Asymmetric monetary policy: empirical evidence for Italy

Abstract

A growing body of empirical work has examined the potential asymmetry in the effects of monetary policy on United States real activity. This study looks for such an empirical evidence for Italy in the period 1982–1998. Monetary shocks are obtained as residuals from a central bank reaction function where the three-months interbank rate is taken as the indicator of the monetary policy stance. The effects of these positive and negative shocks on output are statistically different from zero and the null of symmetry between the two is rejected in favour of negative shocks having a greater impact on real output growth, thus confirming an asymmetric effect of monetary policy even for Italy.

Notes

As will be seen later, the study refers to the asymmetric adjustment of prices (or convex aggregate supply curve) and the credit market imperfections (or pushing on a string view).

A different stream of economic literature that links monetary policy effectiveness to the phase of the business cycle is not reviewed here. Garcia and Shaller (2002) and Kakes (1998) stress that monetary policy – either tight or easy – should be more effective during recessions than during expansions.

Such an exercise has been reproduced successively with quarterly US data from 1960–1993 period and with quarterly data from the 1963–1993 period for a panel of 12 OECD countries by Karras and Stokes (1999b,c).

See Ball and Mankiw (1994).

In Karras and Stokes (1999c) there are two points that are skipped for the sake of simplicity: they endogenise the breakpoint of the asymmetry instead of imposing it at zero and they consider the possible existence of aggregate supply shocks introducing oil shocks.

In Karras and Stokes (1999c) there is some weak evidence in favour of the pushing on a string view but the authors are again inclined for an almost symmetric effect of monetary shocks on prices.

This work uses annual data from the 1953–1990 period for 18 European countries.

The use of monetary aggregates to identify a shift in monetary policy has been disputed by Sims (1980) who discovered that interest rates had a better predictive power for real activity: ‘when a short interest rate is added to the vector autoregression (…) the central role of the money stock surprises evaporates’ (p. 250).

Bernanke and Blinder (1992) emphasize the role of the federal funds rate as an indicator of the stance of monetary policy at least when the period 1979–1982 is excluded (during 1979–1982 the Federal Reserve explicitly abandoned use of the Federal funds rate as intermediate target).

In this choice De Arcangelis and Di Giorgio (1999) is followed.

Before the ‘divorce’ the Bank of Italy was forced to underwrite unsold Treasury bills at auction and it did not have much independence in the use of its instruments.

De Arcangelis and Di Giorgio (1999) note that the responses to shocks defined in this way (to ‘unforecastable innovations’) are less subject to the Lucas' critique.

Batten et al. (1990) and Sims (1992) stress the central role of short-term interest rates for the monetary authorities of France, Germany, Japan, the UK and the USA. Grilli and Roubini (1995) find the same evidence for Italy.

The auction rate on repo operations is used in the Bank of Italy monthly model of the money market (1988); the Treasury bill and the overnight rate are used, among the others, by De Arcangelis and Di Giorgio; Gaiotti (2000) uses the interest rate on the operations of the Bank of Italy.

The dates were taken from Gaiotti (2000) who uses the narrative accounts of the last 20 years by Angeloni and Gaiotti (1990).

The consumer price index and the real exchange rate have been seasonally adjusted using the X-11 ARIMA Seasonal Adjustment Program by the US Census Bureau.

A regression has even been run without the contemporaneous value of the output growth rate obtaining the same results.

In the robustness checks the output gap will be employed, instead of the output growth rate, obtaining the same results.

Usually, transformations of independent variables are used to correct for non-linearity and transformations of dependent variables to correct for non-constant variance of the variable itself or non-normality of the error terms.

The Pr(chi-square) for the skewness/kurtosis tests for normality passes from 0.0006 to 0.0109.

One can find the exclusion of the contemporaneous values of the shocks even in the output equation of Garibaldi (1997) and Morgan (1993).

The inclusion of the contemporaneous values of the shocks with the optimal lag structure for that particular case gives a result similar to the one by Karras (1996a). You can find it in Table 13.

See Buttiglione and Ferri (1994), Bagliano and Favero (1996), Bertocco (1997),

He detects, instead, the working of a credit channel in the 1960s and 1970s when the central bank was more interested in credit flows.