Testing the 'Inaction Corridor' in a Three-Regime Threshold Error Correction Model with an Application to a Buffer-Stock Model for US Money Demand

In this paper, we develop a test for the existence of a middle 'band of inaction' in a three-regime threshold vector error correction model (TVECM). A Wald statistic is proposed for this purpose, its limiting distribution is derived (which is non-standard) and critical values calculated through simulations. Our methodology is applied to estimate a buffer-stock model for the US money demand and to test for the existence of an inaction middle range within which the money balance is free to fluctuate without portfolio adjustments. Our data seem to corroborate this hypothesis