No-Arbitrage Near-Cointegrated VAR(p) Term Structure Models, Term Premia and GDP Growth
35 Pages Posted: 10 Feb 2009 Last revised: 20 Jan 2011
Date Written: January 2011
The recent macro-finance yield curve literature has focused on the extraction of a reliable measure of term premia (on long-term bonds) and on their relevant relationship with future economic activity (GDP), because of the practical implications of this relationship for the conduct of the monetary policy. However, the associated empirical findings do not agree neither about term premia empirical properties nor about the importance or even the direction of the above mentioned relationship.
The present paper proposes a two-step approach to handle both problems. First, in a VAR setting, we extract a reliable measure of the term premia by means of averaging estimators techniques aiming at optimally solving prediction problems when highly persistence processes are present and, thus, providing a so called Near-Cointegrated VAR(p) approach. Second, we analyze the dynamic response of the GDP to shocks on the term premia by using the New Information Response Function concept allowing, in particular, to deal with shocks on variables which are filters of the basic ones in the model. Results in both steps are quite different from those appearing in the literature because of a careful treatment of persistence, in the first step, and of the number of lags in the second one.
First, we find, coherently with the typical macroeconomic view, and in contrast with OLS-based VAR decompositions, that the NCVAR-based term premium measure is rather stable and contra-cyclical, with the expectation part accounting for most of the yield variability.
Second, we find that an increase of the long-term spread caused by a rise of a term premium induces two effects on future economic activity: the impact is negative for short horizons (less than one year), whereas it is positive for longer ones. Therefore, this result suggests that the above mentioned ambiguity could come from the fact that the sign of this relationship is changing over the period that follows the shock.
Keywords: Averaging Estimators, Persistence Problem, Near-Cointegration Analysis, No-Arbitrage Affine Term Structure Model, Term Premia, GDP Growth, New Information Response Functions
JEL Classification: C51, E43, E44, E47, G12
Suggested Citation: Suggested Citation