The relative efficiencies of market and planned economies.

The advantage to efficiencies of the decentralization of decisions and of individual responsibility is even greater, perhaps, than the nineteenth century supposed; and the reaction against the appeal to self interest may have gone too far . . . The authoritarian state systems of today seem to solve the problem of unemployment at the expense of efficiency and freedom. John Maynard Keynes [16, 380-81]

. . . it was not the possibility of planning as such which has been questioned . . . ., but the possibility of successful planning. . . . What we should anticipate is that output, where the use of available resources was determined by some central authority, would be lower than if the price mechanism of a market operated freely under otherwise similar circumstances. Friedrich A. Hayek [11, 203-04]

I. Introduction

Are planned economies systematically more or less productive than market economies? This question provoked heated intellectual debate since the time of Hayek [11], Keynes [16], Lange [19], Lange and Taylor [20] and Lerner [22]. The collapse of the communist systems of East Europe since 1989 closed the debate: planned economies were without doubt less efficient in utilizing economic resources than market economies. But how much less efficient? This question is of obvious scientific interest, but it has not been answered.

Some work addressing the broad question of comparative efficiency has been done by Bergson [4; 5; 6], Mater [25; 26], Moroney [28], and others. All indicate comparatively lower efficiency in planned economies. But none addresses the question "by how much?" Our contribution here is to answer this question using a novel methodology. We estimate aggregate frontier production functions for two groups of countries: 7 centrally-planned economies in East Europe and 17 predominantly market economies in West Europe.

Estimating frontier production functions enables two types of comparisons that have not before been possible. First, we can compare the overall efficiencies of East European countries relative to West European countries. We find that, overall, the planned economies were about three-fourths as efficient as the market economies. Second, we can compare relative economic efficiencies among countries within East Europe. Here we find moderate variation, with all 7 countries operating within 70% to 83% of the aggregate East European frontier.

II. Brief History of the Debate

The argument favoring central planning is that competitive markets are unstable, leading to swings of overinvestment in some sectors and underinvestment in others, thereby generating a wasteful use of resources. Such waste could in principle be avoided by central planning of production, prices and resource allocation.

Hayek [11; 12], Mises [27] and other proponents of market systems acknowledged the theoretical possibility of rational centralized price setting and resource allocation. But they reasoned that the stupendous volume of information required for centrally directed resource allocation is impracticable. No central authority could assimilate it, much less act on it. Instead, they forcibly argue that prices set in decentralized markets prompt greater overall economic efficiency.

The computational feasibility of centralized resource allocation in a flexible price setup was later analyzed formally by Arrow and Hurwicz [1] and Malinvaud [24]. And the strictly technical feasibility of allocating resources without either market or accounting prices was demonstrated by Heal [13; 14]. But as Heal readily acknowledges, the information requirements on the part of the Central Planning Board and of plant managers are more stringent than in either the Arrow-Hurwicz or Malinvaud models.

The collapse of the centralized economic systems of East Europe since 1989 shows the serious flaws that plagued them. And the fact that real per capita incomes grew much faster in West Europe than in East Europe after World War II also signifies productivity advantages of decentralization.(1)

Bergson [4; 5] found four planned economies to be generally less efficient than seven market economies in their use of capital and agricultural land in 1975. Moroney [28] found seven East European planned economies to use capital and energy generally less efficiently than seventeen West European market economies during the 1978-1980 period. Both Bergson and Moroney used ordinary least squares to estimate a constant-returns-to-scale Cobb-Douglas production function, with a dummy variable to identify the planned economies. Neither Bergson nor Moroney could address the degree of technical inefficiency among the economies of Eastern Europe. We do so here.

Our objective is to estimate stochastic production frontiers. We identify the separate production frontiers of the two types of economies, and position different countries in East Europe relative to their own frontier. We use Moroney's data, which comprise a 24 x 3 panel of 17 market and seven planned economies over the three year period 1978-1980. Countries are listed in Table III. They include three of Bergson's four planned economies plus four others, and five of Bergson's seven market economies plus twelve others. These panel data enable us to exploit econometric techniques unavailable to Bergson and Moroney. We use maximum likelihood techniques to construct a stochastic production frontier, relative to which the "firm effects" provide country-specific measures of productive efficiency. The model allows for the possibility of neutral and non-neutral (in the sense of Hicks) differences in technical efficiency between the groups of countries.

