Course Outline

Course Syllabus
Eco 5375
Economic and Business Forecasting
Fall 2014
Prof. Tom Fomby
TTh 5:00 – 6:20 PM
250 Maguire
Purpose of Course:
This course is dedicated to teaching students tools in econometrics that are especially useful in
forecasting economic and business time series such as sales, expenditures, and macroeconomic
variables such as GDP, interest rates, inflation, stock market, etc.
Student Learning Objectives:
The student will learn the essentials of and demonstrate proficiency in
Decomposition of Times Series into trend, seasonal, cyclical, and irregular components
Deterministic Trend/Seasonal Forecasting Models
Box-Jenkins Forecasting Models
Exponential Smoothing Forecasting Models
Unobservable Component Forecasting Models
Vector Autoregressive Time Series Models
Evaluation of the forecasting accuracies of competing forecasting methods
Evaluation of the usefulness of a proposed leading economic/business indicator
Forming efficient “combination” forecasts
Running SAS, EVIEWS, and R Computer programs
Textbook for Course: The textbook for this course is Forecasting Examples for Business and
Economics Using the SAS System (1996). In addition to this book I will be relying heavily on class
notes, handouts, and posting on the class website. Note: You can ftp any of the programs
contained in this textbook by going to .
Supplementary textbook (student’s choice): The Little SAS Book (5th edition) by L.D. Delwiche
and S. J. Slaughter. This book is one of the classics for learning SAS and preparing for
certification exams in SAS. See
Certification in SAS: If you are an Applied Masters student in our department you might want to
consider becoming certified in SAS. There are two levels of certification: Level I - SAS Certified
Base programmer ( and Level II – SAS Certified
Advanced Programmer ( If you take and pass
either of these tests, the Richard B. Johnson Center for Economic Studies will cover the costs of
the exam (approximately $90). Having a SAS certification on your resume can help you find a
job in quantitatively oriented fields. All SMU students have access to free e-learning courses for
the purpose of preparing for the certification tests. The URL for these free e-learning courses is
at You will need to set up an account with SAS and then contact
the Office of Information Technology (OIT) (
to obtain an activation code to have free access to the e-learning courses.
Computer Usage: We will mainly be using the computer program called SAS (Statistical Analysis
System) produced by the SAS Institute located in Cary, North Carolina. There are two major
ways to access SAS for your homework problems and instruction in class. First, for students who
do not have personal computers or laptops, you can access SAS in the computer labs on campus.
The computer labs of note are the ones in our department (301W in the Umphrey Lee Building),
the computer labs located in Fondren Library West, Rooms 103B and 103C, and the computer
lab located in the Cox School’s Business Information Center (BIC). Second, for students who
own a laptop or personal computer, you can access SAS and other software programs through
Access.SMU (SMU’s Virtual Computer Lab system). The delivery system to your computer is via
Citrix. You can go to the website
and then look on the right of the page to find a link for instructions on how to install the Citrix
Receiver to your computer or laptop. Before you can run SAS on your computer, you have to
install Citrix Receiver on your machine. Citrix provides you with “virtual” access to the SAS
software in that Citrix makes it appear that you have SAS installed on your own computer when,
in fact, it is being accessed from an SMU server on campus. After you install the Citrix Receiver
on your computer, you can then logon to the Citrix Receiver by entering your student ID and
personal password. You will be asked if you wish to permit Citrix to access your computer. Each
time Citrix asks you for permission to access your computer, you should give full permission to
do so. Otherwise, the Citrix Receiver software will not be permitted to connect you to VCL and
you will not be able to access SAS remotely.
Evaluation of Student:
The evaluation in the class consists of four parts:
• Quick Quizzes (20%)
• Exercises (20%)
• Mid-Term Exam (30%)
• Final Exam (30%)
The Quick Quizzes (QQs) will consist of a short answer and/or multiple-choice quiz that will be
administered in the first five minutes of the class. The QQs are designed to see if you have
retained the information of the previous lecture and if you have done any assigned readings that
I may have asked you to do. In addition to keeping the students current in the class and
providing review material for the mid-term and final exams, the QQs allow me to keep track of
student attendance. It has been my experience that for each Quick Quiz a student misses
before the mid-term exam the student, on average, loses 3% on his/her mid-term score. The
bottom line is that it pays to come to class! To reflect the fact that not every day is a good day,
I will be dropping your lowest QQ grade before calculating a QQ average.
With respect to homework exercises, students can confer with each other with respect to
programming advice and discussion of basic ideas but in the final analysis each student is
expected to write up his/her own homework answers and not make copies of others’
homework. Copying someone else’s homework to hand in as one’s own work is a violation of
the SMU Honor Code and will be dealt with according to the rules of the SMU Honor Code. It is
important to know that the homework assignments are very important in that the basic ideas
covered by them invariably show up on the mid-term and final exams. If you know you are
going to be missing a class on the day a homework exercise is due, hand in your homework in
advance to receive full credit for your work. Any homework that is handed in late will be given a
one letter grade reduction for each day of tardiness. It is my policy to drop your lowest exercise
score before calculating your exercise average.
