Central bank independence and macroeconomic performance: a survey of the evidence. Friedrich Kißmer and Helmut Wagner. Diskussionsbeitrag Nr.

Transcription

1 Central bank independence and macroeconomic performance: a survey of the evidence by Friedrich Kißmer and Helmut Wagner Diskussionsbeitrag Nr. 255 Juni 1998 Fachbereich Wirtschaftswissenschaft FernUniversität Hagen Postfach 940 D Hagen

2 Prof. Dr. Helmut Wagner Lehrstuhl für Volkswirtschaftslehre, insb. Makroökonomik Fachbereich Wirtschaftswissenschaft FernUniversität Hagen Eugen-Schmalenbach-Gebäude Universitätsstraße 41 D Hagen Dr. Friedrich Kißmer Lehrstuhl für Volkswirtschaftslehre, insb. Makroökonomik (Prof. Dr. Helmut Wagner) Fachbereich Wirtschaftswissenschaft FernUniversität Hagen Eugen-Schmalenbach-Gebäude Universitätsstraße 41 D Hagen

3 Central bank independence and macroeconomic performance: a survey of the evidence Friedrich Kißmer and Helmut Wagner FernUniversität Hagen 1 Introduction In the 1990s numerous countries provided their central banks with greater legal independence from the government. This trend to an increase in the degree of central bank independence can be seen not only in highly industrialised countries but also in transition countries, as well as in developing countries, in particular in Latin America. Cukierman (1996) lists 25 countries in various parts of the world which strengthened the independence of their central banks between 1989 and the beginning of Seen from today, this list is incomplete, because changes in central bank statutes have taken place since 1996 as well. For example, in May 1997, the Chancellor of the Exchequer, Gordon Brown, announced that the Bank of England could decide on the use of its monetary policy instruments without any interference from the government. 1 The concrete reasons for institutional changes to central bank structures vary from country to country. In the Member States of the European Union, some of the amendments to national central bank statutes must be seen in the context of the requirements of legal convergence which result from the Maastricht Treaty and the Statute of the European System of Central Banks (ESCB). An example of this is the amendments of the German Central Bank Act of , which, among other things, increase the minimum term of office of members of the decision-making bodies of the Bundesbank from two to five years and abolish the government's right of suspensory veto. In addition, price stability was laid down as the 1 However, the goal of monetary policy is still set by the government in the form of a direct inflation target.

4 2 primary objective of monetary policy (Deutsche Bundesbank 1998, pp. 25f.). With the establishment of a European Central Bank, monetary policy in the participating countries will be delegated to a supranational institution which will be provided with a large extent of legal independence from the interests of national governments (Alesina/Grilli 1992; De Haan 1997). The transformation process in the former socialist countries of Central and Eastern Europe has also been accompanied by substantial central bank reforms. Some countries (e.g. Czech Republic, Hungary) have taken the statutes governing independent central banks in industrialised countries in Western Europe as a model for the reform of their own central bank statutes. As several countries would like to become members of the European Union, the requirements of the Maastricht Treaty will make further reforms necessary in future here as well (Radzyner/Riesinger 1997, p. 57). In addition, some developing countries have enforced institutional reforms concerning the relationship between government and central bank. Countries in the Western Hemisphere for example, such as Argentina, Chile, Colombia, Mexico and Venezuela, have been attempting to secure their stabilisation progress by providing their central banks with more legal independence (Siklos 1995; Cukierman 1996). The institutional reforms to establish independent central banks vary from country to country in essential details. Apart from creating independent central banks by law some countries enforced additional specific mechanisms to make central banks explicitly accountable. Countries like New Zealand, Canada, United Kingdom, Sweden, Australia, Finland, Spain have adopted an inflation targeting framework for monetary policy. The announcement of an inflation target for monetary policy is aimed in the first place at fulfilling the function of a nominal anchor. Therefore this strategy is an alternative to monetary targeting or to exchange rate pegging. For these purposes direct inflation targeting should signal a restriction of the central bank's ability to behave in a discretionary manner and in addition ensure transparency of monetary policy. If the inflation targets are set by the government for a defined period, or agreed between the government and the central bank, inflation targeting forms an alternative institutional design for an (instrument-) independent central bank which stresses the accountability of the central bank. In this sense, inflation targeting is aimed at preventing a

