CFE 2015: Start Registration
View Submission - CFE
A1572
Topic: Contributions on evaluation of forecasting Title: Micro information dynamics: Decomposing the forecasting power of aggregate indicators Authors:  Klaus Wohlrabe - ifo Institute for Economic Research (Germany) [presenting]
Abstract: Many popular business indicators, such as the Ifo Business Climate, are based on surveys. Surveys collect information from often very heterogeneous respondents. In aggregate indicators, however, this rich panel information remains mostly unused. Panel information is used, in particular, differences in assessment and predictive ability between groups, in two ways: First, to examine which subset of respondents contributes most to the forecasting performance of an aggregate indicator and whether focusing on certain subsets can improve it. Second, to explore how panel information can improve the forecasting performance of an aggregate indicator based on the full sample. We classify respondents into subgroups based on size, sector, and characteristics of their past survey responses. Furthermore, we introduce dispersion and disagreement measures based on micro survey data. Our focus is forecasting industrial production in Germany with the large micro data set underlying the Ifo Business Survey. We demonstrate that well-defined subgroups of firms convey a large share of the information embedded in the aggregate indicator. Most importantly, exploiting the heterogeneity in responses between subgroups improves the forecasting accuracy.