Invariance
16th April 2009, 13:53
Discussing the Reliability of Soviet Statistics:
To what extent can we trust statistics published by the former USSR? I would like to address this topic, and dispel the myth that the Soviet Union engaged in falsifying statistics for propaganda purposes, but still support a critical approach to the reliability of those statistics (and indeed, the economic statistics for any other country, or for any other topic). Statistics are often presented as irrefutable truth of an argument. And I am not denying their empirical worth. I do think they can be misleading, however.
Were Soviet Statistics Falsified?
Some Cold War economists, such as Naum Jasny and Nicolas Spulber, argued that the Soviet Union did indeed falsify statistics for propaganda purposes (see “Soviet Statistics.” The Review of Economics and Statistics, February 1950, pp. 92-99 by Naum Jasny and The Soviet Economy: Structure, Principles, Problems (1962), pp143-44 by Nicolas Spulber). Generally, however, even Western academics support the view that whilst the USSR did withhold statistical information, as well as their methodology, it did not actively falsify statistics. Alexander Gerschenkron, Abram Bergson, Harry Schwartz and Alec Nove fall within this category.
Indeed, Gerschenkron pointed out that Jasny’s own use of Soviet statistics indicate that they hold some usable and empirical worth. Likewise, Robert W. Campbell, in Soviet Economic Power (1960) p.35, wrote:
‘By and large, our experience has been that the mass of Soviet statistics are more or less internally consistent, and this reinforces the doubt that the Russians are engaged in a process of double bookkeeping or outright statistical falsification.’
Nove also wrote:
‘Very few people believe that they are invented. The evidence against such a view is very strong. Despite captured documents, despite the presence in the West of Soviet officials who had defected, no evidence exists that the central Soviet statisticians invent figures to order, to propaganda effect. By this is meant that no one issues orders to print a figure of 400 knowing that the correct one is 350.’
Nove, quite correctly, points out that figures were suppressed – and later an explanation given that production fell in that particular year or period. Such actions would be inconsistent with falsifying.
Hastily proclaiming that the USSR falsified statistics has embarrassed some Western academics. For instance, one American commentator compared the cotton and wool cloth physical output which were below the previous year, and compared them to the official statistical reports which claimed an increase. The commentator jumped to the ‘obvious conclusion’ of cheating/falsification on behalf of the Soviet government. It wasn’t. There had been a change in definitions from linear metres to square metres, the object of which was in fact to stop cheating by those who sought to fulfill plans by making cloth narrower. Far from the government being engaged in falsification, it was enacting practices to prevent the statistical distortion which they were concerned with.
Those who claim that Soviet statistics are falsified, generally have an axe to grind against the economic success of the USSR.
Availability of Statistics:
Generally after the post-Stalin era, there was a liberalization of the publication of statistics and a greater academic discussion of statistical problems. Some academics are quick to point this out as evidence of the totalitarian nature of Stalin’s Russia and so forth. And, to an extent, they are correct; Stalin was fairly adamantly opposed to mathematical planning of the Soviet Union, and numerous heads of statistical departments were accused of ‘wrecking’ and ‘sabotage.’
However, part of it also has to do with the precarious situation the USSR was faced in – facing a potential invasion by Germany, or indeed by other Western powers. Statistical information of the country’s economy could provide excellent strategic advantages to Nazi Germany. And this line of logic wasn’t unfounded – the 1941 Plan was captured by the Germans, and later ended up being published in America by the American Council of Learned Societies). Alec Nove points out in “A Note on the Availability and Reliability of Soviet Statistics”, The Soviet Economy (1961) pp. 307-314,that statistical information of numerous areas of the economy were lacking. For example, output figures for various industrial products, nonferrous metals, ships, aircrafts, various chemicals, various machines and military weapons. Understandably enough, data on military weapons was classified. However, considering the (increasing) proportion which military spending enjoyed, leaving this out of economic data would naturally distort any model of the USSR’s economy. Nove also points out the lack of data on labour forces in particular industries, and little data on worker or peasant earnings.
Nevertheless, by 1960 the USSR Statistical Handbook alone contained over a thousand pages of statistical tables. A marked improvement from the era of Stalin. I now turn to the question of their reliability.
General Statistical Reliability:
One problem with reliability is the definitions employed in gathering and outputting that statistical data. As Nove points out, when Soviet authorities point out that 300 million pairs of shoes have been produced in x period, then that accords with data from the Central statistical offices in Moscow.
Yet, what defines footwear? Sometimes it only includes leather footwear, sometimes all footwear, sometimes handicraft production of shoes is ignored (but not stated to be ignored). This all leads to problems of reliability; the, sometimes, arbitrary nature of definitions. The methodology of aggregation is often omitted. Definitional changes can distort the true picture of things, particularly when rewards are based on output.
