First-difference estimator for panel censored-selection models

We study how processes with infrequent regime switching may generate a long memory effect in the autocorrelation function. In such a case, the use of a strong fractional I(d) model for economic or financial analysis may lead to spurious results.