1. Your boss wants you to produce some forecasts. You are first going to analyze

1. Your boss wants you to produce some forecasts. You are first going to analyze the data.
Select some macroeconomic data that are either monthly or quarterly and that at least appear to exhibit trend and seasonality. Be sure to get at least ten years of data. Do not select a series that has already been seasonally adjusted. Try to select data that are different from the set you chose last time. For example, choose a different frequency (quarterly versus monthly), different type of data (trade versus GDP, international versus country-regional, etc.), etc. Use the data to do the following exercises.
a. (5 points) Submit a plot of the percent change of the series.
b. (15 points) Submit the correlogram of the percent change of the series and discuss what you can learn from the correlogram. (Two or three sentences at most.)
d. (5 points) Using the correlogram as a guide, run some simple ARMA regressions on the percent change of the series (no more than 20). Submit a well organized table in grid format (see the book for an example) with the AIC and SIC values for the models you estimated.
e. (10 points) Determine which model is the best in your opinion. Briefly discuss why you chose the model you selected. (Two or three sentences at most.)
f. (5 points) Submit the estimation output of the model you selected as well as a plot of the actual, fitted, and residuals. (Do NOT submit output from all of the regressions you attempted.)
2. Your boss wants you to produce some forecasts. You are first going to analyze the data.
Use the same macroeconomic data as in problem 1. Instead of selecting your model based on its in sample performance, use the techniques you have learned to select the best model based on forecasting performance. In addition to looking at simple out of sample forecasting performance, use recursive cross validation . If you are an adept programmer, you can program a loop to make recursive cross validation faster to perform. If you are not an adept programmer, start with your preferred model and select only one other candidate model – no more.
(40 points) Briefly discuss how you selected the model. Did the model you chose change from last week, explain why that might be?
3. Estimating VARs
Use the macroeconomic data you used in Question 1 and Question 2. Now estimate some VAR models. This will involve getting other data that may be related to your original series. Using the techniques you have learned, select the model that is, in your opinion, the best model for the data.
a. (15 points) Briefly discuss how you selected the model.
Using your best model, create a forecast for your series of interest (the length of the forecast should be determined by the data, e.g. for quarterly data perhaps a year, while for monthly data perhaps six months).
b. (15 points) Briefly discuss the fit of your forecast.
c. (25 points) Create the impulse response functions for your VAR and interpret a few of the more interesting ones.
d. (25 points) Create the variance decomposition for your VAR and interpret a few of the more interesting ones.
4. Comparing Forecast Accuracy
a. (20 points) Now, using the VAR you chose in the previous question and the model you chose in question 2, compare the forecast accuracy of the two models for the original series you selected.