Economics 221
Time-Series Econometrics: Forecasting
Professor F.X. Diebold
Fall 2015


This course provides an upper-level undergraduate / masters-level introduction to forecasting, broadly defined to include all aspects of predictive modeling, in economics and related fields.

Prerequisites: Courses in (0) math (MATH 104 plus either MATH 114 or 115) (1) intermediate economics (ECON 101 plus ECON 102; Wharton students can satisfy the ECON 101 prerequisite with BEPP 250 honors -- the regular BEPP 250 course does not count as a substitute for ECON 101), (2) probability/statistics for economists (ECON 103) and (3) introductory econometrics, including basic time-series econometrics (ECON 104). We will also use some ideas from financial economics, and some elementary matrix algebra, and students must be willing/able to learn them as necessary. Finally, students should be able to program (including simulations) in languages like EViews, R, Python, etc. We will emphasize EViews and R.

Although we will make heavy use of (and assume significant background in) general econometrics/statistics, this course is much more sharply focused. It explicitly and exclusively about economic prediction, or forecasting, as opposed to general econometrics/statistics, or anything else. Emphasis will be on forecast construction, evaluation, and combination (point, interval, density).

Relevant topics but are not limited to: regression from a predictive viewpoint; conditional expectations vs. linear projections; decision environment and loss function; the forecast object, statement, horizon and information set; the parsimony principle, relationships among point, interval and density forecasts; statistical graphics for forecasting; forecasting trends and seasonals; model selection for forecasting; characterizing, modeling and forecasting cycles with ARMA and related models; Wold’s theorem and the general linear process; nonlinearities and regime switching; the chain rule of forecasting; optimal forecasting under symmetric and asymmetric loss; recursive and related methods for diagnosing and selecting forecasting models; formal models of unobserved components; conditional forecasting models and scenario analysis ("stress testing"); vector autoregressions, predictive causality, impulse-response functions and variance decompositions; use of survey data; business cycle analysis using coincident and leading indicators: expansions, contractions, turning points, and leading indicators; incorporation of subjective information; Bayesian VARs and the Minnesota prior; evaluating a single forecast; comparing forecast accuracy; encompassing and forecast combination; combining forecasts; preliminary series, revised series, and the limits to forecast accuracy; prediction markets; unit roots, stochastic trends, stochastic trends and forecasting; unit roots; smoothing; ARIMA models, smoothers, and shrinkage; using stochastic-trend unobserved-components models to implement smoothing techniques in a probabilistic framework; cointegration and error correction; evaluating forecasts of integrated series; volatility forecasting via GARCH, stochastic volatility and realized volatility.

Initial lecture videos: here1 and here2 and here3 and here4 and here5 (no more after this one).

Diebold's Forecasting.
Silver's The Signal and the Noise.

We will read and discuss a significant number of research journal articles.

Software: EViews, R, perhaps others if you wish. There are issues and tradeoffs, which we will discuss, and which you should consider before deciding what to do.

Supplementary materials:
No Hesitations blog.
Other books: (1) Econometrics and (2) Elements of Forecasting (4e)
Software intros: EViews Intro; R Intro; Python/Julia Intro (Sargent and Stachurski), Python Intro (Sheppard)
Weekly meeting of the Research/Reading Group
27 August 2015 slides and full paper.

Piazza: The system will get you help quickly and efficiently from classmates and TA's. Rather than emailing questions, simply post them directly on Piazza. Our class page is: If you have any problems or feedback for the developers, please email them at

Grading: Consistent class attendance and participation are crucial for good performance. Performance will be assessed by N standardized problem set scores (P's), a standardized final exam score (E), and class participation (C). (Regarding class participation, I intend for this to be a highly-interactive class.) The final score will be .60*Pbar + .25*E + .15*C. P's are due one hour before the start of class on the assigned day. Under no circumstances will late P's be accepted, so be sure to start (and finish) them early, to insure against illness and emergencies.

Office hours: Posted on web at

TA: Ross Askanazi,
Office hours: W, Th, F 10:30-11:30, McN 526
Weekly review sessions: Wed. 7-8, McN 169

Important dates:
P 1 due Sept 17. Do Ch. 3, EPC 1. Data here.
P 2 due Oct 6. Do section 4.2. Data on book site.
P 3 due Oct 27. Do Ch. 7, EPC 1.
P 4 due Nov 17. Here.
P 5 due Nov 24. Read and report on Brownlees, Engle and Kelly (here).
P 6 due Dec 8. Read and report on Gillen, Plott and Shum (here).

Final exam: Standard university-scheduled day/time/location.

Important administrative policies here. (READ CAREFULLY!)

Note well: Modifications and adjustments to this outline are inevitable and may be implemented at any time. Check frequently for updates.