A NEW TEST FOR CHAOS

Whether certain financial and economic time series are better described by linear stochastic models or are appropriately characterized by deterministic chaos is an issue of great current interest. Empirical research to detect the presence of chaos in such time series has been hampered by lack of an adequate test for chaotic behavior in small, noisy data sets. This paper presents a new, topological test for chaos, demonstrates its advantages over current testing methodology, and compares the results with earlier analyses based on metric tests of economic and financial series. A short-term forecasting approach for chaotic processes is also described.