcomesmimicunion
largethreateffects
ofsufficiently
outcomes).The equalitystatedbyReynoldsin equation(7) needno longer
will be less than the
differential
hold; the union-nonunionproductivity
productionfunction
the
of
The
appropriateness
unionrelativewage effect.
issues.
on
econometric
rest
largely
to
continues
test,therefore,
of identicalproin
studies
made
most
is
the
assumption
Also troubling
sectors
and
(i.e.,
in
union
nonunion
the
functions
duction
a,, = ao). Such
studiesbytheabsenceofseparate
in aggregate
is necessitated
an assumption
data on the union and nonunionsectors,althoughunion densitycan be
sidevariablesin aggregatemodels.Brown
withotherright-hand
interacted
theconsequencesofrelaxingthisassumption
and Medoff(1978) investigate
sensitiveto assumeddifferences
and findthe estimatedunion coefficient
in au and ac, beingsharplylowerin mostcases.The assumptionof equivis probablylessserious-and lessnecessaryalentproductionparameters
studies(see Clark1980a,1980b).In addition,estimation
inindustry-specific
of the productionfunctionin log-linearform,made possibleby use of a
biasesupwardtheabsolutevalue
approximation,
first-order
Taylor-series
effect(Lovell, Sickles,and Warren
of the estimatedunion productivity
1988). The implicationis thatexistingCobb-Douglas studieshave overstated(by an apparentlymodestdegree)both positiveand negativeestieffects.
matesof unionproductivity
The productionfunctiontestalso may depend cruciallyon the ability
to controlaccuratelyforall inputsin theproductionprocess.Union and
in the qualityof unmay differsystematically
nonunionestablishments
arenotindependent
measuredorganizationalfactorsso that"firmeffects"
of union status.For example,inputssuch as managerialsupervisionand
andomission
withunionism,
thequalityoflaborrelationsmaybe correlated
That being said, neither
of thesefactorsmay bias the union coefficient.
Brown and Medoffnor Clark findtheirestimatesto be sensitiveto the
exclusionof laborqualitycontrols.
functionalform
One responseto the broaderproblemof a restrictive
has been to assume a translogratherthan a Cobb-Douglas technology
Union Effects
77
(Boal 1985; Bemmels 1987). The translogformallows one to test for
restrictions
on technologymade by Cobb-Douglas-for example,unitary
elasticity
ofsubstitution
betweeninputsand homogeneity.
Translogstudies
havedecisively
rejectedtheCobb-Douglasspecification.
Yet as Allen(1987)
pointsout,estimation
ofanyproductionfunction
byordinaryleastsquares
(OLS) is strictlyappropriateonly when inputquantitiesare exogenous
(see Nerlove[1965], pp. 29-34, fora moregeneralanalysisof thispoint).
This is clearlynot the case in the productionfunctiontest,posing a simultaneity
problemforall suchstudies.One responseto theproblemhas
been to measurethe relativeefficiency
of union and nonunionestablishmentsusing (translog) cost and profitfunctions(e.g., Allen 1987), in
whichtheunioneffectis capturedvia interceptand slope parameters
that
varyby union status.Both functionshave the advantageof permitting
a
directestimateof the net effectof unionson profitsor costs,in contrast
to the alternative
practiceof comparingunion coefficients
fromseparate
productionand wage equations.2
Evidenceand Interpretation
In lightof theconcernsregardingtheproductionfunctiontest,critical
evaluationof the empiricalevidenceis required.We make no attemptto
providea completereviewof what is now a largeliterature
(for surveys,
see Freemanand Medoff[1984];Hirschand Addison [1986]); rather,we
examinecloselywhatwe believearethemostimportant
studiesandpatterns
evidentin this literature.
The originalBrown and Medoff(1978) paper
will be our benchmark.
Usingaggregatetwo-digitmanufacturing
industry
data cross-classified
by stategroupsfor 1972,Brown and Medoffobtain
coefficients
on uniondensityof from.22 to .24, implyingvaluesof c - 1
(obtainedby dividingtheunion coefficient
by 1 - a) of from.30 to .31.
In a separateanalysisusingthe 1973-75CurrentPopulationSurvey(CPS)
files,theyestimatea logarithmic
union-nonunion
wage differential
of .23
(averagedovermalesand females).
Most ofthepotentialmeasurement
problemspreviouslydiscussedapply
with some forceto the Brown-Medoff
study.While theseproblemsare
largelyunavoidableand recognizedby the authors,any generalization
of
theirresultsmust meet the dual criteriaof plausibilityand consistency
2Analternative
approach
tomeasuring
productivity
differences
istheestimation
ofnonstochastic,
nonparametric
frontier
production
functions
(Forsund,
Lovell,
andSchmidt
1980).Theseestimation
procedures
allowproduction
efficiency
to
be decomposed
intoallocative(price)and technical
efficiency,
thelatterbeing
further
decomposed
intoa Farrellmeasure
ofefficiency
(a firmis technically
inefficient
ifit operateson theinterior
of itsproduction
set),a measure
of input
"congestion,"
anda measure
ofscaleefficiency.
In principle,
suchestimation
results
wouldfacilitate
inferences
as tohowunionsaffect
productivity.
Thisapproach
may
provepromising
forindustry-specific
unionproductivity
studies.
78
Addison/Hirsch
withsubsequentfindings.By eitherstandard,theirestimateof the union
is too high.Labor's sharein value added is .53 (Brown
productivity
effect
and Medoff1978,p. 367), implyingan increasein unitcost of 12% (.23
times.53) stemmingfromtheunionwage premium.A .22 to .24 increase
in totalfactorproductivity,
however,in turnwould decreaseunitcostsby
over 20%. Hence, the net effectof union wage and productivity
effects
would be to decreaseunitcosts substantially.
Althoughunion effectson
the profitratecannotbe ascertainedwith precisionfromthe change in
unitcosts(Clark 1984), theseparameterestimateswould almostcertainly
implyan increasein profitability
resultingfromunionism.But theprofits
literature
providesunambiguousevidenceoflowerprofitability
in unionized
industriesand firms.
Wessels (1985) casts furtherdoubt on the plausibilityof large union
estimatesbecause of inconsistencies
betweenthisresultand
productivity
employmenteffects.Specifically,he arguesthat if wage settingis constrainedby thedemandfunctionand laboris measuredin efficiency
units
in the productionfunction,it is impossibleto reconcileequal wage and
increaseswith conventional(below unity)estimatesof the
productivity
elasticityof labordemand.The possibleexceptionis whereunionsexclubut even here it is shown that the
sivelyenhancecapital productivity,
betweencapitaland laborwould haveto fallconof substitution
elasticity
siderablybelow unityto producetheresultthatunionsraiseproductivity
substantially.
Again,conventionalestimatesof the substitution
elasticity
are too highto reconcileapparentlysmallunionemployment
effects
with
the measuredimpactof unions on productivity.
Wessels concludesthat
mustbe wrong:eitherunionsdo notsubstantially
one oftheeffects
increase
reduceemployment.3
or theysubstantially
productivity
Thereare surprisingly
fewmanufacturingand economy-wideproductivitystudiesand,exceptforBrownand Medoff,nonereportsevidenceof
a positiveoverallunion productivity
effect.In perhapsthe best studyto
date,Clark (1984) uses theProfitImpactof MarketStrategy
(PIMS) data
for902 manufacturing
businessesfrom1970-80(over 4,600 usable observations)to estimatevalue-added(and scales) equationssimilarto (2). He
obtainsmarginally
coefficients
on theunionvariablefrom-.02
significant
to.-.03, in sharpcontrastto theresultsin Brownand Medoff.The Clark
studyhas theadvantageofa largesamplesize overmultipleyears,businesson unioncoverage,and a detailedsetof controlvarispecificinformation
ables (althoughthe union coefficient
is littleaffectedby the inclusionof
hisbasicpointis sound,limitations
attachtoWessels'sanalysis.
He
3Although
ofefficient
doesnotaddressthepossibility
bargaining
outcomes
offthedemand
curve,he appearsto say(Wessels1985,p. 103) thatunitcostsdo notchangeif
andwagesincrease
productivity
byequalpercentages
(thisignoresthefactthat
ofresearch
labor'sshareis lessthanone),andherather
glossesoverthediversity
findings.
Union Effects
79
the latter).The studycan otherwisebe subjectedto many of the same
criticismsthatapply to the Brown-Medoff
analysis.In Clark's separate
two-digit
industry
regressions,
positiveunionproductivity
effects
arefound
and petroleum(estimatesrange from 6%
only for textiles,furniture,
to 17%).
StudiesbyBemmels(1987), who estimates
a translogproductionfunction
usinga limitedsampleof46 surveyed
manufacturing
plantsin 1982,Hirsch
(in press),who estimatesa variantof equation(2) usinga sampleof 315
Fortune1,000 companies,and Lovell, Sickles,and Warren(1988), who
estimatean economy-wideannual time-series
model of equation(2), all
concludethatunionsdecreaseproductivity
(the estimateof c in thelatter
studyis implausiblylow). Based on theextantevidence,we concludethat
theaverageunionproductivity
effectis probablyquite smalland, indeed,
is just as likelyto be negativeas positive.
As expected,resultsfromindustryand firm-specific
productivity
studies
producea variedpicture.The primaryadvantageof thesestudiesis that
manyof the econometricproblemsinherentin the aggregatestudiesare
avoided. For example,outputcan be measuredin physicalunitsrather
thanvalue added, information
on firm-level
union statusis morereadily
can be allowed to differ
betweentheunion and
available,and coefficients
nonunionsectors.Moreover,recentstudieshave moved away fromthe
or have attemptedto avoid
overlyrestrictive
Cobb-Douglas specification
inputendogeneityproblemsby estimationof cost ratherthanproduction
functions.But theseadvantagesare achievedat thepriceof a loss in genof
erality.Taken as a group,however,we believethatwithinthediversity
theseresultsare to be foundsystematic
patternsthatincreaseour understandingof how unionsaffecttheworkplace.
