EPP – Interpreting macroeconomic time series

Master2 Economics and Public Policy, (Polytechnique, Ensae, Sciences-Po) (from 2009 to 2012)

Teaching assistant to Benoit Mojon, with Guillaume Cléaud

The purpose of this course was to introduce different techniques and tools to analyse the business cycle and economic time series. The main class was given by Benoit Mojon. Guillaume and I were giving the computer based classes. The computer based classes were applying standard quantitative techniques using Octave-Matlab and Dynare: time series analysis, filters, VAR, small DSGE models…

You can download the Exercises imagined for this class, including home assignments and exams.

Description of the computer based classes:

After TD1, students should be abble to program with Octave or Matlab (editor, .m file, saving and loading data, matrix calculations, graphs…).
They should have intuitions about time series framework and know some definitions (stationnarity, ergodicty, AR, MA, randow walk, drift….)

After TD2, they should be abble to code and use Octave-Matlab functions, plot nice graphs with legend, title, colors… Theyshould be abble to comment the business cycle from filtered economic time series.

After TD3, they should be abble to solve (small) general equilibrium models. They should be abble to use dynare to simulate a model and extract usefull information from its output.

After TD4, they should be abble to solve an aggregation problem (Dixit-Stiglitz or other), to log-linearize properly (using Taylor first order approximations) any usual equation and be abble to explain the derivation of the Phillips curve.

After TD5, if they were not experts in solving (small) general equilibrium models after TD3, they should really be. Log-linearizing should be very easy and simulating models with dynare should raise no problem they can’t solve. They can comment on the results from the simulation compared to real data properties.

After TD6, they should be abble to estimate models (univariate and VAR) using OLS and compare their fit both in and out of sample. They should be abble to extract, interpret and use the main factors from a database.

After TD7, they should understand the necessity of structuring VAR residuals. They should be familiar with three different methods of structuration (Choleski, structural decomposition, Blanchard and Quah). They should at least be abble to use the simplest method : Choleski.

After TD8, they should be abble to run the estimation of a DSGE model using Dynare. They should understand the difference between maximum likelihood and bayesian estimation. They should be abble to use the options mode_check, mode_compute, the blocks varobs, estimated_params_bounds, estimated_params_init, and partially calibrate a model, define priors, run a Metropolis-Hastings.