Created: 6/1/1974

OCR scan of the original document, errors are possible

wcia historical review program . release as sanitized

Estimating Military Hardware Production from Soviet Industrial Data


June Copy Nn

f i1



I The Soviet Union nat not published details of its expenditures on national defense since. National income snd budgetary data released by the Soviet government since Wortd War II have provided only one summary figure for Defense and published data on -industrial production have not explicitly included the output of the defense industries. There have been large, unexplained residuals in official Sov,et budgetary and national income data, however, and authoritative Sov.ct statements have suggesled lhat defense expenditures arc included in apparently civilian budgetary and national income accounts. '

estern interest in determining the sue and trend of Soviet defense spending ongsnally stemmed from the issues of Soviet military capability during the post-World War II period. In thehe question or using finance! daia to monitor an arms control agreement heightened the interest in cstaNish>n< the magnitude of Soriet defense expenditures.

eport prepared by the Stanford Research Institute (SRI)8 fo. Ihe Arms Control andurveyed the feasibility andhe methodologies to beomlonne an arms limitation agreement from financialhis report developed several defense-related residuals in an attempt lo estimate Soviet military hitdwate procuicment. The estimate, in ihe SRI report (the work ofc) gained additional exposure whenmtncr introduced them at the hearings of the Joint Econonm Committee of the Conpesshen,ompendium ofubmitted to the Jointotnmiiiceeport on "The Technological Base of Soviet Military Cower" inhael Borctsky presented lus estimates of military hardware production, which were also derived from residual industrial data. More1 SRI publication' included an update ofstimates, androup of SRI economists and consultants prese-itcd

- ae Ah-,

1 WT SW- *,

i , IVl. UNTlASSiriED

Note: Comments and queries retardme this publication are welcomed They may be directed

estimates atlriuutcd to Leeaper prepared foi the Joint Economic Committee of the Congress.4

his publication assesses these attempts to estimate the production or procurement of defense hardware from official data on the machine building and metalworking (MBMW) branch of industry. The reasonableness of the basic assumption-oMhe Lee and Boretsfcy sftdtcs ^thatproducnorrof defense rrardwafT is included in official MBMWs examined first. Then, the validity of estimates made by Lee. presently at GeneralEMPO, Center for Advanced Studies, and Boretsky. of the US Department of Commerce, is assessed.inal section, the lessons learned from the review of Ihe work of LeeSky 'n'n8machinery outputnd


Several estimates of Soviet military hardware production have been basedidely acceptedhat Soviet data on the MBMW sector of the economy include the output of the defense industries. Soviet sources as well as Western analysis of Soviet data support this proposition. Thus, defense production should be obtainableesidual of MBMW output after subtracting estimates of non-defenso production.

Published Soviet data on MBMW output, however, do nottraightforward determination of defense production by subtraction of non-defense output. Western analysis have tried lo use Soviet data on total machinery production and the purchases of machinery for crvilian uses to obtain defense purchasesesidual. Data on neither production nor purchases can bf used without adjustment because they are reported in different prices, and Soviet descriptions of economic data are so imprecise thai substantial error remains in any adjustment. Moreovei, Soviet data collected for specializedoreign trade or interindustryusl be introduced into the calculations because of the paucity of published Soviet data on uses of machinery. Relhnce on such data, compiled from different statistical systems and in different prices, weakens the estimates lo an uncertain degree. Abo. published Soviet growth indexes, which contain well-known sources of possible upward bus. must be used. The use of such indexes, of course, will distort trends in the residual. Finally, in using data on production and purchases to derive residual machine building (MB) purchases. Western analysts must deal with the lime lag between production and sale.

Two quite different calculations of an MBMW residual show in sharp relief the ^conclusiveness of the residual technique.8 SRI publication, William T. Lee estimated Soviet procurement of military hardwareesidual calculation. Michael Boretsky0eries of military hardware ' MM Evfc. fa.ud Ftxto,thototofMol


tell duals ihal resembled Lee's residuals conceptually, but whose values were substantially different from the Lee estimates. Indeed. Lee and Boretsky derived quite different values for each of the important steps of Ihe residual processthe calculationross value of output ofinal output of MB.he presumed military residual left after civilian uses are deducted from the

output of MB. as follows:

Cross output of machine building

Final outpul of machine building


Lee ran afoul of most of the pitfalls cited above. He mixed data prepared for specialized interindustry studies with other data prepared for the Soviet Central Statistical Administration (CSA) without reconciling the two sets of data for differences resultinghe accounting concepts used (CSA data include some non-machinery production, but does not include all machinery produced)he prices used (retail prices versus factory or producern estimating the tiend in his defense-related residual, moreover. Lee did not use Soviet growth indexes where their application seemeds inime series for the gross output of MB and for consumernd instead created his own indexes, Lee also erred in subtracting all unfinished production from annual gross output of MB instead of merely the change, and in failing to deduct MB shipments to other industries, which are not counted as "final" output His separate estimate of Soviet rcscarch-and-dcvelopment expenditures is also suspecte did not include social security payments in production costs,is assumption that MBMW employment data excludeorkers is open lo question.

ichael Botelskyethodology similar lo Lee's in obtaining his estimate of Soviet spending on military and space programs. Boretskyeasure of total machinery output in the economyingle benchmark year. The estimates for olhcr years were then made using the CSA index of MBMW output, which includes some non-machinery outpul and which does not include all machinery outpul. Moreover, the gross value of output for MBMW which Boreisky used9 is inconsistent with the values of MBMW output reported tn Soviet sources.

oretsky eliminated indirect taxes and tiadc-disliibulion markups from producer and consumer durables to pul machinery production and civilian purchases on the same price basis, but his adjustments were incurred, lie also used data that included imports in order lo make estimates from data that did not include

he prospecls for using residual output of MB to determine Soviet expenditures on military hardware are therefore dim. While some of the estimative and accounting errors in the work of Lee and Boretsky can be corrected, data are still insufficient and Soviet explanations of their national statistics still too vague to enable estimates to be made with confidence. Crude estimates will continue to be needed in calculating the residual.mall percentage error in such crude estimates made al high levels ofn deducting metalworking output from total gross output ofan leadery large percentage error in the much smaller, supposedly defense-related residual. In short, the data and informationoo formidable with present information to producelausible value or trend of military-space hardware outpul from Soviet data on the MBMW scctot of industry.