We test the hypothesis that there is no systematic difference in performance across the two economic systems. If the hypothesis is rejected, and if the market economies are systematically more productive, then the Bergson-Moroney hypothesis is both confirmed and quantified to a degree that heretofore has not been possible. We do find systemic differences. We find a large gap in technical efficiency between the two systems. The lower efficiency of the centrally-planned economies was present a full decade before the turmoil that rent them politically and economically. The differences we find in economic efficiency are in no way ascribable to the disruptions encountered during "the transition." These countries have deteriorated sharply since their political disintegration. The reasons for their economic collapse during the transition are a very different matter from the reasons for lower efficiency when central planning was intact.(2) We focus strictly on comparative efficiency when Eastern Europe was governed by central planning Soviet style.

III. The Stochastic Frontier Model

We follow Bergson and Moroney by assuming a constant returns to scale Cobb-Douglas production technology,(3) except for the fact that now these features characterize the deterministic kernel of a stochastic production frontier, which we write as

ln[(Y/L).sub.it] = [a.sub.0] + [a.sub.e] ln[(E/L).sub.it] + [a.sub.k] ln[(K/L).sub.it] + [a.sub.D]D

+ ln [a.sub.DE]D [multiplied by] ln[(E/L).sub.it] + [a.sub.DK]D [multiplied by] ln[(K/L).sub.it] + [v.sub.it] - [u.sub.i], (1)

where i = 1, . . . , 24 indexes countries and t = 1, 2, 3 indexes the years 1978-1980. (Y/L), (E/L) and (K/L) are output per worker, energy per worker, and capital per worker, respectively, and are described below. D is a dummy variable that takes a value of zero for market economies and one for planned economies. Statistical noise is represented by the error component [Mathematical Expression Omitted], and productive inefficiency is represented by the error component [Mathematical Expression Omitted]. We assume that noise and inefficiency are uncorrelated, and that each is uncorrelated with the regressors.(4)

Equation (1) is interpreted as follows. The first three terms on the right side describe the deterministic kernel of the stochastic production frontier. The next three terms allow planned economies to have different regression intercepts, conditioned in part by energy per worker and capital per worker (hence non-neutrally different production frontiers) than market economies. The intercept for the market economies is [a.sub.0], which is constant across countries and over time, while that for the planned economies is [[a.sub.0] + [a.sub.D]D + [a.sub.DE]D ln[(E/L).sub.it] + [a.sub.DK]D ln[(K/L.).sub.it]], which varies across countries and over time. This last set of terms thus allows individual planned economy frontiers to be close to or far from the market economy frontier and to either close or widen its gap over time, depending on their input mixes.

Adding the error component [v.sub.it] makes allowance for the usual sources of statistical noise, and generates the stochastic production frontier. Subtracting the error component [u.sub.i] allows for technical inefficiency in production, and imposes the requirement that actual output in a country cannot exceed maximum output defined by its stochastic production frontier. Note that statistical noise varies across countries and over time, but inefficiency varies only across countries. This seems a reasonable restriction in so short a time span.

[(Productivity).sub.it] = exp{[a.sub.0] + [a.sub.D]D + [a.sub.DE]D x ln[(ElL).sub.it] + [a.sub.D]D x ln[(K/L).sub.it]}. (2)

The performance of a planned economy is evaluated in two ways: (1) its productivity relative to other economies, and (2) its efficiency in relation to its own group's stochastic production frontier. Productivity in country i in period t is measured by and so the ratio of the productivity of a planned economy to that of the market economies is given by

[(Productivity Ratio).sub.it] = exp{[a.sub.D] + [a.sub.DE] ln[(ElL).sub.it] + [a.sub.DK] ln[(K/L).sub.it]}. (3)

This ratio varies across planned economies and over time. The efficiency of a planned economy is given by

[(Efficiency).sub.i] = E[exp{[u.sub.i][where][v.sub.it] - [u.sub.i]}]. (4)

Efficiency varies across the seven planned economies, but not over time.(5)

Maximizing the likelihood function for equation (1) yields estimates of the parameters of production technology ([a.sub.0], [a.sub.K], [a.sub.E], [a.sub.D], [a.sub.De] and [a.sub.DK]:), the noise parameter ([Mathematical Expression Omitted]), and the technical efficiency parameters [Mathematical Expression Omitted]. [Mu] is a placement parameter allowing the modeal value of is a spread parameter allowing efficiency values to inefficiency to be zero or positive, and [Mathematical Expression Omitted] be clustered closely together or widely dispersed. The results reported below are based on a re-parameterization [Mathematical Expression Omitted] and [Mathematical Expression Omitted] Details of the model and the maximum likelihood estimation procedure can be found in Battese and Coelli [3] and Coelli [8].

