Econ 427, Spring 2009

Problem Set 2 -- Due Tuesday, February 16.

Ch 6. Problems and Complements:

2. (p. 108) [Hint: you will need the data in Diebold data file ...\Data Sets\fcst4-06\fcst6input.wf1]

6. (p. 109) [Hints: you will need the data in Diebold data file ...\Data Sets\fcst4-06\PC6_06_web.dat. Start by creating a new workfile for "Daily - 7 day week" data beginning in 1/1/1998 and ending in 12/31/1998, and then import the data. See the Eviews page of our website for step-by-step instructions on "How to Import a Diebold Dataset into Eviews."

Since you have less than a year of data, it would not make sense to include monthly seasonal dummies (there is no possibility of recurring monthly patterns with only 9 months of data). So the "seasonal" factors that you will want to consider are day-of-week effects only. When selecting the final model in (d), use the MSE, AIC and SIC criteria. Remember that for comparability with linear and quadratic trend models, you must esimate the exponential model in levels, rather than the log linear form. See p. 91 for how to do this.]

Note: You can download an Excel file where I created day-of-the-week dummies from the EViews page.

Extra Problem:

Go online and download a monthly or quarterly time series for the U.S. or Hawaii economy. You can get quarterly data for components of GDP and other data at the BEA web site, http://www.bea.gov/ . You can get monthly data on employment, jobs and other series at the Bureau of Labor Statistics web site, http://www.bls.gov/ . You can find data for a large number of Hawaii indicators at the UHERO Data Portal, http://uhero.prognoz.com/ . The St. Lous Fed FRED site, http://research.stlouisfed.org/fred2/ , has quite a lot of data on many topics. Be sure that you do not get a seasonally adjusted series!

Using all but the most recent year of data for your time series, select a forecasting model. You should allow for both trend and seasonal factors, and you should consider alternative types of trend modeling. Comment on your model's residual correlation, normality, equation fit, etc.

Use your model to forecast over the most recent year. Create a graph of forecast and actuals, with error bands, similar to Fig. 6.9 on page 108. Remember that there is example EViews code for such figures under "Forecast figure with confidence intervals" on our EViews page online.