Inflation culture, central bank independence and price stability

Transcription

1 European Journal of Political Economy Ž. Vol Inflation culture, central bank independence and price stability Bernd Hayo ) Department of Economics, UniÕersity of Bamberg, Feldkirchenstr. 21, Bamberg, Germany Revised 1 January 1997; accepted 1 May 1997 Abstract This paper proposes that central bank independence alone is insufficient to explain low inflation. It argues instead that central bank independence may be interconnected with public attitudes towards inflation via a historical feedback process that has led to an anti-inflation culture and public consensus on monetary stability in countries with low inflation rates. Using survey data from European Community members over the period , I find supportive evidence of the existence of a stability culture in low-inflation countries. In their assessment of the importance of price stability, people living in countries with a low inflation record appear to be more sensitive to an increase in the actual inflation rate than people in higher-inflation countries. q 1998 Elsevier Science B.V. All rights reserved. JEL classification: E42; E31 Keywords: Inflation; Inflation culture; Central bank independence; European Community 1. Introduction Price stability is a central issue in the discussion of European Monetary Union Ž. EMU. Many of the studies in the literature on EMU use an institutional or ) Corresponding author at present address: Rheinische Friedrich - Wilhelms - Universitat Bonn, ZEI, University of Bonn, Walter-Flex-Str. 3, D Bonn, Germany, Tel.: q ; fax: q ; hayo@sfbb4.econ1.uni-bonn.de r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved. Ž. PII S

2 242 ( ) B. HayorEuropean Journal of Political Economy game-theoretic framework to analyse the optimal design of a European Central Ž. 1 Bank ECB that would ensure stable prices. The German Bundesbank Law is often cited as an example for the design of a ECB and has influenced the relevant articles Ž Article 105 and following. in the Treaty of Maastricht: the Treaty postulates an institution which is at least legally as independent as the Bundesbank Ž see Eichengreen, Such independence is often considered as sufficient to overcome problems of time-inconsistency and seigniorage Ž see Fischer, 1995., although McCallum Ž argues that time-inconsistency and optimal contracts are not the most important issues, but rather the independence of the central bank from everyday political pressures. In this paper, I will refer to this selection of arguments, centering on central bank independence for achieving stable prices, as the orthodox view. The theoretical literature on the orthodox view is supported by empirical studies. Alesina Ž and Grilli et al. Ž investigate the causal relation between the degree of central bank independence and the rate of inflation Žalso see Alesina and Grilli, 1992., and observe that the greater the independence of a country s central bank, the lower its inflation rate. Similar results are obtained by De Haan and Sturm Ž Yet although this orthodox view prevails within mainstream macroeconomics, there remain some doubts in the profession that this is the whole story. 2 One strand of thinking within the orthodox view questions the exogeneity of central bank independence. As suggested for instance by Cukierman Ž 1994b., there may exist political and economic factors influencing the degree of legal independence granted to the central bank. The main trade-off here is between credibility and flexibility of monetary policy for the incumbent governing party. A lack of credibility results in expectations of rising prices, causing an inflationary bias in the economy. The inflationary bias leads economic agents to adjust their behaviour accordingly, with the outcome being higher interest rates and wage increases. Granting more independence to the central bank raises credibility, but reduces the discretionary use of monetary policy by the government, and vice versa. Two hypotheses relating to the inflationary bias are tested by De Haan and van t Hag Ž using cross-section data from 19 countries. First, they investigate the relationship between central bank independence as dependent variables and proxies for the inflationary biases as independent variables. Second, they test the hypothesis that governments planning to incur higher deficits may wish to increase credibility, to reduce the interest premium associated with expectations of high future inflation rates. No supporting evidence is found for either of the two 1 See, e.g., Bean Ž 1992., Cukierman Ž 1994a., Treutler Ž 1993., von Hagen and Suppel Ž Obviously, I build up some kind of straw-man for a better focus of the discussion. As everyday experience shows, most economists are not that fundamentally oriented anyway.

