Econometric forecasting: A brief survey of current and future techniques

Amongst the techniques discussed are (a) for univariate series, generalisations of Box-Jenkins auto-regressive-moving average models including intervention analysis, non-linear models, and models with time varying parameters; (b) for single output, multiple input situations, causality testing, and diagnostic testing of alternative specifications using Lagrange multiplier tests; and (c) for multivariate cases, vector autoregressive models, order-selection criteria, Bayesian priors, and factor models. Co-integration and error-correction models are also introduced.Some of these techniques are illustrated using a forecast comparison exercise concerning forecasts of monthly electricity demand per customer.