¹ Bergson, Abram, “Development under Two Systems: Comparative Productivity Growth since 1950,” World Politics 23 (July 1971), 579–617.CrossRefGoogle Scholar Using data for the period 1950 through 1967, Bergson showed that aggregative growth and aggregative growth per worker were about equal between O.E.C.D. and East European states, but that in the East the ratio of investment to GNP was higher and the incremental capital productivity ratio was lower (which suggests lower dynamic efficiency). My results are quite similar in these respects.

² I have chosen 1979 as my end point for the calculations because it appears to have been an average year. As a result of the second oil shock, many countries in the sample experienced recessions from long-term growth paths. Even though most of the growth-rate figures in this essay are calculated by fitting regression curves to the data, this procedure does not (contrary to popular belief) eliminate difficulties in selecting end points. For instance, fitting an exponential growth curve to a GDP series that manifests a perfectly even 5 percent growth rate for 29 years and then has a temporary leveling off of its GDP for the next three years results in a calculated growth rate of 4.7 percent. If a quadratic term is added to the regression, an even greater distortion in the growth rate is obtained, not to mention a statistically significant rate of growth retardation.

Note: Gross domestic products (GDP) are roughly equal to gross national products minus income from abroad. For all countries in this study, the differences between the two aggregates are slight.

³ First, most of the official East European aggregate series for the volume of production omit services; they focus exclusively upon “material production.” Second, many components of these volume series are weighted by distorted final prices: the overall effect of the bias that is induced is difficult to judge because some of the most slowly growing sectors (e.g., agriculture) and industrial branches, as well as some rapidly growing sectors (e.g., investment goods), are underweighted. Third, in some countries, for certain periods, gross production for each branch was aggregated without any attempt to eliminate intermediate products. Since these three factors appear to give an upward bias to the official production series, numerous Western economists have attempted to recalculate the indices so as to make them more comparable with similar series for the West. The most difficult problem is the weighting problem: it has been most commonly solved by use of “adjusted factor cost weights,” a procedure that has been defended in, e.g., Bergson, Abram, The Real National Income of Soviet Russia since 1928 (Cambridge: Harvard University Press, 1961).Google Scholar

⁴ As far as I have been able to determine, the most careful Western recalculation of the Soviet GNP and its components for the period from 1950 through 1979 is the Greenslade index (exact sources are in Appendix A). For various East European countries, the only Western recalculations of the GNPs and their components for almost the entire postwar period have been prepared by Thad P. Alton and his associates at the Research Project on National Income in East Central Europe. The results of both are presented periodically in various publications of the Joint Economic Committee of the U.S. Congress.

Both the Greenslade and Alton estimates were made by calculating indices based on very large samples of production series of individual goods and services (usually in physical units). These series are first combined into a number of industrial branches using several different kinds of weights; the branches are then combined using adjusted factor-cost weights. The methodology of the Greenslade index is described in U.S. Congress, Joint Economic Committee, USSR: Measures of Economic Growth and Development, 1950–1980 (Washington, D.C.: G.P.O., 1982).Google Scholar

⁵ Because of lack of comparability with countries in Eastern Europe and also because of problems with the data, I have excluded two very small members (Luxembourg and Iceland) and one underdeveloped country (Turkey) from the O.E.C.D. list.

⁶ Similarly, given the first interpretation, the heteroskedasticity test (described in fn. 8) is necessary for the statistical tests to be carried out properly. Given the second interpretation, the heteroskedasticity test merely assures us that this particular statistical screen is of roughly equal severity across equations.

⁷ Classification problems of a less serious nature arise for certain other countries, such as Hungary and France. Hungary adopted some market elements in its reforms of January 1968; these have lasted to the present day. France has had sufficient dirigistic elements in its economy at certain times in the recent past to be considered as having a marked degree of central administration in certain sectors. However, these changes do not represent a transformation of the essential nature of the economic system, and therefore I do not believe that further adjustments to the systems variable are warranted. Sufficient data are presented below so that those disagreeing with this empirical judgment can redefine one or more systems variables and recalculate the regressions.