5 3 democratic deficit which can emerge when political decisions are taken by an authority which is not directly legitimated by elections (Briault/Haldane/King 1997; Nolan/Schaling 1996). Mainly, the three following reasons are put forward for the establishment of independent central banks (see Debelle/Fischer 1994 as well, p. 195): a) Central bank independence is justified theoretically by the existence of incentive constraints for politicians which force dependent central banks to realise suboptimally high rates of inflation (Persson/Tabellini 1997). Since the study by Kydland/Prescott (1977) this inflation bias has been discussed primarily against the background of the credibility problems of monetary policy caused by dynamic inconsistency (see Section 2). In addition, the political independence of the central bank is aimed at protecting against partisan or electoral cycles in monetary policy (Alesina 1988, p. 40; Alesina/Roubini 1997, p. 212), which in combination with ex ante uncertain election outcomes in turn may lead to suboptimally high output volatility and contribute to a higher inflation bias (see Alesina/Gatti 1995). b) Numerous empirical studies (see Eijffinger/De Haan 1996 for a survey) claim that in industrial countries legal measures of central bank independence are inversely related to the average inflation. In contrast, there is no systematic relationship between central bank independence and economic growth. In the case of developing countries it is observed that (in contrast to developed countries) legal independence of central banks is not a good measure of the actual central bank independence. The negative correlation between central bank independence and inflation is confirmed, however, if so-called behaviourally oriented indices like the actual turnover rates of central bank governors or the political vulnerability of the central bank governor are used to measure the degree of central bank independence (Cukierman 1994; 1996). c) Moreover the track record of the German Bundesbank which is regarded as a particularly independent central bank is referred to. The empirical and theoretical justifications for central bank independence are not undisputed in the literature. In some cases, the consistency of indices based on interpretations of central bank statutes is disputed in general and in particular concerning some concrete indices

6 4 (Mangano 1997; Neumann 1996; Eijffinger/Schaling 1995). In addition, the originally maintained correlations between central bank independence and macroeconomic variables are not always confirmed (Campillo/Miron 1996; Fuhrer 1997), the causality between central bank independence and inflation is disputed (Posen 1993), and higher disinflation costs as a result of a higher sacrifice ratio with central bank independence are also claimed (Debelle/Fischer 1994; Fischer 1996; Jordan 1997; Hutchison/Walsh 1997). Furthermore the theoretical foundations of central bank independence, as far as they are based on the problem of time inconsistency, have not yet been fully researched. In this context there is increasing criticism that institutional reforms do not really solve the credibility problem, but merely relocate it (McCallum 1995; al-nowaihi/levine 1996; Jensen 1997; Illing 1998). The remainder of this paper is organised as follows. Section II provides a brief overview of the theoretical foundations of central bank independence. In section III the criteria used to measure central bank independence in the empirical literature are presented first. The following overview of the macroeconomic consequences of central bank independence focuses on the empirical evidence regarding inflation, output (economic growth, respectively), and disinflation costs. This procedure can be justified by the small number of empirical studies dealing with the relationship between central bank independence and other macroeconomic variables such as interest rates, budget deficits etc. (see Eijffinger/De Haan 1996, p. 38). Furthermore many economists consider inflation and output as the main determinants and indicators for the level of social welfare. Therefore the empirical literature is evaluated in particular against the background of Grilli/Masciandaro/Tabellini`s free-lunch hypothesis. This hypothesis states that central bank independence entails social benefits in terms of low inflation rates but no apparent costs in terms of real macroeconomic performance (like increased output volatility or reduced economic growth). Section IV concludes and suggests topics for further research.

7 5 2 The theoretical basis of central bank independence The modern theory of central banking stresses the importance of institutions and incentive structures for monetary policy. According to Grilli/Masciandaro/Tabellini (1991), monetary institutions can be described by the political and economic independence of the central bank. Economic independence is defined as the ability of the central bank to determine the use and the choice of its monetary policy instruments autonomously and without interference from the government. 2 Economic independence may be adversely affected by central bank s obligations to finance the government budget, to supervise commercial banks and by a lack of freedom to set interest rates (Grilli/Masciandaro/Tabellini 1991, p. 368; Grilli/Alesina 1992, p. 56). In addition, the question of who (the government or the central bank) is in charge of exchange rate policy has to be addressed in this context. Political independence is defined as the ability of the central bank to choose monetary policy goals autonomously and without interference from the government. The basic determinants for this ability are found in personal independence (e.g. procedures for appointing and dismissing central bankers, terms of office), in the government's rights to give instructions to the central bank as well as the right to veto, to suspend or to defer central bank s decisions, and its right to be present on the central bank board. Furthermore the degree of political independence depends on the formal responsibilities of the central bank. In this context, however, different interpretations are placed on the role of statutory requirements that the central bank pursues price stability. Grilli/Masciandaro/Tabellini (1991), Alesina/Grilli (1992) and Cukierman (1992) see an explicit mandate for the central bank to restrain inflation as a strengthening of central bank independence. In contrast to this, Debelle/Fischer (1994) and Fischer (1995a; 1995b) characterise this as a reduction of the central bank's "goal independence". The reason for these deviating interpretations is found in the different accentuation of two aspects: Debelle/Fischer (1994) and Fischer (1995a; 1995b) emphasise that the central bank s discretionary powers may be restricted, while 2 This definition of economic independence is very similar to the meaning of instrument independence introduced by Debelle/Fischer (1994). Debelle/Fischer distinguish between instrument independence and goal independence. For more details see pp. 28f. of this paper and Debelle/Fischer (1994), p. 197.

