Economic Forecasting Seminar

1
Economic Forecasting Seminar

• Exploratory Data Analysis
• Your First Forecast

2
The First Steps

• Oral Presentations
• Written Reports
• Tools Techniques for an exploratory analysis of
  your data for the purposes of modeling and
  forecasting

3
Oral Presentations

• Four Parts
• Introduction You your audience
• Overview Tell em what youre going to say
• Important Features of your Analysis Say it
• Conclusions Tell em what you said

4
Oral Presentation Introduction

• Your name
• Who you are speaking to. Examples Board of
  Directors of Enron the Forecasting Team for the
  Energy Division of Charles River Associates the
  Marketing Division of AFLAC.
• Why are you here?

5
Oral Presentations Helpful Hints

• PowerPoint useful for organizing your talk and
  communicating lots of information in a succinct
  fashion.
• Try to have eye contact with your audience, not
  your slides.
• Avoid distracting gimmicks and mannerisms.
• Practice your talk before you present it. Your
  goal for this class should be a 7 minute
  presentation.

6
Oral Presentations More Hints

• Dont include too much text on any one slide.
  More than two sentences per bullet / more than 4
  or 5 bullets per slide is too much.
• Use charts and simple tables.
• Think about format use color for important
  features/data. Consider the number of decimal
  places implications on accuracy.
• Know your audience. Dont assume a lot of
  technical expertise.

7
Oral Presentations Overview

• Key features of your data time range prominent
  features (e.g., strong trend seasonality).
• Brief summary of the methods of your
  analysis/forecast.
• One or two key results hints as to the forecast.

8
Oral Presentations Important Features of Your
Analysis

• Decompose your series. Whats the single most
  important feature?
• Consider the presence of outliers or
  irregularities.
• Describe your forecasting method in your own
  words.
• Illustrate how you applied this technique and
  your choice of model.
• Assessment of accuracy or your method.

9
Oral Presentations Conclusions

• Present your forecast in an appropriate context
  (e.g., recent trends).
• Review of your presentation.
• Linchpin assumptions or key factors on which
  your forecast depends.
• What role did your judgment play in this
  forecast?
• Email me your PowerPoint presentation no later
  than 1 hour before class.

10
Written Report

• 1 to 2 pages in length, including charts
  tables.
• May include a technical appendix of any length.
• Should mirror your oral report in terms of
  content.

11
Written Report Format

• 12 point text minimum. 1 inch margins on all
  sides.
• Section headers in bold.
• Charts with wrap-around text help tell your
  message. Think about the context of your charts.
• Think about the informational content of the
  headers in your report.

12
Written Report Overview

• Provides an Executive Summary of your entire
  report, including a teaser about the results.
• Data Analysis section illustrates key features
  of your data. Decomposition of your series and
  one or two key features.

13
Written Report Forecast Method

• Describes succinctly and in your own words the
  forecast method you have used and the model you
  chose.
• Apply selection criteria in a logical fashion.
• Provide an ex post assessment of accuracy.

14
Written Report Forecast

• Ex ante forecast.
• Level and/or rates of growth terms depending on
  the context of your story.
• Anything interesting about your forecast?

15
Written Report Conclusions

• Same as for the oral reports.
• What are the linchpin assumptions?
• What judgmental adjustments would you make to the
  models unadjusted forecast (if any)?

16
Tools for Exploratory Data Analysis

• Graphs give them a title and identify the
  units.
• Consider using gridlines to help identify levels.
• Consider labels for important points/features.
• Consider subsets of the data.
• Report summary statistics
• Calculate rates of growth

17
Evaluating Forecast Accuracy

• Involves calculating forecast errors and
  modifying your methods.
• Several criteria exist which can be used to
  assess the adequacy/accuracy of the method you
  choose.

Ex post simulation "Historical Simulation"
Ex post forecast
Ex ante forecast
Backcasting
Time, t
Estimation period
T3 (Last Observation)
T1
T2
18
Ex Post Forecasting

• Withhold 12 observations from the end of your
  sample (T3-T2).
• Estimate your model (T2-T1).
• Forecast the observations you withheld.
• Calculate summary statistics for the resulting
  forecast errors.

19
Mean Error

• A measure of bias in the forecast.

20
Mean Absolute Error
21
Mean Absolute Percentage Error
22
Mean Squared Error
23
Root Mean Squared Error
24
Assignment Your First Forecast

• Forecast Method Linear trend projection.
• Select a regression model for your series.
• Use only a linear trend seasonal dummy
  variables and/or a recession dummy as candidate
  explanatory variables.
• Forecast your series through 200812.
• Evaluate the forecast.