Abstract
Instead of exploring the uncertainty about the existence of a unit root in the long-span U.S. real GDP series as inprevious studies, e.g., Rudebusch (1993), inthis study we investigate the uncertainty about the state (permanence vs. transitoriness) of the output shock period by period by using the "innovation regime-switching" (IRS) model. In this model the effect of a shock may be permanent or transitory in different time periods. By applying the IRS model to the 1870-2008 annual U.S. real GDP data, we find that the output shocks in the periods of the 1893 depression, the 1907 financial panic, the two World Wars and the Great Depressionare likely to have had a large but transitory effect, whereas the output shocks inthe remaining periods are likely to have had a permanent effect. This result suggests that the longspan real GDP is neither a unit-root series nor a trend-stationary series.