Froma methodologicalperspective,
two ofthebestanalysesareClark's
studiesof the cementindustry(Clark 1980a, 1980b) and Allen's most
recentanalysisof the construction
industry(Allen 1987). Clark's studies
arenotablefortheuse ofphysicaloutputmeasures,forallowingproduction
functionparametersto varybetweenunion and nonunionplants,in controlling
forfirmeffects
throughthestudyofplantschangingfromnonunion
to unionstatus,and in introducing
a supervisory
labor inputmeasure.In
hiswiderstudy,Clark(1980b)investigates
between
productivity
differences
ninenonunion(29 observations)and 119union(436 observations)cement
plantsovertheperiod1973-76.He findsmarginally
significant
(usingonein the range.07-.10. An interesting
tailedtests)union coefficients
result
is thatunionproproductivity
effects
aremostpronouncedin theSouthwest
regionof his sample,wherenonunionfirmsare mostprevalent,although,
as before,theunioncoefficients
are estimatedwithconsiderableimprecision. In a novel studyof six cementplantsthatchangedfromnonunion
to union statusbetween1953 and 1976,Clark (1980a) reportsa positive
differential
of from6% to 8%. Althoughsignificance
union productivity
80
Addison/Hirsch
levelsareagainweak,evidencefromhis fixedeffects
modelindicates
a
effect
smallpositive
productivity
thatis apparently
sustained
through
time.
In perhapsthemostambitious
studyto date,Allen(1987)attempts
to
therelative
measure
efficiency
ofunionandnonunion
establishments
using
functions.
areestimated
costandprofit
Translogcostfunction
forthree
office
separatesamplesof buildingconstruction
projects(commercial
buildings,
schools,andhospitals).As notedearlier,
thisapproachallows
and also sidesteps
fortheendogeneity
of inputquantities
atproblems
betweenunionproductivity
and wageeffect
tachingto thecomparison
difestimates
fromseparate
andearnings
functions
production
containing
levelsofaggregation.
ferent
control
variables
orestimated
atdifferent
Allen
finds
inschoolandhospital
higher
costsforunioncontractors
construction,
butlowercostsin commercial
office
buildings,
ceterisparibus.The hyis soundlyrejectedacrossall
pothesisof equal costfunction
parameters
betweencoefin theschoolssample;however,
differences
specifications
arelessevident
in thecommercial
office
ficients
buildings
sample(while
thenumber
ofnonunion
is toolimited
to allow-suchtestsinthe
hospitals
hospitalsample).
A striking
result
thisstudyisthefinding
from
ofgreater
scaleeconomies
in theunionsector;diseconomies
appearin thenonunion
sectoratlower
morerapidly
withoutput.
Unionoffice
outputlevelsandincrease
building
exceeda threshold
construction
costslessoncebuildings
sizeand,thereafter,
becomessteadilylessexpensive
withoutput.For schools,on theother
costsfallwithoutputin theunionsample
hand,eventhoughmarginal
aremorecost
contractors
(and risein thenonunion
sample),nonunion
efficient
at all outputlevels.Takeninconjunction
withhisprofits
results,
a competitive
thesefindings
unionexplanation
support
foroffice
buildings
anda monopoly
unionexplanation
forschoolsandhospitals
(see,relatedly,
Allen1986a).Office
sector
whileschools
arebuiltfortheprivate
buildings
andsomehospitals
Allenarguesthatmarket
arebuiltforthepublicsector.
characterizes
segmentation
publicsectorcontracts,
allowingunionconIn commercial
struction
to havehighercostswithout
in profits.
sacrifice
office
theabsence
andgreater
ofmarket
buildings,
bycontrast,
segmentation
cost consciousness
requirecompetitive
unionism.Scale economiesin
areattributed
unionized
to theeffectiveness
ofunionhiring
construction
hallsinguaranteeing
of(alreadyscreened)
skilledlaborand,
largesupplies
in part,to theirrelevance
of certainrestrictive
practices
(e.g.,minimum
crewsizes)onlargeprojects.
He isunabletoquantify
theseeffects,
although
itdoesappearthathisresults
reflect
morethanjustcompositional
effects
(see Allen[1986b]forfurther
discussion
ofscaleeconomies).
We havedwelledat somelengthon theAllenstudybecausenotonly
doesitprovide
a glimpse
insidetheunionproductivity
blackbox,butalso
demonstrates
palpablythattheunionimpactis nota datum,hintsat the
andaddresses
a number
roleofmarket
of
potentially
important
structure,
Union Effects
81
in theliterature.
We notethattherationaleforpositiveunion
deficiencies
because
productivity
effects
is notrootedin collectivevoiceconsiderations
relationthatobtainsin construction.
of theshort-term
employment
studies,severalsysDespitesubstantial
diversity
in theindustry-specific
tematicpatternsare revealed.As previouslynotedin Hirschand Addison
effects
tendto be largestin those
(1986, chap. 7), estimatedproductivity
whereunion-nonunion
aremostpronounced.
industries
wage differentials
Of course,such a patternis exactlywhatwould be predictedby criticsof
theproductionfunctiontest.We contendthattheseresultsalso supporta
of unionization.In industriesor
traditional"shock effect"interpretation
firmswhereunions obtain largewage premiums,managementmustrereducing
spondto theincreasein laborcostsbyorganizingmoreefficiently,
slack,and increasingmeasuredproductivity.4
effects
are largelyrestricted
to the
Second,positiveunion productivity
wherecompetitive
pressures
privatesectorand appearto be mostsignificant
are intense.This latterpointwas takenup by Clark (1980b), who found
effectin the regionwherenonunioncomthe largestunion productivity
petitionwas most pronounced.Similarly,in an earlyand widely cited
in two cities,Mandelstamm(1965) idenstudyof residentialconstruction
in a nearbymetropolitan
tifiedcompetitionfromoutsidecontractors
area
observedin his more highly
as the chiefsourceof the greaterefficiency
unionizedcity.And in Clark's (1984) analysisof lines of business,the
differentials
were foundin
largestpositiveunion-nonunionproductivity
and
industries
havingrelatively
largenonunionsectors(textiles,furniture,
in public
effect
petroleum).The apparentabsenceofa sizableproductivity
bureaus
libraries(Ehrenberg,Sherman,and Schwarz 1983), government
(Noam 1983), schools(Ebertsand Stone 1987), and hospitals(Sloan and
Adamache1984), despitesignificant
unionwage premiumsin all but the
librarysample,supportsthe thesisthatproductmarketcompetitionis a
in unionizedmarkets(this themeis examined
spur to greaterefficiency
directlyby Allen [1987,1986a]).
Perhapsthemostdirectevidenceof a shockeffectis reportedby Clark
tounionismexhibitfairly
(1980a), who findsthatmanagement
adjustments
consistentpatternsin the wake of unionization.In all cases new plant
managerswerehired,and in mostcases therewas increasedemphasison
cuttingcosts,establishingproductiontargetsand goals, and improving
remains.
and communications.
Yet a problemofinterpretation
monitoring
4Causation could be arguedto work in the opposite direction,with unionThis appears
inducedproductivity
increasesleadingto largeunionwage premiums.
unlikely.If unionismraisedproductivity
directly,therewould be mutualgains
fromunionization,with labor and shareholderssharingin the gains. Thus we
in nonunionfirmsencouragingunionorganizingand
would observemanagement
in unionfirms.
higherprofits
Neitheroutcomeis observed.Moreover,firm-specific
market-clearing
wages.
productivity
changesshouldnot affectcompetitive,
82
Addison/Hirsch
cost-consciousecoThe evidenceindicatesthata relativelycompetitive,
nomicenvironment
is a necessaryconditionfora managerialresponseto
gain. Such a response
unions thatwill produce a positiveproductivity
shouldalso be larger,ceterisparibus,the largertheunion wage gain (or,
the greaterthe pressureon profits).It is preciselyin such
alternatively,
however,thatthereshould be relativelylittle
competitiveenvironments,
managerialslack and theleastscope forunionorganizingand wage gains.
This reinforces
our earlierconclusionthatthepossibilityforsizableunion
effects
is limited.Clearly,anyjudgmenttakenon thepotential
productivity
enhancement
hingescruciallyon one's priors
scope forunionproductivity
and slack in theeconomy.5
regardingthedegreeof competitiveness
and perhapsmore appealingexplaFor thisreason,a complementary
nation of the empiricalregularitiesin the literaturewould be that the
studiessuffer
froma selectivity
or survivalbias. As discussed
productivity
in
in the previoussection,union firmsunable to increaseproductivity
responseto unionwage increasesare less likelyto surviveand be included
datasamples.Thus,estimates
in researchers'
ofunioneffects
on productivity
(or profitsand costs) are biased estimatesof the union effecton a representativesampleof nonunionfirms.Not all firmsare "shocked" or able
unioncostincreases.In thelongrunthese
to offset
to respondsufficiently
firmswill not survive.
theworkplace
Directevidenceon theroutesthroughwhichunionsaffect
remainsmeager.As Brownand Medoff(1978,p. 374) state:"The idea that
unionsmake firms. . . moreproductivewould be morepersuasiveif the
is improvedcould be isolated."A cenmechanismsbywhichproductivity
traltenetof the collective-voiceframework
motivatingtheseanalysesis
ofvoiceforexitwill reducequitsand improvemorale
thatthesubstitution
and cooperationamongworkers.Relatedarguments
suggestthatimproved
structures
can lead to unionproductivity
gains.Yet theevidence
governance
evidencerevealsthatquits
and time-series
is opaque. While cross-section
are significantly
lower among union workers,ceterisparibus(Freeman
1980;Blau and Kahn 1983;and Mincer1983), thereis littleto suggestthat
thisadvantagesignificantly
Clark(1980a) recontributes
to productivity.6
portseitherno changeor evenhigherquitsfollowingunionizationin three
of thesix plantshe followedovertime.One routethroughwhichunion
is via the establishment
of procedures
voice mightimproveproductivity
powerduringa periodwhere
S Unionsinitially
mayorganizeand gainbargaining
is
marketpower.Union impacton profitability
a firmor industryhas significant
increase
likelyto intensify
overtimeas foreignand domesticnonunioncompetition
and unionwage increasescannotbe passed throughto consumers.
