Testing Change in Time Series

Atsushi Inoue - University of Pennsylvania

This paper proposes nonparametric tests of change in the marginal distribution of a financial time series. Because the null limiting distributions of the KolmogorovSmirnov and the Cramer-Von Mises statistic depend on a nuisance parameter, we cannot tabulate critical values a priori. In order to circumvent this problem, we develop a new simulation-based statistical method, called the "wild block bootstrap". We prove the validity of our bootstrap approach in terms of size, local power and test consistency. We evaluate the finite-sample properties of the proposed tests in a set of Monte Carlo experiments, and we examine structural stability in financial markets, including equity and foreign exchange. We find evidence of structural instability in equity and interest rates.

Scheduled for Session 1.2 Time Series - I