Forecasting Macroeconomic Variables Using Disaggregate Survey Data

We propose to construct factor models based on disaggregate survey data to forecast national aggregate macroeconomic variables. We apply our methodology to Norges Bank’s regional survey, which allows to construct regional and sectoral factor models, and to the Swedish Business Tendency survey, which allows to derive sectoral factor. The analysis identifies which information extracted from the regions and the sectors performs particularly well at forecasting different variables and horizons. Results show that several factor models beat an autoregressive benchmark in forecasting inflation and unemployment rate. However, the factor models are most successful in now casting and forecasting GDP growth. Forecast combinations of regional and sectoral factor models based on past performance give in most cases the most accurate forecasts.