A reexamination of the Nelson-Plosser data set using recursive and sequential tests

Abstract

The present paper applies to the Nelson-Plosser data set the recursive, rolling, and sequential tests proposed by Banerjee, Lumsdaine and Stock (1992) for unit roots in the presence of mean or trend breaks. Unlike Perron's method, these three types of test endogenize the break point in the mean or trend and thus are more appealing in empirical studies. The (reverse) recursive test indicates rejection of the unit root null in industrial production and unemployment rate. The sequential test indicates that nominal GNP and common stock prices are stationary with a break in the mean.