The Basicefense Production and National Economic Accounts

estern efforls lo use Soviet national accounts to obtain estimates of defense spending assume that these accounts include defense expenditures. The rationale for this assumption is as follows: Soviet national accounting practice follows the Marxist division of economic activity into spheres of material and non-material production. Accordingly, the Soviets claim that their statistics on "national income" include only materialhe production of tangible goods, as opposed to the production of intangible services. Moreover, all material production is presumed tooviet national income. Asa result, the production of uniquely military hardware (missiles, ships, tanks,s therefore assumed to be included in the sphere of material production and in official Soviet data on national income.

Ihe Soviet statistical construct "national income" apparently docs include only material product. The derivation of this construct in the Soviet statistical yearbooks and Other Soviet sources includes Consumption and accumulation, subtotals which arc consistent with other Soviet data on retail sales of material' products and capita! investmenl

n addition, the assumphon that all material production is included in published Soviet national income data seems acceptable. The CSA could compile Soviet economic data net of defense production. But this argument would imply that the CSA keeps two sets of national accounts and withholds Ihe set including defense production from publication. Mosl Western observers have concluded thai the CSA docs not maintain iwo such sets of books, bul rather simply does not publish sensitive, defense-relatedence, the existence of large, unexplained residuals in Soviet data and the occurrence of apparently irreconcilable data.

LT Ab'U,BmCiWomi. hew.

iv6v. p. is'. UNCl)

Soviet descriptions of the contents of national accounts and the administrative subordination of the defense industries lo the MB ministries suggest lhal defense hardware production is included in MBMW, the USSR's largest industrial branch. Major support for this supposition comes from the list of MBMW industries enumerated in the forms used for drawing up the seven-year* In this classification,ranch of MBMW. includes warships, and another branch designated "defense industry" includes MBMW enterprises for the "production and repair of military equipment and munitions, including all aircraft plants but excluding naval shipyards, and specialized plants for the production of instruments andn the classification, It is also noted that general explosives and some nuclear materials for military use are included in Ihe chemical industry. Except for shipments to stockpiles, these materials should be embodied in the final output of the warhead assembly plants included in MBMW.

That the published Soviet dala on gross value oulput (GVO) of MBMW do, indeed, include output of military hardware is supported by the behavior of the MBMW index in the immediate post-World War II years. The output of MBMW fell. This sudden decrease (comparedncrease) probably reflects the drop in military-related production as the economy retooled for civilian production.

hus the published MBMW index almost CCi(airily includes the production of military hardware. Insofar as the reported gross outputs are consistent wilh ihe published growth indexes, values for the GVO of MBMW would also include Ihe value of military hardware. If Ihe GVO of MBMW includes Outptil of the defense industries, it should then be possible lot >Oine level of accuiaCy -ihe production or procurcmenl of uniquely military products of MBMW by subtracting ihe value of civilian production from total MBMW production.

Lee's Estimateachinery Purchases Residual

tanford Research Institute teporl by William T. Lee8 analyzed the possibilities of financial verification of Sovicl militaryne of (he approaches to verification used in lhal studyachinery purchases residual (MPR) (hat has also appeared in subsequent SRI publications

Military space procurement of durables (hardwaicl is estimatedesidual in the final value of oulput of the Soviet machine building and metal working industries "

I j:i i .

thoiyaritHi SSSHody,ince tins puttbonon wis ippmvcd. iddiiwiui official eiissir.fc.iiom hive irrtuded defense indusiiiesheii dcsciiptiofts ol ihe MBMW bnnch: Covins. Me/wftrfceittye ukoinmyv.. Guryev.huuyoortnot* Hhaiyayim SSSR..

op. tU.

8. i. tit. Sonei milOmtl Sfturily. op. ill

A paper prepared by several SR1 economists and consultants foi3 compendium of the Joint Economic Commillce of the Congress also included estimates of total defense spending attributed lo Lee.'

The methodology that Lee used was based on official Soviet growth indexes of the output of the MBMW sector. These indexes contain well-known sources of possibleew product pricing'o and changing amounts ofhich can distort trends. The first step in the methodology after converting the indexes into values, called for the subtraction of the GVO of melalworking (including repair) from the GVO of MBMW. Hie resulteries of GVO for MB which arc then adjusted for double-counting and changes in unfinished production to estimate the final outpul of the MB branch. The final outpul series represents output available for investment, consumption, export, and defense. Ice subtracts his estimates of producer and consumer durables from final output to obtain the residual. This residual, which is assumed to be defense-related production, growsillion rubles% of the GVO of MBMW)billion rubles% of the gross output of MBMW).

The purpose of this section is to evaluale Lee's methodology in some detail.resents each step, and the discussion which foUows considers these steps under three general headings: (a) the denvation of the GVO of MB from the GVO of MBMW: (b) Ihe derivation of the value of the final outpul of MB: and (C) the subtraction of producer and consumer durables toilitary hardware residual.

fXrivation of the Gross Value of Output of Machine Building

Lee found official Soviet estimates oi the GVO of MBMW5 prices5 and calculated values for other years by means of ihe Soviet index of growth in MBMW gross output Next, he subtracted annual estimates of output of Ihe melalworking sectors (metal articles and capital repair) to obtain the gross outpul of MB.

The time series for ihe GVO of MB. which Lee estimates by this methodology, contains many errors which together result in an overstatement of the value of the GVO of MB rangingillion nibles5illion

*. RobtM W. Ca-pbel. m. at. a*c-rtiiu priorl

*u.ccenturtrd br (he

hit ol iKCl ( or>dUrtion nmi

and miiim Ike-tueh would ippir aflti theNrtM pioduclion. Tliehi powih

II. Inoru.n, aptchlllaliM wdhin ihrl*l, roulci in in* lh*of*kioi of ihe induiliy connii ihe mine ot thcten lit okuutior,mu(?

<hcm i. of it* mmmof ih* mtoi.




mi im

mi m

ml in


rublespecifically, Ue failed lo recognize ihe differences between the commodity-based Soviet data used in theirnd (he establishment-based data used in other Soviet reporting systems; he subtracted turnovernd trade and distribution markups and transpoilalion costs in the meialworking industries from MBMW which did not include such markups, and heublished Soviet index of growth in MB and created an index of his own.

