We now establish a framework for analysing some of the problems that arise in using large-scale macroeconometric models for forecasting. A number of facets of the real-world forecasting venture complicate the picture compared to the analyses of chapters 4 and 6, and each may contribute to forecast errors. A taxonomy of sources of forecast error is developed to enable a systematic treatment of these issues, along with the sources of error covered in earlier chapters. The taxonomy comprises five major categories, namely parameter non-constancy, model mis-specification, sampling variability, variable mis-measurement and error uncertainty, and suggests partial remedies. The potential impact of forecasts, or expectations, on outcomes (the ‘bandwagon effect’ of section 1.3) arises naturally in the taxonomy as a form of variable uncertainty. We also discuss some of the techniques commonly used to evaluate large-scale macroeconometric models.