The RATS Software Forum

A Forum for RATS Software and Econometrics Discussions
https://www.estima.com/forum/

VECMCAUSE.RPF—Causality tests in error correction model

https://www.estima.com/forum/viewtopic.php?t=2912

Page 1 of 1

VECMCAUSE.RPF—Causality tests in error correction model

Posted: Mon Jun 26, 2017 2:39 pm
by TomDoan
VECMCAUSE.RPF is an example of testing for causality in a VECM. This has separate tests for causality and long-run causality.

Detailed description

w-tb3n6ms.txt
Data file
(62.86 KiB) Downloaded 1454 times

Re: VECMCAUSE.RPF—Causality tests in error correction model

Posted: Thu Jun 21, 2018 6:45 am
by upani
Hi,
I want to know whether this programme is suitable for checking weak exogeneity test.

With regards,
Upani

Re: VECMCAUSE.RPF—Causality tests in error correction model

Posted: Thu Jun 21, 2018 8:13 am
by TomDoan
X being weakly exogenous (for the beta in the cointegration vector) in the (X,Y) relationship is another way of stating that Y doesn't cause X long-run.

Re: VECMCAUSE.RPF—Causality tests in error correction model

Posted: Thu Jun 04, 2020 8:04 am
by upani
Dear All,

I am estimating the VECM with spot and future price. I am using vecmcause.rpf function to estimate this.

Code: Select all

*VECM Short-run Causality and Long-run Causality
@johmle(lags=5,det=rc,cv=cv)
# LFP LSP
LINREG DLFP
# Z{1} DLFP{1 to 4} DLSP{1 to 4}
set z = LFP-LSP
linreg dlfp
# z{1} dlfp{1 to 4} dlsp{1 to 4}
exclude(title="Test for Spot Price causing Future Price")
# z{1} dlsp{1 to 4}
exclude(title="Test for Spot Price long-run causing Future Price")
# z{1}
*
* Futue Price causing Spot Price
*
linreg dlsp
# z{1} dlfp{1 to 4} dlsp{1 to 4}
exclude(title="Test for Future Price causing Spot Price")
# z{1} dlfp{1 to 4}
exclude(title="Test for Future Price long-run causing Spot Price")
# z{1}
After running the code i got the result which is giving the following result.

Code: Select all

Linear Regression - Estimation by Least Squares
Dependent Variable DLFP
Weekly Data From 1959:01:16 To 1972:11:17
Usable Observations                       723
Degrees of Freedom                        714
Centered R^2                        0.0183605
R-Bar^2                             0.0073617
Uncentered R^2                      0.0184954
Mean of Dependent Variable       0.0005295308
Std Error of Dependent Variable  0.0451958154
Standard Error of Estimate       0.0450291490
Sum of Squared Residuals         1.4477237217
Log Likelihood                      1220.2576
Durbin-Watson Statistic                1.9975

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  Z{1}                         -0.137533229  0.213611823     -0.64385  0.51988172
2.  DLFP{1}                       0.055408510  0.196688977      0.28171  0.77825040
3.  DLFP{2}                      -0.045075649  0.176082637     -0.25599  0.79803117
4.  DLFP{3}                      -0.298758427  0.149946042     -1.99244  0.04670317
5.  DLFP{4}                      -0.200389465  0.108827208     -1.84135  0.06598428
6.  DLSP{1}                      -0.011645722  0.195881453     -0.05945  0.95260801
7.  DLSP{2}                       0.073686676  0.175396297      0.42012  0.67452759
8.  DLSP{3}                       0.319689325  0.146596548      2.18074  0.02952816
9.  DLSP{4}                       0.192572149  0.102848127      1.87239  0.06156071


Linear Regression - Estimation by Least Squares
Dependent Variable DLFP
Weekly Data From 1959:01:16 To 1972:11:17
Usable Observations                       723
Degrees of Freedom                        714
Centered R^2                        0.0183605
R-Bar^2                             0.0073617
Uncentered R^2                      0.0184954
Mean of Dependent Variable       0.0005295308
Std Error of Dependent Variable  0.0451958154
Standard Error of Estimate       0.0450291490
Sum of Squared Residuals         1.4477237217
Log Likelihood                      1220.2576
Durbin-Watson Statistic                1.9975