3 ( ) B. HayorEuropean Journal of Political Economy hypotheses relating central bank independence and inflationary bias. This suggests that the story of Cukierman Ž 1994b. may not suffice to explain why some countries have independent central banks and low inflation rates, and others do not. A different approach, set out for example by Posen Ž 1993., questions the credibility of the mainstream answer as to why central bank independence causes lower inflation rates. His view is that political interest groups have specific inflation preferences, in particular the interest of the financial sector lies in keeping inflation low. Posen thus proposes that it is inappropriate to focus on institutional design in the creation of independent central banks, while neglecting the political self-interest calculations. Hence the political-interest group hypothesis is a challenge to the economic orthodox view. But there clearly exist problems in identifying agents preferences. Does the financial sector indeed benefit from low inflation rates? There are of course the usual collective action problems as well. Changes in the structure of the financial sector should influence its effectiveness as a lobbying group, which should then be revealed by fluctuations in inflation rates. Based on some empirical tests, De Haan and van t Hag Ž also raise doubts about the generality of Posen s claims. Another approach, associated with political business cycles, suggests that central banks, even if ostensibly independent, can be captured. Vaubel Ž proposes that since overseers of central bank policy are political appointments, central bank policy may be politicised. Vaubel presents evidence with regard to the Bundesbank, which is critically evaluated by Berger and Woitek Ž Sieg Ž formulates a formal model of such opportunistic behaviour by a captured independent central bank. The approach which I wish to consider here takes as its point of departure a still different literature which points to a deep-rooted aversion to inflation in the population Ž see, for example, Bofinger, 1992 or Treutler, It is argued, for example, that the experience of two hyperinflations during the 20th century have created a strong fear of inflation in the German population. The corresponding hypothesis is that people s preferences with respect to price stability matter at least in Germany in explaining low inflation rates. Collins and Giavazzi Ž try to identify different inflation preferences Žin the form of attitudes. in Europe. Using time-series data of inflation and unemployment expectations to explain the public s assessment of the overall economic situation, they attempt to measure the corresponding weights of a widely used economic loss function Ž see, e.g., Barro and Gordon, They claim that in the 1980s a structural change occurred in the weights, towards increased preferences for lower inflation. Moreover, they propose that their results indicate convergence of public attitudes on inflation in Europe over time. Their approach and findings are however controversial and not without critics Žsee Fratianni, 1993; Thygesen, In this paper, I maintain only some parts of the orthodox framework, but

4 244 ( ) B. HayorEuropean Journal of Political Economy emphasise instead the role of a consensus on attitudes towards inflation. 3 The alternative hypothesis put forward is based on the idea of a specific inflation culture in the respective societies. This will allow for a somewhat different and, arguably, more convincing reading of the various inflation histories of Ž European. countries. The formal-institutional design of a central bank is thus only one aspect of a stability regime, to be supplemented with public attitudes towards the goal of price stability in explaining inflation differentials among countries, and public perception that price stability is important becomes a significant precondition for achieving a low inflation rate. The empirical literature on central bank independence cannot be used to contradict this position, since it cannot by itself answer the crucial question of causality Žsee also Forder Ž That is, is the prime reason for low inflation an independent central bank? Or is it the case that countries with a population which places a high value on price stability have designed their central banks accordingly? With regard to the EMU, the implication is that ECB independence is not enough to achieve low inflation rate in Europe. A public consensus on the need for price stability is required to make the formal independence of a central bank effective. In the absence of such a consensus, interest groups may influence monetary policy or decision makers may interpret fixed objectives as soft constraints Ž e.g., when is price stability endangered?.. 4 My principal aim is to cast doubt on the adequacy of the orthodox model as a sufficient interpretation of recent monetary history. 5 The data sources are European public opinion polls on how people in the different countries of the European Community Ž EC. value price stability Žannual observations from 1976 to 1993 taken from the Eurobarometer surveys.. In Section 2 I consider the relationship between the orthodox and my model, and present possible transmission mechanisms from preferences to monetary policy. Section 3 explains the research design, provides a short summary of the data, and compares the importance of inflation in the members of the EC. Section 4 investigates the relation between the inflation goal and actual inflation rates using alternative estimators. Section 5 discusses the 3 For convenience, preferences and attitudes are used interchangeably, though in principle, I regard attitudes as empirically measurable proxies of preferences. 4 From a different perspective, the importance of preferences towards inflation is stressed by Morales and Padilla Ž Based on their theoretical model, they maintain that a majority voting procedure in the governing council of a ECB leads only to an efficient outcome, if the members have similar preferences on monetary policy. If one assumes, as is usually done, that the members of the governing council reflect the public view towards inflation that prevails in their home countries, then looking at the preferences of people in Europe might be important in this respect too. 5 For the sake of simplicity, I do not take into account that inflation is not just a monetary phenomenon and that it is influenced by other factors as well. I would maintain, though, that monetary factors are generally the most important ones, at least in the longer run.

5 ( ) B. HayorEuropean Journal of Political Economy relation between my results and indicators of central bank independence. Section 6 provides a summary and suggests directions for further research. 2. Price stability, inflation culture and central bank independence The attraction of the orthodox model lies in its combination of rigorous theoretical foundations and empirical support, and that there is no other equally coherent theory to explain diverging inflation records in the literature. I will transform the orthodox view by placing public attitudes towards inflation at the centre of the argument. The rationale for my public-preference explanation is based on how central bank independence and public attitudes towards inflation fit together. A straightforward interpretation regards the independence of a central bank as a means to achieve the goal of low inflation rates as embodied in people s preferences. This type of interpretation seems to be, at least implicitly, present in many specifications of public objectives functions in the literature Že.g., Barro and Gordon, 1983 or Cukierman, I refer to this as the preference-instrument interpretation, and it can be regarded as being built on the literature on economic policy in a democracy and political business cycles. 6 Two main submodels based on the preference-instrument view can be identified. In an election-oriented scenario, we assume that people have for whatever reason a preference for low inflation. The voting process allows people to express their preferences for low inflation, either directly or indirectly. In the former case, voters would directly elect the central bank governor Žsee Downs, In the latter case they elect a government with low inflation policies Ž political parties as described by the partisan view, see Alesina, 1988., and the government appoints a conservative central banker who chooses an independent Ž. 7 anti-inflation monetary policy as envisaged by Rogoff 1985 and others. A constitutional-oriented approach emphasises the institutional setting of the central bank. We can assume a contract-based scenario, or a more evolutionary one. In the contract-based scenario, we can imagine people deciding on the creation of a central bank while being in, following the phrase used by Rawls Ž 1971., the original position. If the public s preferences favour low inflation, people will choose an independent central bank removed from direct political pressures, like demands for financing higher government expenditures via increasing the money supply. In the evolutionary constitutional scenario, people also value the achievement of price stability from the beginning, but because of difficulties in understanding complexities of inflation, it takes some time for 6 Ž. Ž. A recent introductory summary of this literature is given in Keech 1995 ; also see Frey Preferences of voters may determine the weights of an economic policy welfare function as Ž. assumed in most of the models in the tradition of Barro and Gordon 1983.

