Sentiment Indicators Based on a Short Business Tendency Survey

18/09/2019 |  Suhoy Tanya, Roash Daniel
All Press Releases In Subject:
The Economy and Economic Activity

Abstract

Updated version May 2020  

In the post 2008-09 crisis period the variance of real growth has declined and the information content of qualitative surveys has changed. We report on a recent attempt to assess the reliability of the Business Tendency Survey, conducted in Israel since 2011, and its usefulness for nowcasting.

At the firm level, the relationship between soft and hard data has been explored through logistic regression with firms' random effects, based on matched datasets that merge firms' qualitative evaluations of past sales with their revenue, collected from administrative sources. We document an improvement in the reliability of qualitative responses since the questionnaire was focused on month-over-month evaluation. Estimated marginal effects show that the probabilities of the "Up" response have been influenced the most. Our results suggest a significant effect of intra-sectoral firms' heterogeneity, particularly in Services.

On a macro level, we aggregate sectoral balances of opinions with monthly updated regression-based weights, thus enabling early estimates of underlying monthly GDP growth. Mutual correlations between survey covariates are handled through PLS-regression and variable selection based on relative importance scores. This Sentiment indicator explains between 21 and 48 percent of in-sample GDP-variance and suggests a reasonable out-of-sample fit, if we target the first GDP estimate. Regarding final GDP estimates, the model does not outperform an assumption of average growth over the test period between 2016 and 2019.

Keywords: Business-tendency survey, Sentiment indicator, Partial Least Squares, Monthly GDP

JEL: E32, E37​


To the​ Full Article as PDF​ file​​ ​