8 6 Grilli/Masciandaro/Tabellini (1991), Alesina/Grilli (1992), Cukierman (1992) mainly stress the partial commitment of the government. The following sections 2.1 and 2.2 deal with the meaning of these institutional details. 2.1 The inflation bias of monetary policy A starting point for the theoretical foundations of central bank independence consists in the consideration that the behaviour of politicians is also greatly influenced by existing rules of the game. Even for the simple case that politicians do not pursue their own opportunistic or partisan interests in the implementation of monetary policy, existing incentive constraints can lead to suboptimal policies. 3 The fundamental reason for this is that policymakers operate in a discretionary regime, i.e. monetary policy decisions are taken sequentially over time in a second-best world and therefore a socially desirable monetary policy may suffer from a lack of credibility caused by time-inconsistency (Persson/Tabellini 1997). According to Blanchard/Fischer, a policy is time inconsistent when a future policy decision that forms part of an optimal plan formulated at an initial date (ex ante) is no longer optimal at the time the policy is implemented (ex post), although there is no relevant new information (Blanchard/Fischer 1989, p. 592; Schaling 1995, p. 25). Various economic decisions are based on agents expectations of future monetary policies if we assume that the monetary authority is able to influence the inflation rate. For instance, when deciding on labour supply, wage contracts, investments or the portfolio allocation, agents have to form expectations of the future inflation rate. In a discretionary regime, as described above, policymakers can make revisions of ex ante announced policy decisions and therefore may create more inflation than forward-looking agents expect. Knowing this sequential timing of policy decisions as well as the policymakers objectives, rational agents do not expect the inflation rate to be in accordance with the ex ante optimal monetary policy if they anticipate that the monetary authority, once expectations are formed, has an incentive to deviate from the ex ante optimal 3 In contrast to this case of benelovent policymakers (i.e. policymakers are assumed to behave like social planners) models of political cycles highlight the role of policymakers opportunistic and partisan motivations. Politicians are said to be opportunistic if they only care about winning elections, and they are said to be partisan if they, once in office, pursue ideological goals (For an overview see Alesina/Roubini (1997) or Persson/Tabellini (1997), Chapter 3).

9 7 policy. The consequence of the time-inconsistency of the (ex ante) optimal but not enforceable monetary policy consists in the realisation of a credible policy which leads to a suboptimally high rate of inflation (inflation bias). Essentially, the literature refers to three reasons for this inflation bias (Cukierman 1992): a) Employment motive: In accordance with the short-run Phillips curve, monetary policy can only influence the level and variability of employment by means of a surprise inflation. If the private agents expectations are correct, output fluctuates around the natural rate level simply as a result of supply shocks. The basic assumption of the game theoretic approach to monetary policy, which is based on work by Kydland/Prescott (1997) and Barro/Gordon (1983), now consists in claiming that the natural rate level is lower than the socially desired output because of distortionary taxes or distorting wage-setting behaviour. If the government does not have sufficient non-distorting policy instruments, it has an incentive to surprise the private sector with an unexpected inflation, if, besides of inflation, society and the government regard deviations of output from distortion-free output as social costs, and non-indexed wages are fixed at the time the monetary policy is implemented. b) Revenue motive: In the presence of nominal debt the government has an incentive to generate an unexpected inflation. Inflation surprises have the consequence that nominal interest rates on government bonds remain unchanged in spite of inflation, and therefore not only the real debt burden declines but also real balances as the basis of seigniorage revenues remain unchanged. It need not necessarily be assumed here that the government's preferences differ from the social welfare function. For example, government revenues could also be used to finance public goods which may be arguments of the social welfare function. However, it must be assumed that the government can use only a limited amount of non-distorting taxes (see also Debelle 1996; Debelle/Fischer 1994; Beetsma/Bovenberg 1997). 4 c) Balance of payments motive: Cukierman (1992, Chapter 5) argues that policymakers of countries with persistent current account deficits may have an incentive to resort to an 4 Notice that because we focuse on the problem of time-inconsistency just the incentives creating a surprise inflation are considered. However, an anticipated inflation can also influence the government budget constraint. For example, even with anticipated inflation seigniorage revenues can increase.

