Building confidence intervals and reassessing the past with the Mésange model

The Mésange model, a macro-econometric model co-developed by INSEE and the French Treasury, simulates the response of the French economy to different types of changes in its environment. The aggregate behaviors of agents are described by about forty estimated econometric equations, which determine the dynamic response of the model to a shock, in deviation from the balanced growth path. This study analyzes the uncertainty of the model arising from its estimation through two approaches. The first one aims at constructing confidence intervals for the response functions of the model using a non-parametric bootstrap method. A resampling algorithm – which randomly draws from the series of econometric residuals to simulate alternative paths for the model – is used to quantify the uncertainty linked to the estimation of certain coefficients and the misspecification linked to the model itself. This first approach confirms the significance of the dynamic effects of the simulations carried out with Mésange for most macroeconomic variables. The second approach offers a reassessment of the 2008 economic crisis based on alternative scenarios randomly simulated. These results allow us to analyze and confirm the robustness of the Mésange model, by making a methodological contribution to measuring the uncertainty arising from the simulation when compared with a past trajectory. They are based on a thorough preliminary study of the statistical properties of the residuals of the econometric equations, in order to justify the use of the bootstrap. This study also highlights some points of attention for a future re-estimation of the model.