Thomas H. Naylor

Computer Simulation Experiments with Economic Systems: The Problem of Experimental Design

Abstract Experimental design considerations have been virtually ignored by economists who have conducted computer simulation experiments with models of economic systems. The objective of this paper is to spell out in detail the relationship between existing experimental design techniques and techniques of data analysis and the design of simulation experiments with economic systems. We begin by defining the problem of experimental design as applied to computer simulation experiments. With the aid of an example model, we explore several techniques of data analysis and a number of specific experimental design problems. Although this paper is oriented towards the design of computer simulation experiments in economics, the techniques which are discussed are of a general nature and should be applicable to the design of simulation experiments in other disciplines.

SPECTRAL ANALYSIS OF DATA GENERATED BY SIMULATION EXPERIMENTS WITH ECONOMETRIC MODELS

This paper is concerned with the use of spectral analysis to analyze data generated by computer simulation experiments with models of economic systems. An example model serves to illustrate two different applications of spectral analysis. First, spectral analysis is used to construct confidence bands and to test hypotheses for the purpose of comparing the results of the use of two or more alternative economic policies. Second, spectral analysis is employed as a technique for validating an econometric model. 1. INTRODUCTION DURING THE PAST decade computer simulation experiments with econometric models have become a commonly employed tool for analyzing the behavior of complex economic systems. While economists have improved the estimation process and have considerably enhanced the descriptive power of their econometric models, there have been fewer imposing gains made in the statistical analysis of the resulting output. The major impetus behind the use of simulation by econometricians and economic policy makers is the possibility (and opportunity) of validating econometric models and testing and evaluating alternative economic policies before they are put into effect on actual economic systems. Complete exploitation of simulation experiments with econometric models implies a thorough analysis of the data so generated. Yet as Burdick and Naylor [5, 31, 32] have pointed out in recent articles, a preoccupation with model building among many econometricians simulating economic systems has unduly diverted attention from experimental

Design of computer simulation experiments for industrial systems

Computer simulation experiments design for industrial systems, considering variance techniques, multiple ranking procedures, sequential sampling and spectral analysis

Methods for analyzing data from computer simulation experiments

This paper addresses itself to the problem of analyzing data generated by computer simulations of economic systems. We first turn to a hypothetical firm, whose operation is represented by a single-channel, multistation queueing model. The firm seeks to maximize total expected profit for the coming period by selecting one of five operating plans, where each plan incorporates a certain marketing strategy, an allocation of productive inputs, and a total cost. The results of the simulated activity under each plan are subjected to an F-test, two multiple comparison methods, and a multiple ranking method. We illustrate, compare, and evaluate these techniques. The paper adopts the position that the particular technique of analysis (possibly not any one of the above) chosen by the experimenter should be an expression of his experimental objective: The F-test tests the homogeneity of the plans; multiple comparison methods quantify their differences; and multiple ranking methods directly identify the one best plan or best plans.

Experience with corporate simulation models—A survey

Abstract In this article the authors argue that if corporate simulation models are going to help management meet their objectives then changes are necessary. These changes require that (i) the models should be more user oriented, (ii) there should be more emphasis on production modelling, (iii) there will be an increased use of optimization techniques linked to corporate models, (iv) there will be an increased effort to integrate finance, marketing and production, (v) the more sophisticated application will be concerned with the external environment and (vi) model builders should become more aware of the importance of corporate politics.

A Computer Simulation Model of the Textile Industry

Abstract A nine-equation econometric model of the U. S. textile industry is presented to explain the behavior of the industry during the period 1953 through 1962. The endogenous variables of the model include apparel output and demand, textile mill products output and demand, employment, wages, profit, and investment. A computer program was written to simulate the behavior of the textile industry between 1953 and 1962 on the basis of the behavioral relationships implied by the nine-equation model. Three different techniques were used to compare the simulated data generated by the model of the textile industry with actual observed data—graphical analysis, spectral analysis, and total variance analysis.