Commodity-Establishment Problem

An input-out put iibfe iitool otanalyUi wht'h oepmi the flow ul twod! and nrrrSwi beiwNn productiveiuu.

Tlicimmini lory indued laa ox Had* lu/oo.ei levied (hiaflr on coniumei foodi

he values for GVO of MBMW that Ue uses are compiled by Soviet statistical agencies on an "establishment" basis. The GVO ofhe sum of the value of output of all enterprises whose primary production is machinery, metal articles, or machinery icpairoes not include machinery producedecondary product in other branches of the economy. These data do include, however, the value of non-machinery products produced in MBMW. Commodity-based data, on the other hand, classify data mio simdar product groups. Thus, machinery produced outside the MB industries would be included with (he

muchincry output of MB industries, bul the non-machinery output of MB industries would not be included. To convert GVO of MBMW from an establishment basis to abasis, non-MBMW production must be removed from the establishment output and MBMW-type production carried out elsewhere must be added. Two Soviet authors have slated that9 commodity-based gross output of MBMWf the establishment-based grosshis was the result of removingf the establishment-based gross output that represented non-MBMW products and adding secondary MBMW output equal% of the establishment-based gross output.9 the commodity-based gross output of MBMW can be calculated2 billion rubles. Lee's estabUshment-hased value6 billion rubles, therefore, overstates Ihe gross output and Ihe resulting residualillion rubles.

the commodity-establishment adjustment involvesubtraction, the possibility that it varies over time isnet adjustment9 rs the only one known, thefor other years have not been published. The inability of alo make the proper cominodity-estabJishment adjustment over timestumbling block in this type of methodology.

Purchaser rs Producer Prices

Soviets have never released data on the value of output of theof MBMW. To reduce his MBMW series to MB output alone, Leegross outputs of the metalworking and repair sectors available in thetable. But the gross outputs in theables are in purchaserhence are not comparable with the gross outputs reported in constantprices- Therefore, when Lee fails to adjust9 GVOfor the turnover tax collected and the transportation, trade,costs incurred in marketing the product, he is subtractingalue that is atillion rubles too large.

Incorrect Index for Metalworking GVO

M.lnft/iAcunp.ILVi. Bern, edi. arVwf* ptmwiiravopertn.HS.

runci ass mm




an index of metalworking output exists for all" Lee used9Serived frombetween the MBMW and MB indexes reported in the Soviet- and assumed lhat the index grewonstant rale in theThis index was applied9 base value for three MWable, thai include sanitary engineering equipment which may not be in


melalworknig This procedure distorts the ticnd in the GVO of MW that Lee estimates85 and therefore distorts the residual

The Problem in Capital Repair

he index of "production of repairhich is part of MBMW. is published in the Soviet statistical yearbooks, but Lee could find no ruble valueenchmark. Instead, he used the commodity-based gross output for "repair of all machinery" fromable as his benchmark and applied the establishment-based growth index to this value touble series for "production of repairhe coverage of the two repair branches is not comparable and the differences go beyond the commodity-establishment adjustment The establishment-based gross output of "production of repair work" isillion rubles less than the value Lee used for the benchmarkhis means the residual is understated by this amount

The Indexes for MBMW and MB

establishment-based GVOs which lee estimated0 andbe in error on their own account. The GVO of MBMW is based onindex published in official yearbooks The price weights of thischanged over time as the Soviets have linked constant-price segmentsprices used haveurrent wholesale prices2 prices5 prices7 prices. The error introduced by the useprice weights cannot be detcimined but could be sizable, particularlyIn addition, any extension of the series7 could alsoerrors. Lee's procedure for converting the published MBMW index intoresulted in overestimates for the value of MBMW output in allhe overestimates rangedillion rublesillion rubles4 and were the result of using the Soviet values5 (which were published with two-digit accuracy) as thethe index rather than58 Soviet values (which werethree digithe sum of Lee's errors in estimating the GVOcan be seen in Tablehere Lee's series and the index derived fromeries based on the official index. The indexes growin several years, and the value error resulting from errors inGVO of MB growsillion rubles5 to

Derivation of the Value of Final Output of Machinebuilding

moving from the GVO of MB lo an estimate of the value ofof MB, it is necessary to subtract the changes in the amount of unfinished

lT"implied br th* powifi MMtl at CVO ol MB. CVO of nwuJ wort rot, aad GVO of ntpek utfii thai Um ntoc at npm iamma lobtn.

II. Statnl Ihr tin indxakd by ukr official inoti. piodvcei<ii ndhich (fund lo Ihr raluei pubhihed by the SoOcls fat Hick yean

TiMe 2

Compuiio- ol Lee and Official Sonet Machine Buildine Indcae* and Vahiei

illion Ruble.)


1 VVmnmlmw* ,to, lu niwCVO or MfeU w

Id TI, foitheolumn ?.

S- tonected Lee value leu oHIcujI

production and the value of transfer, between enterprises. Ue. however overestimates unfinished production and undent unites interindustry transfers The net result is an ovenlalcmcnt of final oulputillion rubles9 that is then carried down to the residual

77ie Treatment of Unfinished Production

ecause MB gross output includes (he increase in the value of unfinished production during the year, Ue makes adjustments io exclude such changes In dotng so. however, he overestimates the necessary adjustments byillionillion rubles.9 the gross output of MB fromillion rubles, from which Ue8 billion rubles, the total costinished production on hand a( the end of ihe year. He should have subtractedillion rubles, ihe difference bc.wccn (he value of unfinished production at (he end8 and the endimilarly, Ue should have sublracted onlyillion rubles0 instead7 billion rubles5 the oversiatemenl of tins adjustment amountsillion rubles.

l hum

Interindustry Deliveries

Lee estimates that transfers between machinebuilding enterprisesercentage of the gross output of machinebuilding varyn thehese figures are based on fragmentary MBMW data for thehich indicate that the rate was constant in that period and accounted for no moref MBMWcc notes that "the most precise figure available9 which is based on data from the: interindustry matrices."'* Nevertheless, Leeigureor the9 The percentages wtuch appear inre based on this single observation and Lees notions as to the direction of change. Lee believes that intraindustry deliveries have increasedroportion of total output because the proportion of material purchases in MBMW prime costs has risen.