⁸ The analytical methods used in this article are simple, but raise a problem that should be mentioned. I have taken time-series data for the various countries, estimated particular types of statistics for each country (e.g., growth rates), and then tried to explain these statistics by a second set of regressions. Such a two-stage procedure introduces a particular type of heteroskedastic disturbances that, in turn, raise questions about the calculated standard errors. This problem has been discussed in detail by Saxonhouse, Gary R., “Regressions from Samples Having Different Characteristics,” Review of Economics and Statistics 59 (May 1977), 234–37CrossRefGoogle Scholar, and White, Halbert, “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity,” Econometrica 48 (May 1980), 817–38.CrossRefGoogle Scholar White has proposed a rather simple test to determine whether or not such heteroskedasticity is sufficiently important for any given data set to require adjustments to the standard errors in the second set of regressions. These tests were carried out on all important regressions presented in this paper; it does not appear that heteroskedasticity raises any serious problems in interpreting the regressions in the two-step procedure followed below; i.e., the standard errors appear adequate for my intended purposes. Although these tests for heteroskedasticity also indicate the occurrence of certain types of specification errors, they are by no means a cureall for detecting specification problems; for this reason, I have tried to present as much of the basic data as feasible. In this way, readers not only can see what is happening at the country level, they also have the information to calculate alternative regressions that are specified differently or that include other explanatory variables.

⁹ One apparent anomaly of the data can be eliminated quickly. The participation ratio is the ratio of the economically active (those in the work force or looking for work) to the relevant population between 15 and 65. If a considerable number of people over 65 or under 15 are participating in the labor force (e.g., in agriculture), it is possible for the participation ratio measured in this way to be greater than unity—as in the case of Portugal.

More serious problems arise in the assumption that labor stocks can be used as an approximation of labor flows. First, the length of the working day has decreased in almost all countries. Since I am using the number of economically active (at a given time) rather than the number of labor hours, my calculations of growth of the labor input have an upward bias. Second, because the reduction of seasonality is not taken into account, the data on labor input growth have a downward bias. Reduction of seasonality means that the ratio of annual labor hours to the stock of labor at any given point rises over time; therefore, the growth of the stock of labor understates the growth of the flow of labor services. The quantitative importance of this downward bias is unclear, but it probably does not offset the upward bias arising from the first problem. Third, the degree to which the various national statistical agencies handle certain tricky problems (e.g., women in agriculture) in a similar fashion is unknown. Fourth, the labor force data for certain socialist states are incomplete in the mid- and late 1970s, so that I had to make some estimates.

¹⁰ Official statistics in Eastern Europe distinguish several different types of investment series. See Dyker, David A., “A Note on the Investment Ratio in Eastern Europe,” Soviet Studies 35 (January 1982), 95–106.Google Scholar The two best-known ones are “accumulation” and “gross fixed capital investment.” The former is broader and apparently includes inventory investment, investment in livestock, and (in some countries), capital repair. As a share of national income (Marxist definition), it has increased in most East European states. The information is conveniently summarized by Feiwal, George, “A Socialist Model of Economic Development: The Polish and Bulgarian Experiences,” in Wilber, Charles K. and Jameson, Kenneth P., eds., Socialist Models of Development (Oxford: Pergamon Press, 1982), 929–51Google Scholar. In these official statistics, the aggregate of gross fixed capital investment also appears to be increasing considerably faster than net material product, so that its share is increasing as weli.

¹¹ The recalculations of East European national account data by Thad Alton and his associates do not include data on the final product side except for estimates of final private and public consumption. However, the difference between their calculated GNP values and their estimates for consumption represent a residual, the primary components of which are defense and gross fixed capital investment; from the behavior of this residual certain inferences can be drawn about the nature of the GFCI series. From 1965 to 1980, the ratio of this residual to the GNP has declined in Bulgaria, Hungary, and Poland, and has risen slightly in Czechoslovakia, East Germany, and Romania. If we assume that defense expenditures and GNP have risen at roughly the same rate during this period, then Alton's estimates suggest that in this period the actual growth of gross investment was roughly similar to the growth in GNP (i.e., a significant percentage below that derived from the GFCI series). Thus, the assumption that gross investment grew at roughly the same rate in the two parts of Europe is plausible. See Alton, Thad and others, Eastern Europe: Domestic Final Uses of Gross Product, 1965, igyo, and 1975–1981, Occasional Paper No. 72, Research Project on National Income in East Central Europe (New York: 1982).Google Scholar

¹² See Gregory, Paul, Socialist and Non-Socialist Industrialization Patterns (New York: Praeger, 1971Google Scholar, and Ofer, Gur, The Service Sector in Soviet Economic Growth: A Comparative Study (Cambridge: Harvard University Press, 1973).Google Scholar

¹³ These results falsify the argument sometimes found in the East European literature that dynamic efficiency is greater in the East because of the Wests growing unemployment and more violent business cycles. The counterargument that the East European countries have a lower dynamic efficiency because of their growing underemployment of labor has received no empirical verification in this study either.

