Predicting bear and bull stock markets in Thailand with dynamic binary time series models

This paper aims to study the predictability of the future bear and bull in Thailand stock markets using dynamic binary time series models. The monthly data sets of financial and macroeconomic variables ranging from 1990 and 2013 comprise of money supply, inflation rate, crude oil price, manufacturing production index (MPI), interest rate (RP1Day), term spread and real effective exchange rate (REER). The bear and bull periods in stock market can be identified by applying nonparametric methodology based on Bry-Boschan (1971) turning point dating rule. We compare the performance between static probit model and dynamic probit models and find out that interest rate is the most powerful variable for predicting bear and bull stock market in Thailand and dynamic autoregressive probit model is the best in term of forecasting accuracy among alternative models.

本研究旨在采用动态二元时间序列模型，探究泰国股票市场未来熊市与牛市的可预测性。本研究使用1990年至2013年间的月度金融与宏观经济变量数据集，涵盖货币供应量、通货膨胀率、原油价格、制造业生产指数（MPI）、利率（RP1Day）、期限利差以及实际有效汇率（REER）。泰国股票市场的熊市与牛市周期，可通过基于Bry-Boschan（1971）转折点定年规则的非参数方法予以识别。本研究对静态概率单位模型（probit model）与动态概率单位模型的表现进行了对比，结果表明，利率是预测泰国股市熊市与牛市行情的最具效力的变量，且动态自回归概率单位模型在各类备选模型中预测精度最优。