FORECASTING ECONOMIC GROWTH RATE WITH MIXEDFREQUENCY DATA

In time series methods, both dependent and independent variables must be at the same frequency. This requirement could especially cause some serious problems in forecasting macroeconomic series because of publishing macroeconomic data at different frequencies to public. However, in their study of 2004, Ghysels, Santa-Clara and Valkanov have eliminated this problem by developing the MixedData Sampling (MIDAS) approach. MIDAS approach is a method which allows us to use data sampled at different frequencies together. In this study, MIDAS application was carried out to forecast economic growth with alternative yield spreads. In the empirical study which covers the period of 2010-2017, economic growth rate is at quarterly frequency whereas yield spreads are at monthly and weekly frequencies. Because of the differences at the frequencies of the variables, MIDAS approach was used. In this context, economic growth rates of the first three quarters of 2017 were forecasted for Turkish economy. Economic growth forecasts were evaluated in terms of performance values of the specified model