I am trying to calculate VaR using historical time series for FX CCY pairs that are pegged or managed float. Can you suggest how can I quantify and incorporate probability of peg breaking (given that it is a rare event).
The old peso problem... The answer is most likely that you cannot get useful results from the time series of the individual currency pairs you care about. Maybe look at a very long time series of a large number of exchange rates and estimate some general frequency of breaks, perhaps conditional on the amount of carry on offer? Or, if you have access to option prices, try to imply a jump intensity parameter?