The Surprise Effect
Most econometric research (Sims 1980; Canova and De Nicolo 2002; Giannoni and Woodford 2005; Bernanke, Boivin and Eliasz 2005; Christiano, Eichenbaum and Evans 2005) do not distinguish between anticipated and unanticipated components of monetary policy.
However, generally speaking, they find that monetary policy shocks are effective over a period of some years. This would confirm the relevance of the surprise effect, in contrast with the opinion of Lucas (1996: 679) himself. However, it mainly raises the need, underlined by Mishkin (1982), to introduce the possibility that monetary action could be partly anticipated but still effective. Thus, the reason why what appears to be a relevant surprise effect lies in misspecification of the model. In fact, Mishkin found that a lag specification of anticipated policies leads to results that tend to negate their ineffectiveness, thus invalidating Lucas' proposition.This finding has been confirmed by later contributions. Cochrane (1998) again underlined the need to distinguish between anticipated and unanticipated components of monetary policy in all types of models. Anticipations can derive from explicit central bank communication, especially through forward guidance on its future action (we will deal extensively with this in Chapter 5 and 6) or simply reflect the private sector's expectations. This kind of analysis cannot be done through a vector autoregression (VAR), and various authors estimate a New Keynesian model in order to compare the response of output to policy surprises and news. The estimated effects of monetary policy shocks are very different from those usually found in the literature. Now the contribution of news is less than 2 per cent of fluctuations, whereas the effect of anticipated monetary policy varies from 15 to 25 per cent of medium-run output fluctuations and is more persistent. Real effects of systematic monetary policy depend on the shares of frictionless competitors having REs and rules- of-thumb agents or agents facing other frictions. Hoover, Jorda (2001), using a VAR, confirmed and strengthened these findings of significant real effects deriving from anticipating systematic monetary action (see also Flaschel, Franke and Semmler 1997: chap. 8).
Most recently, by estimating a benchmark New Keynesian model, Milani and Treadwell (2012) find that the contribution of monetary policy news to output fluctuations is larger than that of surprise shocks. This stresses the relevance of the communication policy followed by the central bank in recent times, especially of forward guidance.
3.4