The intraindustry data used by Lee was taken from the published incomplete version of9 matrix which omitted two MBhe percentages of mtcrplant transfers in the completed version of the table are-% for MBMWor MB. Both percentages exceedimit inferred by Lcc from the data in the published tabic. Data fromable, however should not be directly used in making this adjustment because of the commodity-establishment problem and because input-output data reflect purchaser pnees rather than producer prices. The net effect of these differences wouldorrection for double-countingnstead ofmplied bv the input-output *

The error in calculating mtcrplant based on the incorrect ratio would have been significantillion rublesf Lcc had not subtracted far loo much unfinished production from gross output before applyinp the percentage of interplant transfers. Because changes in unfinished production almost ccrtatnly were included in gross output when the base year percentage was calculated, this procedure led to an understatement of interplant deliveries. Because of these errors the error in the value derived by Lee for interplant deliveriess small In other years, however, the errorn overestimateillion rubles

Ixc also failed to take account of deliveries of MB enterprises to other sectors of the economy. This value is available4 billion rubles) and7 billion rubles) from the reconstructedableshis correction alone would reduce the Lee residualillionillion rublesis oversight is partly offset by errors in estimating deliveries among MB cntcrpnscs and in handling the value of unfinished production. Thus9 Lee does not deductillion rubles of inlcrinduslry deliveries, but he


19. Ibid.

cleuly mu4 io th* tource urn but an be deduced fromol

h*ili tbe published

ht Srrucrurt ,keconomyhtTeblr. New Yoik, TD*

illion rubles more than he should have in lhc ad)uslmenl for unfinished production.

Subtraction of Producer and Consumer Durables

final steps in the derivationachinery purchases residualsubtraction of estimates of deliveries of producer and consumer durablesthese estimates. Lee adjusts Ihe producer durables component ofimports but not for its non-machinery element or for the trade,transporlation charges, which are also included. The chief problem inof consumer durables is lhat he accepted an estimateingleand relied on arbitrary growth rates wiih no link to Soviet data toseries.

The Producer Durables Adjustment

Ue uses the value of equipment included in published data on capital investment to estimate domestic production of producer durables. Lee notes that Soviet investment data include (I) some non-machinerymports,rade, distribution, and transportation costs. All three elements should be removed to derive domestic production of producer durables in producer prices. However. Ue removes only the surplus of imports over exports, converted to domestic prices. This procedure removes imports from the investment data and implicitly includes exports in the producer durable category so thato components of final use are subtracted logcther. Ue explains his failure to remove the other elements by staling that these "adjustments are purposely not made in an attempt to keep the producer durables entry conservatively on ihe high side so as not to exaggerate the military/spacee acknowledges that the difference between producer and purchaser prices overstates the value of equipment in capital investment by" accordingoviet source In addition, non-machinery products may account for at leastf equipment data. According lo these crude factors. Lee may have understated his residualillion rubles9illion rubles

Ue assumes zero lag between production and purchases of both producer and consumer durables. However, if this year's purchases include part of last year's production, the resulting residual represents neither output nor purchases of military/space hardware.

The Consumer Durables Adjustment

e calculates the oulput of consumer durables by inflating aof consumer durables purchased5o account for theincluded in the CIA sample and then moves this value over timehegrowth Of the GVO ofnderUe rejects the CIA growth index (or consumer durables becauseJS.


doi include consumer electronics. Neither of the growth rales that Lee uses can be supported as indexes of growth in consumer durables. For ative-year periode could have used the comprehensive official index of deliveries of consumeror the entire period he could have derived an index from sales data available yearly in the statistical handbooks for consumer durables.


The methodology which Lee used to derive his .estimates incorporated most of the pitfalls in Soviet data that were cited at the outset. In his attempts to estimate some of tho missing Soviet data, heumber of errors. He mixed input-output data with other data prepared for the CSA without (I) first estimating total machinery output in the economy by removing non-machinery production from the CSA data and including machinery produced outside the MBMW branchemoving turnover taxes and trade distribution markups from the input-output and Soviet investment data. In confronting the trend problem, moreover, Lee did not use Soviet indexes where their application seemeds inime series for the gross output of MB and for consumernd instead created his own indexes, often with, no basis in fact. Lee also erred in subtracting all unfinished production from gross output of MB instead of merely the change, as even Western accounting practice would do, and in failing to deduct MB shipments to other industries. Lee's inflation of the CIA estimate of the value of consumer durables5 toase-year estimate for consumer durablesrude adjustment for the omission of consumer electronics from the CIA sample.

Finally, Lee seems to estimate the value of his residual without conducting sensitivity tests on his estimates. Estimates made at very high levels of aggregation in Ihe data, such as in the value of mctalworking, for example, maymall percentage error, but the ruble equivalent of that error canarge percent of the residual that is, the errorillion rubles which Lee made in estimating the machinery componcnl of MBMW is% of the value of MBMW. but itf the residual value of machinery which Lee shows

Lee's Estimate of Research and Development Expenditures

2J. TiSU. Nvodnoyt thoiyaruvoodu. p.h* published index, baud on Soviei uiput-outpul "ork. ejowi% per yen} whOt Lee'i index Uieieweiec veai. Tne formci it gencially accepted as an accurate measure ol contumpnon ol contumei Ihe hypothesis had been outlined In go culler unit

tudy alsoalculation of the value ofdevelopment (USD) expenditures included in MBMW oulput. Thisbased on the following hypothesis: the gross outputrojects isthe published value of MBMW output, but MBMW employmentersonnel.1*

44. The procedure can be illustratedimple formula:

( MBMWveiafic wage In

CVO of ^

- The value of RAD

Gross outpuls thus defined as cost plus profit and is subtracted from the official Soviet gross output value (seehe difference is labeled

omponent of MBMW. Since the residual represents gross output. Lee adjusts it to exclude unfinished production and intraindustry use so as to obtain an estimate ofinallthough the methodology appears straightforward, the residual obtained can be questioned on several points:

security payments were not included in Lee's wageaic part of the wage portion of Soviet cost data. Inclusion ofalone would reduce the residual9illion tublcs.

