Simulation with large econometric models: The quest for a solution

Abstract

This paper describes the development of algorithms to solve large-scale macroeconometric models. The bulk of this paper was written in 1970 when the problem of finding solutions for these large-scale models was still an issue. During the 1960s, large powerful (for that time period) computers became available for use by economists who proceeded to use this powerful new tool to develop and apply econometric software to the problem of estimating large-scale macroeconometric models such as the Wharton Quarterly Model and the Brookings Model. Having succeeded in estimating these models, the next problem was to find a way to solve them either in the context of simulation experiments or forecasting exercises. During the late 1960s, the author investigated the problem of solving these large-scale models and found several acceptable methods which are presented in this paper.

Econometric Software

“…On the one hand, at the end of the 1960s, the first econometric modeling languages [203] began to be written, in order to support the creation, maintenance, and use of large macroeconometric models. The creation of multi-hundred equation models, such as the Brookings and Wharton models in the USA, the Candide model in Canada, the Cambridge growth model in the UK, and others elsewhere [31], implied the need to manage time-series databases containing thousands of variables, to create model solution algorithms [197,216], and to develop graphical, tabular, and other easily understood display facilities. An economic consulting and forecasting industry, including such firms as Chase Econometric Associates, Data Resources, and Wharton Econometric Forecasting Associates, and an emerging time-sharing telecommunications technology, both fostered and financed the creation of quasi-natural-language software packages offering substantial database maintenance facilities, such as Damsel, EPS, MODLER, and XSIM [206].…”

“…In the first case, to solve a model usually requires it to be ordered at least quasi-recursively, a somewhat heuristic process considered typically in the specialist literature [32,79,125,126,163,184]. When using Newton techniques, the model need not be reordered but may have to be linearized, at least in part, in order to obtain the partial derivatives needed to compute the Jacobian used in the model solution, although later versions of TROLL and TSP provide nonlinear differentiation procedures that to a degree obviate this need [55,216]. Optimized solutions, particularly in the context of rational expectations or model-consistent solutions, add a further level of complexity to the solution process [75,76,78].…”

Econometric software: The first fifty years in perspective

“…A similar web of contacts existed in the UK and between Canadian and US econometrician-developers. These linkages obviously make it difficult to identify the work occurring at any one place as logically preceding that at any other: early individual contributions were made, such as those by Mark Eisner [86,88], Robert Hall [27], David Hendry [146], Michael McCracken [213], Morris Norman, Ross Preston, George Schink [92,299], Lucy Slater [306-308, 314, 315], and Arnold Zellner [351,352], but the lack of a developed historical literature makes it difficult to establish strict precedence in the context of an organized and evolving body of knowledge.…”

“…[163]. But slowly during this period the method of Gauss-Seidel became generally adopted, based upon recognition by Kuh of the applicability of the method [29, p. 516], first for the Wharton Model [92,117,299], leading to the creation of relatively robust model solution programs by the early 1970s. 14 The concept of an online data base as a central element was transmitted to Lexington, Massachusetts in 1968-69, taking root with the creation of the aptly named Data Resources, Inc.…”

Introduction

“…The Monroematic calculator is historically interesting, in part because not only was it used by economists during the early 1960s and before, in much the way that Desai describes, but also because of its sustained use by some economists, such as Michael Evans, even as late as 1967-1968 to solve even relatively large econometric models, like the Wharton Quarterly Econometric Model of the United States. At about that time, the electronic computer finally came into its own for this purpose, albeit only comparatively speaking in these years [18,20]; more years would pass before it actually became easy to solve such a model. Desai's paper is interesting for its depiction of a transitional time during which both desktop calculators and the electronic computer were used, much as horses and internal combustion vehicles shared the roads even into the 1930s.…”