Policy Design

Mathematical control theory studies the behaviour of dynamic systems and how inputs to the system should best be chosen over time to achieve desired outputs. Engineers use these mathematical results to control chemical processes in industrial plants. Output of the process can be kept at or near its desired level by adjusting the rates at which ingredients are introduced. Rockets can be guided to their destination using the same techniques; control theory suggests how deviations from the desired course can be corrected using the minimum amount of thrust (and fuel) from the rocket's engines.

Since the 1960s and early 1970s, economists have attempted to apply these control theory techniques to the design and conduct of economic policy. Policy makers would specify their objectives in terms of rates of growth, inflation and unemployment. Control theory would then suggest how the 'inputs' to the economic system -the policy instruments -should be chosen so as best to achieve the policy objectives. Until recently, these applications of control theory to economics have been based on the direct analogy between engineering systems and economic systems; this permitted the straightforward transfer of the results of engineering control theory to economics. This direct analogy, however, is ill-founded: in economic systems, agents respond to policy in an intelligent, non-mechanical manner. If control theory is to make a useful contribution to economic policy, it must be supplemented by a serious treatment of expectations and of strategic behaviour by economic agents.

Any realistic economic theory must assume that economic agents base their current behaviour at least in part on what they expect to happen in the future. If forward-looking behaviour and expectations are to be treated seriously, then policy itself must be viewed in a somewhat different manner. The effects of a given change in a policy instrument cannot be analysed in isolation -one must specify the policy rule from which the change derives, For example, the announcement of an unexpectedly large money supply increase may cause the exchange rate to depreciate if the markets expect the change to be permanent. Yet the same announcement will cause an appreciation if the authorities are seeking to target the money supply, and the markets expect the authorities to react to the money supply increase by raising interest rates. The emphasis then is not on a particular setting of a policy variable, but on the rule used by the government to adjust the variable in response to developments in the economy.

The assumption that the private sector forms its expectations in a 'consistent' or 'rational' manner provides the most straightforward way of allowing agents to respond to policy in a non-mechanical fashion. One interpretation of the rational expectations assumption is that agents construct their forecasts or expectations in a manner consistent with the true underlying structure of the economy. Another interpretation is that agents form the best possible forecasts given the structure of the model and the information available to them, and that they do not make systematic errors in forecasting.

How plausible is the assumption that expectations are 'rational', so interpreted? It might be justified on the grounds that agents have available to them published forecasts of reasonable quality, Agents may also learn adaptively cbout the economy, so that their forecasts or expectations gradually converge in a stable world to the consistent ones. There has been much debate over the plausibility of rational expectations, but there is no doubt that the assumption does provide a crucial test-bed for policy evaluation, For suppose a policy performed badly under the assumption that agents' expectations were rational, and in forming their expectations agents understood the structure of the economy and hence the consequences of the policy. Then even if agents were less well informed, the policy could only succeed if agents were to make mistakes in a particular and systematic fashion. This would clearly be an unsatisfactory and unreliable basis for economic policy.

The rational expectations assumption can be implemented in a variety of ways in economic models. Early attempts allowed wages and prices to be perfectly flexible so as to clear the markets for labour and goods. This rendered government stabilization superfluous. Private sector behaviour adjusted more rapidly than government policy, and this gave the private sector effective control over the economy. But more recent models emphasise the constraints and coordination problems within the private sector, arising particularly from contractual wage adjustment and incomplete information, These models give ample scope for government stabilization policies,

The design of policy is greatly complicated, however, by a theoretical issue of potentially great practical relevance: the problem of time inconsistency, There are some otherwise optimal policies which have the property that as time proceeds government has a short-run incentive to renege, on previous policy commitments and to make 'U-turns'. Because of this incentive to renege, the announcement of a time-inconsistent policy may not carry credibility, because the private sector will anticipate the subsequent abandonment of the policy. However determined the government may sound when it announces fixed monetary targets, unions may expect monetary policy eventually to accommodate money wage increases if unemployment rises far enough.

Recent literature may have overemphasized this problem, particularly if governments are not too

myopic. Reneging on optimal, time-inconsistent policies may lead to loss of credibility and force the government to adopt an Inferior policy, By eschewing economy, policy U-turns, a government may over time establish a 'reputation' which makes its policy announcements more credible. Such a reputation for commitment to policy announcements may well be desirable for policymakers, particularly in an environment where policy has to cope with continuing disturbances to the economy. These ideas are being formalised in the emerging literature on 'reputations' and on the design of 'threat strategies' to support a given policy rule.

