Bias correction for the regression-based LM fractional integration test

Abstract

Long memory testing, Small-sample behavior, Bias approximation,

Asymptotic normal tests for integration in panels with cross-dependent units

“…This behavior has two causes. First, the single test statistics are biased in small samples (Demetrescu and Tarcolea 2008), and the bias cumulates over N units. Second, the chosen loadings imply relatively heterogeneous cross-correlations; this departure from Hartung's assumptions adds to the distortions.…”