6 246 ( ) B. HayorEuropean Journal of Political Economy people to get to know the usefulness of different instruments. From the alternative means tried over time, an independent central bank evolves as the best at least for the moment way of keeping inflation low. The question is whether the preference-instrument interpretation based on one of the above transmission mechanisms is a satisfactory explanation of preferences for low inflation. An important problem with the preference-instrument interpretation is its clear-cut causal chain from preferences to outcomes. But it is by no means obvious why preferences towards price stability should exist autonomously, especially since even professional economists do not fully understand the phenomenon and consequences of inflation. It seems unlikely that people have inflation genes that determine their preferences. Preferences about price stability are rather part of the economic culture of a country, which is influenced by historical experience. Economic culture is defined here as the values and attitudes of a population on all aspects related to the economic system. 8 Consequently, the sub-concept of an inflation culture attempts to capture the values and attitudes people have with respect to the economic phenomenon of inflation. For example, in the case of Germany it is often argued that people s attitudes are affected by experience with high inflation. De Haan and van t Hag Ž attempt to capture this by showing that countries with high inflation rates during the period are more likely to create an independent central bank later on. Aversion to inflation may have its origins in past times of hyperinflation. However, historical circumstances in themselves are not enough. The collective economic history of a population must be translated into an economic culture. Indeed, the historical evidence, even for Germany, indicates that central bank independence does not reduce inflation as long as there exists no public consensus on the primacy of this policy objective. 9 Therefore, I propose a different integration of the orthodox model and people s preferences, a transmission mechanism which allows for a feedback relationship between the institutional framework of the central bank and attitudes towards inflation. A historical-feedback interpretation assumes a world of change, where neither preferences nor institutional structures are fixed, and the two elements can reinforce one another. An explanation for the creation of independent central banks based on such an historical-feedback interpretation runs along the following lines. From their everyday experience, people perceive that inflation is bad, for 8 This reflects the seminal definition by Almond and Verba Ž for the concept of a political culture. 9 For example, in Germany, the directorate of the Reichsbank was given independence through the Autonomiegesetz from 26 May 1922 Ž see, e.g., p. 160 and following of Pfleiderer, However, hyperinflation increased further since the directorate considered the monetization of national budget deficits as a necessity. Interestingly, the independence was granted due to pressure from the reparation commission, which acted in the interest of the victorious countries of World War I. Thus it cannot be understood as a reform law based on a public consensus in Germany.

7 ( ) B. HayorEuropean Journal of Political Economy instance by having to adjust their spending behaviour to ever rising prices. They think of ways to achieve more stable prices in a complex world, under conditions where economic policy is subject to short-run political pressures Žcf. the interestgroup argument of Posen, In view of the interest groups, society may wish to tie its own hands with respect to short-run policy considerations. A solution is to create an independent central bank committed to stable prices. Although such an institution may achieve a somewhat lower rate of inflation Žand here is the orthodox aspect., it will not be able to undertake a firm monetary policy without public support. Over time, people begin to give credit to the independent central bank for achieving lower inflation, especially if real costs of a stable anti-inflationary monetary policy such as lower economic growth or more unemployment are low on average: Grilli et al. Ž 1991., for example, show that countries with a relatively low inflation rate do not seem to experience corresponding output losses, and Alesina and Summers Ž conclude that there exists no obvious, direct relation between central bank independence and income growth, unemployment or interest rates. Such outcomes strengthen people s beliefs that fighting inflation is worthwhile, which again feeds back to increase support for an independent central bank. Support for the independent central bank as an institution increases policy effectiveness and the inflation rate can be further stabilised, and so forth. The outcome is a consensus on price stability within the inflation culture of a country. 10 As long as there are no exogenous shocks, society will support the anti-inflationary policies of an independent central bank, even if this harms some people in some circumstances. 11 Seen within this framework, an independent central bank need not be viewed as an undemocratic institution Žcf. p and following of Cukierman, 1994a., even if voters have no direct control over monetary policy. However, can a stability consensus be achieved? In p. 422 and following, Cukierman Ž presents, for instance, a finding that the empirical relationship between central bank independence and inflation is much weaker in developing countries. He interprets this as due to a high turnover-rate of central bank governors, reflecting a weakening of legal central bank independence by high actual inflation. I interpret the empirical relationship differently, namely that a consensus within society on the importance of fighting inflation has not Ž yet. been achieved in these countries. The performance of independent central banks has not been impressive enough for people to support stabilising monetary policies, and, due to a number of rigidities, monetary stabilisation in these countries induces 10 This type of mechanism offers connections to the literature on political culture and constitutional economics. As an example of the former, see the influential work by Lipset Ž 1988., where he attempted to divide political support into aspects of legitimacy and efficacy. For the constitutional approach to economics, see Buchanan Ž This does not, of course, exclude the possibility that everyday politics influences such an institution. Again we can point to the view presented by Posen Ž

