Should the Bank of England be independent?

Blake, Andrew P. ; Westaway, Peter F.

1. Introduction

This note examines the question of whether the responsibility for the operation of monetary policy in the UK should be ceded to an independent Bank of England. While sterling was committed to membership of the European Monetary System, this issue was pre-empted since monetary policy was effectively determined by the actions of the Bundesbank. Now, with the UK's withdrawal from the EMS and no immediate prospect of any return, the question has again returned to the policy agenda. In fact, the arguments used to justify such a delegation of responsibility are closely related to those originally associated with the decision to join the EMS itself; that monetary policy should be controlled by an authority with a strong aversion to inflation and a reputation for sticking to its announced policy. It has been argued that an independent Bank of England would offer precisely these advantages without compromising the domestic (ie UK) objectives which were overridden by the Bundesbank.

It is important to recognise, however, that an independent Bank of England may no more offer a panacea for the UK economy's shortcomings than did membership of the EMS. Nor would such a policy arrangement imply that policymakers either in the Bank or the Treasury could be indifferent to the monetary stance in the rest of Europe. So if such an institutional change is to be contemplated, it is crucial to be able to identify the circumstances under which delegation of policy would be beneficial. To do this, it is necessary to conduct a rather more rigorous examination of the policy problem than is conventionally carried out in general discussion. This must be done by analysing macroeconomic policy choices in the context of dynamic game theory which allows us to investigate how the instruments of monetary and fiscal policy should be set with respect to the particular objectives of the relevant authorities, who must take into account the current and future actions of each other and private sector agents. Only then can we analyse the circumstances in which it is advisable for the UK government to surrender its monetary policy instruments to an independent Bank of England.

We argue that it is necessary to acknowledge three crucial but logically distinct aspects to the policy question. Firstly, what should the objectives of the Bank be compared to those pursued by the government itself? This question is especially relevant if the government has some prior control over the central bank's constitutional requirements or, as in the UK context, is able to appoint its Governor. Importantly, there are circumstances where, even in a democratic society, it may be desirable to surrender monetary policy to an institution with more 'conservative' preferences than the elected government. This occurs because the 'inflationary bias' implied by the optimal monetary stance will be lower with a more conservative decision maker although this can only be achieved at the cost of a reduced responsiveness to unexpected shocks to output. Secondly, what is the ability of the Bank to pre-commit itself to its policy announcements compared to the government itself? If the policy intentions of the monetary authority are credible in the sense that the private sector believe their policy announcements will be adhered to, then this would be sufficient to remove the inflationary bias. This puts a premium on policymakers having a reputation for keeping their word. Thirdly, it is necessary to know the relationship between the fiscal authorities (which we assume to remain within the control of the elected government) and the Bank in its role as monetary policymaker; Is co-operation necessary for the delegation of authority to be beneficial or can the Bank be completely independent to follow its objectives? All three of these interrelated aspects will be clarified in the next two sections.

Ultimately, of course, the relative importance of these factors can only be determined by empirically based work. Consequently, in section 4 of this note we use the National Institute UK econometric model to illustrate some of the issues raised. Section 5 draws some general conclusions.

Before proceeding, it is worth emphasising that the arguments in this paper are predicated on the assumption that policymakers are attempting to manipulate the instruments of policy to maximise some form of representative social welfare function. It is obvious that in practice, these objectives are likely to be somewhat more complex than the simple performance criteria (inflation and output) adopted in the following sections. Nevertheless, we would argue that the methods described in this note and elsewhere represent the only systematic procedures for evaluating the desirability of a change in institutional structure such as the setting up of an independent Bank of England. Of course, in practice, politically motivated policymakers may adopt a much less consistent approach to policy choice, instead changing policy in response to every short term political opportunity. In such circumstances, the case is much stronger for delegation of policy to an independent Bank whose objectives more closely reflect the electorate's preferences.

2. A monetary policy game

We begin by reviewing the most basic monetary policy game on which it is possible to make the argument for delegating control to an independent central bank. The key feature of the model that we use is that private sector expectations of policy have an influence on the outcome. This renders the optimal monetary policy time inconsistent, that is, policymakers will have an incentive to renege on the strategy once it has been announced. We can illustrate this using the simple static framework originally described by Kydland and Prescott (1977) and Barro and Gordon (1983).

