Clear long-range whether forecast

The pace of development in econometrics over the past ten years has led to textbooks on the subject becoming almost unmanageably long as they attempt to keep students - and their lecturers - up to date.

At the same time, there has been an increase in the number and size of related journals, which means that professional economists seeking to use techniques outside their immediate specialisation have access to something of an overload of information. But since the teaching term has not been extended and there are few opportunities to expand the topic list taught, the pressure is on lecturers to select themes that define their course.

To have authoritative summaries in well-defined areas would therefore be very useful.

A Companion to Economic Forecasting is part of the series Companions to Contemporary Economics, which aims to provide accessible introductions to key topics in economics and econometrics for both students and professional economists.

In fact, the book provides the equivalent of an Ordnance Survey map as specialists guide the way into central topics where either the immediate goal will be achieved or clear pointers will be found to a more dedicated search.

The book's editors have made significant contributions in economic forecasting. Their opening chapter - not to be ignored in the rush to get to later sections - is a useful summary of essential concepts: how economists forecast; how to measure success and failure in forecasting; the main problems; its future.

It addresses the debate over the validity of macro-economic and financial forecasting, and is a sound starting point towards understanding the nature of the uncertainty faced by a forecaster.

Uncertainty relates to the accuracy of a forecast, although one has first to decide the definition of an "accurate" forecast. To paraphrase the editors, a key distinction is between "we know what we don't know" and "we don't know what we don't know". Examples of the former are uncertainties arising from the need to estimate the parameters of a model. Examples of

the latter are structural changes in economies such as those in the aircraft manufacturing and passenger airline industry following the events of September 11 2001.

Other topics covered in the book's 23 chapters include the nature and sources of uncertainty; density forecasting (of particular relevance given the Bank of England's well-known fan charts for inflation forecasts); different forms of models for forecasting (such as structural time-series, co-integrating vector-autoregressive, and non-linear); forecasting and economic policy; the methods and practice of evaluating forecasts; the combination of several forecasts; forecasting competitions; the use of leading indicators; periodic effects in forecasting; and forecasting financial variables.

The text contains many practical examples, including forecasting national income, inflation, leading indicators, non-durable consumption, auto-sales, unemployment and exchange rates.

The contributions are generally concerned with model-based forecasting but an interesting chapter by Dilek Önkal-Atay, Mary Thomson and Andrew Pollock summarises research on judgemental forecasting. Model-based forecasting also uses judgement, for example intercept adjustments to estimated equations to capture specification errors and structural change, but much forecasting practice in business is not based on formal statistical or econometric models. The prevalence of judgemental forecasting raises some challenging and only partially resolved questions about if, when and how judgemental rather than formal models should be used.

Given that the concept of rationality underpins much recent economic theory, the chapter by Herman Stekler on the rationality and efficiency of individuals' forecasts is of interest. Relevant questions are whether individuals make systematic errors (weak rationality) and whether they make efficient use of information available at the time of making their forecasts (efficiency). Based on studies of forecasts of US inflation and growth, and some from the UK, the jury seems still to be out on weak rationality, while there is evidence from several studies against efficiency. At the least this suggests a need to re-evaluate the dominance of the stronger form of rational expectation in economic models.

Overall, the Companion achieves its aim admirably. The essays form a coherent whole, and are written at a similar and accessible level. They are pertinent to the current state of development of economic forecasting.

To some economists the practice of forecasting is akin to alchemy but this book should do something to persuade even sceptics that the construction and analysis of forecasts is a matter for serious study.

Kerry Patterson is professor of econometrics, University of Reading.