Elder-Serletis(2010) VAR-GARCH-M

[TomDoan]

Hi Tom,

In March 2013 I asked about the correct expansion of the model. However I think this could be wrong. I feel that this is the correct expansion of the model:

BVEC(1)(*)'s correspond to the oil equation, while B(2)(*) correspond to the GDP equation, and
y1=Oil price change
y2=GDP growth
y1(t)=BVEC(1)(9) + BVEC(1)(1)y1(t-1)+BVEC(1)(5)y2(t-1) + BVEC(1)(2)y1(t-2)+BVEC(1)(6)y2(t-2) + BVEC(1)(3)y1(t-3)+BVEC(1)(7)y2(t-3) + BVEC(1)(4)y1(t-4)+BVEC(1)(8)y2(t-4) + 0*H_{1,1} +\epsilon_{t1}
y2(t)=-B*y1(t)+ BVEC(2)(9) + BVEC(2)(1)y1(t-1)+BVEC(2)(5)y2(t-1) + BVEC(2)(2)y1(t-2)+BVEC(2)(6)y2(t-2) + BVEC(2)(3)y1(t-3)+BVEC(2)(7)y2(t-3) + BVEC(2)(4)y1(t-4)+BVEC(2)(8)y2(t-4) +BVEC(2)(10) *H_{1,1}^{1/2} +\epsilon_{t2}

Could you check if this is right?

Thanks

Miao

Your order is a bit confusing, but yes, that's correct.

[TomDoan]

How could we see the estimation results for the corresponding SVAR model without GARCH? Thanks,
Miao

It's just a standard VAR with a Cholesky factor in the order oil-GDP.

[miao]

Hi, Tom,

Sorry I am a little new to RATS.
Could you give the codes that produce the estimation results for the corresponding SVAR model (estimator, standard deviation, T, p-value)?

Thanks,

Miao

[TomDoan]

Hi, Tom,

Sorry I am a little new to RATS.
Could you give the codes that produce the estimation results for the corresponding SVAR model (estimator, standard deviation, T, p-value)?

Thanks,

Miao

There's nothing really to "estimate". The coefficient on which they concentrate in the paper is the "M" effect on the GARCH, which isn't included in the standard VAR since there isn't any GARCH effect. There are also no "GARCH" coefficients. The standard lag coefficients aren't interesting and are effectively never reported. The VAR is estimated at the start of the program to provide the information criterion for comparison.

Again, the "structural" part is simply a Cholesky factor model. That's very, very simple in the VAR case (no one would even call it a "structural" VAR). It does, however, become a rather complicated model in the GARCH case because it tightly restricts the multivariate GARCH process.

[ecofin]

hello Tom

I am trying to make a study of contagion between different stock market. I plan to use a VAR GARCH-M model. the methodology Elder Sender and the uncertainty of oil price responds to the methodology that I want to follow my questions are:
1 - what is the parameter that measures the effect of uncertainty in the price of petrol on GDP
2 - why they isolate the effect garchp GARCH (1) (3) = 0.0? and I did not understand the provennce 3 because it has only 2 variables.
3 - How do you interpretation of other parameters ie bvec (1) (1) ......... bvec (1) (10) and bvec (2) (1) ..... ...... bvec (2) (10)?
4 - How to estimate taking into account the effect M?
ewise varcoeffs(i,j)=bvec(j)(i)
compute %modelsetcoeffs(basevar,varcoeffs)
*
* Extract the coefficients of the VAR with lag.
*
ewise phi(i)=%modellagmatrix(basevar,i)
if I understand well the parameters from the code are in the same order as writing the VAR model ? the GARCH parameters : You can write to me in the form of the model Based on GARCH parameter from the code?
thank you for understanding

[TomDoan]

If you're doing two stock markets, you would not want to use this model the way it's set up, since GDP is treated quite differently from the oil price variable (it has a very weak GARCH effect and the M effects are assumed to run only from oil to GDP). The "structural VAR" part is also based upon a recursive ordering from oil to GDP.

There's a 20 page chapter in the GARCH e-course that goes through this example in considerable detail if you really want to understand it. However, again, I'm not sure this is the model on which you want to base your analysis.

[ecofin]

If you're doing two stock markets, you would not want to use this model the way it's set up, since GDP is treated quite differently from the oil price variable (it has a very weak GARCH effect and the M effects are assumed to run only from oil to GDP). The "structural VAR" part is also based upon a recursive ordering from oil to GDP.

There's a 20 page chapter in the GARCH e-course that goes through this example in considerable detail if you really want to understand it. However, again, I'm not sure this is the model on which you want to base your analysis.

hi tom

the fact I wanted this model VAR GARCH-M to see the effect of a shock from a market i to market j. if I understand the METHODOLOGY of sender and ender on the uncertainty of oil prices on economic activity USA. is type METHODOLOGY I want to use to see if indeed there has been contagion of sovereign debt in the European market. To do this I want to divide my sample into three study periods (pre-crisis, crisis, post-crisis) to see if the shocks are different between countries and also to see the countries that responded more to shock Greece crisis.

is that I can have on your approach?

thank you for your comprehesion.

[TomDoan]

Please understand that this is not "our" model or "our" approach. It's the model of Elder and Serletis. We have simply provided a program to do their analysis using RATS.

However, from the description of what you are doing, it sounds like you need a somewhat different model. As I have said, this was designed for studying the relationship between oil prices and GDP (or some other macro aggregate). Bond prices have a different dynamic than oil prices and have a very different dynamic than GDP. Haven't you done a literature review on multivariate GARCH models applied to bond prices? That's where you should start.

[cczzwhy]

Excuse me,Mr Tom.I got a problem when I applied this code to my own data,I can get every steps perfectly,except the last impulse response.

I get different paths each time ,and the correct one just showed one time.I wonder if you know how to solve it.

[TomDoan]

Download a new copy of the two .RPF files from the original post in the thread. It should work better.

[cczzwhy]

Thank you very much,it works fine now.

[cczzwhy]

Hello,Mr Tom

It`s me again,I got weird impulse responses as follow ,the paths do not start from zero,can you help me?

[cczzwhy]

Sorry,here is the graph

[TomDoan]

They look fine. I'm not sure what you mean by "start at zero". The initial shocks are mirror images of each other---the subsequent behavior isn't a mirror image because the model is non-linear but they certainly look reasonable.

[cczzwhy]

Hi,Tom!
I got some details need to confirm about this model...I did the arch effect test as you pasted above,and got the results like this

Code: Select all

Test for ARCH in
Using data from 1991:06 to 2015:02

Lags Statistic Signif. Level
   4     0.543       0.70445


Test for ARCH in
Using data from 1991:06 to 2015:02

Lags Statistic Signif. Level
   4     4.873       0.00082

Is it fine?