Forecasting the Philippine Stock Exchange Index Using Time Series Modeling Techniques

Abstract

Forecasting stock market indices like the Philippine Stock Exchange Index (PSEi) is crucial for investors, economists, and policymakers in understanding market behavior and making strategic decisions. This study aimed to examine the historical trend of the PSEi from 2004 to 2023 and determine which among various time series models best predicts its value in 2025. Specifically, the study evaluated polynomial regressions, logarithmic, power series, moving averages, exponential smoothing, and autoregressive models to identify the most suitable forecasting approach. A quantitative research design was employed using secondary data collected from Yahoo Finance and Investing.com. Monthly PSEi closing prices were compiled, averaged annually, and analyzed using Microsoft Excel and the Data Analysis Toolpak, which enabled trendline generation, smoothing applications, and lag-based regression modeling. The results showed that the PSEi experienced an overall upward but volatile movement over two decades, with notable dips during global crises. Among the models tested, the quintic polynomial regression achieved the highest explanatory power, but its predicted value of 15,872.78 suggests potential overfitting. Moving average models effectively smoothed short-term fluctuations but tended to underpredict future growth, while autoregressive models captured significant temporal dependencies, with higher-order lags revealing delayed market responses. The study concludes that while polynomial and curve-fitting models can capture nonlinear behavior, they should be used cautiously due to overfitting risks. It is recommended that future forecasting efforts explore hybrid models that combine polynomial trends, autoregressive structures, and smoothing techniques for improved accuracy.