A Study on spurious long memory in nonlinear time series models

This paper discusses the existence of spurious long memory in common nonlinear time series models, namely Markov switching and threshold models. We describe the asymptotic behavior of the process in terms of autocovariance and autocorrelation function and support the theoretical evidences by providing Monte Carlo simulation. The existence of long memory in these nonlinear processes is induced by the nature of the process in certain conditions. In addition, GPH estimator itself introduces bias.