Testing for short-run threshold effects in a vector error-correction framework: a reappraisal of the stability of the US money demand

Abstract

Abstract In this paper we analyze the stability of the money demand system in the US. To this aim, we develop an estimation and testing framework for a threshold vector error-correction model (VECM), where short-run dynamics are regime dependent and are driven by an exogenous, stationary and ergodic threshold variable. We modify a traditional Wald-type test for linearity and derive its asymptotic distribution, which turns out to be non-standard, but similar to the one proposed by Andrews [Andrews, D. 1993. “Tests for Parameter Instability and Structural Change with Unknown Change Point.” Econometrica 61 (4): 821–856]. Accounting for the diversity of the potential determinants of money holdings, a broad scope of money demand systems is analyzed. The linearity test confirms that money demand is not only suffering from structural breaks, but is driven by regimes governed by business cycles. Moreover, a similar pattern and a robust dating of the regimes is observed no matter the money demand system considered, offering hence a reconciling picture of previous empirical studies. From a policy perspective, our findings supports the implementation of cyclical monetary policy by the Federal Reserve.