We analyze the issue of the impact of multiple breaks on monetary neutrality results, using annual data on real output and monetary aggregates for Argentina (1884-1996), Australia (1870-1997), Brazil (1912-1995), Canada (1870-2001), Italy (1870-1997), Mexico (1932-2000), Sweeden (1871-1988), and the UK (1871-2000). In particular, we empirically verify, whether neutrality propositions remain addressable (and if so, whether they hold or not), when unit root tests are carried out allowing for multiple structural breaks in the long-run trend function of the variables. It is found that conclusions on neutrality are sensitive to the number of breaks allowed. In order to interpret the evidence for structural breaks, we utilize a notion of deterministic monetary neutrality, which naturally arises in the absence of permanent stochastic shocks to the variables. We utilize a resampling procedure based on the fact that changes in the trend function bias unit root tests towards a non-rejection. In particular, using a dynamic programming algorithm to obtain global minimizers of the RSS for locating breaks, we simulate the distribution of the t-statistic for the null hypothesis of a unit root, under the hypotheses that the true models are both a Trend Stationary (TS) model with up to four structural breaks, and a Difference-Stationary (DS) model, both estimated from the data. We then compare the position where the sample estimate of the t-statistic for testing a unit root lies relative to the empirical densities of the t-statistic, under both the estimated TS and DS models. We present evidence in favour of models in which the cycle fluctuates in a stationary way around a broken trend. In other words, the (unit root) permanent stochastic changes vanish, giving rise to stationary behaviour affected by infrequent structural breaks. This leads to interesting questions about the testing for monetary neutrality, and allows us to introduce the concept of deterministic monetary neutrality.
