engleski

Forecasting Croatian quarterly real GDP with monthly monetary aggregate M1 data: A mixed frequency approach (MIDAS)

Forecasting macroeconomic variables such as gross domestic product (GDP), consumption, investments etc. is one of the most important tasks in predicting business cycles. Problems arise from the fact that many macroeconomic indicators are available only as lower frequency data and forecasting models generally require data to be of the same frequency. For example, GDP data is sampled quarterly, money and inflation are sampled monthly, while most interest rate data is sampled daily. Mixed Data Sampling (MIDAS) regression estimation technique overcomes these problems, enabling data sampled at different frequencies to be used in the same regression. Therefore, in this paper, Croatian quarterly real GDP is forecasted using monthly monetary aggregate M1 data and the MIDAS approach with the Beta and Almon weighting methods. An estimated MIDAS model using the Beta weighting method shows that monthly monetary aggregate M1 has an overall positive effect on quarterly real GDP, but the effect disappears quickly in lags. On the other hand, the estimated MIDAS model using the Almon weighting method shows that the overall effect of monthly monetary aggregate M1 on quarterly real GDP is also positive and decreasing in lags. However, in contrast with the Beta weights estimate, the lag coefficients show a linear pattern. In addition, computed static forecasts from MIDAS models show that monthly monetary aggregate M1 can be used for predicting Croatian quarterly real GDP.

business cycles, MIDAS approach, mixed frequency data, monetary policy

nije evidentirano