Public Disclosure Authorized
Public Disclosure Authorized
POLICY RESEARCH WORKING PAPER
The Soviet Economic
Decline
Stanley Fischer
ledto therelatKaSoviet
'WVhat
on
declinewas reliance
anda
capitalaccumulation
of substitution
low elasticity
betweencapitalandlabor.
are
economies
Planned
at
lesssuccessful
apparently
laboreffortwith
replacing
capital.Tentativ"evidence
thatthe burdenof
indicates
defensespendingalso
to theSoviet
contributed
debacle.
Public Disclosure Authorized
Public Disclosure Authorized
Historical and Republican Data
1284
The WorldBank
PolicyResearchDepartnent
and Growh Division
Macroeconomics
AprilI1994
POLICYRESEARCHWORKINGPAPER1284
Summary findings
Sovietgrowth for 1960-89 was the worst in the world,
after controllingfor investmentand human capital.And
relativeperformanceworsensover time.
Easterlyand Fischer explain the decliningSoviet
growth rate from 1950 to 1987 by the declining
marginalproduct of capital.The rate of total factor
productivitygrowth is roughly constant over that period.
Althoughthe Sovietslowdownhas conventionallybeen
attributed to extensivegrowth (risingcapital-to-output
ratios), extensivegrowth is also a feature of marketoriented economieslike Japan and Korea. One message
from Easterly'sand Fischer's resultscould be that Sovietstylestagnationawaitsother countriesthat have relied on
extensivegrowth. The Sovietexperiencecan be read as a
particularlyextreme dramatizationof the long-run
consequencesof extensivegrowth.
What led to the relativeSovietdeclinewas a low
elasticityof substitutionbetween capital and iabor, which
causeddiminrishing
returns to capital to be especially
acute. (The natural question to ask is why Sovietcapitallabor substitutionwas more difficult than in Western
market economies,and whether this difficultywas
related to the Soviets'plannedeconomicsystem.)
Tentative evidenceindicatesthat the burdeniut defelns
spendingalso contributedto rhe Sovietdebacle.
Differencesin growth performancebetweerntile SuvOe
republicsare explainedby the same factorsthat tigure in
the empiricalcross-sectiongrowth iiterature:i, .al
income,human capital populationgrowth, anOthe
degree of sectoraldistortions.The resultsEasteilyand
Fischergot with the SovietUnion in the internaiurial
cross-sectiongrowth regressionindicatethat the planneui
economicsystemitselfwas disastrousfor long-iun
economicgrowth in the SovietUnion.
This point may now seem obviousbut was not so
apparent in the halcyondays of the 1950s,wliei,the
Sovietcase was often cited as support for the neuLiassical
model'sprediction that distortionsdo not havL5Leadystate growth effects.Sincea heavydegree of piloining
and governmentinterventionexistsin many countries,
e, eciallvdevelopingcountries,the ill fated Soviet
experiencecontinuesto be of interest.
This paper - a product of the Macroeconomicsand Growth Division,PolicyResearchDepartment- is part of a larger
effortin the departmentto studythe determinantsof long-rungrowth.Copiesof the paper are availablefreefrom the Worid
Bank,1818 H Street NW, Washington,DC 20433. PleasecontactRebeccaMartin, room NI 1-043,extension 3 1320(56
pages).April 1994.
ThePolicyResearch
WorkingPaperSeries
disseminates
the findingsof workin progress
to encourage
theexchange
of ideasabout
development
issues.
Anobjective
of theseries
is to getthefindingsout quickly,evenif thepresentations
arelessthanfully polished.
The
arethc
of theauthorsandshouldbeusedandcitedaccordingly.
Thefindings,interpretations,
andconclusions
papers
carrythenames
Boardof Directors,
or anyof itsmember
countries.
authors'ownandshouldnot beattributedto theWorldBank,itsExecutive
Producedby the PolicyResearchDisseminationCenter
THE SOVIET ECONOMIC DECLINE: HISTORICAL AND REPUBLICAN DATA'
William Easterly
WorldBank, Room N-i 1-043
1818H Street NW
WashingtonDC 20433
and
StanleyFischer
Departmentof Economics
MIT
Cambridge MA 02139
Easterlyis inthe Macroeconomics
andGrowthDivisionattheWorldBank,and Fischerin theDepartment
of
Economics
at MIT,and a ResearchAssociateof theNBER.Viewsexpressedherearenotto be attributedto the
WorldBank.Wearegratefulto KarenBrooks,AlanGelb,EugeneGravilenkov,
BarryIckes,BarryKostinsky,
Marthade Melo,GurOfer,SergioRebelo,BryanRoberts,MarcRubin,RandiRyternan,MartinSchrenk,Marcelo
Selowsky,MichaelWalton,andAlwynYoungforcommentsandusefuldiscussions
whilewritingthisdraft,to
seminarparticipants
at MIT,theUniversity
of Michigan,theFederalReserveBoard,andtheWorldBank,to
ProfessorMarkSchafferof theLondonSchoolof Economics
forkindlymakinghishistoricalWestern,official,and
Khanindataavailablein machinereadableforn, andto ElanaGoldandMaryHallwardfordiligentresearch
assistance.
2
Whilethe finalcollapse of the Soviet Unionand Sovietcommunismnow appeanto have been
inevitable, It is esendal to try to pinpoint the causes of the economicdecline, withoutwhich the Soviet
Union would still exist. The differentaccountsof the causes of decliningSovieteconomicgrowth
developedby Bergson(1387b), Desai (1987), Ofer (1987), Weitzman(1970)and others emphasize:the
Soviet reliance on extensivegrowth which, giventhe slow growth of the labor force and the failing
marginalproductivityof capital, eventuallyran out of payoff; the decliningrate of productivitygrowth
or technicalprogress associatedwith the difTicultiesof adoptingand adaptingto the sophisticated
technologicsbeing introducedin the West (includingEast Asia as part of the West);the defense
burden; and a variety of special factorsrelatingto the absenceof appropriateincentivesin the Soviet
system, includingcorruption and demoralization.
In this paper, we first place the Sovietgrowth performance in an internationalcontext using the
empirical cross-sectiongrowth literature. In section 1, we start with an overviewof the data and of the
Sovietgrowth record, comparingit to other countriesusing a standarcdgrowth regression. We compare
the Sovietpattern of extensivegrowth (rising capital to output ratiom)to other countries. We reexamine
and update Soviet-levelproductionfunctionestimatesbased on the official,unofficial(by the Russian
economistKhanin)l and western data, as well as examiningother historicalindicatorsof the Soviet
growth pattern. In section 11,we turn to a new data set: officialdata on republicanoutput, capital
stocks (both in constant, or as the Sovietscalled it, comparableprices), and employmentby sector.
These data have not previouslyfigured in the Western literature on Sovietgrowth2 . The fact that the
republics will now have to operateas independenteconomicunits adds interestto our republican-level
1Ericson(1990)arguesthatthe Khanirn
dataarepreferableto the westerndatacreatedby Bergsonandothers;
Bergson(199la, 1987a)criticizestheKhanindatafora poorlydocumented
methodology
andtheapparentuseof
unweightedaveragesof physicalindicators.
Harrison(1993)providesa moresympathetic
analysisof Khanin'sdata,
emphasizing
hisattemptsto adjustforthe reportingbiasesinherentin theSovietstatisticalsystem.Wearegratefulto
ProfessorMarkSchafferformakingtheKhanindataavailable.
2Therepublicandatawereprovidedby Goskomstat
of theCommonwealth
of Independent
StatesandbytheCenter
forEconomicAnalysisand Forecasting
attheMinistryofEconomics,
andare availablefromEasterlyattheWorld
Bank,1818H StreetNW,Washington,
DC20433, Werescaledthedatato reconciledifferentbaseyearsforthe data
incomparebleprices.
3
reso.
We disus the patternsof growthby sectorand republic,exploringcrou-sectioncorrelations
beow growthof the Sovietrepublicsor sectorsandconventionalright-handsidevariablesusedin
growthregressions.In the conclusion,we offersomethoughtson interpretation
of our results.
1. Ihe Soviet Growth Record
lhe ftndamentalproblemin evaluatingSovietgrowthis dataqualir.3 Ofcial Sovietoutput
dataoverstategrowth,as a resultof bothmethodological
problns - particularlyin deflatingnominl
d"a, andinentivesto mis eportoutputwithinthe Sovietsystem. Westernanalysisof Sovietgrowth
relieson the elsic studiesof Bergson(1961),andother., as wellas theCIA,whichmakesthe
workingassumptionthat physicalquantitiesas presentedin theotlicialdatawerenotsystematically
misreported.Thusthe differencebetweenthe westernestimatethatper capitaSovietGNPincreased
between1928and1987by 3.0 percentper annum(4.3 percentfor aggregateGNP)andthe official
estimatefor NMPper capitaof 6 percentper snnumresultsmostlyfrompricingcorrections,andalso
fromdifferencesm the coverageof NMPandGNP.4 The classicwesternestimatesgenerallyassume
thatSovietinvetment nd capitaldataare moreaccuratethanoutputdata(Bergson(1987a),a view
summarized
in Ofer
disputedby Wiles(1982)).The westerndatathrough1985are conveniently
(1987).
Weuse fourdifferentdatasets in theempiricalworkin thispaper:
(1)
dataon real output,industrialproduction,employment,andthe
TheofficialSovietUnion-wide
capitalstockin thematerialsectorin 1973rubles,takenfromofficialsources;
(2)
Westerndata on output,industrialproduction,employment,andthecapitalstock,for the
SovietUnionas a whole:including
(a) the Powell(1963)/CIA(1982)/CIA
(variousyears)serieson valueaddedandcapitalstocks
in indunry,and
3 T1his
discussiondrawson materialin Fischer(1992).
4 The Sovietconceptof
Net MaterialProductomittedfrom GNP servicesnot directlyrelatedto production,such as
pasenger transportation,housing,and the outputof governmentemployeesnot producingmaterialoutput.
4
(various years)/Kellogg(1989;
(b) the Moorsteen-Powell(1966)/Powell(1968)/CIA(1982)/CIA
series on GNP, labor input, and capital stock for the entire economy. These series are chain linked,
in 1970rubles for 1960-80,and 1982rubles for the 198ts.
using 1937 rubles for 19289-60,
(3)
Khanin's data, from Khanin(1988), also at SovietUnion-widelevel, for output, employment
and the capital stock in the material sector; and
(4)
Republic-leveldata on aggregateand sectoral output and inputs in the materialsectors for
1970-90in constantrubles, which were made availableby (ioskomstat.
The direct source of our datasets(1) through (3) is Gomulkaand Schaffer(1991), who spliced
together series from the sources described. Note that the Khanindata are presentedfor the material
sectors (i.e. not includingconsumerservices), as are the officialdata. Our preferred dataset for the
aggregatedata will be (2); the others are presentedto test the robustnessof the conclusionsto
alternativeestimatesof outputsand inputs.
COMPARSON
A) SOURT GROWTHIN INTERMATIONAL
Growth rates of series in the first three data sets for different periods are presented in Table 1.
The Western outputper worker growth rates are well below the officialrates, with the Khanindata in
turn below the Westem data. All series show growthdecliningsharply since the 1950s.
How does the Sovietgrowth record compare to the rest of the world? We use the Western GDP
series to compare Sovietper capita growth over 1960-89with World Bank per capita growth rates for
102 countries(we look here at per capita ratinerthan per worker growth to enlarge the sampleof
comparatorsand make it consistentwith the cross-sectiongrowth literature). The first column of Table
2 shows that Sovietper capita growth has been slightlyabove the global average over both 1960-89and
1974-89.
However, Sovietgrowth no longer looks respectableonce we control for the standardgrowth
detemiinantsfrom the empirical literature. The last column of Table 2 shows the residual from
insertingthe SovietUnion into the core regressionof Levine and Renelt (1992), which relates growth
to initial income, populationgrowth, secondaryenrollment,and the investmentratio to GDP. The
Levine-Reneltregressionincludingthe Soviet Unionis as follows:
S
Percaphagrowth60-89--0.83+ 17.49Inwstment60-89-. 35 GDPpercavita1960+
(85) (2.68)
(14)
3.16 * Secondary
enrollment1960-.38 Populationgrowth6089 - 2.34DummyforUSSR
(1.29)
(22)
(1.43)
103observatioas,
R2-. 46.(standarderrorsin parentheses)
Except for populationgrowth,Levineand Renelt showedthese variablesto be robust to alternative
specificationsin growthregressions(althoughconcernsabout endogeneityremain). The regression
resultsare identicalto the Levine-Reneltoriginalsince we are dummyingout the Sovietobservation.
Exceptinginitial income. the values of the Sovietright-handside variablesshould have implied
very rapid growth-populationgrowth was low, and secondaryeducationand the investmentratio were
near the top of the distribution. Growth was only average, hence the large negativeresidualof 2.3
percentagepoints in 1960-89.It is notable that the only countrieswith worse residualsare generally
both small and poor: Surinaine,Jamaica, Guinea-Bissau,Liberia, Zambia, and Peru. Sovietper capita
income in 1989was only half of what it would have been if the average relationshipbetween growth
and the right-handside variableshad held over 1960-89.
The Soviet residual in this OLS regression is not actuallysignificantin a two-sidedtest at the 5
percent level. However, the presence of so many small and poor countriesamongthe large outliers
makes us suspectheteroskedasticity. The suspicionis justified. We split the 1960-89sampleinto thirds
on the basis of total real GDP (i.e. populationtimes PPP per capita income)and rerun the above
regression for the top and bottomthirds ranked by total GDP. (The USSR is includedin the top third
ranked by total GDP and we continueto dummnyit out.) The Goldfeld-Quandttest statisticfor
heteroskedasticity- which is equal to the ratio of the sum of squared residualsin these two subsample
regressionsand is distributedas an F statisticwith the numberof degreesof freedomof the numerator
and denominatorcorrespondingto the degrees of freedom in the subsampleregressions- indicatesthat
we can reject homoscedasticity.The test results are as follows:
Sum of squaredresidualsin third of sample with lowestreal GDP:88.3
Sum of squaredresidualsin third of sample withhighestreal GDP: 33.9
F (29, 28) = 2.61 (significantat 1 percent level)
6
3aod on the test results, we now perform weightedleast square usingthe lo of total real GDP u the
weightingseries. The results are now as follows:
Per capita growth 60-89- -0.43 + 15.93 Investment60-89 -. 28 GDPper capita 1960 +
( 73) (2.19)
(.08)
2.56 * Secondaryenrollment1960- .24 Populationg&owth60-89- 2.28Dummyfor USSR
(0.73)
(.16)
;3.48)
102 observations,R2 (weighted)-. 84. R2 (unweighted)=-41.
