Comparison of system dynamics calibration and econometric estimation

System dynamics and econometrics have traditionally been seen as contrasting and sometimes conflicting approaches for deriving parameters for economic models. System dynamics as a simulation method uses repeated runs to achieve calibration through optimisation, whereas simulation as an estimation procedure is much less common in econometrics . A comparison of system dynamics calibration with econometric estimation is undertaken applied to technology adoption data. Values of the same order of magnitude were achieved for parameters for the same model, with numerical differences and because models are not well identified. However, different model specifications using the same method, which represent various behavioural factors, exhibit wider differences. The comparison also revealed some additional process factors, such as whether to select population (stock) or derived sales (rate) variables to calibrate against data. System dynamics takes an operational approach to economic modelling, which may lead to overspecified models from an econometrics perspective, where the typical criterion is parsimony. However, there is recognition in econometrics teaching and practice that models need to have a more behavioural basis, to be functionally richer to address the ‘non-stationarity revolution’ and to permit better identification and selection of causal factors.