Output is determined by a simple surprise supply curve

|Mathematical Expression Omitted~

where |y.sub.t~ is output, ||Pi~.sub.t~ is inflation, |Mathematical Expression Omitted~ is expected inflation and ||Epsilon~.sub.t~ is an independently normally distributed, mean zero disturbance term with a variance ||Sigma~.sup.2~. The monetary authority controls the inflation rate by choice of money supply so that the monetary instrument is inflation itself. It is assumed to act to minimise a well defined quadratic objective function:

|Mathematical Expression Omitted~

|Mathematical Expression Omitted~

where the target level of inflation is zero while |Mathematical Expression Omitted~ is the target level of output which is greater than the natural rate, reflecting tax or labour market distortions in the economy. |Lambda~ represents the relative priority attached to inflation and output, thus a higher |Lambda~ implies a greater degree of conservatism.

Expectations of inflation formed by the private sector, wage setters for example, are assumed set prior to the realised inflation rate and in equilibrium will be fulfilled except for a random error. Thus any predictable element of the monetary authorities policy rule should be incorporated into the expectations formation mechanism. The crux of the problem, however, is that wage setters know that the policymaker has an incentive to stimulate output above its natural rate once wage contracts have been agreed to for that period. As a consequence, they will build an inflationary mark-up into their wage demands which is just high enough to prevent the policymaker from finding the boost to output worthwhile. The resulting inflation rate will be given by

|Mathematical Expression Omitted~

while output will remain at its natural rate. The second term,

|Mathematical Expression Omitted~

represents the inflationary bias that Kydland and Prescott identified. This outcome is known as the discretionary or time consistent equilibrium.

By contrast, if the policymaker can pre-commit to an announced policy rule despite the ex post incentive to renege, then expected and actual inflation can effectively be chosen together, resulting in a co-operative solution where the inflationary bias is set at zero ie

||Pi~.sub.t~ = -|Lambda~||Epsilon~.sub.t~ (4)

and output is again (on average) at its natural rate. Note that the optimal rule does not preclude a response to the unexpected output shock. If the monetary authority could be tied to implementing (4)--in other words have their discretion taken away from them so that they cannot reoptimise--then this is welfare improving. This is the origin of the 'rules versus discretion' debate in the modern macropolicy literature. The rule (4) is superior to the time consistent policy (3) which is the only sustainable equilibrium if the policymaker retains the ability to reoptimise.

Barro and Gordon (1983) suggest mechanisms whereby this outcome can be achieved noncooperatively, but legislative changes or institutional reform are required to guarantee the monetary authority the required pre-commitment. This is an argument put forward to justify delegating monetary policy to an authority not subject to the same political pressures that policymakers are. But is central bank independence necessarily sufficient to endow the reputation for sticking to its guns and not reoptimising? Although this is typically thought to be true of the Bundesbank, it may not be a reasonable characterisation of a newly independent central bank such as the Bank of England.

There is, however, a further justification for appointing an independent and, moreover, conservative central banker. Rogoff (1985a) showed that the inflationary bias can be diminished by appointing a central banker which was more conservative than the policymaker. Imagine that a central banker was appointed whose preferences were represented by |Mu~ |is less than~ |Lambda~ ie acts to minimise the cost function

|Mathematical Expression Omitted~

The consistent policy associated with this is

|Mathematical Expression Omitted~

which reduces the inflationary bias since

|Mathematical Expression Omitted~

Appointing an independent central banker with preferences different to those of the policymaker ensures a lower discretionary inflation rate. An independent 'conservative' central banker acting to minimise (5) can therefore be welfare improving relative to the policymaker setting monetary policy itself with reference to the 'true' objective function (2).

This view is widely held. For example, Mankiw (1990, p. 1650) on discussing the best choice of a central banker states that 'an alternative to imposing a fixed rule is to appoint individuals with a fervent distaste for inflation', implying that the time inconsistency can be circumvented by an appropriate institution. However, Rogoff (1985a) noted an important caveat to this result. Consider (6). Although the inflationary bias is reduced so is the response to the output shock ||Epsilon~.sub.t~. If the variance of the shock, ||Sigma~.sup.2~, is great enough then the benefit of a lower inflation rate can be outweighed by increased expected costs associated with insufficient stabilisation of output.

