By Professor George Skiadopoulos, Queen Mary University of London and University of Piraeus A common finding when it comes to options trading is that selling index put option contracts (such as those written on the S&P 500) tends to offer average returns and Sharpe ratios which seem to be excessive when compared to equities. Academic literature has well documented that selling index put options yields high average returns. This is also well known among practitioners: these trades have been highly profitable, except in instances where sales coincide with spikes in volatility, such as when Lehman Brothers crashed, or in the initial heat of the pandemic in March 2020. This raises the question whether seemingly high average put option returns are abnormal, or they can be justified when taking into account the risks of these option strategies. The answer to this long-standing question is not trivial: one needs to account for all possible risks, and measure them accurately. This task becomes more difficult in the case of options compared to equities, because option returns are related to risks in a non-linear fashion. For instance, option price is not a linear function of the underlying put price. The price of a put option on the S&P 500 increases when the S&P 500 price goes down, yet the relationship is not linear. This has been a puzzle for academics and practitioners for years. To date, the economic sources driving this abnormal behavior have remained relatively unexplored; until now. In the paper, “Learning and Index Option Returns,” researchers Alejandro Bernales from University of Chile; Gonzalo Cortazar and Luka Salamunic, both from Pontificia Universidad Católica de Chile, and I, George Skiadopoulos from Queen Mary University of London and University of Piraeus, offer a new economic explanation for what appear to be factors contributing to perceived high abnormal risk/return performance in selling put index option contracts. We provide a new option pricing model that incorporates learning for economic fundamentals. In our model, the agent learns about how the average growth rate of dividends evolves over time, as the growth rate of dividends may change abruptly from time to time due to the occurrence of shocks. Our model extends the Black-Scholes model to explain the observed empirical patterns in the average returns of short index put options. The proposed economic mechanism encompasses risks which affect option returns, and it also takes into account the non-linear relation of option returns and their risk factors. Learning is modeled in a Bayesian setting, where the investor formulates an initial belief about what the expected dividend growth may be, and then he/she updates her initial belief as he/she receives more information. To verify that our model can generate the empirically observed patterns in average short index put options, we first computed the average returns for index put options using actual option data on the S&P 500 across different strikes. To this end, we used IvyDB US Optionmetrics. While there are many sources one can use when it comes to performing academic research, we, have done considerable work with OptionMetrics’ IvyDB US database, and it appears to be the most commonly used database for academic research on U.S. index and stock option markets over the past two decades. We, then, compute returns for the same set of strikes with the learning model. Finally, we compare the calculated average returns from the learning model to the average returns from the historical data. What we find is that the patterns calculated with our learning model are similar to the ones obtained from the historical data from OptionMetrics. This showcases that the empirically observed patterns in short index option returns can be generated by learning about fundamentals. What should institutional investors take from this research? How should they apply it to their work today? The main implication of this study is that participants in option markets, including traders, institutional investors, and policy makers, should be cautious in asserting that specific option strategies provide abnormal returns. To make this claim, one needs to take all risks and non-linearities into account. One way to do this is with the learning model. To see the full paper, please visit https://doi.org/10.1080/07350015.2018.1505629. George Skiadopoulos is Professor of Finance in the School of Economics and Finance, Queen Mary University of London, and in the Department of Banking and Financial Management, University of Piraeus. He is also Director of the Institute of Finance and Financial Regulation (www.iffr.gr) and an Honorary Senior Visiting Fellow at Business School, City University of London.
And this can also happen: https://www.washingtonpost.com/arch...ped-out/090c57f7-fc75-4f02-9d22-585cf53c8548/ And an update: This investor has now deceased and I don't think he's ever recovered his losses.
Lol, the findings of our study is that you need to pay for our research to achieve outstanding returns
Yeah something that they can easily be found by doing google research or by their own backtesting or by least favourably their own live trading when they earned all the abnormal returns for a whole entire year only to see it completely wiped out by one trade. But they need to cover for the $$ that they spent on that learning model and all the manhours for all those "computing". IvyDB US Optionmetrics is expensive.
Contradictions are meant to happen. A particular strategy works fine for one and for others it can work just the opposite way. I sometimes wonder whether we can promote studies or journals like this. Guess we can. But again, if many such online courses are already available on Google, why would one want to pay for learning and trying out strategies.
CBOE took a look at why their cash reserved put index indicated that put writing achieved greater returns than their buy-write index. They found that most of the "excess return" was attributed to the modeling assumptions used to roll the contracts (SOQ to VWAP). In my own experience differences in return between buy-writes and covered calls revolve around differences in bid/ask spreads, if the contracts are expiring OTM or ITM.
the presumed reason selling puts on indexes works is because the volatility in indexes is skewed to the put side.. people are hedging index longs by selling covered calls and buying puts.. this pushes the price of calls down and the price of puts up.. as a result, the puts are overpriced compared to their probability of expiring ITM and calls are underpriced compared to their probability of expiring ITM.. go look at a put and call both $10 OTM on SPY.. you'll notice a huge difference in price.. that's the skew.. in markets, the probabilities tend to hold true in the long run.. so, if you are consistently selling overpriced puts, over time you'll see that difference show up as outperformance against the index.. however, if you sell puts and then start selling covered calls once you get assigned you will essentially negate the positive expected returns you created by selling the puts in the first place.. there are ways around this, but it gets super technical..
Interesting, shorter term ATM SPY calls and puts trade at parity. OTM prices show the put skew and 30 day ATM pricing shows a put skew.
skew increases as you go OTM.. also, the vol curve flattens the closer you get to expiry so its neutral at the time of expiry.. so the closer to expiry you are the more neutral the skew will be.. if you're trading skew, that mean reversion of the skewed vol curve to the neutral vol curve is essentially the edge you're capturing.. check out a vol smile, it looks kinda like a backwards nike check.. when there's put skew, the OTM puts are on the slope and the OTM calls are in the dip..
Fascinating, the 30-day atm prices I was looking at happened to fall on an ex-dividend date. So, that's why the puts and calls were not at parity. If you're opening your trade by selling at any given price and checking for call/put parity, which is what the CBOE study was addressing, you wouldn't notice skew. I have to wonder if there are that many collars being done that the skew is created, but maybe. I'm wondering if part of the skew pricing is due to the general tendency of equity market to rise. When I look at TLT and GLD the skew effect seems to disappear. Don't know, just wondering.