INCORPORATING GOOGLE TRENDS AS BIG DATA FOR ENHANCED INFLATION FORECASTING: EVIDENCE FROM KAZAKHSTAN

Simple item page
dc.contributor.authorBeisenbek, Rakhat
dc.date.accessioned2024-12-23T09:16:46Z
dc.date.available2024-12-23T09:16:46Z
dc.date.issued2024-12-11
dc.description.abstractAccurate inflation forecasting is essential for policymakers and businesses, allowing for informed decision-making and economic stability. Most historical forecasting methods are based on a few key macroeconomic factors while leaving a wide variety of more detailed data from the big data sources unused. This paper aims to analyze the performance of using Google Trends data on improving the inflation forecast of Kazakhstan and compare the result with the traditional macroeconomic models. The research develops two primary models: a baseline model that includes conventional macroeconomic indicators such as GDP growth, oil prices, NEER, and inflation expectations, and an enhanced model that integrates Google Trends data for key terms like "inflation," "GDP," and "exchange rate." These data are preprocessed with standardization, lagging, and percentage change transformations. Machine learning techniques, specifically random forest and gradient boosting regressors, are applied to evaluate model performance. Statistical validation includes Likelihood Ratio tests for out-of-sample density forecast accuracy, as well as the Mean Squared Error (MSE), Mean Absolute Error (MAE) evaluation metrics calculation, Mincer-Zarnowitz regression for bias, and the Diebold-Mariano test for forecasting accuracy. Findings show that incorporating the set of variables of Google trends improves the forecasting accuracy of the enhanced model by making relatively small MSE and MAE compared to the baseline. The Likelihood Ratio test supports the improvement of the models for density forecasting, and in terms of feature importance, Google Trends data turn out to be critical for the enhanced model. While the result of the Diebold-Mariano test turned out to show the marginal significance, extending the dataset period and applying advanced techniques further maintained the robustness of the enhanced model. This research proves that Google Trends as a big data contributes to enhancing the accuracy of the inflation forecasts for developing economies.Although sentiment analysis was initially considered, it was excluded from the study due to data limitations in Google news and the absence of an access tokens for media sources like Facebook.
dc.identifier.citationBeisenbek, R. (2024). INCORPORATING GOOGLE TRENDS AS BIG DATA FOR ENHANCED INFLATION FORECASTING: EVIDENCE FROM KAZAKHSTAN. Nazarbayev University Graduate School of Business
dc.identifier.urihttps://nur.nu.edu.kz/handle/123456789/8366
dc.language.isoen
dc.publisherNazarbayev University Graduate School of Business
dc.rightsAttribution-NonCommercial-ShareAlike 3.0 United Statesen
dc.rights.urihttp://creativecommons.org/licenses/by-nc-sa/3.0/us/
dc.subjectGoogle Trends
dc.subjectInflation Forecasting
dc.subjectForecasting Accuracy
dc.subjectPredictive models
dc.subjectType of access: Embargo
dc.titleINCORPORATING GOOGLE TRENDS AS BIG DATA FOR ENHANCED INFLATION FORECASTING: EVIDENCE FROM KAZAKHSTAN
dc.typeMaster`s thesis

Files

Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
RAKHAT_BEISENBEK_MSF_THESIS.pdf
Size:
822.29 KB
Format:
Adobe Portable Document Format
Description:
Master's thesis
Download
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
6.28 KB
Format:
Item-specific license agreed upon to submission
Description:
Download

Collections

02. Master's Thesis