Économie et Statistique n° 451-452-453 - Macroeconomic modeling : continuity, tensions

Economie et Statistique
Paru le : 03/01/2013
Consulter

The New Version of MZE, a Macroeconometric Model for the Euro Zone: Confidence Intervals for Checking Simulation Results

Muriel Barlet, Marie-Émilie Clerc, Marguerite Garnero, Vincent Lapègue et Vincent Marcus

MZE (French acronym for “Modèle Zone Euro” [Euro Zone Model]) has evolved since its initial version in 2003. Our article presents the new version. By backcasting the series to 1980-1991, the designers have lengthened its estimation period from 1991-2001 in the previous version to 1980-2008Q2. The model's structure remains neoclassical in the long term but contains Keynesian elements in the short term. In the short run, activity is determined by demand, and prices and wages adjust gradually. In the long run, the aggregates corroborate relationships derived from neoclassical theory. As the equation coefficients are estimated imprecisely, coefficient uncertainty is reflected in the simulation results. To obtain confidence intervals for assessing their degree of precision, this article proposes an initial application for an operational macroeconometric model of Kilian's non-parametric bootstrap method (1998). The first step in the method is to reveal the existence of coefficient estimation biases by calculating confidence intervals. The model then uses this information to re-estimate the coefficients and new confidence intervals centred on them. The analysis of simulation results becomes more nuanced as a result, since we can assess the significance of their divergence from the baseline scenario over all horizons. Simulated with the coefficients estimated as described above, the model's new version offers less inflationary responses to standard macroeconomic shocks than the 2003 version. Taking the example of the response to a 1% growth in world demand, GDP rises 0.18% in the first quarter, with the 95% confidence interval ranging from 0.13% to 0.26%. Nominal wages increase 1.87% in the long term (versus 4.44% in the 2003 version), with the 95% confidence interval ranging between 1.67% and 2.10%.

Economie et Statistique
No 451-453
Paru le : 03/01/2013