The next problem is the reliability of the statistics themselves – whether enterprises are prone to exaggerate output. Distortion of production data has been a crime since the days of the first Five Year Plans. There is a limit to the extent to which output can be exaggerated; the ratio between production and output puts a limit – one cannot say to have produced x amount, then when it comes time to deliver that amount the enterprise is lacking. Conversely, there may be a temptation to conceal output in order to keep extra stocks to meet next period’s production quota, or to conceal selling goods via the black market or privately. A problem which the post-Stalin government faced was the poor quality of various goods – if output was merely based on the number of x commodities produced, then obviously quality may be downgraded in priority. From memory, various tables/furniture were measured in weight. Hence, some enterprises purposely made heavy tables to meet quotas – an obvious example of not utilizing resources as effectively as possible. Measures were taken to fight these sorts of practices, where product quality became more important.
Nevertheless, some academics have employed an ‘equal law of cheating’ approach – unless it can be shown that exaggeration is geometrically increasing, the rate of growth must remain constant. (Some have reasoned that the law of equal cheating doesn’t apply to agriculture – owing to the difficulty in keeping track of products and the series of campaigns which put pressure on various products (grain, maize, meat, milk campaigns) to exaggerate their output. Khrushchev pointed out in 1961 of the practice of some collectives purchasing butter from shops and then redelivering it to meet their required output. Further, the increase in prices paid for produce probably led to a discouragement of evasion).
Definitional problems can have significant repercussions, however. Until 1953, crop data was published in terms of biological yield, that is, ‘on-the-root crop.’ Jansy was quite right to point out the exaggeration such a definition brought; it ignored the considerable harvest losses, was distorted by inspectors who exaggerated on the root crop estimates, since delivery obligations of kolkhozy depended on them (payments to the MTS).
In 1952 the grain harvest was said to have been 130 million tons. This was officially revised to 92 million tons!
Accuracy of Indices and National Output:
According to official statistics, gross industrial output rose almost 21 times between 1928 and 1955. The highest western estimate by F Seton allows for a 12 fold rise. The lowest, by W. Nutter, allows for a five and a half fold increase.
Most (all) Western academics rejected the Soviet output figure. However, as Nove points out it is not that the indexes are deliberately ‘cooked’ but that indices by definition are aggregations lacking accuracy. Much depends on price weights. For example, A Gerschenkron’s calculations showed that from 1899 to 1939 American production of machinery increased 15 fold with 1899 price weights, but less then doubled with 1939 price weights and he emphasized the enormous difficulties due to changes in type and design. Once again, definitions and methodology are all important.
There are some serious defects in the official indices.
Firstly, that until 1950 the weights used were those of 1926-27, which essentially gives pre-industrialization weights to the fastest growing area of the economy! Further, the big rise in costs in the 30s means that the prices at which new products were introduced into the index were higher than they would have cost in 1926-27. There was then a tendency to concentrate on the production of times bearing high 1926-27 prices, even at the cost of under fulfilling plans for the less highly valued items, because plan fulfilment was measured in 1926-27 prices. This led to rising inflation in the index.
Hence, Soviet statisticians can select the base and weights which help show very large increases. For instance, base-year weights give a larger increase in output than end-year weights. When Soviet statisticians calculate a cost of living index, however, they are careful to use end year weights which minimise the increase in prices and so represent real wages in a more favourable light.
National Income Problems:
Briefly, the official claim is a 17 fold increase in national income in the period 1913-1955. This is quite a staggering figure. It is widely agreed amongst Soviet economists that the national income of the Russian empire in 1913 was approximately a fifth of that of the US.
However, if that was correct, the Soviet national income would now be well ahead of that of the United States by the early 1960s. Further, it is odd that the national income can increase 17 fold when its principal component, agriculture, showed a rise in gross output of only 70 percent.
Clearly the methodology used to calculate the increase in national income is inaccurate.
Conclusion:
I hope I have shed some light on some of the problems facing the accuracy of Soviet statistics. I will make a further post later on referencing some of the Soviet views of a new approach to statistical recording which was really interesting even in the 1960s. V. N Starovskii, Director of the Central Statistical Administration USSR, envisioned a type of automatic statistical computer system which would be highly accurate and avoid many problems of exaggeration and so forth. The methods they used then were quite crude compared to today’s standards (essentially a ‘hole in the card’ system); the internet could be of immense use in aggregating data and allowing whatever government to make an optimal plan which suits its purposes. I welcome any debate, criticism or comments.
