The price stability oriented monetary policy of the ECB: an assessment

Abstract

The definition of price stability adopted by the ECB has recently been criticized in the literature, particularly for being unable to fully anchor inflation expectations and creating a deflation risk. In the paper empirical evidence is provided against these claims. Despite the unfavourable macroeconomic conditions for the euro area since 2001, monetary policy management has lead to the setting of the policy rate at levels compatible with trend inflation (the long-run inflation forecast) in the range 1–3%, and therefore without affecting negatively the inflation outlook.

Notes

¹ See Alesina et al . (2001) and Gros et al . (2001).

² See Fitoussi and Creel (2002), Fitoussi (2003), Svensson (2003), Wyplosz (2003), and von Hagen and Hofman (2004).

³ See IMF (2002) and Svensson (2002a, b, 2003).

⁴ See ECB (2003).

⁵ See Morana (2004) for evidence in favour of the inflationary effects that monetary base growth may exercise also when the economy is in a deflation trap, with short-term nominal interest rates at the zero lower bound.

⁶ In this study quarterly euro-12 area data from 1980:Q1 through 2003:Q4 have been used. As a measure of M3, seasonally adjusted quarterly averages of the month-end stocks of M3 are used. Nominal GDP is in millions of euro and has been seasonally adjusted and converted to euro via the irrevocable fixed conversion rates of 31 December 1998. The real and nominal GDP series are used to construct the GDP deflator. The short-term nominal intertest rate has been computed using 3-month money market interest rates for the euro area countries, while the long-term nominal interest rate has been computed from 10-year government bond yields or close substitutes. The stock market index is taken from Datastream (TOTMKEM) and converted into euro using a synthetic US$/EUR exchange rate series.

⁷ Note that this is not a restrictive assumption, since the actual specification of the long-run money demand equation is obtained by the error correction terms entering the real money balances dynamic equation in the VECM model, and may include, for instance, a proxy for wealth or the opportunity cost of holding money.

⁸ This relationship can be derived from the Gordon growth model, assuming a proportional relationship between real dividends and output.

⁹ The motivation for a time-varying parameter inflation risk premium can be found in the process of economic convergence, in the countries currently belonging to the Euro area, following the introduction of the Maastricht Treaty in 1992. See Section III(a) for additional details.

¹⁰ Although recent evidence would point to long memory and structural breaks as the cause of persistence of the euro area inflation rate (see for instance Morana, 2005), in the light of the econometric framework employed, modelling it as an I(1) process may be considered preferable to the modelling of the variable as an I(0) process, given the strong shock persistence shown by long memory processes.

¹¹ Critical values for the trace test statistic for the model with the step dummy have been computed using DISCO and are reported in Table 1.

¹²Bruggeman et al . (2003) consider a slightly different system from the one analysed in the paper, also including the own return for M3. The time period investigated is also different (1980:2–2001:4) and they do not allow for a break in the Fisher parity equation. On the basis of Bartlett corrected statistics, they conclude in favour of two cointegrating vectors. Despite these results, also on the basis of the error correcting behaviour of the whole system, it is considered that the assumption of four cointegrating vectors is appropriate for the present data. Results for the Ericsson and MacKinnon (2002) test are available upon request to the author. See also Bagliano et al . (2002) and Brandt and Cassola (2004).

¹³ Once the restrictions implied by cointegration and the orthogonality of the permanent shocks have been taken into account, additional restrictions need to be imposed on the long-run impact matrix, on the basis of economic theory, for exact identification. See the methodological Appendix.

¹⁴ In addition to the restrictions implied by the orthogonality of the transitory shocks and the orthogonality between the permanent and transitory shocks, r(r − 1)/2 additional restrictions, based on economic theory, need to be imposed on the contemporaneous impact matrix for exact identification. See the methodological Appendix.

¹⁵ The suggested interpretation for the structural shocks is also supported by the estimated impulse response functions, which are not reported for reasons of space. They are available upon request to the author.

¹⁶ Some experimentation showed that using 20 lags in the historical decomposition was sufficients to achieve a full reconstruction of the cyclical components.

¹⁷ Disequilibrium dynamics is termed as the one associated with the transitory innovations, and equilibrium dynamics the one associated with the permanent innovations. See the methodological Appendix.

¹⁸ Overvaluation of the stock market is measured relatively to the trend component determined by productivity shocks.

¹⁹ See ECB(2004) for similar findings.

²⁰ See Bagliano et al . (2002).

²¹ See also Altavilla and Landolfo (2005) for similar findings concerning the implicit inflation reference value.

²² In the preset application s_t  = 1, 2, i.e. high and low inflation risk premium regime.

²³ The p-value of the tests are 0.1176, 0.3765, and 0.6447, respectively.

²⁴ The p-values for the LR linearity test are equal to 0.0730 and 0.0552 for the three-regime and the two-regime models, suggesting that the two-regime model should be preferred to the three-regime model. Moreover the two-regime model should be preferred to the linear model, albeit the null of linearity is not rejected at a significance level lower than 6%. However, evidence of instability in the mean component are also provided by the Hansen instability test: the test statistic is equal to 0.505, allowing to reject the null of stable intercept at the 5% significance level. Moreover, the estimated means in the two regimes are statistically different. The test for the equality of the two means is equal to 9.079, with p-value equal to 0.