Rational expectations introduces a crucial element of game theory into macroeconomic policy formulation. On choosing a particular policy or 'strategy', the government must take into account the effects of its choice on the expectations and the choices made by other players. This is particularly important in the international economy, where economic interdependence means that policy formulation must be seen in terms of a game with many players, the governments, and interacting strategies, My recent research, carried out jointly with Paul Levine, emphasises the dangers arising from policy optimization by single country which ignores the effects of this policy on other countries.

Prisoners' dilemma

This may usefully be compared to the' Prisoners Dilemma' in game theory- Suppose we have two prisoners, A and B. Their jailer tells prisoner A that he will receive a light, 5-year sentence if he confesses and B does not; prisoner B will then receive a heavy 25-year sentence. On the other hand. if prisoner B confesses and A does not, B gets off lightly with a 5 year sentence and prisoner A receives the heavy 25 year sentence. If neither prisoner confesses, they will both receive a moderate, 10-year sentence. If both confess, both will receive a 20-year sentence, What should prisoner A do? One thing is quite clear: A is better off confessing, no matter what B does. For suppose B confesses; if A confesses as well, his sentence is reduced from 25 to 20 years. Suppose B does not confess; A is still better off confessing, since this reduces his sentence from 10 to 5 years. The argument is exactly the same for B -confession is his best policy irrespective of whether or not A confesses. So acting individually, both will confess -but both are clearly better off if they can agree not to confess and then stick to the agreement. Yet each will always have an incentive to abandon cooperation. In this situation cooperation (neither confesses) leaves both prisoners better off, but whether cooperation is viable or sustainable depends on the social or institutional framework within which cooperation arises.

Advantageous for one country ... disastrous if implemented by all

In policy terms, countries have a strong incentive to engage in such non-cooperative behaviour, even though the outcome is worse for all countries than if all countries agreed on a cooperative strategy. In particular, our research has shown that there is a strong incentive for countries acting alone to rely. excessively on exchange rate appreciation to control domestic prices, while using expansionary fiscal policy to sustain output. While such policies may prove advantageous for the single country, they may well harm others and prove disastrous if implemented by all countries. This highlights the need for research on appropriate cooperative rules or mechanisms for international policy coordination. We also need better understanding of the incentives or penalties required to sustain such cooperative behaviour among governments.

The Prisoners' Dilemma is usually analysed as if the prisoners are only faced with a choice at one point in time. When applied to economic policy this is clearly unrealistic governments formulate policies repeatedly over time. Economists have therefore devoted more attention recently to the study of 'repeated' Prisoners' Dilemma games. This research suggests that when such games are repeated rather than one-off, the cooperative strategy will be more attractive to players, provided that they are sufficiently concerned about the future. This is because the cooperative strategy is likely to elicit a cooperative response: once again, it pays to invest in a reputation for cooperation even if one's motives are wholly selfish. The lesson of this literature too is that sustaining cooperative behaviour may only be difficult if governments are unduly myopic.

Control theory and game theory can yield useful general insights into policy formulation if  expectations are modelled properly and If agents are allowed to respond to policy In a non-mechanical fashion. These techniques can also Illuminate the conduct of recent UK macroeconomic policy. The government's Medium-Term Financial Strategy (MTFS) contained clear technical flaws that might have been avoided had greater attention been paid to the literature on. policy design. In particular, forward-looking behaviour in the foreign exchange market generated a marked sterling appreciation as a result of anticipated high real interest rates. This converted a gradualist strategy of disinflation into a much more abrupt process. Moreover, the emphasis on the Public Sector Borrowing Requirement (PSBR) as an intermediate policy target rendered inoperative the usual automatic fiscal stabilizers. The fall in aggregate demand which accompanied the severe disinflation reduced government revenues and increased transfer payments. The PSBR target obliged the government to increase taxes or cut government spending in a procyclical (and in Keynesian terms perverse) direction. This exacerbated rather than dampened the fall in demand and increased the volatility of the economy in the face of demand disturbances. In addition, the over-emphasis on monetary targets led to a neglect of important external considerations, and this was reflected in the behaviour of the exchange rate. These dangers were known at the inception of the MTFS, and are not criticisms made with the benefit of hindsight.

There is an evident need for greater partnership between the practice and the theory of policy formulation, though at present the prospects of realising such a partnership seem remote.

David Currie