For much of the summer the market had largely been looking at the Scotland referendum as a relatively low risk event. Common wisdom was that the vote would likely be "no". What binds is more than what separates? Perhaps another factor was simply that it was just too far away for the market, as evidenced by the relative lack of media coverage outside of Scotland until recently.
However, market expectations clearly changed with the YouGov poll in last week's Sunday Times (7th Sep), which showed the "yes" vote in the lead. The immediate market reaction was to dump GBP and buy GBP vol as soon as markets open on the Sunday evening (see chart which illustrates the different reaction of GBP/USD implied to Scotland news to EUR/USD implied). Since then GBP has managed to recover to some extent, helped by some new polls which suggested "no" was in the lead. This leads us to the question of how precisely can we quantify the market risks of such a scheduled event? One way we can do this is by looking at implied volatility as a measure of the market's expectations around the future risk of any asset. Implied volatility is after all a function of not just the general volatility environment but also the expectations of realised volatility around future scheduled events.
Typically, before scheduled events, vol market makers will mark add-on weights for these events. For the US employment report this amounts to around 3-4 vol points on an overnight EUR/USD. The precise figure obviously depends on the market demand around dates and the general importance attached to the event by the market. Generally speaking, if say the US employment report has been a non-event the last few times, the market might not be as keen to bid up vol around an event. The converse can also occur. Given the impact of the scheduled event will be unknown, the market's event implied vol add-on associated it will obviously be an estimate.
As a result, implied volatility over such events generally tend to exhibit a large risk premium. In other words, implied vol tends to be higher over these events then realised volatility.More broadly, implied vol tends to be higher than realised volatility, but the presence of scheduled events can widen this difference (see Thalesians - What's in an event - Understanding FX implied vol event add-ons - 03 Nov 2013). We could argue that because the scheduling is known, market participants might overemphasis hedging its risk over these dates. (You won't get fired for paying up vol over payrolls, but you might be fired that one time realised explodes!)
We can attempt to monetise risk premium via gamma trading (see What's gamma trading - 22 Jul) from the short side (see Thalesians - Gamma, gamma, gamma - Explaining gamma trading in FX markets - 14 Apr 2014). We must be very careful how we do this, given that the returns from such a strategy can be skewed. We also need to consider all sorts of trading issues, such as delta hedging, when we run such a strategy. A failure to do could result in much deeper drawdowns on those occasions when realised volatility really does outperform implied.
In the case of Scottish referendum, unlike the US employment report, there are very few comparable precedents. We can for example look to UK general elections. However, the nature of this referendum and general elections. One of the closest similar events is the Quebec referendum in 1995. Admittedly, even the Quebec vote was very different and clearly a sample of 1 is not enough to make a proper statistical comparison.
The difficulty in pricing the event means that traders will demand a much higher event vol add-on than for an "ordinary" event. Even if the probability might appear "low" of a "yes" vote, the tail risk is clearly something that traders needs to be compensated for. Complicating matters is also the way that poll results have become "vol" events, which traders should mark in their vol calendars (if you're a vol trader and you don't do this, someone else will and you'll be left holding the bag!). To some extent, this can also help us, giving us more "similar" events to extrapolate from. We could also examine realised volatility over the release of various polls, to give us a hint about how price action could behave (with the caveat that a poll is not as important as the actual referendum).
Whichever way you look at it, pricing vol over the referendum date and even leading up to it, is extremely challenging, given the lack of past comparisons, but we can at least try!
To read more about this topic please see the below work I've done on systematic gamma trading and quantifying the value of events in vol terms. My book Trading Thalesians also has some colour on this topic (mixed in with a bit of ancient history).
Thalesians - Gamma, gamma, gamma - Explaining gamma trading in FX markets - 14 Apr 2014 (available for Thalesians clients only)
Thalesians - What's in an event - Understanding FX implied vol event add-ons - 03 Nov 2013 (available for Thalesians clients only)