10 8 unexpected devaluation (given purchasing power parity this is equivalent to an unanticipated increase in the price level). The main idea behind the time-inconsistency theory consists in the assumption that private agents are aware of the government's incentives and expect that rate of inflation which, they believe, the government no longer has a systematic incentive to deviate from. The rational expected inflation rate will be higher than the socially optimal rate of inflation if the social welfare function, apart from inflation, depends on one or more motives for unexpected monetary expansion and the government cannot enter into any binding commitment to realise the (ex ante) socially optimal rate of inflation. Facing these rational expectations the government in fact does no longer have any incentive to deviate from the expected rate of inflation and the expectations will be correct on average. Expectational errors caused by monetary policy can only stem from policy reactions to shocks which are known at the time the policy is implemented but not at the time rational expectations are formed. So there are no systematic output effects or reductions of the government debt at all. In comparison with the ex ante optimal policy, the time-consistent policy leads to additional social costs caused by suboptimally high inflation rates. 2.2 Institutional solutions to the time inconsistency problem One possible way to deal with the credibility problem consists in removing all discretionary power from the government. 5 The establishment of an independent authority would then be unnecessary if a strict, legally embedded simple k% money supply rule is reasonable. 5 In the following we ignore the repeated nature of policymaking, i.e. the fact that monetary policy is an ongoing process that involves repeated interactions between the policymakers and the public. However, these repeated interactions may lead to reputational effects which can mitigate the time-inconsistency problem. With repeated interactions policymakers have to balance the current benefits of a surprise inflation against the future costs of a higher expected inflation, if the private sector expects a similar monetary policy in the future. There are some (well-known) problems with reputational solutions regarding the expectaction formation mechanism and the multiplicity of equilibria. Since we cannot discuss these problems in detail, we simply assume that policymakers discount the future highly. This means, that there is still a need for other solutions to the credibility problem (Herrendorf 1997, p. 26; for an overview of reputational solutions see Persson/Tabellini 1997).

11 9 Government would only have to pass a law requiring the government to fix the growth of money supply at a steady rate. However, studies on the employment motive for monetary expansion show that when stochastic shocks are taken into account the optimal monetary policy does not conform to a simple rule but also includes an optimal shock absorption (Rogoff 1985; King 1996). By following a simple rule the government might be able to eliminate the inflation bias, but would produce suboptimally high output fluctuations. On the other hand, statutory entrenchment of the optimal state-contingent rule appears to be extremely difficult, because it is hard to imagine how all contingencies might be described ex ante and verified ex post. What remains is the choice between simple rules 6, which are inflexible, and discretionary policies which display an inflation bias. It is this trade-off between credibility and flexibility which has led to a game theoretic foundation of central bank independence. Here two approaches can be differentiated: 7 on the one hand, Rogoff s (1985) proposal to delegate monetary policy to an independent conservative central banker and the contracting or targeting approach on the other hand. What the theories have in common is that they propose the establishment of central bank structures which permit monetary policy to react to economic disturbances independently without interference from the government. However, they differ in their policy advises regarding the determination of central bankers objectives and incentives. i) The conservative central banker: Rogoff (1985) deals with the employment motive in the framework of a Barro/Gordon model with stochastic supply shocks. Rogoff shows that social welfare can be improved if the government delegates monetary policy to a "conservative" central banker who agrees with the social preferences regarding the target values of inflation and output, but places a greater weight to the inflation target than the government (society). Once appointed, the conservative central banker operates under discretion and is independent to pursue an activist policy (Fischer 1994, p. 290). By appropriate choice of the 6 We should stress, that some authors propose to pursue a rule with an escape clause under extreme circumstances (for example Flood/Isard 1989). Note, however, that simple rules (whether with or without escape clauses) may not be credible even if they are legally embedded. In this case, there is no choice between rules and discretion. 7 The following division does not do full justice to all theoretical approaches in every detail. For example, the Rogoff approach can also be interpreted in the sense of an inflation targeting (Rogoff 1985) or in the sense of a central bank contract with a quadratic transfer function (Persson/Tabellini 1993).

12 10 degree of conservativeness society realises a better equilibrium position than the government itself can achieve following inflexible rules or discretionary policies. In comparison to the discretionary solution without delegation the inflation bias is reduced but not completely eliminated. In the presence of supply shocks, this welfare gain is reduced by higher output volatility. Because the solution is based on the assumption that the conservative central banker optimises a preference function with, in comparison to the social welfare function, distorted (i.e. greater) inflation aversion, it presupposes central bank independence and the binding commitment of the government not to renege on delegation ex post. The interpretations of Rogoff's solution are quite different in the literature. 8 For example, the individual preferences of central bankers can be emphasised. In accordance with this view, the government selects a central banker type with an optimal degree of inflation aversion. This banker should have a greater inflation aversion than the government (in other words, he should be more conservative ), but he should not only be concerned with inflation, and in this sense should not be an "inflation nutter" (King 1996; p. 56). However, individual preferences regarding macroeconomic variables are often unknown and only vaguely defined (Persson/Tabellini 1997, p. 39). Furthermore, monetary policy decisions are not taken by just one person. One interpretation of the Rogoff solution which is a supplement, rather than an alternative, is that society has a central bank statute, or has agreed to a statute, which gives priority to price stability (Persson/Tabellini 1997, p. 39). The statutory requirement to pursue the goal of price stability is seen not only as legal indication of the corresponding central bankers preferences (Cukierman 1992, p. 377), but also in the sense of an indicator for the political independence of the central bank. Seen in this way, Rogoff's analysis becomes a theoretical foundation for a (legally based) political and economic independent central bank. For some economists the German Bundesbank can be taken as a prototype for the Rogoff model (Fischer 1995b, p. 276). Rogoff's findings imply that, ceteris paribus, in comparison to a dependent central bank, an independent central bank realises a lower rate of inflation, a lower variability of the inflation 8 For example, see Svensson 1997, p. 104; Persson/Tabellini 1997, p. 39; Neumann 1995, p. 11; Fischer 1995b, p. 277.

