Department of Economics and Business Economics

Generalized Efficient Inference on Factor Models with Long-Range Dependence

Research output: Working paper/Preprint Working paperResearch


  • rp16_05

    Final published version, 565 KB, PDF document

A dynamic factor model is considered that contains stochastic time trends allowing for stationary and nonstationary long-range dependence. The model nests standard I(0) and I(1) behaviour smoothly in common factors and residuals, removing the necessity of a priori unit-root and stationarity testing. Short-memory dynamics are allowed in the common factor structure and possibly heteroskedastic error term. In the estimation, a generalized version of the principal components (PC) approach is proposed to achieve efficiency. Asymptotics for efficient common factor and factor loading as well as long-range dependence parameter estimates are justified at standard parametric convergence rates. The use of the method for the selection of number of factors and testing for latent components is discussed. Finite-sample properties of the estimates are explored via Monte-Carlo experiments, and an empirical application to U.S. economy diffusion indices is included.
Original languageEnglish
Place of publicationAarhus
PublisherInstitut for Økonomi, Aarhus Universitet
Number of pages27
Publication statusPublished - 1 Feb 2016
SeriesCREATES Research Papers

See relations at Aarhus University Citationformats

Download statistics

No data available

ID: 96566316