During the 1960s most of the countries of Eastern Europe experienced a visible retardation of economic growth. The investigation of this phenomenon lead many Eastern as well as Western economists to the conclusion that the retardation was caused primarily by declining rates of growth of the total factor productivity. Originally, the factor productivity was measured by traditional index methods, however, more recently several attempts were made to estimate the rate of growth of the total factor productivity from macroeconomic production functions (see Tlusty-Strnad , Hajek-Toms , Rychetnik , BrownNeuberger-Licari , Desai , Weitzman , Toda , Hocke-Kyn-Wagener . Incidentally the production function approach lead some authors (Weitzman, Desai, Brown-Neuberger--Licari) to the suggestions that the low elasticity of substitution may be an alternative explanation of retardation.
Most of the work on the estimation of production functions was done for the Soviet Union. Those few studies which concentrated on the other East European countries came frequently to statistically insignificant results. The reasons are well known: small number of observations and an extremely high degree of multicollinearity. In one of our previous papers (Kyn-Kyn ) we tried to demonstrate, that this obstacles can be overcome by estimating the production functions from the pooled cross-section and time series data. By doing so we received economically reasonable and statistically significant estimates of the capital and labor elasticities, and of the rate of technical change for Poland, Czechoslovakia, Hungary, Bulgaria and Rumania. Our estimates did show a clear declining trend in the rate of technical change at least for some of the mentioned countries, but it did not unambiguously support the hypothesis of low elasticity of substitution.
The estimation from pooled cross-section and time series data has two obvious advantages: the number of observations is considerably increased and the multicollinearity is for all practical purposes avoided. However, one new problem arises: the traditional assumptions of homoscedastic and serially uncorrelated random disturbance cannot be maintained. In the above mentioned paper we did not pay any attention to this problem and applied the ordinary least squares method. In this paper we present the estimated macroeconomic production functions from the pooled data using the stepwise procedure in which the autocorrelation and heteroscedasticity is eliminated by transformation of variables.
Our intention to use the “combination approach” was facilitated by the fact that the official statistical sources in the East European countries provide almost all needed data in the similar classification into sectors and industrial branches1. Our basic observations are for each country cross-sectionally divided into 15—16 industrial branches and are available for approximately 20 post-war years as shown in Table 1.
The observations are:
The subscripts i and t refer to branch and year respectively.
All the data are related only to the “socialist” or “state” sector and cover the industrial production in the East European definition, i. e. including mining but excluding the building industry. Naturally, agriculture, transportation, trade and services are also excluded.
The data on gross value of industrial output and on labor could have been taken directly from the official East European statistical sources without great problems. Only the standard adjustments for organizational and price changes and estimation of few missing observations was needed. Much greater problem was with capital stock data, which in some cases had to be reconstructed from gross investment, without knowing the