A0188
Title: Combining caterpillar-SSA methods and mixed frequency data regression for inflation forecasting
Authors: Elena Zarova - Tashkent State University of Economics (Uzbekistan) [presenting]
Abstract: Obtaining reliable inflation forecasts is important for any type of economy and level of economic development. Officially published monthly consumer price indices lag behind the actual market situation for a period of 1-2 months, which is critical for decision-making in the financial and real sectors of the economy, as well as for the competent economic behavior of households. Reducing this gap as much as possible is very important in conditions of economic instability. A possible approach to solving this problem, which has scientific novelty, is based on a combination of the Caterpillar - SSA (Singular Spectrum Analysis) method and the MIDASR (Mixed Frequency Data Sampling Regression Models) method. The results of multivariate forecasting of weekly consumer price indices for individual goods using the SSA method are the input for constructing a regression model of the monthly CPI for food products using MIDASR methods. Based on statistical information criteria, it is concluded that this method provides a more reliable leading estimate of the CPI relative to other multifactor modeling methods. Practical examples of solving this problem using data from a number of countries and R packages are given. The proposed method has practical importance for many forecasting problems.
