international macroeconomics: msc economics week 1: question 1 peter stanley, david glover, daniel...

International Macroeconomics: MSc Economics

Week 1: Question 1

Peter Stanley, David Glover, Daniel Funge and Bruce Moniri

(a) Import the data into EViews 6.

• Go to file – open – Foreign Data Workfile• Choose relevant file/data source (pppdata.xls)• Format data as desired on ‘Spreadsheet Read’

menu

Data• 91 products representing a good for each

country• Q - represents the log of relative prices

Q = In([p*e]/q)• p – price of an unspecified good in a non US

OECD country (x)• q – price of the same good in the US• e – nom exchange rate between the US and x.• Monthly data from Jan’81 – Dec’95 (180

observations) for each good.

Data

-7.50937808291180

………

-7.145439293912

-7.061058539911

………

-3.5839355182180

………

-3.74961164222

-3.69617048121

-3.6603777711180

………

-3.7164177212

-3.64243684911

QPanelIDDATE

(b) Conduct unit root tests on Q reporting results with and without a

trend.

• Process: View – Unit Root Test – choose root test (Lin-Levin or Im, Pesaran and Shin).

• Check menu box corresponding to trend/no trend choice.

Results

0-18.8305trendIm, Pesaran and Shin W-stat

0-7.04755no trendIm, Pesaran and Shin W-stat

14.61616trendLevin, Lin & Chu t*

0.78340.78362no trendLevin, Lin & Chu t*

Prob.**Statistic Trend/No TrendMethod

Does the real exchange rate have a unit root? (LL test)

• Interpretation of Lin Levin results: for 5% significance level.

• If P > 0.05 Cannot reject unit root in all series.• If P < 0.05 Reject.• Our P (no trend): 0.7834• With trend: 1• Therefore, we cannot reject the existence of a

unit root in all series.

Does the real exchange rate have a unit root? (IPS)

• Interpretation of Im, Pesaran & Shin test results: for 5% significance level.

• If P > 0.05 Cannot reject unit root in all series.• If P < 0.05 Reject.• Our P (no trend): 0• With trend: 0• Therefore, we can reject the existence of a unit

root in all series, i.e. IPS suggests that at least one series is stationary.

(c) Estimate an AR(1) Model for Q using the fixed effects estimator.

• Fixed effects requires the creation of a dummy for each panel cross section. EViews will do this for you if you ensure your data is set in a recognised panel format.

• Process: Quick – Estimate Equation – Type equation as Q = C(1) + C(2)*Q(-1) - select fixed effects in panel options.

Results

0715.09810.0013780.985667Q(-1)

0-10.31190.005007-0.05163C

Prob. t-StatisticStd. ErrorCoefficient 

What is the half-life for the real exchange rate?

• Take coefficient 0.985667 from fixed effects panel regression.

• Using 1 as the t=1 value, calculate number of periods required to take series value to 0.5.

• Using formula:

log(0.5)/log(p)=log(0.5)/log(0.985667)=48.01

Half-lifeDecay of real exchange rate

0

0.2

0.4

0.6

0.8

1

1.2

1 6 11 16 21 26 31 36 41 46 51 56 61 66 71

q

(d) Estimate an AR(1) model for each good separately.

• Process: un-stack the data using the ‘reshape current page’ function in EViews.

• This splits panel data into separate cross-sections.

• Then run an AR(1) on each separate cross-section.

• Results – Mean: 0.9811,

Standard Deviation: 0.02339

Results

Distribution of AR(1) Coefficients

010203040506070

0.87

3065

0.88

7114

0.90

1162

0.91

5211

0.92

9260

0.94

3308

0.95

7357

0.97

1406

0.98

5454

0.99

9999