عدم تقارن متغیر با زمان در تنظیمات بانک مرکزی : مورد بانک مرکزی اروپا

This paper examines the asymmetric preferences of the European central bank (ECB) as identified by Surico, 2007a and Surico, 2008. Under asymmetric preferences, a central banker places different weights on the losses associated with positive and negative deviations of economic variables such as inflation or output from their target values. Although asymmetry is conventionally estimated by the generalized method of moments, we use the bias correction Kalman filter suggested by Kim (2006), introducing the concept of time-varying asymmetry in central bank preferences. Estimates of the interest rate reaction functions suggest asymmetries in preferences for both output gap and interest rate. These asymmetries indicate that the ECB increases its interest rate aggressively when there is a surge in output but does not sustain an interest rate above its reference value.

In conventional monetary policy analysis, the central banker’s objective – in particular, price and output stability – is often specified using a quadratic functional form. Linear-quadratic analysis is a tractable framework that can be used to examine the dynamic features of an optimal monetary policy. In addition, Rotemberg and Woodford (1998) and Woodford (2003) provide a micro-foundation for the quadratic objective as a second-order approximation to a representative agent’s utility function. With the justification of a micro-foundation, a large number of studies have employed the quadratic objective. On the other hand, an increasing number of studies attempt to introduce more flexibility into the central bank’s objective function despite the theoretical weakness of the lack of a micro-foundation. These studies provide empirical results for preferences that have been estimated using a reduced form derived from the central banker’s optimization problem and are more general than the quadratic objective. Moreover, practitioners often do not favor the use of a quadratic objective for the central bank and suggest that the symmetry of the quadratic form around the origin is not realistic as it implies that central bankers put equal weights on losses from output expansion and contraction.1 In reality, it is also likely that policymakers adjust their attitudes to the losses associated with various economic variables over time. Surico (2007a) investigated the asymmetric preferences of the ECB. The asymmetric preference model provided by Surico, 2007a and Surico, 2008 assumes a Linex (linear exponential) loss function, which includes the quadratic objective as a special case. The Linex function allows for asymmetry in the loss function around the target values of the central bank. As a result, the central banker sets different weights for losses associated with positive and negative deviations of economic variables from their targets. The remarkable advantage of this approach is that the hypothesis of symmetric preferences can be tested by imposing restrictions on the estimated Euler equation of the central banker’s optimization problem. If the restriction is rejected, the assumption of asymmetric preferences is statistically justified. With this approach, Surico (2007a), using a sample from January 1999 to December 2004, highlighted the ECB’s aversion to output contraction and to the interest rate exceeding a reference level but found no asymmetry for inflation. Although there is a large body of empirical literature on the ECB, only Surico, 2003 and Surico, 2007a and Aguiar and Martins (2008) have examined its asymmetric preferences. Aguiar and Martins (2008) utilized a threshold quadratic loss function, which also allows for the application of different weights to the positive and negative deviations of economic variables, and found that the ECB has an aversion to higher inflation. The differences between their results and Surico’s (2007a) findings stem from the differences in the data sets and assumed models (Aguiar and Martins, 2008, 1660).2 Their approach differs from ours in that the threshold loss function limits its functional form to a quadratic. This paper adopts Surico, 2007a and Surico, 2008 framework in order to investigate the ECB’s policy, using a more general form of the objective. The most important contribution of this paper is the estimation of time-varying asymmetry in central bank preferences using the bias correction Kalman filter (Kim, 2006). In previous studies, average asymmetries in preferences were estimated using the generalized method of moments (GMM). In contrast, the time-varying asymmetry introduced in this paper changes over time; the estimator can describe how the central bank updates its preferences in each period. This paper obtains the following three results. First, the estimated ECB’s objective function exhibits a positive asymmetry for output over the whole sample. In contrast to the results in Surico (2007a), this implies that the ECB places a larger emphasis on losses from output expansion than on those from output contraction. Second, the result in Surico (2007a) remains intact if the same sample period as in his study is considered. Third, the interest rate policy of the ECB is characterized by bilateral asymmetry for both output and interest rate. The ECB increases its interest rate aggressively when there is a surge in output but does not sustain interest rates above a range of 3.5–4%. This paper proceeds as follows. Section 2 reviews the earlier contributions to the literature on asymmetric preferences of central banks, including investigations of other general specifications of the loss function. Section 3 introduces Surico, 2007a and Surico, 2008 model, consisting of inflation and interest rate reaction functions. Section 4 describes the strategy for estimating time-varying asymmetry in central bank preferences. Section 5 reports the results of both GMM estimation and Kalman filtering. In particular, we follow the time-varying nature of the ECB’s policy preferences during the first decade of its existence. Section 6 concludes the study.

This paper estimated the asymmetric preferences of the ECB based on Surico, 2007a and Surico, 2008. First, asymmetries in the ECB’s preferences for output, inflation, and interest rates are estimated with GMM using inflation and interest rate reaction functions. Then, the time-varying asymmetry in central bank preferences for the interest rate reaction function is estimated with the bias correction Kalman filter. The results of the GMM were as follows. First, a preference for output expansion could be found for the sample period also used in Surico (2007a). Second, estimation with an extended sample showed a reversed asymmetry, i.e. an aversion to output expansion. In summary, the ECB’s interest rate setting can be characterized by bilateral asymmetries for output and the interest rate – an aversion to output expansion as well as interest rate above its reference value. The ECB increases its interest rate when output expands; on the contrary, it reduces the interest rate immediately once the interest rate exceeds its reference value. Time-varying asymmetry for output was estimated to be positive and increasing over time indicating an increase in the ECB’s aversion to output expansion. The time-varying estimate for asymmetry in the ECB’s preferences towards inflation is not significant. Asymmetry in the ECB’s objective function with respect to the interest rate is estimated to be positive but not significantly different from zero.