Facebook released disappointing earnings yesterday (July 25, 2018) after the close and fell about 20%. While a -20% jump due to earnings is actually quite routine, the associated market value loss was a record. According to Bloomberg,

The company’s shares fell the most in its history as a public company, wiping out more than $120 billion in market value. It marks the largest ever loss of value in one day for a U.S. traded company.

Here is a chart of the stock price move after about 2 hours into today’s regular session:

In financial modelling, there are various approaches to handling jumps. The most common is probably “Poisson-driven” jumps,
which are also known as compound Poisson models. In those, you don’t know either when the jump will occur or what magnitude it
will be.  When you add stochastic volatility, such models are often called SVJ models, such as the ones I discuss here.

Such models are inappropriate for jumps associated to earnings releases because you know exactly when the jump will occur.

Instead, drop the Poisson factor: all you need to model is the jump distribution. (And often a log-normal suffices for the price jump). I like to call these types of events “scheduled jumps”. There are several pages of discussion in “Option Valuation under Stochastic Volatility II”, mostly using the stocks with “equity VIXes”, as examples.

From an investor’s point of view, the good thing about (bad) earnings release jumps is that they are largely idiosyncratic. In this case,
although there was some spill-over to other internet stocks (and QQQ), the broad US equity market was largely unchanged by the event. As always, an investor’s first line of defense in the stock market is diversification.

While there’s no Facebook VIX, as I point out in the mentioned book, one could always construct one historically — perhaps easiest would be to use the “old VIX” methodology. It would be interesting to do so and see how that behaved through this event. The point of any VIX method is the benefit of a ‘standardized’ implied volatility when studying how the options implied vol changed across the jump event. In general, earnings releases impose  a “sawtooth-like” seasonal pattern upon the (otherwise noisy) time series of option implied volatility.