This course considers econometric inference from time-series data. A crude outline is as follows. We discuss univariate and multivariate linear processes (ARMA, VAR), markov-switching models, and statistical concepts such as stationarity, ergodicity, and feedback. These tools are used to extend the applicability of least-squares and maximum-likelihood estimation to deal with dependent data in systems of equations. We also consider such topics as forecasting, seasonality, autoregressive conditional heteroskedasticity, spectral analysis, and the Kalman filter. Finally, we look at non-stationary time series (ARIMA) and derive statistical tests for unit roots and cointegration.


Lectures :
The main reference text for the course is Time Series Analysis by J.D. Hamilton (1994, Princeton UP).

Slides, class notes, and additional material will be made available.


Lectures complémentaires

A useful complement is Econometrics by F. Hayashi (2000, Princeton UP).

Some other texts that contain treatments of the material covered are (i) Applied Econometric Time Series by W. Enders (2009, John Wiley & Sons); and (ii) Time Series: Theory and Methods by P.J. Brockwell and R.A. Davis (2009, Springer).

Discussion that are more specialized to the econometrics of financial time series can be found in (i) Analysis of Financial Time Series by R.S. Tsay (2010, Wiley-Blackwell); and (ii) The Econometrics of Financial Markets by J.Y. Campbell, A.W. Loo, and A.C. MacKinlay (1996, Princeton UP).

A technical monograph on the estimation of linear processes is Asymptotic Theory for Econometricians by H. White (2000, Academic Press).

Course taught in English at Sciences Po

Niveau requis : Introduction to statistics and econometics at the graduate level. Experience with statistical software.

Dernière mise à jour : mardi 25 septembre 2012