Estimations of Russian Historical National Accounts

Alexey Ponomarenko


The calculation of long-term historical statistics for Russia has been included as a part of the Asian Historical Economic Statistics Project as the Russian economy is closely linked to those of countries that one normally associates with the word "Asia." The general objective of this Project is to reconstruct long-term statistical time series for various Asian countries, data that will ideally facilitate academic research and government planning.

Long-term time-series statistics are useful not only for historians but for economists as well. For example, our understanding of the current economic situation in a given country, such as Russia, improves if the analysis takes into account long-term dynamics and structural changes. Furthermore, in order to construct mathematical forecasting models, it is necessary to have a long-term series of indicators. Thus, there is a wide range of possibilities for academic and policy applications of the results of this Project. Consequently, the Russian government statistical service has decided to support the Russian estimation work of the Asian Historical Statistics Project.

Each country possesses unique challenges and problems for researchers attempting to generate historical estimations. Some countries, due to political instability, wars, boundary changes, and/or poor levels of public administration, never possessed strong statistical systems. For other countries, the methodological problems are more significant, since national statistics were often based on different systems of estimation.

Both types of problems are relevant to the Russian case. In this paper, I highlight the distinctive features and problems involved in generating long-term economic statistics for Russia. The three main characteristics of Russian economic statistics are as follows. First, during the 20th century, Russia experienced two revolutions, extremely bloody civil and world wars, and a number of other great changes and collapses. Second, it cannot be forgotten that for much of the 20th century, Russia was part of the USSR and that the Russian economy and corresponding statistics formed only one portion of the Soviet Union's total economy and statistics. Third, the System of Material Products (MPS), a system quite different from the System of National Accounts (SNA), was used by the Russian national statistical service for macro-economic estimations until very recently.

Therefore, it is very difficult to compile a long-term Western-style statistical series for Russia. It is almost impossible to find a unified approach to estimation of statistical indicators for all of Russia's modern history owing to strong differences in sources of information and quality of statistical data for every period. The various historical stages must be clearly distinguished and original methods of estimation for every stage must be reconstructed. I use the following periodization: the pre-revolutionary period (1900-1915); the Inter-War period from after the civil war to World War II (1925-1940); the Post-World War II period (1945-1959); and the final period of the centrally- planed economy (1960-1990). Any statistical data for the highly unstable points in Russian history (wars and revolutions) can only be evaluated through indirect methods.

In the Project, we are at present calculating the statistical time series for Russia for the period thirty-year period, 1960 -1990. This is the first step of an international joint project that was begun in June 1989.

There are estimations of Soviet GDP for past years--some made by the CIA, and some by Soviet and foreign economists. However, there are several caveats in employing these estimates. First, these GDP estimations were for the Soviet Union as a whole, and not necessarily focused solely on Russia. While these days, many researchers require separate statistics on the modern Russian economy, in the Soviet period, the Russian economy by itself was not a common research area. Second, there was not enough direct information for estimations in the past, because a significant portion of the statistical data was never published or made public. Now we have the full cooperation of the official Russian statistical service, and information from Russian archives is becoming increasingly available. Third, methods employed for alternative estimations were at times too primitive; for example, the previous methods did not take into account secular, long-term qualitative changes. Our intention is to employ more accurate methods which compensate for these oversights. Fourth, the previous statistics were mostly estimations of real growth of the Soviet GDP. Our plan is to construct a full set of national accounts for Russia at current prices, and also to produce estimations of real indexes by several alternative methods. Therefore, the long-term economic statistics for Russia generated by this Project will be groundbreaking in many respects.

This is not to deny the utility of official Soviet and Russian statistics. Generally speaking, Soviet (and Russian) statistics were very accurate. As the Soviet economy was centrally planned, statistics were useful tools for controlling and supervising the implementation of government economic plans. Due to this reason, the Soviet government paid a great deal of attention to statistics and reporting discipline. Under Soviet law, if a Soviet firm sent-in false reports, this would make the corporation open to criminal charges, and the director or accountant could be arrested as a result. There was a thorough coverage of firms through regular statistical surveys in industry, agriculture, transportation, construction, and other areas of economic activity.

Starting from the early 1960s, the Soviet government began publishing macro-economic indicators of MPS both for USSR as a whole and for every republic-member of the Union. The MPS is a system of macro-economic indicators that cover the so-called "material sphere" of the national economy. The "material sphere" includes industry, agriculture, construction, cargo transportation, trade and other activities, all related to the direct production of "real" goods, or services involved in the production of goods.