The Design of Computer Based Planning and Modeling Systems.

Abstract Today there are nearly 2000 corporations in North America and Europe either using, developing, or experimenting with some form of corporate planning model. With the emergence of this new and rather substantial interest in the methodology of corporate planning modeling, there appears to be a definite need for a conceptual framework which can be used to design and implement computer based planning and modeling systems. In this paper the authors describe a collection of elements which they believe to be of critical importance in designing a corporate planning model. Their objective is to develop a set of criteria for not only designing a planning and modeling system, but a set of criteria which can also be used to facilitate the evaluation and comparison of alternative planning and modeling systems. There are over 50 planning and modeling software packages on the market today. These include systems such as BUDPLAN, COMOS and SIMPLAN. This paper attempts to provide the reader with a convenient checklist of possible features to consider in either designing ones own system or selecting an appropriate software package.

The application of sequential sampling to simulation: an example inventory model

Four different sequential sampling procedures are applied to the analysis of data generated by a computer simulation experiment with a multi-item inventory model. For each procedure the cost of computer time required to achieve given levels of statistical precision is calculated. Also the cost of computer time using comparable fixed sample size methods is calculated. The computer costs of fixed sample size procedures versus sequential sampling procedures are compared.

A POLICY MODEL OF THE UNITED STATES MONETARY SECTOR

Monetary policy actions in the United States are presumed to influence a set of ultimate target variables, such as real income, price levels, and the balance of international payments, through their effects on a set of proximate and intermediate targets, such as bank reserves and deposits, business and consumer credit, and security prices and yields. The Federal Reserve System conducts daily operations with these variables, involving billions of dollars in financial assets and affecting the economic welfare of every American. Therefore, an understanding of the nature of the medium through which these operations are conducted is a high priority concern of monetary economists.

An Econometric Model of the Textile Industry in the United States

T HE model presented in this paper is a system of recursive linear regression equations, the parameters of which are estimated from monthly series of data covering the period, January, 1951, through December, 1962. The parameters relate certain exogenous and predetermined endogenous variables to apparel demand, apparel output, textile demand, textile output, and employment, earnings, prices, profit, and investment for the textile industry. Elsewhere [9] the authors use this model to simulate the aggregate behavior of the industry for the 1951-1962 period and for 1963 through 1964. There are a number of problems and characteristics which are unique to textile manufacturing in the United States. They are frequently brought to the attention of the public as a result of the industrys extreme sensitivity to such things as changes in tariffs, changes in wage levels, and changes in government price policies on cotton. Imports of textiles, for example, have been troublesome to American textile producers due to low wages abroad and the development of a large and efficient productive capacity in many foreign countries. This situation has been the subject of considerable controversy in recent months due to the expiration of the Long Term Arrangement on trade in cotton textiles [7, 21]. Related to the import problem is the problem of cotton prices. The introduction of one-price cotton in April, 1963, ended a dual-price system, in effect since 1956, which had caused producers of the United States to pay higher prices for inputs than foreign producers. In addition to these economic variables which are subject to control largely by public policy or by bargaining, there are other characteristics of the industry which have contributed to its historical instability. Evidence of this has been the two-year textile cycle which persisted in the output of textile mill products, even in spite of a relatively stable end-use demand.2 Several previous studies have noted the damaging effect of this cyclical pattern upon the industry [1, 6, 24, 25]. Stanbacks analysis of the textile cycle [14, 15] provided reasons for this cyclical pattern, and also provided measures of the cycle for the period 1919 through 1956. In developing an econometric model of the textile industry it was necessary to determine whether the cyclical behavior, which influenced this industry in earlier years, still prevailed. Preliminary investigations led the authors to the conclusion that there had been basic changes in the nature of the industry, such as increasing concentration, diversification of inputs, wider ownership, and more scientific management which had led to greater stability.3 These are among the features of the United