8 248 ( ) B. HayorEuropean Journal of Political Economy high real costs, at least in the short-run, which works against the stability consensus. 12 Some of the countries in the empirical analysis which follows have more or less tied their currencies to the German Mark, notably the Netherlands. One may argue therefore that the Netherlands imports price stability via the exchange rate and not through central bank independence. This argument does not constitute a problem for my hypothesis. The exchange rate link Ž external anchor. is just another way to foster price stability. But it is only sustainable because it is supported by a majority of people in the Netherlands who believe that fighting inflation is important. To conclude, even if one is not persuaded that attitudes towards inflation and central bank independence are interrelated, it is apparent that we know less about these issues than the economists talking solely of time-consistency problems and central bank reputation would have us believe. What needs to be stressed is not that these issues are generally irrelevant for understanding inflation, but rather that they cover only some parts of the story, and perhaps not the most important parts for explaining differences in inflation rates across countries. Section 3 is concerned with the construction of an empirical indicator for the importance people in the EC member states place upon the achievement of low inflation. 3. Database and descriptive analysis The database consists of public opinion polls Ž Eurobarometers. from 1976 to 1993 in nine countries of the EC. For Greece we have observations since 1980 and for Spain and Portugal from 1985 onwards. From the often large surveys, we focus on the so-called Inglehart Index, which attempts to measure the change in people s values. Four goals of a state are presented to the respondents: Ž. 1 Maintain order, Ž. 2 More democracy, Ž. 3 Fighting rising prices, and Ž. 4 Protecting freedom of speech. 13 The respondents are asked to name the goal which they consider to be the most important for their country in the long-run. In another question, they choose the goal which they consider to be of second priority. The goal of inflation is measured relative to the other objectives. One could argue that this is a disadvantage, since we are interested in an absolute indicator of 12 Cukierman Ž does not propose simple causal explanations like my sketch of the orthodox view. On the other hand, in his account of central bank independence on p. 445, a mechanism as outlined above is not discussed. However, towards the end of his book Ž p. 455., he mentions the consensus within a society as a possible influence on successful stabilisation programmes in countries such as Argentina or Brazil. 13 A detailed discussion of the two positions, materialism and postmaterialism is given, e.g., in Inglehart Ž The items Ž. 1 and Ž. 3 should capture materialist thinking and Ž. 2 and Ž. 4 postmaterialist ideas.

9 ( ) B. HayorEuropean Journal of Political Economy importance of the inflation goal. But it is unlikely that a question like how important do you consider the achievement of price stability? will yield more insight. In answering this type of question, people experience no sanctions; it is a soft question. Although the Inglehart Index does not really use hard questions, it entails at least a trade-off. People have to decide on a specific item. If they choose inflation, it is at the expense of other goals. Importance can a priori only be measured on a relative scale, and the Inglehart Index sets out the underlying scale. One potential problem with this framework is that an abnormal situation in a country might bias people s decisions. Since we cannot allow for special circumstances, apart from actual inflation, we confront possible measurement errors. An example is Northern Ireland, where people put a much stronger emphasis on maintaining order than in other countries. It is clear that the political situation in this region cannot be called normal and I have excluded observations from Northern Ireland from the analysis. 14 We also cannot be sure whether people, when naming the fight against inflation as the first priority, express their aversion to rising prices or just their dislike of decreasing real incomes. As argued by Cukierman Ž on p. 22, it may be the case that historical experiences of high inflation only lead to institutional reforms if they were accompanied by large losses in real income. The data at hand do not permit us to disentangle these conceptually different types of motivations. Another problem is the selection of alternative goals given in the question. Maybe these are not the most significant goals. Then, even if people say that the fight against inflation should be the first priority, it may not be worth much, since the importance of the inflation goal is measured against low-valued alternatives. For instance, it would have been interesting to have unemployment as an alternative item. However, Hibbs Ž and Fischer and Huizinga Ž report that inflation is perceived by the U.S. population as a more important issue than unemployment. I maintain that the above goals are useful and any bias introduced through this channel is likely to be small. In the following sections I refer, for reasons of brevity, to the importance of the inflation goal, but what is actually meant is the relative importance of fighting rising prices versus the other objectives. I will look at two types of variables. One is the share of people who consider price stability as the most prominent long-term goal. The other includes the share of respondents who either believe fighting inflation is the first or second most important objective. 15 The second variable might give a more stable measure over time, since it is less influenced by current events than the first. 14 In any case, including the appropriately weighted observations on Northern Ireland to the ones used here for Great Britain does not change the country average very much. 15 See Tables A1 and A2 in Appendix A for a summary of the shares of the inflation goal in public opinion.