6 Brown and Medoffintroducea quits variableinto theirproductionfunction
of lowerquitson
to theeffects
and attribute
thereductionin theunioncoefficient
theresidual
byone-fifth;
productivity.
Thisprocedurereducestheunioncoefficient
comto "bettermanagement,
morale,motivation,
four-fifths
arelooselyattributed
munication,
etc." (Brown and Medoff1978,p. 74).
Union Effects
83
but thereis no directevidenceto supportthisview.
to handlegrievances,
Ichniowski(1986), using data from10 unionizedpaper mills,findsthe
numberof grievancesfiledto be inversely
relatedto productivity
(see also
Katz,Kochan,and Gobeille 1983); leftunansweredin thistreatment,
howof union voice in reducingand/or arbitrating
ever,is the effectiveness
grievancesrelativeto implicitgrievanceproceduresoccurringin similar
nonunionplants.
unobserved
In sum,we should expectto findrathermorepositiveand directindicationsoftheoperationofcollectivevoice.Furtherambiguity
is occasioned
by Allen's (1984a) findingthatabsenteeismis at least30% higheramong
union workers(even thoughhe reportsthatthe decline in productivity
fromthissourceis onlya fraction
of 1 percent).The greaterdissatisfaction
apparentlyexpressedby union workersneed not be inconsistentwith
union-inducedimprovements
in efficiency
via collectivevoice (Freeman
and Medoff1984,chap. 9; Leigh 1986). But neitheris it inconsistent
with
a shockeffect
explanationwherebytheworkpace is increasedin unionized
settings,
leadingto greaterworkerdissatisfaction.
Finally,thereremainsthe unresolvedproblemof union endogeneity.
We knowthatunionworkplacesdiffer
fromnonunionworksubstantially
places(Duncan and Stafford
1980;Freemanand Medoff1984). Can unions
moresuccessfully
organizewherethereexistspecial cost advantages?Or,
as suggestedbyBrownand Medoff(1978,p. 368), areunionsmoreheavily
Do U.S. companies'
organizedin less productive(two-digit)industries?
newerand largelynonunionplantshave higherproductivity
than their
older and morehighlyorganizedplants?We know of no studythathas
in estimating
satisfactorily
addressedtheissueofunionendogeneity
unionnonunionproductivity
differentials.
We concludethatthereis no compellingevidencethat,in general,the
net effectof unions on productivity
is positiveor negative.Ratherby
default,theexplanationforpositiveunionproductivity
wherethey
effects,
are observed,is best explainedby eithershock effectargumentsor by
in theproductionfunction
deficiencies
test.Actualunionproductivity
gains
appearto be generated
bymanagement
responsesto sizablewagepremiums,
significant
nonunioncompetition,and deteriorating
profitexpectations.
Estimatedunion-nonunionproductivity
differentials
may resultfroman
entanglingof union productivity,
Or we may be
price,and wage effects.
observingtheoutcomeof a competitiveprocessin whichless productive
unionizedenterprises
have been selectedout of the system.Accordingly,
our analysisof union effects
mustexaminedynamicaspectsof economic
performance.
III. Unionsand Profitability
of theproductivity
evidenceis the
Closelyrelatedto theinterpretation
issueof how labor unionsaffectprofitability.
One of the moretroubling
Addison/Hirsch
84
betweenthe
puzzles in theliterature
has been theseeminginconsistency
unionproductivity
ofa substantial
effect
and mounting
empirical
suggestion
reducesprofitability.
Information
on
evidencethatunionismsignificantly
unionwage and productivity
effects,
coupled withknowledgeof labor's
on unitcosts.For exshare,makespossiblea calculationof union effects
effect
ample,a productivity
of9% wouldleavea firm'sunitcostsunchanged
iflabor'ssharewere .60 and theunionwage premium15% (it is assumed
thatthefirmis fullyunionizedor,alternatively,
thatnonunionemployees
also receivea 15% premium).The linkbetweenunitcostsand profitrates,
In general,it dependson the natureof the
however,is not transparent.
the
bargainingsettlement
(e.g., on thedemandcurveor strongefficiency),
productdemandelasticity,
marketstructure,
and the elasticityof substitutionbetweencapitaland labor (Clark 1984).
Perhapsthe simplestcase is thatof sequentialwage and employment
and
productivity
determination
on thelabordemandcurve,withoffsetting
wage effectsthatleave unitcosts unchanged.For a productmarketmonopolist,Clark (1984,pp. 896-97) showsthattherateof returnon capital
willdecrease(increase)iftheelasticity
ofsubstitution
is greater(less) than
one, whereasthe returnon sales will remainunchanged(because Q and
P do not change).
is thestrongefficiency
case. For purposesof comparMore interesting
ison, examine,initially,the case of inefficient
sequentialwage and emon the demandcurve,in which(subjectto conploymentdetermination
straints)theunionmaximizes"rents,"R, and thefirmmaximizesprofits,
11,givenwu. That is,
max R = (wu - wo)L
(union maximand),
(5)
and
max 11 = PQ
-
rK - wuL (firmmaximand),
(6)
whereQ is output,P productprice,K capital,r thepriceof capital,wu is
the realizedunion wage,wo is the wage in the absenceof unions,and L
is employment.
A measureof the excessof the union wage bill over the
competitive
wage bill,R, has been a commonmaximandassumedin the
literature
(e.g., Rosen 1969).
Such an outcome is inefficient:
settlements
by both parties
preferred
existoffthe demand curve,at a lower wage, and a higheremployment
level.An efficient
bargainingsituationon a verticalcontractcurveimplies
thatthetwo partieswill maximize(borrowingterminology
fromAbowd
[1987]) the totalvalue of the enterpriseV, beingthe sum of firmprofits
(11) and union rents(R), and thenbargainover divisionof the surplus.
MaximizingV resultsin thesameoutput,price,and inputusageas obtains
85
UnionEffects
in thecase wherethe firmmaximizes11 subjectto the competitivewage
cost wage,wo; thatis,
or opportunity
max V = 11 + R = PQ
-
rK
-
waL + (we
-
wo)L = PQ
-
rK
-
wOL.
(7)
costwage
accordingto theopportunity
Thus,thefirmadjustsemployment
and not its "own" wage. Here, the union has no short-runreal effects
to whichwe turnsubsequently).Since
(unionswill have long-runeffects
in thiscase, Q, P, K, and L are not affectedby the union,we can state
unambiguously
thata unionwage increase(wa > wo), withno offsetting
increase,will decreasethe profitrate,whethermeasuredby
productivity
ratesof returnon sales or capital,the price-costmargin,or market-value
measures.
A positiveunionproductivity
effect,
takenin conjunctionwitha union
wage increasethatleavesunitcostsunchanged,will also resultin a lower
profitrate(Clark 1984). Continuingto assumethatthe firmacts as if it
effect
union productivity
facescompetitivefactorprices,a factor-neutral
would have the same effectas a decreasein marginalcosts, increasing
optimaloutputand loweringprice.Input usage will increaseor decrease
(K/L remainsconstant)dependingon whethertheproductdemandelasticityrj is greateror less than unity(Clark's monopolymodel implies
1 > 1). The profitratewilldecreasesinceP falls,unitcostsremainconstant,
withoutput.
and theuse of K variesproportionately
The pointof the above discussionis thata comparisonof union wage
and productivity
effects
estimateoftheunion
providesonlyan approximate
effecton profitrates.By the same token,evidenceon union wage and
profiteffects
does not allow a preciseestimateof the union productivity
effect.Empiricalevidence on union-nonunionprofitrate differentials,
estimates.
ofproductivity
however,does permitus to gaugetheplausibility
is importantin itsown rightsinceit is
More fundamentally,
profitability
thatshouldmostaffectfirmsurvival,managementbehavior,
profitability
and long-runresourceallocation.
Beforeturningto the empiricalevidenceon union profiteffects,
some
commentsarein order.The empiricalstudiesare surprisingly
preliminary
thelong-standing
attention
givenunionwage
recent,especiallyconsidering
studiesfrom
effects
by laboreconomistsand the plethoraof profitability
theindustrialorganizationliterature.
Studieshave employedas theirunit
of observationeitheraggregateindustry,firm,or line-of-business
(LB)
measureshave been examined.The industry
data. Severalprofitability
use theprice-costmargin(PCM), definedas (total
studiesmostfrequently
revenue- variablecosts)/totalrevenue,typicallymeasuredby
PCM = (value added - payroll - advertising)/shipments.
(8)
86
Addison/Hirsch
Firm-levelanalyses,by contrast,are likelyto use accountingmeasuresof
the rateof returnon sales (r, = earnings/sales)or the rateof returnon
capital (rk = earnings/assets).
Severalof thefirm-level
analysesfocuson publiclytradedfirmsand use
the most frequentbeing
a stock marketvalue measureof profitability,
Tobin's q, definedas
q
=
marketvalue/replacement
costsof assets.