13 11 rate, and greater output volatility. Although it is true that the social welfare in a discretionary policy regime is improved, the trade-off between credibility and flexibility still remains. In an extension to Rogoff's approach Lohmann (1992) was able to show that Rogoff s solution can be improved through a partial reduction of central bank independence. The idea is that by establishing institutional features of the central bank (for example long tenure of central bank governors or majority rules which make it difficult to change a legislated law or override the decisions of the central bank board; see section 3.1) the government also fixes the size of the costs incurred by the incumbent politicians when they override the decisions made by the conservative central bank (Lohmann 1992, p. 277). Rogoff's analysis assumes that these costs are infinite. Lohmann shows that these costs should be finite (i.e. the government should retain the option of overruling or dismissing the central banker ex post), because in the case of very large supply shocks it would be optimal to override the inflation-averse central bank. In equilibrium, however, the conservative central banker is never overridden, because she prefers to react stronger to very large shocks than to small shocks in order to prevent interference from the government. Alesina/Gatti (1995) extend the Rogoff approach by taking partisan interests into consideration. In the two-party system assumed here, in which the preferences of parties differ merely in the degree of their inflation aversion, ex ante uncertainty regarding the outcome of elections leads to political uncertainty in monetary policy (see also Wagner 1990). However, both parties can increase their welfare if they cooperate and agree to the delegation of monetary policy to an independent central bank. The inflation bias will be lower than in the case of non-cooperative behaviour of the parties, but, in contrast to the Rogoff solution, output volatility will not necessarily be greater. Again an independent central bank stabilises pure productivity shocks suboptimally, but output variability caused by political uncertainty will be reduced (Alesina/Gatti 1995, p. 200). ii) The targeting or contracting approach: Starting from a principle-agent approach Walsh (1995a) and Persson/Tabellini (1993) come to the conclusion that, even though the government should transfer control over monetary policy instruments to an independent central bank, the government should also provide the central bank with incentives to optimise

14 12 the social welfare function. This will be done in the form of a (performance) contract between the government (as the principal) and the central bank (as the agent). Under the assumption that the preferences of the government and the central bank coincide, Walsh (1995a) shows that a simple contract which makes the central banker's remuneration linearly dependent on the realised rate of inflation eliminates the inflation bias without any sacrifice in stabilisation efficacy. The trade-off between credibility and flexibility would be solved by this and the welfare properties of the ex ante optimal rule mimicked. In addition, it was shown that the incentive structures of optimal performance contracts can also be generated through the implementation of inflation dependent dismissal rules (Walsh 1995b). Such dismissal rules come close to the corresponding rules of the Price Targeting Agreement in force in New Zealand (Walsh 1995b, Persson/Tabellini 1997, p. 42). As incentives in the standard approaches depend exclusively on deviations between realised inflation from the socially desirable rate of inflation, performance contracts are frequently interpreted in the sense of a direct inflation targeting as well. Svensson (1997) shows that, under conventional assumptions, the result of an optimal central bank contract can also be achieved when the government imposes an inflation target on the central bank which is below the socially desirable inflation target. 9 Seen in this way, the trade-off between credibility and flexibility would be solved by having an economic independent central bank (Debelle/Fischer (1994) use the term "instrument independence"), which would obtain clear inflation targets from the government and would therefore not be provided with any "goal independence" in the sense used by Debelle/Fischer (Svensson 1997 p. 99; Cukierman 1996, p. 26). The Bank of England might be seen as a real-world example for this institutional design in the sense that the Bank of England Act 1998 gives the Bank of England the responsibilitiy for setting policy instruments (i. e. interest rates) to meet the government s inflation target. 9 Svensson s solution can also be understood as the delegation of monetary policy to a conservative central banker with an extra low personal inflation target. However, because the personal preferences of the potential central banker are difficult to identify ex ante, Svensson prefers to impose a direct inflation target on central bank (Svensson 1997, p. 109; see also Cukierman 1996, p. 26). Green (1996) stresses the credibility problems of the Svensson solution, because the rationally expected rate of inflation is on average larger than the target rate set by the government (see also Svensson 1997, p. 108 f.; King 1996, p. 64).