The Sovietdummnybecomeshighlysignificantwith weightedleast squares, with a t-statisticof
4.8. Taking into account that only countriesdoing worst than the USSRwere small economiesmakes
the Sovietperformance look even worse. After correctingfor heteroskedasticity,the Sovieteconomnic
performanceconditionalon investmentand human capitalaccumulationwas the worst in the world
over 1960-89.
How does the comparativeSovietperformanceevolve over time? Sincethe World Bank data
used by Levine and Renelt begins only in 1960, we comparethe Sovietperformancealso with the
cross-countrySummers-Heston(1991)dataset that extendsback to ' )50. We perform a pooledtimeseries, cross-sectionregressionusing decade averages for the same specificationas before (except that
we have to unfortunatelyomit the secondaryenro!lmentvariable for lack of reliable Sovietdata for the
1950s).We use the same Sovietdata as in the previous regression, but now broken down by decade.
We put interceptdummiesfor each decade, as well as a separate Sovietdummy for each decade. We
continueto use weightedleast squareswith the weightingseries being the log of total GDP, as the
Goldfeld-Quandtstatisticstill indicatesa significantlylarger variance for small economies.5 The results
are:
5 TheF-statisticfor the ratio
of the sum of squaredresidualsin the bottomthird to that in the top third of the sample
rankedbytotalGDP(in PPPpricesfromSummers-Heston
(1991))isF(124,121)=2.03,
whichis significantat the I
percentlevel.
7
bydck
Pr. cpia roih bydwca - 0 022+.120InvestmenGDP
(005) (016)
- I 5E06GDPp capiet,initialye.y
(3.6E-07)
.626Popidauongrowh
bydwde + .00560sdummy.005 70sdummy-.015
80 dummy
(003)
(143)
(.004)
(.O33)
USSR
60s- 017Dummyfor
USSR
70s
+ .024Dummyfor
USSR
50s- 008Dummyfor
(009)
(010)
(01))
-.
023 DumyYfor USSR
80J
(009)
-. 26
391obueruao.,R2(wighte) -. 54,R2(unweighied)
As is well known, world economicgrowth deceleratedin the 70s and even more in the 80s. However
the Sovietgrowth decelerationis notableeven by comparisonwith the worldpattemn:Scvieteconomic
mrowth
was significantlyabovethe world average in the 1950s,and significantlybeloweven the poor
world growth of the 1980s. Note especiallythe good performanceof the Soviet Union in the 1950s,
even controllingfor high investment:it suggeststhat whateverthe weaknessesof Sovietcentral
planningin hindsight,these weaknesseswere unlikelyto have been apparentprior to 1960.
5) POSSIBLEEXPLANATIONS
FORPOORANDDECNING SOVIETGROWTH
We now considerother possible factors in the relativeSovietdecline, includingthe defense
burden, demoralization,and Sovietdisincentivesfor innovation.Could the poor and declininggrowth
performancebe explainedby the burden of defenseon the Sovieteconomy?Althoughmeasurementis
problematic,the burden seems to have been high and rising. In Table 3, we show some estimatesof the
Sovietdefenseburden as a share of GDP. Over the entire periodsince 1928,Soviet defensespending
has risen from 2 percent of GDP to the much higher levelsof the mid-andlate-1980s,of around 15-16
percent of GDP. Over the period 1960-89in which the Soviet growthdecline occurred, the rise in the
defense burden is more modest- from 10-13%in 1960 to 12-169%in the 1980s.
The internationalevidence for adverseeffectsof defensespending on growth is ambiguoussee Landau (1993)for a recent survey. Landau(1993)himself finds an inverted U relationship:military
spending below 9 percent of GDP has a positiveeffect on growth, but defenseabove9 percent of GDP
has a negativeeffect on growth. To see whetherthis affects the Sovietdummy in the growth
regressions,we insert defense spendinginto the decade-averagegrowth regressionsperformedearlier.
We alsoincludea variablemeaurin war casualtiesper capitaon nationaltorritoryto inswuthat the
datais only
militaryspendingvariableis notsimplyproxyingfor wars.Becausethemilitaryspending
availablefor recentperiods,we use datafromthe 1980.only.6 The regressionincludinga quadratic
functionof militaryspendingis as follows:
GDPpercapita1980+
1980-882.7E-064
Percapitagrowth198C-88- -0.003+ i 27 InvwtmeP/GDP
(. IE-06)
(017) (038)
enrollment1970+ .0081MilItory spending/GDP
growth,1980-88 + .007Secondary
-1.34 Population
(.0024)
(017)
(38)
2 -0.746Warcaswaltipercapita-.0155Dummyffor
USSR
-.00041(Miltarysp.ndlng/GDP)
(.0268)
(0.343)
(0001)
ecorsin
-. 30 Standard
-. 59,R2(unrnwighted)
R2(weighted)
bylogoftotalGDP.77observation,,
IS wigshted
penthes.
We confirm Landau's resultof an inv'ted U-shapedrelationshipbetween growth and defense
spending.Military spenduig reducesthe magnitudeand significanceof the Sovie;dummy. However, as
Landaualso noted '-is result is not very robust - omittingSyria and Israel from our sample eliminates
the significanceof military spending. The defenseexplanationfor the Sov.et declineis plausiblebut
not firmly establishedwith cross-sectiondata. We will test the defensehypothesisfurtherwith the
Soviettime series in the productionfunctionestimatesbelow.
Another hypothesisaboutthe Sovietgrowth declineis that it was related to the increasing
demoralizatioi.of the population,or alternativelyto the increasingbreakdownof worker discipline.
This breakdown of disciplinecould have resultedfrom the gradual openingup of the Sovietsystern,
and the decliningreliance on state terror.
Demoralizationis obviouslyhard to measure, but we present some fragmentsof evidence.One
statisticrelevant to demoralizationis shown in Figure 1, which representsthe results of a survey of
emigreswhich asked how satisfiedthey had been with the standardof living in the USSR. The young
had been less suisfied than the old. Amongthe many possibleexplanationsfor these results is that
6 Landauonlycoven developing
countries,soweuseinsteaddat
fromHewitt(1993)dut coversallcountries
Intemational
(includingtheUSSRitself).ThedataforbothLandauand Hewittis mainlyfromSIPRI(theStockholm
is fromEaterly,Kremer,Pritchett,andSummers(1993).
PeaceResarch Institute).Thedataon warcasualties
9
declininggrowth and disappointedexpectationsamongthe young were mutuallyreirfbrcinr.7
Other indicatorsof life in the Soviet Unionalso support the idea of a system breakingdown.
Westem specialistswere amazedto leam that Sovietmale life expectancyactuallydeclinedin the 1970s
while other countries' (male) life expectancyrates were rising(Fiiure 2). Sovietlife expectancywu
decliningeven though per capita incomegrowth was slightlyabove the world average, as we have
seen. There was a recovery in Sovietlife expectancyin the 80s, but the USSRwas stUisupassed
during the decade by developingcountrieslike Mexico.'
Another possibleexplanationfor poor and decliningSovietgrowth couldbe adverse incendves
under central planningfor technologicalinnovation(Berliner(1976)). A recent theoreticaland
empirical literature argues that endogenoustechnologicalinnovation,as measuredby resourcesdevoted
to research and development(R&D), significantlyexplainsreladve growthperformanceacross
countries(Coe and Helpman(1993), Lichtenburg(1992), Romer (1989); see Birdsalland Rhee (1993)
for a dissentingview).
Westem estimatesof the Sovietresearch effort, presented in Figure 3, show R&D spending
rising as a share of GNP. The R&D share is abovethe 2-3 percent in the leading industrialized
countries. In 1967, about 1.5 percentagepoints of this was estimatedto be for defenseand space
(Bergson, 1983). The share of defenseand space R&D in total R&D is believedto have fallen over
1959-84(Acland-Hood(1987)), implyingan even steeper rise in civilianR&D. The data on Soviet
R&D thus go in the wrong directionto explain either poor Sovietgrowth on average or the fall of
Sovietgrowth over timne.
7 Amongthe otherexplanations:
theoldremembertheperiodof muchlower
theyoungarechroniccomplainers;
consumption
beforetherapidSovietgrowthof the 1950s;theauthoritiesresistedemigrationby theyoung,sothat
anyyoungemigrehadto be moredeterminedanddisgruntledthantheaverageemigre.Al>, sincetheoriginal
sourcedid notreportstandarddeviationswithinthesamplegroups,we areunsurewhetherthedifferencesare
statist,callysignificant.
'One factorcouldhavebeenthesharplyrisingconsumpdon
of alcoholinthe 60s,70s,and early1980s,whichitself
(TremI1991).However,wearereluctantto maketoo muchof
maybe an independent
indicatorof demoralization
thissincesomecountrieswithrapidincomeincreases- likeKorea- ilso hadshaply risingalcoholconsumption.
10
C) THEEA7ENSIVE
GROWrHHYPOTHESIS
As noted, in the introduction,the conventionalhypothesisfor the Sovietgrowth decline is the
pattem of extensivegrowth, definedby Ofer (1987) as a rising ca; ital-outputratio, Figure 4 showsthe
evolutionof the capital-outputratios impliedby the alternativedata series for 1950-87.' The wstern
series shows the capital-outputratio increasingtwo and &half times between 1950and 1987. The
official series also rises steadilybeginningat the end of the S0s, more thar.doublingtetween 1958ad
1987. The Khanindata, by contrastwith the other two series, show only a small increase in the
capital-outputratio betweenthe early 1950sand 1987. The capital-outputratiosin industry first
decline in the S0s and then rise sharply after 1960, accordingto both Westernand officialestimates.
The behaviorof the capital-outputratio is central to the debate about whetherreliance on
extensive growth was the Achilles' Heelof Sovietindustrialization,as the conventionalwisdomhas it.
In the neoclassicalmodel, a rising capital-outputratio impliescapital deepeningduring the transition to
a higher steady state, but this capital deepeningwill sooner or later run intodininishing returns that
will cause growth to slow or stop. The Sovietrelianceon capital deepeningis implicitlycontrastedwith
market economies,where according to the famous Kaldor stylizedfact, capital-outputratios remain
relativelystable (recentlyreaffirmedby Rorrer (1990)). A constantcapital-outputratio is consistent
with neoclassicalsteadystate growth with labor-augmentingtechnical change. King and Rebelo (1993)
argue that capital deepeningcannotaccount for much of sustainedeconomicgrowth in the neoclassical
model withoutimplyingimnplausibly
high rates of return to capital early in the transitionprocess.
Nevertheless,recent researchon capital accumulationin market economiescasts doubt on
whether the Sovietextensivegrowth experiencewas unique. Appendix 1 lists the per annumgrowth
rates of the capital-outputratios in a selectionof recent growth accountingstudiesand a few older ones.
All studies agree that the capital-outputratio in the U.S. has remainedremarkablyconstant, which
9 We beginthe graphsin 1950becausewewrepuzzledbythe extremevolatilityof allof the capital-output
series
before 1950.We concludethatmore eventhan the usual cautionshouldbe attachedto resultsthat rely on pre-1950
data.
11
perhaps accountsfor the conventionalwisdomthat Kaldor's stylizedfact holds. However,a numberof
recent studies point to capital-outputratios rising at Soviet-stylerates in Jpn and in some of the E'st
Asian NICs such as Korea (Young(1993b),Kim and Lau (1993), King and Levine (1994),Benhabib
and Spiegel(1992), Nehru and Dhareshwark,.993)).
Moreover, the latter three, cross-country,studiesshow that rising capital-outputratios are a
feature of growth for many countries.10 The three studies computecapital stocksfor a large sampleof
countries,using a variety of data sources(Summers-Hestonversus World Bank)and a variety of
assumptionsabout initialcapital stocks and depreciationrates. The three concur that rising capitaloutput ratios are by no means rare: the median capital-outputratio growth of their respectivesamplesis
around 1 percent per annum, and fully a quarter of the samples' capital-outputratio growth rates are
over 1.7 percent per annum.11 Nor is it only developingcountriesthat are shown to have rapid
capital-deepening. For example,the studiesconcur that capital-outputratios in Austria and France
increased at over 1.5 percent per annum. The literature on extensivegrowth as the bane of Soviet
developmentdid not recognizethat extensivegrowthalso occurred in market economies,and
sometimeswith strikingresults as in Japan and Korea. What is notable aboutthe Sovietexperiencewas
not the extensive growth, but the low payoff to the extensivegrowth.
As either a cause or a consequenceof the low payoff, the level of the Sovietcapital-output(KY) ratio had become extreme by the 1980s.The K-Y ratio as measuredby the Westem GDP and total
capitalstock series was 4.9 in 1985, which is higher than any of the 1985K-Y ratiosin the BenhabibSpiegeland King-Levineexercises. In the Nehru-Dhareshwarsample, there are only four countries
with a K-Y ratio above the Sovietsin 1985, none of which seem especiallyrelevantas comparatorsGuyana, Zambia, Jamaica, and Mozambique.
One other implicationof the extensivegrowth model is that investmentratios have to rise over
withincome.
ratiorisingsystematically
I SeealsoJudson(1994),whoshowsthecapital-output
data(BenhabibandSpiegel1993andKingand Levine1994),we
IIForthetwostudiesthatuse Summers-Heston
extreme(bothhighand low)inthe
to GDPratiosareimplausibly
omitAfricafromthesamplebecauseinvestment
1950s.
12
time if growth is to be maintainedwhile the capit%loutputratio rises. As has previouslybeen
highlightedin the literature(see Ofer (1987))the Sovietinvestmentshare doubledbetween 1950and
1975. as can be seen in the CIA estimatespresentedin Figure 5. After 1975,the investmentshare
continuedto increase, but more slowly.
