Options on USO when oil went negative

Normal conditions

During normal market conditions, USO moves at the same volatility as the weighted average volatility of its holdings. Before 2020, USO held the front-month future and therefore its realized volatility matched the reference futures’ realized volatility. If you looked at close-to-close realized volatility you’d get different measures but that’s for technical reasons:
  1. USO and the futures have different closing times.
  1. Because of the roll, sometimes USO is holding a blend of first and second-month futures. In that case, USO’s volatility would match the weighted average of its holdings’ volatility.
Pricing the options on the ETF given that the futures options are the bigger and more liquid market is an exercise in arbitrage replication. For ETF options that expire more than 1 month in the future, the problem is actually quite tedious and not a pure arbitrage. I won’t explain how you do it, but suffice to say, it makes a good interview question and requires pulling together a lot of knowledge about how futures curves work.
Pricing a near-dated USO option is straightforward. It will hold the same contract for most of its life and it can easily be arbitraged against an option on that same near-dated future.

When oil went negative

When oil went negative, pricing options on USO became more difficult. But an opportunity arose if you thought about the problem qualitatively. We’ll get to that in a moment.
  1. First, why did pricing USO options get more difficult? You had to change your models from ingesting implied volatilities and think in terms of dollar premiums. Both implied and realized volatilities are usually represented as annualized standard deviations in percent. If an asset has 25% implied volatility, the market is implying that its 1-year standard deviation is 25%. But if oil is 0 or negative, the notion of percent volatility is nonsensical. Furthermore, the typical option pricing models such as Black-Scholes that presume lognormality are worthless — a lognormal distribution is bounded by zero. For both of the above reasons, the option models we were forced to use had little intuitive context. The numbers you’d stick in the model to get the straddle to line up to the market were gibberish. Instead, I rely on some ad-hoc translations that converted futures options premiums to USO option premiums just like previous question where we turned the $1 put premium into a fair $2 NAV premium. This was fine enough because there was a trade that made sense without needing models at all.
  1. Just sell the USO options In the preceding questions, I reversed the order of how I encountered the market in Spring of 2020. In the above examples, we built up to the idea that the ETF should trade at a premium to NAV and then we quantified by how much. In 2020, I first observed that the ETF was trading premium to NAV as oil was dropping but before it went to and through zero. My absolute first reaction was, “there’s a lot of people who can’t trade the futures that are trying to buy oil cheap and that demand is causing a premium to NAV in USO”. It wasn’t until they started listing strikes below zero on the futures that I realized that the futures can go negative and therefore the ETF has become the 0-strike call. Using the market in the 0-strike puts for the futures, I computed a “fair premium” for USO. It was actually trading even fatter than that premium by about a 10-15 cents at one point. But this insight wasn’t as valuable as the next one… The USO options were likely very overpriced. I had 3 reasons for this opinion: a) There was a non-zero chance that a fast enough sell-off would take USO to zero. If the fund liquidated, all extrinsic option value in the market would go to zero (kind of like a cash takeover situation). And of course, implied volatility in USO was quite fat at the time. b) The USO fund managers made the quick decision to exercise their right to roll into longer-dated futures at their discretion. At first, they rolled immediately into second and third-month futures that didn’t appear to have a risk of going to zero. But they had shown their hand — they were not going to let the fund go to zero even if it meant changing its complexion dramatically from something that was created to track near-dated oil. I thought there was a real possibility, that they would roll into even more deferred futures. Alas, they did! In fact, they rolled into futures as far out as 1 year. But why is this significant? Because the realized volatility of futures declines the further you go out in time. When you see the news, they talk about the prompt future’s price change. But the 5-year future barely moves. A 1-year future might move at about 50 or 60% of the volatility of a prompt future. And that percentage falls even more when the front month future gets extra-volatile as it did in 2020. USO was now rolling into contracts that had reliably lower realized volatility. Combined with (a) these USO options are looking more and more juicy! c) Finally, we come to the implication of our work on the ETF NAV. Remember the premium to NAV reflects the extrinsic value of the 0-strike. And when the futures are near zero, USO has less than a 100% delta. So it’s mechanically lower volatility than the futures just as a hedged portfolio is less volatile than an unhedged one. The options on this ETF which had to overcome the drag of its premium to NAV eroding as it rallied were now looking certifiably expensive.

The final upshot

The USO options had become a hassle to price, but reasoning through the situation it was clear that on an expectancy basis there were sound arguments for thinking they were just too expensive.
Combining that with a model for the fair NAV premium (remember there should be a NAV premium, but the one that was trading was still too high), the game plan was to sell calls and just delta hedge them with futures. There was no need to be surgical about this…you could just stream offers or run electronic eyes to sell calls on the screens. As a matter of strategy, I remember there not being any real liquidity on the bid side in the voice market which was a clue that other sharps understood what was going on. So it wouldn’t make sense to just napalm all the screen bids because there was no size behind them and you’d be better off chipping away mid-market on smaller lots and on the off-chance that a buyer showed up (maybe somebody would want to buy calls to bet on an equally sharp recovery) you could let the screens frame up a nice sale (even if that meant smoking a customer order near the bid…size not price was the priority given how dislocated the options looked.