15 13 However, an optimal central bank contract becomes considerably more complex if we consider distorted or uncertain preferences of the central bank and/or the government, central banker's risk aversion, or a state-contingent inflation bias. If optimal contracts are very complex, problems with regard to their implementation are raised because it becomes more difficult (i.e. there are transaction costs for the government as the political principal) to review the compliance and to design the incentive structures (see Persson/Tabellini 1993; Illing 1998 on the consequences of incomplete contracts). In addition, the equivalence of an optimal central bank contract and strict inflation targeting is destroyed. Walsh (1995a) shows as well, for example, that the incentive structures of an optimal central bank contract are not dependent solely on inflation but also on output if central bankers try to maximise their income (Walsh 1995a, pp. 158ff.; see also Walsh 1995c, pp. 239ff.). Svensson (1997) highlights the importance of persistence in the labour market which, among other things, lead to a discretionary monetary policy not only entailing an increased inflation bias but also a stabilisation bias (i.e. employment is too stable and inflation is too variable relative to the exante optimal policy). The reason for this is that with persistence, surprise inflation also leads to real economic effects in subsequent periods. The optimal central bank contract is then statecontingent, and cannot be mimicked solely by a state-contingent inflation target. Optimal inflation targeting now requires the appointment of an independent weight-conservative central banker à la Rogoff (Svensson 1997, p. 107). Assuming a stochastic inflation bias Herrendorf/Lockwood (1996) come to a similar conclusion. They show that, if trade unions have private information about supply shocks prior to setting nominal wages and if the choice of central banker s preferences cannot be made state-contingent, weight-conservatism is part of the optimal institutional design. 10 In addition, Beetsma/Jensen (1997) and Schaling/Hoeberichts/Eijffinger (1998) show that the equivalence between linear inflation contracts and inflation targets breaks down if central banker s uncertain preferences are considered. Despite differences in details both studies on the implications of uncertain preferences support the view that the appointment of a central banker that is more inflation averse than society is part of the best institutional design. 10 In this case, optimal inflation targeting, optimal employment targeting, or signing optimal linear central bank contracts lead to the same results (Herrendorf/Lockwood 1996, p. 16).

16 14 Summing up the theoretical approaches considered here, they show that central bank independence by itself does not improve the social welfare in a discretionary regime. To do this, either the preferences of central bankers must differ from those of the government in an appropriate manner and/or the central bank must be provided with policy targets or incentive structures.

17 15 3 Empirical evidence on central bank independence The empirical substantiation of central bank independence is based on studies which see a "free lunch" in the existence of an independent central bank (Grilli/Masciandaro/Tabellini 1991, p. 375): on average, countries with more independent central banks realise comparatively low inflation rates without real economic costs in terms of lower economic growth or higher output volatility. In fact, many empirical studies on the relationship between central bank independence and inflation confirm a negative correlation between inflation and central bank independence. The derivation of these findings is based on the use of different measures for central bank independence (Section 3.1). For industrial countries, legal indices are mainly used as proxies for central bank independence (Section 3.1.1) whereas for developing countries more behaviourally oriented indicators (Section 3.1.2) are used (Cukierman 1996, p. 15; Eijffinger/De Haan 1996, p. 33). In addition, there appears to be hardly any correlation between central bank independence and long-run growth or output volatility, but some studies come to the conclusion that the disinflation costs are related positively to central bank independence. It is questionable, however, whether the abovementioned empirical findings regarding the relationships between central bank independence and inflation, output and disinflation costs can be seen as stylised facts. More recent studies criticise not only the consistency of the indicators for central bank independence but also the robustness of the statistical associations and as well the causality. 3.1 Measures of independence To examine the links between central bank independence and economic performance verifiable criteria for measuring central bank independence have to be found. Measuring central bank independence is based for industrial countries mainly on the interpretation of central bank laws and therefore concerns legal independence only. The construction of socalled legal indices is done by structuring those criteria which are regarded as being relevant and valuating the compliance with them on a numerical scale. Section is concerned with the indices most frequently used in empirical studies.

18 16 Actual central bank independence can, however, differ in a large extent from legal independence. For example, some central banks in the transition countries are confronted with problems of an adequate implementation of the central bank law and its possibly deviating public assessment (Wagner 1998, pp. 15f.; Radzyner/Riesinger 1997, p. 75). The personal characteristics and abilities of central bankers as well as informal relations between the government and the central bank and subtle methods of political influence may also lead to a deviation of actual from legal independence (Cukierman 1992, pp. 393ff.; Eijffinger/De Haan 1996, p. 22). Furthermore, the extent with which legal independence becomes actual independence may also be characterised by historical experiences which lead to specific "stability cultures" and a general readiness to comply with valid laws or to a respect for laws in general (Radzyner/Riesinger 1997, p. 76; Cukierman 1996, p. 37). In addition, in spite of a wide legal independence from the government there may in fact be political dependence on national parliaments if central bank laws can be amended, abrogated or suspended by simple majorities (see Cukierman 1996, p. 40 on the Central Bank of Russia and Radzyner/Riesinger 1997, pp. 82f. on other examples in Central and Eastern European countries). Along with legal indices of central bank independence additional, behaviourally oriented indices have therefore also been developed, in particular by Cukierman (1992) and Cukierman/Webb/Neyapti (1992). These indices include a questionnaire-based index for identifying deviations of the legal position from actual practice. In addition, the actual turnover rate of central bank governors and the political vulnerability of the central bank governor are used as proxies for the actual independence of central banks, in particular in developing countries Legal indices of central bank independence Legal indices are used as noisy indicators of actual central bank independence. Because the legal foundations of monetary policies in most countries change relatively rarely, legal indices possess practically no explanatory power for the economic development within countries. Most of the empirical studies on the relationships between central bank independence and macroeconomic performance is therefore cross-sectional (Cukierman 1996, p. 8; Eijffinger/De Haan 1996, p. 28). Evaluation of central bank laws is done by calculation of numerical index values. These values are used to rank central banks in accordance with their independence.

