EERC Methodological Seminars Series

 

ECONOMETRIC SEQUENCE

Time Series Analysis

 

Instructor: Stanislav Anatolyev, New Economic School, Moscow

 

 

This seminar is the second part of the econometric sequence. It is devoted to time series analysis as seen by an applied researcher. We will see how estimation of structural time series models differs from that in cross sectional analysis. Then we will turn to nonstructural modeling. First we will study popular models of the conditional mean dynamics, such as linear ARs and VARs as well as nonlinear structures like threshold models, Markov switching models and the like. We will also review such issues as stationarity vs. nonstationarity, unit roots and seasonal adjustment. Then we will turn to modeling of the conditional variance as represented by the class of ARCH models. Theoretical and empirical examples will be abundant throughout. Beside a chapter in Greene's book, several theoretical surveys and applied papers will be assigned as readings.

 

ORGANIZATION

 

Along with lectures, there will be separate computer sessions. The statistical working tool is Econometric Views. The students are encouraged to work in groups. The instructor will be available during office hours.

 

LITERATURE

 

Greene, W. (2000) Econometric Analysis, 4th edition, chapter 18

Mishkin, F. (1990) What Does the Term Structure Tell Us About Future Inflation? Journal of Monetary Economics 25, 77-95.

Potter, S. (1999) Nonlinear Time Series Modelling: An Introduction. Manuscript, Federal Reserve Bank of New York.

Evans, G. (1989) Output and Unemployment Dynamics in the United States. Journal of Applied Econometrics 4, 213-237

Diebold, F. and Lopez, J. (1995) "Modeling Volatility Dynamics," in K. Hoover (ed.), Macroecono­metrics: Developments, Tensions, and Prospects, Boston: Kluwer Academic Press, 427-472.

Hsieh, D. (1989) Modeling Heteroskedasticity in Daily Foreigh-Exchange Rates. Journal of Business and Economic Statistics 7, 307-317.

 

SYLLABUS

 

I. Stationarity and Structural Time Series Modeling and Estimation

        Stationarity and ergodicity, mean reversion, loss of memory

        Asymptotics and bootstrap in time series

        Serial correlation, conditional heteroskedasticity and robust inference

        Structural Rational Expectations models

 

II.Univariate time series: modeling the mean

        Stationary and nonstationary variables: totally different behavior

        Trends versus random walks: is GNP trend deterministic or stochastic?

        Seasonal adjustment

        Autoregressive linear processes: estimation, testing and model selection

        Prediction and prediction errors

        Testing for unit roots: the augmented Dickey-Fuller test

        Nonstationary univariate time series: stochastic and deterministic trends

        Nonlinear time series modeling of the mean: thresholds, structural breaks, chaos

 

III. Multivariate time series: modeling the mean

        Vector autoregressions: simultaneous determination of money and prices

        Identification of VAR: structural vs. reduced forms, the simultaneity problem and identifying restrictions

        Estimation of VAR: single equation OLS estimators

        Granger causality: does printing money cause GNP growth?

        Impulse response functions: which shocks are more long-lived?

        Variance decomposition: which shocks are more important?

        Nonstationary multivariate time series: spurious regression, cointegration, common trends

IV. Modeling the variance

        Stylized facts about financial data: volatility clustering, volatility comovements, leptokurtosis

        Engle's ARCH and Bollerslev's GARCH. The properties of ARCH and GARCH processes: stationarity, existence of moments, thickness of tails, persistence

        Estimation of ARCH and GARCH models. The Maximum Likelihood estimator and its asymptotic properties

        Testing for ARCH effects

        Conditionally non-normal innovations and Quasi-ML estimation

        ARCH-M models: time varying risk premium

        IGARCH models: long memory in variance