Scribd
Upload
6 views10 pages

Cointegration

The document provides an overview of cointegration in multivariate models, explaining the process of testing for cointegration using the Engle and Granger two-step method. It details the steps involved in estimating long-run and short-run models, including the importance of the error correction term. Additionally, it discusses model diagnostics and the significance of the estimated coefficients in relation to the long-run equilibrium.

Uploaded by

alemenjivar03
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
6 views10 pages

Cointegration

The document provides an overview of cointegration in multivariate models, explaining the process of testing for cointegration using the Engle and Granger two-step method. It details the steps involved in estimating long-run and short-run models, including the importance of the error correction term. Additionally, it discusses model diagnostics and the significance of the estimated coefficients in relation to the long-run equilibrium.

Uploaded by

alemenjivar03
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

Cointegration Overview

In univariate models, we have seen that a stochastic trend can be removed


applying first differences. Then, the stationary series can be estimated and
forecasted using the Box-Jenkins 3 steps method.
Treating Non-Stationary variables in multivariate models is not that
straightforward.
Why? There can be a linear combination of integrated variables that is
stationary, in such case we say the variables are ”cointegrated”.
xt = β0 + β1 yt + t
t = xt − β0 − β1 yt
Where t = I (0)
Cointegration - Engle and Granger 2
steps method
Step 1
Test the variables for their order of integration. NOTE: The variables have
to be of the same integration order.
Estimate the long-run equilibrium model by OLS, save the residuals and
test if the residuals are stationary.
yt = β0 + β1 zt + et

If the variables are cointegrated, the OLS estimates will yield ”super
consistent” parameters β0 and β1 , as they converge faster than they do
using stationary variables. (Watson-1987)
The Augmented Dickey Fuller Statistics are not valid for the residuals unit
root test. Why? The residuals are not observable.
Step 2
Estimate the Error correction model - Short Run Model
Model Diagnostics
Example - Trade: Method 1
Overview
Variables: Exports (Xt ) - Imports (Mt ) for Canada . 1961Q1 - 2005Q4
Step 1:
Check for Stationarity:Both variables have to be integrated of the same
order.
After checking stationarity:
Estimate the long run model

Xt = β0 + β1 Mt + t

Save the residuals


Perform ADF test on the residuals and use the residual regression test
table.
t = Xt − β0 − β1 Mt
Where t is I (O ) if variables are cointegrated
Recall the statistics of the ADF are not appropriate. Use the table values.
Regression-Residuals Test Table
Graph
Example - Trade: Method 2

We will put aside the table and implement the cointegration tests provided by
Stata.
Conduct the Engle and Granger test using the egranger command in stata
you need to install the option in stata
Some Observations
Granger test is a Residual-Based Test for Cointegration
The test is simply a unit root tests applied to the series residuals
The null hypothesis is: ”Series are NOT cointegrated”. Therefore,
rejecting the Null hypothesis will result in our series being cointegrated.
Error Correction Model
We have already estimated the Long Run Model
Variables are non-stationary in levels, but are cointegrated:

Xt = β0 + β1 Mt + t

Now lets estimate the Short Run Model - Error Correction Model
For the short run model, the variables need to be in differences - stationary
form.
We need to incorporate to the model the error correction term, which are
the residuals of the long run regression but lagged one period.

t −1 = Xt − β0 − β1 Mt −1 (1)

4Xt = β0 + β1 4Mt + β2 t −1 + νt (2)


plug (1) in (2)

4Xt = β0 + β1 4Mt + β2 (Xt − β0 − β1 Mt −1 ) + νt


Error Correction Term - Considerations

4Xt = β0 + β1 4Mt + β2 t −1 +νt


| {z }
ECT

ECT = Error Correction Term


β2 is the error correction term estimated coefficient, where −1 < β2 < 0
Values out of the range are explosive results. You need to
review/re-estimate your model.
The coefficient determines the ”speed of adjustment” towards the long run
equilibrium.
The deviations from the long-run equilibrium are corrected gradually by the
ECT through a series of partial short-run adjustments.
Short Run and Long Run Estimated
Models
Long Run Model
Xt = −0.45 + 1.02Mt + t
Short Run Model
Xt = 0.009 + 0.64Mt + −0.11t −1 + νt

Model
Variable
Long Run p-value Short Run p-value
C -0.4536 0.0041 0.0009 0.0000
Mt 1.0204 0.0000 0.6381 0.0000
ECT n/a n/a -0.1165 0.0006
The ECT is statistically significant.
The coefficient of the ECT is -0.1165, suggesting that almost 12% of the
discrepancy between the long run and the short run is corrected within a
quarter.
Short Run - Model Diagnostics

Residuals Normality Test


Shapiro-Wilk W test for normal data
H0 = Residuals are Normally distributed
If p > 0.05, residuals are normally distributed
Serial Correlation
We conduct the Portmanteau test for white noise
H0 = Residuals are White noise
if p > 0.05, the residuals are white noise

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy