FORECASTING INNOVATION DIFFUSION WITH NEAR-OPTIMAL BERTALANFFY-PÜTTER MODELS

Using a classical example for technology diffusion, the mechanization of agriculture in Spain since 1951, we considered the forecasting-intervals from the near-optimal Bertalanffy-Pütter (BP) models. We used BPmodels, as they considerably reduced the hitherto best fit (sum of squared errors) reported in literature. And we considered near-optimal models (their sum of squared errors is almost best), as they allowed to quantify model-uncertainty. This approach supplemented traditional sensitivity analyses (variation of model parameters), as for the present models and data even slight changes in the best-fit parameters resulted in very poorly fitting model curves.