Although econometric estimation is straightforward, the interpretation of technical inefficiency is not. It is best to be clear about why it is not. What we identify as technical inefficiency actually contains two elements of allocative inefficiency. The first is attributable to an economy's failure to use the least-cost combination of energy, capital, and labor. If one had reliable estimates of the opportunity costs of these inputs, one could estimate a set of cost-minimizing input demand equations, thereby separating allocative inefficiency in input use from strictly technical inefficiency (see Schmidt and Lovell [31] and Forsund, Schmidt, and Lovell [10]). This procedure is not possible for two reasons: because reliable opportunity costs of capital are not available for the planned economies, and because the price of energy to all CMEA countries was heavily subsidized by the Soviet Union (energy prices determined by central planners did not reflect true opportunity costs).

A second sort of allocative inefficiency creeps in since the CMEA countries produced a GDP mix dictated by planning boards rather than by market prices. Thus even if we ignore system-wide differences in the qualities of goods and services (taken into account as best they could by Robert Summers, Alan Heston, the late Irving Kravis, and other designers of the International Comparison Project) the GAPS of the CMEA countries embody a mix of products that may be allocatively inefficient. What we call technical inefficiency has a pretty broad interpretation. Even so, it is interesting and important and worth knowing. At this level of aggregation, it seems best not to worry (or worry much) about a strict interpretation.

IV. The International Sample(6)

We link real GDP per worker to aggregate energy consumption per worker and net capital stocks per worker using cross section samples for the most recent years for which complete, comparable data are available, 1978, 1979 and 1980. Labor input is adjusted, in each country, for labor force participation rates. These rates are generally higher in Eastern Europe (ranging for example in 1980 from 47.4 percent in Hungary to 54.9 percent in Romania) than in Western Europe (from 35.1 percent in Spain to 48.9 percent in Switzerland). Labor is not adjusted for average weekly hours, or cross-country differences in education levels. Such adjustments are not very important.(7) So our ignoring these cross-country differences does not affect one's interpretation of the statistical results.

Real GDP estimates were prepared under Phase III of the United Nations International Comparison Project for three CMEA countries: Hungary, Poland and Romania. The more extensive analysis in this paper is made possible by the meticulous estimates of real GDP per capita for the former Soviet Union, Bulgaria, Czechoslovakia and East Germany published by Summers and Heston [34], and based on International Comparison Project methodology. A thorough appraisal of this work is given by Marer [26]. These are the most closely comparable estimates of real output per head ever assembled. They are adjusted annually for changes in purchasing power parity and expressed in 1975 U.S. dollars per capita; hence they are directly comparable across countries and time. The methodology used in their construction is described fully in Kravis, Heston and Summers [18], Kravis [17] and Marer [26].

Some authorities, Bergson [5, Ch. 3] for example, believe the purchasing power parity ratios for planned economies overstate the dollar value of their currencies, and thus their national output. If so, two consequences follow: (1) the estimated position of the planned economies' frontier relative to the market economies' frontier is overstated, hence (2) the estimated relative inefficiency of planned economies is understated.

Energy consumption per worker is measured comprehensively. It includes all energy attributable to hydrocarbons: crude petroleum, refined petroleum products, natural gas liquids and natural gas; solid fuels: anthracite and bituminous coal, lignite and brown coal, coke, charcoal, fuelwood and bagasse; and electricity (whether generated hydroelectrically or from fossil fuels or from nuclear fuel). All forms of energy are converted to kilograms of standard coal equivalency, 7,000 calories per gram, and hence they are fully comparable across countries. These data are obtained from the 1981 Yearbook of World Energy Statistics [35]. The coefficients used to convert the alternative forms of energy to standard coal equivalency are published in this volume.

For all countries capital is measured by economy-wide net fixed capital stocks, expressed in 1975 U.S. dollars. West European capital stocks are developed from benchmark estimates published by Leamer [21]. Those of the East European countries are based on official government estimates of each country, compiled by the L. W. International Financial Research, Inc., Research Project on National Income in East Central Europe. These estimates are then converted to 1975 U.S. dollars by applying purchasing power parity ratios from the International Comparison Project.

Two points concerning the capital stock estimates bear mentioning. The first is an accounting matter. The net capital stocks for the East European countries are based on the assumption of 15-year asset lives, the same as for the market economies. Some authorities believe that assets are longer-lived in the CMEA than in most market economies. Thus a potential error of too-rapid depreciation may understate the East European capital stocks. Quantitatively, however, this potential error is small.