(9)
Firm-levelstudiesalso have employed"events" analysesthat calculate
changesin a firm'smarketvalueattributable
to notfullyanticipatedevents
involving
unionization(e.g.,representation
electionsor changesin contract
provisions).A typicalmeasurehereis thatof cumulative
abnormalreturns,
CAR, definedas
CAR
=
ract- rexp,
(10)
whereractis a firm'sactualrateof return(includingdividendsand capital
gains) and rexp is the firm'spredictedrateof returnbased on a "market"
equation(which regressesa firm'sr on a broad-basedmarketportfolio's
r) estimatedover some period well beforeand/or afterthe event.CAR
are summedover the period forwhich it is believedpublic information
about, or anticipationof, the eventmightaffectthe stock price. CAR
typicallyare averagedover a sampleof firmsin orderto draw inferences
as to averageunioneffects.7
is entirely
PriceNone ofthemeasuresusedin theliterature
satisfactory.
cost margins(PCMs) are readilymeasurablewithindustrydata and correspondmost closelyto the excess of price over variable(or marginal)
cost,on whichso muchindustrial
organizationtheorycenters.In practice,
however,PCMs may not correspondcloselyto the profitmeasuresthey
are intendedto represent(Liebowitz 1982). At thefirmlevel,accounting
ratesof returnare readilyavailablefromfinancialreportsand measure
actualor realized,as opposed to expected,earningsat particularpointsin
time.Unlike accountingmeasures,stockmarketindicatorssuch as q are
ratherthan historical,measureperformanceover time
forward-looking
ratherthanforsingleperiods,adjustforriskdifferences
acrossfirms,and
areless likelyto be influenced
byendogenousaccountingprocedures.The
eventstudiesalso use stockmarketmeasuresand havetheadded advantage
of identifying
theeffects
of specificeventson marketvalue. Yet isolating
the effectsof unanticipatedeventsis oftenelusive(Binder 1985), and it
see BeckerandOlson (1987).
'For a detailedreviewofunioneventsstudies,
ofstrikes
onsecurity
examining
theeffect
analyses
events
Here,wedo notconsider
prices.
Union Effects
87
(e.g., certifito generalizefrom"marginal"unioneffects
maybe difficult
cation electionsand concessionbargaining)to averageunion effectson
profitability.
is presented
on profitability
ofunioneffects
ofU.S. estimates
A summary
in table 1. The moststrikingresultof thestudiesis thecommonthemeof
among
differences
in unionregimes,despitesubstantial
lowerprofitability
analysesin methodology,data source,unit of observation,timeperiod,
Moreover,themagnitudeof thereductionin
and measureof profitability.
profitsis large.Althoughobservedpercentagereductionsare not strictly
Of the aggregatemanucomparable,thereare some strikingsimilarities.
andAnnualSurvey
studiesusingCensusofManufactures
industry
facturing
data,Freeman(1983) findsthatmovingfrom0% to 100%
ofManufactures
union coveragelowersprice-costmarginsby 13%-19%. Corresponding
estimatesby Voos and Mishel (1986a) and Domowitz, Hubbard, and
Petersen(1986, 1987) point to reductionsin PCMs of 23% and 22% to
25%, respectively.
analyses,Clark(1984) reportsthatunionbusinesses
Amongthefirm-level
19% and
have ratesof returnon capital and sales thatare, respectively,
while Hirsch and
18% lower thanthoseof theirnonunioncounterparts,
Connolly (1987) estimateunion effectson r, rangingfrom11% to 17%
and on Tobin's q from13% to 20%. Hirsch (in press) reportssizable
in rk and q betweencompanieswith low and high levels of
differences
collectivebargainingcoverage.Ruback and Zimmerman(1984) obtain
butfindon averagethatunionwinsin representation
highlyvariableresults,
electionsreducemarketvalueby3.8%,whereasunionlosseslowermarket
value by 1.3%. Abowd (1987) calculatesthe cost of union contractsfora
sample of publiclytradedfirmsand findsthatunanticipatedchangesin
roughlydollarfordollar,bychangesin shareholder
unionwealthareoffset,
wealth.An anomalyamongthestudiesis Allen's(1987) findingthatprofits
projectsare at leastas highas thoseon nonunion
on union construction
projects,despitehigherunioncostsin two of his threesectors.8
Some perspectiveon themagnitudeof theseestimatesis in order.Typically,the reportedestimates(or partialderivatives)are evaluatedon the
basisofa changefromzero to completeunioncoverage,whenin factmost
industriesand largefirmshave partialcoverage.On the otherhand,the
maybe biased upward(towardzero) in many
negativeunioncoefficients
bias, measureof thesestudiesforreasonshavingto do withsimultaneity
to
more
are
Unions
bias.
likely organizein
menterror,and selectivity
8 Allenattributes
pricesto
higher
to passthrough
ofunionprojects
theability
unionandnonunion
whereby
segmentation
market
ofgeographical
theexistence
eachother.
against
directly
compete
rarely
construction
firms
inschoolandhospital
inthesesectors
is possiblebecauseofthelackofcostpressure
Suchsegmentation
appliesherebutnotin hisprivate
thattypically
wagelegislation
andprevailing
sample.
office
building
0
C
U
CZ>
CZC
~~~~~~~~~~1
CZ
~~~~~~~C)
U
U~~~~~~~~~~
o
L) S
g -~~~~~~~~~~~-o
4,0
(JS~~~~l
4- t
,UU
4-
0~~~~~~~~~~~~~~
CZC
E C)CL
o
o
00o-
-
U
-~~~~~~~~C
1C0
-
I
0
CsCL0
88
C:
C) U
C
-1Z
-'
C'
C'S
Ct 0
0
o
uC~-~U
C
Q<
Rt
-.
o5
>
4.
C
U
.
C0)
-
s. ,*
;C
c
:
^
U
t
.
5C .aW
-r
A
4W
-
C
Co
CI
*~~~C0)
c) wno)
0)*0)
U
U
Ur.
C's0
E-r,
sU
U
r.
$00
-,
aE
-
0)
,-3.
sr W-Coo
4io
_
o Z
x
H
>
c
~
;
H
,
C'
sE
va~~~~~)
009
~~~~~~~C)
0
P4~~~~~~~~~~(
4
'S89'
cn
~~~9o~~~~~
m
C'SQ
71
t
0
C'S
I
4
7:
.y-C'
Co
-
-
7:1 55-U
$.
*:1
C'S
(
U
.
C
00 v
*
-U
1o.
U
>
0)
U
z
t
:
0>71
U(n
7
C's
C's~
Qj
C
- Q
I
0)-
C's
U
~~~~~~UC
U~~~~~~~~~
-U
C'
-
C~~~~~~~~~~~'
~
s
>
U
OL..
,
t~~~u~~
U~~~~~~r
U
U
cl
00
z
j
lZ
~.
o ~.-C'
U
-
U
'90
C
U 4-,
C
c
7:1
A
0)~~~~~~c
o
)
CIS
(n
U
a
~U
~
U
ctU
(n
ON
0)
CIS
(-1
~
C't
(f
~~~~~~~~~~~
~
-.
'S
x
a
.-C~~~~~~~~j
C
tC
a
x
U
ucs,
c
CIC-
C~~
_
c
U
C
I
C-~
$
U
0
CIS~~~~~~~~
U
O
CI
4
(n~~~
>
v
'Ct
:$C'S
>
~~~~
)0)~
-~~~~~~~
~ ~
C'~~~~~,,U
0)
~
CIS
U
X
CX.C
C'
U
U
~
Ct
C's
>
C~~~~~~~~~~~~~~~~'
91~~~~~~~~~~~~~~.
-
z -c
'
92
Addison/Hirsch
industries
and firmswherepotentialprofitsare high,biasingtowardzero
in OLS profitequations.Voos and Mishel
the(negative)unioncoefficient
bias, obtainingestimatesof a 35%
(1986a) findevidenceof simultaneity
reductionin PCM whenunionismis endogenous,as comparedwith23%
withOLS (Hirsch and Connolly [1987] providesimilarqualitativeevidence). Substantialmeasurementerrorin the union variablemay arise
industry
industry
unionmeasuresto four-digit
frommatchingthree-digit
results
or to individualfirms.Confidencein theunion-profits
observations
has been enhanced,however,by studiesmeasuringunion coverageat the
line of businessor companylevel (Clark 1984; Abowd 1987; Hirsch,in
the
problem.The data reflect
press).Finally,thereis a potentialselectivity
thathavesurvivedunionization;
performance
onlyofthoseestablishments
thus,union effectson survivingfirmsmay be a biased estimateof actual
firmsand potentialeffects
on currentnonunioneffects
on nonsurviving
union firms.
Rubackand Zimmerman's(1984) eventsanalysis,and thecloselyrelated
studybyBronarsand Deere (1987), provideperhapsthemostdirectmeaon
suresof the expectedeffectof union organizingand representation
marketvalues. Inferencesdrawn fromthese studiesmust be qualified,
however.The unionelections(coveringat least750 workers)typicallydo
notincludetheentirefirm,theprofiteffects
fornew (marginal)unionized
of union effectsin existingcompanies,
firmsmay not be representative
and resultsare highlyvariableacrossfirms(Ruback and Zimmermanfind
thatequityvalues increasedin 35% of the firmswhereunions won representation).
Althoughthe magnitudeof the profitreductiondue to unionization
itdoes appearto be substantial.
This in turn
cannotbe identified
precisely,
issueofthemeansthroughwhichsuchsignificant
profit
raisesthedifficult
those
To addressthisissue,we mustfirstidentify
reductionsare effected.
mostaffected
byunionsand,second,locatethesources
firmsor industries
ofunionrents.To theextentthatunionsreduceprofitsbelow competitive
For thisreason,
levels,thelong-runsurvivalof unionfirmsis threatened.
of
are
economic
profits
derivingfrom
rents
themostlikelysources union
marketpower and quasi rentsassociatedwithspecial firmadvantagesor
Note thatmonopoly
capitalinvestments.
long-livedphysicaland intangible
at low
profitsneed not be fullyappropriablewhere firmscan transfer,
On theotherhand,approcost,productionto a nonunionenvironment.
priablequasi rentsneed notbe associatedwithmonopolypoweror suprareturns;rather,theymaysimplyrepresentnormalreturnsto
competitive
(Klein, Crawford,and Alchian 1978).
long-livedinvestments
Much attention
has focusedon therelationship
betweenunions,profits,
and marketstructure.Clark (1984) findsforthe PIMS sample thatthe
associatedwithunionizationis entirelyrelargereductionin profitability
strictedto businesseswithlow marketsharesdespitemuchhigherprofit
ratesin firmswithhighmarketshares.Clark rationalizesthissurprising
UnionEffects
93
mechanismby whichroughlyequal
resultby postulatinga transmission
union wage increasesare generatedthroughoutan industry,with large
on low marketshare,low-marginfirmsand negpercentageprofiteffects
firms.This resultis not
ligibleeffects
on highmarketshare,high-margin
elsewhere(e.g., Hirschand Connolly1987) and would seem
corroborated
to be specificto thePIMS sampleof businesses.