19 17 Many econometric studies go beyond a purely ordinal aspect because they use the numerical values of indices as explanatory variables in econometric regressions. It is therefore particularly important to know how the legal measures for central bank independence also differ with regard to their numerical index values. Primarily there are three problem areas which have led to different index values in the literature. The index values are dependent (Eijffinger/Schaling 1993, p. 50; Mangano 1997, p. 6) on (i) (ii) (iii) the criteria contained in the index, the interpretation and evaluation of the law with regard to each individual criterion, the way in which evaluations are aggregated into an overall index, including the weighting of the criteria. These problems make it clear that the construction of indices for measuring central bank independence inevitably contains subjective and arbitrary aspects (Eijffinger/Schaling 1993, p. 51; Mangano 1997, p. 6; Cukierman/Webb/Neyapti 1992, p. 356). In the following, some frequently used indices are introduced and compared with each other. A starting point for further research was without doubt the construction of two indices by Bade/Parkin (1988). The authors differentiate between an index for political independence and one for financial independence. We will be looking below at the index for political independence only because the differentiation between political and financial independence in the definition of financial independence selected by Bade/Parkin has not been followed in the academic literature on this subject (see Appendix A) ). To describe the extent of political independence Bade/Parkin examined the legal provisions of 12 countries for the period with regard to the three following criteria: (+) Is the central bank the final policy authority? (++) Are more than half of the policy-board appointments made independently of the government? (+++) Is there a government official (with or without voting power) on the policy board?

20 18 As Bade/Parkin only permit yes/no decisions, eight (2 3 ) policy types can be distinguished. However, the authors find that just four types occur in reality. Accordingly, the index values run in integers from 1 to 4, where a higher value characterises greater political independence of the respective central bank (Eijffinger/Schaling 1993, p.53; Appendix below as well). As Akhtar (1995) remarks, Bade/Parkin appear to classify an explicit legal specification of monetary policy objectives as being unimportant with regard to the degree of central bank independence: legal specifications of policy goals in central bank laws are not weighted (Akhtar 1995, pp. 432 f.). This procedure stands in contrast to the indices in Grilli/Masciandaro/Tabellini (1991) and Cukierman (1992), which also take into account whether price stability or low inflation are referred to explicitly as goals of monetary policy in the central bank law. In contrast to this, Bade/Parkin regard the question of the exchange rate system for so decisive that they do not analyse the period of the Bretton-Woods system. Alesina (1988, 1989) has extended the Bade/Parkin index by considering additional industrial countries, while Eijffinger/Schaling (1993) place greater emphasis on the significance of policy independence by modifying question (+) in the Bade/Parkin index (see Appendix B)). They take into account, that both the central bank and the government might have some but incomplete policy authority. Grilli/Masciandaro/Tabellini (1991) have developed legal indices which, in comparison to Bade/Parkin, permit a more extensive consideration of details. As we have already stated (see p. 5 of this paper) the authors distinguish between political and economic independence.

21 19 Table 1: The Grilli/Masciandaro/Tabellini index of political independence Political Independence (P1) Governor not appointed by government * (P2) Governor appointed for >5 years * (P3) All the Board not appointed by government * (P4) Board appointed for >5 years * (P5) No mandatory participation of government representative in the Board * (P6) No government approval of monetary policy formulation is required * (P7) Statutory requirements that central bank pursues monetary stability * amongst its goals (P8) Legal provisions that strengthen the central bank s position in conflicts with the government are present * Overall index of political independence, constructed as the sum of the asterisks Table 1 above shows the criteria of political independence of the central bank. The Grilli/Masciandaro/Tabellini index of political independence corresponds to the number of all fulfilled criteria from (P1) to (P8). Criteria (P1) and (P3) assume that central bankers behave independently if they are not appointed directly by the government. Appropriate rules should prevent that stable political majorities lead to political one-sidedness of the monetary policy decision-making body in the course of time. Neumann (1996) emphasises that these criteria can be misleading if the nomination committees are formally but not actually independent of the government (Neumann 1996, p. 5). Long terms of office (P2) and (P4) are aimed at preventing the government from carrying out short-term changes to the composition of the decision-making body. Criteria (P5), (P6) and (P8) refer to the relationship between the government and the central bank with regard to the formulation and implementation of policies. Monetary policy decisions are naturally no longer independent of the government if government representatives have voting rights or if the government must approve monetary policy decisions. However, even when fulfilling these criteria ((P1) to (P6) and (P8)), the political independence of central