It is helpful to explain these arguments graphically. Suppose that the true preferences of the policymaker are given by |Lambda~ = 0.3. Charts 1 and 2 illustrate the perceived welfare losses for different values of |Mu~. This allows us to read off the range of values for which the outcome will be preferable; these will occur where the distortion to the output response is outweighed by the fall in the inflationary bias. This is simply a graphical exposition of the Rogoff (1985a) result. For ||Sigma~.sup.2~ = 0, we reproduce the deterministic case where it will always be optimal to cede control to a conservative banker, indeed the more conservative the better. This is shown by the locus of discretionary welfare losses which always lies below the original welfare loss for all values of |Mu~ below 0.3. As the variance of the output shock ||Epsilon~.sub.t~ increases however, so Rogoff's point is that if |Mu~ is chosen too low (ie if the central banker is too conservative) then it is may be preferable for the policymaker to maintain control of monetary policy despite the inflationary bias; Chart 2 illustrates for ||Sigma~.sup.2~ = 2 where |Mu~ must be chosen above 0.13.

Now consider how these conclusions are affected when the chosen central banker is also assumed to be able to pre-commit itself. Chart 1 confirms the point noted above that in the deterministic case, it would be worth surrendering control to any central banker, conservative or liberal, with pre-commitment capability since any preference parameter |Mu~ would deliver the zero inflation outcome. As the variance of the output shock increases, however, so the scope for choice narrows although it is always greater than in the cases when the banker lacks leadership. For example, consider Chart 2 where ||Sigma~.sup.2~ = 2. If the central banker has no pre-commitment capability, then any value of |Mu~ between 0.13 and 0.3 would guarantee a superior expected outcome with |Mu~ yielding the optimal degree of conservatism. But if the central banker can be assumed to have leadership, then a value of |Mu~ between 0.11 and 0.48 would be preferable to policymakers keeping their sovereignty (note that even a less conservative central banker would be superior). Of course, the optimal degree of conservatism when the central bank does have pre-commitment is to be equal to that of the policymakers themselves, ie |Lambda~ = 0.3.

One final point worthy of particular emphasis is the fact that the policy conclusions reached from the analysis of this simple model are importantly affected by the existence of stochastic shocks. Here, the greater the degree of noise, the less likely it is to be advantageous to surrender the instruments of policy to an independent central bank. This influence of the stochastic environment occurs despite the fact that the optimal policies derived are themselves certainty equivalent. Rather it is the design of the institutional structure which is affected by the effect of disturbances. This point is important to bear in mind when considering the merits of policy regimes on more sophisticated models for which a full stochastic analysis is not possible.

An interesting development in the literature has been to consider policies that are conditioned on the size of shocks. Lohmann (1992), for example, elegantly suggests how an 'escape clause' can be built in. The policymaker overrides the bank in the face of an extreme shock and imposes greater output stabilisation. Although this seems a promising avenue for future research the analysis of such equilibria quickly becomes very complicated when we introduce two additional factors--dynamics and fiscal policy. These complications also make the two justifications outlined above to appoint independent central bankers rather less convincing. We discuss the implications of both in the next section.

3. Monetary and fiscal policy in dynamic rational expectations models

So far, in the context of a simple static model of monetary policy alone, we have concentrated on two aspects influencing the decision to delegate sovereignty to an independent Bank of England; the objectives of the Bank and its pre-commitment capability compared to the government itself. Introducing fiscal policy brings a new complication. Assume that the government retains control of fiscal policy and has a well defined objective function. Appointing an independent central banker increases the number of decision makers from one to two.(1) Now, there is scope for the monetary and fiscal authorities to act either cooperatively or noncooperatively. When acting cooperatively they minimise a weighted average of their respective objective functions. Noncooperatively they act to minimise their cost functions separately. Of course, the question to be addressed in choosing whether to delegate the instruments of monetary policy remains whether welfare is higher on the grounds of the government's own criteria, not on the basis of any weighted average cost function.

