Modeling and estimation of economic time series primarily concern making inferences about the intertemporal law of motion of economic variables. The general objective of this project is to refine and apply the seminonparametric approach (SNP) for estimation of the underlying law of motion of a stationary process. The SNP approach was initially developed by Tauchen and Gallant for an asset pricing application. It can accomodate just about any arbitrary deviations from linearity and normality and, unlike other methods, provides a compact summary of the data in the form of a fitted parametric density which can be used for the purpose of the formal testing of interesting economic hypotheses. The project consists of a methodological section and an empirical section. The methodological work is aimed at gaining further understanding of the statistical properties and the reliability of SNP estimators. This work entails both Monte Carlo simulation and derivation of analytical results. The empirical work entails application of the SNP methodology to issues related to labor market dynamics, short-term price movements on financial markets, and nonlinear causality between money and income. The main research tool for the empirical work is nonlinear maximum likelihood estimation. The results of the project are expected to be useful to empirical economists and to econometricians and statisticians interested in time series density estimation. Their particular appeal is that they are not linked to a specific family of nonlinear models; rather the approach is nonparametric, in the sense that the procedures are in principle capable of capturing a wide variety of nonlinearities that have previously been examined only parametrically or only with very limited tools.
