Serial number: 01/2021
Author(s): Nazrin Ramazanova, Ramiz Rahmanov
Language: English
Date: 2021
Abstract: The paper investigates the performance of various combinations of the single predictor MIDAS models and DFM model to nowcast household consumption in Azerbaijan. We evaluate the out-of-sample forecasting performance at horizons of up to four quarters ahead. All models use the information on monthly indicators released ahead of quarterly household consumption. The results show that both DFM model and single predictor MIDAS models with high-frequency regressors of retail trade turnover and Google Trends (arts & entertainment, sports, travel, hobbies & leisure) outperform the univariate benchmark model for one quarter ahead forecast. However, in comparison with MIDAS models, DFM shows better forecast performance.
Key words: nowcasting, household consumption, MIDAS, DFM
JEL classification: C32, C38, C52, C53, E21, E27