¹⁴ Economists in both East and West have been paying increasing empirical and theoretical attention to problems of cyclical fluctuations in planned economies. An extensive history of doctrine on the subject is provided by Sabov, Zoltan, Zyklische wirtschaftliche Aktivitaetschwankungen in sozialistischen Planwirtschaften [Cyclical variations in economic activity in socialist planned economies] (Berlin: Duncker & Humblot, 1983).Google Scholar Most of the theories of cyclical behavior in socialism are based either on an investment cycle approach Goldmann, and Kouba, Karel, Economic Growth in Czechoslovakia [White Plains, N.Y.: International Arts and Science Press, 1969])Google Scholar or on a planning cycle approach (e.g., Kyn, Oldrich and others, “Simulation des Einflusses der Planung auf die sowjetische Wirtschaft” [Simulation of the influence of planning on the Soviet economy], Schriften des Vereins fuer Sozialpolitik N.F. [West Berlin, 1978] 98).Google Scholar A number of others have contributed to the theoretical debate: Bajt, Alexander, “Investment Cycles in European Socialist Economies: A Review Article,” Journal of Economic Literature 9 (March 1971), 53–63Google Scholar; Cobeljić, N. and Stojanović, R., “A Contribution to the Study of Investment Cycles in the Socialist Economies,” Eastern European Economics 2 (Numbers 1–2, 1963–64)CrossRefGoogle Scholar; Horvat, Branko, Business Cycles in Yugoslavia (White Plains, N.Y.: International Arts and Science Press, 1971)Google Scholar; Hutchings, Raymond, “Periodic Fluctuations in Soviet Industrial Growth Rates,” Soviet Studies 20 (January 1969), 331–53CrossRefGoogle Scholar; Kyn, Oldrich and others, “Growth Cycles in Centrally Planned Economies: An Empirical Test,” in Kyn, O. and Schrettl, W., eds., On the Stability of Contemporary Economic Systems: Proceedings of the 3d Reisenburg Symposium (Goettingen: 1979)Google Scholar; Montias, John M., “Socialist Trade and Industrialization,” in Brown, Alan and Neuberger, Egon, eds., Foreign Trade and Central Planning (Berkeley: University of California Press, 1968)Google Scholar; Olivera, Julio H. G., “Cyclical Economic Growth Under Collectivism,” Kyklos 13 (No. 2, 1960), 229–55CrossRefGoogle Scholar; Uffhausen, Richard, “Simulation von Investitions- und Planungszyklen in der sowjetischen Wirtschaft” [Simulation of investment and planning cycles in the Soviet economy], Working Paper 74 (Osteuropa-Institut Muenchen, October 1980)Google Scholar; Wiles, Peter, “Are there Any Communist Economic Cycles?” ACES Bulletin 24 (Summer 1982), 1–21.Google Scholar

A number of empirical studies have also been made: Neuberger, Egon, “Is the U.S.S.R. Superior to the West as a Market for Primary Products?” Review of Economics and Statistics 46 (August 1964), 287–93CrossRefGoogle Scholar; Staller, George J., “Fluctuations in Economic Activity: Planned and Free-Market Economies, 1950–60,” American Economic Review 54 (June 1964), 385–95Google Scholar; Staller, George J., “Patterns of Stability in Foreign Trade: OECD and Comecon, 1950–63,” American Economic Review 57 (September 1967), 879–88Google Scholar; Wiles (see above).

¹⁵ Another possible measure of fluctuations is (1 - R)², which is just the SEE squared and divided by the variance of the dependent variable (a function of its growth rate). For a number of reasons, the SEE seemed a more appropriate measure for exploration of fluctuations. As one might expect, in the experiments using the (1 - R)² measure, the economic systems variable played roughly the same role.

¹⁶ “Cobwebs” (so called because the diagram to explain them looks like a cobweb if drawn correctly) are instabilities arising because production takes a long time (e.g., a growing season) and producers do not know the market price they will finally obtain and must estimate it. Famous cobweb phenomena include the hog cycle and the hog-corn cycle.

¹⁷ Pryor, Frederic L. and Solomon, Fred, “Commodity Cycles as a Random Process,” European Journal of Agricultural Economics 9 (No. 3, 1982), 327–47.CrossRefGoogle Scholar