Scribd
Upload
28 views8 pages

9 Cointegration and Error Correction Model

Uploaded by

Rahul Singh
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
28 views8 pages

9 Cointegration and Error Correction Model

Uploaded by

Rahul Singh
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/ 8

Cointegration and Error

Correction Model
COINTEGRATION
• We explained earlier that regressing two nonstationary time series
may lead to spurious regression
• We do a D-F test for unit root, and use a difference model if necessary
• However, regressing two nonstationary time series may not always
lead to spurious regression. If this happens, we say that the time
series under study are cointegrated
• Two time series variables are cointegrated mean that they have a
long-run or equilibrium relation between them
When a spurious regression may not be spurious?
• Consider the PDI and PCE series. Both are stochastic nonstationary series (as per
the D-F test) and are I(1)
• You run the regression: LPCE t = b 0 + b 1 LPDI t + b 2 t + u t
• Predict the residuals and perform unit root test (D-F test) for the residuals.
• If it turns out that the residuals are I(0), that means stationary, then we say that
the regression of LPCE and LPDI is not spurious. We say that LPCE and LPDI are
cointegrated.
Error Correction Mechanism
• After allowing for deterministic trend, we have shown that LPCE and
LPDI series are cointegrated. That is they have a long-term or
equilibrium relationship: LPCE t = b 0 + b1 LPDI t + b 3 t + u t

• But how is this equilibrium achieved, for in the short run there may
be disequilibrium?
Error Correction Mechanism
LPCE t = b 0 + b 1 LPDI t + b 3 t + u t DLPCE t = a 0 + a 1 DLPDI t + v t

The above model gives us the long The above model gives the short
run relation between the LPCE run relation between the LPCE
and LPDI. and LPDI. α1 gives the immediate
or short run impact of change in
LPDI on change in LPCE
We can treat the error term as the “equilibrating” error term that
corrects deviations of LPCE from its equilibrium value

DLPCE t = a 0 + a 1 DLPDI t + a 2 u t -1 + v t

error correction model


Error Correction Mechanism
• Granger Representation Theorem: If two variables Y and X are
cointegrated, the relationship between the two can be expressed as
an error correction mechanism (ECM).

• The ECM postulates that changes in the dependent variable depend


on changes in the independent variable and the lagged equilibrium
error term, ut-1.
The coefficient of the error term suggests that DLPCE t = a 0 + a 1 DLPDI t + a 2 u t -1 + v t
about 6% of the discrepancy between long term
Regression Statistics
and short term PCE is corrected within a given Multiple R 0.437545
time period. R Square 0.191445
This means that PCE(Y) and PDI(X) have a long Adj R Square0.180736
Standard Error
0.006138
term relationship which arises out of short term Observations 154
error correction mechanism. If the predicted Y in
ANOVA
the previous period is lesser than the actual Y in df SS MS F SigF
the previous period (positive residual) then this Regression 2 0.001347 0.000673 17.87649 1.08E-07
Residual 151 0.005689 3.77E-05
error would be corrected in the current period Total 153 0.007036
and Yt would be induced to increase because of
error correction. Similarly, if the predicted Y in Coeff Std error t Stat P-value Lower 95%Upper 95%
Intercept 0.005545 0.000627 8.836984 2.36E-15 0.004305 0.006784
the previous period is greater than the actual Y in dlpdi 0.308224 0.051729 5.958471 1.72E-08 0.206018 0.410429
the previous period (negative residual) then this lagehat -0.06193 0.033862 -1.82902 0.069369 -0.12884 0.00497

error would be corrected in the current period Lagged residuals

and Yt would be induced to decrease because of


error correction.
Steps for estimating an Error Correction model
• Run the level regression: Yt=bo+b1Xt+b2t+ut
• Predict residuals (et)
• Perform ADF on residuals to confirm that residuals are stationary
• If residuals are stationary, then Yt and Xt are cointegrated.
• To run the error correction model, take difference of Yt and Xt
• Run the following regression (error correction model):
ΔYt=bo+b1ΔXt+γet-1+vt

You might also like