INTRODUCTION

1. Introduction

   a. The importance of forecasting.

   1. Decision making base on expectations about the future.

   2. Hence, forecasting is an unavoidable activity.

   b. Why forecasting?

   c. What is a forecast?

   d. Hierarchial forecasting.

2. A time series

   a. Time series vs. cross-sectional data

   b. Trend, seasonal, cyclical and random components.

   c. Autocorrelation

   d. Outliers or interventions

3. Overview of time series forecasting methods.

   a. Univariate methods

   b. Multivariate or casual methods

   1. Outliers or interventions

   2. Independent variables

   c. Point and probability forecasts

   d. Most basic assumption is that history repeats itself.

   1. Statistical techniques identify the historical relationships in a series or between series.

   2. Once the historical relationships are identified correctly, then forecasts can be made.

4. The application of sound theory identifies the important right-hand-side independent variables.

   a. Appeal to economic, financial, accounting, engineering or a natural science to tentatively identify the important independent variables.

   b. Avoid ad hoc or spurious relationships between the dependent and independent variables.

   1. Spurious relationships will not forecast.

   2. Spurious relationships will not meet the theory test.

5. If the data are time series, then time series techniques must be used to summarize the relationship between lagged values of the series and independent variables

   a. Classical regression techniques (OLS) will not summarize the relationship among time series variables (except by chance alone).

   b. If classical regression techniques are used to summarize the relationship and these techniques usually fail, then the analyst will spend an inordinate amount of time identifying 'fix-up' routines to repair the failures.

   c. Most econometric textbooks devote most of their pages to showing the application of various 'fix-up' techniques to overcome the failure of OLS to correctly summarize the relationship embedded in a univariate series or between series.

   d. Classical regression techniques are applicable when the analyst is summarizing cross-sectional data series.

Return to Econ 3320 homepage