Effects of a signal-to-noise ratio on finite sample inference for cointegrating vectors

This paper investigates effects of a signal-to-noise ratio on finite sample inference for cointegrating vectors. The ratio is defined as a measure of the magnitude of a permanent shock relative to a transitory shock. According to Monte Carlo experiments conducted in this paper, a high signal-to-noise ratio tends to reduce size distortions of a likelihood-based test statistic for a hypothesis on cointegrating vectors; a low signal-to-noise ratio is, in contrast, prone to amplify the size distortions. The experiments demonstrate that the performance of a bootstrap method also depends on the volume of the signal-to-noise ratio. Finally, an empirical illustration is presented.