Analyzing the relationships between survey forecasts for different variables and countries

This paper evaluates the quality of survey forecasts, their accuracy and unbiasedness, and their overall consistency. The paper also tries to find out whether the relationships between economic variables are the same in survey data and in the actual data. In other words we analyze whether the data generating mechanisms of forecast values and actual data coincide. The analysis deals with three countries/economic areas: the Euro Area, Japan and the US and makes use of different surveys and data frequencies. Since the results are somewhat blurred by the recent 2008-2010 financial crisis thus inclusion of the crisis period makes a lot of difference in main results. Even so, we find that the basic features of the data have quite few alarming features. Different surveys come quite close to each other and results for different countries/economic areas are reasonably similar. It is only that we find some evidence that the relationships between economic variables in the survey data are different from actual data. Moreover, we find that forecast errors are quite closely related to dispersion of survey respondents’ forecasts. Thus, increased forecast uncertainty seems to be positively related to size of forecast errors.