Assessing the predictive capability of google search based negative sentiment indices on Philippine inflation
Date
2025-01
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This study evaluates whether internet search-based sentiment indicators can improve monthly inflation forecasts in the Philippines. We construct four Negative Sentiment Indices (NSIs) using Google Trends data on inflation-related search terms, namely, “price increase and price decrease”, “price hike” and “price rollback”, “pagtaas ng presyo” and “pagbaba ng presyo”, and “inflation and deflation”. These NSIs are designed to capture public concern over rising prices. Additionally, we generate a composite index (PC1) through principal component analysis to summarize the common trend across the four NSIs. Using Autoregressive Distributed Lag (ARDL) models, we assess the forecasting power of these sentiment measures. Cointegration tests show that NSI_2 is cointegrated with monthly inflation at the 10% significance level. However, over a 12-month rolling forecast horizon for 2021, ARDL models that include these variables generally fail to outperform a simple AR(1) benchmark. In terms of RMSE and Diebold-Mariano test results, most NSI-based models yield weaker predictive performance. These findings suggest that while sentiment data may contain useful information, their forecasting value in the Philippine context remains limited. Future research could explore mixed-frequency models using weekly Google Trends data or incorporate richer sentiment sources such as social media and news analytics to improve inflation forecasting in emerging markets.
Description
Keywords
inflation forecasting, Google Trends, negative sentiment index, ARDL, Diebold-Mariano Test