How unusual is the doublingof the investmentrate over a 25 year period? In the Summers
and Heston (1991) internationaldatabasefor 1950-75,8 out of 52 countries- most notablyJapan and
Taiwan - had a doublingor more of investmentrates.12 Shiftingthe sampleperiod forwardby 10
years to expand the sample,6 out of 72 countrieshad a more than doublingof investmentover 196085, amongwhich Korea and Singaporeare of particularinterest. Sovietinvestmentmobilizadonwa at
a level that was above average, but not unknown,amongmarket economies.
The stand-byof Soviet industrialization,machineryand equipmentinvestment,also increased
sharply as growthdeclined. The importanceof this sector to growth has been emphasizedby de Long
and Sununers(1991, 1992, 1993); the Sovietdata suggesta high ratio of machineryinvestmentto GNP
is not sufficientto generate growth.
GROWTH
D) PRODUCTIONFUNCTIONSAND EXTENSIVE
Another way to evaluatethe extensivegrowthhypothesisis to do the traditionaltotal factor
productivitycalculation. For the TFP calculation,there is little differencebetweenthe officialand
westem data on factor inputgrowth whileKhanin showssubstantiallylower rates of growth of capital
(Table 1). This is a consequencein part of Khanin's view that hidden inflationis as serious in capital
goods industriesas in consumergoods, a view shared by the "Britishschool" of Hanson (19U), Nove
(1981), and Wiles(1982).
In Table 4 we show summarystatisics for productivitygrowth for the USSR, ;lculated
assuminga Cobb-Douglasproductionfunctionwith labor's share equal to .6 and the share of capital
equal to .4 (slightlyabove that used by Bergsonand the CIA, but within the conventionalrange for
12
We contiue to excludeAfiica.-countriesfromthisandthe following
sample.
13
developingcountries)for alternativedata series.13 With the assumptionof Cobb-Douglasproduction
(unit elasticityof substitutionbetweencapital and labor) we see a strongly decliningtrend in TFP
growth after the 1950's.
The most interestingaspect of Table 4 is that the 1950sonce more stand out as an exceptional
period in Sovietgrowth. It is especiallystrikingthat even the westerndata for the industrialsector
Lnplyproductivitygrowth in that decade of more than 6 percent per annum. Note the remarkable
divergencesof views about perfornance in the 1930sthat emerge from Table 4, with officialSoviet
data showingextraordinaryrates of productivitygrowth and Khaninand westernGNP data imnplying
negativerates.
Westem GNP data present the most pessimisticassessmentof Sovietproductivityperformance,
implyingthat productivitygrowth in the Soviet Unionwas positiveonly in the 1950s. The Khanin data,
which uniformilyexhibit lower overall growth than western GNP data, by contrast imply positive
post-1950productivitygrowth, a result of the lower rates of growth of capital in the Khanin series.
The data in Table 4 point to one extremelyimportantfeature of the Sovietgrowth slowdown:
estimatedproductivitygrowth for the industrialsector was positiveuntil the 1980s. This locatesthe
major slowdownof productivityin the non-industrialsector. Lookingahead to Table 7, usingthe
aggregateof the republicandata for 1970-90,the biggestproblem was in agriculture,where
productivityappearsto have declinedby 4 percent per annum, with constructionand trade and
procurementshowingsmall positiverates of productivitygrowth.14
Followingthe pioneeringwork of Weitzman(1970, 1983)and later zontributions(Desai (1976,
literaturethatcapitalsharesarehigherin developingthan
13t haslongbeena stylizedfactin thedevelopment
(1992)estimatethatthecapitalshareis between.4 and.55for
developedcountries(seeforexampleDeGregorio's
LatinAmerica).Westemestimatesof Sovietpercapitaincomesuggestit wasa developingratherthana developed
country.
14 Previousestimatesof productivitygrowthin agriculturewerelessdrasticbutstillshowedpoorperformance.
growthin agricultureof 0.2percentover 1971-79.Brooks
Diamond,Bettis,andRamson(1983)showproductivity
1960-80.Wearenotsurehowourcalculationofnegative
over
growth
productivity
agricultural
zero
show
-d
(1983)
grainyieldswererising
growthrelatesto Desai's(1992)evidencethatweather-adjusted
productivity
agricultural
rapidlyin the 1980s,unlessthe increasein yieldswasobtainedthroughmassiveincreasesin inputs.
14
1987),Bergson (1979), and others), we also investigatewhetherCES functionsprovide a boeer
representationof the data than the Cobb-Douglasproductionfunctioninposed in calculatingthe TFP
estimatesin Table 4.
Weitzman's basic findingwas that a CES producdonfunctionwith a low eluticity of
substitutionof 0.4 fit the data better than the Cobb-Dougla, and that the hypotheis that the ebsticity
of subsdtutionwas one could be rejected. Bergson(1983) cridcized this result, on the groundsthat it
impliedimplausiblyhigh estimatesof the marginalproductof capital in earlier yea.
Desai (1987)
concurred with Weitzman's findingfor aggregateindustry,but claimedthat Cobb-Douglaswas an
adequaterepresentadonfor some branchesof industry.
Estimationof productionfunctionsin industrialcountriesis the subject of a large literaure.
The usual method is to estimateparametersof factor demandsderived from the cost funcdon, the dual
of the productionfunction(see Jorgenson(1983) for a survey). This is obviouslyinappropriatefor a
non-marketeconomy like the SovietUnion. Direct estimadonof productionfuntions is usually
thoughtto be taintedby endogeneityof the factor supplies, particularlycapital; we believethis would
be much less of a problem in the non-marketsystem of the USSR. Table 5 showselasticitiesof
substitutionestirnatedby nonlinearleast squares (see Appendix2 for the regressions),and recalcubted
TFP growth rates for 1950-87(assumingHicks-neutraltechnicalprogress) for subperiodswith the CBS
form:
ln(Y/L) = cl *Time*D5059+ c2*Time*D6069+ c3*Time*D7079+ c4*Time*D8087
+ c5* ln[c6*(K/L)lI/c5+ (1-c6)] + c7
We find indeed that, with the exceptionof the estimateabased on the Khanindata, all of the altermative
estinates of Sovietoutput and capitalgrowth per worker lend themselvesto the CES form with low
elasticitiesof substitutionbetweencapital and labor (signiflcantlybelow one).l5 The results are less
ISAclasic uticle by Diamnond,
McFadden,
and Rodriguez
(1967)showsthatit is in generalimpossible
to identify
sepurely a time-varying
elasticityof substitudon
pamneterandthebia of technicalchange(neutralverus laboraugmenting
etc.) Weidentifythesubstitution
paraneterby presumingthatit is constantovertimeandthattechnical
changeis neutral.
1s
sharp when we use the entire sample 1928-87,where as indicatedearlier the data before 1950are
volatile. Serial correlation is a problemfor most of the estimates,with the significantexceptionof the
resultsusing our preferred WesternGDP series for 1950-87.
The results with the Khanindata are intriguingbecause they support a story of unit elaticity of
substitutionand poor (though not necessarilydeclining)productivitygrowth. Accordingto the Khanin
data, growth declines mainly becausecapitalgrowthslowed (see Table 1 again). Given Bergson's
criticismsand the limited informationaboutthe methodologybehind the Khanindata, these differing
results can only point to the need for furtherresearch into Khanin's approachto see wheher his work
representsa valid criticism of the Westernestimates.For the moment, we are forcedto regard the
Khlaninstory as unproven.16
The low elasticity of substitutionfrom the other data series gives us an importantinsightinto
the lack of success of the Sovietextensivegrowth strategycompared to the high payoffsfrom capital
deepeningin Japan, Korea, and other market economies. The literature does not find the elastdcityof
substitutionbetweencapital and labor in market economiesto be greatly below one (see for example,
Berndtand Wood (1975) and Prywes (1983)for discussionof theirs and other results for U.S.
manufacturing).A recent study estimatingthe elasticityparameter from the convergencebehavior of
the cross-sectionalnational per capita incomedata even argues that the elasticityof substitutionis
slightlyABOVEone (Chua (1993)). Diminishingreturns to extensivegrowthwere much sharper in the
USSR than in market economiesbecausethe substitutabilityof capital for labor was abnormallylow.
In the concludingsection, we will speculatewhy substitutabilitymay have been low in a planned
economy.
Another striking feature of Table S is that the impliedrates of TFP growth show no significant
declinebetween the 50s and 80s. Thus, freeingup the functionalform of the productionfunctionrules
16Wewouldhavelikedto examinetheimplications
of the"Britishschool"of Hanson(1984),Nove(1981),and
Wiles(1982),whosomewhat
similarclaimsto Khanin's.However,we cannotdo sosincethoseresearchers
do not
providealternativetimeseriesforoutputandcapital.Notethata lowerestimaeforthegrowthrae of capitalover
theentireperiod,asimpliedbythe"British"arguments,wouldimplyhigherTFPgrowthbut doesnot implyanything
abouttheestimatedelasticityof substitution
thatwouldresukfromusingsuchdata.
16
out the collapsingproductivitygrowth explanationfor declininggrowth: in Table 4, both extensive
growth and declini productivitygrowthaccount for the overall fall in growth; in Table 5 the fault lies
endrely with dlminishingretuan to capital.
Table S also showsthat the level of TFP growth is more plausibleafter we control for the
shrply falling marginalproduct of capital with a low elasticityof substitution. Our preferred estimates
are toe yielded by the WesternGDP estimatesin the last column in Table S for the 1950-87sanple.
Those estimatesyield a constant TFP growth rate of I percent per annum, in contrast to the negative
TFP growth impliedby the Cobb-Douglasestimatesin Table 4 for the 60s through the 80s. We find a
positiverate of TFP growth with fallingreturns to capitalmore plausible than a negativerate of TFP
growth.
In Figure 6 we examinea second implicationof the estimatesin Table 5: these are estimatesof
the *shareof capital" impliedby the alternativecolumnsin Table 5 for 1950-87,assumingmarginal
productivitypricing. In the graph, we presentonly the more reliable westerndata. For total GNP, the
share of capital fallssteadily throughout. For industrialoutput, the impliedshare of capital wouldhave
been close to one until the mid-50s, and it then would have begun a sharp declineto close to zero by
1980.
In Figure 7 we present closelyrelated data, on the marginalproductof capital impliedby the
CES estimates. The western GNP data imply high rates of return to capital in the early 50s, declining
to about 3 percent in 1987. The 1950marginalproductof capital in industry is lower than that implied
by the GDP estimates. It stays constantthroughoutthe 50s, then declinessharply to zero by the late
70s.
The data presentedin Figures 6 and 7 suggestthat a market economycould not have gone
through the growth process of the Sovieteconomybetween 1928and 1987. The very low wage shares
in the early period wouldprobably have preventedany but a subsistencewage equilibriumin those
periods. The essentiallyzero marginalproductof capitalin industry (estimatedusing western data) by
the mid-80s would have been inconsistentwith equilibrium,and wouldhave meant that investmentin
industryand the capital-laborratio would have been lower.
17
What would have happenedin the early years if there had been a market economy? One
posibility is that some method-such as trade unions-would have been foundto divorce factor
payments from marginalproductivities. Anotherpossibilityis that different technologieswouldhave
been adopted. Similarly,in the later period, there may well have been other technologiesavailablethat
yielded a positive return to capital. It is also possible that if the extensivegrowthroute had been closed
off in a market economy,there wouldhave been more incendvefor Sovietentrepreneursto attemptto
imnprovetechnology.
The high capitalshare in the CES productionfunctionsbefore 1960has one other implicadon
we find interesting. A CES functionwith a high capital share acts much like a linear functionof
apital, so that the marginalproduct of capital can stay flat for as long as the capital share is high (see
the line for indutrial capital's marginalproduct in the SOsin figure 7). With a very capital-intensive
producdon of goods, includingcapitalgoods, the Soviets were close for a while to the model of
growth through rapid reproductionof capital- describedby Feldmanin the 1920sas using "machines
to make more machines"(see Dornar(1957)).17
However, as the capitalshare begins to fall, the marginalproductwill begin to decline. The
decline can be precipitouswhenthe elasticityof substitutionis particularlylow (see the industrial
marginal product in the late 50s and early 60s). While we find the extremevaluesof the marginal
product of capital and capital's share in Figures 6 and 7 surprising, they do not logically rule out the
CES form-the capital-laborratio in a non-marketeconomycould be drivento levels that would not be
observed in a market economy.
E) COMBININGREGRESSIONEVIDENCEWITHPRODUCTIONFUNCTIONESTIMATES
As a final exercise, we insertthe other apparent correlate of declininggrowth - defense
spending _ into our productionfunctionestimates(we take the midpointof the Brada and Graves
estimatesin Table 3 spliced together with the Steinbergestimatefor 1985-87). Specifically,we allow
1 7 Rebelo(1991) shows formallythat constantretums to reproduciblefactorsin the capitalgoods sector is sufficient
to geneate a constant,sustainedrate of growthevenwithoutTFP growth.
18
the Hicks-neutralrate of technologicalprogress to dependlinearly on the share of defensespending in
GDP in the productionhnction estimatedwith WesternGDP and capital stock data:
3 + (1-c4)]+ C5
ln(Y/L) - cl*nme + c2*nme*(DefenreSpending)+ c3* ln[c4*(KfL)l/C
Theresults are shown in Appendix2. We find defensespendingdoes indeedhave a significantand
negativeeffect on the rate of incr.ase in the total productivityterm in the productionfunction.
However, the effect is not very quantitativelyimportant:every additional1 percent of GDP spent on
defenselowered productivitygrowth by .07 percent. The increase over 1960-87of 2.2 percentage
points in the defenseshare thus would have loweredgrowth by .15 percentagepoints. Moreover, the
parametersof the CES functionare virtuallyunchangedfrom our earlier regressionso the low
substitutability,diminishingreturns story still holds We also tried equipmentinvestmentand R&D
spendingas independentinfluenceson the technicalprogress term, but both gave insignificantresults.