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  Z{1}                         -0.137533229  0.213611823     -0.64385  0.51988172
2.  DLFP{1}                       0.055408510  0.196688977      0.28171  0.77825040
3.  DLFP{2}                      -0.045075649  0.176082637     -0.25599  0.79803117
4.  DLFP{3}                      -0.298758427  0.149946042     -1.99244  0.04670317
5.  DLFP{4}                      -0.200389465  0.108827208     -1.84135  0.06598428
6.  DLSP{1}                      -0.011645722  0.195881453     -0.05945  0.95260801
7.  DLSP{2}                       0.073686676  0.175396297      0.42012  0.67452759
8.  DLSP{3}                       0.319689325  0.146596548      2.18074  0.02952816
9.  DLSP{4}                       0.192572149  0.102848127      1.87239  0.06156071


Test for Spot Price causing Future Price

Null Hypothesis : The Following Coefficients Are Zero
Z                Lag(s) 1
DLSP             Lag(s) 1 to 4
F(5,714)=      2.37598 with Significance Level 0.03753272


Test for Spot Price long-run causing Future Price

Null Hypothesis : The Following Coefficients Are Zero
Z                Lag(s) 1
t(714)=  -0.643847 or F(1,714)=   0.414538 with Significance Level 0.51988172


Linear Regression - Estimation by Least Squares
Dependent Variable DLSP
Weekly Data From 1959:01:16 To 1972:11:17
Usable Observations                       723
Degrees of Freedom                        714
Centered R^2                        0.1189251
R-Bar^2                             0.1090531
Uncentered R^2                      0.1190266
Mean of Dependent Variable       0.0005046662
Std Error of Dependent Variable  0.0470507803
Standard Error of Estimate       0.0444112244
Sum of Squared Residuals         1.4082627910
Log Likelihood                      1230.2479
Durbin-Watson Statistic                2.0008

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  Z{1}                          0.661788913  0.210680477      3.14120  0.00175205
2.  DLFP{1}                       0.280735548  0.193989859      1.44717  0.14828930
3.  DLFP{2}                       0.115947684  0.173666296      0.66765  0.50457504
4.  DLFP{3}                      -0.177690602  0.147888368     -1.20152  0.22994854
5.  DLFP{4}                      -0.106650721  0.107333797     -0.99364  0.32073674
6.  DLSP{1}                      -0.202259818  0.193193417     -1.04693  0.29548654
7.  DLSP{2}                      -0.075602209  0.172989374     -0.43703  0.66221896
8.  DLSP{3}                       0.218149954  0.144584837      1.50880  0.13179163
9.  DLSP{4}                       0.125687247  0.101436766      1.23907  0.21572675


Test for Future Price causing Spot Price

Null Hypothesis : The Following Coefficients Are Zero
Z                Lag(s) 1
DLFP             Lag(s) 1 to 4
F(5,714)=     18.68860 with Significance Level 0.00000000


Test for Future Price long-run causing Spot Price

Null Hypothesis : The Following Coefficients Are Zero
Z                Lag(s) 1
t(714)=   3.141197 or F(1,714)=   9.867120 with Significance Level 0.00175205
I am bit confused over the interpretation as spot is not having any long-run relationship with future but future is having a long run relationship with spot.
I will grateful if you help me to understand the issue over here. Because Weak exogeneity test suggest the reverse.

Thanks and Regards,
Upananda

Re: VECMCAUSE.RPF—Causality tests in error correction model

Posted: Thu Jun 04, 2020 8:41 am
by TomDoan
I'm not sure what you mean by "weak exogeneity test suggests the reverse". What you're citing IS the (or at least a) test for weak exogeneity. The interpretation is that the forward price is weakly exogenous in the cointegrating relationship.

