Réponses individuelles aux enquêtes de conjoncture et prévision de la production manufacturière

In this paper, we compare the performances of balances of opinion with those of competing dis-aggregate indicators derived from the Mitchell, Smith and Weale (MSW) methodology as concerns the one-quarter forecasting of the manufactured production growth rate. The data used are the Business Tendency Survey in Industry carried out by INSEE and the French Quarterly Accounts. The MSW indicators are derived from the entrepreneurs' individual responses to the tendency-survey questions relating to either past or expected production. They are of two kinds: the parametric indicators are based on ordered discrete-choice models at the production-unit level (polytomic logit models); the other indicators are derived from a simpler non- parametric methodology. MSW's applications to British and German data suggest that their indicators perform better than a set of three competing aggregate indicators, but the authors do not obtain the same result on Portuguese and Swedish data. Based on a large panel and a thorough out-of-sample analysis, the application to French data concludes that the forecast performances of the MSW indicators are less good or, at best, equivalent to those of the balances of opinion, depending on the models used.