- Vinnie.
To what extent can we trust statistics published by the former USSR? I would like to address this topic, and dispel the myth that the Soviet Union engaged in falsifying statistics for propaganda purposes, but still support a critical approach to the reliability of those statistics (and indeed, the economic statistics for any other country, or for any other topic). Statistics are often presented as irrefutable truth of an argument. And I am not denying their empirical worth. I do think they can be misleading, however.
Were Soviet Statistics Falsified?
Some Cold War economists, such as Naum Jasny and Nicolas Spulber, argued that the Soviet Union did indeed falsify statistics for propaganda purposes (see “Soviet Statistics.” The Review of Economics and Statistics, February 1950, pp. 92-99 by Naum Jasny and The Soviet Economy: Structure, Principles, Problems (1962), pp143-44 by Nicolas Spulber). Generally, however, even Western academics support the view that whilst the USSR did withhold statistical information, as well as their methodology, it did not actively falsify statistics. Alexander Gerschenkron, Abram Bergson, Harry Schwartz and Alec Nove fall within this category.
Indeed, Gerschenkron pointed out that Jasny’s own use of Soviet statistics indicate that they hold some usable and empirical worth. Likewise, Robert W. Campbell, in Soviet Economic Power (1960) p.35, wrote:
‘By and large, our experience has been that the mass of Soviet statistics are more or less internally consistent, and this reinforces the doubt that the Russians are engaged in a process of double bookkeeping or outright statistical falsification.’
Nove also wrote:
‘Very few people believe that they are invented. The evidence against such a view is very strong. Despite captured documents, despite the presence in the West of Soviet officials who had defected, no evidence exists that the central Soviet statisticians invent figures to order, to propaganda effect. By this is meant that no one issues orders to print a figure of 400 knowing that the correct one is 350.’
Nove, quite correctly, points out that figures were suppressed – and later an explanation given that production fell in that particular year or period. Such actions would be inconsistent with falsifying.
Hastily proclaiming that the USSR falsified statistics has embarrassed some Western academics. For instance, one American commentator compared the cotton and wool cloth physical output which were below the previous year, and compared them to the official statistical reports which claimed an increase. The commentator jumped to the ‘obvious conclusion’ of cheating/falsification on behalf of the Soviet government. It wasn’t. There had been a change in definitions from linear metres to square metres, the object of which was in fact to stop cheating by those who sought to fulfill plans by making cloth narrower. Far from the government being engaged in falsification, it was enacting practices to prevent the statistical distortion which they were concerned with.
Those who claim that Soviet statistics are falsified, generally have an axe to grind against the economic success of the USSR.
Availability of Statistics:
Generally after the post-Stalin era, there was a liberalization of the publication of statistics and a greater academic discussion of statistical problems. Some academics are quick to point this out as evidence of the totalitarian nature of Stalin’s Russia and so forth. And, to an extent, they are correct; Stalin was fairly adamantly opposed to mathematical planning of the Soviet Union, and numerous heads of statistical departments were accused of ‘wrecking’ and ‘sabotage.’
However, part of it also has to do with the precarious situation the USSR was faced in – facing a potential invasion by Germany, or indeed by other Western powers. Statistical information of the country’s economy could provide excellent strategic advantages to Nazi Germany. And this line of logic wasn’t unfounded – the 1941 Plan was captured by the Germans, and later ended up being published in America by the American Council of Learned Societies). Alec Nove points out in “A Note on the Availability and Reliability of Soviet Statistics”, The Soviet Economy (1961) pp. 307-314,that statistical information of numerous areas of the economy were lacking. For example, output figures for various industrial products, nonferrous metals, ships, aircrafts, various chemicals, various machines and military weapons. Understandably enough, data on military weapons was classified. However, considering the (increasing) proportion which military spending enjoyed, leaving this out of economic data would naturally distort any model of the USSR’s economy. Nove also points out the lack of data on labour forces in particular industries, and little data on worker or peasant earnings.
Nevertheless, by 1960 the USSR Statistical Handbook alone contained over a thousand pages of statistical tables. A marked improvement from the era of Stalin. I now turn to the question of their reliability.
General Statistical Reliability:
One problem with reliability is the definitions employed in gathering and outputting that statistical data. As Nove points out, when Soviet authorities point out that 300 million pairs of shoes have been produced in x period, then that accords with data from the Central statistical offices in Moscow.
Yet, what defines footwear? Sometimes it only includes leather footwear, sometimes all footwear, sometimes handicraft production of shoes is ignored (but not stated to be ignored). This all leads to problems of reliability; the, sometimes, arbitrary nature of definitions. The methodology of aggregation is often omitted. Definitional changes can distort the true picture of things, particularly when rewards are based on output.