10 250 ( ) B. HayorEuropean Journal of Political Economy Ž. Fig. 1. Development of share of people in % who consider the fight against inflation as the most important goal of the state.

11 ( ) B. HayorEuropean Journal of Political Economy Table 1 Mean Ž in %. and standard deviation of the share of people who consider the fight against inflation as the most important goal Ž , Greece Ž , Spain and Portugal Ž Fra Bel Net Ger Ita Lux Den Ire GB Gre Spa Por Mean Standard deviation Variables containing only the fight against inflation as a first priority are marked with an index 1, e.g., Fra1 stands for the share of people in France who consider keeping prices stable as the most important long-term goal of the state. Correspondingly, variables including people who say the inflation goal is their first or second priority, are indexed by 12, e.g., Ger12 denotes the share of the public in Germany which believes that anti-inflation policy should be either the first or the second most important objective. In Fig. 1, the indicator for the importance of inflation in the EC countries is plotted over time. Nine graphs Ž Greece and Spain excluded. suggest a negative trend. In Greece and Spain the series looks stationary. A similar structural picture emerges if we look at the variables containing first and second priorities. To describe the series in somewhat more detail, we find in Table 1 the mean and standard deviations for the top priority variables. On average, about 25% of the people in these European countries Žwithout taking account of the relative population size. considered the fight against rising prices to be the most important goal of the state. The inflation objective plays a relatively prominent role in Belgium, Portugal and Ireland, but is relatively unimportant in Denmark, the Netherlands and Germany. Table 2 gives the corresponding numbers for the variables containing first or second priority. Here the simple mean across the EC countries is about 57%, i.e., more than one half of Europeans believe that the fight against rising prices should be the first or second priority of the state. The highest proportion is found in Portugal, followed by Greece and Belgium. Once again in Denmark, the Netherlands and Germany less importance is placed on this objective. These results do not indicate that the absolute shares reflect the actual inflation performance of the countries. It rather seems to be the case that the weight given Table 2 Mean Ž in %. and standard deviation of the share of people who consider the fight against inflation as the first or second priority Ž , Greece Ž , Spain and Portugal Ž Fra12 Bel12 Net12 Ger12 Ita12 Lux12 Den12 Ire12 GB12 Gre12 Spa12 Por12 Mean Standard deviation

12 252 ( ) B. HayorEuropean Journal of Political Economy Table 3 Mean and standard deviation of inflation rates in the EC Ž , Greece Ž , Spain and Portugal Ž a FraIn BelIn NetIn GerIn ItaIn LuxIn DenIn IreIn GBIn GreIn SpaIn PorIn Mean Standard deviation Source: OECD Economic Outlook Ž and OECD Economic Outlook Ž a The numbers given for mean and standard deviation are based on an arithmetic mean. The use of the Ž correct. geometric mean would not change any of the following results to the goal of stable prices is relatively small in countries with a low inflation rate and vice versa. To make this point, in Table 3 the inflation averages of the EC countries are recorded. 16 One can show the actual numbers are omitted here that there exists a significant positive correlation between the importance of the goal of price stability and actual inflation performance. Since it does not make any sense to postulate a theoretical relation where a higher importance of the inflation goal leads to higher inflation, it is presumably actual inflation that drives public opinion on this issue. Hence in Section 4 we investigate the influence of actual inflation on the two importance variables. 4. Importance of the inflation goal and actual inflation It is interesting to assess the effect of actual inflation on the public opinion variables. We would like to obtain a fundamental indicator of priorities, independent of everyday changes in inflation. To achieve this, we regress the importance variables on actual inflation. The basic estimated equations are Eqs. Ž 1a. and Ž 1b.: Importance1its a1q b1 Inflationitq u1it Ž 1a. Importance12 its a12q b12 Inflationitq u12it Ž 1b. where: isfrance,..., Portugal; Importance1itsshare of people who consider the fight against inflation as the first priority of the state; Importance12its share of people who consider the fight against inflation as the first or second priority of the state; u1i and u12is white noise error terms. 16 The numbers given for mean and standard deviation are based on an arithmetic mean. The use of the Ž correct. geometric mean would not change any of the following results.

