New Research: Finance prof Dr. Gradojevic explores financial option pricing before and during COVID-19

New research by Lang finance professor Dr. Nikola Gradojevic explores how COVID-19 impacted the global financial option market and the most effective ways to forecast its price. In his recent paper published in the journal Annals of Operations Research, Gradojevic and his co-author Dr. Dragan Kukolj (University of Novi Sad, Serbia) state that sophisticated machine learning methods in option pricing is the best determinant of future pricing, regardless of the state of the market. 

Financial options belong to a special class of financial instruments called financial derivatives. Options are traded in both organized exchanges (e.g., stock markets) and decentralized markets (e.g., foreign exchange market). The buyer of an option has the right to buy or sell the underlying security in the form of a stock, currency, bonds, or commodities at a fixed price within a specified period of time, called the maturity of an option. Movements in option markets are often used as a gauge for the expectations of market participants (or market sentiment) about the future. 

Dr. Gradojevic answers more details about his research below: 

Why focus on financial options? 

Similar to insurance contracts, financial option contracts provide their buyer with a payment for the loss that may be incurred in an investment. For this protection, the buyer of an option pays a price called the premium. Calculating a fair premium price is a complex task. If the premium is set too high given the amount of risk in the marketplace, there will only be a few buyers and the issuer may have difficulties selling options. On the other hand, if the premium is set too low, the issuer may not have sufficient funds to pay for investment losses once many options (or “insurance claims”) are exercised at the same time. 

In insurance, an actuary will utilize various statistical techniques in order to calculate premiums for a client. In business finance, a financial engineer will typically use computers to solve the problem of calculating a fair option premium. In recent years, AI (machine learning) has been employed for this purpose. Financial engineers rely on AI-based machine learning models called artificial neural networks, which are able to learn patterns from data. These models are called non-linear and are especially useful when financial markets are exposed to systemic shocks, such as the COVID-19 pandemic. 

What were the findings of your research? 

In our paper, we identified three distinct market phases during the COVID-19 pandemic in 2020:  

• January 1, 2020 to February 19, 2020 (pre-COVID-19); 

• February 20, 2020 to March 23, 2020 (COVID-19 market crash); 

• March 24, 2020 to June 15, 2020 (post-COVID-19 recovery). 

We studied how option pricing was impacted by these market regimes and, also, used explainable artificial intelligence (XAI) to understand the factors that were at play on each regime. First, we found that the first month of pandemic turmoil changed the investors’ behavior who concentrated on purchasing cheaper, more liquid options as though they expected a large market recovery swing. Thus, the traders were, to a certain extent, correct in predicting the impending market movements towards recovery. It is important to note that option traders are typically more sophisticated than stock traders and their behavior at the beginning of the pandemic clearly shows it. 

Then, our research findings suggested that the option pricing model’s accuracy worsened from the pre-COVID-19 to the recovery period. Complex effects emerged from the pandemic and the most accurate pricing models that were able to capture such effects were the advanced AI (machine learning) models. This confirmed the usefulness of sophisticated machine learning methods in option pricing, regardless of the market regime.  

What can we take away from these findings? 

During the COVID-19 market crash, we observed an increased importance of liquidity and demand shocks that were reflected in the open interest factor (i.e., the number of open option contracts available for trading) of the option price formula. Apparently, when the market is collapsing and is in a state of panic, open interest changes affect option demand imbalances, which, in turn, impacts option prices substantially. This factor is not included in standard option pricing formulas and the evidence of its increased importance during market distress represents a novel contribution to our understanding of option pricing mechanisms.  

Various market participants can benefit from this research. First, investors and scholars can gain a better understanding of the dynamics of option prices and their underlying risk factors during times of market volatility and distress. Then, it is of interest to policy-makers to measure how much variability in option (or, more generally, asset) prices can be attributed to behavioral (non-traditional) factors at crisis times. Ignoring such behavioral effects that can be inferred from our clustering method may lead policy-makers to wrong conclusions about the effectiveness of a particular policy aimed at calming volatile markets.