estimate of employment begins with published data onof workers in MBMW. To obtain total employment. Leefragmentary data on the number of engineers and support staffthat were (hen available. Although Lcc allowsighestimate of employment, and avetages the result, figures fordata that were published later indicate5 to0 men higher than Lee'sfor that year, while the actual employment0 waslower than his low estimatehe estimates for otherbe similarly distorted, and the actual figuresuchin employment than implied by Lee's estimates.

n calculating the share of labor costs in total costs. Leef whal the Soviets designate as "other costs" in their cosl breakdown



foi MBMW. While il is true (hat these costs are partly labor and partly material costs, the "other cost* categoryatchall including costs lhal arc difficult for the Soviets to assign Io labor or material cost categories. The labor component of "other costs" would not be included by the Soviets in calculating the annual wage bill. Therefore, when Leef "other coats" inabor cost percentage, the total costs are understated. An overstatement of the residual results.

n the basis of Lee's hypothesis that MBMW data do notersonnel, if the industrial gross output of scientific organizations is added to branch gross output, then employment data should also include the workers in these establishments. While inconsistencies in Soviet reporting techniques areirect statementoviet source confirming Lee's hypothesis on the exclusionersonnel would be necessary for most Western analysts IO accept an inconsistency as gross as thai which he is assuming. What Lee may have rediscovered is an inconsistency between the prime cost data and the gross oulput data, which may or may not be relatedroduction.

Borctsky's Estimateachinery Purchases Residual

ichael Boretskyesidual for machinery purchases, which he equated with military hardware procurement,0is methodology, as set out in Tablean be Summarized as follows:

Rubin in IMS

tfinwr ol Ihr

. until)


tot CVO of

i. Inii.inuuuiv an In

f> Valuecimrdnl< prodwn sold

7 ValM d( ItfiH ouipui of


21. Michael Bo.eisfcy. The Technologicalr SovietWs-t.

I SMrntrl.Alilf IIP.

The GVO for MBMW (establishment definition is estimated95 prices by mean* o( ihe official indexes of GVO growthoviet-reported figure ofillion rublesJ*

This value is reducedommodity-based estimate using9 commodity-establishment ratio, previously

The MB portion of this outpul is estimated by using the MB-MBMW relation in the9 IO

The bcnchmarlc value estimated ins extrapolated to other years ineriod by means of the official Soviet index of MB growth;

ntraindustry use in each year is estimatedhanging share of the estimates made in;

he value of intermediate ptoducts sold to non-MB sectors is determined fromable and the share of this value in total MB.

The value of the final output of MB is derived by subtracting the estimates ofrom the estimates of;

Private and public consumption of durables are estimated from Soviet data on retail trade sales.

) Soviet data on investment in fixed capital and inventories are used to adjust the series for producer durables:

estimate* of exports snd imports are made and exports are subtracted from final output while imports are subtracted from producer durables.

The residual remaining is identified as "military and space programs "

oretsky's methodology is superior to Lee's in its treatment of commodity-establishment conversions, turnover taxes, and trade and distribution costs. However, it still employs benchmark estimates which are moved forward and backward through time by inappropriate official indexes or by his own rough estimates of growth.

lb clhodolorv detcribed by Bo.ti.l, ,noo,

Tlx <aUie he inet ThedcunbtdMetIn the GVO thai Boritk, tat9 It. See paiwi* is.


Derivation of ihe Gross Value of Ouiput of Machine Building

Lee. Boretsky first had loVO of MB from theMBMW. Borelsky's estimate of the GVO of MBMW, however, is in error as

thef "BMW output between MB and MW. While aware of 'the commodity-establishment problem. Boretsky nevertheless used the official Soviet estabhshmenf-based index of growth in MB.

The Value of Ihe GVO of MBMW

derivation9 base year value of MBMW grossbasts) is unclear. The value of MBMW output9 thatwith6 billion rubles. Bul this value is not the result of thehe sets forth andillion rubles less than Lee'sstimate6 billion rubles is inconsistent with other Sovielthis time period. Unofficial Soviet sources report the GVO of MBMWillionalue which is consistent wilh the growth indexvalues published by the Soviets for thef the reported increase of GVO of MBMW9 loigure83 billion rubleshich is dose8 billion rubles thai can be calculated using Borelsky'sBorelsky's GVO for MBMW in his benchmark year, therefore,by atillion rubles

The Division of MB and MW Output

detive the GVO of MB. Boretsky calculates the ratio of MB loan early reconstruction of theablend applies thehis estimate of the GVO of MBMWommodity basis (lineabledoing so, he is on shaky ground-

data of theeconstruction were basedVOimports, whereas the adjusted establishment dala loapplies the ratio cover domestic outpul only;

ata include turnover tax. trade and distribution costs,costs, whereas the adjusted establishment data loratio is applied arc in enterprise wholesale prices.

H. inweentiKt of( Ind-Mrr In.ihe pdcen.ite mmutcoviet wucee. Bwi ippiMatLO* at thi. meihodoloev noafecM u> aemiMiboa

29.. Cj-rtto.. Hy* *etvnte*.yr wnur. ntmohHl <

it" p.

M Th* IHSfvuMot i.f 14


Jr VUdWht ItStew*rcu*wl


c. ata are9 puces and the adjusted establishment data arc in5 prices.

s estimate of the GVO or the MW sectors (includingI9S93 billion rubles. The value based on Ihe reconstruction of thetable adjusted to producer prices,9 billion rubles. Thus, the sumerrorsillion ruble underestimate of GVO of MW.

The Growth Index for Machine Building

calculate MB output for years otheroretsky usedindex for establishment-based GVO of MB, thus implicitly assumingcommodity and establishment GVOs grow al the same rate. Because thisthe GVO of MB so derived can not be relied on toommodity-based residual

Derivation of the Value of Final Output of MB

to derive the value of final output of MB from theMB center on the elimination of interindustry and intralndustry uses of output.