How do we reconcileour productionfunctionestimateswith our earlier cross-sectiongrowth
regressionevidenceusing the Levine-Reneltspecification? The Soviets' high capital-outputratio and
low substitutabilityof capital for laborimplies a lower derivativeof growth with respectto the
investmentrate than in other countrieswith lower K-Y ratios and more substitutablecapital for labor.
To see this, assume zero depreciationand labor growth for simplicityand set labor =1 by choice of
units. Assumea CES functionY=A(yKP+ I-y)(l/P). Growthwill be givenas a functionof the
investmentratio (I/Y = AK/Y)as follows:
AY/Y = y (I/Y) [ (K/Y) (y+(l-y)K-P) ]-I
As is well known, a higher K/Y impliesa lower marginaleffect of the investmentratio on growth
simply becausea given investmentrate translatesinto lower capital growth. With a unit elasticityof
substitution(p=O), this is the only way that the level of capital influencesthe marginaleffect of
investment.With a less than unit elasticityof substitution(p < 0), higher capitalhas an even stronger
negativeeffect on the coefficienton investmentin a growth equation. Althoughobviouslynot the only
explanation,this is consistentwith the large negativeresidualfor the USSR-- and increasinglynegative
residualsover time - in the cross-sectionregressions. (With only one observation,we cannot
distinguishbetweena Soviet slope dununy on the investmnent
coefficientand a Soviet interceptdummy.)
19
We concludefromour reexamination
of the aggregatedatathatthe originalWeitzmanstory
holdsup. Sovietgrowthdeclinedbecauseof diminihig returnsto capitalaccumulation,
andnot
becauseof a slowdownin TFP growth. The averagegrowthperformancewaspoorwhenwe takeinto
accountthe rapidcapitalgrowthandhigheducationlevels. The generalextensivegrowthhypothesisof
theliteratureon Sovietgrowthis notsufficientexplanadonby itself,becausein a comparativecontext
we findthat Sovietextensivegrowthwasnotthat unusual.It wasthe low substitutability
of capitalfor
labor,ratherthan extensivegrowthper se, thatwas thefatalweaknessof theSovietdevelopment
strategy.
11.R
_Ub_ican
RMlts
-
Capitl Growthand
la
Growth
_ow
The republicantimeseriescovertheperiod 1970-90,andprovidedetaileddatato describethe
economicdeclinein the finalyearsof the Sovietsystem,by republicandby sector. The dataare for
NMP. Table6 showsleast-squares
estimatesof real NMPgrowthper workerin the USSRand the
republics,overall,andby branchof industry,for theyears 1970-90.Growthratesin the CentralAsian
republicswerewellbelowthosein the restof the SovietUnion,withBelarusandGeorgiahavingthe
highestratesof outputgrowth. Amongthe CentralAsianrepublics,Turkmenistan
experienced
negativegrowthof per workeroutputover the20-yearperiod;the growthrate of per workeroutputof
Kyrgyzstan,the mostrapidlygrowingof theCentralAsianrepublicswas nonetheless
a full percentage
pointbelowthat of the slowest-growing
of theotherrepublics,Azerbaijan.Growthperformancein the
Balticsdoesnot standoutrelativeto theSovietaverage.
Acrossbranches,outputper workergrewmostrapidlyin industry,andat a negativerate in
agriculture;outputgrowthin theservicesectors,transportandcommunications,
andtradeand
procurement,waspositive. Belarusshowsthehighestrate of growthof outputper workerin industry;
Lithuania,whereaggregategrowthwasrelativelylow,alsoshowsrapidoutputgrowthin industry.
The slowgrowthof outputin the CentralAsianrepublicsclearlyowesmuchto thepoorperformance
of agriculturein theserelativelyagriculturalrepublics.
Table7 showsestimatesof TFP growth(computedassuminga 0.4 capitalsharefor all sectors
20
and a Cobb-Douglasform) by sector and by republic.18 Judgingby TFP growth, industrydid
relativelybetter than other sectors in the EuropeanUSSR Oustas industrialproductivitygrowth is
usually higher than other sectors in the West), while agriculturewas a disaster everywhere.Transport
and communicationsdid well in the border regionsof Belamsand the Baltics. Productivitygrowth in
constructionand in trade is uneven and generallycloseto zero. Central Asia is an almostunrelieved
tale of woe for all sectors, with Kyrgyzstanstandingout again as havingthe best performance in that
region.
For the entire material sector's TFP growth, Georgiaand Belarusdid the best over 1970-90,
Armenia, Azerbaijan,Latvia, and Estonia, the next best, and Central Asia the worst. The relative
successof Belarusand Georgiawas due entirelyto industry, with productivityperformance in
sectors
agriculturestill disastrous,and performancein the transport/communications/construction
generallypoor.
The relativeperformanceof the republicsshownhere is broadly consistentwith previous
studies focusingon earlier periods. The rankingof TFP growthby republic for the period 1960-75in
Koropeckyj(1981)is similar to that in Table 7: Belarusis at the top, and Central Asia at the bottom.
Whitehouse(1984)presents similar findingsfor 1961-70:Belarusand Georgiaare third and fourth in
productivitygrowth(Latvia and Estoniaare at the top), with Central Asia again firmly ensconcedat the
bottom. The bad Central Asian outcomeis well knownin the literature (for example, Rumer(1989)).
A) EXPLWNINGRELATIVEPERFORMANCEWITHINTHE USSR
Growthby republic is correlated with some of the samefactors - human :apital, initial income,
and populationgrowth- that have been singledout in recent cross-sectionalgrowth regressions(Barro
(1991), Barro and Lee (1993), Levine and Renelt (1992)). We first examinesimp'e correlations
betweenestimatedproductivitygrowth over the period and these factors, and then present the resultsof
a multipleregression.
formfortherepublicansectorsafterrejectingit forthe
to resortto theCobb-Douglas
I tWeareratherembarrassed
aretooshortto lendthemselvesto CESestimation
series
SovietUnionas a wholein section1. Therepublicandata
ratesarestillusefuldescriptivestatistics.
growth
TFP
of individualproductionfunctions.TheCobb-Douglas
21
Pigue 8 shows the associadonbetweenone measureof humancapial - the percentageof
specblists with higher educationper capita - and productivitygrowth. The negativeproductivity
growth of the Central Asian republicsis associatedwith a low level of higher education,whilethe
relatively high growth of Georgia, Latvia, Estonia,and Amenia is strikinglycorrelatedwith a high
proportion of highly-trainedspecialists. Belarusis well above the impliedregression line, reflectingits
relatively low dwhreof higher educationspecialists. Sinilarly, the Central Asianrepublics' poor record
is associatedwith rapid populationgrowth (Figure 9).
We have also run a standardcross-sectionalgrowth regression for the fifteen republics.While
we have not found republicandata to match the internationaldata in the Levine-Reneltregressionwe
presnted In Section 1, we can do a similar cross-sectionacross republics. The results, presentedin
Table 8, are similar to those obtained in the standardcross-countryregressions. The educational
variablehas a posidve coefficient,while those on initialincomne(relativeto the Sovietaverage) and
populationgrowth are negative. The coefficienton populationgrowth is much larger than is normalin
cross-countryregressions.
The coefficienton initial relativeincome impliesthat the rate of convergencebetweenthe
Soviet republicsis over 4 percent per year (takingthe derivativesat the Sovietaverage). This impliesa
rate of intermalconvergencefor the Soviet republicsconsiderablefasterthan the convergence
coefficientsfound by Barro and Sala i Martin (1992),which are around 2 percentper annum for both
US states an* a sample of 98 countries. Chua (1993)showsthat convergenceis more ranid the lower
is the elasticity of substitutionbetweencapital and labor. The rapid convergenceof Soviet republicsto
each other (though admittedlyb. ed on the tenuous evidenceof 15 observations)is yet another
confumation of the stronger force of dirninishingretums (and possiblythe lower elasticityof
substitution)in the Soviet Union comparedto market economies.19
19 An altemnative
explanationfor faster convergenceamongSovietrepublicsis that Sovietpolicymakersplacedmore
emphasison regionalredistributionthan did Westempolicymakers.
22
IN THEREPEUCS
PATTRNSANDPRODUCV/7 GROWTH
B) SECTOPiAL
Table 7 showsenormous variationsin productivitygrowth betweensctors and republics; in
this section, we examinewhetherthese variationsare related to degreesof sectoraldistortion. It is well
known that the Soviet Union (and socialisteconomiesin general) had distortedsectoral structuresof
production(Ofer (1987)). Ofer comparedthe sector employmentand outputshares in the Soviet Union
to those that would have been expectedfor a country of its pce capita incomelevel, using the patterns
of sectoral shares and income establishedby Chenery, Robinson,and Syrquin(1986). He showedthat
the Sovietservices sector was smallerthan normal, Sovietagriculturewas larger than normal, and
Soviet industry roughly normal for a country of its per eapita income level. The atrophied service
sector has been documentedalso with recent data (Easterly, de Melo, and Ofer (1994)).
Table 9 shows the differencebetweenthe sectoral shares of employmentthat would have been
predictedby the republics' respectiveper capita incomesand their actual sectoralshares. The per capita
incomesare derived from the estimatesof relative incomesby the World Bank, and then appliedto the
per capita income for the SovietUnionas a whole in Bergson(1991b). The employmentshares of the
comparatorsare taken from InternationalLabor Organization(various years) for a sampleof about 70
developingand developedcountries. We see the basic pattern confirmed:agricultureis larger than
average in all of the republicsfor their respectiveincomelevels, and trade is smaller. Transportis
larger than expected in the republics.Industryand constructiondo not divergeas sharply from the
expectedpatterns.2 0
Are the sector imbalancesrelated to relativeproductivitygrowthperformance?The answer is
yes - sectors that were "too large" had poor productivitygrowth. Figure 10 showsa scatter of
productivitygrowth 1970-90for the 5 sectors and 15 republicsagainstthe sectoral employment
20 Thesemeasuresare extremely
Forexample,the
crudeandobviouslyreflectotherfactorsbesides"distortions".
fertilesoilsof Ukraineand Moldovamightimplya largeragriculturesharethanpercapitaincomealonewould
dataset is enormousandtheresiduaisshownin
sharesinthe intemational
predict.Thevariationin employment
significantin OLSregressions(withtheexceptionof manyof thedeviations
Table10aregenerallynotstatistically
remainusefuldescriptivestatisticsfor
deviationsnevertheless
in republicantradeshares).Thesectoralemployment
statisticscovertheentireeconomy,andso are
thenatureof therepublicaneconomies.Notethatemployment
preferableto NW sharesthatonlycoverthematerialsector(notto mentionthepricingproblems).
23
deviationsin 1970. Republicanagriculturalsectors that were above the predictedemploymentshares
also had poor (negative)TFP growth, while industrialand trade sectors belowpredictedshares of
employmenthad positiveTFP growth. The slope coefficientin figure 10 is -.11 (ten percentagepoints
excess employmentshare lowers TPP growth in that sector by 1.1percentagepoints). The coefficientis
strongly significant. This result is reminiscentof the finding in market economiesthat distorted
sectoral price incentivesor other measuresof departurefrom comparativeadvantageare negatively
related to growth (Barro (1991), Easterly(1993), Edwards(1989), Fischer (1993)).
III. Conclusion:internretin the results on SovietGrowth
Our results confirmand updatethe results of Weitzman(1970) on the low Sovietelasticityof
substitutionbetween capital ard labor, which combinedwith the Sovietattemptat extensivegrowth, is
sufficientto explain the declineof Sovietgrowth. The natural questionto ask is why Sovietcapitallabor substitutionwas more difficultthan in Westem market economies,and whetherthis difficultywas
related to the Soviets' p!annedeconomicsystem.
Recent work on modelsof endogenouseconomicgrowth stressesthe notion of a broad concept
of capital, includinghumancapital, organizationalcapitol,and the stock of knowledge,which can
substituteeasily for raw labor and perhaps replaceit altogether(Rebelo 1991,Jones and Manuelli
1992, Parente and Prescott 1991). Conversely,one possibleexplanationfor the Soviets' substitution
problemswould be that, under an autocraticallydirected economicsystem, they accumulateda narrow
rather than a broad range of capital goods. Some formsof physicalor humancapital that were missing
wouldhave been market-orientedentrepreneurialskilis, marketingand distributionalskills, and
information-intensivephysicaland humancapital (becauseof the restrictionson informationflows). It
is more difficultto substitutemore and more drill pressesfor a laborer than it is to substitutea drill
press plus a computerizedinventoryand distributionsystem for a laborer. There is nothingthat
explicitlysupports this conjecturein our results, but it is an interestingdirectionfor further research.
The other messagefrom our results couldbe that Soviet-stylestagnationawaitsother countries
that have relied on extensive growth, a point that has been made forcefullyfor those extensivegrowers,
the East Asian Tigers, in several articlesby Young(1992, 1993a, 1993b). After all, the USSRhad its
24
period of rapid growth in the 30s through S0s when it appeared to be followinga linear output-cpital
production function,as we have shown. If East Asiancapital-outputratios keep rising until they reach
the extreme Sovietlevels, they too could experiencea drastic slowdown.Even if diminiing returns
are weaker in the East Asian economies(if, followingour conjecture, they have been accumulatizna
broader range of capital goods and experiencinghighersubstitutabilitybetweencapital and labor),
diminishingreturns would still eventuallycause a growthslowdown. The Sovietexperiencecan be
read as a particularlycxtreme dramatizationof the long-run consequencesof extensivegrowth.21
The cross-sectionresults on republics,althoughbased on a small numberof datapoints,support
the idea that some cf the same factorsthat are argued to determinegrowth in the recent empirical
cross-sactionliterature- humancapital, populationgrowth, initial income, sectoral distorticns- also
matteredunder Sovietcentral planning. Our results with the SovietUnion in the internationalcrosssection growth regression indicatethat the plannedeconomicsystem itselfwas disastrousfor long-run
economicgrowth in the USSR. While this point may now seem obvious, it was not so apparent in the
halcyondays of the 1950s,when the Sovietcase wasoften cited as support for the neoclassicalmodel's
predictionthat distortionsdo not have steady-stategrowtheffects. Sincea heavy degree of planning
and govermnentinterventionexists in many countries, especiallyin developingones, the ill-fatedSoviet
experiencecontinuesto be of interest.
