Traditional system dynamics models are subject to error in the estimat
ion of performance measures because of the method of calculation over
discrete time intervals. It is shown that such errors may be significa
ntly reduced by dividing the computation between two runs with differe
nt solution intervals and applying Richardson's extrapolation method.
Examples are given for a variety of different published models and per
formance measures, (C) 1998 John Wiley & Sons, Ltd.