Testing for efficiency and non-linearity in market and natural time series |
| |
Authors: | Teo Ja i Douglas Wood |
| |
Affiliation: | a Institute of Econometrics, Statistics and Mathematical Finance, School of Economics and Business Engineering, University of Karlsruhe, Germanyb Manchester Business School, Manchester, UK |
| |
Abstract: | Time series in traded markets such as currencies and securities involve supply/demand interaction, so they might be expected to contain distinctive and identifiable structures in comparison with data based on natural phenomena such as river flows or sunspots. This paper tests this proposition using standard econometric tests including variance ratios, modified rescaled range (R/S) ratios and BDS statistics together with non-linear prediction models. Four time series of each type (market or natural) are subject to a battery of tests for random walk and non-linear dependence. Surprisingly, the tests provide no reliable discrimination between the two types of series or reveal any embedded specification differences. |
| |
Keywords: | Efficiency tests non-linearity neural networks market versus natural data |
本文献已被 InformaWorld 等数据库收录! |
|