COMBINING INPUTS TO SECURE A MEASURE OF TOTAL FACTOR INPUT.

An underlying theme in this paper is that the differences in approach in this area arise partly from the complexity of the phenomena in the real world being studied, the implications of this interrelated complexity for unbiassed and efficient estimates of the structural relations, and the problems of getting an adequate number of observations of the required form. Two distinct approaches have been used in the study of marginal and total factor productivity. One approach is to use the factor shares in national income as weights to combine the individual factor inputs to make an index of total factor input, and to use a factor's relative share to measure its marginal contribution. The second approach is to estimate the production relation from the data being used, and derive the marginal contribution of the productive factors to output from the estimated relation. The longest parts of the paper review the procedures followed to cope with the main problems in the real world, and the strengths and limitations in the two approaches. The discussion emphasizes the issues for the economy as a whole, and touches only briefly on the issues in disaggregation. Three major themes are emphasized in the conclusions to the paper. One is that many of the problems, the differences in view, and the controversies grow out of the range of interrelated issues in practical applications. A second major theme is that most of the attempts to solve particular issues by those using the factor shares approach are rather similar to those followed by researchers estimating the production relations directly. A third theme is to encourage more studies that will look at the interconnections between production relations and income distribution, from the points of view of both economic theory (and its predictions about the relevance to concrete applications) and statistical estimation. [ABSTRACT FROM AUTHOR]