THE COMBINATION OF TIME SERIES AND CROSSSECTION DATA IN INTERINDUSTRY FLOW ANALYSIS
Abstract
A problem which arose in the course of a larger study to explain the variations of input-output ratios over time is considered. (By the input-output ratio of industry 'i' to industry 'j' is meant the ratio of that part of the output of industry 'i' used by industry 'j' to the output of industry 'j'. For such a study, there are two types of data available. For all years, there are available (ideally) outputs and final demands (the final demand for an industry consists of all uses of its product other than in other industries or itself) for all industries. The 'balance equations' of input-output analysis form an (incomplete) system of simultaneous relations which may be estimated by some version of the method of maximum likelihood (for computational reasons, the single-equation limited-information method is the only one likely to be used). However, for some years, additional information is gained in the form of knowing the actual interindustry flows, which clearly should substantially increase the accuracy of the estimates. The simplest technique is, of course, to assume that the 'true' inputoutput ratio for any year for which flow data are available is exactly equal to the observed input-output ratio for that year. The assumption behind this is, however, contradictory to the basic postulate that all the relations involved are valid only up to a stochastic term. It is therefore of interest to consider more explicitly the interindustry flow model implied in the use of both time series and interindustry flow data for estimating the parameters involved.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 18, 1956
- Accession Number
- AD0605093
Entities
People
- Kenneth J. Arrow
Organizations
- RAND Corporation