The case for coordinating policies is usually stated in terms of efficiency gains: it is possible to design a set of coordinated policies which makes at least one country better off (in terms of its own policy preferences), without any being made worse off, than the best outcome possible without coordination. Coordination, because it takes into account economic interdependence, is beneficial.

Calculations of such efficiency gains usually assume that coordinated policies are designed using correctly specified models of the economies concerned and/or perfect foresight regarding the future movements of exogenous variables. This is seldom a realistic assumption, and econometric policy projections are often criticized for their sensitivity to specification errors and to 'information errors'. These are errors made by the policy-maker when forecasting the future path of variables outside his control, such as oil prices, but which directly affect his own policy targets or the behaviour of other interdependent economies. In Discussion Paper No. 151, Research Fellow Andrew Hughes Hallett argues that, ironically, the principal attraction of coordinated economic policies may be their robustness to 'information errors'. This is an important practical consideration, according to Hughes Hallett, if the theoretical analysis of policy coordination is to influence policy formulation. Policy-makers are loath to commit themselves to a sequence of actions which may later need substantial or frequent revisions. Their scepticism is heightened by some empirical studies, such as that by Oudiz and Sachs, which suggest that policy coordination between industrial nations is likely to yield only small efficiency gains.

Hughes Hallett notes that it is difficult to establish whether cooperative or non-cooperative policies are more robust to information errors. There are, however, reasons for thinking that cooperation increases robustness. Most performance indices are constructed from combinations of the policy targets. Consequently, if a given shock or error produces a positive disturbance in one target and a negative disturbance in another target and/or the same target at a later date, then the performance index will be disturbed by less than either of the target variables. Cooperative policies are designed by optimizing a performance index in which every country's domestic objectives are combined. This combined index is necessarily more robust to unforeseen disturbances than any of its individual components, because cooperation pools the risks between countries. This suggests that cooperative solutions will be more robust to information errors than non-cooperative solutions.

This theoretical conclusion is confirmed by Hughes Hallett's empirical research, at least as long as policy-makers exercise their freedom to revise their decisions over time. He first simulates a modified version of the European Commission's 'Comet' model over the period 1974-8. The greater robustness of cooperative over non-cooperative policies is found to be stronger for the US economy than for the EEC and, more significantly, it increases with the degree of uncertainty. That result is repeated in a second exercise using the EEC's 'Compact' model and involving explicit cooperation between the United States, the EEC and Japan in the period 1977-81. The greater robustness of cooperative policy-making does not appear to be model-dependent, since the 'Comet' and 'Compact' models incorporate different assumptions, for example regarding wealth effects and market- clearing. There is, he concludes, a strong practical argument for cooperation as a way of overcoming some of the uncertainties inherent in econometric policy analysis.