^{note 1 in page 60} It is assumed that forecasters within the policy-making govern ment departments do not have access to much useful information which is not available to outside forecasters such as those at the Institute.

^{note 2 in page 60} That rule will be most helpful if it refers to the movement of direct policy instruments such as tax rates, not to intermediate targets such as the PSBR or the money supply.

^{note 1 in page 61} The standard error is rather larger in the equation using revised data for the CPI and the suggestion of upward bias a little more pronounced.

^{note 2 in page 61} The idea derives from Brainard, W. C., ‘Uncertainty and the effectiveness of policy’, American Economic Review, May 1967. One needs to distinguish, however, between the case where the effect of policy is constant but not known with precision and the case where the effect of policy is believed to vary over time. Our discussion in this note is concerned only with the former.

^{note 1 in page 62} These regressions are not therefore to be interpreted as a serious attempt to estimate ‘reaction functions’. In the present context we are interested in regressing P only on such variables as appear in equation (1) (or might appear in a generalisation of it).

^{note 2 in page 62} An additional arguement leading to a broadly similar conclusion would be in line with the familiar ‘Lucas critique’ of modelbuilding. If a policy rule is changed then the way private agents form expectations will change and, in consequence, the behaviour of the economy will then change (to an unknown ex tent). Such changes in behaviour can only increase the uncertainty of all forecasts.