The first step in modelling households consumption is to determine the aim of the model. Different structures are appropriate if we try to understand the behaviour of households or build a forecasting framework. In this paper we present those models which are able to support forecasting of Hungarian consumption at the National Bank of Hungary. As our primary goal is forecasting, we evaluate the different models by their forecasting accuracy and stability. The expected accuracy depends on the information available at a given time. In order to exploit all the relevant information we use proxy variables to avoid publication delays and problems of original variables and at the same time incorporate unobserved factors as well. Liquidity constraint and uncertainty of households can not be observed directly, but we argue that these factors can be proxied by other time series. Our results show that these variables help predicting household consumption in Hungary.

Available only in Hungarian