21 Weitzman
(1990)describesSovietgrowth(asanalyzedbyOfer(1990))asthebestapplicationof theSolow
neoclassical
modeleverseen.
25
BIBLIOGRAPHY
page
Acurid-Hood,Mary. 1987. 'EstimatingSovietMilitaryR&D Spending' in Jacovsen,ed., Tle Snviet Def1EePM
136-147.
Vol. 6, No. 3,
Anderson,Barbaraand Brian Silver(1990). 'Trends in Mortalityof the SovietPopulation",Saxig.Ecin=.
pp. 191-251.
Barro, Robert J. (1991). 'EconomicGrowth in a CrossSectionof Countries',
407-444.
,
ru Journualnf
-- and Jong-whaLee (1993), "InternationalComparisonsof EducationalAttainment",
3?(3', 363-393.
105, 2 (May),
et
_Mn
of
ami
Barro, RJ. and X. Sala-i-Martin.1992. "Convergence',Journalof Poliical Economy100, 223-251.
Benhabib,Jess andMark Spiegel. 1993. "The Roleof Human Capitalin EconomicDevelopment:Evidencefrom Agregte
InumalofMgna,
Cross-country
andRegionalU.S. Data",forthcoming,
Berndt,Ernst and D-wid0. Wood. 1975. 'Technology, Prices,and the DerivedDemandfor Energy". Ile Review
EconoMiCsand Statistics,Vol. LVII. No. 3, 259-268.
aLLa
ProductionFunctionin SovietPostwarIndustrialGrowth",
'.
Bergson,Abram (1979). 'Noteson the~
'.connnmics,
3, 2 (June), 116-126.
C
(1983). 'TechnologicalProgress" in A. Bergsonand H. Levine(eds), 'TbeSovietEconomy:Towardthe Year 'Q.
London:Allen & Unwin.
(1987a). "On Soviet Real InvestmentGrowth', SgyieLStudine,39, 3, (July),406-424.
(1987b). 'Comparative Productivity:The USSR, Eastem Europe.and the West. Americmn.Economic
Review,
vol. 77,
no. 3, pp. 342-357.
-
(1991a). "Real NationalIncomeMeasurementIn Soviet Perspective",paper preparedfor the Bureauof Labor Staistia
(BLS)and Eurostatconferenceon economicstatisticsfor Economiesin Transition.WashingtcnDC, February.
(1991b). "The USSRBefore the Fall: How Poor and Why",
Iournal
of EconomicPersectives, 5, 4 (Fall), 29-44.
Berliner, JosephS. (1976). The _nnovatiorDecision in Soviet Tndus . Cambridge,MA: MIT Press.
Birdsall,Nancy and ChangyongRhee, 1993, "Does ResearchawdDevelopmentContributeto EconomicGrowth in Developing
Countries?', Policy ResearchWorkingPaper 1221(WorldBank:WashingtonD.C.)
Brada, Josef C. and RonaldL. Graves. 'Slowdown in Soviet DefenseExpenditures." Southeconmic
aurN. vol. 54 pp.
969-84,1988
Brooks, Karen McConnell.1983. "Productivityin Soviet Agriculture".in D. Gale Johnsonand Karen Brooks,eds.,
Prospectsfor SovietAgriculturein the 1980s.
NewYork:OxfordUniversity
andMosheSyrquin(1986).IndusMiaL4&aisnlidGrowh.
Chenery,Hollis,ShermanRobinson
Press.
GrowthCenterDiscussionPaper,YaleUniversity.
andSlowConvergence'.Economic
Chua,Hak. 1993."FactorSubstitution
PrintingOffice)
CIA (1982),USSR Measuresof EconomicGrowthandDevelopmnen
(WashingtonDC: USGovernment
CIA(variousyears), Hsndbookof EconomicStktistics(WashingtonDC: US GoverrnnentPrintingOffice)
Coe, DavidT. and ElhananHelpman(1993), "InternationalR&D Spillovers".NBER WorkingPaperNo. 4444, Augm.
De Gregorio, Jose. 1992. "EconomicGrowth in Latin America" urnslo_lm,
Vol. 39 No.1, pp. 59-85.
26
De Long, J. BradfordandLawrenceSummers. 1991. 'Equipment, Investment,RelativePrices,and EconomicGrowth.
Quarteriylounrl of Ecorgomic,May 1991, Vol CVI, Issue 2, pp. 445-502.
and
on
(1992). 'Equipment Investmentand EconomicGrowth: How Robustis the Nexus?'
rcnmicActivity,2, 157-211.
Ri
r
and_
(1993). 'How StronglyDo DevelopingCountriesBenefitfrom EquipmentInvestment?7,InurnLnf
Mnncsarj,Ecorimiza
32, 395-415.
Desai, Padma (1976). 'The ProductionFunctionand TechnicalChangein PostwarSovietIndustry:A Reexawnatioa',
AmericanEconomicReview,66. 3 (June). 372-381.
Oxford: BasilBlackwell.
-
(198T). The SovietEconomy. ProbILm a nd PrQtpets.
-
(1992).'Reforming
the SovietGrain Economy:Performance,Problems,and Solutions',
ricn EonomiC Reviaw,
vol. 82, no. 2, pp. 49-54
Diamond,Douglas,Lee Bettis,and RobertRamsson.'AgriculturalProduction' in A. Bergsonand H.S. Levine, ods.. Mm
,Soviet Fronomy:Toward the Year 2000, Allen& Unwin, Boston.
Diamond,Peter, Daniel McFadden.and MiguelRodriguez.1967, 'Measurement
of the Elasticityof FactorSubstitutionand
Biasof TechnicalChange", in...
Domar,Evsey (1957). 'A SovietModelof Growth', in Essaycin the Theory of Economic rrowth. New York: Oxford
UniversityPress.
Easerly,William, Marthade Melo, andGur Ofer. 1994. 'Servicesas a Growth-PromotingSector in the Former Soviet
Union', in Mario Blejer, ed. _Macroeconomic
ad Strumral Anpets of Ru_aiq c miCReform, Intenutjoal
Center for EconomicGrowth.
EAsterly,William. 1993. 'How Much Do DistortionsAffect Growth?' InalofMD
i
32: 187-212.
MichaelKremer, Lant Pritchea, and LawrenceH. Summers,1993, 'Goodpolicyor goodluck? Countrygrowth
performanceand temporaryshocks', Journalof MonetaryEconnmics 32, 459-483.
Edwards, Sebastian.1992. 'Openness,
OutwardOrientation,Trade Liberalization,and EconomicPerformance', NBER
workingpaper.
Elias, Victor J. Sourcesof Growth!a Studvof Seven latin AmericanEconomies,InternationalCenter for Economic
Growth,
1992.
Ericson, Richard E. (1990a). 'The SovietStatisticalDebate:Khaninvs TsSU', in H.S. RowenandC. Wolf, Jr. (eds) Mm
ImpoverishedSuperpower. San Francisco,CA: ICS Press.
Feschbach,Murray. 1983. 'Populationand LaborForce' in A. Bergsonand H.S. Levine,eds., The So
the Year 2000, Allen & Unwin,Boston.
Toward
Fischer,Stanley(1992). 'Russiaand TheSovietUnion:Then andNow', NBER Conferenceon Trasition in Eastern Europe.
_____.
(1993) 'TheRole of MacroeconomicFactors in Growth', Jour-a of Mn
om
32 (3), 485-511.
Gomulka,S. andM. Schaffer'A NewMethodof Long-RunGrowthAccouuing, withApplicaios to the SovietEconomy
1928-87and the US economy1949-78.
for EonomicPerforinne DiscuscionPaper 14. Jazuay 1991.
Haon., Phillip(1984). 'NewEvidenceaboutSoviet Investment',Ri
h;
13 September.
Harrison,Mark(1993). *SovietEconomicGrowthSince 1928:The AlternativeStatisticsof G.1. Khanin', EwzmAsAa
Sidi, 45,1, pp. 141-167.
Hewitt,Daniel (1993). 'Miliary Expenditures1972-1990,The ReasonsBehindthe Post 1985Fall in World Militay
27
Spending,IMF WP/93/18
InternationalLaborOrganization(variousyears), Handbwkof LAhorStatics
IMF, World Bank, OECD, and EBRD. 1991.A Studynf the SokietPconomy(3 volumes).
Jones,L. andR. Manuelli. 'A ConvexModelof EquilibriumGrowth",
cion
of PolitinAL
1 _urnal
98, 1008-1038.
Joint EconomicCommittee(1990), Measuresof SovietGNP (U.S. GovernmentPrintingOffice:WashingtonD.C.)
Jorgensen,Dale. 1986"EconometricMethodsfor ModelingProducerBehavior", in Z. Griliches nd M. Intriliptor,
Hand,bookof Roonomerrica,Vol. 3. Amsterdam:North-Holland.
Judson, Ruth. 1994. "Do Low Human CapitalCoefficientsMakeSense?A Puzzle andSomeAnswers", mimeo, Departnent of
Economics,MIT.
Kellogg,Robert L. 'Inflation in Soviet Investmentand CapitalStockdata and its impacton meaurement of total factor
productivity",CIA Office of Soviet Analysis,May 1989.
Khanin,G. 1988. *EkonomicheskiiRost: Al'ternativnaiaOtsenka",Kammnais, (No. 17)
Kim, Jong-l1and LawrenceJ. Lau. 1993. "The Sourcesof EconomicGrowthof the East Asian Newly Industrialized
Countries', StanfordUniversityMimeo.
King, Roben and Ro3sLevine. (1994). 'Capital Fundamentalism,EconomicDevelopment,and EconomicGrowth',
forthcoming.
Carnegie-RochesterSeries on PublicPolicy, forthcoming.
King, Robert and SergioRebelo(1993). "TransitionalDynamicsand EconomicGrowth in the NeoclassicalModel',
FconomicReview, September.Vol. 83, No. 4. pp. 908-931.
&Mariin
Kingkade.W.W. (1987) 'DemographicTrendsin the Soviet Union", in Joint EconomicCommittee,Gorbachev'sFcnomie
liAn. Volume 1, U.S. GovernmentPrintingOffice,WashingtonD.C.
Koropeckyj,I.S. (1981). "Growthand Productivity' in I.S. Koropeckyjand GertrudeSchroeder(eds), ERornmic.q
of Snviet
Regions. New York: Praeger.
Landau, Daniel. 1993. "The EconomicImpactof MilitaryExpenditures'.Policy ResearchWorkingPaper 1138, WorldBank.
Levine, Ross and David Renelt(1992). "A SensitivityAnalysisof Cross-CountryGrowth Regressions",An
Review,82, 4 (Sept), 942-963.
Lichtenberg,Frank R., 'R&D Investmentand InternationalProductivityDifferences,"NBER WorkingPaper#4161,
September1992.
A. Maddison,The World Economyin the 20thCentury. DevelopmentCenter of the Organizationfor EconomicCooperation
and Development,Paris, 1989
Marer, Paul, Jaos Avray, John O'Connor, Martin Schrenk,and Daniel Swanson.1992,
adg theDam, World Bank: WashingtonD.C.
n_ie
Millar, J.R. and E. Clayton. 1987.Qualityof life: subjectivemeasuresof relativesatisfaction,in J.R. Millar, ad. Politics,
work, daily life in the USSR. CambridgeUniversityPress.
Moorsteen,RichardH. and Raynond P. Powell(1966). rne SovietCaitl Stock 1923-1962.Honewood. IL: Irwin.
Nehru, Vikram andAshok Dhareshwar(1993), "A New Databse on PhysicalCapitalStock:Sources,Methodology,and
l Vol. 8, No. 1, 37-59.
Results." Revistade AnaiV Eo
uSlim,
Nove, Alec (1981). "A Note on Growth,Investment,and Price Indices", Soviet
Ofer, Gur (1987). "Soviet EconomicGrowth: 1928-85-, 1a
u_ L
_,
23, 1 (Jan), 143-145.
25, 4 (Dec), 1767-1833.
A
28
-
Issues of SovietReforns',
(1990). Macroeconomic
_
, 297-339.
A
Work Pper No. 3733.
N13ER
andGromwth,
Adoption
StephenL., andEdwardC. Prescott,1991."Technology
Parente,
P. (1968)*TheSovietCapitalStockandRelatedStuaical Seriesforthe WarYean', Me Econom
Powell,Raymond
GrowthCenter,YaleUniversity.
Production'in AbramBergsonandSimonKuzneu,eds.,
(1963).'Industrial
MA:HarvardUniversityPress.
Canbridge
T
in the Soviet huani,
Substitution'.FederalReserveDankof NewYork.
Prywes,Menahem.1983.'A NestedCESApproachto Capital-Energy
Policy AnalysisandLong-RunGrowth', u
Rebelo,Sergio(1991). 'Long-Run
f
a
99, 3 (Jun),
500.521.
Growth', in R. BDra(ed),Mmax
in the Tneoryof Long-Run
Romer,Paul(1990). 'CapitalAccumulation
MA:HarvardUniversityPress.
M=. Cambridge,
______
(1989). 'What DeterminestheRateof GrowthandTechnicalChange?', World Bank PRE WorkingPaper#279,
September1989.
Rumer,Boris Z. (1989). Soviet Central Asia A TragicFaeiMe.
Boston,MA: UnwinHyman.
Solow, RobertM. (1956). 'A Contributionto the Theoryof EconomicGrowth". f
(Feb). 65-94.
rterlX oumnalOf9knomia. 70, 1
Roben andAlan Heston(1991). The Penn WorldTable (Mark5): An ExpandedSetof Internaional Comparisons
Summers,
1950-88'. QuarterlyJournalof EconomicsCVI, 369-406.
Dmitri. 1987. 'EstimatingTotal SovietMilitaryExpenditures:an Alterntive ApproachBasedon Reconstructed
Steinberg,
SnvietDefe
Soviet NationalAccounts.' in Jacobsen,ed.,*Te
Treml, VladimirG. 1991.'AlcoholAbuse'.in Jones, Connor ndPowell, eds. Soviet SocialPrhlems
UNESCO(1975),UNESCOStaigicil Yearbook(UnitedNations:NewYork)
Weitznan,MartinL. (1970). 'Soviet PostwarEconomicGrowthand Capital-LaborSubstitution', AzjiaEc4
Review, 60, 5 (Dec), 676-692.