22 20 bankers need not coincide with political neutrality. Vaubel (1997) for instance stresses the view that independent central bankers have partisan preferences of their own and may wish to support their preferred parties. According to his party preference hypothesis an independent central bank tries to improve the government s electoral chances if the majority in the central bank council shares the government s partisan views and it tries to prevent the incumbent politicians re-election if the opposition parties command the majority on the central bank s board (see the discussion between Vaubel (1997) and Berger/Woitek (1997) on the question, whether the German Bundesbank has partisan preferences). Criterion (P7) has to be seen against the background of the definition from Grilli/Masciandaro/Tabellini of political independence as being the ability of the central bank "(..) to pursue the goal of low inflation" (Grilli/Masciandaro/Tabellini 1991, p. 367). With this definition the authors cover more than the independence of the central bank from the government. It is true that a legal mandate for monetary stability is supposed to reduce the possibility of short-run political considerations leading to changes in the objectives of monetary policy. But at the same time this restricts the discretionary power of the central bankers with regard to the choice of the monetary policy goals. It cannot be concluded from the concrete formulation of criterion (P7) chosen by Grilli/Masciandaro/Tabellini that price stability must be the sole goal of monetary policy, but no weight is placed on other possible goals (see also Akhtar 1995). The ability of the central bank to choose its own monetary policy instruments without interference from the government is measured by Grilli/Masciandaro/Tabellini with the help of the following criteria, which form the index of economic independence.

23 21 Table 2 The Grilli/Masciandaro/Tabellini index of economic independence Economic Independence (E1) Direct credit facility: not automatic * (E2) Direct credit facility: market interest rate * (E3) Direct credit facility: temporary * (E4) Direct credit facility: limited amount * (E5) Central bank does not participate in primary market for public debt * (E6) Discount rate set by central bank * (E7) Banking supervision not entrusted to the central bank (**) or not entrusted to the central bank alone (*) (**) (*) Overall index of economic independence (being the sum of the asterisks) Criteria (E1) to (E4) refer to characteristics of the direct central bank credit to the government. Under the aspect of preventing a direct financing of budget deficits, in particular close limits (E4) on credit facilities or a complete prohibition of direct central bank credits to the government appear to be effective. Criterion (E5) refers to the purchase of government bonds in the primary market, but not to sale and purchase in the secondary market. Therefore, even if (E5) is fulfilled the central bank could finance budget deficits (indirectly) through purchases of government bonds in the secondary market. Criteria (E6) and (E7) concern the independence of monetary policy from an administrative fixing of the discount rate (E6) and from influences which can arise from bank supervision (E7) (for a criticism of the economic criteria see for example Neumann 1996, pp. 5ff.; Eijffinger/De Haan 1996, pp. 57f.). As with other indices (e.g. those from Cukierman, Cukierman/Webb/Neyapti, Alesina or Eijffinger/Schaling), the way central bank independence is influenced by the exchange rate system is not taken into account explicitly (an exception is an extension of the Cukierman/Webb/Neyapti indicator by Siklos (1995) for five countries in Latin America). This procedure can probably be justified by the fact that in nearly all countries the government is responsible for the choice of the exchange rate regime

24 22 (as well as for determining exchange rate parities or upper and lower limits in systems of fixed exchange rates) wheras the central bank normally has to manage the chosen exchange rate system. The total index for central bank independence results from the sum of the index values for political and economic independence. The most detailed indices for measuring the legal independence of central banks have up to now been developed by Cukierman (1992) and Cukierman/Webb/Neyapti (1992). Indices LVAU (unweighted index of legal independence) and LVAW (weighted index of legal independence; see Appendix C)) take into account 16 characteristics for central bank independence which are divided into 4 clusters. LVAU and LVAW differ from each other through their different weighting of some components. The main groups cover characteristics of the personal independence of the chief executive officer (CEO) (term of office, appointment, reasons for dismissal), aspects of the central bank's policy independence (policy formulation, resolution of conflicts - PF), the final objectives (OBJ) laid down in the central bank statute and the legal restrictions on the ability of the government to borrow from the central bank (LL) (see Cukierman 1992, p. 372). In contrast to Grilli/Masciandaro/Tabellini (1991), graded evaluations are provided for with detail questions as well. For example, there are six possible ratings for the question concerning the objective of the central bank, according to the weight given to price stability compared with other stated objectives. In concurrence with Grilli/Masciandaro/Tabellini (1991), here too a statutory mandate to pursue the goal of price stability is seen against the background that delegating monetary policy to an independent central bank is aimed at mitigating the credibility problem (Grilli/Masciandaro/Tabellini 1991, p. 367). "In Rogoff s (1985) terminology, the objectives variable measures the strength of the 'conservative bias' of the bank s charter" (Cukierman/Webb/Neyapti 1992, p. 357). The following table is taken from Eijffinger/De Haan (1996, p. 23) and shows the numerical index values for 22 industrial countries.