Cyclically-adjusted measures of structural trend breaks: an application to productivity trends in the 1990s

This paper compares several linear trend break models of labor productivity. One empirical problem that arises when estimating trends in macroeconomic data is the influence of cyclical behavior on tests of trend breaks. Using various methods to correct for cyclical influences, this paper finds that a simple linear trend model with a small number of breaks aptly characterizes aggregate productivity. Moreover, this paper confirms previous research that has found that the aggregate productivity trend has not steepened in the 1990s. Econometrically, this paper shows the benefits of using nonparametric bootstrap methods. In particular, a phase-dependent moving block bootstrap method appears well suited to generate small-sample critical values to test for trend breaks in macroeconomic time-series.