Freeman(1983),
In whatappearsto be the leitmotivof thisliterature,
Karier(1985), and Salinger(1984) conclude that negativeunion profit
arerestricted
to highlyconcentrated
industries
(but see Domowitz,
effects
Hubbard,and Petersen1986;Voos and Mishel1986b).Freemanand Karier
base theirconclusionson industryPCM equationsin which an industry
withconcentration
(or concenunionizationvariableinteractsnegatively
profitreductrationdummies).Both authorsreportlargeand significant
industriesand an absenceof profitreduction
tionsin highlyconcentrated
in lowlyconcentrated
industries.
Salinger'sevidenceis lessstraightforward.
He estimatesa nonlinearleastsquaresprofitequation(withq as theprofits
measure)in which industryunionismentersin a three-wayinteraction
His union coefwithconcentration
and a vectorof profitdeterminants.
ficientcan be treatedas the"tax rate"on thereturnsproducedjointlyby
and theprofitdeterminants.
He, too,concludesthatunions
concentration
capturemostof themonopolyprofitsassociatedwithconcentration.
The abovestudiesappearto showthatunionlaborcapturesa significant
whilehavinglittleeffectin relatively
shareofpotentialmonopolyprofits,
industries.
Theyalso establishthattheomissionofunion
morecompetitive
variablesin traditionalprofitmodelsis a seriousflaw.Despite the appeal
of thesefindings,
theconclusionthatindustryconcentration
providesthe
majorsourceforunionrentswould appearto be incorrect(fora detailed
critique,see Hirsch and Connolly [1987]). For example,it is far from
does in factproduceanyexcessprofitsfrom
settledwhetherconcentration
whichunionscan extractrents.Few firm-level
analysesfindconcentration
and manyfinda negativerelation
determinant
ofprofits,
to be a significant
betweenthe variables(Bothwell,Cooley, and Hall 1984).9 Indeed, Raon
venscraft
(1983) and othershave shown thatthe positivecoefficient
in industry-level
concentration
profitstudiesmayinsteadcapturetheeffect
offirm-level
marketshare.Usingfirmor line-of-business-level
data,market
ofprofits,
whereasconcentration
shareis an important
positivedeterminant
is, ifanything,
negativelyrelatedto profits.
is the major source forunion gains,there
Moreover,if concentration
evidencefromthe labor marketindicatinga
should existcorroborative
forunionsto
findslittleevidence
generates
profits
thatconcentration
9Salinger
toFreeman
(1983)andKarier
ina manner
similar
hisresults
capture
yetinterprets
leastsquares
Salinger's
nonlinear
(1985).HirschandConnolly(1987)reexamine
formis
functional
results
usinga similardatabase.TheyshowthatSalinger's
thatconcenforthecontention
andthatthereis littlesupport
undulyrestrictive
a sourceforunionrents.
profits
provide
tration-related
94
Addison/Hirsch
differential
in moreconcentrated
compensation
largerunion-nonunion
studies
genfunction
industries
(HirschandConnolly1987).Yetearnings
increases
either
unionornonthatconcentration
erallyfindlittleevidence
oflarger
microlevel
evidence
unionwages,andthereexistsno significant
in moreconcentrated
industries
(Lewis 1986,pp. 154unionpremiums
smaller
inmoreconunionpremiums
appearsomewhat
55). Ifanything,
on Toevidence
industries
(e.g.,Mellow1983).Andfirm-level
centrated
obtained
thattheresults
bin'sq andr,(HirschandConnolly1987)suggests
and omitted
to specification
in earlierstudiesmayhavebeensensitive
10
variables.
structure
hasshifted
awayfroma focuson market
Recently,
attention
In
as a sourceof uniongainstowardquasi rentson firminvestments.
model,itisarguedthatwhenthecapitalcycle
Baldwin's(1983)theoretical
the
thesurplusthatprovides
is longrelative
to theunion'stimehorizon,
to capturebyorcapitalis vulnerable
return
on durableand specialized
willsharequasirents
ganizedlabor.Oncesuchcapitalis inplace,thefirm
to shutthefirm
down.Baldwin
thanallowa strike
withtheunionrather
islikelytokeepinplaceinefficient
contends
a unionfirm
that,inresponse,
unionwagedemands.
Subsequent
papers
capitalas a meansofmoderating
with
to be compatible
thatunionbehavioris unlikely
haveemphasized
thatis,thejointwealthofshareholders
andpotential
pluscurrent
efficiency;
as longasunionmembership
willnotbemaximized
future
unionmembers
is nontransferable.
Hirsch,andHirschey
(1986)extendBaldwin'srent-seeking
Connolly,
unionsas a taxon the
capital,treating
modelto thecase of intangible
research
and development
fromnontransferable
(R&D) capital.
returns
arkfoundto add lessto
withthatmodel,R&D investments
Consistent
industries
inhighly
thanto thevalue
unionized
valueoffirms
themarket
Lawrence
and
Lawrence
in
industries.
unionized
of firms lowly
(1985)
in a declining
withhighfixed
industry
developa modelin whichfirms
unionbargaining
power.
capitalwillfaceincreased
costsfromlong-lived
in the
be
enhanced
is
union
that
powermay
Theirargument
bargaining
the
in
a
decline
decrease
demand.
Such
run
a
decline
may
secular
short by
in
of
substitution
the
due
to
a
decrease
of
labor
demand
elasticity
elasticity
to substitute
fromtheinability
capitalfor
lumpylabor-saving
stemming
toexplain
tothesteelindustry
unionlabor.Themodelisappliedspecifically
steeloutput
unionwagepremium
accompanied
falling
whyan increased
'? Hirschand Connolly(1987) concludethatreturnsto marketshare,R&D, and
limitedforeigncompetitionmay be more likelysourcesof union rentsthan is
TheirmarketshareresultsconflictwithClark's (1984) rathercounconcentration.
Note thata likelysourceforuniongainsarerentsmadepossible
findings.
terintuitive
suchas postal
protectionand regulation;forexample,in industries
bygovernment
airlinesand trucking.
deliveryand, priorto deregulation,
UnionEffects
95
duringthe seventies,followedby rapidemploymentdeclinesduringthe
Linnemanand Wachter[1986]).
eighties(see, relatedly,
modelsis thattheuniontax on
A predictionof theunionrent-seeking
investment
behaviorand,subsequently,
that
a firm'squasi rentswill affect
We returnto thesethemesin the
firm'slong-runeconomicperformance.
or value
nextsection.But notethatevidencesupportingstrongefficiency
givenlevelsof quasi rents(Abowd 1987), does not imply
maximization,
if union bargainingpower affectsfirminlong-runallocativeefficiency
vestmentbehavior.
to effecta comparisonof
illuminating
Finally,it would be particularly
virtually
survivalratesbetweenunionand nonunionfirms.Unfortunately,
is availableto us in thisarea.An exceptionis the limited
no information
analysisby Kaufmanand Kaufman(1987) of the automotiveengineand
concludingthatfirmsorganizedby theUnitedAuto
bodypartsindustry,
morelikelyto havefailedthannonunion
Workers(UAW) aresignificantly
and non-UAWunion firms.This findingis consistentwithestimatesdefirmsof a largeUAW compensation
rivedfromtheirsampleof surviving
premiumrelativeto non-UAWand nonunionfirms.
(but notproductivity)
to lowerprofitability
under
The above evidencepointsunambiguously
unionism.Despite a varietyof statisticalproblemsthattendto bias estimatedprofiteffectstowardzero, unions' trueimpactis likelyto reside
profits
withintherangeofestimatespresentedin table1. The productivity
puzzle statedat the outsetof this section-that evidenceof largeunion
effectsis incompatiblewith evidenceof large profitreducproductivity
tions-may be more apparentthan real. Our readingof the evidenceis
seem likelyto resultfroma combination
thatreductionsin profitability
of union wage increasesand rathersmall (and possiblynegative)union
effectson productivity.
(Note again the variabilityin each effectacross
firms,
industries,
and time.)Althoughit is notin doubtthatunionsreduce
over long periods of time,competitivepressuresin labor,
profitability
limittheprofitspreadbetween
product,and capitalmarketsmustultimately
union and nonunionfirmsor lead to reductionsin the size of the union
sector.The steepdeclinein privatesectorunionismmayhave resultedin
ofU.S. companies
no smallpartfromtherelatively
poorprofitperformance
in ecoduringthe 1970s.The magnitudeof union-nonuniondifferences
nomicperformance
duringthe 1980sawaitsdetailedempiricalscrutiny.
IV. Unions, Investment Behavior, and Productivity Growth
on-the-demand-curve
modeltreatsa unionwage increase
The traditional
as an exogenousincreasein the priceof labor and a relativedecreasein
the price of capital. Profit-maximizing
firmsrespondby freelyvarying
theirinputmix,increasingthe relativeuse of capital and higher-quality
labor.Investment
activitymayincreaseor decrease,dependingon therelof scale and substitution
theunionrentativestrength
effects.
By contrast,
96
Addison/Hirsch
seekingmodel views union wage increasesas an outcome made possible
by a tax on thefirm'sreturnsor quasi rentsaccruingfrommarketpower
A firmfacinga powerfulunion is less likely
and long-livedinvestments.
to putin place newlong-livedcapital(evenofa labor-saving
type),because
a unionwitha crediblestrikethreatcan appropriatea shareof thenormal
In the faceof effective
returnsfromthatinvestment.
rentseeking,firms
in tangibleand intangiblecapital whose returnsare
reduceinvestments
mostvulnerableto unioncapture."
