5
$\begingroup$

While pulling some reports from Bloomberg today I came across the volatility surface for NYSE: ONDK, set for an earnings call next Monday. From what I've seen before (not much, 1 month on the job and counting) volatility should trend upwards with longer terms, and present in the form of a "smile" when you plot volatility against moneyness. The surface for ONDK had volatility dropping down with term, and presented more of a volatility frown than smile.

What exactly does it mean when volatility's highest at close-to-the-money options? And what could drive decreasing volatility with term?

$\endgroup$
  • 1
    $\begingroup$ I currently don't have access to the implied volatility smile of ONDK. However, concave implied volatilities around the forward are a common pattern for short-term options when a special event with a known time of occurrence is priced in (e.g. earnings, French election, ...). See e.g. these answers: quant.stackexchange.com/questions/30749, quant.stackexchange.com/questions/31368. $\endgroup$ – LocalVolatility May 5 '17 at 20:18
  • $\begingroup$ Concave is more likely for single stocks - not so common for macro names. I think it is probably due to retail players, but I cannot prove that. One way to get evidence of it would be to see if the concave down shapes around the money only occur for the hot retail investor names for single stocks - typically tech stocks or biotechs where the gamma seems really sexy. For this particular ticker, that is not the case - but good links - you provided! $\endgroup$ – FinanceGuyThatCantCode May 5 '17 at 20:33
  • 1
    $\begingroup$ I disagree on concavity being due to retail players and mostly a single stock phenomenon. It is a direct consequence of a bi-modal jump being priced in - i.e. there is a good vs. bad outcome. Yes - these events are usually more extreme for stocks. However, around big elections you find significantly concave smiles on indices as well. Past recent examples include i) CAC (but also EuroStoxx, DAX) before the first round of French election end of April and ii) FTSE (and again also EuroStoxx, DAX) before the Brexit vote. $\endgroup$ – LocalVolatility May 5 '17 at 20:37
  • 1
    $\begingroup$ I agree on indices usually staying monotonic and that you see W shapes mostly for single stocks. However, this is to some extend owed to the fact that for the above mentioned macro events, the outcomes are not very symmetric. There is usually one very likely outcome (at least in the perception) - no Brexit, Macron wins. This gives rise to an asymmetric but locally concave implied vol. smile - much like in the DAX plot at the bottom of my second link. When the outcomes are more symmetric as is often the case for earnings, you can observe the W-shapes. $\endgroup$ – LocalVolatility May 5 '17 at 20:59
  • 1
    $\begingroup$ I disagree with your last statement though. Let's say you have a symmetric jump priced in and the volatility is otherwise constant. In that case the vol. is highest near the forward and (locally) concave. This is exactly the situation in the first plot of my second link. $\endgroup$ – LocalVolatility May 5 '17 at 21:01
3
$\begingroup$

I just checked this one. I saw the picture you are referring to and I think the frown is not real. Much of it looks like noise created by wide bid ask spreads and extremely high implied vols. The data here is very poor and the options are spread very far apart. The underlying is $4.60 and the options are struck a dollar apart. That would be like SPX having strikes every 500 points more or less. So the smiles only have 0-2 useful data points in the front months and then BBG's algorithms created a smile off of those points. I know very few people that are detail oriented enough to make good smiles with such sparse data. I am probably the only person I ever met anal enough to try to do this right though I am sure there are others I have not met.

I like to fit the various smiles in a very specific order. I assign a metric to each expiration that measures how much "information" there is in the smile. For that metric, I consider the range of useful strikes in delta space, the number of strikes available to me in that range, and how tight the bid ask spreads are roughly on average. Then I first fit the smile that has the most "information" according to my metric. Then given the last smile that I fit in the surface, I inductively fit the next smile while trying to maintain forward volatilities between this smile and already fit smiles that are sensible and without arbitrage. By enforcing no arbitrage, we actually get smoother term structures cross sections as we vary the strike by moneyness - that to me is very valuable even though I don't price too many exotics. The smoothness of these term structure cross sections makes the term structure of my parameters fitting each smile much smoother. Smoothness of parameters make the parameters meaningful and useful for other things.

My point - doing this kind of detailed work would give you the smiles you are more familiar with.

Another potential issue - and this is a big one - I see the short interest in BBG says that 12.52% of the float is currently short. This means that the cost of borrow is likely to be high and BBG is probably calculating vols under the assumption that borrow cost is not unusual. A very expensive borrow cost will change the forwards - think of the borrow cost as a continuous dividend yield. This can distort vols a lot, but I cannot confirm how expensive this borrow is.

As far as term structure vols go, usually term structures of vols are increasing as you say, but when there is event risk (like earnings) the shorter dated vols might be higher than the longer dated vols. Also, during periods of extreme stress, the vol term structures tend to be decreasing due to the mean reverting nature of vol. In other words, usually the period of stress is short lived so the short dated vol should be very high - and the longer dated vol starting a few months down the road should revert to more normal level - i.e. the forward vols starting a few months out are much lower.

$\endgroup$
  • $\begingroup$ I disagree on a couple of things you say - first off, the borrow cost will not be used to calculate the forward, it will be implied for the difference between the forward (or if it's missing the fwd implied from out call parity and then used), and second, I don't think there's much value in talking in terms of the term structure of term vol - talk in terms of total variance, where the condition is obvious. $\endgroup$ – will May 6 '17 at 0:07
  • $\begingroup$ Also, where you say that you're the only one to be careful in the fitting - I do not think that's true, but I will definitely say that all data companies that publish vol surfaces ***k up from time to time (probably more regularly that anyone would hope) and publish complete garbage surfaces... $\endgroup$ – will May 6 '17 at 0:09
  • $\begingroup$ Absolutely the borrow cost is part of the forward calculation - just follow a replication argument - buy a forward and short the stock which includes you paying the borrow (or sell the forward and buy and lend the stock). No put call parity here with the American exercise. Even with no divs - delta hedging a call with extreme borrow costs when deep ITM might mean early exercise to avoid the borrow costs incurred. Term structure of vol is useful for sure - total variance is useful for no arbitrage considerations, but TS of vol is still useful for the consideration of trading strategies. $\endgroup$ – FinanceGuyThatCantCode May 6 '17 at 14:19
  • 1
    $\begingroup$ As for fitting - let me tell you...so sick and tired of asking countless people to do this for me...in the end I have to do it myself. Thing is - it is not a big intellectual exercise...few people are good at cleaning the data and handling the wings the way I want them handled. A lot of quants just want to get the data into the right format, turn the crank on the optimizer and be done - particularly at hedge funds where resources compared to a bank are limited. Banks can be better, but even there I found the way the wings and bad data were handled to be weak. Probably GS is ok. $\endgroup$ – FinanceGuyThatCantCode May 6 '17 at 14:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.