Documents / FOIA release
This Joint Committee Print, dated December 8, 1982, gathers studies that the Central Intelligence Agency's Directorate of Intelligence prepared for the Joint Economic Committee of the US Congress. It gives estimates of Soviet gross national product for 1950 to 1980, with indexes of industrial production, agricultural production and consumption, and explains the methods and data behind them. The foreword says official Soviet statistics inflate growth and that the CIA measures offer a more accurate picture. The document contains no UFO material.
Approved for Release: 2019/07/19 C05210421 general, the larger the sample’s share of the total, the greater the likelihood that its characteristics are similar to those of the full population.” Using the 1972 input-output table as a point of reference enables us to measure the sample coverage and to identify those areas of industry where a relatively small sample threatens the accuracy of the indexes. If we sum the value of all the products within each sector’s sample for 1972 and compare this with the reported GVO in the input-output table, we get a rough notion of the percentage of industrial output covered by SPIOER, at least for that year. The measure is rough because differences in accounting practices and pricing make the industrial sample and the input-output table incompatible in some respects. For example, the production of commodities pub- lished in physical units includes intraplant turnover, but the value of production as reported in the input- output table presumably removes intraplant turnover in measuring gross output.” Production data ex- pressed as a CSA index and perhaps as a value series are also usually reported on an establishment basis, and the input-output data are compiled on a commod- ity basis. A commodity-based classification scheme rebases the secondary production of an enterprise to the industry to which it ostensibly should belong, but establishment-based classification makes no attempt at such a correction.” Another important difference is the SPIOER sample’s use of 1 July 1967 prices and the input-output table’s use of 1972 prices. To the extent that there has been inflation—concealed or otherwise—our estimated sample coverage may be understated. 1‘ A large share—but less than 100 percent—does not ensure that the sample represents the total if the products included are atypical. For example, Soviet production data may be upward biased because the CSA publishes only data that reflect favorably on Soviet accomplishments. On the other hand, some published data under- state growth because the Soviets generally measure output in natural units, like the number of bulldozers. Such treatment fails to capture changes in product assortment and quality. Therefore, because of the peculiarities of Soviet data, a large sample share does not guarantee the representativeness of the sample. ’° Narodnoye khozyaystvo v I 978 g, p. 580 and Treml, Gallik, Kostinsky, and Kruger, Structure, pp. 46-47. Also see the earlier discussion of gross output measures in this paper. ’° Treml, Gallik, Kostinsky, and Kruger, Structure, pp. 123-146. In SPIOER’s coverage of industrial production, 87 percent of the industrial sectors reported in the recon- structed 1972 input-output table have some form of representation in the sample (table 13). The machin- ery branch has the most sectors and also the most gaps in coverage, although the missing sectors are relatively unimportant. Of the represented sectors the median coverage is 71 percent with the degree of coverage ranging from 8 percent in one machinery sector to more than 100 percent in certain sectors in the construction materials, light industry, and proc- essed food branches.“ Coverage in four of 10 branches is more than 70 percent, and all but two have at least 50-percent coverage. The most restricted samples are found in nonferrous metals (29 percent) and chemicals (42 percent); statistics on production of nonferrous metals and many chemicals are considered classified by Soviet authorities and hence are not reported. Machinery coverage would have been low—about one-half—except that we added our estimates of military machinery to the value of the sample. Overall in this benchmark year the SPIOER sample ac- counted for about 60 percent of Soviet industrial production. A Rough Measure of Sample Representativeness. The fact that SPIOER product samples represent substan- tial shares of total output in 1972 does not guarantee that the samples are representative in terms of their growth. The best available benchmarks of representa- tiveness are the various input-output tables. One rough measure can be performed computing gross value of output in current rubles from the 1959, 1966, and 1972 input-output tables for 10 branches of " The sectors where this occurs are: logging (115 percent), cement (106 percent), concrete (106 percent), asbestos cement (1 18 percent), linen (159 percent), hosiery and knitwear (125 percent), flour and cereal (142 percent), and fruit and vegetables (155 percent). There are several possible explanations for the seeming anomaly. First the average wholesale prices used by SPIOER could be too high. Second, in some cases prices may have been reduced between 1967 and 1972. Third, different definitions of output and differences arising from the conflict between commodity versus establishment classifications also play a role. 208 Approved for Release: 2019/07/19 C05210421
Not linked to a story yet.
FOIA release, from the cia-readingroom collection. The PDF is mirrored here; the original link is above. 399 pages are in the text index: search them above, or from the library's search.