An Economic Analysis of Numerical Data Processing in the Firm

This paper provides a coherent framework within which to understand the economics of data processing in enterprises. The computational processes in the firm are described with the help of a dynamic parallel processing model of associative computation. The basic model is extended to include the assumption that the speed of information processing in each individual computational center depends upon the capital and labor allocated to it. In such a model, the conditions for the efficient organization of numerical data processing are defined, and the architecture of efficient structures is analyzed. It is shown that, similarly as in the computer systems, the so-called "skip-level reporting" structures are efficient for decentralized numerical computation in business firms. However, if processing elements in the skip-level reporting structures cannot be equally loaded, then their computational power has to be adjusted to the given information workload. The method of adjustment of the resources allocated to processing elements to the information workload of the one-shot skip-level reporting structures of information processing is presented, and the efficiency frontier is characterized. Furthermore, the optimal organization of numerical data processing in enterprises is analyzed on the example of predicting demand in the firm