The second point concerns differences in the quality of capital. The CMEA capital stocks are reported in national currencies. Most East European experts believe the purchasing power parities used to convert these stocks to 1975 U.S. dollars tend to overstate CMEA capital stocks because the lower quality of East European capital goods is inadequately captured by the conversion, as discussed by Marer [26], and by Bergson [4], and by Schroeder and Edwards [32]. On the other hand, the principal designers of the International Comparison Project contend that the purchasing power parity ratios adequately account for such quality differences [26]. In any event, any depreciation and quality errors are qualitatively offsetting and likely to be quantitatively small.

V. The Empirical Results

Empirical findings are collected in Tables I and II. Table I reports the MLE regression. The economic implications are clear. Both energy per worker and capital per worker contribute positively to output per worker, although the contribution of energy is small and not statistically significant. The planned economy dummy variable terms are individually and collectively statistically significant).(8) The estimated value of [a.sub.D] implies that the planned economies have significantly inferior technologies, with maximum potential equal to exp {[Mathematical Expression Omitted]} = 47% of that of the market economies at the same input mix. This is the Hicks-neutral difference in technical efficiencies.

Table I. Production Frontier Parameter Estimates (t-statistics),
1978-1980

                                         Maximum Likelihood
Variable                  Parameter      Parameter Estimate

Intercept                  [a.sub.0]        2.445
                                          (19.337)

ln(E/L)                     [a.sub.E]       0.033
                                           (0.653)

ln(K/L)                     [a.sub.K]       0.440
                                           (7.068)

D                           [a.sub.D]      -0.751
                                          (-2.669)

D [multiplied by]In(E/L)    [a.sub.DE]      0.255
                                           (3.242)

D [multiplied by] ln(K/L)   [a.sub.DK]     -0.147
                                           (2.180)

                       [[Sigma].sub.2]      0.077
                                          (2.614)

                            [Gamma]         0.990
                                         (282.511)

                            [Mu]          -0.727
                                         (-2.137)

log likelihood                           121.973
# iterations                                  32
Table II. Productivity Ratios for Planned Economies, 1978-1980

Country           1978     1979     1980     Mean

Bulgaria          0.828    0.825    0.823    0.826
Czechoslovakia    0.759    0.764    0.772    0.765
East Germany      0.765    0.767    0.767    0.766
Hungary           0.684    0.688    0.693    0.688
Poland            0.767    0.764    0.766    0.766
Romania           0.765    0.772    0.780    0.773
USSR              0.737    0.742    0.742    0.740
mean              0.757    0.759    0.763    0.759

However this differential is partly offset by the fact that the planned economies have technologies that are significantly more energy intensive ([Mathematical Expression Omitted]) and significantly less capital intensive ([Mathematical Expression Omitted]) than the market economies. The two efficiency error component parameters are statistically significant, and the estimated value of [Gamma] implies that most of the deviation from the fitted production frontier takes the form of deviation beneath it attributable to inefficiency. The stochastic production frontier framework is clearly applicable, even after allowance is made for a productivity differential across the two types of economic system, and for time-varying productivity differentials among the planned economies.

Table II reports productivity ratios for the seven planned economies. These ratios are calculated using equation (3), and show the ratio of the production potential of a planned economy to that of the group of market economies. The planned economies have an overall productivity potential (technical efficiency, broadly conceived) 76% as high as the market economies. This ratio edged up imperceptibly over the three-year period, and masks some variation across countries.

Table III. Seventeen Market Economies and Seven Centrally Planned
Economies, 1978-1980

Market Economies      Planned Economies

Austria               Bulgaria
Belgium               Czechoslovakia
France                East Germany
West Germany          Hungary
The Netherlands       Poland
Switzerland           Romania
The United Kingdom    USSR
Denmark
Finland
Iceland
Norway
Sweden
Italy
Malta
Portugal
Spain
Turkey

VI. Conclusions

Over the period 1978-1980 real GDP per worker averaged about $12,600 for the seventeen European market economies, but only $7,775 for the seven CMEA countries, using 1975 U.S. dollars. This 38% difference is explained in large part by the 24% productivity gap we have identified. The planned economies used their economic resources about 76% as efficiently as the market economies of Western Europe.