13 ( ) B. HayorEuropean Journal of Political Economy Two interpretations of the regressions can be considered: First, we could argue that the intercepts a1 and a12 capture the fundamental evaluation of the inflation goal. If this explanation were correct, the orthodox view would imply that we should not find any pattern in the data with respect to the estimators for the constant term. My hypothesis would argue that the countries with a low inflation rate should instead have a larger constant term than the high-inflation countries. Second, we could concentrate on the sensitivity of the importance variables with respect to actual inflation by looking at the estimates of the slope parameters b1 and b 12. The orthodox view would maintain that there should not be any significant differences in the parameters, whereas my hypothesis predicts higher parameter estimates for the low inflation countries. Table 4 summarises the estimates of the importance variables. The t-tests indicate highly significant parameter estimates. For both importance variables, the results are very similar. A glance at the intercept estimators makes clear that countries with a high constant term are those countries exhibiting high inflation. Neither the orthodox nor my hypothesis are supported by this measure. The use of the ˆ a1 and ˆ a12 parameters does not appear to be successful. With respect to the slope parameter, we do find an interesting relation: Countries with a relatively low inflation rate tend to have a higher slope parameter estimate. This result contrasts with a naive version of the orthodox view and corresponds to what one would have expected under my alternative hypothesis. In other words, countries where the population reacts rather sensitively to an increase in the inflation rate are performing better in keeping prices stable. This result points to a stability consensus within the economic culture of low-inflation countries. Table 4 OLS-regression of first priority on inflation and first and second priority on inflation Ž , Greece Ž , Spain and Portugal Ž State ˆ a bˆ R DW State ˆ a bˆ R DW Fra 18.4)) 1.81)) )) Fra 46.7)) 2.05)) )) Bel 30.8)) 2.09)) )) Bel 59.2)) 2.06)) )) Net 8.3)) 2.09)) ) Net 27.0)) 3.59)) )) Ger 8.7) 3.28)) )) Ger 28.8) 5.81)) )) Ita 16.2)) 0.62)) Ita 45.0)) 1.28)) ) Lux 14.7)) 1.78) )) Lux 39.7)) 2.84)) )) Den 4.9)) 1.30)) Den 17.2)) 3.4)) Ire 27.4)) 0.82)) ) Ire 59.4)) 0.92)) )) GB 11.9)) 1.47)) GB 36.5)) 1.82)) Gre 25.5)) Gre 61.8)) Spa1 19.6)) Spa )) Por1 27.6)) 0.92) )) Por )) )Rejection of the null hypothesis at the 5% significance level. ))Rejection of the null hypothesis at the 1% significance level.

14 254 ( ) B. HayorEuropean Journal of Political Economy The finding that all of the slope coefficients are positive could also be regarded as in accord with the often-found quadratic specification of the public objective function within the framework of the orthodox view Že.g., Barro and Gordon, 1983 or Cukierman, The implication is that a society s aversion to inflation rises with the inflation rate. However, typically preferences are assumed to be given in these models and as has been outlined above this cannot be considered a satisfactory explanation for the existence of differing inflation rates and the creation of independent central banks across countries. To analyse the robustness of the estimates, diagnostic tests were performed, namely the Durbin Watson test Ž DW., an LM-test for autocorrelation of first and second order, an LM-Test for ARCH, a test for normality, the White-test for heteroskedasticity, and the RESET-test. The DW test statistics are given in the table, while most of the other tests, which do not indicate a serious problem, have been omitted. However, the DW statistics point at first-order autocorrelation for a number of regressions, which, even if OLS continues to be unbiased, invalidates the tests of significance. To account for that problem, the same regressions were reestimated allowing for an autoregressive error term Ž RALS.. The error term in Eqs. Ž 1a. and Ž 1b. is now assumed to be of the form u1itsd1u1it 1q 1t and u12itsd12 u12it 1q 12t, with 1t and 12t being white-noise errors. The results are displayed in Table 5. It is clear that the estimates with low DW-statistics in the OLS-regression have a significant autoregressive error term. Substantial changes in the slope parameter estimates can be found, especially for Belgium and Luxembourg and in the intercept for Belgium and Denmark. In general, the RALS-estimators of the Table 5 RALS-regression of first priority on inflation and first and second priority on inflation Ž , Greece Ž , Spain and Portugal Ž State ˆ a bˆ ˆ d Pseudo R State ˆ a bˆ ˆ d Pseudo R Fra 19.2)) 1.49)) 0.52) 0.83 Fra 47.4)) 1.80)) 0.46) 0.84 Bel y )) 0.78 Bel y )) 0.82 Net 9.3)) 1.3) 0.70)) 0.81 Net ) 0.90) 0.86 Ger ) 0.90)) 0.89 Ger ) 0.89)) 0.91 Ita 16.3)) 0.59)) Ita 47.4)) 0.92) 0.72)) 0.80 Lux )) 0.79 Lux 32.9 y )) 0.84 Den 5.0)) 1.27)) Den y )) 0.88 Ire 27.8)) )) 0.61 Ire )) 0.72 GB 11.4)) 1.53)) GB 35.9)) 1.93)) y Gre 27.5)) 0.21 y Gre 64.4)) Spa 20.3)) y0.11 y Spa 54.3)) y Por 27.6) ) 0.72 Por 73.6)) y )Rejection of the null hypothesis at the 5% significance level. ))Rejection of the null hypothesis at the 1% significance level.