Intraindustry Uses of MB9

calculates the intraindustry use of MBf sales inmultiplying the ratios of (a) interindustry use tonduse to interindustry, both derived by Sovietable. This apparently siraighiforward calculation contains threetogetheret overstatement of this adjustment equal

ratios that Boretsky assumes arc calculated from data ondemonstrated in Vladimir Trend's reconstruction of the Sovietto be based on data for all MBMW:"

assumed lhal the ratios were calculated from theand could therefore be used together. But the ratio oftoas calculatedable in producer pricesturnover lax and transportation and trade costs) while the ratiouse to interindustryas basedableprices;

ratio of interindustry use lo GVO) was based onconsumption. Boretsky knew that gross output inexcluded exports, but he did not realize that importsadded and that the adjustment madeet foreign trade (imports

il. Srr. for example, Vladimirreml, ri. at. .Vimdia-e,, UNCLASSIFIED. SI. VtwtimU C. Treml. opVol. II.. UNCLASSIFIED

minus exports) adjustment ratherubtraction of exports Accordingwn estimates of the foreign tradeis adjustment is overstated

these errors affect the accuracy of Ins calculation, and theintraindustry use should be rejected. The value of intraindustry salesn producers prices3 billion rubles.stimaten overstatementillion rubles.

Intraindustry Uses of MB5

estimates intraindustry use of MB5 by increasingintraindustry percentage useactor. This factor was basedhe believed were expost data from5able. Thehowever, was an exante planning tabic whose price base is unknown;derived from it represented planned (or expected) changes innot actuallso, the data from which the increasewere based on material costs rather than GVO. Only if thematerial costs to gross output did not changeould thethe ratio based on material costs reflect the change in the ratio based onComparison of96 IO tables shows that this was notThe value5 is therefore of questionable realiabiliiy. andre even less reliable because there is no reason touniform trend in the rate of change in intraindustry use. Intraindustry6 in producer prices7 billion rubles. Borctsky's valueillion rubles. After adjusting for price differences, Boretsky'sis overstated byillion rubles.

Interindustry Deliveries9

o derive MB interindustry deliveries to other branchesoretsky multiplied the GVO of MB. an adjustment factor derived by subtracting the share of intraindustry uses of) from the interindustry-use4 reducedo allow for the exclusion ofs noted above, this factor was based on data for total MBMW9 producer prices (with the net foreign trade balanceoietsky applied this ratio, adjusted only for exports, to MB measured5 prices. The value derived byillion rubles)illion rubles higher than the value in9 Soviet IO tabic, thus understating the residual by at least this amount.

Interindustry Deliveries

o estimate sales of MB to other sectors of the economy. Boretskyoviet statement that5 the proportion (relative importance) of icpair in the overall volume of machine building output declined by almostn

comparisonn ihecar period this proportion will declineoretsky applies these declining ratios to his estimate of interindustry usen so doing, Boretsky assumed analogous trends in sales of MB to other productive sectors and in the proportion of repair in MB output.

declining proportion of repair, however, has little to douse of MB. The repair GVO referred to in the Soviet statementgross output of specialized enterprises whose primary function is repair.weight of purchases by these enterprises from the MB sector (partuse) would be influenced by the declining share of repair inof MBMW. However, changes tn current and capital repair performedon their own equipment and MB sales to other sectors oftools would not be reflected in the declining share in total MBMWrepair enterprises. The magnitude of the error is indicated by anthe data in6able. According to the Boretskyof MB to sectors other than MBf gross outputtheable for that year theyf grosserror amountsillion rubles in the year

The Subtraction of Producer and Consumer Durables

final step inesidual value of MB final outpul isof estimates of producer and consumer durables. Here,the turnover taxes and used an average value for investment inyear that probably understates investment. His inventory seriesonsequence, the residual is overstated. Boretsky also assumes abetween production and purchases of producer and consumer durables. Ifexists between the purchases dala and the output dala, then theis neither the purchases nor the output of military and space hardware.

Producer Durables' Equipment and Inventories

The machinery investment data used by Boretsky can be criticizedumber of points. Boretsky cites the Soviet statistical yearbooks for his gross investment data, but8 yearbook that he cites docs not have data in the form in which It appears in Boretsky's paper. Investmentkey years in the Boretsky calculations) is not given separately but as partive-year aggregate figure for the. The value used by Boretsky8 appears loimple average of the five-year total. The margin for error inalculation is sizable: when the same method was applied to the totaleriod for which separate yearly values arc availableater yearbook, the average annual investmentillion rubles greater lhan the actual value

Boretsky docs not show his calculation of investment values for theowever, if his source and methodology are used. Ihe value9 investment appears to be understated. Given the value of investment1 and

his estimatehe value of investment9 would have been altlhon lubles, wheieas the gross inveslincnl value implicit in Boretsky's net value8 billion rubles,*1

eries for investmeni in inventories of MBassuming that working capital grew at Ihe same rate as gross output.single published value ofillion rubles0 toeriescapital in MBMW. he assumedf this working capital wascapital ofrobably on the basts of gross output of MB asof GVO of MBMW. Working capital is then divided intounfinished production with the help of the published percentageworking capital. Because this percentage distribution is available only forhe had to estimate it for the earlier years in his. Hisinvestment in inventories arc the yearly changes in these derived seriesand unfinished production.

Consumer Durables: Private and Public Consumption

Netofnail ran rabinullwo ,ubk. dmW.

16 TinW liom Ptm&ip Minor. Jfcr Comumr- ft) ,hr SoWer lco*omf. Ml


il. VUdimo C. Treml. ft. ml. Omfito* of Sowm /unw'-Owpix Tabla io frodutmWoTable. BE A. Foieirnejwiii No. I. WaiNrunon. DC. lv7J. UNCLASSiriLD JB. Bird on Hie dlllnbotloo of ihe luinmei In. inabk Inroduce* pcicci) between the MB and MW teciou


Boretsky uses Soviet data on private and public consumptiono estimate the value of consumer durables. In the conversion of the data from retail prices to enterprise wholesale prices, die turnover tax and the cost margins for retail trade, distribution, and transportation musi be removed.

In removing the turnover tax. Boretskyurnover tax rateor the entirehisrude weighting of many individual rates, appears to be high on the basis of other sources of information: Vladimir Tremlaten Ihe reconstruction of9abic, andabic has an implicit rateorn addition, the turnover tax collections from MBMW94 areillionillion rubles, respectively. These values arc equal to the Boretsky estimates for just MB, indicating that the Boretsky values arc overstated byl1

Boretsky's estimates of the retail trade margins and transportation costs are based on their share in total material costs in MBMW9he use of this overall MBMW ratio for private and public consumption is wrong because retail margins and transportation costs are not distributed proportionally among either the MBMW sectors or among the various final demand and interindustry uses. Boretskydjustment for retail margins and transportation costs combined5% of retail trade sales The implicit rate calculated fromable in producer prices was close. He has therefore understated these charges by.

gaps in Soviet data with others in his treatment af the machinery components of fixed investment and workinghich weakened the reliabtlily of the residual

Can the Residuals be Trusted?