Empiricalevidencewithdirectbearingon the union rent-seeking
hypothesisremainsmeager.Baldwin (1983) contendsthat her theoretical
well to the steelindustry(see, also, Lawrence
model applies particularly
and Lawrence1985). Connolly,Hirsch,and Hirschey(1986) hypothesize
thatfirmsin highlyunionizedindustrieswill investless in R&D because
such investments
add less to the marketvalue of union than nonunion
firms.Based on evidencefroma 1976 sample of 367 Fortune500 firms,
havesignificantly
lower
theyfindthatfirmsin highlyunionizedindustries
R&D intensity(R&D investment
divided by sales). Union and unionprofitability
interaction
variablesenternegativelyin theirR&D intensity
bothdirectly
equation,suggestingthatunionsdecreaseR&D investment
and also througha redirection
of earningsaway fromvulnerableR&D.
Two recentstudiesalso conclude thatinnovativeactivityis less pronouncedin unionizedfirms.Acs and Audretsch(1987) use 1982 manudataand findsignificantly
fewerlarge-and small-firm
innovations
facturing
in highlyunionizedindustries.12Hirsch and Link (1987) providesimilar
evidencefroma surveyof 315 small-and medium-sizedNew York state
businessesin 1985.Theirorderedprobitmodels,in which
manufacturing
Likert-scaleresponsesto surveyquestionson the importanceof product
" In additionto thepapersby Baldwin(1983), Lawrenceand Lawrence(1985),
and Connolly,Hirsch,and Hirschey(1986), see Grout (1984) and Tauman and
Weiss (1987). Using a Nash bargainingmodel,Grout shows thatfirmsfacinga
union will have lower levelsof investment
when contractsare nonbindingthan
when theyare binding.Tauman and Weiss develop a duopoly model in which
eitherthe union or nonunionfirmmay choose the mostproductivetechnology,
dependingon thesetof assumptionschosen.In thecase wheretheunionchooses
a wage afterfirmsselecttheirtechnologies,
theunionfirmis unlikelyto adopt the
mostproductive
technology.
Note thatevenifcooperative(i.e.,jointlymaximizing)
will be below the competitivelevel if the union's
settlements
obtain,investment
timehorizonis shorterthantheexpectedlifeof capital.For a fullerdiscussion,see
Hirsch(in press).
12 The numberof innovations
are calculatedseparatelyforlargeand smallfirms
withinindustries.
The data wereprovidedby the Small BusinessAdministration,
whichdefinesan innovationas "a processthatbeginswithan invention,
proceeds
withthedevelopment
oftheinvention,
and resultsin introduction
ofa newproduct,
processor serviceto themarketplace"(Acs and Audretsch1987,p. 110,n. 1).
Union Effects
97
innovation(relativeto theircompetitors)are thedependentvariables,inis significantly
dicatethatinnovativeactivity
lessimportant
forunionthan
fornonunionbusinesses.
Recentwork by Bronarsand Deere (1986) and Hirsch (in press) offer
more comprehensiveanalysesof union effectson firminvestmentand
financialbehavior.Bronarsand Deere developa dynamicmodelin which
joint wealthmaximizationby shareholdersand currentunion members
and all current
does notmaximizerentsaccruingto shareholders
and future
unionmembers.Because currentunion memberscannotbe compensated
by futureunion members(e.g., throughthe purchaseof union memberhave an incentiveto shiftpositivecash
ship), boththeyand shareholders
flowscloserto thepresentand defernegativecash flowsfurther
intothe
future.Hence, Bronarsand Deere predictthatunionizedfirmswill have
lower tangibleand intangiblecapital investments,
offera higherpayout
and relymoreheavilyon debt financing.Empirical
rateto shareholders,
resultsfroman unbalancedpanelof756 publiclytradedfirmson theCompustattapesbetween1972 and 1976,matchedto industry
union coverage
data,providesupportforthesehypotheses."3
Otherthingsequal, theyfind
in moreunionizedindustries
thatfirms
have(1) lowerinvestments
in capital
and lowercapital-to-labor
ratios,(2) lowerR&D investments,
(3) lower
advertising
expenditures,
(4) higherdebt and debt-equityratios,and (5)
higherpayouts(dividendsand stock repurchases)and a higherratio of
payoutsto retainedearnings.All union resultsare highlysignificant,
the
in theR&D and debt-equity
magnitudeoftheunioncoefficients
equations
beingparticularly
large.
Hirsch (in press) providesa similaranalysisfordata on 315 Fortune
1,000firmsduringthe 1970s.This studyemploysa firm-level
measureof
union coverage(based on a 1972 survey)and includesdetailedindustry
controlsin the regressionequations.Hirsch findsthatunionized firms
investsignificantly
less in researchand developmentand physicalcapital
thando similarnonunionfirms.And in a recentstudyusingthe Ruback
and Zimmerman(1984) electiondata,Bronarsand Deere (1987) findthat
firmsdecreasecapitalinvestment
and employment,
whileincreasingdebt,
elections(regardlessof the outcome) but
followingunion representation
findno conclusiveevidencethatR&D or advertising
expenditures
are affected.
and decreasedtangibleand intangiblecapital inLower profitability
vestments
amongunion firmsare likelyto have a negativeimpacton the
long-runproductivity
growthof firms.This bringsus to what at first
the sharplydifferentiated
results
appearsto be a puzzle in the literature:
13 Bronars
dummiesor detailedindustry
andDeere (1986) do notincludeindustry
controlvariablesin theirregressions.
98
Addison/Hirsch
ofproductionfunction
studiesindicating
positiveunionproductivity
effects
on the one hand,and the R&D-based productivity
growthliterature
on
theother.Most R&D studiesthatincludea union argumentreporttotal
factorproductivity
growthto be negativelyassociated(but not always
significantly
so) withthelevelofunionism(e.g.,Terleckyj1974;Kendrick
and Grossman1980;Mansfield1980;Link 1981;and Sveikauskasand Sveikauskas1982;but foran exception,see Clark and Griliches[1984])."4
But thereneed be no inconsistency
betweenthe two sets of findings.
Fromequation(3), a positivecoefficient
on theunionvariablein theproductionfunctiontestimpliesthatunionizedestablishments
have higher
totalfactorproductivity.
an attemptedreconciliationof the
Accordingly,
productivity
and productivity
growthliteratures
mightproceedas follows:
unionsincreasethe levelof productivity
(perhapsthroughshock effects)
butsubsequently
retardproductivity
growth,possiblyreflecting
long-run
responsesto decreasedprofitexpectations.An alternative
but moreconventionalargumentwould be thatthe legacyor heritageeffects
of union
work rulesare negative.This is the argumentespousedby Maki (1983),
in an analysisof thegrowthof totalfactorproductivity
in Canadian manufacturing
duringthe period 1926-78. Ignoringthe errorterm,Maki estimates the equation
p = ao + aP
+ a2AP + a3S,
(11)
wherep is the annualgrowthrateof totalfactorproductivity,
P is union
density,and S is days lost throughstrikesdividedby total employment
(we ignorecontrolvariables).The firstuniontermis supposedto pick up
the long-termeffectsof unionismon growth,while firstdifferences
in P
are intendedto proxyimpactor shock effects.
Maki findsthatthe shock
effects
are positiveand thelong-term
effects
are negative.For thevarious
it takesfrom5 to 8 years(coefficient
specifications,
a2 dividedby a,) for
theone-timepositiveeffectof an increasein unionizationto be offsetby
thecontinuinglong-term
effectof unionismin slowinggrowth.
In a relatedanalysisof U.S. productivity
growthin two-digitmanufacturingindustriesfrom1957 to 1973,Hirschand Link (1984) show thatif
one specifiesthe standardCobb-Douglas model in difference
form,then
changesin totalfactorproductivity,
p, shouldbe a functionof changesin
union density,AP (see eq. 3). Since Hirsch and Link employ a three14 Hirsch
(in press)findsproductivity
the1970stobe slowerfor
growth
during
companies
withhighlevelsoffirm
unioncoverage
in 1972,although
muchofthe
slowergrowth
canbe accounted
forbycontrolling
in detailforindustry
effects.
Theunioncoefficient
in a productivity
growth
equationmeasures
onlythedirect
unioneffect,
however;
unionsaffect
growth
indirectly
through
theirimpacton
investments
incapitalandR&D.
99
Union Effects
to includetechnicalcapital,theirestimating
factorCobb-Douglasfunction
equation(ignoringcontrolvariables)is
p = X + 'D(dT/dt)/Q
+ (1 - a)(c - 1)AP,
(12)
whereX is the rate of disembodiedgrowth,ID is aQ/OT (the marginal
into
productof technicalcapital), dT/dt approximatesnet investments
stock T, and (dT/dt)/Q is proxied by R&D intensity.They reporta
on the union changevariable,
negativeratherthana positivecoefficient
to thepredictionof thecollectivevoice model.WhenHirschand
contrary
Linkalso includea levelofunionismvariable,bothP andAP arenegatively
growth.The inferenceis, then,
relatedto productivity
and significantly
(as reflectedin
thatunionismnot only reducestotal factorproductivity
increase."
thesignon AP) but also slows therateof productivity
growth
on productivity
unioneffects
Finally,notethatstudiesestimating
attemptto hold constantchangesin stocksof R&D and physicalcapital.
growthhave
The specificroutesthroughwhichunionsslow productivity
not been examinedthoroughly.To the extentthatunions decreaseinthe
vestmentin tangibleand intangiblecapital,these studiesunderstate
growth.A fullaccounting
negativeimpactofunionismon firmandindustry
not onlyof unions' directeffects
of theunionrole requiresmeasurement
derivingfromtheir
growth,butalso ofanyindirecteffects
on productivity
decisions(Allen 1986c; Hirsch,in press).16
impacton investment
The ratherdivergentconclusionsreachedin the productionfunction
thatunionsraise
doubton thehypothesis
castfurther
and R&D literatures
moreproIf unionizedfirmsand industriesare significantly
productivity.
