Published online by Cambridge University Press: 28 November 2008
The paper considers the size and the implications of the errors made by the British Treasury in forecasting the budget over the period 1951–84. On average the Treasury underestimated the public sector fiscal deficit by 0.5 per cent of GDP; this is about half the average amount by which Chancellors attempted to change aggregate demand by means of their budgets. The general pattern was for tax revenue to be under-predicted, but for public expenditure to be under-predicted by even more. The government's fiscal deficit and thus the reflationary effect of government policy was therefore greater than the Treasury intended.
A decomposition exercise carried out on the Treasury's own model for the fiscal year 1980/81 suggests that only about half of this error is due to errors in the specification of the model; the other half is due to errors in data estimation and in forecasting exogenous variables. Hence, the scope for improving forecasts by improvements to the model is limited. The paper considers three alternative methods by which forecasts could be improved: the application of realisation functions relating forecast to actual values; more frequent budgeting; and an attempt to derive more tax revenue from those categories (such as excise duties on alcohol, petrol and tobacco) whose yield is easiest to predict.