Lee and Boretsky estimates have been reviewed in detail -quite technicalore important criticism is that any estimate ofencounters too many imponderables toirm calculation ofas important as production of "military and spaceof machinery residual still produce inconclusive results, evencalculations can build on the Lee and Boretsky studies and have themote recent information. Some object lessons are described below.

9 Residual

The9 is one of the most promisingesidual estimate because the Soviets prepared theirable for that year. Vladimir Treml and his associates have reconstructedable on the basilreat deal of research. This reconstructed tableeemingly unmatched opportunity toesidual for military-related machinery.

But not all of the reconstructed table can be used. The GVO of the Soviet MBMW sector of industry compiled on an establishment basis was reportedS billion rubles in enterprise wholesale pricesilue9 can be estimated by using the announced growthor MBMWonverting the resultommodity basis (by multiplying by the commodity-establishmentalue0 billionhis value in enterprise wholesale prices, however, exceeds the Treml estimate ofillion rubles, which is in purchasers' prices. Sinceable was exploratory, something could have been excluded In its preparation. Moreover, the announced GVO ofillion rubles8 (establishment definition) is consistent with other Soviet data. Therefore, we begin with the establishment-based figure and convert itommodity basis before bringing the LO data to bear on the residual problem

Theable reports data for the GVO of melalworking andillionotal interindustry deliveries byillionnd consumption of MBillionhese values include turnover taxes and trade distribution markups. Ihe sum of these three items must he subtracted from the GVO of MBMW. But these deductions are too large because they are valued in purchasers' prices.1 billion rubles for turnover taxes1 billion rubles, representing distribution markups, must be added at this stage of Iheheillion

e tnWfptiM wfcoatiak prionMS.

Except foi Inn collection, thur dita are ftom Ok itcomuiicied Sonal IO labtt9 tare Treml rl.i col ledre. Ooionklair*Infill it. Obmomiiyrmmyihlfnnosli,. UNCLASSIFIED.

Residual Machinery9



(.tots value of MW and repay outputotal interindustry uses deltvertci1 MB output used for consumption1 Plus

Turnover taxes on MBMW output MBMW trade-distribution markups Equals



Deliveries of machinery to Axed mTesimect. michinery exports, and deliveries to other uses1 Leu

Machinery component of investment Machinery exports Ilui

MachineryHquall Residual MB deliveries1

I" wrpmeei _ '

1. radtum' price*.

epresents output used for investment, exports, and other purposes such as defense and additions to inventories of machinery.

Investment and net exports of machinery arc not identified in theable Indeed. Soviet investment data9 are incomplete and include distribution charges and some investment in furniture, plumbing, and the like. The value for investment originating in MB shown in the tabulation Ls the sum of (I) Ihe published value of state investment in machinery and otherollective farm investment in machinery and equipment (derived by subtracting the value of construction and other types of investment from total collective farm investment)eduction ofBoretsky's factor) to allow for equipment such as furniture thai did not originate in MB.

The determination of the domestic value of Soviet exports and imports is uncertain because they must be converted from foreign trade prices to domestic prices. There is little reason to choose among the several conversion ratios that Western researchers have developed; the tabulation uses the values of imports and exports fromable as reconstructed by Vladimir Trend and associates

he residual6 billion rubles includes production of military hardware and any items unaccounted for in the calculation of Ihe residual. For

example .he residual certainly includes changes in invenlorie* and machinery losses More important, il may well include Ihe value of capital repair performed by MB enterprises on their "own

The residual also reflects any errors made in estimating the various values used in the tabulation. The primary source or error is the commodity^tablishmenl ratio. Although published by the Sovietshe ratio was based on sample survey data and may therefore be incorrect. Il is impossible to quantify this uncertainty. Price differences undoubtedly are another source of error. Cross output of MB is reported5 prices.ata are9lthough it has often been assumed that average MB prices did not changeS9 this remains an assumptionossible source of error.

Pie overall error in the Loc and Boretsky estimates is further illuminatedomparison with9 residual from Tablehis residual8 billion rubles larger than Boretsky's estimate4 billion rubles lower than Lee's estimate.

6 Residual

he CSAable togetherhe quality of the research reflected in this table is believed by Western experts to be vasdy superior lo lhat supporting9 table The full table. like9 table was reconstructed by the Research Analysis Corporation.6 is another vintage year for the calculation of residuals, although the procedure followed is somewhat different from thai used in9 residual

Table 6

ResidualMachinery Output.Rubles0 Enterprise eVhpleale prices

Value of final output of.g


MB output used for

Machinery component of6j





o(rs fo. (Hoducoe* snd wrfpenuied fromThe, Includehorn oncembon. b, Ihethe upwdiiuH up lo use discontinuation cf II. pio-cts.end.iu.e, lolthesnd minus ihe value of ulvspd

apital npau .orulucied "hi" should hiveemo.ed liom MB with the cornnsoditvestaWishmcMThe Sovttla are amhtuou, In ,hei. dcac.mi.on of ihi. adms-nscm



he methodology, shown in Tableses dala on final outpul and consumption directly from the reconstructed portions of6 Soviel table in producers' prices (enterprise wholesaleence, difficulties arising from converting establishment-based data to commodity-based dala are avoided while turnover taxes and trade distribution margins have been removed from both the value of MB output and the value of deliveries of MB output to consumption.

Investment in machineryillion rubles) is estimated by reducing the published value of investment in machinery and equipment% to allow for non-machinery products (Boretslcy's factor) andor distribution charges (the average distribution charge on MB deliveries to final demand. Dala on foreign trade in machinery from the Soviet foreign trade yearbook arc converted to domestic prices by coefficientsS for imports0 for exports. These coefficients were developed by ihe Foreign Demographic Analysis Division of Ihe US Department of Commerceesult of research on theables.

