Modelling Anti-inflationary Monetary Targeting in Romania

Abstract

This paper characterises Romania's experience with anti-inflationary monetary targeting over the period 1999–2005 prior to the country's switch to inflation targeting. We uncover the National Bank of Romania's preferences, conditional on an estimated macro-model. We find that Romania's monetary targeting regime can be characterised by a concern for price stability and an additional role for smoothing of the central bank's instrument (base money growth). Exchange rate variability and output gap stability appear not to significantly enter the National Bank of Romania's objective function.

Keywords:

• Monetary targeting
• optimal monetary policy
• Romania

Acknowledgements

This paper benefited from discussions with Máximo Hemingsen. The views expressed here do not necessarily reflect those of the European Central Bank. The usual disclaimer applies.

Notes

¹See, for example, Gerlach & Svensson (2003), Laubach & Posen (1997) and Mishkin (2001).

²See, for example, the description in Guttmann (2005) for the Australian case.

³Price stability has instead been made the primary objective of Korea's inflation targeting regime since 1999 (Sánchez, 2009).

⁴The loss function parameters indicate how different goals are traded off in response to shocks. They are estimated under the assumption that the NBR sets monetary policy optimally, conditional on economic outcomes and an empirical macro-model.

⁵A weaker currency is generally found to be contractionary in the empirical literature, even after including a number of different controls. Concerning liability dollarisation, see for example Céspedes et al. (2004). Those discussing the implications of contractionary depreciations for monetary policy include Eichengreen (2005), and Sánchez (2007, 2008).

⁶Technically, we are not influenced by other attempts at characterising monetary policy intentions, such as Favero and Rovelli (2003) and especially Cecchetti and Ehrmann (2002).

⁷These conditions were given by drastic macroeconomic adjustment in the context of an IMF-backed stabilisation programme and an adverse international environment (Gabor, 2008).

⁸The growth rate of Romanian foreign reserves induced by capital flows remained high at the start of the IT regime. This is partly due to substantial liberalisation of the capital account in early 2005 and anticipations of EU accession in 2007.

⁹See section 4 for a description of the main data series used in Figure 1 and Table 1. Additionally, BIS data are employed for core inflation measures, real GDP and bilateral exchange rates. Two core inflation measures are used: core 1 excluding volatile (vegetables, fruit, eggs, fuel) and administered prices, and core 2 excluding only administered prices. The external variables included in Table 1 are a subset of all those considered in our analysis.

¹⁰The reason for dropping year 1999 from this period is that we lack quarterly data for some variables (M2, core inflation measures and real GDP). The results reported here for other variables would be substantially unchanged if we included year 1999.

¹¹More specifically, we define variable , where z=m; y; e. Notice that this averaging simply amounts to attaching weights in the loss function to objective variables over more than one period. This does not affect the time series properties of the errors in the macromodel equations (2) through (5) below. The estimation of these equations employs no moving averages of the data.

¹²Note that we allow for contemporaneous effects of the nominal interest rate and NEER on money demand. The model is closed with a PPP constraint: .

¹³The Appendix contains the details of our estimation. Estimating the model until the announcement of the IT regime (August 2005) did not substantially change any of the results reported below.

¹⁴Our use of a two-step method (first estimating the macro-model, then, conditional upon the latter, obtaining the policy coefficients) allows us to consistently compare the optimal central bank's loss function with rule-of-thumb benchmarks for the policy weights, conditional on the same empirical macro-model.

Underlying all base money growth paths are the set of weights reported in Table 2. Concerning the weights that are based on optimisation and thus estimated (lines 3 and 4 in Table 2), the empirical results involved are reported in Table 4. The estimate for ω_π is 0.455. The null hypothesis that ω _dm equals zero is rejected (Wald test: ; ); the coefficient's point estimate is 0.254. The table shows that ω _y and ω _e are each insignificant at the 5% level. Since this is also true for them jointly (Wald test: ; ), we set both coefficients to zero. The final estimates for ω_π and by ω _dm are thus re-normalised to 0.641 and 0.359, respectively.

We thus find that the NBR attached a negligible emphasis on both output and the exchange rate. The latter two variables however play indirect roles in the determination of inflation, and thus of monetary policy itself. Inspection of our estimated macro-model allows us to see that inflation is directly affected by REER (see equation (A.2) in the Appendix). Output influences inflation indirectly, given that it enters money demand function (A.4), thus affecting interest rates and – indirectly via equation (A.3) – the exchange rate, which in turn impacts inflation, as we have just seen.16 In sum, even if they are not monetary policy objectives, output and exchange rates indicators are worth being monitored on a regular basis by the NBR.

¹⁵Which specific type of normalisation we adopt is inconsequential. The only thing that matters for representing central bank preferences are the relative weights between all possible objectives.

¹⁶The demand for output reacts (negatively) to the exchange rate (see equation (A.1) in the Appendix). This can be rationalised in terms of the adverse effect of the exchange rate on competitiveness.

¹⁷More specifically, actual base money growth rates depicted in Figures 2 through 4 correspond to smoothed (six-month moving averages) transformations of annual growth rates. Simulated paths are not smoothed, as they are already derived from an optimisation involving smoothed goals.

¹⁸This analysis concerns Romania's actual experience with MT. Sterilisation of capital flows did not become a destabilising force before the regime's replacement by IT.

¹⁹Despite being parsimonious, these exogenous processes were found to pass standard misspecification tests.