Une comparaison des prévisions des experts à celles issues des modèles B VAR

1999 
[eng] A Comparison of Expert Forecasts with BVAR Model Forecasts by Sandrine Lardic and Auguste Mpacko-Priso . This paper checks whether economic and financial experts forecast macroeconomic and financial variables «better» than alternative techniques and in particular the Bayesian method. The BVAR methodology, presented in detail in Lardic and Mpacko-Priso (1996) and summarised in this paper, is used to generate six-month and twelve-month forecasts of the Consumer Price Index, Industrial Production Index, Standard and Poors 425, and Standard and Poors 500 for two samples. These forecasts are then compared with economic and financial expert predictions as well as with forecasts derived from traditional techniques for the same periods of time. Statistical and economic criteria are used to gauge the different forecasts. We show that the BVAR forecasts are generally better than the individual expert forecasts. We conclude that the BVAR methodology merits being used more than it is at present. [fre] Une comparaison des previsions des experts a celles issues des modeles BVAR par Sandrine Lardic et Auguste Mpacko-Priso . Cet article a pour objet de tester si les previsions des variables macro-economiques et financieres faites par des professionnels de l'economie et de la finance sont «meilleures» que celles issues de techniques alternatives, notamment bayesiennes. Suivant la methodologie BVAR, les previsions sur les horizons de six et douze mois des variables Indice des Prix a la Consommation, Indice de la Production Industrielle et Indice Standard and Pors 425 ou Standard and Poors 500 sont generees pour deux echantillons distincts. Ces previsions sont ensuite comparees a celles des experts de l'economie et de la finance d'une part et a celles issues des techniques traditionnelles de prevision d'autre part. Cette comparaison repose a la fois sur des criteres statistiques et economiques.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    17
    References
    0
    Citations
    NaN
    KQI
    []