Balance of opinion What about missing the weights?

Due to their early release, Business Tendency Surveys (BTS) are widely used in short term forecasting. Their questions are mainly qualitative; answers are most often used to calculate balances of opinions, which are defined as the difference between the proportions of positive answers with respect to the negative ones. These indicators are then used by forecasters as explanatory variables in econometric models. The balances of opinions are generally weighted with the firm size. However, there is no theoretical evidence of the efficiency of this kind of weighting. We propose here a model which aims at determining optimum weights; these weights should allow us to optimize the forecast of the macroeconomic variable. According to our analysis, the weights have to grow less than proportionally with the firm size. This conclusion is empirically tested through several examples derived from the French Industry BTS.