The empirical role of money growth in the macroeconomy has been studied by many researchers using different data and methods. The most prominent studies focusing on the quantity theory of money are those of Friedman and Schwartz (1963, 1982) using data averaged over the phases of the business cycle in order to capture longer-run movements with U.S. and U.K. data back to 1867. The study by McCandless and Weber (2001) is a representative example for a different approach for 110 countries in the post-WWII period: longer-run growth rates are approximated by growth of a variable over a 30-year period (average growth rates). (1) The problem with phase-averaging is that it does not deal with possible nonstationarity of the data. (2) The problem with growth averages over long periods, while informative to a degree, is that they have limitations, too. It is impossible to assess the timing of movements in these variables. (3)
In the frequency domain, the "long run" is associated with frequency zero, which in our terminology is the long-run trend. Previous studies in the frequency domain on the long-run neutrality of money at zero frequency include the studies of Lucas (1980) and Geweke (1986). The theory of spectral analysis provides a rigorous framework for extracting specific frequency bands from the data, instead of considering only frequency zero. Stock and Watson (1998) applied the band-pass filter of Baxter and King (1999) to extract cycles (frequency bands) for U.S. data and focused on the business cycle and on mostly the post-WWII period. Christiano and Fitzgerald (2003) proposed an improved band-pass filter and provided an application to the longer-run relation of money and inflation, for bands of 8-20 years and 20-40 years, in the United States from 1900 to 1997. (4)
The usefulness of monetary aggregates for conducting monetary policy has been seriously called into question because of the instability of money demand. A prominent empirical study is that of Estrella and Mishkin (1997). Recent mainstream New Keynesian models have no explicit role for money (Woodford 2008). A contrasting view is that the longer-term role of money is still of relevance for monetary policy. The European Central Bank explicitly emphasizes a supporting role that monetary aggregates should play in order to provide an anchor for inflation expectations (Hafer, Haslag, and Jones 2007; Issing 2006; King 2002; Nelson 2003). Christiano, Motto, and Rostagno (2007) provided a theoretical New Keynesian model where inflation expectations lose their anchor, unless money is monitored and reacted to. Also, Beck and Wieland (2008) presented a New Keynesian model, where cross-checking with money growth movements improves inflation control. Similarly, Assenmacher-Wesche and Gerlach (2006, 2007) found a significant role for a low frequency (low-pass filtered) component of money in the empirical New Keynesian Phillips-curve equation. (5)
In the spirit of Friedman (1961), we use long time spans so that our results are unlikely to be because of a specific policy rule in place in a given period. We have 122 years of annual data, from 1880 to 2001, before the introduction of the euro in 2002. We have real gross domestic product (GDP), money, and consumer price indices (CPIs), with no gaps, for 11 industrialized countries. (6) The countries included in our study differ in such terms as their institutions, fiscal and monetary policies, and economic growth rates. Nonetheless, at least in peace time they were similar in generally having comparatively free markets in which prices and interest rates were determined. Our aim is to apply analytic methods that are free from economic and statistical models as much as possible in order to explore empirical regularities, or stylized facts, in the data. Also, the focus of our research is the longer run.
We employ band-pass filters to pick out cycles from each time series that last 8-40 years, that is, the longer-term component. …