Establishing a hawkish reputation: interest rate setting by newly appointed central bank governors

Abstract

In this article, we explore the interest rate setting behaviour of newly appointed central bank governors. We estimate an augmented Taylor (1993) rule for 15 OECD countries and the period 1974 to 2008. We find that newly appointed governors fight inflation more aggressively during the first four to eight quarters of their tenure in an effort to establish a reputation for being inflation averse.

Keywords:

• central bank governors
• credibility
• inflation
• monetary policy
• reputation
• Taylor rules

JEL Classification:

• E31
• E43
• E52
• E58

Acknowledgements

Thanks to Frederike Anika Engel, Edith Neuenkirch, Florian Neumeier, and Peter Tillmann as well as seminar participants at the University of Trier for their helpful comments on earlier versions of the article. The usual disclaimer applies.

Notes

¹ We use the term ‘governor’ to refer to the central bank’s head, even though the actual job title sometimes is ‘president’ or ‘chairman’.

² Currently, it is typically committees that decide on appropriate interest rates. However, empirically, the governor is found to have a huge influence on them. For instance, Blinder (2007) concludes that Alan Greenspan was influential enough to almost always impose his view on the Federal Open Market Committee. Although it is doubtful that the governor has complete discretion in setting the interest rate all the time, she/he is almost never outvoted in monetary policy decisions (Claussen et al., 2012). This implies that the governor should have at least some agenda setting power when it comes to a vote in the monetary policy committee.

³ Hansen and McMahon (2011) study the voting behaviour of individual members of the Bank of England’s Monetary Policy Committee. However, their analysis focuses on the preferred level of interest rates and dissents throughout the central banker’s tenure, rather than on the reaction to inflation.

⁴ All countries that joined the OECD after 1974 are omitted from the sample as are countries with no scope for independent monetary policy. For example, Austria pegged its schilling against the German mark and Luxembourg had no central bank before 1998 and relied on the Belgium National Bank’s monetary policy. Interestingly, Belgium itself pegged its franc against the German mark.

⁵ For instance, if the inauguration date is May 1981, the governor can influence the central bank rate in Q2-1981. In contrast, central bankers who take office in June 1981 will first affect the central bank rate in Q3-1981. We explored the robustness of this assumption with three additional settings: the governor had to be in office in the respective quarter for at least (i) 1 day, (ii) 2 months or (iii) a full quarter. However, the differences across indicators are small as we focus on a period of four to eight quarters after inauguration. This implies that the indicators partly overlap and reduce differences across the indicators. All omitted results are available on request.

⁶ Data sources: central bank rates (IMF), consumer price indexes (OECD) and industrial production (OECD). Note that the data used in this analysis are ex post data due to the lack of real-time data for all 15 countries and the complete sample period.

⁷ One could also test for differences in the reaction to the output gap in Equation 2. However, as the regression estimates indicate no change in all cases, we focus on differences in the reaction to inflation.

⁸ Note that the panel is unbalanced as monetary policy in Finland, France, Germany, Italy, Portugal and Spain has been conducted by the European Central Bank since Q1-1999. Furthermore, there are some missing observations for industrial production at the beginning of the sample in case of Australia and New Zealand.

⁹ Note that the inclusion of cross-sectional dependence in the weighting matrix can also be interpreted as a proxy for time fixed effects.

¹⁰ Note that standard econometric software is not able to invert the matrix of instruments when using all valid lags to define moment conditions (Arellano and Bond, 1991). Furthermore, simulation studies show that there is a trade-off when increasing the number of lags: although efficiency increases, so does the finite sample bias of the GMM estimates (Judson and Owen, 1997). To ensure the robustness of our findings and address Kiviet’s (1995) criticism of dynamic panel GMM estimators, we additionally estimate Equations 1 and 2 using panel generalized least squares and weights based on the assumption of contemporaneous correlation between the cross sections. Results are available on request.

¹¹ It is well known that monetary policy was considered ‘passive’ during the 1970s in many Western economies (see, e.g., Lubik and Schorfheide, 2004) leading to such estimates. Not surprisingly, the reaction to inflation is significantly larger than 1 as soon as the starting point of the sample is restricted to 1983 or later.

¹² Note that formal testing is not required as Table 1 shows that the even more conservative two-sided null hypothesis is rejected in all cases.