One-step-ahead Rolling Forecast

[Todor Penchev]

Hello guys,

I am a MSc student in England and i am studying Financial Risk management. Currently I am working on a project and I need to produce one-step-ahead forecast for an out-of-sample period but I have no idea what to do. I have some basic STATA knowledge but haven't been that deep yet. Basically the thing I want to do is the following:

I want to use different GARCH models (for the moment just the standard symmetrical GARCH (1,1) and the asymmetrical EGARCH (1,1)) to forecast one step-ahead daily Value-at-Risk. My in-sample period starts in 2000 until 31st Dec 2009, and i want to use it to forecast the VaR for the next 5 years, which will be the out-of-sample (up to the end of 2014). Then i will generate the actual returns for the out of sample, and compare them with the forecasted VaR, and see which model has the fewest violations depending on the confidence interval.

I know that i need to program STATA to do a rolling window forecast, which moves one step ahead then includes the new observation and removes the first etc. I need to forecast the variance of the model (say GARCH 1,1- i know i can obtain the current conditional variance with the predict cv, variance command) and its mean return, which i will later apply to the VaR formula to calculate it.

I have found a lot of studies that do this method but none of them explain how to do the actual forecasting part for the out of sample period. I also read some STATA books but am not able to understand much. I am new to this program and still have a lot to learn.

Can anyone tell me how i can do a one-step-ahead rolling forecast like that, or direct me to some good articles that i can read to understand this better?

Thanks in advance,
Todor

[Scott Merryman]

Perhaps these programs by Robert Yaffee would be useful:

https://ideas.repec.org/p/boc/usug10/10.html

Abstract:
"Forecasters are expected to provide evaluations of their forecasts along with their forecasts. The forecast assessments demonstrate comparative, adequate, or optimal accuracy by common forecasting criteria to provide acceptable credence in the forecasts. To assist the Stata user in this process, Robert Yaffee has written Stata programs to evaluate ARIMA and GARCH models. He explains how these assessment programs are applied to one-step-ahead and dynamic forecasts, ex post and ex ante forecasts, conditional and unconditional forecasts, as well as combinations of forecasts. In his presentation, he will also demonstrate how assessment can be applied to rolling origin forecasts of time-series models."

[Todor Penchev]

Thank you very much Scott, will take a look at that.