Author: Kristine Joy E. Carpio

Year: 2018, Volume 28 No. 1
Pages: 49-66


The future of the stock market may never be predicted consistently, nor its past behavior understood entirely, but any knowledge gained from observing it could help decide on a sound investment strategy. In this study, I looked at the daily returns of the Philippine Stock Exchange index (PSEi) from March 1, 1990, to January 31, 2017, and see how the data relates to the mathematically verifiable aspects of the noise theory and efficient market theory (EMT). In relation to the noise theory, I looked at the occurrences of anomalies. For the EMT, I made use of discrete-time Markov chains to determine some trends. The study results showed that most stock market anomalies are present while persistent behavior is hardly present in the dataset. Furthermore, I applied day ahead time domain forecasting methods starting with the simple moving average models to autoregressive moving average models. The augmented Dickey-Fuller test indicate that the daily returns are a stationary series although the ACF and PACF plots have consistently shown non-zero correlations for lags 1, 9, 12, 13. I have obtained AR(1) and ARMA(1,2) processes for the data and both models indicate the same forecasting accuracy via the Diebold-Mariano test. Although these time domain processes were unable to predict the random noise in the data, these processes were accurate in predicting the signs of the values as supported by the Pesaran-Timmermann test.

Keywords: PSE index, Markov chains, stock market anomalies, long memory