Is M3 an appropriate aggregate for guiding ECB monetary policy?

This paper utilises an approach to long run modelling proposed by Pesaran, Shin and Smith (2001) to develop empirically weighted broad monetary aggregates for the Euro area. Two alternative aggregates are derived based upon the long run relationship between the M3 component assets and nominal income or prices respectively. The empirical results do not support the use of M3 as the key monetary indicator, as both aggregates accord a very low, or zero weight, to the broadest of the M3 component assets. The implied optimal weights are very different from those that would be implied by either simple sum or Divisia aggregation. Furthermore, recursive estimation reveals that the optimal weights do evolve over time in response to financial innovation and changes in wealth holder preferences. This implies that an aggregate such as M3, with fixed and equal weights of unity on all component assets, may not be a reliable leading indicator for inflation. Out of sample forecasts confirmed that the optimally weighted monetary aggregates have superior predictive content for inflation at longer forecast horizons such as 12 quarters.