Date of Award
Program or Major
Doctor of Philosophy
Marc W Herold
The study begins with explanation of I-O tables, intertemporal stability of its coefficients, and logic of updating techniques. Following a literature review, nine non-survey updating methods are selected and utilized to update the actual 1966 table of the Soviet Union to the target year of 1972. Next, the data and simulation procedure are specified and justified. The concept of matrix comparison along with methods to accomplish this task are discussed. Then, the resultant updated matrices are compared with the actual data, via employment of 25 criterions. Accordingly, RAS and Friedlander procedures are ranked as top performers. The results, while reasonable in holistic sense, are not impressive partitively.
Exogenous estimation of a subset of coefficients is considered next. Several possibilities for "selective targeting" are investigated, and three such criterions, namely "key," "most important," and "largest" coefficients are adopted. These criterions, then are used to modify RAS, Friedlander, and NAIVE methods via incorporation of exogenous data. An additional approach, Residual Minimum method, is also employed. Thus, ten "modified" estimates are obtained and compared with the actual table. The outcome indicates substantial improvements in the RAS and Friedlander updates, particularly when exogenously estimated subset consists of the largest, or the "most important" coefficients. NAIVE method displays reasonable improvement. Inclusion of prior information, however, in some instances leads to deterioration of individual estimates for the remaining coefficients.
Finally, sectoral aggregation and its effects on the performances of updating methods are addressed. Aggregated estimates at three levels of sectoral details, i.e., 36, 16, and 6 sectors, are obtained and comparisons are performed. The results indicate that, generally the performances of the updating methods, as well as the intertemporal stability of coefficients, are direct functions of the level of sectoral detail. No change in ranking occurs due to aggregation,
Conclusions of this research may be used for selection of updating methods, as well as in construction phase of tables, to identify and focus on the most influential coefficients. Throughout, a rather detailed presentation of methods and statistical tools are offered. All experiments are conducted for both direct and inverse matrices. NAIVE, (constant coefficient), is added for comparative purposes.
Jalili, Ali Reza, "An inquiry into non-survey techniques for updating input-output coefficients: Comparative experiments with data from the Soviet Union" (1994). Doctoral Dissertations. 1817.