We view our estimate of 24% lower technical efficiency in the CMEA countries as an upper bound on their relative performance for two reasons. First, the purchasing power parity rates used to convert East European currencies to 1975 U.S. dollars probably overstate the real GDP per capita of the CMEA countries as stressed by Schroeder and Edwards [32], and Bergson [4; 5]. SEcond, many experts, especially Bergson [6], now believe that the International Comparison Project discounted inadequately for the lower quality of CMEA goods and services. If so, real GDP per capita in the planned economies is overstated in the published data we have used. These reasons reinforce each other to overstate CMEA technical efficiency.

That East Europe was, at best, 76% as efficient as West Europe is a worthwhile pre-transition benchmark. Hayek [11] and Keynes [16] put clearly their conjectures about comparative efficiency. But in 1935 and 1936, before the postwar Soviet consolidation of East Europe, they could do no more than conjecture. This paper provides a convincing element of empirical confirmation 60 years later.

1. Bergson [6] focuses on the much lower consumption per worker in the former USSR than in a broad sample of 13 OECD countries, in 1985, and the sharply inferior quality of Soviet goods.

2. The major problems encountered by East European countries during their transformation to market-type systems are reviewed by Dornbusch and Wolf [9], Schleifer and Vishny [33], Hitchens, Wagner and Birnie [15], Murphy, Schleifer, and Vishny [30], and in several papers published in the book edited by Aslund [2].

3. Although aggregate production functions can be a subject of controversy, there seems to be no better framework for estimating aggregate economic efficiency. At the aggregate (economy-wide) level, constant returns-to-scale is virtually compelling. Caves and Barton [7, 21-25] recommend the use of stochastic translog production frontiers rather than the Cobb-Douglas. Their suggestion has much merit for estimation with firm-specific or industry-specific data (in their study), since the translog need not be homogeneous nor homothetic. At the aggregate level, however, the simple Cobb-Douglas may well be preferable. For comparison of aggregate translog and Cobb-Douglas production functions, see Moroney [29].

4. Stochastic independence between the composite disturbance term [v.sub.it], and the regressors holds true if managers cannot know the effect of the technical disturbance until after the inputs have been preselected. See Zellner, Kmenta, and Dreze [36].

5. The expected value in equation (4) depends on the parameters [Mathematical Expression Omitted] and [Mu], estimates of which are obtained, along with estimates of the technology parameters, by MLE. These parameter estimates impose structure on the normally distributed noise error component and the truncated normally distributed inefficiency error component. The expectation operator then selects the "most likely" (maximum likelihood) degree of inefficiency from the regression residual for each observation.

6. This section summarizes a much more detailed discussion of the variables and how they were constructed. For full details, see Morohey [28].

7. Neglecting to adjust labor input for average weekly hours and labor quality probably makes little difference. Bergson [5, Chapter 2, Appendix Table 7] found average weekly hours in 1975 to be quite similar among several market economies (the U.S.A., Germany, France, Italy, the U.K.) And four planned economies (the U.S.S.R., Hungary, Poland, and Yugoslavia). Japan and Spain displayed higher weekly hours than the U.S.A. (by 15 percent for Japan and 19 percent for Spain). Adjusting the labor input for indexes of educational quality and average weekly hours made a difference of four or five percent, with the exceptions of Japan and Spain. See Bergson's Appendix Table 7, last column.

8. The likelihood ratio test statistic for the restrictions [a.sub.D] = [a.sub.DE] = [a.sub.DK]: = 0 is 31.196. This statistic is distributed as [Mathematical Expression Omitted], and is significant at the 99% confidence level.

References

1. Arrow, Kenneth J. and Leonid Hurwicz. "Decentralization and Computation in Resource Allocation," in Essays in Economics and Econometrics, edited by R. W. Pfouts. Chapel Hill: University of North Carolina Press, 1960.

2. Aslund, Anders, editor. Economic Transformation in Russia. New York: St. Martin's Press, 1994.

3. Battese, George E. and Tim J. Coelli, "Prediction of Firm-Level Technical Efficiencies With a Generalized Frontier Production Function and Panel Data." Journal of Econometrics, 38, 1988, 387-99.

4. Bergson, Abram, "Comparative Productivity: The USSR, Eastern Europe, and the West." American Economic Review, June 1987, 342-57.

5. -----. Planning and Performance in Socialist Economies. Boston: Unwin Hyman, 1989.

6. -----, "The USSR before the Fall: How Poor and Why." Journal of Economic Perspectives, Fall 1991, 29-44.

