Hypothesis testing with error correction models

Abstract Grant and Lebo (2016) Keele et al. clarify the conditions under which popular general error correction model (GECM) can be used interpreted easily: In a bivariate GECM data must integrated in order to rely on coefficient, $\alpha _1^\ast$ , test cointegration measure rate of between single exogenous x dependent variable, y . Here we demonstrate that even if are all integrated, is misunderstood when there more than independent variable. The null hypothesis no any but correct alternative cointegrated with at least one—but not necessarily one—of 's. A significant occur some I (1) regressors equation balanced. Thus, limiting distributions right-hand-side long-run coefficients may unknown. We use simulations problem then discuss implications for applied examples.