Economics 537 Final Exam 2011 Dr. Stokes

Economics 537 Final Exam 2011 Dr. Stokes

The final exam will count 20% of the grade. It must be typed. It is due by 6:00 Thursday of Final's Week. The goal of this exam is to introduce you to major datasets.

Answer all questions. Total points = 100.

1. ARIMA Forecasts of Long Run Weather. (50 points) The dataset hinich.mac found on smith

/usr/local/lib/b34slm/datafiles/hinich.mac

contains a file DARWIN that consists of monthly atmospheric pressure readings at the Darwin station in the pacific in the period 1/1/1882 to 1/12/1996. Means obtained should be:

Variable Mean Standard Deviation Variance Maximum Minimum

PRESSURE 1 0.79710145E-03 1.0467080 1.0955976 3.7000000 -3.7000000

CONSTANT 2 1.0000000 0.0000000 0.0000000 1.0000000 1.0000000

Data file contains 1380 observations on 2 variables. Current missing value code is 0.1000000000000000E+32

Data begins on (D:M:Y) 1: 1:1882 and ends on 1:12:1996

hinich.mac is a large file of over 143,000 lines and cannot be edited except with powerful editors.

25 points. You are asked to Filter the series PRESSURE using an appropriate ARIMA model and forecast at origin 1381 and 1300 for 50 periods out. How good is your model.

25 points Estimate the spectrum of this series using weights 1 1 1 1 1 and discuss what you find.

2.High Frequency Financial Data. (50 points). The file hinich.mac contains a file MDC that contains data on log price relatives and log volume relatives for McDonald Douglas stock. Means obtained are:

Variable # Label Mean Std. Dev. Variance Maximum Minimum

MDC 1 log(price(t)/price(t-1)) 0.195226E-04 0.205014E-02 0.420309E-05 0.574306E-01 -0.300012E-01

VOL 2 log(vol(t) /vol(t-1))) 0.106870E-03 3.23181 10.4446 9.97516 -8.38773

TIME 3 time as hour:minutes 12.9167 1.73247 3.00144 16.0000 10.1000

DATE 4 yr:mo:day example 800101 825652. 17075.0 0.291556E+09 851231. 800102.

DAY 5 Day of week Monday = 1 3.00916 1.40209 1.96586 5.00000 1.00000

CONSTANT 6 1.00000 0.00000 0.00000 1.00000 1.00000

Data file contains 54612 observations on 6 variables. Current missing value code is 0.1000000000000000E+32

Part a. (10 points) Attempt to filter MDC and VOL using ARMA models using either manual or automatic methods.

Part b. (15 points) In the finance literature there is a debate whether volume is associated with increases in log price relative or decreases in price relative. Using a one hour window model:

estimate

using the B34S REG command. Use the TEST option to test if VOL(1 to 6) and VOL(1 to 12) granger causers MDC. Conversely test whether MDC(1 to 6) and MDC(1 to 12) impacts VOL.

In view of your findings above is it appropriate to attempt to estimate a transfer function model of Model 1, of Model 2? Discuss your reasons, You will be graded on how carefully you layout your analysis and present your findings.

Part c. 25 points. Attempt a GARCH model of MDC. You will have to use RATS or the B34S Matrix command. Try a two pass and a one pass approach.

Note: These are big datasets and as a result you will have to add space on the command line if running batch.