--
(1983) "IndustrialProduction' in A. Bergsonand H.S. Levine, eds., TheSovietRconomycToward the Year I=,
A Unwin. Boston.
Allen
(1990). 'Comment', NR,R MacroecgRaMiGLAnna, 339-342
Aspectsof RegionalEconomicDevelopmentin the USSR' in V.N. Bandera
Whitehouse,F. Douglas(1973). 'Demographic
and Z.L. Melnyk (eds). The Soviet Rconomyin ReganaIPersecixe. New York: -raeger.
Wils, Peter (1982). 'SovietConsumptionandInvestmentPrices ndthe Meaningfiunessof Real Investment', Snai%t.Si%d,
24, 2 (April), 289-295.
Young,Alwyn. 1993b,'TheTyrannyof Numbers:ConfrontingtheStatisticalRealitiesof the EAt AsianGrowthExperinc'e',
MIT
m-imeo
from the East AsianNlCs: A Contrrian View', fortcoming,
1993s,''Lessons
in Hong Kong andSingapore', In
1992,
M 'A Tale of Two Cities: Factor Accumulatiob ndTecbial Chaunge
CambridgeMA: MITPreus.
A,
M
_
eds.,
Ficher,
OlivierJ. BlinchardandStanley
29
Figure 1: Index of satisfaction with standard of living in USSR
reported in survey of emigres, 1983
2.8
2.7
Index ranges from 4 (very satisfied)to 1 I
(very unsatisfied)
2.6
5^2.5**_
2.4
2.3
Over 54
41-54
31-40
Age group in survey
Source: Millar and Clayton (1987)
Under 31
Figure 2: Male life expectancy at birth: USSR and world median
68
64
62
USSR(Source:Anderson
andSilver(1990))
60
58
56
Medianfor 150countries(Source:WorldBankEconomicandi
SocialDatabase)
52
50
%n
!,------ -
/
-
-
-
%O
- El !L
O7
%O
…--------
_O
- % -r .1 12O
~00ac
%
0 0t
t toF0
@
!S0000^
et%
!e00 CP40
1950
---
1952
1954
c
19560
1958
1960
1962
1
0
1964
1966
0r
C.
0
a~~~~~~~~~~~~~~~
0
0~~~~~~~~~~~~~~~~~~~~~~~~
o1968
()
1970
1972
1974
1976
II-
1978
1980
1982
1984-0
1986
a
32
Figure 4: Capital-output ratios, AlternativeEstimates
5,2
4.5
CIA(GDP)
I
4
Khanin(NMP)
Ofricial (NMP)
3.5
__
_
_
_
_
_
_
3.
2.5
1.5
ob
3.2
N
°h
~I, ~ri
C
°@
0
'C
i
',~
"0
~C
N
"0
I~
qw
'
~
1-
@0
'
N
@00
000
-
3
Industry
-
(official)
2.8
Industry(West)
2.6
2.4
2.2
2
1.8
_
.
*.
_
1.6
o
N
t
'C
eo o
WI
N
§
C@0
o
N
q
0
N
q
0
0
un
0
b
OV
1950
1952
1954
1956
\
1958
CA
n
1960
1962
\
1964
1966U
1968
o
S
1970
1972
0
1974
nr
2-
1976
1978
.
1980
1982
1984
1986
B~~
U~~
Shareof capitalin output
c -
1950
_
i
0
00
Lo
_
0v
.
0
0
0t
Ub
0J
00
@z0
0
vo
_
_
1952
1956
1956
4
0
1958
r~~~~~~~~~~A
3
1960
c
1962
1964
1966
e0.
1968
C,
u
1970
1972
0
1974
jCJ
1976
1978
1980
1982
1984
1986__
~
~
~
~
.~
J
Marginalproductsof capital(includingrisingTFPlevel)
o
1950
o
0
0
kA
t
,
_~
-
U
o
,o
"j
o
o
o
w>
Uh
4i
w
C.
t
.
IA
,
1952
1954
1
~
1956
CL
1958
I
1960
U
.
1962
1
1966
0
196_
00
1970
1972
0
1974
1976
1973
.
1980
O
1982
1984
1986
0
/
_
Ci
Un
36
Figure 8: Growth and Human Capital, Soviet Republics
3%_
g2%
Belarus
;
1%
Moldova
Kyrzstan
E0%
1
*° -1%1
0%
AmiaG
GF
~~~~~~~~Azerbaijan Latvia 8
a
U
Lithuania
Esoi
*
Uknaine* Russia
Estonia
zl
a
fajikstan
Uzbekistan
Kazakhstan
a
Turkmenistan
-2%
-3%
120
Anena
140
160
180
200
220
240
260
280
300
Specialists with higher education per 10,000inhabitants, 1965
320
37
Figure 9: Growth and Population Growth, Soviet Republics
3%_
Belarus Georgia
n
_2t%
%
,Latva
Estonia
Ukraine
a
Aglenia
U
KyrgyZstan
Russia LMIuania
0%
h
Azerbaijan
Moldova
-t
iikt
°
I*
!
1%|
°
Uzbekistan
Kazakhstan
.-2% I
Turkmenistan
0%
2%
1%
rate of natural incrase (percent), 1970
3%
Figure 10: Sectoral employment sbares and TFP growtb, 5 sectors and 15 republics
3%
a
- .%
a
~~~~-S.ckn
~wxjb
rI
a
Gwga a~ab
we apcu**e. cinshucfKit uak'y.
and
Sckr
-W. dmvegs
Gowmn Wed
U
1%
'mu |
%~~~~~~~~~~
i-1%
1970per cape*swm
*t
(Summs4iosn)
gm fw4o~
e
TI*
'
w*
--5%
14%
-15%
-10%
5%
0%
5%
10%
15%
Deviationfrom sector shae predictedby per capita incomw
20%
25%
30%
co
39
Table 1: Soviet Growth Data, 192847
Period
Industry, official | Industy,
| Material sectors, | Materialsctors,
|hsnin
official
Westerm
1Total economy,
Western
Growth rates of output per worker, alternative estimates
1928-87
6.3%s
3.4%
2.1%
6.0%
3.0%
1928-39
12.S%
5.0%
0.9%
11.4%
2.9%
1940-49
0.1%
-1.S%
-1.0%
2.1%
1.9%
1950459
8.9%
6.2%
5.3%
8.3%
5.8%
1960-69
5.7%
2.8%
2.7%
5.4%
3.0%
1970-79
5.2%
3.4%
1.2%
4.1%
2.1%
1980-87
3.4%
1.S5%
0.2%
3.0%
1.4%
Growth rtes of capital per worker, altermativeestimates
1928-87
6.2%
3.2%
2.3%
6.1%
4.9%
1928-39
11.9%
6.5%
S.9%
8.7%
S.7%
1940-49
1.5%
-0.1I%
-1.3%
2.7%
1.5%
1950-59
8.0%
3.9%
3.5%
7.7%
7.4%
1960-69
6.1%
3.4%
3.8%
7.1%
5.4%
1970-79
6.3%
4.1%
1.9%
6.8%
5.0%
1980-87
S.6%
4.0%
Note: growth rates ar- logntbmc wrt-squares estimates.
-0.1 %
5.3%
4.0%
Table 2: The Soviet Union in the Levine-Renelt (1992) Growth Regression, 1960-89
Average for sample excluding
Soviet Union
Soviet Union
Rank of Soviet Union in
sample (out of 103
observations)
Per capita
income, 1960
Per capita
(Summers growth, 1960-89 Heston PPP)
Population
Investment
growth,1960- Secondary
ratio to GDP, Growth
89
enrollment, 1960 1960-89
residual
2.00
1792
2.07
21%
21%
2.36
2796
1.05
58%
29%/e
45
24
81
10
7
-2.34
97
Sources: Datafor all countries except Soviet Unionfrom Levine and Renelt (1992).
Soviet data: Per capita growth--Western GDP described in text, updated to 1988-89 with Marer et al. (1992)
Per capita income--Bergson(1 991b)for 1985 PPP, bockcast to 1960 with per capita growth infirst column
Population growth: Feschbach (1983), Kingkade (1987), IMF et al. (1991). Marer et al. (1992)
Secondary education: UNESCO (1975 Statistical Yearbook), 1970from Marer et al. (1992)
Investment rate: Joint Economic Committee (1990), updated for 1988-89 with Marer et al. (1992) (JEC series available at 5 yr
intervalsfrom 1960-75, interpolated in between)
Growth residual: Residualfrom Levine-Renelt regression offirst column on other columns
0
41
Table3: Sovietdefeme burdena share of GDP
Jwa6 andGmws Brda andGmvs
Oir (1987)
(1988)-High
(1988) - Low
1928
1950
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
Steinberg(1990)
(cW'rentrubles) (constantrubles) (constant
rubles) (constantrubles)
2%
9%
12%
13.34%
9.90%/
13.86%
10.60%
14.93%
11.39%
15.49%
12.32%
15.03%
12.17%
14.49%
11.79%
14.11%
11.54%
14.40%
11.95%
14.45%
12.14%
14.61%
12.08%
13%
13.83%
11.48%
13.28%
13.56%
11.30%
13.76%
13.80%
11.34%
13.61%
13.33%
11.03%
13.14%
13.71%
11.28%
13.15%
14.14%
11.53%
13.57%
14.32%
11.62%
13.30%/e
14.07%
11.26%
12.98%
14.00%
11.09%
13.08%
14.53%
11.43%
13.05%
16%
15.06%
11.82%
13.91%
15.48%
11.75%
14.03%
15.36%
11.70%
14.58%
15.51%
11.63%
14.36%
15.55%
11.57%
14.37%
14.79%
14.49%
14.63%
42
Table4: Toa factorproductivitysrowthrmtu,alteative smin, USSR
period
Khanin
material acton
Official
material
sectors
1928-40
-1.7
7.2
1940-50
-0.2
19S040
Westemot
industrial
sctor
Official
indutrial
s_ctor
Weder at.
GNP
1.7
7.2
-1.2
2.5
-1.1
1.7
-0.2
3.8
6.0
6.1
4.1
1.3
1960-70
I.S
2.9
1.9
3.q
-0.1
1970-80
0.4
1.4
2.4
1.7
-0.8
1980-87
0.4
0.7
ounces:seW
earlierdescriptionin text
-0.1
1.1
-1.2
43
Table 5: Elasitkis of substtuton and TFP growth with utmated CEShmcdows
Watern
Xanln
material
sectors
Official estfmata Ocfidal
material Industrial Indiustrial Western
sectors
sector
sector
GNP
For 1950L87sampk:
Eaticity of subsitudon
1.11
0.37 *
0.13 *
0.40 *
0.37 *
TFPgrwth In:
1950-59
0.11%
2.93% *
2.40% *
3.72% *
1.09% *
1960-69
-0.07%
2.88% *
2.36% *
3.60% *
1.10%*
1970-79
-0.30%
2.98% *
2.51% *
3.74% *
1.16%*
1980-87
-0.35%
2.92% *
2.43% *
3.62% *
1.09%*
Forentir sampleperiod, 1928-87
Easticityof substitution
1.11
0.38 *
0.22
0.45 *
0.81
TFPgrowthIn:
1928-39
-2.03%* 3.34% * -1.38%
0.72%
-0.52%
1940-49
-1.17% * 2.18% * -0.72%
0.51%
-1.32%*
1950-59
-0.18%
2.96% *
0.36%
1.48%* -0.21%
196069
0.33%
2.97% *
0.40%
1.27%
-0.15%
1970-79
0.30%
3.05% *
0.43%
1.38%
-0.18%
1980-87
0.22%
2.97% *
0.37%
1.28%
-0.33%
Notes: * indicateselasticityof substitutionsignificantlydifferentthan oneor TFP growth
rates significantlydifferentthanzero. Full regressionresultsgivenin appendix.
44
Table6: Growtbrats of NM? pewworker1970-90
cmoutt prkc
_
Total
Industry
Trasortn&
Agriculture
_
USSR
2.8%
3.4%
Slavic:
-1.3%
________
ConstrucimonTrade&
Commurncation
Procurelment
3.1%
2.7%
_
2.1%
.
