One part of my dissertation involves forecasting the realised volatilities of interest rates. I know that I have to first model the realised variance as a garch process.
So, what I have done so far is, I got the swap rates from 2000-2006. Then, I calculated the logarithmic daily changes of them. Afterwards, I estimated a GARCH 1,1 process (although I am not sure if that was done right).
Then, my next forecast equation (in theory) is rt+h = c0 = c1*rt-1+h + e
rt = logarithmic daily changes
h= forecast period
How do I now forecast the realised variance for the next 10 years e.g.?
Since I am not the best at using EViews, I would appreciate any help/hints/assistance!
What are the steps of forecasting realised volatility of swap rates using EViews?
Moderators: EViews Gareth, EViews Moderator
Re: What are the steps of forecasting realised volatility of swap rates using EViews?
can you post the workfile?
Is easier to see what you need to do and understand what are you doing.
Is easier to see what you need to do and understand what are you doing.
Re: What are the steps of forecasting realised volatility of swap rates using EViews?
Sure, Sorry I didn't do that earlier!
- Attachments
-
- ForecastSwaps.wf1
- (96.57 KiB) Downloaded 240 times
Re: What are the steps of forecasting realised volatility of swap rates using EViews?
Hello David,
After checking the correlogram of rt it seems to follow an ARMA(1,1) which you have to estimate the "rt c ar(1) ma(1)".
After that you need to conduct the Heteroskedasticity Test, which confirms the ARCH effects. so You need to estimate the ARMA(1,1)-GARCH(1,1).
After that to perorm the out-sample forecast, You need to rearrange you range size and than perform the forecast with the method you want and for the sample you want.
I will attache the screenshot explaining and the workfile.
After checking the correlogram of rt it seems to follow an ARMA(1,1) which you have to estimate the "rt c ar(1) ma(1)".
After that you need to conduct the Heteroskedasticity Test, which confirms the ARCH effects. so You need to estimate the ARMA(1,1)-GARCH(1,1).
After that to perorm the out-sample forecast, You need to rearrange you range size and than perform the forecast with the method you want and for the sample you want.
I will attache the screenshot explaining and the workfile.
- Attachments
-
- david.jpg (153.83 KiB) Viewed 5045 times
-
- forecastswaps.wf1
- (104.81 KiB) Downloaded 239 times
Re: What are the steps of forecasting realised volatility of swap rates using EViews?
Hi pvestia,
Thank you for your quick answer!
I adjusted the range size, however, when I do the dynamic forecast, there is only a straight line from 2006 onward. Any idea why?
If I'm right, I can't use the static forecasting, since I don't have the actual values, correct?
Thank you for your quick answer!
I adjusted the range size, however, when I do the dynamic forecast, there is only a straight line from 2006 onward. Any idea why?
If I'm right, I can't use the static forecasting, since I don't have the actual values, correct?
Re: What are the steps of forecasting realised volatility of swap rates using EViews?
Hello, Yes with the static forecast you are only abble to forecast One-period because you only have an auto-regressive lag. The result of the dynamic forecast in Eviews I don't understand them quite well. They should converge to zero, but is kind of odd being straight line, I don't know is suppose to be like that.
One thing you can do, I'm trying something similar, is to perform multiples static forecasts in a loop. I don't know with this procedure I'm proposing you doesn't generate regressor bias, but you can give it a try.
I will give you a code, I didn't resize the data in my workfile because is irregular and I don't know the dates you want. To use the code you just go to Files\New\Program and then copy and run it.
After you done this, you estimate again the ARMA(1,1)-GARCH(1,1) and do a static forecast for the all range and see what you got.
One thing you can do, I'm trying something similar, is to perform multiples static forecasts in a loop. I don't know with this procedure I'm proposing you doesn't generate regressor bias, but you can give it a try.
I will give you a code, I didn't resize the data in my workfile because is irregular and I don't know the dates you want. To use the code you just go to Files\New\Program and then copy and run it.
Code: Select all
' number of obseration you added to the range for your forecast
' based on you screen shot the number is 5211-2600=2611
!x = 2611
'THE LOOP STARTS
'define loop for the forecast
for !i=1 to !x
'make the ARMA-GARCH
smpl @first @first+2598+!i
equation arma_garch.arch(1,1) rt c ar(1) ma(1)
'perform the static forecast
arma_garch.makemodel(mod1)
smpl @first+2599+!i @first+2599+!i
mod1.solve(d=s)
smpl @all
series rt = rt_0
'delete the auxiliar
delete rt_0
next
After you done this, you estimate again the ARMA(1,1)-GARCH(1,1) and do a static forecast for the all range and see what you got.
Who is online
Users browsing this forum: No registered users and 34 guests