Nonparametric evidence on asymmetry in business cycles using aggregate employment time series

Abstract

This paper presents nonparametric evidence on asymmetry in business cycles using time series on production worker employment in the U.s. durables and nondurables manufacturing sectors. A seminonparametric method for time series is used to estimate the conditional distribution of each of the two data series given its past. Durables employment is found to display a larger conditional variance following a contraction in the series and a smaller conditional variance following an expansion. Separate nonparametric kernel estimates of conditional variances verify this finding. This property is consistent with existing evidence of cyclical asymmetry in unemployment rates. Cyclical asymmetry is not evident for nondurables.