ductive,ceterisparibus,shouldnottherebe evidenceoffasterproductivity
growthin industriesand firmswhere unionizationhas grownmost or
reducedinnovativeactivity,
declinedleast?Evidenceof lowerinvestment,
growthin moreunionizedsectorsshouldnotonly
and slowerproductivity
engendercaution in evaluatingthe evidencefromproductionfunction
studies,but also shouldgiveriseto concernabout thelong-runimpactof
Perhapsthe long runhas arrived.
unionson economicperformance.
on theHirschandLinkfindings
to placeundueemphasis
15 We arereluctant
thepossible
ofobservations,
number
thelimited
becauseofthelevelofaggregation,
function
production
oftheconventional
andthefragility
ofunionism,
endogeneity
Allen(1986c)
however,
inconstruction,
growth
ofproductivity
test.In ananalysis
toboththeinitiallevelandchange
related
changeinversely
productivity
alsofinds
(1984,pp. 169-71)contendthattherelaFreemanand Medoff
in unionization.
istooweaktodrawanydefinite
andunionism
growth
productivity
between
tionship
conclusions.
impact,
economy-wide
16 It isdifficult
aboutunionism's
conclusions
todrawfirm
frees
fornonresources
amongunionfirms
andgrowth
investment
sincedecreased
however;
norcostless,
instantaneous
isneither
movement
Suchresource
unionfirms.
between
union
differences
profit
suchsignificant
ifitwere,therewouldnotexist.
companies.
andnonunion
100
Addison/Hirsch
V. Conclusions
oftheburgeoning
Ouranalysis
literature
unioneffects
on
investigating
and growthhas exposeda numberof unresolved
productivity,
profits,
issues.Statistical
shortcomings
and considerable
ambiguity
attachto the
production
function
testsadoptedinthebulkoftheproductivity
studies.
Whileempirical
evidence
is mixed,
thecontention
thatunions,
on average,
raiseproductivity
cannotbe sustained.
Severalempirical
significantly
regularities
inthisliterature,
however,
maybeexploited.
First,
positive
union
effects
inindustries
productivity
tendtobemostpronounced
whereunionnonunion
wagedifferentials
are large.Second,positiveeffects
are more
evidentwherecompetitivepressuresare present.And, relatedly,positive
effects
areabsentorquitesmallinthepublicandlesscompetitive
sectorswithorwithout
a unionwagepremium.
thelinkbetween
these
Although
is somewhat
observed
regularities
shaky,sinceunionsareleastlikelyto
inmorecompetitive
obtainlargewagepremiums
thefindings
are
settings,
withbothshockeffect
broadlyconsistent
and selectivity
explanations.
Firmsin competitive
environments
are"shocked"intoproductivity
improvements
inthefaceofwageincreases
anddecreased
profit
expectations.
Andcompetitive
ensurethatin theverylongrun,firms
pressures
must
increase
in orderto survive.
productivity
Thus,anycross-section
sample
willbe nonrepresentative,
of firms
sinceunionizedfirms
notincreasing
theirproductivity
areleastlikelyto survive.
The finding
thatunionsreduceprofitability
further
callsintoquestion
thesizableunionproductivity
differentials
in someof theproreported
ductionfunction
studies.Freemanand Medoff
arguethatunionsreduce
in general
profitability
becausetheirproductivity
effects,
thoughsubstaninsufficient
inwagecostsandgreater
increases
tive,arenevertheless
tooffset
To thisis addedtheriderthatsinceunionprofit
capitalintensity.
effects
are largelyrestricted
to highlyconcentrated
industries
thelonger-term
consequences
ofthistransfer
arebenign.
Ourownreading
ofthisliterature
is thatmanyoftheproductivity
results
aresimply
andprofit
inconsistent;
"true"productivity
effects
aremorelikelytobe closetozeroandnegative
profit
effects
intheempirical
maybe lessseverethansuggested
literature.
The conclusion
thatunionscapture
concentration-related
is based
profits
on a fragile
betweenprofitability,
largely
econometric
conrelationship
and unionismat theindustry
centration,
level.Corroborative
evidence
fromthelabormarket
is lacking.Returns
accruing
fromothercorrelates
ofmarketpower(e.g.,marketshare,foreigncompetition,
and government
andfrom
entry
restrictions)
long-lived
capitalappeartobemoreimportant
sourcesofunionrents.
Themanner
inwhichunionrentseeking
isofsomeimportance.
proceeds
Suchevidenceas we havebeenable to uncoverdoes notencourage
an
viewofunionism's
Unionrentseekoptimistic
longer-term
consequences.
andintangible
ingat theexpenseoflong-lived
tangible
capitalappearsto
UnionEffects
101
in physicalcapital,as well as to decreaseR&D
lower firms'investment
activities.As a consequence,proand otherinnovativeand risk-taking
ductivitygrowthtendsto be slower in unionized firmsand industries.
oppositionto unions,and decliningunioncoverage
Increasedmanagement
withinmostsectorsof theU.S. economy,appear to be
and employment
of highlyunionpoor performance
predictableresponsesto therelatively
ized companiesduringthe 1970s.More workis requiredhere,particularly
of improveddata sets. But alreadythe locus of the
in the construction
debatehas shiftedtowarda considerationof unionism'sdynamiceffects
on long-runeconomicperformance.
References
Abowd,JohnM. "CollectiveBargainingand theDivision of theValue of
DiscussionPaperno. 2137.Cambridge,Mass.: National
theEnterprise."
Bureauof EconomicResearch,January1987.
Acs, ZoltanJ.,and Audretsch,David B. "Innovationin Large and Small
Firms."EconomicsLetters23 (1987): 109-12.
and Exit-Voice."Industrial
Allen,StevenG. "TradeUnions,Absenteeism,
and LaborRelationsReview37 (April 1984): 331-45. (a)
".UnionizedConstructionWorkersAre More Productive."QuarJournalofEconomics99 (May 1984): 51-74. (b)
terly
and Publicly
in Privately
. "The EffectofUnionismon Productivity
Owned Hospitals and NursingHomes." Journalof Labor Research7
(Winter1986): 59-68. (a)
in OfficeBuildingand School
. "Unionizationand Productivity
andLaborRelationsReview39 (January1986):
Industrial
Construction."
187-201.(b)
Change underUnionism."
Levelsand Productivity
. "Productivity
Mimeographed.Raleigh: North Carolina State University,March
1986.(c)
Journal
ofEconomics
. "Can Union LaborEverCost Less?"Quarterly
102 (May 1987): 347-73.
and Labor Unions: An Applicationof
Baldwin,CarlissY. "Productivity
Contracts."JournalofBusiness56 (April
theTheoryof Self-enforcing
1983): 155-85.
Becker,BrianE., and Olson, Craig 0. "Labor Relationsand FirmPerforoftheFirm,editedby
and thePerformance
mance."In HumanResources
M. Kleiner,R. Block,M. Roomkin,and S. Salsburg,pp. 43-85. Madison,
Wis.: IndustrialRelationsResearchAssociation,1987.
Bemmels,Brian. "How Unions AffectProductivityin Manufacturing
Plants."Industrialand LaborRelationsReview40 (January1987): 24153.
Binder,JohnJ."Measuringthe Effectsof Regulationwith Stock Price
Data." RandJournalofEconomics16 (Summer1985): 167-83.
Blau, FrancineD., and Kahn, LawrenceM. "Unionism,Seniority,and
Turnover."IndustrialRelations22 (Fall 1983): 362-73.
in WestVirginiaCoal
Boal, WilliamMarch."Unionismand Productivity
102
Addison/Hirsch
Mining."Paper presentedat the EconometricSocietywintermeeting,
New York,December1985.
Bothwell,JamesL.; Cooley, Thomas F.; and Hall, Thomas E. "A New
View oftheMarketStructure-Performance
Debate."JournalofIndustrial
Economics32 (June1984): 397-417.
Bronars,StephenG., and Deere, Donald R. "The Real and FinancialDecisionsofUnionizedFirmsin a DynamicSetting."Mimeographed.
Santa
Barbara:Universityof California;College Station:Texas A&M, May
1986.
. "Union Threats,Unionization,and Firm Behavior:A Decision
EventStudy."Mimeographed.SantaBarbara:University
of California;
College Station:Texas A&M, September1987.
Brown,Charles,and Medoff,James."Trade Unions in the Production
Process."JournalofPoliticalEconomy86 (June1978): 355-78.
Brown,JamesN., and Ashenfelter,
Orley."TestingtheEfficiency
of EmploymentContracts."
JournalofPoliticalEconomy94, suppl. June1986):
S40-S87.
Card, David. "EfficientContractswith Costly Adjustment:Short-Run
Employment
forAirlineMechanics."AmericanEconomic
Determination
Review76 (December 1986): 1045-71.
Clark, Kim B. "The Impact of Unionizationon Productivity:
A Case
Study."Industrialand Labor RelationsReview 33 (July 1980): 45169. (a)
. "Unionizationand Productivity:
Micro-econometric
Evidence."
Quarterly
JournalofEconomics95 (December 1980): 613-39. (b)
. "Unionization and Firm Performance:The Impact on Profits,
Growth,and Productivity."
AmericanEconomicReview74 (December
1984): 893-919.
Clark,Kim B., and Griliches,Zvi. "Productivity
Growthand R&D at the
BusinessLevel: Resultsfromthe PIMS Data Base." In R&D, Patents,
editedby Zvi Griliches,pp. 393-416. Chicago: Uniand Productivity,
versityof Chicago Press,1984.
Connolly,RobertA.; Hirsch,BarryT.; and Hirschey,Mark."Union Rent
Seeking,IntangibleCapital,and MarketValue of theFirm."Reviewof
Economicsand Statistics
68 (November 1986): 567-77.
Domowitz, Ian; Hubbard,R. Glenn; and Petersen,Bruce C. "The IntertemporalStabilityoftheConcentration-Margins
Relationship."Journal
ofIndustrialEconomics35 (September1986): 13-34.
