Nonlinearity, nonstationarity, and thick tails: How they interact to generate persistence in memory

We consider nonlinear functions of random walks driven by thick-tailed innovations. Nonlinearity, nonstationarity, and thick tails interact to generate a spectrum of autocorrelation patterns consistent with the observed persistence in memory of many economic and financial time series. Depending upon the type of transformation considered and whether it is observed with noise, the autocorrelations are given by unity, random constants, or hyperbolically decaying deterministic functions, possibly with some independent noise, and thus may decay slowly or even not at all. Along with other sample characteristics, such patterns suggest that these three ingredients may generate the ubiquitous evidence for long memory.