7. Caves, Richard E. and David R. Barton. Efficiency in U.S. Manufacturing Industries Cambridge, Massachusetts: The MIT Press, 1990.

8. Coelli, Tim J., "A Computer Program for Frontier Production Function Estimation." Economics Letters, 39:1, 1992, 29-32.

9. Dornbusch, Rudiger, and H. Wolf, "Economic Transition in Eastern Germany." Brookings Papers on Economic Activity, 1992 No. 1, 235-72.

10. Forsund, Finn R., C. A. K. Lovell, and Peter Schmidt, "A Survey of Frontier Production Functions and Their Relationship to Efficiency Measurement." Journal of Econometrics May 1980, 5-25.

11. Hayek, Friedrich A., editor. Collectivist Economic Planning. London: Routledge, 1935.

12. -----. The Road to Serfdom. Chicago: The University of Chicago Press, 1944.

13. Heal, Geoffrey M., "Planning Without Prices." Review of Economic Studies, July 1969, 347-62.

14. -----. The Theory of Economic Planning. Amsterdam: North-Holland, 1973.

15. Hitchens, D. M. W. N., K. Wagner, and J. E. Birnie. East German Productivity and the Transition to the Market Economy. Brookfield, Vermont: Ashgate Publishing Company, 1993.

16. Keynes, John Maynard. The General Theory of Employment, Interest and Money. London: Macmillan, 1936.

17. Kravis, Irving B., "Comparative Studies of National Income and Prices." Journal of Economic Literature March 1984, 1-39.

18. -----, Alan W. Heston and Robert Summers. World Product and Income: International Comparisons of Real Gross National Product. Baltimore: Johns Hopkins University Press, 1982.

19. Lange, Oskar, "On The Economic Theory of Socialism." Review of Economic Studies October, February 19361937, 53-71, 123-42.

20. ----- and Fred M. Taylor. On the Economic Theory of Socialism. Minneapolis, Minnesota: University of Minnesota, 1938.

21. Learner, Edward E. Sources of International Comparative Advantage. Cambridge, Massachusetts: MIT Press, 1984.

22. Lerner, Abba P. The Economics of Control. New York: Macmillan, 1944.

23. L. W. International Financial Research. Occasional Paper Number 124, "Economic Growth in Eastern Europe, 1985-1992." New York, 1993.

24. Malinvaud, E. "Decentralized Procedures for Planning," in Activity Analysis in the Theory of Growth and Planning, edited by E. Malinvaud and M. Bacharach. London: Macmillan and International Economic Association, 1967.

25. Marer, Paul. "Economic Performance and Prospects in Eastern Europe: Analytical Summary and Interpretation of Results." Papers submitted to the Joint Economic Committee, Congress of the United States. Washington, D.C.: U.S. Government Printing Office, 1981, pp. 19-94.

26. -----. Dollar GNPs of the U.S.S.R. and Eastern Europe. Baltimore: Johns Hopkins University Press, 1985.

27. Mises, Ludwig von. Socialism. New Haven: Yale University Press, 1951.

28. Moroney, J. R., "Energy Consumption, Capital, and Real Output: A Comparison of Market and Planned Economies." Journal of Comparative Economics, June 1990, 199-220.

29. Moroney, John R., "Energy, Capital, and Technological Change in the United States," Resources and Energy 14, 1992, 363-80.

30. Murphy, Kevin M., Andrei Schleifer, and Robert W. Vishny, "The Transition to a Market Economy: Pitfalls of Partial Reform." Quarterly Journal of Economics, August 1992, 889-906.

31. Schmidt, Peter and C. A. K. Lovell, "Estimating Technical and Allocative Inefficiency Relative to Stochastic Production and Cost Frontiers." Journal of Econometrics February 1979, 343-66.

32. Schroeder, Gertrude E. and Imogene Edwards. Consumption in the USSR: An International Comparison. Joint Economic Committee, United States Congress, Washington, D.C.: Government Printing Office, 1981.

33. Schleifer, Andrei, and Robert W. Vishny, "Improved International Comparisons of Real Product and its Composition." Review of Income and Wealth, June 1991, 207-62.

34. Summers, Robert and Alan W. Heston, "Improved International Comparisons of Real Product and Its Composition." Review of Income and Wealth June 1984, 207-62.

35. Yearbook of Worm Energy Statistics. New York: United Nations, 1983.

36. Zellner, Arnold, J. Kmenta, and J. Dreze, "Specification and Estimation of Cobb-Douglas Production Function Models." Econometrica, October 1966, 784-95.