Unfortunately, this new estimate of the residual is not as solid as it looks.ubstantial difference was observed between the gross values of MBMW output calculated9 from establishment-based data andata,6 data should be tested for the same discrepancy. But the required information does not existoreover, the residual (a) is affected by all of the inevitable errors in reconstructing the MB portion ofable and (b) mcludes losses and inventory charges.4*

7 Residual

illustrate some of the pitfalls encountered in estimatingoutpul for years whenala ate not available, anmade toachinery residual7ecauseassumptions arc so critical to the estimate of the machineryresults are presented in terms of low, central, and high estimates. "Iheis based on the best available point estimate for each value in the

calculation of the residual. The high and low estimates use values in each of the

steps thaieasonable range of uncertainly. For the high estimate all of the values chosen tend lo drive up the residual while for the low estimate the

reverse is true. In addition, three different price deflators arc used to permit

comparison of the residual with Ihal found

large differences in the estimates shown ineflect(I) estimating the ruble value of commodity output ofdeliveries ofonvening foreign trade in machinery

foreign Dtrriopiplik AnalrfU utfi.on,f Own merer,t Soviet Input Output TibUi to rroduftn' Meet:6 BreoiutnittrdNCLASSIFIED.

*i. For rumple, do Kin of aimmotity-hnd to nuSkt-wmanBl ' -n

In cunnccuon vilC9 IO Utile.


In houtf" cap.til repair, ho-ever. iceini lo be(tinted. In cont.att lo9 iriidutl

AlremativeheU( Reiidual faRubles7 pIKCS

CVO of MBMW. establishment basis GVO of MBMW, commodity baits CVO of MB. commodity bash Less Interindustry use Exports

MB oulput for consumption Machinery component of investment


Import! fcquah

Residual MB deliveries?

enterprise who leak prices) Residual MB deliveries

5 prices, assuming:

ncrease in7

no chanRe in7

ecline In

from the foreign trade rubles in which it is reported to domestic prices,stimating consumption and investment data in enterprise wholesale prices.

published for the first time3alculation ofof MBMW7illioneast-squaresthe relation between indexes of growth in the GVO of MBMW. MW. andcan be used to estimate (he share of GVO of MB in lhat of.

The analysisarrow margin of error of thisow of


is no commodity-esublishment ratio for any year other thanis there reliable information about changes in this ratio over time. Thisaccording to changes in the importance of production ofby machinery enterprises and in the amount of machinery producedenterprises Soviet policy has been lo increase the degreespecialization. Because of the uncertainty about the success of thisdifferent assumptions about7 ratio are used: (I) Ihe ratioI9S9 valuehich implicitly assumes that the two components ofdid not change or experienced offsettinghe ratio equalsimplies that Ihe Soviet specialization policy was so successful lhal the

i'-i'rc I

commodity- establishment problem has disappeared,he ratioliich implies that the Soviets succeeded in eliminating non-machinery outpul at machinery enterprises but could not reduce the relation between machinery produced by non-machinery enterprises and the production of MBMW enterprises. Although higher or lower ratios could be assumed for sensitivity testing, they would be even more arbitrary. But the higher the ratio, the greater would be the machinery residual.

[jb>.i'i". t MvnoK nn, oc FtenotHyt

khoiyayitw, no.. UNClASSintl).

ht mioi calculated (or heavy tnduilrr product!ioduntV pii'.ei utileJAD. op.3 roi impotU5 tor eipom.

Interindustry use is estimatedf GVO, based on recently availablehis share agrees closely with that embodied inable in producers' prices. Therefore, the range of uncertainty ins limitedor the low estimateor the high estimate.

Soviet imports of machinery75 billion rubles in foreign trade rubles, while exports of machinery1 billion rubles. The best available ratios for converting these values to domestic prices are those used for6or imports0 Torherereat deal of uncertainty in theseange for the low and high estimate is calculated by assuming an error ofin the ratios.

The value of MB production allocated to consumption7 was extrapolated from6 billion rubles inable (producers' prices) using the Soviet production index "household and culturalhis index includes non-MB products but is dominated by consumer durables manufactured in MBMW.

7 the machinery component of investment is reported in "delivered" prices ofillionlthough these prices refleel the general price changes inhe CSA claimed that average machinery prices changed very little.y case,9 prices arc almosl certainly closer to the7 prices lhan5 prices in which investment had been reported.

The average trade and transportation share of the value of final products calculated from data inableas used to remove these charges from the "delivered" price of investment Non-machinery products arc netted out using Boretsky's factor ofnd% margin on either side for uncertainty.

The residual7 pricesange ofillioniliion rubles above and below ihe "central" estimateillion rubles. The low valueillion rubles is less lhan6 residual calculated in Tablehile the high value isimes as large as6 residual.

Convening icsidual output oi" machineryiven year's prices io constant prices is essential to obtain results thai are comparable over time. The official Soviet price index for MBMW declinesSn contrast, some Western (and some Soviet) analysts believe that machinery prices have actually beennsofar as the residual represents military hardware, the appropriate deflator could be quite different from either the official or unofficial price indexes for MBMWhole. Therefore, the upper limit for price increases57 was arbitrarily limitedower limit to price changes was assumed to berop in the official Soviet index for MBMW. No price change was taken lohird alternative.

The values for residual MB output75hown in Tableangeillion3 billion rubles. Even the central estimate rangesillion9 billion rubies.

Billion Rubles5 Prices

Table 3

The range in these values is more than enough to encompass the Lee estimate5 and the Boretsky estimate

The residual, calculated by comparing9 estimate with the central estimate7 (assuming no pricencreases% per year, compared with the Lee estimateer yearnd the Boretsky estimateer year. Their higher growth can be matched only by making the extreme assumptions involved in the highest estimate5 prices.

The low level of reliability of present estimates of MB residuals seemsingle answer is not possible, and the range of results obtained from reasonable sensitivity tests is too wide. These considerations, and the fact that residuals still contain non-defense elements of machinery production, demonstrate the inconclusiveness of the residual approach, given the existing state of the art and of the statistics.

lot example. Abraham S. Becker. Ruble fitee Lerelt and Dolia Ruble Railoi of So'iti Machinery. RAND.DRE.KCLASSIFIED.

Original document.

Comment about this article or add new information about this topic: