Does industry survey data improve GDP forecasting?

Abstract

This study assesses the integration of industry survey data into Bayesian Vector Auto Regressive (BVAR) models for GDP forecasting in Sweden. Analyzing a combination of macro economic indicators, CPI and unemployment rates, with survey data from NIER, it explores the effects of different variable combinations on the forecasting ability of different models. The research concludes that some forward looking survey data boosts short term forecasting performance in BVAR models, especially expected sales price in the private sector and expected sales in the trade sector. Key findings include the superior predictive capability of certain variable combinations, most significantly the model consisting of expected sales price in the private sector, expected number of employees in the private sector and expected sales in the trade sector. The research offers insights for refining BVAR models and the incorporation of survey data to achieve more precise GDP forecasts.