You are on page 1of 1

Econ 483, Spring 2017

Problem Set 10
Due: Thursday, April 20 at the start of class

Instructions:

This problem set has 4 questions. List the STATA commands you use (ideally by printing out the
STATA do-file) and report results and print graphs as requested in the questions. You may work
in groups, but you should write up your own answers in your own words. For the last question,
write this on your own.

1. Suppose we start with an AR(2) model:

y t = + 1 y t 1 + 2 y t 2 + t

However, we have identified two possible break dates: s 1 and s 2 at which all the and
parameters change.
Explain how to set up this model using dummy variables, and explain in words how to
interpret each of the parameters in your model.

2. For this question we will use the US Consumer Price Index (CPI, label CPIAUCSL in Fred),
and the AAA and Baa corporate yields (AAA and BAA in Fred).

(a) Load the data, create an appropriate time index, and create the variables:

y t = C P I t
xt = B A At A A At .

(b) Plot both y t and x t .


(c) Using 12 lags of y t and 12 lags of x t , test for Granger causality of x t on y t .

3. For this question we will use the yen-dollar exchange rate, series EXJPUS on Fred.

(a) Download the data and provide a time-series plot.


(b) We will consider linear trend models. Using a window length of 200 (months), cre-
ate rolling estimates of the linear time trend model y t = + t + t . Plot the rolling
estimates of the intercept and of the slope coefficient (use separate plots).
(c) Repeat part (b) but using recursive estimates, starting with a window of length 200.
(d) Discuss the results from parts (b) and (c). Is there evidence to suggest a break? If so,
around what date?

4. For the Silver reading, you may read Chapter 10 or Chapter 11. Write 2-5 sentences on what
you found most surprising in the chapter.

You might also like