Essays on the decomposition of macroeconomic time series into permanent and transitory components

This dissertation is comprised of three essays on modern macroeconometrics. The first essay takes issue with the conclusion of recent papers that U.S. output is trend stationary. It is shown that tests of the null hypothesis of a unit root against the alternative of trend stationarity are sensitive to data-based lag selection and departures from the maintained hypothesis of temporal homogeneity.;Specifically, time series which contain a unit root may appear to be trend stationary if they are perturbed by large additive outliers. This generates the false appearance of trend stationarity. There is overwhelming evidence against the hypothesis of trend stationarity in the post-war data. Also, the implied business cycle is implausible.;The second essay extends the framework of dynamic factor models by specifying two common factors. Both common factors are subject to changes in regime. This allows for asymmetry in both the common permanent and common transitory components of time series. Previous work with dynamic factor models has restricted asymmetry to only the permanent component of a time series. This assumes a priori that recessions cause permanent damage. In light of recent evidence which suggests that recessions only temporarily lower output, we allow for the possibility that the source of business cycle asymmetry is the common transitory factor. With the exception of the 1990-91 recession, which appears to be entirely permanent, we find that five of the six recessions from 1959 to the present are comprised of both permanent and transitory variation. Our parameter estimates imply that a six month recession permanently lowers the level of industrial production by 2.86%.;The third essay proposes a variety of statistics which test the hypothesis that a time series contains a point of structural change and/or has a unit root. Regarding tests for structural change, the literature is incomplete. A complete cataloging of these tests is undertaken. Also considered are statistics which test the joint hypothesis of a unit root and no structural change. While these have the potential to offer an increase in power over statistics which ignore the unit root hypothesis, this is generally not found to be the case.