A short-term Forecasting Model for GDP
Published on December 10, 2021
Published by Macro-Economic Analysis Division
During times of economic instability in and out of the country, an accurate assessment of the economic state is essential for policy responses to be effective. Although Gross Domestic Product (GDP) is the most indicative measure of macroeconomic performance used to make economic projections, its limitations are that it lacks timeliness, because GDP figures are released two months after the end of every quarter. This study sought to improve the accuracy of economic projections by creating a model that forecasts quarterly GDP growth in advance using monthly GDP data.
To this end, based on Principal Component Analysis (PCA) and the Dynamic Factor Analysis (DFA) model, in which monthly data are used to make quarterly data approximations, a quarterly GDP growth estimation model was devised. The duration of analysis was configured from year 2000 to year 2019, and by adding GDP data from the year 2020 - a year of extreme volatility - to present the results of analysis, the impact of COVID-19 on the projection performance was also gauged. The analysis results indicate that when compared with the AR model, which is a benchmarking model that was used to forecast GDP growth with quarterly data, the PCA & DFA model, which was used to estimate GDP growth with monthly data was found to better project GDP growth, because a smaller forecasting error occurred. As such, the PCA & DFA model is expected to contribute to improving the accuracy of economic projections as this shorter-term GDP estimation model makes quarterly approximations of real GDP growth rates prior to the release of actual quarterly GDP growth rates.