Download or read book Nonparametric Efficiency Testing of Asian Stock Markets Using Weekly Data written by Cornelis A. Los and published by . This book was released on 2001 with total page 30 pages. Available in PDF, EPUB and Kindle. Book excerpt: The efficiency of speculative markets, as represented by Fama's 1970 fair game model, is tested on weekly price index data of six Asian stock markets - Hong Kong, Indonesia, Malaysia, Singapore, Taiwan and Thailand - using Sherry's (1992) non-parametric methods. These scientific testing methods were originally developed to analyze the information processing efficiency of nervous systems. In particular, the stationarity and independence of the price innovations are tested over ten years, from June 1986 to July 1996. These tests clearly show that all six stock markets lacked at least one of the two required fair game attributes, and, accordingly, Fama's Efficient Market Hypothesis must be rejected for these Asian markets. However, Singapore emerged from these tests as the most efficient regional Asian stock market. A tentative ranking in order of stock market efficiency is: Singapore, Thailand, Indonesia, Malaysia, Hong Kong and Taiwan. Singapore's stock market pricing is closest to the speculative market behavior which can support stock options. Our tests show both Hong Kong and Taiwan to be inefficient markets. Both exhibit non-stationary (likely because of continuing institutional changes) and dependent price innovations, making them particularly unsuitable for stock option pricing. In Taiwan the weekly price innovations show even higher order (Markov) dependencies. Although the price innovations in Malaysia, Thailand and Indonesia are at least stationary at the weekly level, they exhibit regular higher-order transitions and the large sustained movements in both bull and bear markets, which are so characteristic for illiquid emerging markets. All six Asian stock markets exhibit strong price trend behavior, which, perhaps, can be profitably exploited by technical analysis with first-order Markov filters (e.g., Kalman filters) in windows of between a week and more than a month.