Re: VECMCAUSE.RPF—Causality tests in error correction model

Posted: Thu Jun 04, 2020 12:46 pm
by upani
Sir,

In the first equation results

Test for Spot Price causing Future Price

Null Hypothesis : The Following Coefficients Are Zero
Z Lag(s) 1
DLSP Lag(s) 1 to 4
F(5,714)= 2.37598 with Significance Level 0.03753272


Test for Spot Price long-run causing Future Price

Null Hypothesis : The Following Coefficients Are Zero
Z Lag(s) 1
t(714)= -0.643847 or F(1,714)= 0.414538 with Significance Level 0.51988172

As per the above results, there is a short-run causality from spot to future price but there is no long-run causality from spot to future. Am i correct?

With regards,
Upananda

Re: VECMCAUSE.RPF—Causality tests in error correction model

Posted: Thu Jun 04, 2020 12:55 pm
by TomDoan
There is fairly weak evidence of causality (.03 isn't all that significant at a data set that size), but yes, you would conclude that there is causality but not long-run causality.

Re: VECMCAUSE.RPF—Causality tests in error correction model

Posted: Thu Jun 04, 2020 1:23 pm
by upani
Dear Sir,

Thanks a lot for clarifying my doubt. Similarly in the second equation

Code: Select all

Linear Regression - Estimation by Least Squares
Dependent Variable DLSP
Weekly Data From 1959:01:16 To 1972:11:17
Usable Observations                       723
Degrees of Freedom                        714
Centered R^2                        0.1189251
R-Bar^2                             0.1090531
Uncentered R^2                      0.1190266
Mean of Dependent Variable       0.0005046662
Std Error of Dependent Variable  0.0470507803
Standard Error of Estimate       0.0444112244
Sum of Squared Residuals         1.4082627910
Log Likelihood                      1230.2479
Durbin-Watson Statistic                2.0008

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  Z{1}                          0.661788913  0.210680477      3.14120  0.00175205
2.  DLFP{1}                       0.280735548  0.193989859      1.44717  0.14828930
3.  DLFP{2}                       0.115947684  0.173666296      0.66765  0.50457504
4.  DLFP{3}                      -0.177690602  0.147888368     -1.20152  0.22994854
5.  DLFP{4}                      -0.106650721  0.107333797     -0.99364  0.32073674
6.  DLSP{1}                      -0.202259818  0.193193417     -1.04693  0.29548654
7.  DLSP{2}                      -0.075602209  0.172989374     -0.43703  0.66221896
8.  DLSP{3}                       0.218149954  0.144584837      1.50880  0.13179163
9.  DLSP{4}                       0.125687247  0.101436766      1.23907  0.21572675


Test for Future Price causing Spot Price

Null Hypothesis : The Following Coefficients Are Zero
Z                Lag(s) 1
DLFP             Lag(s) 1 to 4
F(5,714)=     18.68860 with Significance Level 0.00000000


Test for Future Price long-run causing Spot Price

Null Hypothesis : The Following Coefficients Are Zero
Z                Lag(s) 1
t(714)=   3.141197 or F(1,714)=   9.867120 with Significance Level 0.00175205
Here there is a short-run causality from future to spot and there is a long-run causality from future to spot. In the first equation there is a weak long-run relationship between spot and future. But Here there strong long-run causality between future and spot. Is not it the same inference given that in the long-run equilibrium spot price will be equal to future price or vice versa.

With sincere regards,
Upananda

Re: VECMCAUSE.RPF—Causality tests in error correction model

Posted: Mon Dec 30, 2024 9:42 am
by TomDoan
The equilibrium is Spot=Future. They are never equal, but the process is pushed towards that. Weak exogeneity of the futures price (i.e. if the actual loading on the disequilibrium is zero in the futures equation) means that the long-run adjustment is entirely through the spot price. Which isn't that hard to imagine given that the alternative in the spot market is to take physical delivery, which is expensive.
All times are UTC-06:00
Page 1 of 1

Powered by phpBB® Forum Software © phpBB Limited