GARCH-ARCH relating conditional volatility to unconditional volatility

2018-03-09 08:43:59

After comparing the inferred conditional volatilities from GARCH models (using Matlab) with the unconditional volatilities from the actual training set, I noticed that although the general trends matched, the inferred conditional volatilities were always lower. Was just wondering if this is an error on my modeling end, or whether this is the actual case? If it's the latter, was just wondering how quants map forecasted conditional volatilities to unconditional volatilities? Is it enough to say that vol will increase/decrease and make a bet one way or the other? Thanks in advance.