15 ( ) B. HayorEuropean Journal of Political Economy Table 6 SURE-regression of first priority on inflation and first and second priority on inflation Ž State ˆ a bˆ s State ˆ a bˆ s 1 12 Fra 20.4)) 1.51)) 5.31 Fra 48.6)) 1.79)) 5.08 Bel 31.4)) 1.96)) 5.91 Bel 62.2)) 1.43)) 6.05 Net 8.5)) 2.03)) 2.82 Net 28.3)) 3.22)) 5.99 Ger 8.3)) 3.4)) 6.53 Ger 29.1)) 5.71)) 9.77 Ita 16.2)) 0.62)) 2.59 Ita 48.1)) 0.99)) 4.99 Den 4.6)) 1.33)) 2.12 Den 20.4)) 2.93)) 5.86 Ire 28.1)) 0.74)) 5.07 Ire 62.5)) 0.56)) 6.21 GB1 11.9)) 1.46)) 2.90 GB )) 1.81)) 3.99 log-like y log-like y VecARŽ VecARŽ VecNorm 19.6 VecNorm 20.9 )Rejection of the null hypothesis at the 5% significance level. ))Rejection of the null hypothesis at the 1% significance level. s is the standard error of the respective equation, log-like the value of the log-likelihood function for the system. sensitivity parameters are lower than their OLS-counterparts. Therefore, the force of the OLS-results seems to be somewhat reduced but not necessarily contradicted. We cannot be sure that by using RALS we have treated the cause of the autocorrelation. Other possible reasons, like omitted variables or a lacking dynamic adjustment, may be important too. Since the sample size is very small, there is no scope to estimate a dynamic model with any precision. Accordingly, preliminary tests with first-order lags indicate that in some cases the autocorrelation cannot be fully removed from the residuals. An analysis using unit-root tests and graphical inspection of the spectrum points towards possible non-stationarity in the series. Since the sample size is much too low to distinguish between the processes, no detailed analysis will be presented. 17 Repeating the regressions in Eqs. Ž 1a. and Ž 1b. for the variables in first-differences leads to similar results as in the RALS-estimation. Only the coefficients for the Netherlands, Germany, and Great Britain turn out to be significant. One should note that the interpretation of the regression in differences would entail a reaction of the change in inflation attitudes caused by a change in actual inflation. However, Hibbs Ž on p. 725 presents results showing that the respondents look at the level of inflation when forming their attitudes, and not at the changes in the growth rate of prices. The correlations between actual inflation rates and 17 See, for example, the survey by Campbell and Perron Ž and especially the comments by Cochrane Ž and Miron Ž

16 256 ( ) B. HayorEuropean Journal of Political Economy estimated parameters from the equations in differences are somewhat higher than the ones computed for the RALS-regressions Ž given later in Table Further analysis of the OLS-residuals reveals contemporaneous correlation among the residuals from the different equations. To enhance the efficiency of the estimates, the SUR-estimator has been employed Žsee, e.g., p. 363 and following of Darnell, In Table 6, the SUR-estimates for only eight countries are given to avoid degrees-of-freedom problems. The resulting parameter estimates are different but still relatively close to the OLS-estimates. While we do not discover autocorrelation ŽVecARŽ or non-normality Ž VecNorm. for the vector of residuals, we find evidence of first-order autocorrelation for all of the single equations Ž not given in the table., and thus are still confronted with a possible bias in the variance estimators. 5. Evaluating the results with respect to central bank independence To summarise the discussion of the previous sections, I will demonstrate how the estimated indicators relate to the indices of central bank independence used in the literature. The utilised indicators for central bank independence have been proposed by the following: Bade, R. and Parkin, M., 1982 in Central Bank Laws and Monetary Policy Ž unpublished., Grilli et al. Ž on p. 366, and Alesina and Summers Ž on p As is described by Cukierman Ž on p. 422, his index for legal independence is very similar to these indicators. Therefore, we do not explicitly consider Cukierman s index in the following analysis. While Bade and Parkin and Alesina and Summers try to construct an overall indicator of central bank independence, Grilli et al. distinguish between political and economic independence. Political independence Ž PolIn. is defined as the autonomy of the central bank to pursue a policy of price stability without being influenced by the government. Economic independence Ž EcoIn. reflects the ability of the central bank to select its monetary policy instruments. The index of Alesina and Summers Ž AleSu. is a combination of Bade and Parkin s Ž Bade. and Grilli et al. s index. In Table 7, correlation coefficients between the estimated indicators for importance of the inflation goal and the average inflation rates are calculated. As is usually the case, the Spearman rank correlation estimates are lower than the ones calculated using Pearson s statistic. From the table, we can draw the following conclusions. Ø The central bank indices are negatively correlated with the average inflation rate. A significant correlation exists only for economic independence. 18 More information about the estimation in differences can be obtained from the author upon request.

