Nonlinear ARDL using Eviews

TIME SERIES ECONOMETRICS WORKSHOP:

"Asymmetric Co-integration, NARDL and Structural VAR"

by Professor Mansor Ibrahim

Session 2: Nonlinear ARDL using Eviews

1. Use this dataset: LINK,
2. Type "program" in the command area to open program window
    [or Menu -> File -> New -> Program],
3. Copy, paste and run the following code:

GENR X=LWTI

GENR DX=X-X(-1)

GENR POS=DX>=0

GENR DXPOS=POS*DX

GENR DXNEG=(1-POS)*DX

GENR XPOS=@CUMSUM(DXPOS)

GENR XNEG=@CUMSUM(DXNEG)

GENR LWTI_POS=XPOS

GENR LWTI_NEG=XNEG

DELETE X DX POS DXPOS DXNEG XPOS XNEG
GENR DWTI_POS=LWTI_POS-LWTI_POS(-1)
GENR DWTI_NEG=LWTI_NEG-LWTI_NEG(-1)

4. Run the following OLS regression:

DFP c LFP(-1) LY(-1) LWTI_POS(-1) LWTI_NEG(-1) DFP(-1 TO -3) DY DY(-1 TO -3) DWTI_POS DWTI_POS(-1 TO -3) DWTI_NEG DWTI_NEG(-1 TO -3)

5. Generated output

 6. Use General-to-Specific approach do identify the right number of lags, i.e. trim down lags if higher lags are found to be insignificant. So, we find lag 3 insignificant. Next, we define 2 lags:

DFP c LFP(-1) LY(-1) LWTI_POS(-1) LWTI_NEG(-1) DFP(-1 TO -2) DY DY(-1 TO -2) DWTI_POS DWTI_POS(-1 TO -2) DWTI_NEG DWTI_NEG(-1 TO -2)

and get the following results:

Again, 2 lags are insignificant. Next, trying with 1 lag only.

DFP c LFP(-1) LY(-1) LWTI_POS(-1) LWTI_NEG(-1) DFP(-1) DY DY(-1) DWTI_POS DWTI_POS(-1) DWTI_NEG DWTI_NEG(-1)

DFP(-1) and DWTI_POS(-1) are significant at 10% s.l. --> we retain. Next, run:

DFP c LFP(-1) LY(-1) LWTI_POS(-1) LWTI_NEG(-1) DFP(-1) DY  DWTI_POS DWTI_POS(-1) DWTI_NEG

Next, after the adjustment, run the following:

DFP c LFP(-1) LY(-1) LWTI_POS(-1) LWTI_NEG(-1) DFP(-1)  DWTI_POS DWTI_POS(-1) 

Finally:

DFP c LFP(-1) LY(-1) LWTI_POS(-1) LWTI_NEG(-1) DFP(-1)  DWTI_POS

Do not close this output.

Next, we test for joint significance of lagged level variables (i.e. cointegration):

View --> Coefficient Diagnostics --> Wald test --> C(2)=C(3)=C(4)=C(5)=0

Output:

Next, to calculate Beta 1 = -(LY(-1)/LFP(-1)) = - 0.262796 / - 0.508185,

Go to View --> Coefficient Diagnostics --> Wald test --> -C(3)/C(2)=0

Output:

Next, to calculate Beta 2 = -(LWTI_POS(-1)/LFP(-1)),

Go to View --> Coefficient Diagnostics --> Wald test --> -C(4)/(2)=0

Output:

Recommended literature:
- Learner, "Let's Take Cons Out of Econometrics"
- Zellner (2007). JEconometrics
- Prof. Hendry's Econometric Methodology