On the other hand, these were not ideal statistics; the main problem was that much of it was propaganda. The Soviet statistical office did not publish any statistical data that could be used to discredit the Soviet government. This, however, does not necessitate the conclusion that every statistical data was used on behalf of the government. For instance, macroeconomic indicators at current prices, such as National Income or the main aggregates within it, have little practical utility as tools for propaganda or counter-propaganda. Only indicators that were comparable with the same indicators as those of the West were deployed as propaganda. These were National Income and GDP in dollars terms, indexes of real GDP growth, and the share of military spending in total expenditures. Naturally, it would be a mistake to use these official statistics without any adjustments for these politically-based biases.

Generating GDP in real terms is our objective. GDP in dollar terms, while a very interesting issue, is not included within this Project due to the number of specific problems that must be solved on order to generate the appropriate calculations. It is a very difficult process to generate accurate estimations of real GDP indexes. Theoretically, there are two different approaches to the estimation of real indicators. One approach uses a number of deflators, which are based on direct information related to price changes such as CPI or other price indexes, for estimation purposes. The second approach involves using volume indexes (such as indexes of production in physical term - tons of steel or coal, kilowatts of electricity, even number of employees in some industries). In this case an implicit deflator may be compiled during the last stage.

The first approach is more traditional and is used in many countries for estimation. However, the problem is that there may be errors in the official Soviet price indexes; therefore, we cannot use them without controls. In contrast, Soviet statistics in physical terms appear to be more reliable. This is the main reason why most alternative estimations (CIA and others) are based on the second approach using volume indexes. Unfortunately, this approach also has some problematic features.

First, paucity of information from certain industries causes some difficulties. A large number of indicators in physical terms were published, but for some branches, such as military production, machinery, chemistry, non-ferrous metallurgy and others, the data is insufficient for accurate estimations. Moreover, volume indexes do not take into account any changes in the quality of goods produced and services. This is not such a serious issue for short-term estimations, such as estimations of dynamic indexes for one year. But for long-term estimations (30 years as in our case), the quality changes are very significant especially in some sectors such as machinery. For instance, according to CIA estimations, the cost of one jet fighter rose about ten times over a long-term span. The fact that the volume indexes ignore such drastic changes places clear limits on the utility of the data. Second, there is the problem of so-called "double deflation." Theoretically, for accurate GDP recompilation into constant prices, output and intermediate consumption should be deflated by two different deflators because different price changes may take place. In the Soviet case, double deflating is especially significant because state price policy for raw material production controls served as a form of subsidy for final producers. Methods of estimation using volume indexes cannot take this into account.

Thus, both these approaches have strong and weak features. A third "mixed" approach, used in the US for estimation of indexes in housing services, may be employed. This third method was based on using volume indexes for living space but with adjustments for the rise of quality (e.g., utilities). This is admittedly a more complicated approach but it can produce useful results. In our project we intend to use all three approaches for estimation. Of course, the results generated by each method will be different. However, users of our data can select from three options for data in accordance with their research methods, knowledge, and concrete purposes.

Estimation of military spending for USSR is another major problem. There is general agreement among specialists that official published figures are for the most part incorrect. Reasonably accurate data on military spending was published officially only one time: in 1989, during Gorbachov's PERESTROIKA. Thus, the challenge is to produce accurate estimations using this one publication in combination with CIA estimations of USSR military spending.

In several other aspects, as acknowledged even by CIA experts, the quality levels of Soviet statistics were very high. The first basis for the CIA's conclusion was that during WW II, when some Soviet documents containing classified statistical data fell into the hands of the Germans. At the end of the war, American experts used this data to analyze the Soviet statistical system. Except military production, which was never published, all other indicators matched exactly the officially published data. The second basis for the CIA's assessment was related to the curious entry in Soviet statistics called the "index of plan execution." This indicator was used for internal propaganda during the Soviet period. In accordance with Soviet ideology, the state plans were executed and even "over-executed" by enterprises. If in practice, some enterprises or the economy as a whole could not produce the goods in question, the figures for the published state economic plans (but never the actual reported statistics) were "corrected" to "prove" the efficacy of the plan. This practice indicated that the Soviet statistical reporting system was quite strong in actuality, and that while the publicized statistics were altered for propaganda purposes, the actual reported data was never doctored for propaganda needs.