The next problem is the reliability of the statistics themselves – whether enterprises are prone to exaggerate output. Distortion of production data has been a crime since the days of the first Five Year Plans. There is a limit to the extent to which output can be exaggerated; the ratio between production and output puts a limit – one cannot say to have produced x amount, then when it comes time to deliver that amount the enterprise is lacking. Conversely, there may be a temptation to conceal output in order to keep extra stocks to meet next period’s production quota, or to conceal selling goods via the black market or privately. A problem which the post-Stalin government faced was the poor quality of various goods – if output was merely based on the number of x commodities produced, then obviously quality may be downgraded in priority. From memory, various tables/furniture were measured in weight. Hence, some enterprises purposely made heavy tables to meet quotas – an obvious example of not utilizing resources as effectively as possible. Measures were taken to fight these sorts of practices, where product quality became more important.
Nevertheless, some academics have employed an ‘equal law of cheating’ approach – unless it can be shown that exaggeration is geometrically increasing, the rate of growth must remain constant. (Some have reasoned that the law of equal cheating doesn’t apply to agriculture – owing to the difficulty in keeping track of products and the series of campaigns which put pressure on various products (grain, maize, meat, milk campaigns) to exaggerate their output. Khrushchev pointed out in 1961 of the practice of some collectives purchasing butter from shops and then redelivering it to meet their required output. Further, the increase in prices paid for produce probably led to a discouragement of evasion).
Definitional problems can have significant repercussions, however. Until 1953, crop data was published in terms of biological yield, that is, ‘on-the-root crop.’ Jansy was quite right to point out the exaggeration such a definition brought; it ignored the considerable harvest losses, was distorted by inspectors who exaggerated on the root crop estimates, since delivery obligations of kolkhozy depended on them (payments to the MTS).
In 1952 the grain harvest was said to have been 130 million tons. This was officially revised to 92 million tons!
Accuracy of Indices and National Output:
According to official statistics, gross industrial output rose almost 21 times between 1928 and 1955. The highest western estimate by F Seton allows for a 12 fold rise. The lowest, by W. Nutter, allows for a five and a half fold increase.
Most (all) Western academics rejected the Soviet output figure. However, as Nove points out it is not that the indexes are deliberately ‘cooked’ but that indices by definition are aggregations lacking accuracy. Much depends on price weights. For example, A Gerschenkron’s calculations showed that from 1899 to 1939 American production of machinery increased 15 fold with 1899 price weights, but less then doubled with 1939 price weights and he emphasized the enormous difficulties due to changes in type and design. Once again, definitions and methodology are all important.
There are some serious defects in the official indices.
Firstly, that until 1950 the weights used were those of 1926-27, which essentially gives pre-industrialization weights to the fastest growing area of the economy! Further, the big rise in costs in the 30s means that the prices at which new products were introduced into the index were higher than they would have cost in 1926-27. There was then a tendency to concentrate on the production of times bearing high 1926-27 prices, even at the cost of under fulfilling plans for the less highly valued items, because plan fulfilment was measured in 1926-27 prices. This led to rising inflation in the index.
Hence, Soviet statisticians can select the base and weights which help show very large increases. For instance, base-year weights give a larger increase in output than end-year weights. When Soviet statisticians calculate a cost of living index, however, they are careful to use end year weights which minimise the increase in prices and so represent real wages in a more favourable light.
National Income Problems:
Briefly, the official claim is a 17 fold increase in national income in the period 1913-1955. This is quite a staggering figure. It is widely agreed amongst Soviet economists that the national income of the Russian empire in 1913 was approximately a fifth of that of the US.
However, if that was correct, the Soviet national income would now be well ahead of that of the United States by the early 1960s. Further, it is odd that the national income can increase 17 fold when its principal component, agriculture, showed a rise in gross output of only 70 percent.
Clearly the methodology used to calculate the increase in national income is inaccurate.
Conclusion:
I hope I have shed some light on some of the problems facing the accuracy of Soviet statistics. I will make a further post later on referencing some of the Soviet views of a new approach to statistical recording which was really interesting even in the 1960s. V. N Starovskii, Director of the Central Statistical Administration USSR, envisioned a type of automatic statistical computer system which would be highly accurate and avoid many problems of exaggeration and so forth. The methods they used then were quite crude compared to today’s standards (essentially a ‘hole in the card’ system); the internet could be of immense use in aggregating data and allowing whatever government to make an optimal plan which suits its purposes. I welcome any debate, criticism or comments.
- Vinnie.