Page 21 -

14 K.G. Troitzsch

two decades of computational social science saw mathematical models and their
inelegant solutions, microsimulation and the first agent-based models before the
name of this approach was coined.
Among the first problems tackled with the help of computer simulation, there
were already predictions of the future of companies (“industrial dynamics”,
Forrester 1961), cities (“urban dynamics”, Forrester 1969) and the world as a
whole (“world dynamics”, Forrester 1971) in the early 1960s and 1970s by Jay W.
Forrester as well as predictions of the consequences of tax and transfer laws for both
the individual household and the national economy in microanalytical simulation,
an attempt that started as early as in 1956 (Orcutt 1957). Other early attempts at
the prediction of election and referendum campaigns also became known in the
1960s, such as Abelson’s and Bernstein’s simulation analysis of a fluoridation
referendum campaign of the Simulmatics Project directed by de Sola Pool. What all
these early simulations have in common is that they were aimed at predicting social
and economic processes in a quantitative manner and that computer simulation was
seen as a “substitute for mathematical derivations” (Coleman 1964, p. 528), and
although Simon and others had already taught computers to deal with nonnumerical
problems as early as in 1955 (“the Logic Theorist, the first computer program that
solved non-numerical problems by selective search”, Simon 1996, pp. 189–190),
Coleman still believed in 1964 that “the computer cannot solve problems in algebra;
it can only carry out computations when actual numbers are fed in” (Coleman 1964,
p. 529).
As system dynamics and microanalytic simulation—simulation approaches that
continue to be promoted by learned societies such as the System Dynamics Society,
which celebrated its 50th anniversary with an international conference in Boston in
July 2007, or the International Microsimulation Association, which also celebrated
50 years of microsimulation with an international conference held in Vienna in
August 2007—are not the focus of this handbook, this chapter will only give a short
overview of these two approaches and go into the details of some other early models
that remained more or less isolated and were even more or less forgotten.
System dynamics was developed by Jay W. Forrester in the mid-1950s as a tool to
describe systems which could have modelled with large systems of difference and
differential equations containing functions whose mathematical treatment would
have been difficult or impossible. The general idea behind system dynamics was
and is that a system, without considering its components individually, could be
described in terms of its aggregate variables and their changes over time. The best
known examples of system dynamic models are Forrester’s (1971) and Meadows
et al.’s (1974) world models which were inspired by the Club of Rome and won
public attention in the 1970s when they tried to forecast the world population, the
natural resources, the industrial and agricultural capital and the pollution until the
end of the twenty-first century by describing the annual change of these aggregate
variables as functions their current states and numerous parameters which had some
empirical background.
Microsimulation was first described in papers by Orcutt (1957) who designed a
simulation starting with a (sample of a) given population and simulating the indi-