Another complication which arises when we consider more realistic models is that behaviour is likely to have effects which last for longer than the current period. Such a move from static to dynamic models brings further considerations to the macropolicy game. So far we have only discussed credibility in the context of pre-commitment of policy announcements in the current period. For static games, of course, this amounts to complete pre-commitment since the 'game' only lasts for one period. Once dynamic games are considered, however, policymakers must be able to re-commit themselves to policy intentions over current and future periods. If a policymaker is able to make a credible policy announcement which takes into account the strategic response of the other players in the game, then it is said to be acting as a Stackelberg leader. If no player can exercise this type of leadership, we revert to the Nash equilibria analysed in section 2. In fact, the ability of a particular policymaker to assume leadership may in part depend on the particular policy instrument adopted; for example it may be easier for the fiscal authority to assume leadership since the fiscal instrument may be difficult to alter once set; by contrast, the monetary instrument, interest rates say, can be altered at a few hours notice. Secondly, there is commitment over time. In general, policymakers will have an incentive to re-optimise period-by-period even in the absence of unanticipated shocks; this is the dynamic manifestation of the time inconsistency problem. The longer the period over which a policymaker can be committed, the more desirable will be the policy outcome. In the context of the desirability of an independent Bank of England, it seems reasonable to assume that such an institution is more likely to commit itself than a politically motivated policymaker (although it is important not to confuse the pre-commitment period with the rate of time discounting assumed by the policymaker). In fact, the number of possible equilibria to the dynamic macropolicy game is even greater than this, depending for example whether policy is set down in terms of paths for policy instruments, known as open loop equilibria, or whether they commit themselves to policy rules, known as feedback equilibria. These further complications, and more complicated examples associated with differing pre-commitment periods will not be dealt with here; for further details see Blake and Westaway (1992), Blake (1992a, b), Westaway (1990).

To recapitulate, when we consider both fiscal and monetary policy in a dynamic model there are four interrelated dimensions to the problem.

* The objective functions of the policymaker and candidate independent central bankers. This can include differences in the weightings on policy targets as we assume in this note, but also differences in the targets themselves; for example, within the EMS, the Bundesbank were concerned with German rather than UK inflation.

* The degree of co-operation between the fiscal and candidate monetary authorities.

* The length of period over which commitments can be made by the policymaker and candidate monetary authorities. For dynamic games this typically reflects the choice of either open loop or feedback strategies.

* The leadership that the (potential) monetary and fiscal authorities have both with respect to each other and with respect to the private sector. For open loop strategies this means that a policymaker exercising leadership is pre-committing to a time inconsistent strategy.

So how do we decide under what circumstances it might be advantageous for the Treasury to cede control of monetary policy to an independent Bank of England? It should be clear that this question is by no means straightforward. Rather than exhaustively list the possible games that can be played we first focus on four stylised regimes and give a qualitative account of the issues that arise.

(1) Identical objective functions; differing degrees of leadership and/or commitment.

If the objective functions of the government and the Bank are identical then the degree of cooperation is unimportant. It would only be worthwhile for the policymaker to hand over control of monetary policy if the independent central banker either has leadership with respect to the private sector or has a greater degree of pre-commitment than the policymaker acting alone (or both). In this circumstance leadership with respect to each other makes no difference. Cohen and Michel (1988) argue that, in general, fiscal authorities have per period leadership but monetary authorities do not. Currie (1992) argues the converse where there are independent central banks--fiscal authorities are by nature profligate and the independence of a central bank is a natural restraining feature. Blake (1992a) for a small numerical model shows that for time consistent policies where one of the authorities has leadership then it is the leadership of the monetary authority that is most beneficial.

(2) Differing objective functions; commitment period and leadership unaffected by ceding control.

What if we ignore the question of reputation in the sense that we assume that ceding to a central bank does not change the leadership structure or commitment period? This may be realistic if we argue as Currie et al. (1992) that a new independent European central bank would not have the commitment that, say, the Bundesbank has, at least in the short term. In these circumstances, turning over to an independent Bank of England would only be beneficial if it helped reduce the inflationary bias without imposing both output costs or inducing unacceptably high variability in the fiscal instrument. This could potentially be achieved by the two authorities playing either cooperatively or noncooperatively. Playing cooperatively they minimise a weighted average of the two objective functions, noncooperatively they act independently. We need not rule out leadership with respect to the private sector or open loop commitments. Currie et al. (1987) give an algorithm where both authorities have open loop commitments and leadership with respect to the private sector and Currie et al. (1992) an open loop solution where only one authority has leadership.