Students will be excused from taking the mid-term exam or the final exam only with a note from
a physician, or in the case of a death in the family, with a note from a parent or guardian. Even
with an excused absence, either of these exams must eventually be taken before a course grade
will be assigned to the student.
If you must miss a class due to legitimate circumstances beyond your control, be sure and
contact me beforehand so that I will know of your circumstances. If excused, I will
correspondingly excuse you from any QQ that is given that day. I want to emphasize that
diligent attendance in this course is essential because a lot of the course material presented in
class will be from my personal class notes and can’t be found in any textbook per se. Note:
After 4 unexcused class absences, I reserve the right to administratively drop students from the
My grading scale in this course is as follows:
Classroom Website:
Office: Room 301M, Umphrey Lee, 214-768-2559. E-mail address: [email protected]
Office Hours: 3:00 – 4:30 PM TTh or by appointment.
My Graduate Teaching Assistant: Yixiang Zhang. His E-mail address is: [email protected]
If you should need extra tutorials or help outside of my office hours, contact Mr. Zhang and he
will be happy to go over concepts that you may not fully understand.
Important Dates to Remember:
First Day of Class: August 26
Fall Break: Monday – Tuesday, October 13 – 14
Last Day to Drop Classes: Friday, November 7
Last Day of Semester in this Class: Thursday, December 4
Final Exam Date: Wednesday, December 17, 3:00 – 6:00 PM in Room 250 Maguire.
General comments on work and class etiquette:
In order to succeed in this class, constant work is essential. Come to class. Read all assigned
readings and prepare for the Quick Quizzes. Don’t get behind. If there is something in class
discussion or homework assignments that you don’t understand, don’t hesitate to ask me in
class, after class, during office hours, or through e-mail.
Obviously, general rules of etiquette apply: cell phones are to be turned off during class and
miscellaneous reading material stowed away.
Some Standard Stuff You Should Know
Excused Absences for University Extracurricular Activities:
Excused Absences for University Extracurricular Activities: Students participating in an officially
sanctioned, scheduled University extracurricular activity should be given the opportunity to
make up class assignments or other graded assignments missed as a result of their participation.
It is the responsibility of the student to make arrangements with the instructor prior to any
missed scheduled examination or other missed assignment for making up the work. (University
Undergraduate Catalogue)
Disability Accommodations:
Disability Accommodations: Students needing academic accommodations for a disability must
first contact Disability Accommodations & Success Strategies (DASS) at 214-768-1470 or to verify the disability and to establish eligibility for
accommodations. They should then schedule an appointment with the professor to make
appropriate arrangements. (See University Policy No. 2.4)
Religious Observance:
Religious Observance: Religiously observant students wishing to be absent on holidays that
require missing class should notify their professors in writing at the beginning of the semester,
and should discuss with them, in advance, acceptable ways of making up any work missed
because of the absence. (See University Policy No. 1.9.)
Honor Code:
All SMU students are bound by the Honor Code (see SMU Student Handbook for a complete
discussion of the SMU Honor Code). The code states that “any giving or receiving of aid on
academic work submitted for evaluation, without the express consent of the instructor, or the
toleration of such action shall constitute a breach of the Honor Code.” A violation can result in
an “F” for the course and an Honor Code Violation on your transcript.
I. Introduction to Course
A. Focus of this Course: Time Series Forecasting
B. Field of Forecasting is meeting the Market Test
C. Example 1: What is a p-value?
D. Example 2: Sales Forecasting and Optimal Inventory
E. Example 3: Four Competing Forecasting Models
F. Example 4: Leading Indicators and Out-of-Sample Experiments
Reference: Class Notes and various SAS programs
II. A Brief Introduction to SAS
A. APPS.SMU and Accessing Computer Programs on SMU’s Virtual Server – Downloading
Citrix Receiver
B. Introduction to SAS (SAS = Statistical Analysis System)
i. Program Editor in SAS 9.4
ii. Data Steps and Procedure Steps
iii. Log and Listing Files
C. Inputting Data
i. Direct Input
ii. Infile Statement
References: Class Notes and Chapters 1 and 2 in The Little SAS Book
III. Preparing Time Series Data for Forecasting
A. Proc Expand in SAS
B. Interpolating Missing Observations
C. Changing the Frequency from Monthly to Quarterly
D. Changing the Frequency from Quarterly to Monthly
E. Transforming the Data
Reference: Example 8 in Forecasting Examples
IV. Additive Decomposition of Time Series
A. Y = T + S + C + I (Additive Decomposition)
B. Trend, Seasonal, Cycle, Irregular Components
C. A Stylized Decomposition of a Time Series
D. It is important to know which components are in your time series and
to properly account for them. Otherwise, you will sacrifice forecasting
References: Class Notes
V. A First Generation Forecasting Model – The Deterministic Trend/Deterministic
Seasonal (DTDS) Model
A. The Simple Trend Model – A Deterministic Trend
B. Trend Model with Seasonal Dummies
C. DTDS plus Autocorrelated Errors
D. An Example: The Plano Sales Tax Data
E. Tests for Trend and Seasonality – F-tests
Reference: Class Notes and Examples 10 and 13 in Forecasting Examples
VI. Some Important Concepts in Time Series Forecasting
A. Mean, Variance, and Autocorrelation in Time Series
B. Definition of Covariance Stationarity
C. Example of a Stationary Time Series: the AR(1) model
i. AR(1) Time Series Model  = ∅0 + ∅1 −1 +  when |∅1 | < 1
ii. Mean, Variance, Autocovariance, and Autocorrelation
iii. The Special Case of ∅1 = 1. The Random Walk model.
