| This dissertation investigates long memory modeling of conditional means and variances in high frequency asset pricing data and monthly inflation series. Chapter 2 introduces a "generic" long memory model, or ARFIMA - FIGARCH model which captures the property of long memory in both the conditional mean and in the conditional variance of a univariate time series. The properties of the process are discussed and parameters of the model are estimated by QMLE. A detailed simulation study reveals the QMLE procedure to work well. Chapter 3 discusses the application of the methodology to monthly US CPI inflation and to weekly Canadian/US forward premium series. Chapters 4 and 5 examine temporal aggregation of assets returns. Chapter 4 reports estimation of the FIGARCH models at daily and weekly frequencies and then discusses theoretical results concerning the properties of aggregated data. Chapter 5 also provides details of the estimation of long memory volatility, FIGARCH models to two, three, six, and twelve hour exchange rate return series. |