Note that if the policymaker had the ability to lead and commit itself to an open loop strategy for both monetary and fiscal policy then it would never pay to cede control to a central bank whose objectives were different from its own as by definition it can implement the best possible policy. It is also not necessarily the case that acting noncooperatively is welfare reducing (as noted in other contexts by Rogoff, 1985b, Kehoe, 1989) although in general it will be.

(3) Differing objective functions; leadership and/or commitment affected by ceding control; cooperative play.

Assuming that the two authorities play cooperatively and the monetary authority is more conservative than the policymaker then the different degrees of leadership and commitment become important. If the monetary authority has either more commitment or leadership (or both) this is likely to be welfare improving as long as it is not too conservative. As with the simple static example described in section 2, excessive care about inflation could impose such output costs that the gain associated with the change in leadership or commitment is insufficient.

(4) Differing objective functions; leadership and/or commitment affected by ceding control; noncooperative play.

Switching from the previous case to include noncooperative play even when cooperative policies are welfare improving changes the outcome again, and possibly for the worse. The monetary authority in such circumstances would need to be even less conservative.

Having discussed these issues in general terms, it is obviously necessary to assess the relative importance of the different aspects of the problem by conducting an empirical investigation which is relevant for the particular example of the UK fiscal and monetary authorities. We do not propose to investigate all the possible equilibria but instead investigate a subset of the first two in the next section.

4. An empirical investigation

There are substantial technical and computational problems when we attempt to calculate the equilibria described in section 3 using a nonlinear macroeconometric model, such as the National Institute UK model. Most of the pioneering analytical work uses linear models for two reasons. Firstly, algorithms for calculating the equilibria for linear models are easily and cheaply implemented. Nonlinear models can only be solved numerically which can be computationally expensive. Secondly, nonlinear models are usually solved over finite time horizons because they require a database. When the model embodies forward looking behaviour the existence of a 'terminal time' can introduce significant distortions in numerical solutions (see Blake and Westaway, 1992a). Linear (or linearised) models can be solved over an infinite time horizon, and are not subject to such induced errors. Therefore we have investigated the importance of these effects using a linearised version of the National Institute macroeconometric model of the UK. For details of how the linearisation was obtained see Blake and Westaway (1992a, b).

Before calculating the outcomes associated with the possible regimes, we must specify the two authorities' objectives. These are represented by discounted infinite horizon quadratic loss functions. They must be sufficiently general to allow both to share the same objectives and to allow for marked differences. Define the two per period loss functions by

|Mathematical Expression Omitted~

and

|Mathematical Expression Omitted~

Here |Mathematical Expression Omitted~ is the target level of output in period t, |Mathematical Expression Omitted~ is the target inflation rate, |Delta~|r.sub.t~ is the change in the real interest rate (our monetary instrument) and |g.sub.t~ the ratio of government spending to GDP (the fiscal instrument). In all cases zero represents the base value. We need to weight the instruments to prevent excessive movements which are either infeasible or undesirable.

The 'true' objective function of the fiscal authority is (8). As in section 2, |Lambda~ measures the 'degree of conservatism' and can vary between zero and unity. If |Lambda~ = 0 the fiscal authority cares only about inflation, |Lambda~ = 1 means only output matters. In all our examples |Lambda~ = 0.3.

The objective function of the monetary authority is (9). A more conservative central banker than the policymaker would be reflected by |Mu~ |is less than~ |Lambda~. We used three values of |Mu~: 0.3 (the same as the fiscal authority), 0.2 and 0.1. Additionally, |Beta~ measures whether the monetary authority attaches costs to movements in the fiscal instrument. For the two authorities to share identical loss functions requires |Mu~ = |Lambda~ = 0.3 and |Beta~ = 2.

If the monetary and fiscal authorities do not cooperate they minimise the discounted infinite sum of their own per period loss functions, ie the fiscal authority acts to minimise

|Mathematical Expression Omitted~

while the monetary authority minimises

|Mathematical Expression Omitted~

The common discount factor of 0.99 implies an annual discount rate of approximately 4 per cent. That they are the same for both authorities is a convenient simplification. If they do cooperate they minimise a weighted average of the two objective functions. The cooperative loss function is then

|Mathematical Expression Omitted~

In all cases the 'true' loss function against which we should measure the alternative regimes is that of the fiscal authorities, (10).

