| |
|
|
|
Term Projects |
||
THE Forecasting Project(FIN 40230 / FIN 70230)Term Projects for 2016 StudentsThe PlanThe only satisfactory way to learn how to forecast is to actually do it. It is also better to form forecasts and then to evaluate them in real time rather than using a contrived situation. A partial requirement for this course is to undertake a forecasting project of the type outlined below. You will present your results late in the semester to the class. We have essentially an fourteen-week college semester (actually, twenty-nine class meetings). The outline we will follow for the timing of the project will be as follows: Week 1-3:Decide what real economic variable(s) to forecast and what other variables are to be used to form the forecast. Denote the variable being forecast by Yt and an explanatory variable (or variables) by Xt. These should be variables whose values are available weekly or monthly so that X1 is the value taken by X in week one or month one (do not use yearly data). Gather past data for the variables. Steve Hayes is the business librarian in the Business Information Center (BIC) located in the lower level of the College of Business Administration; he will be a valuable resource to you as you begin to select variables and search for data. Week 4:Forecast the value Yt, that is, the value that will be taken by Yt in Week t or Month t using moving average and exponential smoothing methods (Chapter 3). Week 5:Again forecast Yt using more recent data and some form of simple regression (Chapter 4). Week 7:Again forecast Yt using more recent data and some form of multiple regression which will involve the use of some causal variables (Chapter 5). Week 9:Use time series decomposition to forecast Yt if that method seems appropriate (Chapter 6). Week 10:Identify and Estimate an appropriate ARIMA (Box-Jenkins type) model and forecast Yt using the most recent data (Chapter 7). Week 14:Combine some or all of your previous forecasts in an attempt to improve your forecasting procedure using the techniques in Chapter 8. Present a multimedia presentation (with appropriate graphics) on diskette with your data and command files on a separate diskette to the instructor. The DataThe most important decision to be made is what variables to use in the project. The series should be available monthly or quarterly. In some rare cases daily or hourly data may be appropriate for the project (but monthly or quarterly date is highly recommended). In any case, a good sequence of previous values must also be available. At least the previous 24 to 48 periods should be found. If the series is suspected to have strong seasonality, values for the corresponding periods for the last three years will be required. If possible it is best to forecast a series in which you have some real interest. Examples could be monthly sales in some industry, the number of telephone calls made from an office per quarter, or hours of mainframe computer time used monthly. The financial press is an excellent source of monthly data. Journals and newspapers such as Barrons, The Economist, The Wall Street Journal, and The Financial Times have many pages of suitable series, such as automobile sales, airline miles travelled, or deliveries of some produced items. It is strongly recommended that the series to be forecast is not a price taken from a highly speculative market such as stock or bond prices, the prices of major commodities, such as copper, silver or gold, an exchange rate or an interest rate. Although such series are certainly readily available and are of general interest, it has been well established that changes in any price or interest series are extremely difficult to forecast and so are inclined to make disappointing projects (i.e., do not use prices or interest rates as your data to be forecasted). You will need some data other than the values of the series to be forecasted in order to complete the project. This other (explanatory) data to be gathered is information that might be helpful in forecasting the series of interest. For example, if one is attempting to forecast electricity sales to households in some region, then a measure of local temperature could be a good explanatory variable. The search should be for a "causal" series or a "leading indicator" of the series to be forecast. All series should have data available over the same period (and of the same periodicity). One can use common sense or economic theory to choose likely explanatory series. Of course, several possible explanatory series could be utilized, and this would make the project rather more rewarding and, perhaps, produce a better forecast. Initial AnalysisEach series should first be plotted through time. These plots give a good indication of the general properties of the series: are they smooth or jagged; is there a monthly or annual swing in the series; and does the series contain a trend in mean? It is also very important to ask if the series contain strange values, called outliers. If they exist, they can be seen easily from the plot. Outliers can arise either from a genuine strange period in the data--due to a strike, typhoon, or heavy snow storm, for example--or can be found because the data was misrecorded or incorrectly plotted. If an outlier is found, it is suggested that for the purposes of the project, a simple method of removal should be used, such as replacing the outlier by the previous, acceptable value or by the average of the two terms on each side of the extraordinary value (be certain to advise the instructor that you are handling any outlier this way). The Various ForecastsThe text contains discussions of several forecasting methods. Some use just the past values of the series to be forecast as discussed in Chapters 3, 6, and 7, and others also use other series, as in Chapters 4, and 5. Chapter 8 has a discussion about combining the results of the various forecasts you will make throughout the semester for the project. Project Structure and TimingYour project should contain at least the following elements:
The project will be graded using the following categories:
|
||