Therefore, we can conclude that Soviet MPS statistics can be used as a base for the estimation of western-style GDP and other SNA indicators. The main approaches to the conversion of MPS indicators into corresponding indicators of SNA were worked out by UN experts many years ago, and it has been employed by many researchers for estimations. For instance, in the Asian Historical Statistics Project, Chinese MPS indicators were converted into SNA indicators by Professor Masaaki Kuboniwa of Hitotsubashi University and a number of Chinese researchers. But Russian statistics contain some problems quite different than those associated with Chinese statistics. First, Russia was not independent country in Soviet times but merely one region within the Soviet Union. As a result, we do not have any separate information on certain kinds of government spending for each region, such as Russia, Ukraine, or others. For instance, some parts of government expenditures for state administration were financed by the separate budgets of each region, such as those of the Russian republic, but other portions were financed through the Federal (Union's) budget. For estimations of Russian GDP, the government expenditures in the Federal budget (not only on state administration, but also for defense, science and some other non-market services) must be divided among the former republics of the Soviet Union. Therefore, new methods for this distribution and recalculation must be established.

Moreover, although the general approaches to conversation of MPS indicators into SNA standard are well-known, concrete methods of estimation depend considerably on the type and quality of statistical data available. Traditionally, researchers paid more attention to the estimation of GDP from the production side for Russia and USSR, then GDP for income and expenditure. This was primarily due to the fact that the published statistical information for production was much better than for those of other areas. However, GDP from the expenditure side is also very important for analysis of Russian economic trends. Recently, we were able to access statistical data from Russian archives that allow us to make estimations of GDP from the expenditure side. Further, this data will enable us to generate estimations of Russian historical GDP for income, not only at current and constant prices, but also at the aggregate level, and in great detail. As far as I am aware, there have been no similar publications or projects undertaken in the past. Consequently, the COE project constitutes the first time that a systematic effort has been made to compile a full set of National Accounts for Russia for period prior to the 1990s by non-government researchers.

A full set of National Accounts opens up new possibilities for analysis. Nevertheless, this is not the only reason why we are constructing this statistical model. Another objective is to improve possibilities for controlling the quality of results. If only one indicator or time series (for example, GDP from the production side) is compiled, it is difficult to determine the level of accuracy of the data. But if we compile a set of related indicators, we can make comparisons and make better assessments regarding quality. There is a special indicator for measuring accuracy, which is called "statistical discrepancy." This indicates the discrepancy between GDP by value added (production approach) and GDP by expenditures in terms of percentages of GDP. In ideal cases, discrepancy will be near zero, but in practice it usually numbers in single digit percentages. In our case, the statistical discrepancy has been reasonable thus far.

Another method for quality control is comparison with official publications. Russian statistical service began official estimation of GDP after the collapse of the Soviet Union. The first published results were for 1989. Some international public agencies, such as OECD, World Bank and IMF, provided technical assistance and supervision, with the Russian official statisticians providing all the statistical information. Thus, the officially published Russian GDP after 1989, while not at ideal levels, are quite good. We did not limit our estimations to methods used for 1988, but calculated GDP for 1989 and 1990 using common methods and data sources that are available for later years. When we compared our estimates and official calculations of GDP for these years we discovered that there were no significant differences. Thus, we concluded that the methods and data sources currently in use in Russia are applicable for estimations for other years.

The estimations are still in progress. At the end of our project, we will prepare a special report that includes detail statistical tables and explanations of methodology. As the initial step for the project, preliminary estimations were made and first results discussed with Japanese experts in the spring of 1998. We were able to collect a number of useful comments and remarks. As a result, we improved methods of estimation, and additional sources of data were added. The new estimations were prepared this autumn. The second workshop, at which Russian experts participated, was held this past September. In addition, some details of the methods used in our project were presented and discussed at the International Seminar on Inflation Measurement in Great Britain this past August. At this point, our aim is to have the final report ready by the start of January 2000 and present the results at International Symposium in Tokyo.

We hope that results of our work prove to be of great interest for many researchers in Japan, Russia and other parts of the world. After this stage of Project is finished, the next stage will be estimations of Russia historical statistics for earlier years. Although there is a great paucity in statistical data for earlier periods, other methods may be developed and employed for the historical estimations. Also our experience of conversion of MPS data into SNA indicators may be applicable for estimations of the same type of historical time-series analysis for other countries of the former Soviet Unions, such as Uzbekistan, Kyrghyzstan, Turkmenistan, and perhaps Mongolia as well. These countries have expressed strong interest in our project, and future cooperation may be coordinated in various forms. In short, we have many possibilities for continuing and extending of our current project in the near future.

(Institute of Economic Research, Hitotsubashi University)