Apart from defining the objectives of the fiscal and monetary authorities, we also need to specify their pre-commitment capabilities. In our empirical examples, we assume that neither is able to make binding commitments and hence they are restricted to following time consistent strategies; we include, however, the optimal time inconsistent policy as a measure of the time inconsistency problem. Remember, this comparison is important since the delegation of control to an independent central banker is only welfare improving if time inconsistency exists.

Ideally, to analyse the implications of an independent Bank of England, we would need to compute variants of the National Institute forecast not only for the different assumptions discussed but also taking into account the range of unexpected disturbances which might impinge on the UK economy over the policy horizon. For simplicity, however, we focus on one representative scenario which might be seen to contain the main elements of the overall policy problem; we assume that both government policymakers and the central bank wish to reduce inflation towards its target level, which is 1 per cent below the prevailing rate, leaving the level of output unchanged. The linearised model is implicitly defined as deviations from base, so that to drive inflation down by 1 per cent sets |Mathematical Expression Omitted~, with |Mathematical Expression Omitted~ implying output stays at base.

In Table 1 we report the loss associated with each regime as a ratio of the loss associated with the policymaker retaining control of both monetary and fiscal instruments but with no commitment, our benchmark. In the first row we report the loss ratio for the optimal but TABULAR DATA OMITTED time inconsistent reduction of inflation and in the second the benchmark loss ratio, which is of course unity. Importantly, the time inconsistent policy does not yield much of an improvement. While this may be slightly surprising given the prevalence of forward-looking behaviour in the National Institute model, it should be recalled that the model also embodies a considerable degree of both real and nominal inertia. For example, the effect of an interest-rate change may have a rapid effect on the exchange rate itself but the consequences for prices may be much more protracted. In fact, this accords with earlier empirical results obtained by Christodoulakis et al. (1991). With very little scope for improving the outcome by giving the policymaker such pre-commitment ability it is important to find which regimes are seriously welfare reducing.

We consider two classes of time consistent policies. Firstly, we give both authorities identical objective functions but allow the monetary authority but not the fiscal authority to lead the private sector. This is an example of case 1 in section 3, where delegating control alters the leadership structure. The result is shown in row 3 of Table 1. It shows a marginal improvement, in line with the modest gains for the fully optimal time inconsistent policy.

Secondly, we allow for leadership to remain unchanged by the delegation but let objective functions differ, case 2 in section 3. In rows 4 and 5 of Table 1 both authorities share the same objective function except that the monetary authority sets |Mu~ = 0.2, implying slightly more concern for inflation than the policymaker. In row 4 they act cooperatively, row 5 separately. Only the cooperative case is welfare improving. Additionally, compare this with rows 8 and 9, where the central bank is even more conservative. The outcome if they act cooperatively is still marginally better than not delegating control but worse than for the less conservative monetary authority. Appointing the most conservative central banker would be a mistake. Equally, if the fiscal and monetary authorities do not cooperate the loss ratio deteriorates substantially.

In rows 6 and 7 and rows 10 and 11 we report the loss ratios for equilibria where the monetary authority does not attach costs to movements in fiscal instruments for both degrees of conservatism. Now even the cooperative cases are welfare reducing. This indicates that it is not enough for the monetary authority to care about lost output. It also needs to be aware of how much it burdens the fiscal authority by policy choices.

Three conclusions can be drawn from these results. Firstly, the lack of much time inconsistency in the model indicates that there may be very little gain possible by creating an independent Bank of England. This lack of inflationary bias immediately weakens the case for such an institutional reform as neither commitment nor conservatism can deliver much in the way of increased welfare. Secondly, it is much better for the two authorities to act cooperatively. A central bank that pursues its own goals with no regard for the social welfare function is frequently harmful. Thirdly, even if the authorities cooperate they must take proper account of the costs of the other's policy instruments.