. "MarketStructure
and CyclicalFluctuationsin U.S. Manufacturing." Mimeographed.Northwestern
University,
January1987.
Frank P. "Do Union MembersReceive
Duncan, Greg J.,and Stafford,
AmericanEconomicReview70 (June
Compensating
Wage Differentials?"
1980): 355-71.
Eberts,RandallW., and Stone,JoeA. "On theContractCurve:A Test of
Alternative
ModelsofCollectiveBargaining."
JournalofLaborEconomics
4 (January1986): 66-81.
"The Effect
ofTeacherUnionson StudentAchievement."
Industrial
and LaborRelationsReview40 (April 1987): 354-63.
Union Effects
103
Ehrenberg,Ronald G.; Sherman,Daniel R.; and Schwarz,Joshua C.
in the Public Sector:A Studyof Municipal
"Unions and Productivity
Libraries."Industrialand Labor RelationsReview 36 (January1983):
199-213.
Forsund,F.; Lovell,C. A. Knox; and Schmidt,Peter."A SurveyofFrontier
MeasureProductionFunctionsand ofTheirRelationshipto Efficiency
13 (May 1980): 5-26.
ment."JournalofEconometrics
Freeman,Richard B. "The Exit-VoiceTradeoffin the Labor Market:
Journalof
Unionism,Job Tenure,Quits, and Separations."Quarterly
Economics94 (June1980): 643-74.
Freeman,RichardB., and Medoff,
JamesL. "Unionism,Price-CostMargins,
and the Returnto Capital." Discussion Paper no. 1164. Cambridge,
Mass.: National Bureauof EconomicResearch,July1983.
. WhatDo UnionsDo? New York: Basic, 1984.
Grout,Paul A. "Investmentand Wages in theAbsenceof BindingCon52 (March 1984):
tracts:A Nash BargainingApproach."Econometrica
449-60.
Growth,and Firm
Hirsch,BarryT. "InnovativeActivity,Productivity
Are Labor Unions a Spuror a Deterrent?"In Advancesin
Performance:
AppliedMicroeconomics,
vol. 5, editedby AlbertN. Link and V. Kerry
Smith.Greenwich,Conn.: JAI Press,in press.
Hirsch,BarryT., and Addison,JohnT. The EconomicAnalysisof Unions:
New Approaches
and Evidence.Boston:Allen & Unwin, 1986.
Hirsch,BarryT., and Connolly,RobertA. "Do UnionsCaptureMonopoly
Profits?"
Industrial
and LaborRelationsReview41 (October 1987): 11836.
and ProHirsch,BarryT., and Link, AlbertN. "Unions, Productivity,
ductivity
Growth."JournalofLaborResearch5 (Winter1984): 29-37.
. "Labor Union Effectson InnovativeActivity."JournalofLabor
Research8 (Fall 1987): 323-32.
Ichniowski,Casey. "The Effectsof GrievanceActivityon Productivity."
Industrialand LaborRelationsReview40 (October 1986): 75-89.
Karier,Thomas. "Unions and Monopoly Profits."Review of Economics
and Statistics
67 (February1985): 34-42.
Katz,HarryC.; Kochan,ThomasA.; and Gobeille,KennethR. "Industrial
and QWL Programs:
RelationsPerformance,
Economic Performance,
An Interplant
Analysis."Industrialand LaborRelationsReview37 (October1983): 3-17.
Kaufman,RobertS., and Kaufman,RogerT. "Union Effectson ProducPersonnelPractices,and Survivalin theAutomotivePartsIndustivity,
try."JournalofLaborResearch8 (Fall 1987): 333-50.
in the United
Kendrick,JohnW., and Grossman,Elliot S. Productivity
States:Trendsand Cycles.Baltimore:JohnsHopkins UniversityPress,
1980.
Klein,Benjamin;Crawford,RobertG.; and Alchian,ArmenA. "Vertical
Integration,
AppropriableRents,and theCompetitiveContractingProcess."JournalofLaw and Economics21 (October 1978): 297-326.
Wage DisLawrence,Colin, and Lawrence,RobertZ. "Manufacturing
104
Addison/Hirsch
persion:An End Game Interpretation."
Brookings
Paperson Economic
Activity,
no. 1 (1985), pp. 47-106.
Job Satisfaction,
and the UnionLeigh, Duane E. "Union Preferences,
Voice Hypothesis."IndustrialRelations25 (Winter1986): 65-71.
Lewis,H. Gregg.UnionRelativeWageEffects:
A Survey.Chicago: Universityof Chicago Press,1986.
Liberty,SusanE., and Zimmerman,
JeroldL. "Labor Union ContractNegotiationsand AccountingChoices." Accounting
Review 61 (October
1986): 692-712.
Liebowitz,S. J."What Do Census Price-CostMarginsMeasure?"Journal
ofLaw and Economics25 (October 1982): 231-46.
Link,AlbertN. "Basic Researchand Productivity
Increasein Manufacturing:AdditionalEvidence."AmericanEconomicReview71 (December
1981): 1111-12.
Linneman,Peter,and Wachter,Michael L. "RisingUnion Premiumsand
the Declining Boundariesamong NoncompetingGroups." American
EconomicReviewPapersand Proceedings
76 (May 1986): 103-8.
Lovell, C. A. Knox; Sickles,Robin C.; and Warren,Ronald S., Jr."The
Effectof Unionizationon Labor Productivity:
Some AdditionalEvidence."JournalofLaborResearch9 (Winter1988): 55-63.
McDonald, Ian M., and Solow, RobertM. "Wage Bargainingand Employment."AmericanEconomicReview71 (December 1981): 896-908.
MaCurdy,ThomasE., and Pencavel,JohnH. "TestingbetweenCompeting
Models of Wage and EmploymentDeterminationin Unionized Markets."JournalofPoliticalEconomy94, suppl.(June1986): S3-S39.
ofUnionsand Strikeson theRateofGrowth
Maki,Dennis R. "The Effects
ofTotal FactorProductivity
in Canada."AppliedEconomics15 (February
1983): 29-41.
Allan B. "The Effectsof Unions on Efficiency
in theResMandelstamm,
identialConstructionIndustry:A Case Study."Industrialand Labor
RelationsReview 18 (July1965): 503-21.
Increasein ManuMansfield,Edwin. "Basic Researchand Productivity
facturing."
AmericanEconomicReview70 (December 1980): 863-73.
Mellow, Wesley. "EmployerSize, Unionism,and Wages." In New Approachesto Labor Unions,edited by JosephD. Reid, Jr.,pp. 253-82.
Suppl. no. 2 to Researchin LaborEconomics,editedby R. G. Ehrenberg.
Greenwich,Conn.: JAI Press,1983.
Mincer,Jacob."Union Effects:Wages,Turnover,and Job Training."In
New Approaches
toLabor Unions,editedbyJosephD. Reid,Jr.,pp. 21752. Suppl. no. 2 to Researchin LaborEconomics,editedby R. G. Ehren1
berg.Greenwich,Conn.: JAI Press,1983.
and Identification
Nerlove,Marc. Estimation
of Cobb-DouglasProduction
Functions.Amsterdam:North-Holland,1965.
Noam, Eli M. "The EffectofUnionizationand Civil Serviceon theSalaries
and Productivity
ofRegulators."In New Approaches
toLaborEconomics,
editedby JosephD. Reid,Jr.,pp. 157-70. Suppl. no. 2 to Researchin
Labor Economics,editedby R. G. Ehrenberg.Greenwich,Conn.: JAI
Press,1983.
Union Effects
105
Contractsare on theLabor Demand Curve:
Oswald, AndrewJ."Efficient
Theoryand Facts." WorkingPaper no. 178. Princeton,N.J.:Princeton
IndustrialRelationsSection,July1984.
University,
at theLine ofBusiness
Relationships
David J."Structure-Profit
Ravenscraft,
and IndustryLevel." Reviewof Economicsand Statistics65 (February
1983): 22-31.
Reynolds,Morgan0. "Trade Unions in the ProductionProcessReconsidered."JournalofPoliticalEconomy94 (April 1986): 443-47.
Rosen, Sherwin."Trade Union Power,ThreatEffectsand the Extentof
Organization."ReviewofEconomicStudies36 (April 1969): 185-96.
MartinB. "Unionizationand ProfRuback,RichardS., and Zimmerman,
ofPoliticalEconomy
EvidencefromtheCapitalMarket."Journal
itability:
92 (December 1984): 1134-57.
Salinger,Michael A. "Tobin's q, Unionization,and the ConcentrationProfitsRelationship."Rand Journalof Economics15 (Summer 1984):
159-70.
Sloan,FrankA., and Adamache,KillardW. "The Role of Unionsin HospitalCost Inflation."Industrialand LaborRelationsReview37 (January
1984): 252-62.
Sveikauskas,CatherineDefina,and Sveikauskas,Leo. "IndustryCharacEconomicJournal48 (JanGrowth."Southern
and Productivity
teristics
uary1982): 769-74.
Svejnar,Jan."BargainingPower,Fear of Disagreement,and Wage Settlements:Theory and Evidence From U.S. Industry."Econometrica54
(September1986): 1055-78.
Tauman, Y., and Weiss, Y. "Labor Unions and the Adoption of New
Technology."JournalofLaborEconomics5 (October 1987): 477-501.
GrowthofIndusofR&D on theProductivity
Terleckyj,Nestor E. Effects
Study.Washington,D.C.: National PlanningAstries:An Exploratory
sociation,1974.
Voos, Paula B., and Mishel,LawrenceR. "The Union Impacton Profits:
EvidencefromIndustryPrice-CostMarginData." JournalofLaborEconomics4 (January1986): 105-33.(a)
Industry."Re"The Union Impacton Profitsin theSupermarket
68 (August 1986): 513-17. (b)
view ofEconomicsand Statistics
andProductivity:
ofUnionson Employment
Wessels,WalterJ."The Effects
An UnresolvedContradiction."JournalofLabor Economics3 (January
1985): 101-8.