(Consistently) testing strict exogeneity against the alternative of predeterminedness in linear time-series models

A portmanteau and a maximum statistic are proposed to discern strictly exogenous from predetermined regressors. Both test statistics rely on residual cross-correlations, pre-whitened via autoregressive sieves. Limiting normality of the portmanteau statistic and Gumbel-convergence of the maximum statistic are derived. (C) 2020 Elsevier B.V. All rights reserved