Scribd
Upload
392 views

ARDL Model

1) The autoregressive distributed lag (ARDL) model allows for variables that are integrated of order zero or one. It estimates both long-run and short-run relationships in a single equation framework. 2) The ARDL procedure involves testing for cointegration using the bounds test. If cointegration is found, it estimates the long-run coefficients and then generates a dynamic error correction model to capture the short-run dynamics. 3) The error correction model (ECM) contains short-run coefficients as well as the error correction term which represents the speed of adjustment back to long-run equilibrium.

Uploaded by

Nayli Athirah Azizon
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
392 views

ARDL Model

1) The autoregressive distributed lag (ARDL) model allows for variables that are integrated of order zero or one. It estimates both long-run and short-run relationships in a single equation framework. 2) The ARDL procedure involves testing for cointegration using the bounds test. If cointegration is found, it estimates the long-run coefficients and then generates a dynamic error correction model to capture the short-run dynamics. 3) The error correction model (ECM) contains short-run coefficients as well as the error correction term which represents the speed of adjustment back to long-run equilibrium.

Uploaded by

Nayli Athirah Azizon
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 16

ARDL MODEL

Motivation for ARDL


• Some of the variables may be I(0) and some may be I(1)
• There is possibility cointegration among of the variables
• The effect of independent variable may not be instantaneous.
Therefore lag effect can be taken into account –ARDL

ARDL (p,q) regression model:-


ARDL Cointegration Test
• Some advantages of ARDL/Bound Testing methodology of Pesaran
and Shin (1999) and Pesaran et al. (2001) over conventional
cointegration test :-
- The variables can be mixed I(0) and I(1)
- It involves single-equation set up
- Different lag-lengths can be assigned for different variables
ARDL-ECM Representation

𝜀
Steps of ARDL Modelling
ARDL Cointegration Test
• Based on Pesaran et al (2001), ARDL bound test can be conducted
at uniform lag of unrestricted model (UECM). So ARDL (p,p,p,p).
Choose the model with lowest AIC where residuals are tested to be
independent and the model are stable. Compare F-statistics wit =h
critical value

• ARDL level relation: Adopt ARDL (p,q,r,s) approach to estimation of


the level relations. The model can be selected based on ‘AIC/SBC’
criteria or using ‘general to specific approach’
Important Procedures in ARDL Modelling

1. Detect cointegation among the variables using the bounds test


2. Obtain the long-run coefficients based on ARDL level relation
3. Estimation short-run dynamic coefficients (ECM)
Procedures of ARDL Model
1) The variable cannot have I(2), otherwise-> INVALID
-use unit root test to test the stationarity
2) Unrestricted error correction model (UECM) is developed

-coefficients δ,  and β represents short-term effect of independent


variables on dependent variables while θ measures long-run impact of
IV on DV
3) Determine the appropriate lag structure for the EUCM model,
-use AIC or BIC OR Hendry’s General to Specific approach
Procedures of ARDL Model
1) The variable cannot have I(2), otherwise-> INVALID
-use unit root test to test the stationarity
2) Unrestricted error correction model (UECM) is developed

-coefficients δ,  and β represents short-term effect of independent


variables on dependent variables while θ measures long-run impact of
IV on DV
3) Determine the appropriate lag structure for the EUCM model,
-use uniform lag (Pesaran et al, 2001) or AIC or BIC OR Hendry’s
General to Specific approach
Procedures of ARDL Model
4) Ensure residuals of UECM model are not serially correlated and the
model is stable
-use Breusch-Godfrey(LM) Test
- check stability using CUSUM and CUSUMQ squares test
5) Perform a ‘Bound Test’ using Wald test on long-run coefficients.
ARDL level relation
Calculating long-run coefficients
Short-run model: Restricted ECM
ARDL in EVIEWS
• Eviews starts with estimating ARDL level relation model. The best
model with lag optimum will be chosen based on AIC.
• There is no guarantee that the model satisfies diagnostic checking.
Need to check!
• From ARDL level relation, Eviews reduced down to form UECM and
having the same order with ECM
• If optimal model contains 0 order, the corresponding variable
disappear in first difference term in UECM. Pesaran et al. (2001)
suggests UECM must at least contains current first difference term.
Short-run dynamic equation

You might also like