Russia
3.0%
3.5%
-2.1%
3.2%
3.1%
2.4%
Ukraine
2.9%
3.2%
0.3%
3.1%
2.2%
2.2%
Belarus
4.5%
5.4%
0.3%
3.5%
2.9%
2.4%
Estonia
3.1%
3.8%
-1.7%
3.6%
2.6%
2.5%
Latvia
3.3%
4.3%
-0.8%
5.6%
1.0%
2.0%
Lithuania
2.8%
4.9%
-0.6%
3.9%
1.4%
1.0%
Moldova
3.3%
3.3%
0.5%
4.1%
2.2%
2.4%
Georgia
3.9%
4.5%
2.5%
3.1%
2.0%
2.5%
Armenia
3.4%
3.4%
-0.8%
5.2%
2.7%
2.5%
Azerbaijan
2.7%
3.9%
1.9%
0.7%
2.7%
1.6%
Kazakhstan
0.7%
0.6%
-4.4%
1.9%
1.9%
0.4%
Turkmenistan
-0.3%
-0.6%
-2.8%
0.9%
1.6%
1.4%
Uzbekistan
1.2%
2.2%
-1.8%
2.7%
1.0%
1.8%
Tajikstan
1.0%
1.7%
-1.8%
3.1%
0.6%
1.8%
Kyrgyzstan
1.7%
3.1%
-2.4%
3.8%
1.4%
1.1 %
Baltic/Moldavian:
J
Transcaucasian:
CentralAsian:
45
Tablh 7: Totalfactorproductivttygrowthby sector
andrepublik,197O.90
Total
lndusuy
Apictlnws
___________
USSR
Transport&
Conmucdton
______Cormnmunication
Trade&
Procurement
0.8%
1.1%
-4.1%
0.8%
0.2%
0.3%
0.8%
0.9%
-5.3%
0.7%
0.5%
0.4%
Ukraine
1.0%
1.3%
.2.7%
1.0%
-0.4%
0.6%
Belau
2.1%
3.0;
-3.3%
1.3%
1.8%
0.3%
SOWic:
Runsia
|
Blaic/MoldaWan:
Estonia
1.3%
1.8%
-4.4%
1.6_%
0.2_%
O.S_
%
LAuvia
1.4%
2.1%
-3.3%
2.9%
-0.9
%
0.2%
Lithuania
0.6%
2.6%
-4.0%
2.0%
-0.9%
-0.6%
Moldova
1.0%
1.2%
-2.7%
2.0%
-0.4%
0.5%
Georgia
2.3%
2.6%
0.1%
1.3%
0.3%
1.0%
Armenia
1.8% j
1.8%
-3.1%
2.8%
1.1%
0.4%
Azerbaijan
1.4%
2.5%
X
0.1%
-1.2%
0.3%
-0.2%
Kazakhstan
-1.1%
-1.5%
-6.4%
0.2%
-0.3%
-1.2%
Turkmenistan
-2.0%
-3.0%
-4.0%
-2.1%
-0.3%
-0.3%
Uzbekistan
-0.4%
0.5%
-3.7%
0.6%
-0.6%
-0.1%
Tajikstan
-0.4%
-0.3%
-2.9%
1.8%
-1.1%
0.9%
Kyrgyzstan
0.2%
1.1%
-3
.9%
1.7%
-0.2%
-0.7%
Tscaucasian:
|
J
CentralAsian:
-
I
T
46
Table 8: Cross-secdonal growth regresion, 15 Soviet repubUcs
m.uu..uu.m.u.u.
m
u u..mumnm..
m.u..munu..u
LS // Dependent Variable is Total Factor Productivity Growth,1970-90
Number of observations: 15
VARIABLE
COEFFICIENT
C
SPECHI65
INCOM60
NATINC70
0.0303710
0.0001098
-0.0002895
-1.4371064
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
=
0.686424
0.600904
0.007592
54.25'78
STD. ERROR
0.0244123
5.011E-05
0.0001477
0.5552033
T-STAT.
1.2440874
2.1910647
-1.9598548
-2.5884325
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
r.rmnn.i.i.i..r.ir..rnrr..~ur...inrn.ini.imu.nmrm.uumiUU
Variables:
SPECHI65- number of specialistswith highereducationper 10,000inhabitants,1965
INCOM60 - income per capita relativeto USSR, 1960
NATINC70-rate of natural increaseof population, 1970
Source: officialdata.
2-TAIL SIG.
0.2393
0.0509
0.0758
0.0252
0.006709
0.012018
0.000634
8.026420
0.004105
47
Table9: SectoralemploymentImbalancesby republic
DeviationJfom employment
sharespredicted by per capita income,1970
Transport
Industry
USSR
Slavic:
Russia
Ukraine
Belarus
Baltic/Moldavian:
Estonia
Latvia
Lithuania
Moldova
Transcaucasian:
Georgia
Armenia
Azerbaijan
Central Asian:
Kazakhstan
Turkmenistan
Uzbekistan
Tajikstan
Kyrgyzstan
Construction Agriculture and comm Trade
3%
1%
6%
2%
-7%
.1%
1%
0%
0%
1%
11%
17%
2%
1%
0%
-7%
.8%
-8%
7%
7%
1%
.9%
2%
0%
2%
-1%
3%
4%
14%
29%
3%
2%
1%
-1%
-7%
-7%
-9%
-9%
-6%
3%
-4%
0%
3%
1%
16%
4%
10%
1%
0%
2%
-8%
-8%
-7%
-4%
-11%
-9%
-9%
-3%
3%
3%
1%
0%
0%
6%
16%
19°%
20%
11%
4%
2%
0%
0%
1%
-7%
-7%
-7%
-8%
-7%
6%
2%
Sources:InternationalLabor Organization(variousyears)for international
employmentdata; see textfor sourceson Soviet republics
Note: Share deviations are calculated by regressing employment shares in 1970 for
non-socialist countries on 1970 per capita income (Summers-Heston), dummying out
the Soviet republics.
48
Appendix 1: Trends in capital-output ratios in growth accounting atdiu
County
Period
Thispaper
USSR (WesternGDP and Capital StockEstimates)
1950-87
Maddison(1989)
France
1950-84
Germany
1950-84
Japan
1950-84
UnitedKingdom
1950-84
UnitedStates
1950-84
China
1950-84
India
1950-84
Korea
1950-84
Taiwan
1950-84
Argentina
1950-84
Brazil
1950-84
Chile
1950-84
Mexico
1950-84
USSR
1950-84
Yowig (1993)
HongKong
1966-91
Singapore
1970-90
South Korea (excludingagriculture)
1966-90
Taiwan(excludingagriculture)
1966-90
Kimand Lau (1993)
HongKong
66-90
Singapore
64-90
South Korea
60-90
Taiwan
53-90
France
57-90
W. Germany
60-90
Japan
57-90
U.K.
57-90
U.S.
48-90
Ela (1992)
Argentina
1950-80
Brazil
1950-80
Chile
1950-80
Colombia
1950-80
Mexico
1950-80
Per
1950-80
Venezuela
1950-80
Perannwnpercentchangein
capitaloutputratio
2.53%
.0.45%
0.16%
-0.91%
0.62%
.0.07%
2.48%
1.54%
0.03%
-0.34%
0.61Y
0.86%
-0.22%
0.50%
3.75%
0.84%
2.79s
3.62%
2.55%
1.11%
1.38%
3.50Y
3.13%
0.68%
1.16%
3.19%
0.68%
-0.19¶/'
0.39f
-0.54%
-0.39%/0
-0.79%
0.44%
1.22%
0.75%
49
I conL)
aceoutldg
studis(Appeudix
ratdosingrowth
Trads in apital-output
Period
Counny
Per annumpercent changein
capitaloutputratio
Chenery, Robinson,_and Syrquin (1986)
Canada
France
Genmany
Italy
Netherlands
United Kingdom
United States
United States
Japan
Hong Kong
Korea
Singapore
Taiwan
47-73
50-73
50-73
52-73
51-73
49-73
49-73
Benhabiband Spiegel (1992) (usingSummers-Hestondata)
1965-85
1965-85
1965-85
1965-85
1965-85
1965-85
75thpercentile of sample (77 countriesin sample,
excludingAfrica)
50th percentile
25thpercentile
UnitedStates
Japan
Korea
Singapore
Taiwan
1965-85
1965-85
1965-85
Nehruand Dhareshwar(1993) (WorldBank data)
1950-90
1950-90
1950-90
1960-90
1950-90
75thpercentile of sample (72 countriesin sample)
50thpercentile
25thpercentile
1950-90
1950-90
1950-90
KI.ngand Levine (1994)(Summers-Hestondata)
1950-88
UnitedStates
1950-88
Japan
1950-88
HongKong
1950-88
Korea
1950-88
Singapore
1950-88
Taiwan
75thpercentile of sample of 74 countriesexcluding
1950-88
Africa
1950-88
50thpercentile
1950-88
25thpercentile
0.66%
0.13%
0.33%
-0.63%
0.17%
0.69%
0.00%
0.63%
2.56%
-0.20OYo
2.78%
2.41%
2.97%
1.72%
0.80%
0.21%
0.20%
2.70%
3.70%
-1.09%
-2.08%
1.64%
1.06%
0.38%
0.40%
2.33%
-0.80/o
3.05%
2.94%
2.63%
1.69%
0.95%
0.23%
squars esImaon of CU h
Appendix2: Nonlinear1WM
Variablename for nHonllnrregressionl
D2839
dummyfor 1928.39
D4049
dummyfor 1940-49
D5059
dummyfor 1950-59
D6069
dummyfor 1960-69
D7079
dummyfor 1970-79
D8087
dummyfor 1980-87
KLINO
official
Capital-laborratio, industry,
KLINW
Capital-laborratio,industry,
Westernest
KLKHAN
Capital-laborratio,Khanin
KLOFF
sectors,
material
ratio,
Capital-labor
oMcal
KLWES
ratio, wholeeconomy,
Capital-labor
Wtstern
Logof outputper worker,industrial LYLINO
sector,officialnumnbers
Logof outputper worker,industrial LYLINW
sector,Westeu estimates
LYLKHAN
Logof outputper worker,material
sectors,Khanin
LYLOFF
Logof outputper worker,material
sectors,official
LYLWES
Logof outputper worker,whole
economy,Westernestimates
T2839
TIME*D2839
T4049
TIME*D4049
TS059
TIME*DSOS9
T6069
TIME*D6069
T7079
TIME*D7079
T8087
TIME*D8087
DEFGDP
Ratioof defensespendingto GDP
TIME
TIME(1,2,3,4,ETC.)
_om
(1) Results for 1928V7 ample
51
NLS //
SMPL
Dependent Variable is LYLKHMAN
range:
-
1928
1987
Number of observations: 60
LYLKHAN-C(1)*T2839+C(2)*T4049+C(3)*T5059+C(4)*T6069+C(5)*T7079+C(6)*T808
7+C(7)*LOG(C(S)*KLKHANA(1/C(7))+1-C(8))+C(9)
.........
m..
mamma
am....
ma.
C(1)
C(2)
C(3)
C(4)
C(5)
C(6)
C(7)
C(S)
C(9)
- -mmmaumumamma.-
a
m
a
amam..
- mm-.-umm
COEFFICIENT
STD. ERROR
T-STAT.
2-TAIL SIG.
-0.0202626
-0.0116552
-0.0017588
0.0033136
0.0030160
0.0022460
9.8304594
0.6102850
0.0916414
0.0073460
0.0034600
0.0025248
0.0029598
0.0031696
0.0029139
113.76212
0.1464251
0.0363921
-2.7583265
-3.3685319
-0.6966280
1.1195481
0.9515282
0.7707873
0.0864124
4.1678994
2.5181703
0.0080
0.0014
0.4892
0.2682
0.3458
0.4444
0.9315
0.0001
0.0150
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
Durbin-Watson stat
0.983100
0.980450
0.055592
93.12259
1.024675
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
0.471133
0.397586
0.157612
370.8535
0.000000
NLS // Dependent Variable is LYLOFF
SMPL
range:
-
1928
1987
Number of observations: 60
LYLOFF -C(1)*T2839+C(2)*T4049+C(3)*T5059+C(4)*T6069+C(5)*T7079+C(6)*T808
7+C(7)*LOG(C(8)*KiOFFA (1/C(7))+1-C(S))+C(9)
COEFFICIENT
ama
a a a a a....a a a a mamma
C(1)
C(2)
C(3)
C(4)
C(S)
C(6)
C(7)
C(S)
C(9)
a
ama
mamma
0.0333843
0.0218378
0.0296170
0.0296679
0.0304758
0.0297401
-0.6260722
0.5403918
1.3706613
.
R-uquared
Adjusted R-squared
S.E. of regression
Log likelihood
Durbin-Watson stat
a mmmaa
aa amama.
STD. ERROR
mum mmma
mama
am
a mama
T-STAT.
a a aa a a mamma
0.0080809
0.0044657
0.0047890
0.0059473
0.0058937
0.0054773
0.1706363
0.1322528
0.1645207
a a
4.1312477
4.8901520
6.1844430
4.9884344
5.1708799
5.4296908
-3.6690441
4.0860527
8.3312377
0.996612
0.996080
0.066469
82.40014
1.101054
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-scatistic)
a
a
a
a
amm,
a
amm
mmmaa
2-TAIL SIG.
mammaS a..m..mamma.a
mmmaamaa
a a
0.0001
0.0000
0.0000
0.0000
0.0000
0.0000
0.0006
0.0002
0.0000
2.117937
1.061674
0.225327
1875.115
0.000000
a
ma
a
aassssssst=s
NLS // Dependent Variable is LYLWES
52
Date: 7-20-1993 / Time: 23:35
SMPL range: 1928 - 1987
Number of observations: 60
LYLWES -C(1)*T2839+C(2)*T4049+C(3)*T5059+C(4)*T6069+C(5)*T7079+C(6)*T806
7+C(7)*LOG(C(8)*KLWES^(1/C(7))+1-C(S))+C(9)
Convergence achieved after 2 iterations
mumm-m
m.mu......mu.
m....
COEFFICIENT
STD. ERROR
u.mu
u.-
r..
U.UUU
mm......................
mum-- ur......-
T-STAT.
2-TAIL SIG.
m..........m...m.m......................n.m
m...u..................m......
C(1)
C(2)
C(3)
C(4)
C(S)
C(6)
C(7)
C(S)
C(9)
-0.0051988
-0.0131666
-0.0020808
-0.0014649
-0.0018068
-0.0033362
-4.1327746
0.7331452
-0.2293690
mm...un...
mu....
ua
0.0086239
0.0045147
0.0048916
0.0058862
0.0059853
0.0060313
5.8211962
0.1523563
0.0861945
mmm...m...
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
Durbin-Watson stat
m....
0.5493
-0.6028340
-2.9163642
-0.4253758
-0.2488673
-0.3018751
-0.5531465
-0.7099528
4.8120442
-2.6610638
ma mm
m..
0.0033
0.6724
0.8045
0.7640
0.5826
0.4810
0.0000
0.0104
m
Mban of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
0.989484
0.987835
0.060641
87.90639
1.118814
u......
mm
0.934737
0.549799
0.187543
599.8574
0.000000
NLS // Dependent Variable is LYLINW
SMPL
range:
-
1928
1987
Number of observations: 60
LYLINW -C(1)*T2839+C(2)*T4049+C(3)*T5059+C(4)*T6069+C(5)*T7079+C(6)*T803
7+C(7)*LOG(C(8)*KLINwA (1/C(7))+1-C(S))+C(9)
mm..
m
mm.m....
m.
C(1)
C(2)
C(3)
C(4)
C(S)
C(6)
C(7)
C(8)
C(9)
,,,,,m.u
m.....
