Leveraged ETF volatility offers risk modeling lessons
Research on single-stock leveraged ETFs highlights how daily rebalancing and volatility can create a wide gap between mean and median returns, a phenomenon that parallels challenges in insurance risk modeling. Financial researcher Hendrik Bessembinder noted that the daily reset mechanism can amplify or dampen returns depending on underlying patterns, providing a useful analogue for actuaries modeling outlier events and portfolio outcomes.
- The core issue with leveraged ETFs is "volatility decay," where the daily rebalancing mechanism erodes long-term returns, especially in choppy markets; even if the underlying stock returns to its starting price, the ETF can still lose value. This effect is magnified with higher leverage ratios, such as 3x, and during periods of high market volatility. For instance, one study found that from 2009 to 2018, a sample of 2x leveraged ETFs had an average annualized return of -11.1%, while the underlying indexes returned a positive 15.7%. - Hendrik Bessembinder's broader research into stock market returns reveals that a very small number of companies generate the vast majority of wealth creation. One study covering 1926 to 2016 found that just 4.31% of U.S. listed stocks accounted for all of the net wealth created in excess of Treasury bills. A global study from 1990 to 2018 showed this concentration was even more extreme, with only 1.33% of firms accounting for all net global wealth creation. - For actuaries, modeling extreme events and the impact of outliers is a persistent challenge, as traditional models can be distorted by anomalous data like the mortality shock from the COVID-19 pandemic. Actuarial science is increasingly focused on robust forecasting methods that can co-estimate the effects of outliers to avoid biased parameter estimates and inflated variance in risk capital assessments. Key concerns in the field include managing misinformation around new technologies like AI, the difficulty of modeling cyber risks due to limited historical data, and the financial transition risks associated with climate change. - Insurance companies are leveraging the modern data stack, using tools like Snowflake and dbt, to create more efficient and governed data platforms. These platforms support critical functions like fraud detection, risk assessment, and underwriting by enabling secure data sharing and faster analytics. For example, Open Insurance achieved 70% data self-service and halved the time to deliver BI dashboards by modeling its data with Snowflake and dbt Labs and then training GPT models on the improved documentation. - MLOps best practices are critical for deploying and maintaining reliable models in production for tasks like pricing and fraud detection. Key practices include versioning all components (data, code, models), establishing automated CI/CD/CT (Continuous Training) pipelines, and implementing robust monitoring to detect data and model drift. A focus on security and governance from the start ensures compliance with regulations in the financial and insurance sectors. - In consumer retail, AI is being used to deliver hyper-personalized customer experiences, which can increase ad engagement by 25% and reduce product return rates by up to 35%. Brands like Nike and Gucci are using AI for everything from designing new collections to providing personalized recommendations and virtual try-on experiences. This level of personalization is a strategic shift, with retailers seeing a 10% to 25% increase in return on ad spend for targeted campaigns. - New York City has solidified its position as the world's second-strongest tech hub, behind only Silicon Valley. From 2016 to 2021, the city's tech sector employment grew by 33.6% to over 172,000 jobs, even as the overall private sector shrank. This growth is creating strong demand for roles like software developer, cloud engineer, and cybersecurity analyst, with tech occupations accounting for over 40% of job postings with starting salaries of $80,000 or more.