iv. The Random Walk Model in not Stationary
v. Differing Prediction Profiles for the two cases: |∅1 | < 1 versus ∅1 = 1
vi. Do Stock Prices follow a Random Walk?
References: Class Notes, Example 1 in Forecasting Examples, and SAS program
VII. Box Jenkins Models for Stationary, Non-Seasonal Time Series
A. Some Simple Box-Jenkins Models and Their Properties
i. ARMA(0,0)
ii. MA(1)
iii. AR(1)
iv. ARMA(1,1)
v. General Notation
vi. Concepts of Stationarity and Invertibility
B. Identification Tools
i. Autocorrelation Function (ACF)
ii. Partial Autocorrelation Function (PACF)
C. Pattern Table
D. Sample Counterparts
E. Information Criteria
F. P/Q Box
G. Overfitting Exercises
H. Example: Lead Production Data
References: Class Notes and Example 1 in Forecasting Examples
VIII. Box-Jenkins Models – Forecasting for Stationary, Non-Seasonal Time Series
A. Minimum MSE Forecasting
B. Various Forecast Profiles
C. Example: The Forecast Profile and Confidence Intervals for the
Lead Production Data
References: Class Notes and Chapter 1 in Forecasting Examples
IX. Box-Jenkins Models for Non-Seasonal, Stochastically-Trending Time Series
A. Taking the First Difference to Control for Stochastic Trends
B. Taking, On Occasion, Second Differences of the Data
C. Augmented Dickey-Fuller Tests for Unit Roots: To Difference or Not
To Difference?
D. Example: The Dow Jones Index
E. Forecasting Levels Based on Forecasts of Differences
F. The Log Transformation and how to use it
References: Class Notes, Examples 2 and 9 in Forecasting Examples, and SAS program
X. Statistical Tests for Detecting Trend, Seasonality, and Cycle
A. Trend Tests: Hirsch, et al. Non-Parametric Tests for Trend and HAC NonParametric Tests of Mean of Differences
B. Tests for Seasonality: Buys-Ballot Plots, ACF at lags s 2s, etc. and Friedman’s NonParametric Test of Seasonality
C. Test for Cycle: Box-Pierce-Ljung Portmonteau Test for Autocorrelation
References: Class Notes
XI. Box-Jenkins Models for Seasonal, Stochastically-Trending Time Series
A. Year-over-Year Differencing
B. Year-over-Year Differencing Combined with First Differencing
C. The Multiplicative Class of Box-Jenkins Models
D. The ACFs and PACFs of Multiplicative Seasonal Models
E. Examples: Airline Passenger Data and Electricity Production Data
F. Testing for Seasonal Differencing
References: Class Notes and Example 3 in Forecasting Examples
XII. Exponential Smoothing – An Old Favorite (Proc ESM)
A. Simple Exponential Smoothing (No Trend, No Seasonality)
B. Double (Brown) Exponential Smoothing (Trend, No Seasonality)
C. Additive Seasonal Exponential Smoothing (No Trend, Seasonality)
D. Winters Additive Method (Trend, Seasonality)
E. Plano Sales Tax Revenue Data – An experiment showing the importance of
Determining the presence or absence of trend in your time series data
References: Class Notes and Example 3 in Forecasting Examples
XIII. An Alternative to the Box-Jenkins Methods – The Unobservable Components
Model (Proc UCM in SAS)
A. Three Unobservable Components plus Noise
i. Trend
ii. Seasonal
iii. Cycle
B. Tests of the Significance of the Components
C. Example: Airline Passenger Data
D. Forecasting the Airline Passenger Data
References: Class Notes
XIV. Searching for an Extra Variable to Help Us Forecast: VARs (Proc VARMAX)
A. Be careful: The Spurious Regression Problem
B. The Equal-Lag Length Vector Autoregressive Model
C. System-Wide Goodness of Fit Measures to Help Choose the Lag-Length
D. Using Out-of-Sample Forecasting Experiments to Detect Useful “Extra”
Variables for use in Forecasting a Variable of Interest
E. Diebold-Mariano Test for Significant Differences in Forecasting Accuracies
F. Example: The “Series M” Data Set
References: Class Notes
XV. Combining Forecasts
A. Combination Forecasting
i. Some Basic Theorems on Diversification of Forecasts
ii. Nelson Combination Method
iii. Granger-Ramanathan Combination Method
iv. Combinations with Time-Varying Weights
B. Application to Economic Time Series
References: Class Notes