Adaptive Premium Harvesting Offers a Novel Return Stream
The Ever Evolving Portfolio Challenge
Traditional portfolio construction has been challenged over the past three years. Computational power and market access are opening new opportunities for many, while the fundamental building blocks of a robust portfolio are increasingly uncertain.
Fixed income allocations are no longer the ballast they used to be, in an environment characterized by shifting interest rate dynamics and elevated sovereign debt. Generational opportunities are playing out amid structural uncertainty and we find ourselves in a delicate situation, causing many to reassess their allocations.
Rather than beginning with ratios of traditional asset classes, what if we started with the characteristics we actually want? Steady return streams. Known risk parameters. Genuine diversification.
Opportunities in the Options Market
The options market provides an intriguing venue for this approach. It is rapidly expanding and having a growing influence on the assets that underlie derivative contracts, especially equities and their indices. Faster news cycles and simpler access for many market participants are among the factors driving this adoption.
Investors buy options for two core reasons: speculation on price increase, and protection against price decrease. In both of these cases, the option buyer is paying for the right to transact at a price and time of their choosing. This freedom of choice afforded to the buyer has value in itself, creating an opportunity for asset managers to develop a return stream based on demand for optionality.
An analogy is often made between selling options and selling insurance. What makes the insurance business profitable are risk models- quantification of how often and how severely insurable events occur. These models ensure that premiums collected outweigh the payouts delivered to customers.
Covered Calls
Asset managers can use simple versions of the insurance business model to generate returns, and there are already established examples available today in the form of yield-focused ETFs. The most common form employs a covered call strategy, which holds an underlying asset and sells call options against it.
The best case scenario for this strategy is to keep the premium paid for those options, while also enjoying some appreciation in the underlying asset. However, during quick moves in either direction, this strategy becomes vulnerable. On the downside, the underlying depreciates, and the capped premium collected from selling calls leaves some of the downside exposed. On the upside, sold calls become a liability as their value can increase far beyond what they were sold for, forcing the fund to buy them back at a loss that dilutes returns from the underlying.
This specific dynamic, when options rapidly increase in price, leaves some commentators to describe option selling as “picking up pennies in front of a steamroller”, with frequent small gains being counterbalanced with the threat of large losses when the trade goes wrong. For many, this is unattractive.
Machine Learning Mitigates Risk
A more strategic approach to option selling is possible. Like actuarial models for insurance, there are relatively simple and inexpensive algorithms that can help avoid the metaphorical steamroller. Option sellers have advantages that make the task of predicting market movements more approachable.
The first is time. Option sellers have time on their side. As time passes, the value of a buyer’s right to choose decreases, as the range of possible outcomes narrows. Option sellers meanwhile experience the decrease in possible outcomes as profit. The second is a low precision requirement. Option sellers don’t have to be right about the price of an asset at some point in the future, as long as they are not too wrong for long enough.
This characteristic pairs well with simple, inexpensive machine learning. Models don’t have to be extremely precise to be very useful. A model can be trained to simply predict up or down, and as long as sufficient time passes without the prediction being wrong, premium is harvested by buying back the option for less than it was sold. Gaining even a small directional edge from prediction gives the seller a different perspective on option prices than the market implies, allowing them to to write contracts at prices that favor their view.
The Future of Premium Harvesting
Together, pragmatic ML and option selling make a very complementary pair indeed. In a challenging environment for traditional income streams that depend on long term macro fluctuations, adaptive premium harvesting could fill the emerging gap in supply for steady uncorrelated returns. Low implementation costs and quick feedback loops are allowing for rapid development and we may soon see a new category emerge that deserves a place in every savvy investor’s portfolio.