5. Conclusions

This note has attempted to cast light on a question which has often been raised in the policy debate, but has been mooted with increasing frequency since the departure of the UK from membership of the ERM. One of the main intentions of this work has been to clarify the theoretical issues which are involved in addressing this question. Another has been to offer some empirically based conclusions relevant to the current UK policy debate. These suggest that the economic grounds for delegation of monetary policy to the Bank of England may be rather weak, and indicate instead that a unified approach to fiscal and monetary policy is more likely to achieve the government's objectives. This tends to contradict some of the more simplistic propositions which have been associated with the arguments for an independent Bank of England. To take the most common example, it has been suggested that an independent monetary authority should be charged with the sole objective of achieving zero inflation. Our analysis has suggested that this may be a foolish strategy to adopt since it may impose objectives on the Bank which are too different from those of the government who also care about output. Crucially, this concern for real activity may be legitimate even if the natural rate of output is invariant to government policy in the long run.

More generally, we have shown that the question of whether and under what conditions the Bank should be granted independence depends on a number of distinct theoretical concepts which are often confused in popular discussion. For example, it has often been claimed that the 'credibility' of monetary policy would be enhanced by an independent central bank, indeed as it was claimed it would be under Bundesbank control within the EMS. Yet, as we have argued here, the precise meaning of credibility in this context is ambiguous. Thus, it is unlikely to be sensible to appoint a monetary authority with an ability to make credible policy commitments if at the same time it is following objectives which differ markedly from those of the government itself.

This note is intended as a contribution to the growing amount of work on this topic, but ultimately we believe that further systematic empirically based investigations of the question are still required.

NOTES

(1) In fact, in game theoretic terms, the number of players in the macropolicy game increases from two to three since the forward-looking private sector counts as a player.

REFERENCES

Barro, R.J. and Gordon, D.B. (1983) 'Rules, discretion and reputation in a model of monetary policy', Journal of Monetary Economics, 12, 101-121.

Blake, A.P. (1992a) 'Time consistent mixed precommitment macropolicy', National Institute Discussion Paper No. 7, (new series).

Blake, A.P. (1992b) 'Equilibrium monetary policy in a model of inflation and employment', National Institute, Discussion Paper No. 29, (new series).

Blake, A.P. and Westaway, P.F. (1992a) 'An analysis of the impact of finite horizons on macroeconomic control', National Institute Discussion Paper No. 11, (new series).

Blake, A.P. and Westaway, P.F. (1992b) 'Do central bankers have to be conservative?', National Institute Discussion Paper No. 26, (new series).

Christodoulakis, N.C., Gaines, J. and Levine, P. (1991) 'Macroeconomic policy using large econometric rational expectations models: Methodology and application', Oxford Economic Papers, 43, 25-58.

Cohen, D. and Michel, P. (1988) 'How should control theory be used to calculate a time-consistent Government policy?', Review of Economic Studies, 55, 263-274.

Currie, D.A. (1992) 'European Monetary Union: Institutional structure and economic performance', Economic Journal, 102, 248-264.

Currie, D.A., Levine, P. and Pearlman, J. (1992) 'European Monetary Union or hard EMS?', European Economic Review, 36, 1185-1204.

Currie, D.A., Levine, P. and Vidalis, N. (1987) 'International cooperation and reputation in an empirical two-bloc model', in Bryant, R.C. and Portes, R. (eds.) Global Macroeconomics: Policy Conflict and Cooperation, (London: Macmillan Press Ltd.).

Kehoe, P. (1989) 'Policy cooperation among benevolent Governments may be undesirable', Review of Economic Studies, 56, 289-296.

Kydland, F. and Prescott, E.C. (1977) 'Rules rather than discretion: The inconsistency of optimal plans', Journal of Political Economy, 85,473-492.

Lohmann, S. (1992), 'Optimal commitment in monetary policy: Credibility versus flexibility', American Economic Review, 82, 273-286.

Mankiw, N.G. (1990) 'A quick refresher course in macroeconomics', Journal of Economic Literature, 28, 1645-1660.

Rogoff, K. (1985a) 'The optimal degree of commitment to an intermediate monetary target', Quarterly Journal of Economics, 100, 1169-1189.

Rogoff, K. (1985b) 'Can international monetary policy coordination be counterproductive?', Journal of International Economics, 18, 199-217.

Westaway, P.F. (1990) 'Precommitment and sustainability', paper presented at the 1990 meeting of the Society for Economic Dynamics and Control, Minnesota, USA.