Extensions of the Pesaran, Shin and Smith (2001) Bounds Testing Procedure
46 Pages Posted: 22 Feb 2021
Date Written: December 06, 2020
We replicate the Pesaran, Shin and Smith (2001) bounds testing procedure (BTP), and extend it with 6 new cases, 4 of which involve a quadratic trend. We provide critical values for the BTP of the lagged regressors in levels under the framework of unrestricted error-correction models (UECMs) to account for degenerate cases of co-integration. Further, we extend the BTP with 11 cases for the quantile UECMs of Cho, Kim and Shin, and present critical values for inter-decile and interquartile BTPs for the unrestricted cases. We extend the Shin, Yu and Greenwood-Nimmo methodology to account for non-linear, or asymmetric, responses of the dependent variables to its covariates (NARDL) and for distributional, or location, asymmetry (QARDL of Cho, Kim and Shin; of the dependent variable. We call this quantile non-linear ARDL, or QNARDL. We provide codes that generate sample-specific critical values of the BTPs. We utilize these critical values in an empirical application of a dynamic equity valuation model for the S&P Global Index. We find that mis-specifying a non-linear relationship as linear produces misleading results and policy implications. There is strong evidence of: (i) trading activity based on fundamentals and (ii) the existence of a stable equilibrium relationship for the price-to-book (PB) ratio of the market index and its fundamentals. During periods of high PB relative to its fundamental values, convergence to equilibrium is faster than during periods of relatively low PB. There is also evidence of momentum trading, i.e. of traders that rely on positive feedback.
Keywords: Asymmetric Co-integrating Relationships, Bounds Testing Procedure, Quantile Non-linear ARDL (QNARDL), S&P Global Index, Unrestricted Error-correction Models
JEL Classification: C12, C22, E32, G15, G40
Suggested Citation: Suggested Citation