Nonparametric Forecasting of the Manufacturing Output Growth with Firm-level Survey Data

A large majority of summary indicators derived from the individual responses to qualitative Business Tendency Surveys (which are mostly three-modality questions) result from standard aggregation and quantification methods. This is typically the case for the indicators called balances of opinion, which are currently used in short term analysis and considered by forecasters as explanatory variables in many models. In the present paper, we discuss a new statistical approach to forecast the manufacturing growth from firm-survey responses. We base our predictions on a forecasting algorithm inspired by the random forest regression method, which is known to enjoy good prediction properties. Our algorithm exploits the heterogeneity of the survey responses, works fast, is robust to noise and allows for the treatment of missing values. Starting from a real application on a French dataset related to the manufacturing sector, this procedure appears as a competitive method compared with traditional algorithms.