Nowcasting Turkish GDP by dynamic factor model

Abstract

This study aims to nowcast, backcast, and forecast the quarter of Turkish Gross Domestic Product (GDP) growth by the dynamic factor model. Although GDP provides detailed information on economic activities, GDP data is released with a significant delay. This thesis constructs a real-time dataset of monthly economic activity indicators to measure the current GDP. Variables used in the estimation of the Turkish GDP are real, financial, and survey indicators. The dynamic factor model is utilized in the estimation because that model can tackle the issues of mixed frequency (quarterly and monthly variables in the dataset), ragged ends (nonsynchronous data publications), and missing data (the upper side of some indicators in the dataset is not same). The model developed by Giannone, Reichlin, Small (2008) and Banbura, Giannone, and Reichlin (2011) was adopted in the study. The thesis evaluates the performance of the dynamic factor model in nowcasting and the ARIMA Model in forecasting the quarterly Turkish real GDP growth rate for the period 2020-2021. Results show that the more data are released, the accuracy of the model increases in the dynamic factor model. On the other hand, no noticeable improvement was observed in the accuracy of the ARIMA model.