Overview
Helmut Lütkepohl is an econometrician at the Institut für Statistik und Ökonometrie, Humboldt-Universität zu Berlin (as of 1999). He is best known for the textbook Introduction to Multiple Time Series Analysis (Springer, 1991), which remains the standard reference for the asymptotic theory of VAR and VECM estimation with integrated variables. His research covers lag-order selection criteria, Johansen cointegrating rank tests under various deterministic-trend specifications, Granger causality asymptotics, and impulse response inference.
Key Contributions
- Lütkepohl (1991) — Introduction to Multiple Time Series Analysis. Defines the asymptotic distribution of OLS in levels VARs with I(1) variables; establishes that the covariance matrix is singular in the cointegrated case; covers VECM estimation, lag selection, impulse responses, and forecast error variance decomposition.
- Dolado and Lütkepohl (1996) — Shows that a Wald test for zero restrictions on VAR coefficients has standard χ2 asymptotics if at least one complete coefficient matrix Ai is unrestricted under H0. Practical implication: augmenting the VAR by one extra lag makes all Granger causality tests valid regardless of cointegration properties. See Granger Causality.
- Saikkonen and Lütkepohl (1997, 1998, 1999) — Trend-adjustment prior to Johansen rank testing: subtract GLS-estimated trend from yt first, then test on the demeaned/detrended series. Achieves higher local power than standard Johansen tests under trend assumptions in Cases 2–4.
- Lütkepohl (1999) — Textbook-style survey chapter: systematizes the five deterministic-trend cases for Johansen rank tests (Table 1), shows the finite-sample lag-order inequality p^(SC)≤p^(HQ)≤p^(AIC), and distinguishes direct coefficient zeros from true Granger noncausality in higher-dimensional systems.
Related