R-squared
Adjusted
R-squared
S.E. of regression
Log likelihood
Durbin-Watson
stat
u
mmm....
mm.m..
mm.......
rn.....mm..m.m.m...
mm....mm.....m.....
mm......
m.
m
COEFFICIENT
STD. ERROR
T-STAT.
2-TAIL SIG.
-0.0137604
-0.0071615
0.0035802
0.0040201
0.0043616
0.0036720
-0.2743993
0.3542810
-1.2839549
0.0079630
0.0038608
0.0026946
0.0038213
0.0051544
0.0054903
0.3211317
0.3325082
0.3090338
-1.7280338
-1.8549179
1.3286169
1.0520246
0.8461905
0.6688182
-0.8544760
1.0654807
-4.1547401
0.0900
0.0694
0.1899
0.2977
0.4014
0.5066
0.3968
0.2917
0.0001
mm.-.-m---mmmm....m...mum.
mmm.mum.....
0.979397
0.976165
0.094259
61.44156
1.034563
m.m...mm...m...
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
mm...m..mu..
-1.911255
0.610544
0.453126
303.0437
0.000000
53
NLS // Dependent Variable is LYLINO
Date: 7-20-1993 / Time: 23:38
1987
SMPL range: 1928
60
Number of observations:
LYLINO -C(1)*T2839+C(2)*T4049+C(3)*TS059+C(4)*T6069+C(S)*T7079+C(6)*T808
)+1-C(8) )+C(9)
7*C(7)*LOG(C(8) *KLINOA (1/C(7)
achieved after 2 iterations
Convergence
uummmmmmmmmmmmmmmmmmmmmmmminmummmummuummmmmmmminu=uuinmmuuminumumummuummmm
C(1)
C(2)
C(3)
C(4)
C(5)
C(6)
C(7)
C(8)
C(9)
s---
-----------
mwmmmum
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
stat
Durbin-Watson
COEFFICIENT
STD. ERROR
T-STAT.
0.0072341
0.0051121
0.0148348
0.0126798
0.0137831
0.0128241
-0.8243762
0.4472637
-1.4963393
0.0086047
0.0053042
0.0050212
0.0064987
0.0069631
0.0069723
0.3706599
0.1562502
0.37q1015
0.8407205
0 9637879
2.9544124
1.9511147
1.9794368
1.8392831
-2.2240773
2.8624840
-3.9470674
m
2-TAIL
SIG.
0.4044
0.3397
0.0047
0.0565
0.0532
0.0717
0.0306
0.0061
0.0002
mmmm=mmmmuummmmmmmmmmmmmmmuu-w.Sumi==nmmu==
Mean of dependent var
S.D. of dependent ver
Sum of squared resid
F-statistic
Prob(F-statistic)
0.996675
0.996154
0.068948
80.20302
1.317886
-2.191933
1.111757
0.242448
1911.111
0.000000
(2) Results for 1950-87 4mmpIe
NLS // Dependent Variable is LYLKHAN
1987
SMPL range: 1950
38
Number of observations:
LYLKHANmC(1)*T5059+C(2)*T6069+C(3)*T7079+C(4)*T8087+C(5)*LOG(C(6)*KLKHAN
A(1/C(S))+1-C(6))+C(7)
C(1)
C(2)
C(3)
C(4)
C(S)
C(6)
C(7)
mmmm...................
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
stat
Durhin-Watson
COEFFICIENT
STD. ERROR
T-STAT.
0.0011270
-0.0006563
-0.0030198
-0.0034716
9.7629283
1.1237026
-0.1735591
0.0048291
0.0051628
0.0050557
0.0045055
220.16876
0.2228861
0.1109200
0.2333864
-0.1271136
-0.5972997
-0.7705352
0.0443429
5.0415996
-1.5647225
m mu mu
u.............mum..............mum...........
0.987796
0.985434
0.036836
75.39730
0.505965
u
u
=. mum...mu
m
m
2-TAIL
0.8170
0.8997
0.5546
0.4468
0.9649
0.0000
0.1278
= mum.....mu
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
mum mum mmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm...........m...mm.m,
mum...........
mmm
SIG.
m..
mum,.
0.707439
0.305213
0.042064
418.1935
0.000000
m....mmm....mm.m
54
NLS // Depenoent Variable is LYLOFF
SMPL range: 1950 - 1987
Number of observations: 38
LYLOFF -C(1)*T5059+C(2)*T6069+C(3)*T7079+C(4)*T8087+C(5)*LOQ(C(6)*KLOFF*
(1/C(5))+1-C(6))+C(7)
C(1)
C(2)
C(3)
C(4)
C(S)
C(6)
C(7)
COEFFICIENT
STD. ERROR
T-STAT.
2-TAIL SIG.
0.0292335
0.0288169
0.0297372
0.0291584
-0.5812015
0.5715239
1.3912334
0.0027887
0.0029999
0.0030072
0.0028214
0.0565917
0.0456052
0.0869462
10.482690
9.6060223
9.8888315
10.334843
-10.270078
12.531981
16.001090
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
Durbin-Watson stat
na.....
nm.....
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
0.999488
0.999389
0.014707
110.2862
1.390053
m.
ann
mmnam
maanmn
2.812825
0.594913
0.006705
10084.79
0.000000
aama,amm....
an
--.-.
NLS // Dependent Variable is LYLWES
SMPL range: 1950 - 1987
Number of observations: 38
LYLWES =C(1)*T5059+C(2)*T6069+C(3)*T7079+C(4)*T8087+C(5)*LOG(C(6)*KLWESA
(1/C(5))+1-C(6) ) +C(7)
C(1)
C(2)
C(3)
C(4)
C(S)
C(6)
C(7)
m..a
mmm
.mmnan
COEFFICIENT
STD. ERROR
0.0109709
0.0109651
0.0116165
0.0109505
-0.5958245
0.9598616
-0.8162872
0.0032629
0.0035797
0.0035835
0.0034259
0.0966218
0.0134483
0.0832581
m.mnu.a
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
Durbin-Watson stat
a..anm....
mna...
m.
mm..
0.998747
0.998505
0.013777
112.7701
1.922959
m...am....
m
m......
T-STAT.
2-TAIL SIG.
3.3623384
3.0631462
3.2417041
3.1963828
-6.1665625
71.373976
-9.8042949
aaa....,
0.0021
0.0045
0.0028
0.0032
0.0000
0.0000
0.0000
m......
mm....
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
n...
a..aa...
m,.
..
m....,
ns
1.285964
0.356312
0.005884
4119.769
0.000000
naa
wm
SMPL
55
Dependent Variable is LYLINW
NLS //
range:
-
1950
1987
Number of observations: 38
LYLINW .C(1)*T5059+C(2)*T6069+C(3)*T7079+C(4)*T8087+C(S)*LOG(C(6)*KLINWA
(1/C(S))+1-C(6))+C(7)
STD. ERROR
COEFFICIENT
.m...
0.0021467
0.0023585
0.0021509
0.0018522
0.0330823
0.0031944
0.1045291
0.0240169
0.0236218
0.0251244
0.0243103
-0.1441544
0.0020469
-2.4449147
C(1)
C(2)
C(3)
C(4)
C(S)
C(6)
C(7)
2-TAIL SIG.
m.........
11.187968
10.015738
11.681057
13.124830
-4.3574480
0.6407644
-23.389801
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
0.997683
0.997234
0.021971
95.03323
0.506999
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
Durbin-Watson stat
T-STAT.
m......mm...m.........
..............
..........................
-m..........
m...fu....
.m....
-m.
n....mm.
------------------
0.0000
0.0000
0.0000
0.0000
0.0001
0.5264
0.0000
-1.535708
0.417767
0.014965
2224.312
0.000000
NLS // Dependent Variable is LYLINO
SMPL
range:
-
1950
1987
Number of observations: 38
LYLINO -C(l)*T5059+C(2)*T6069+C(3)*T7079+C(4)*T8087+C(5)*LOG(C(6)*KLINO^
(1/C(s))+1-C(6) )+C(7)
STD. ERROR
COEFFICIENT
0.0049959
0.0053985
0.0054797
0.0052913
0.1296726
0.0669595
0.2907247
0.0371878
0.0359650
0.0373821
0.0361438
-0.6618038
0.1022757
-2.7672364
C(1)
C(2)
C(3)
C(4)
C(5)
C(6)
C(7)
................
m_...
m...mmunau....
m..u....m
m...
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
Durbin-Watson stat
mm..um
..
u....m..
0.999228
0.999079
0.019953
98.69529
1.008779
.m.m.mu.mUu
T-STAT.
2-TAIL SIG.
mm.,.,mm.
m...am......
7.4436178
6.6619911
6.8219713
6.8308024
-5.1036531
1.5274265
-9.5184097
0.0000
0.0000
0.0000
0.0000
0.0000
0.1368
0.0000
m.m.m....................mm...................mum
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
-1.481138
0.657494
0.012342
6690.964
0.000000
56
(a) Regrsian
wth Defense qpmndfng/rnP
NLS // Dependent Variable is LYLWES
SMPL
range:
1960
-
1987
Number of observations: 28
LYLWES-C(1)*TIME+C(2)*TIME*DEFGDP+C(3)*LOG(C(4)*KLWESA(1/C(3))+1-C(4))+C
(5)
mmm
u
mm.....
m...
u
mN--m..
m.....m..
mm mmm mmmm am a
COEFFICIENT
STD. ERROR
mm,
mmmm.m.m.,
T-STAT.
mm
2-TAIL SIG.
mmmmmmmmmam...ammmmmmmiam.mmmmmmmammmmmmm
C(l)
C(2)
C(3)
C(4)
C(5)
0.0207134
-0.0727455
-0.5785414
0.9690962
-0.8963781
R-squared
Adjusted R-squared
S.E. of regression
Log likelihood
Durbin-Watson stat
mm...
m.......
am.......
a mam
0.0059611
0.0108381
0.2470464
0.0396033
0.2645463
0.996629
0.996042
0.011310
88.52034
2.232942
mmm
ms.m.m
mm.
...
3.4747771
-6.7120112
-2.3418330
24.470094
-3.3883604
Mean of dependent var
S.D. of dep)endent var
Sum of squaied resid
F-statistic
Prob(F-statistic)
m.
amimammmmma..ammmmmmm
0.0021
0.0000
0.0282
0.0000
0.0025
1.468442
0.179790
0.002942
1699.821
0.000000
Pollcy Research Working Paper Series
Title
Author
Date
Contact
for paper
J. MichaelFinger
WPS1264 A Rockand a Hard Place:The Two
Facesof U.S.Trade PolicyTowardKorea
March1994
M. Patenia
37947
MiguelA. Kiguel
WPS1265 ParallelExchangeRatesin
DevelopingCountries:Lessonsfrom StephenA. O'Connell
EightCase Studies
March1994
R. Luz
34303
WPS1266An EfficientFrontierfor International SudhakarSatyanarayan
PanosVarangis
Portfolioswith CommodityAssets
March1994
D. Gustafson
33732
WPS1267The Tax Basein Transition:The Case Zeljko Bogelic
Arye L. Hillman
of Bulgaria
March1994
F. Smith
36072
ElianaLa Ferrara
WPS1268The Reformof Mechanismsfor
3abriel
Castillo
Allocation:
Theory
ForeignExchange
John Nash
and Lesconsfrom Sub-Saharan
Africa
March1994
N. Artis
38010
WPS1269Union-NonunionWageDifferentials
in the DeveiopingWorld:A Case
Study of Mexico
Alexis Panagides
HarryAnthony Patrinos
March1994
I Conachy
33669
WPS1270How Land-BasedTargetingAffects
Rural Poverty
MartinRavallion
BinayakSen
March1994
P. Cook
33902
March1994
M. Divino
33739
F. DesmondMcCarthy
WPS1271Measuringthe Effectof External
Shocksand the Policy Responseto J. PeterNeary
Them: EmpiricalMethodologyApplied GiovanniZanalda
to the Philippines
WPS1272The Value of SuperlundCleanups:
Evidencefrom U.S. Environmental
ProtectionAgency Decisions
ShreekantGupta
GeorgeVanHoutven
MaureenL. Cropper
March1994
A. Maranon
39074
WPS1273 DesiredFertilityand the Impactof
PopulationPolicies
Lant H. Pritchett
LawrenceH. Summers
March1994
P. Cook
33902
WPS1274The NewTradeTheoryand Its
AsadAlam
Relevancefor DevelopingCountries
March1994
A. Alam
87380
RicardoBarros
WPS1275 Female-HeadedHouseholds,
Poverty,and the Welfareof Children LouiseFox
in UrbanBrazil
March1994
K. Binkley
81143
WPS1276 Is There Persistencein the Growth AshokaMY'y
of ManufacturedExports?Evidence KamilYilmaz
from NewlyIndustrializingCountries
March1994
M. Patefia
37947
WPS1277 PrivateTrader Responseto Market StevenJaffee
Liberalizationin Tanzania'sCashew
Nut Industry
March1994
C. Spooner
32116
Policy Research Working Paper Series
TitIS
Author
Date
Contact
for psper
WPS1278 Regulationand CommitmentIn the
AhmedGalal
Developmentof Telecommunications
in Chile
March1994
B. lMoore
38526
WPS1279 OptimalHedgingStrategyRavisited: YingOian
Acknowledgingthe Existenceof
RonaldDuncan
NonstationaryEconomicTime Series
March1994
S. Lipscomb
33718
WPS1280The EconomicImpactof Export
WendyE. Takacs
Controls:An Applbcatlon
to Mongolian
Cashmereand RomanianWood Products
March1994
M. Pateha
37947
WPS1281 Human and PhysicalInfrastructure: EmmanuelJimenez
Public Investmentand PricingPolicies
in DevelopingCountries
Aprl 1994
L. Longo
37786
WPS1282 Copperand the NegativePriceof
Storage
April 1994
A. Kim
33715
WPS1283 InterestRates in Open Economies: Dipak DasGupta
Real InterestRate Parity, Exchange BetoyDas Gupta
Rates,and CountryRisk in Industrial
and DevelopingCountries
April 1994
B. Kim
82467
WPS1284 The SovietEconomicDecline:
Historicaland RepublicanData
April 1994
R. Martin
31320
Donald FrederickLarson
WilliamEasterly
Stanley Fischer