Syllabus
Lecture 1 - Intro + Review of time series models. AR, MA, ARMA models and their properties. Stationarity: economic and econometric interpretation, unit-root tests.
Lecture 2 - Nonstationary models, structural breaks, and forecasting.
Lecture 3 - Spectra, cycles, and filters. Frequency domain analysis of time series. Spectrum, periodogram.
Lecture 4 - Evaluating business cycles in real-time. Can we predict turning points?
Lecture 5 - State space and dynamic factor models: Synthesize many series for predictions.
Lecture 6 - A primer of Vector autoregressions. Identification of turning points, leading indicators, nowcasting.
Lecture 7 - Identification of VAR models. Structural VARs.
Lecture 8 - Identification of VAR models (Cont.) Sign restrictions.
Lecture 9 - Direct estimation of impulse responses: Local projections and narrative approach
Lecture 10 - VARs with nonstationary variables. Cointegration and VECM.
Lecture 11 - Bayesian VARs and Large VARs. Principles of Bayesian estimation. Bayesian VARs, FAVAR, and alternatives.
Lecture 12 - Recent approaches to identification. External instruments (proxy SVAR) and high-frequency identification. Local projections.
Lecture 13 - Nonlinear models. Univariate and multivariate nonlinear models.
Annotation
Students aiming for a career in central banks, academia, or international institutions will learn methods that are necessary to understand, replicate and conduct empirical research in macroeconomics.
The first part of the course covers modelling univariate time series (stationary and nonstationary models, spectral analysis, regime-shift models). The second part of the semester is devoted to multivariate models, forecasting, and identification of causal relationships in macroeconomics. The recently developed approaches to identification such as external instruments in VAR or high-frequency identification are covered as well.
Our course participants apply all covered methods in regular problem sets that are based on replications of academic papers. These problem sets are presented and discussed in the seminars.
Problem sets shall be prepared in R and delivered as Jupyter notebooks, sample R-codes are provided.