rugarch: Forecast result does not show any AR structure

Question

I am currently working with the rugarch package to forecast the EU-ETS price. While I get reasonable results for the in-sample volatility, the forecast of the of the time series does not look correct at all:

Is this because of the low ar1 and ar2 parameter estimates? If so, is there a way to overcome this problem? I have daily observations (n=3,000)

-----------------------------------
GARCH Model : eGARCH(1,1)
Mean Model  : ARFIMA(2,0,0)
Distribution    : sstd 

Optimal Parameters
------------------------------------
        Estimate  Std. Error  t value Pr(>|t|)
mu      0.000703    0.000282   2.4935 0.012651
ar1    -0.037111    0.014955  -2.4815 0.013084
ar2    -0.027603    0.011804  -2.3385 0.019361
omega  -0.165146    0.020888  -7.9061 0.000000
alpha1 -0.035044    0.010961  -3.1971 0.001388
beta1   0.977160    0.002873 340.1154 0.000000
gamma1  0.212031    0.020791  10.1981 0.000000
skew    0.978063    0.021619  45.2417 0.000000
shape   5.448983    0.471307  11.5614 0.000000

Looking forward to your advice.... Thanks!

Answer 1

Yes, your forecast is most likely almost flat because of the low ar1 and ar2 parameter estimates. But is that a reason for concern? There are many threads on Cross Validated asking about flat forecasts. The answers in these threads explain why they are not a problem. In one of them, Stephan Kolassa writes:

If you are concerned that the forecast does not reproduce the variability in your historical data: don't be. Forecasting models attempt to disentangle the signal from the noise and only extrapolate the signal, because the noise is - by definition - not forecastable. Therefore, any forecast will look smoother than the original time series.