17 Table 7 Pearson s and Spearman s correlation coefficients between central bank independence and the OLS-estimators Ž OLS, OLS. for importance of the inflation goal, the SUR-estimators Ž SU, SU., the RALS-estimators Ž RA, RA. and average inflation Ž Inflat. PolIn EcoIn Bade AleSu OLS OLS SU SU RA RA Inflat PolIn ) ) y0.41 EcoIn ) )) 0.74)) 0.78) y0.72) Bade )) )) 0.91)) 0.90)) 0.92)) 0.91)) ) y0.55 AleSu )) )) 0.98)) 0.89)) 0.99)) ) y0.62 OLS )) 0.88)) 0.76) )) 0.99)) 0.84)) 0.74)) 0.77)) y0.67) 1 OLS )) 0.85)) 0.95)) 0.78)) )) 0.99)) 0.69) 0.77)) y0.73) 12 SU ) )) 0.83)) )) y0.87)) 1 SU ) 0.95)) )) 0.74) ) y0.74) 12 RA ) )) 0.64) ) )) y RA ) ) 0.64) ) 0.87)) 1 y0.46 Inflat y0.38 y0.72) y0.52 y0.60 y0.69) y0.91)) y0.93)) y0.81) y0.40 y Pearson s correlation coefficient is given above and Spearman s below the main diagonal. )Rejection of the null hypothesis at the 5% significance level. ))Rejection of the null hypothesis at the 1% significance level. B. HayorEuropean Journal of Political Economy 14 ( 1998 )

18 258 ( ) B. HayorEuropean Journal of Political Economy Ø The estimated indicators for the importance of price stability are also negatively related with the inflation rate. Except for the RALS-estimates of the sensitivity parameters, all correlations are statistically significant. Ø The highest correlation coefficient with average inflation is computed for SURE Then we have the OLS-estimators and economic independence. The lowest correlations are found for RALS-estimators and political independence. Ø The indicators for economic and political independence of the central banks are positively correlated with the estimated sensitivity parameters. In most cases, though, this correlation is not significant. In Section 6, these findings are put into a broader perspective. It is tempting to include the competing variables together in a regression model to assess their respective shares in explaining the variation of the inflation variable. Unfortunately, the number of observations is much too small for meaningful results from a multivariate analysis involving these highly collinear variables. To summarise the results of such an analysis, most of the central bank indices are insignificant and often display a positive sign when put together with a sensitivity estimator in a regression. The coefficients on the sensitivity estimators are typically more often significant and exhibit a negative sign. This finding does not hold in the case of the index variable capturing economic independence, which tends to be as equally important as the sensitivity estimators, as long as an intercept is included in the model. In a model without an intercept term, the sensitivity estimators show again higher t-values and the expected negative sign of the coefficient. 6. Conclusions We are faced with the problem of observational equivalence, since the empirical results are consistent with both the orthodox and economic culture hypothesis. However, previous findings for the orthodox hypothesis have only been replicated, while the evidence in favour of the economic culture hypothesis is new. As discussed above, the observation of a negative relation between central bank independence and inflation rates does not imply causality. Moreover, the empirical results do not exclude different stability cultures in the EC countries as the prime underlying explanation for observed outcomes. That the sensitivity estimates are more highly correlated with the inflation rate than the independence indices raises further doubts about the dominant role given to the independence argument in the literature. 19 This is due to the fact that only eight countries are included in the SURE-estimate. Computing all correlations for these countries only brings out the OLS-estimators as having the highest correlation with average inflation.

19 ( ) B. HayorEuropean Journal of Political Economy Still, we cannot conclude that public opinion is the most important key to explaining different national inflation records. There may be reverse causation. It could be the case that the design of the central bank influences people s assessment of the importance of keeping prices stable. In other words, people could consider the inflation goal to be of greater importance, if it were demonstrated to them that the achievement of stable prices is feasible. But this brings us back to the historical-feedback interpretation sketched above, only with more emphasis on central bank independence. As outlined in Section 2, a hypothesis based on public s preferences opens up two competing interpretations of how central bank independence and preferences hang together, the preference-instrument and the historical-feedback interpretation. Statistical causality analysis could help to shed some light on which of the positions are preferable on empirical grounds. The finding of a two-way causality between attitudes on inflation and central bank independence would accord with the feedback-mechanism, while a unidirectional causality from attitudes to the legal basis of central banks would support the preference-instrument view. Nocausality relations would fit into the orthodox view. Given the restraints of the data, we cannot meaningfully investigate Grangercausality relationships. We cannot even trust Granger-causality tests between actual inflation and attitudes, since the test is very weak for small samples and there are no degrees of freedom left to perform a specification search over different lag length Ž see, for example, Thornton and Batten, On the other hand, few philosophers would accept Granger-causality as a true representation of causality anyway Ž cf. Holland, Still, with more data at hand, further investigations along these lines could be interesting. We can nonetheless try to say something on this issue using a more heuristic approach. From the results given in Table 7, it can be seen that there sometimes exists a high positive correlation between the indices for central bank independence and the sensitivity indicators. This suggests that a large amount of variation is common to both types of variables. If a correlation coefficient between central bank indices and sensitivity parameters is not that high, for example in the case of OLS1 and political independence, the correlation between the latter and average inflation tends to be rather low. A possible explanation for this finding is that formal and actual independence can drift apart. This could be taken as tentative evidence that, if the negative association of sensitivity parameters and central bank indices with average inflation is high, it is likely to be based on common variation of the two types of variables. This implies that sensitivity parameters and independence indices when both correlate highly with inflation may capture similar influences, pointing towards the position that central bank independence and public attitudes towards inflation are interconnected. These arguments give indirect empirical support for the historical-feedback interpretation. On the issue of EMU, the conclusion from this analysis is that it may not suffice for price stability to create an independent ECB. Also of importance in